Professional Documents
Culture Documents
Advanced Heat Transfer, 3rd Edition
Advanced Heat Transfer, 3rd Edition
Advanced Heat Transfer, 3rd Edition
Third Edition
Greg F. Naterer
Third Edition published 2022
by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742
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, micro-
filming, 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, access www.copyright.com or contact the
Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. For works that are not
available on CCC please contact mpkbookspermissions@tandf.co.uk
Trademark notice: Product or corporate names may be trademarks or registered trademarks and are used only for identifica-
tion and explanation without intent to infringe.
DOI: 10.1201/9781003206125
Typeset in Times
by SPi Technologies India Pvt Ltd (Straive)
In memory of my son Jordan who inspires me by his everlasting zeal of exploring and
climbing to greater heights.
Contents
Preface............................................................................................................................................... xv
Author.............................................................................................................................................xvii
Lists of Abbreviations....................................................................................................................xviii
Chapter 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 Transport Properties............................................................................... 7
1.4 Heat Conduction................................................................................................. 9
1.5 Convection........................................................................................................ 11
1.6 Thermal Radiation............................................................................................ 12
1.7 Phase Change Heat Transfer............................................................................. 14
1.8 Mass Transfer................................................................................................... 15
References................................................................................................................... 17
Problems...................................................................................................................... 17
vii
viiiContents
Chapter 3 Convection................................................................................................................... 73
3.1 Introduction...................................................................................................... 73
3.2 Governing Equations........................................................................................ 74
3.2.1 Conservation of Mass (Continuity Equation)...................................... 74
3.2.2 Conservation of Momentum (Navier–Stokes Equations).................... 75
3.2.3 Total Energy (First Law of Thermodynamics).................................... 79
3.2.4 Mechanical Energy Equation............................................................... 80
3.2.5 Internal Energy Equation..................................................................... 82
3.3 Nondimensional Form of Equations................................................................. 83
3.3.1 Dimensionless Variables...................................................................... 83
3.3.2 Buckingham–Pi Theorem.................................................................... 86
3.4 Convection Boundary Layer............................................................................. 88
3.4.1 Boundary Layer Equations.................................................................. 88
3.4.2 Heat and Momentum Analogies.......................................................... 92
3.5 Evaporative Cooling......................................................................................... 93
3.6 Flat Plate Boundary Layer................................................................................ 94
3.6.1 Scaling Analysis................................................................................... 95
3.6.2 Integral Solution.................................................................................. 96
3.6.3 Similarity Solution............................................................................... 98
3.7 Flow Past a Wedge.......................................................................................... 103
3.8 Cylinder in Cross Flow................................................................................... 105
3.9 Other External Flow Configurations............................................................... 107
3.9.1 Sphere................................................................................................ 107
3.9.2 Tube Bundles..................................................................................... 108
3.10 Internal Flow................................................................................................... 109
3.10.1 Poiseuille Flow in Tubes.................................................................... 109
3.10.2 Non-Circular Ducts............................................................................ 116
3.11 Free Convection.............................................................................................. 118
3.11.1 Vertical Flat Plate............................................................................... 118
3.11.2 Body Gravity Function Method......................................................... 123
3.11.3 Spherical Geometries......................................................................... 127
3.12 Introduction to Turbulence............................................................................. 128
3.12.1 Turbulence Spectrum......................................................................... 128
3.12.2 Reynolds Averaged Navier–Stokes Equations................................... 130
3.12.3 Eddy Viscosity................................................................................... 132
3.12.4 Mixing Length................................................................................... 132
3.12.5 Near-Wall Flow.................................................................................. 133
3.12.6 One and Two-Equation Closure Models............................................ 134
3.13 Entropy and the Second Law.......................................................................... 136
3.13.1 Formulation of Entropy Production................................................... 136
3.13.2 Apparent Entropy Production Difference.......................................... 139
3.13.3 Dimensionless Entropy Production Number..................................... 142
References................................................................................................................. 143
Problems.................................................................................................................... 144
Appendices������������������������������������������������������������������������������������������������������������������������������������ 515
Appendix A: Vector and Tensor Notations............................................................... 515
Appendix B: Conversion of Units and Constants.................................................... 516
Appendix C: Convection Equations in Cartesian, Cylindrical and
Spherical Coordinates......................................................................... 517
Appendix D: Properties of Solids............................................................................. 520
Appendix E: Properties of Gases............................................................................. 523
xivContents
Index���������������������������������������������������������������������������������������������������������������������������������������������� 543
Preface
Since the publication of the Second Edition in 2018, numerous updated and emerging new fields
have motivated the development of this new Third Edition of Advanced Heat Transfer. The Third
Edition aims to provide a more comprehensive treatment of advanced topics and enrich the student
learning experience with more case studies, examples, and problems. It also broadens the scope of
applications of thermal engineering, including but not limited to, biomedical, micro-and nanotech-
nology, and machine learning. This Third Edition was updated based on input from colleagues and
readers over the years since the first edition.
Typically courses on advanced topics in thermal and energy systems have specialized books on
a specific aspect of thermal engineering with a focus on the selected topic in depth. Interactions
among various modes of multiphase heat transfer are not fully explored or connected in a way that
provides a systematic framework of analysis. Advanced methods of analysis are typically covered
in depth for a single specific mode of heat transfer, such as conduction or convection, rather than
an overview of methods that can be applied over a wider range of heat transfer modes. In this Third
Edition, advanced solution methods are presented for a 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 this latest edition. In this Third Edition, the chapter on heat conduction (Chapter 2)
provides a more comprehensive treatment of solution methods including non-homogeneous systems
and time-dependent boundary conditions. Also, new sections are added on emerging topics of active
research such as conduction in porous media, heat transfer in living tissue, and microscale conduc-
tion. In each new section, a brief overview of the topic is outlined, along with references to other
sources for more detailed analysis.
Also, the chapter on radiation (Chapter 4) was significantly updated. In particular, the transport
phenomena of radiation in participating media were expanded to include the processes of attenua-
tion by absorption, scattering from multiple directions, enhancement of intensity by emission, and
the resulting radiation exchange among the participating gases. Fundamentals and the governing
differential equations of energy transfer for participating media are presented. This includes the
general equation of transfer, involving the radiative flux vector and overall conservation of energy.
Approximate solutions of the equation of transfer are also discussed. Simplifications based on the
emission and Rosseland approximations are used to find semi-analytical solutions. Coupled radia-
tion and convection problems are considered and analyzed.
More detailed treatment of transport phenomena is also presented in the updated chapter on
chemically reacting flows (Chapter 8). Diffusion processes during chemical reactions are ana-
lyzed for both heterogeneous and homogeneous reactions, as well as reactions in porous catalysts.
Coupled heat and fluid flow with chemical reactions are formulated and investigated with analytical
solution methods. In addition, the section on fuels and combustion is expanded to analyze burning
fuel droplets and radiation exchange.
Several emerging areas of heat exchanger development are also presented in the Third Edition
(Chapter 9). Further types of heat exchangers include honeycomb, moving bed, micro-scale, com-
pact, helically coiled, and two-phase heat exchangers. Systems with more than two operating fluids
are examined. Also, thermal enhancement with metal foams is analyzed. The method of entropy
generation minimization is applied to the design of heat exchangers, in particular, for counterflow
systems, conditions with flow imbalance, phase change, and finned tube heat exchangers.
Lastly, the chapter on computational heat transfer (Chapter 10) was also significantly updated.
New sections were added on radiation heat transfer, including numerical methods of the discrete
transfer model, discrete ordinates, and finite volume method. Concepts in numerical methods for
two-phase flows (boiling and condensation) are presented. Also, an introduction is provided for
xv
xviPreface
machine learning and artificial intelligence. The fundamentals of artificial neural networks and lin-
ear regression are outlined. As an application of the fundamental principles, case studies of heat
conduction and forced convection are presented and discussed.
This Third Edition aims to cover a range of advanced heat transfer topics of importance in both
undergraduate and graduate-level courses. It can serve both as undergraduate and follow-up gradu-
ate 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.
This book is organized into six overall areas: (1) introduction (Chapter 1); (2) primary single-
phase modes of heat transfer (Chapters 2–4); (3) multiphase heat transfer (Chapters 5–7); (4) chemi-
cally reacting flows (Chapter 8); (5) heat exchangers (Chapter 9); and (6) computational 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 any of the specific multiphase systems (Chapters 5–7),
such as gas–liquid (Chapter 5) or liquid–solid systems (Chapter 7), without a loss of continuity. The
latter chapters on chemically reacting flows, heat exchangers, and numerical heat transfer, can also
be studied independently of the others without a loss of continuity.
I’m grateful to faculty colleagues and students who have contributed in significant ways
to the preparation in this Third Edition. My role as the Editor-in-Chief of the AIAA Journal of
Thermophysics and Heat Transfer has provided a window into the latest and emerging developments
in thermal engineering systems. Also, special thanks to Rhys Northcote at the University of British
Columbia for his contributions to the machine learning section, and Brian McDonald, Nancy Chafe,
Brandon Howell, Guofei Yan, Kheya Zaman, Saleh Elfasay, Mahamudul Hasan, and Logan Hornell
at Memorial University, for their contributions to the problem sets.
Greg F. Naterer
St. John’s, Canada
Author
Greg F. Naterer is a Professor of Mechanical Engineering and the Dean of the Faculty of
Engineering and Applied Science at Memorial University in St. John’s, 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 Engineering
Deans Canada (EDC) and the Discovery Grant Selection Committee of the Natural Sciences and
Engineering Research Council of Canada (NSERC). He is the Editor-in-Chief of the AIAA Journal
of Thermophysics and Heat Transfer.
Dr. Naterer has made significant contributions to the fields of heat transfer, energy systems, and
fluid mechanics. He led an international team that developed and constructed a copper–chlorine
cycle of thermochemical hydrogen production. In addition to this Third Edition, he has co-authored
two other books: Hydrogen Production from Nuclear Energy (with I. Dincer, C. Zamfirescu; 2013,
Springer) and Entropy- Based Analysis and Design of Fluids Engineering Systems (with J.A.
Camberos; 2008, CRC Press/Taylor & Francis).
Among his awards and honors for teaching and research, Dr. Naterer has received the EIC Julian
C. Smith and K.Y. Lo Medals, CNS Innovative Achievement Award, CSME Jules Stachiewicz
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), Canadian Academy of Engineering (CAE), and American Institute of Aeronautics
and Astronautics (AIAA). Dr. Naterer earned a PhD in mechanical engineering in 1995 at the
University of Waterloo, Canada.
xvii
Lists of Abbreviations
List of Symbols
a extinction coefficient (absorption plus scattering), m–1
A area, m2
AF air–fuel ratio
b unfrozen water film thickness, m
B Ice thickness, m
Bi Biot number (hD/k)
BT Spalding number (cvΔT/hfg)
c speed of light, m/s; capacitance matrix
cd drag coefficient
cf skin friction coefficient
cp specific heat at constant pressure, kJ/kgK
cv specific heat at constant volume, kJ/kgK
C concentration; slip correction factor
CC cloud cover fraction
D diameter, m; mass diffusivity, m2/s
e internal (thermal) energy, kJ/kg
E total energy, kJ; emissive power, W/m2
Ec Eckert number (U2/cpΔT)
Ei exponential integral
Eo Eotvos number (ΔρgD/σ)
f friction factor
F force, N; view factor; blackbody function; correction factor
Fo Fourier number (αt/L2)
Fr Froude number (U/g1/2L1/2)
g Gibbs free energy, kJ/kg; gravitational acceleration, m2/s; Lamé coefficient
G irradiation, W/m2; mass velocity, kg/m2s
Ga Galileo number (gμ4/ρσ3)
Gr Grashof number (gβΔTL3/ν3)
h convection coefficient, W/m2K; enthalpy, kJ/kg; Planck’s constant (6.63 × 10–34 Js)
hfg latent heat of vaporization, kJ/kg
hsl latent heat of fusion, kJ/kg
H height, m
i, j unit vectors in coordinate directions
I intensity of radiation, W; turbulence intensity
j diffusion flux; Colburn factor; drift flux, m/s
J radiosity, W/m2; Jacobian determinant
Ja Jacob number (cpΔT/hfg)
k thermal conductivity, W/mK; turbulent kinetic energy, m2/s2; reaction rate coefficient
K permeability, m2; chemical equilibrium constant
Kn Knudsen number (λ/L)
L length, m
Le Lewis number (α/D)
m mass, kg
ṁ mass flow rate, kg/s
M molar mass, kg/kmol; heat pipe merit figure
Mo Morton number (gμ4Δρ/ρ2σ3)
xviii
Lists of Abbreviations xix
Greek Letters
α thermal diffusivity (k/ρcp), m2/s; absorptivity; solar altitude angle
β thermal expansion coefficient, 1/K; flow excess parameter
χk mass fraction of phase k
δ boundary layer thickness, m; film thickness, m; thermal penetration depth, m
xx Lists of Abbreviations
Subscripts
a air; ambient; activation; absorption
b blood
A, B constituents A and B in a mixture
atm atmospheric
b base; blackbody; boiling; Binghman slurry; body force; bubble
bndry boundary
c cross-section; collector; cold; critical; characteristic; combustion
civ civic
conv convection
crit critical
cv control volume
d drift; droplet
D diameter
dif diffuse
dir direct
e east; mean beam length; eutectic
eff effective
ev evaporation
f fluid; fin; fusion (phase change point); final; formation
fg fluid–gas
g gas; ground; glass; generation
gen generation
h hot; high; horizontal; hydraulic
i inner; interfacial; initial; surface index
in inlet
Lists of Abbreviations xxi
ip integration point
j surface index
k phase number; kinetic energy
ko Kolmogorov
l, L low; laminar; liquid
le liquid entrainment
liq liquidus
lm log mean
loc local
m mixing length; mean; melting point; mass transfer; metanbolic
min minimum
mf minimum fluidization
mp melt particularization
n north
nb neighboring
nuc nucleate
o outer
opt optimal
out outlet
p particle; pipe
P product; Planck
r reference; relative; removal
R reactant; Rosseland
rad radiation
ref reflected
res residence
s surface; entropy; solar; solid; south; settling
sat saturation
sg superficial gas; solid–gas
sl superficial liquid; solid–liquid
sol solidus; solar time
sub subcooled
t thermal; turbulent
tp two-phase
U drift velocity
v vapor
w wall; wick; water; west
x, y, z Cartesian coordinates
1, 2 node numbers
∞ ambient; freestream
Superscripts
′ turbulent fluctuating quantity
″ flux quantity (per unit area)
* non-dimensional 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 transfer phe-
nomena. 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 con-
vection (or free convection). 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 abso-
lute zero temperature (zero Kelvin, or –273.15°C). These processes will be described in detail in
following 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 chap-
ters, 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 solidi-
fication 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,
thermal–solutal 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 thermal engineering is how to achieve a specified heat trans-
fer rate as efficiently and economically as possible. For example, in microelectronics assemblies,
designers seek better ways of cooling the electronic circuits and more efficient alternatives to con-
ventional 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 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.
DOI: 10.1201/9781003206125-1 1
2 Advanced Heat Transfer
density by macroscopic averaging of the mass divided by the volume (see Figure 1.1). In this defi-
nition, 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 fluctuations. 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
limited 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 char-
acteristic 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 refer-
ence, 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 trajectories. 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 repre-
sents 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 coordinates 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 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.
Kinematics refers to the study of properties of motion, such as fluid velocity and acceleration.
The kinematic properties involve two components: (i) a time derivative; and (ii) spatial derivatives.
The total rate of change of a scalar quantity, B, denoted by the material derivative, DB/Dt, consists
of both the time and spatial components:
DB B B B B
u v w (1.1)
Dt t x y z
4 Advanced Heat Transfer
where u, v, and w refer to the x-, y-and z-direction velocity components, respectively. For example,
to determine the acceleration field, B represents the velocity. The notation of DB/Dt refers to the
total (or material, or substantial) derivative of B. The portion of the material derivative represented
by the spatial derivatives is called the convective derivative. It accounts for the variation of the fluid
property, B, due to a change of position, while the temporal derivative represents the change with
respect to time.
An alternative useful notation is tensor or indicial notation, especially when expressions involv-
ing vectors and/or matrices become increasingly complex or lengthy. A tensor is a variable with
appropriate subscripts. When a subscript (or index) is repeated in a tensor, it denotes a summation
with respect to that index over its range, for example, i = 1, 2, 3. 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 variable 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 (1.1) can be written more concisely as
DB B B
ui (1.2)
Dt t xi
where i = 1, 2, 3. Further details on tensors and indicial notation are provided in Appendix A.
1.2 CONSERVATION OF ENERGY
The conservation of energy, or first law of thermodynamics, is a fundamental basis of heat transfer
engineering. Two general types of energy balances may be used – either a control mass or a control
volume approach. A control mass refers to a closed system of no inflow or outflow of mass from the
system. In contrast, a control volume refers to an open system consisting of a fixed region in space
with inflows and/or outflows of mass across the boundary surfaces. A general energy balance for a
control volume can be expressed as follows (see Figure 1.2):
E cv E in E out E g (1.3)
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 overdot notation refers to the rate of change
with respect to time. A “steady state” refers to conditions that are independent of time, or in other
words, 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 exam-
ple, 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.
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 E f (1.4)
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 inter-
nally within the domain.
1.3 THERMOPHYSICAL PROPERTIES
Modeling of thermal and fluid transport behavior is inherently limited by an incomplete detailed
understanding of the underlying physical processes at a microscopic level. The governing equa-
tions often introduce averaged properties of the medium in order to relate dependent to independent
variables. For example, in the case of heat conduction, the thermal conductivity is a proportionality
constant that relates the heat flux to the temperature gradient. Transport properties reflect a further
limitation in the ability to accurately measure these transport processes. Variability of experimental
results introduces measurement uncertainties that affect the accuracy of models and correlations
which use transport properties.
In this section, three different types of thermophysical properties of a system will be discussed:
thermodynamic, transport, and other properties. Fundamentals of their associated physical pro-
cesses will be described.
1.3.1 Thermodynamic Properties
Thermodynamic properties or variables include pressure (p), density (ρ), enthalpy (h), specific vol-
ume (υ), temperature (T), specific internal energy (e), and specific entropy (s). The fluid enthalpy, h,
is defined by h = e + p υ. “Specific” properties are those expressed on a per mass basis. A thermo-
dynamic property is any property that is measurable and which describes the state of the physical
system. Some thermodynamic constants, such as the ideal gas constant, R, do not describe the state
of a system, and so these are not properties.
6 Advanced Heat Transfer
Specific or intensive properties are independent of size, whereas extensive properties (such as the
total energy) are dependent on the size of the system. For example, an extensive property of a system
containing two parts, A and B, is the sum of properties of both parts A and B. The state postulate
of thermodynamics states that for a simple compressible substance, the number of intensive inde-
pendent properties of a system equals the number of relevant reversible work modes plus 1. One is
added because even if all properties are held constant within a system, one further property can be
changed, such as temperature, through heat transfer.
Pressure, p, is the normal force per unit area acting on a fluid. It is associated with a momentum
change of a fluid and represents a force applied perpendicular to the surface of an object, per unit
area, over which the force is distributed. Consider the force exerted 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 aver-
age 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 RT (1.5)
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 buoy-
ant 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 balance 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 subvisible
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 char-
acterizes 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 isolated 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
perspective that the direction of energy conversion can be reversed. Examples of irreversible pro-
cesses are dissipation of kinetic energy to frictional heating in a boundary layer and heat transfer
from a higher to lower temperature.
Introduction 7
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 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 (1.7)
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 sur-
roundings, 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 perform 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.
1.3.2 Transport Properties
Molecular properties of a substance, such as thermal conductivity, viscosity, and diffusion coef-
ficient (diffusivity), indicate the rate at which specific (per unit volume) heat, momentum, or mass
are transferred through the substance. Examples of thermal conductivities of various solids, liquids,
and gases are shown in Figure 1.3. Values are shown in SI units. Conversion factors between SI
and Imperial units for a number of transport and other thermophysical properties are presented in
Appendix B. An extensive source of property tables for other solids, liquids, and gases was provided
by Weast (1970).
It can be observed from Figure 1.3 that the thermal conductivity varies widely, depending on the
type of substance, from 0.01 W/mK for gases up to 10 to 400 W/mK for metals. Metals have much
higher thermal conductivities than liquids and gases. In metallic bonds, metal atoms give up their
free outer-shell valence electrons to an electron gas and take up a regular 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 liquids 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 nonmetal ions are attracted to
each other. 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,
8 Advanced Heat Transfer
FIGURE 1.3 Thermal conductivities of fluids and solids. (adapted from Hewitt, Shires, & Polezhaev, 1997)
thereby yielding a lower thermal conductivity than solids with metallic bonds. Also, mechanical
properties of the solids 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
(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 viscos-
ity 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 lin-
early 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.
Introduction 9
In the evaluation of transport properties, macroscopic relationships are often used to approxi-
mate the microscopic transport phenomena. For example, heat conduction involves molecular fluc-
tuations. Tracking of individual molecules, through their rotational, translational, and vibrational
modes of motion, and assessing their energy exchange during these intermolecular motions, would
be impractical. Therefore, a phenomenological 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 viscosity. Phenomenological laws refer to approximations based on experimental observa-
tions that improve theoretical models but which cannot be proven rigorously from mathematical or
first principles. The thermal conductivity is defined through a phenomenological law as a macro-
scopic 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, coefficient 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 compressible substances as partial deriva-
tives of the internal energy, e(T, v), and enthalpy, h(T, p), respectively:
e
cv (1.9)
T
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 attrac-
tion 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 substance is heated, the molecular
kinetic energy, vibrations, and movements increase, thereby creating a larger intermolecular separa-
tion 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 temperature since more thermal energy is required to break molecular
bonds in the phase transition at higher temperatures.
1.4 HEAT CONDUCTION
Conduction is a primary mode of heat transfer. Thermal energy is transported by intermolecular
motion during microscopic particle to particle interactions. The carrier motions are the translational,
rotational, and vibrational movements of molecules. In this way, heat is transferred even though no
net bulk movement of the substance occurs.
Consider heat conduction through a solid or liquid layer contained between a heated left wall
and a cooled right wall in Figure 1.4. Molecular energy is transferred between the hotter side of the
layer and the colder side. Also, consider an imaginary plane dividing the left and right sides of the
layer into equal portions. On average, an equal number of molecules crosses the imaginary plane in
both directions. However, the molecules coming from the hotter section possess more energy than
the right side. As a result, there is a net transfer of thermal energy from the hotter side to the colder
side, without any net transport of mass across the dividing plane.
10 Advanced Heat Transfer
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 typi-
cally approximated by macroscopic averaged quantities, such as temperature and thermal conduc-
tivity, k. In a one-dimensional layer in Figure 1.4, the conduction heat transfer rate is governed by
Fourier’s law, named after Jean–Baptiste Fourier (1768–1830):
qcond T T
qcond k k (1.11)
A x x
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 tempera-
ture gradient in Equation (1.11) can be approximated by the temperature difference 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 compo-
nents 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 positive heat flow in the
positive x-direction. This sign convention is illustrated in Figure 1.5. 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 surface. 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 approximation of linearization of the temperature gradient in Equation (1.11) for steady-
state one-dimensional problems.
In Fourier’s law, the thermal conductivity is a transport property that characterizes the rate at
which a substance can conduct heat. As discussed earlier, the variations of thermal conductivity with
temperature are influenced by the molecular structure and intermolecular interactions. For example,
the thermal conductivity of pure metals is often higher at low temperatures, but then increases to
100–400 W/mK at room temperature. Less uniform molecular structures in alloys often lead to
lower thermal conductivities than pure metals. The thermal conductivity of water rises to about 0.7
W/mK at 200 K, but then gradually falls afterward. In gases, an opposite trend is usually observed
where the thermal conductivity continually rises with temperature.
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 temperature 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 1.6). 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
qconv h Ts T f (1.12)
A
configuration, fluid velocity, surface roughness, and other factors to be discussed 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 velocity and temperature fields
change through a boundary layer, and more generally the entire flow field, the governing differential
equations of mass, momentum, and energy conservation must be solved. These three-dimensional
equations in Cartesian, cylindrical and spherical 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 convective 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 com-
monly used for convection problems involving 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 carefully
specified. The selection of an appropriate temperature depends on the type of problem under con-
sideration. In the case of external flow, such as external flow over a circular cylinder, the fluid
temperature is given by the freestream temperature. However, for internal 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 temperature 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 convection
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, consider external flow over
the top surface of the object in Figure 1.6. 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 coef-
ficient 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, a further detailed analysis of convective heat transfer will be presented.
1.6 THERMAL RADIATION
Thermal radiation is a form of electromagnetic energy emitted by all matter above absolute zero
temperature. These emissions arise from changes in the electron configurations of the constituent
atoms and molecules at an atomic scale. Energy is transported through space as a result of this
Introduction 13
electron activity in the form of electromagnetic waves (or photons). Thermal radiation does not
require the presence of a material medium between the objects exchanging energy by thermal radia-
tion. It occurs most efficiently in a vacuum.
Consider a surface that exchanges heat by radiation and convection with the surroundings (see
Figure 1.7). Incident radiation arrives on the surface from the surroundings. Some of the incident
radiation is reflected from the surface, while other incoming energy is absorbed into the surface and
potentially transmitted through the object. Some of the electromagnetic waves emitted from the
surface may not be absorbed by other surrounding objects if they are not in the direct line of sight
of another surface.
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 pri-
mary governing equation is Stefan–Boltzmann’s law. For a surface at a temperature 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
qrad Ts4 (1.13)
A
where A is the surface area and σ = 5.67 × 10–8 W/m2K4 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 qrad ′′ and qrad, respec-
tively. 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 properties of various surfaces and gases are
presented in Appendix G.
Radiation heat transfer between two surfaces or objects, at temperatures T1 and T2, respectively,
can be obtained based on the net radiation exchange between the surfaces,
qrad
T14 T24 (1.14)
Alternatively, linearizing this expression to follow a similar form as Newton’s law of cooling,
hr T1 T2
qrad (1.15)
FIGURE 1.7 Radiative heat transfer between and a surface and its surroundings.
14 Advanced Heat Transfer
hr 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 exam-
ple, consider the net radiation exchange between a small object at a temperature of T1 and a sur-
rounding 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 A 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 molecules,
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 analy-
sis, physical processes, governing equations, and methods of analysis of radiation heat transfer will
be presented and discussed.
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 1.8, 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 sur-
face. 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 processes will be presented in Chapter 5.
Melting occurs in region 2 of Figure 1.8. 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 freez-
ing 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.
1.8 MASS TRANSFER
Similar to the flow of heat by conduction due to a change of temperature, the driving force for mass
transfer by diffusion is a difference of species concentration in a mixture. In a manner similar to
Fourier’s law for the conduction of heat, Fick’s law of mass transfer states that the species diffu-
sion flux from regions of high concentration to regions of low concentration has a magnitude that
is proportional to the concentration gradient. For example, in a two-component mixture, the solute
diffuses from a region of high concentration to a region of low concentration down the concentration
gradient, where the constant of proportionality is called the mass diffusivity.
16 Advanced Heat Transfer
There is a close analogy between heat and mass transfer by diffusion. When the term “mass
transfer” is used in this context, it refers to the relative movement of species in a mixture due to the
presence of concentration gradients, not the movement of mass due to bulk fluid motion. As with
the transfer of heat and momentum, the extent of mass transfer is affected by flow patterns within the
system and diffusivities of the species in each phase. Also, mass transfer coefficients are calculated
and applied similarly as the convective heat transfer and skin friction coefficients.
Recall that intermolecular energy exchange by heat conduction from a hot to cold side of a layer
was discussed earlier. Consider instead an analogous process of mass diffusion of interstitial atoms
(component A) through a solvent (component B), across an imaginary plane over a distance of ∆x.
A layer of fluid with molecular motion and interstitial atoms migrate across the imaginary plane
(Figure 1.9).
Let ГA, N, N1, and N2 refer to the average number of interstitial atom jumps per second in random
directions, number of adjacent sites (i.e., n = 6 in three dimensions), number of A atoms per unit
area in plane 1 (immediately to the left of the imaginary plane), and the number of A atoms per unit
area in plane 2 (immediately to the right of the imaginary plane), respectively. Then the net flux of
interstitial atoms across the imaginary control surface can be written as
jA 1 1
jA jA,12 jA,2 1 A N1 A N 2 (1.18)
A N N
where the double prime notation (″) represents a flux quantity (per unit area). A positive expression
for jA indicates that there is a net species mass flux from the left side to the right side of the imaginary
plane.
The concentration of A atoms at planes 1 and 2, CA,1 and CA,2, respectively, is
N1 N
C A,1 ; C A, 2 2 (1.19)
x x
Performing a Taylor series expansion of the concentration of A atoms at plane 2, in terms of the
concentration nearby to the left at plane 1,
C A
C A,2 C A,1 x (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 imaginary plane,
C
j A A x 2 A 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):
jA dC
jA 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 temperature gradient. A 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.
REFERENCES
Bird, R.B., W.E. Stewart and E.N. Lightfoot. 2007. Transport Phenomena, 2nd Edition, New York: John Wiley
& Sons.
Hewitt, G.F., G.L. Shires and Y.V. Polezhaev, Eds. 1997. International Encyclopedia of Heat and Mass Transfer,
Boca Raton: CRC Press/Taylor & Francis Group.
Weast, R.C., Ed., 1970. CRC Handbook of Tables for Applied Engineering Science, Boca Raton: CRC Press/
Taylor & Francis Group.
PROBLEMS
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.
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.
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.
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?
5. The walls of a house are made from a half-inch sheet of gypsum sheetrock, 4 inches of
foam insulation, and a half inch of light fiberboard. Heat loss from the house during a win-
ter day occurs with an average convective heat transfer coefficient of 200 W/m2. The out-
side air temperature is −5°C.
a. If the inside temperature is to be maintained at 20°C, how much internal heating on
average is required per square meter of wall area?
b. If the house has 5,000 ft2 of wall area and the floor and ceiling are assumed adiabatic,
how much energy would the house require per day in kWh to stay heated at 20°C?
6. A steel plate in a manufacturing plant is removed from a furnace at a temperature of 650°C.
Hoss 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 coefficient is 30% higher
than the initial value where the plate was initially removed from the furnace.
7. 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/m2K. Estimate the win-
dow temperature when the outside air temperature reaches –10°C.
18 Advanced Heat Transfer
8. An aluminum sheet leaves a hot-roll section (called the entry point) of a mill at a tempera-
ture 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 temperature at a position
where the convection coefficient is 40% higher than the value at the entry point.
9. Heat loss occurs by convection through the surface of an object. The surface temperature
is 1,000°C while the surrounding fluid is at 100°C. The convective heat transfer coefficient
is 2,000 W/m2 K and the surface heat flux is 1,800 kW/m2.
a. Determine the heat flux by first converting both the surface temperature and surrounding
fluid temperatures to Kelvin units prior to finding ΔT and the heat flux. Then find the
heat flux without converting any temperatures to Kelvin. Explain why the results are the
same and correct even though the units are wrong.
b. Find the ΔT value in Celsius degrees and then convert it to Kelvin units prior to finding
the heat flux. Explain why this result is incorrect.
c. Then repeat the above questions but consider the same object in the same surroundings
while emitting radiation as a blackbody. Find the resulting radiative heat flux.
10. 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/m2K, respectively. Compare
the heat transfer by convection and radiation at surface temperatures of (a) 60°C, (b) 400°C,
and (c) 1,300°C.
11. 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 approximated 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?
12. Superheated steam at 240°C flows through an uninsulated pipe in a basement hallway at
20°C. The pipe emissivity is 0.7 and the coefficient of free convection is 20 W/m2K. Find
an expression for the total heat loss from the pipe (per unit length of pipe) in terms of the
pipe diameter.
13. 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 stor-
age 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 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 satu-
ration temperature).
14. 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.
15. 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 hours
of operation, assuming that the storage tank is well insulated, and all heat input is directed
to the vaporization process?
16. 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.
17. A cylindrical spacecraft 7.5 m long and 2.5 m in diameter is orbiting the earth where the
incoming solar radiation is 1,300 W/m2. The surface of the spacecraft has an absorptivity
and emissivity of 0.5 and 0.5, respectively. It can be assumed that the ship’s outer surface
has a uniform temperature.
Introduction 19
a. What is the steady-state temperature of the spacecraft if its longer side is exposed con-
tinuously to the sun?
b. What additional heat generation/removal is needed to maintain an average temperature
of 15°C when the longer side is exposed to the sun?
c. Repeat the analysis if the small circular end is exposed to the sun instead of the longer
side.
18. The wall of a walk-in freezer in a meat processing plant contains insulation with a thick-
ness of 10 cm and a thermal conductivity of 0.02 W/mK. The inside wall temperature is
–10°C. The outside air temperature and convection coefficient (outside) are 20°C and 5 W/
m2K, respectively. What additional thickness of insulation is required to reduce the heat
gain (into the freezer) by 10%?
19. 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/m2K and –5°C, respectively. Determine the outer wall temperature of the
concrete wall.
20. A thin plate is exposed to an incident radiation flux of 2 kW/m2 on its top surface. The bot-
tom 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.
21. 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/m2K, 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 thermocouple wire can be neglected.
22. 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/
m2K, respectively. For a specified copper cable (3-cm diameter), estimate the required heat
generation rate (per unit length of cable) to maintain the surface temperature above 0°C.
23. The top cover of a solar collector plate is exposed to ambient air at 20°C with a convection
coefficient of 6 W/m2K. A surface coating is developed and applied to modify the radiative
properties of this absorbing cover plate. The incident solar radiation 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 sur-
roundings (ambient air) relative to the convective heat loss from the plate?
24. Heating, ventilating, and air conditioning (HVAC) equipment maintains a room in a build-
ing 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.
25. Aircraft icing may occur when an aircraft passes through a cloud containing supercooled
droplets. Describe the heat flows to/from an iced aircraft surface and write an energy bal-
ance for the surface based on these heat flows.
26. 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.
20 Advanced Heat Transfer
27. 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.
28. Various technical challenges are encountered by thermal engineers in the design of more
efficient systems for heating and cooling of buildings. Perform a technical literature
(including journals and magazines) to describe the main technological advances and chal-
lenges facing thermal engineers in the development of next-generation HVAC systems.
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 investigated. Also, advanced
solution methods such as separation of variables, conformal mapping, 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 trans-
lational energies at a molecular scale. The temperature of the substance 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 tem-
perature. 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 Fourier’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 prescribes 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
that can flow across the surface. Also, the properties of the material affect the heat flow, particularly
the thermal conductivity. As discussed in the previous chapter, solids are better conductors than liq-
uids 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.
When graphically displaying temperature results, isotherms (lines of constant temperature) are
often used. Plotting the isotherms within a domain shows the regions of high and low temperatures.
Heat flows by conduction in a direction perpendicular to a local isotherm. In other words, heat flows
in a direction of steepest temperature descent. Thus, the heat flow lines and isotherms are mutually
perpendicular to each other. The heat flow pathlines may be constructed by joining perpendicular
crossing points between successive isotherms.
Topics to be covered in this chapter include the formulation and analysis of the heat equation,
thermal resistance networks, fins and extended surfaces, advanced solution methods for multidimen-
sional problems, and transient heat conduction. The reader is also referred to other classic books on
conduction heat transfer by Carslaw and Jaeger (1959), Yener and Kakac (2008), and Han (2011).
DOI: 10.1201/9781003206125-221
22 Advanced Heat Transfer
E cv E in E out E g (2.1)
From left to right, the terms represent the transient accumulation of energy within the control vol-
ume, 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, Ėcv, 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,
Ėin, 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 generation rate multiplied by the size
of the volume. Substituting these individual terms into the energy balance,
T
c p Adx qx A qx A qx A dx qAdx
(2.2)
t x
T
qx k (2.3)
x
Substituting this expression into the energy balance, canceling terms, and neglecting higher order
terms,
T T
c p Adx k
A dx qAdx (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 2T
cp k 2 q (2.5)
t x
In addition to the above governing heat equation, a problem definition also requires boundary
conditions for completion of the problem specification. For heat conduction 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 condi-
tion becomes
T qw
(2.6)
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 (character-
ized 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
conduction heat flux in the solid from Fourier’s law and the convective heat flux in the fluid yields
T
k
x
h T 0,t T f (2.7)
0
In a similar manner, a linearized radiation condition at a boundary, coupled with heat conduc-
tion in the solid, would lead to a Robin-type boundary condition. As expected, problems involving
a Robin condition are often more difficult since both the temperature and its gradient are unknown
at the boundary. Robin conditions often lead to a conjugate problem. A conjugate problem refers to
classes of problems where convection in the fluid and conduction in the adjacent solid are coupled
and must be solved simultaneously, typically through an iterative procedure involving the Robin
condition at the boundary. Both heat transfer in the solid and fluid must be solved separately and
then coupled together iteratively by the Robin condition.
where the terms represent, from left to right, the rate of energy change of the wall with time, and
rates of heat inflow and outflow from the wall, respectively. Under steady-state conditions, the tran-
sient term becomes zero, and therefore the rate of heat inflow equals the outflow, or in other words,
the heat transfer rate remains constant.
For a wall of thickness, L, and thermal conductivity, k, Fourier’s Law is expressed as
dT
qcond k (2.9)
dx
Since this heat transfer rate is constant, it implies that dT/dx is also constant, and so the tempera-
ture varies linearly with x. For steady-state heat conduction, the temperature distribution is linear
(see Figure 2.3).
Separating variables in Fourier’s Law and integrating across the wall,
L T2
q
0
cond
dx kAdT
T1
(2.10)
FIGURE 2.3 Heat conduction and temperature distribution through a plane wall.
Heat Conduction 25
T T
qcond kA 2 1 (2.11)
L
This indicates that the heat transfer rate is proportional to the thermal conductivity, wall area, and
temperature difference between the surfaces, but inversely proportional to the wall thickness, for
steady-state heat conduction.
V1 V2
I (2.12)
Re
where I, V1 − V2, and Re refer to the electrical current, potential (voltage) difference, and electrical
resistance, respectively. From Ohm’s law, the current flow is proportional to the driving 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 (2.13)
Rt
In this case, the analog of the electrical current is the heat transfer rate, q. Rather than a volt-
age 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 (2.13) imply
that the thermal resistance of heat conduction through a plane layer of thickness L can be written as
L
Rt ,cond = (2.14)
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 tem-
perature difference (between the surface and fluid) divided by the convection thermal resistance,
Rt,conv, where
1
Rt ,conv = (2.15)
hconv A
26 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 = (2.16)
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 demonstrates how the
thermal resistance can be calculated for a cylindrical geometry.
1 T
r 0 (2.17)
r r r
where r refers to the radial position. Integrating this equation twice yields the following general
solution,
T r C1 ln r C2 (2.18)
The coefficients of integration may be obtained by applying specified temperatures at the inner
and outer radii, T(r1) = T1 and T(r2) = T2, respectively, thereby yielding
T1 T2 r
T r ln T2 (2.19)
ln r1 /r2 r2
dT T1 T2 T T
qr kA 1 2 (2.20)
dr ln r2 /r1 / 2 Lk Rt ,cond
where L is the tube length. Adding this conduction resistance and the convection resistance yields
the following total thermal resistance per unit length of tube,
ln r /r1 1
Rt,tot (2.21)
2 k 2 rh
An optimal thickness can be obtained by minimizing this thermal resistance with respect to the
outer insulation radius. Differentiating the expression for the total thermal resistance and setting
the result to zero,
dRt,tot 1 1
0 (2.22)
dr 2 kr 2 r 2h
Solving this equation yields a radius of r = k/h for which the total thermal resistance is a mini-
mum. 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 geom-
etry. 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 temperature 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 cir-
cuits. 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 (2.13) to include multiple thermal resistances,
T 1
q ;U (2.23)
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 UAT (2.24)
Consider heat flow through a composite wall consisting of multiple sections in both series and
parallel as illustrated in Figure 2.5. The total resistance to heat flow through the composite wall, Rtot,
can be expressed by a combination of resistances in series and parallel as follows:
1
1 1
Rtot R1 R4 (2.25)
R
2 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 2.6, 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 asymptotes, 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 mul-
tiplied by the thermal conductivity. The shape factor and resulting heat flow rate are expressed by
1
S= (2.26)
Rt k
q Sk T1 T2 (2.27)
The previous solution method can be applied to different geometrical configurations to determine
their conduction shape factors. Various shape factors of common geometrical configurations are
summarized below in Table 2.1 and Figure 2.7.
Thermal contact between adjoining materials is usually imperfect as some contact resistance
is typically encountered at the interface between the materials. The contact resistance is generally
due to machined surfaces in contact that are not perfectly smooth. Some roughness elements are
formed at each interface. When rough surfaces come into contact, gaps between the contact points
and roughness elements lead to additional thermal resistances of convection and radiation across the
gaps which exceed the thermal resistance of pure conduction alone for perfectly smooth surfaces in
contact.
The thermal contact resistance is influenced by several factors, including contact pressure, inter-
stitial materials, surface roughness, cleanliness, and surface deformations. As the contact pressure
Heat Conduction 29
TABLE 2.1
Conduction Shape Factors for Various Geometrical Configurations
Geometrical Configuration Shape Factor
(a) Plane wall with a width of L and cross-sectional area of A A
S=
L
(b) Cylindrical shell with inner and outer radii of r1 and r2, 2 L
S
respectively ln r2 /r1
(c) Spherical shell with inner and outer radii of r1 and r2, respectively 4
S
1/r1 1/r2
(d) Isothermal horizontal cylinder of length L buried with its axis a 2 L
S
distance z below the surface (z/L ≪ 1 and D/L ≪ 1) cos h 1 2 z /D
(e) Isothermal sphere of diameter D buried a distance z below an 2 D
S
isothermal surface (z > D/2) 1 D / 4z
(f) Isothermal sphere of diameter D buried a distance z below an 2 D
S
insulated surface (z > D/2) 1 D / 4z
(g) Vertical cylinder of diameter D buried in a semi-infinite medium 2 z
S
to a depth of z ln 4 z /D
increases, the thermal resistance becomes smaller since the contact surface area between the adjoin-
ing 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 sur-
faces in contact, which influence heat flows through the gases/fluids filling these gaps. Also, surface
features such as roughness, waviness and flatness may be significant factors. Furthermore, surface
cleanliness and the presence of dust particles can also affect the thermal resistance.
30 Advanced Heat Transfer
FIGURE 2.7 Conduction shape factors for selected two-dimensional geometries [q = Sk(T1 – T2)].
Surface deformations, either plastic or elastic, also affect the contact resistance since the defor-
mation causes the contact area between adjoining surfaces to increase. In some cases, the sur-
face roughness at the contact interface may be quantified in terms of the mean roughness height
and waviness to determine an approximate surface profile. However, in many cases it is difficult
to accurately characterize the microscopic behavior of roughness elements in contact with each
other. Therefore, the thermal contact resistance is often best determined by experimental measure-
ments. Once an estimate of the thermal contact resistance is obtained, it can be included in the
thermal circuit as a component in series at the location of the interface between the two materials. A
Heat Conduction 31
comprehensive review of thermal contact resistance modeling and experimental data was presented
by Yovanovich (2005).
d
qx Ac qx Ac
dx
qx Ac dx dAsh T T (2.28)
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, and general
fin equation becomes:
d 2T 1 dAc dT 1 h dAs
dx 2 Ac dx dx Ac k dx
T T 0 (2.29)
In one-dimensional problems, analytical solutions of this fin equation can often be obtained.
Once the geometrical profile of the fin is known, the derivatives of the cross-sectional and surface
areas can be determined, after which the fin equation can be solved analytically subject to appropri-
ate boundary conditions. The following example presents an exact solution for a rectangular fin with
a specified base temperature.
d 2
m2 0 (2.30)
dx 2
where θ = T(x) – T∞, m2 = 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 specifica-
tion of the boundary conditions.
Applying the boundary conditions of an adiabatic fin tip with a specified base temperature, the
following results are obtained for the temperature excess, θ, and heat flow rate, qf, respectively:
cos h m L x
(2.32)
b cos h mL
d
qf kAc M tan h mL (2.33)
dx 0
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:
e mx (2.34)
b
qf = M (2.35)
Another type of boundary condition is a specified temperature at the tip of the fin, θ(L) = θL. In
this case, the results for the temperature and heat flux become:
Heat Conduction 33
cos h mL L / b
qf M (2.37)
sin h mL
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 effectiveness 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 dif-
ference between the fin and ambient fluid for convection. The fin efficiency is written as
qf qf
f (2.38)
q f ,max hAs b
For the previous example of a uniform cross-sectional area fin with an adiabatic tip, the expres-
sions for the actual and maximum heat transfer rates can be substituted above to yield:
tan h mL
f (2.39)
mL
From this result, it can be observed that the fin efficiency decreases with longer fins, since mini-
mal heat transfer occurs when the temperature along the fin approaches the fluid temperature in long
fins.
The fin efficiency is often depicted in graphical form in terms of an abscissa, Xc, where
1/ 2
h
X c L3c/ 2 (2.40)
kAp
which represents a dimensionless characteristic length of the fin. The fin efficiency, ηf, typically
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 2.10 for various fin geometries. All design parameters that lead to a temperature distribution
closer to the base temperature (i.e., higher fin conductivity, 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 2.10,
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 2.10 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
34 Advanced Heat Transfer
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 coef-
ficient), 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 cri-
teria. 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.
Since heat transfer occurs through the fin and across the base surface, the associated thermal resis-
tance, Rf, and resistance of the base surface, Rb, are also useful in characterizing the fin performance,
1
Rf (2.41)
f h f Af
1
Rb = (2.42)
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 (2.43)
R f Rb hb Ab f h f A f o hA
Heat Conduction 35
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 convection coefficients are
equal,
hb Ab f h f A f A
o 1 f 1 f (2.44)
hA A
This overall surface efficiency can then be substituted back to obtain the overall equivalent resis-
tance of the fin, Re. Using this resistance in Equation (2.13), 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
Kocabiyik (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).
2.5 MULTIDIMENSIONAL CONDUCTION
2.5.1 Cartesian Coordinates
The previous sections have examined heat transfer in one-dimensional systems, where the tempera-
ture variations occur predominantly in only one spatial direction. In practice, heat transfer occurs
in multiple spatial directions simultaneously. Recall the one-dimensional form of Fourier’s law was
presented in Chapter 1. The generalized form of Fourier’s law for three-dimensional heat conduction
can be written as
T T T T T T
q kT k , , k i j k (2.45)
x y z x y z
where ∇ represents the vector gradient (or nabla) operator, and i, j, and k refer to unit vectors in the
x-, y-, and z-directions, respectively. The bold font is used to designate a vector quantity and distin-
guish 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/m2) 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 concentration gradient (called Fick’s law).
Consider a two-dimensional energy balance of conduction heat flows into/out of a differential
control volume as illustrated in Figure 2.11. 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 differ-
ential 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 2.11.
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:
36 Advanced Heat Transfer
T 2T 2T
cp k 2 2 q (2.46)
t x y
where q̇ refers to a volumetric heat source. Under steady-state conditions without any heat genera-
tion, the two-dimensional heat equation is reduced to the following Laplace equation:
2T 2T
0; 2T 0 (2.47)
x 2 y 2
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 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 problem 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 tem-
peratures, 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 Laplaces’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 origi-
nal overall problem. The principle of superposition is a useful tool when analyzing complex linear
problems involving a combination of different types of boundary conditions.
most generalized coordinate system from which other curvilinear coordinate systems are special
cases.
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
r x 2 y 2
x r cos u1 r
1 y
y r sin tan u2 (2.48)
z z x u z
z z 3
x r sin cos r x y z
2 2 2
u1 r
tan y /x
1
y r sin sin u2 (2.49)
z r cos
1
cos z / x y z
2 2 2
u
3
In order to establish the heat conduction equation in curvilinear coordinates, an energy balance
is required, Equation (2.1), as well as the arc lengths for each side of a differential control volume.
Figure 2.13 illustrates a curvilinear coordinate system. Using the chain rule to express each of the
Cartesian coordinates in terms of the curvilinear coordinates, xi = xi(u1, u2, u3),
xi x x
dxi du1 i du2 i du3 (2.50)
u1 u2 u3
FIGURE 2.13 General coordinate system and heat flow through a differential flux tube in curvilinear
coordinates.
ds dx dy dz
2 2 2 2
(2.51)
Substituting the individual arc length expressions, dxi, and rearranging, yields the following arc
length in the curvilinear coordinate direction si,
2 2 2
x y z
gi (2.53)
ui ui ui
= =
dA1 ds2 ds3 g2 g3 du2 du3 (2.54)
= =
dA2 ds1ds3 g1g3 du1du3 (2.55)
= =
dA3 ds1ds2 g1g2 du1du2 (2.56)
= =
dV ds1ds2 ds3 g1g2 g3 du1du2 du3 (2.57)
where g = g1 g2 g3 is called the Jacobian determinant of the transformation from Cartesian to curvi-
linear coordinates.
Then Fourier’s Law and the heat conduction equation can be represented in terms of the above
relations. Define a1, a2, and a3 as the unit vectors corresponding to the u1, u2, and u3 directions. Then
Fourier’s Law can be written as
3
T T T 1 T
q kT k
1s
a1
s2
a2
a2 k
s2
i 1
ai
gi ui
(2.58)
Heat Conduction 39
The heat conduction equation can be obtained by performing an energy balance on a differential
control volume, based on Equation (2.1) for steady-state conditions without heat generation, and
applying Fourier’s Law, thereby yielding
3
T g T
u k g
1
cp q (2.59)
t g i i ui
i 1
Using this general form, the heat conduction equation can then be expressed in any of the curvi-
linear coordinate systems.
The 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 T
cp k y k y x k z q (2.62)
t x x
T k T T
q k ir j k kz (2.63)
r r z
T k T k 2T 2T
cp r 2 k 2 q (2.64)
t r r r r 2
z
T k T k T
q k ir j k (2.65)
r r r sin
T k 2 T k 2T k T
cp r r 2 sin 2 2 r 2 sin sin q
(2.66)
t r 2 r r
A similar approach as the previous section can be used to determine the conduction shape factor
for geometrical configurations in orthogonal curvilinear coordinates.
40 Advanced Heat Transfer
Consider a heat flow rate of Q1 in the u1 direction through a differential flux tube bounded in ui
coordinates (see Figure 2.13). The total thermal resistance through the flux tube is given by
T T dT dT /ds1 ds1 ds
d R1
3
1 (2.67)
Q1 k dT /ds1 dA1 kdA1
where the superscript (3) designates a differential of order 3. Expressing the arc length and differen-
tial surface area in terms of the metric coefficients,
1 g1 du1
d R1
3
(2.68)
k g du2 du3
Then the total thermal resistance of a flux tube between u1 = a and u2 = b becomes
b
1 g1
d R1
3
du1 (2.69)
kdu2 du3 g
a
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
(2.70)
kR1
u3 u2 g1 / gdu1
a
This result represents the shape factor for regions bounded by curvilinear coordinate lines.
Teerstra, Yovanovich, and Culham extended this approach to calculate shape factors for various
three-dimensional geometries (2005).
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 (2.70), the shape factor in each of the cur-
vilinear coordinate directions is determined as follows:
u1 r g1 1
L /2
d dz L /2
Case i : T T r u2 g2 r Sr
2
(2.71)
r2
ln r2 /r1
u z
3
g 1
3
0 0
r1
1/r dr
u1 z g1 1
r2 / 2
d dr
r22 r12
Case ii : T T z u2 g2 r Sz
2
L (2.72)
4L
u r
3
g 1
3
r1 0
0
1/r dz
u1 g1 r 2
L ln r2 /r1
L r2
drdz
Case iii : T T u2 r g2 1 S /2
/2
(2.73)
u z
3
g 1
3
0 r1
0
rd
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 boundary through cylindrical
coordinates, rather than Cartesian coordinates. In numerical methods for heat transfer and fluid flow,
Heat Conduction 41
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.
T x ,y X x Y y (2.74)
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 C X x Y y
i
i i i (2.75)
which is also a solution for suitable choices of the constants, Ci. The separable assumption allows
the governing equation to be reduced from a partial differential equation involving 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 following example
further describes how the two steps of the method of separation of variables are applied.
2T 2T
0 (2.76)
x 2 y 2
42 Advanced Heat Transfer
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 (2.77)
y
y 0
T hb
y
k
T x ,b Tf (2.78)
y b
T h0
x
k
Tb T 0,y (2.79)
x 0
T ha
x
k
T a,y Tf (2.80)
xa
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:
X Y XY 0 (2.81)
or,
X Y
(2.82)
X Y
Since the left side is a function of only x and the right side is a function of y, both sides must be
constant - let the constant equal k = λ2. The constant must be positive because the function Y
should contain a periodic solution in order to satisfy the first two boundary conditions. Setting
each side of the above equation to a constant yields two ordinary differential equations.
Therefore the original heat equation (a partial differential equation) has been reduced to two
separated ordinary differential equations with exponential and trigonometric solutions for X(x)
and Y(y), respectively,
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,
Heat Conduction 43
b
cot b (2.85)
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 boundaries, an
infinite series is formed by superposing all possible solutions to meet the boundary conditions,
T x ,y A cos h x B sin h x cos y
n 1
n n n n n (2.86)
where λn refers to the set of n eigenvalues. Applying the right boundary condition along the sur-
face x = a yields
Bn An n (2.87)
where
At the asymptotes, Bia = 0 and Bia → ∞, the coefficients become ψn = tan h(λa) and ψn = cot
h(λna), respectively. If λna > 2.65, then tan h(λa) ≈ cot h(λa).
The temperature solution can then be rewritten in terms of only one unknown coefficient:
T x ,y A cos h x
n 1
n n n
sin h n x cos ny
(2.89)
A
h0 h
cos ny 0 Tb Tf (2.90)
k
n n n
n 1
k
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 result-
ing 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:
An
Tb Tf sin nb h0 (2.91)
k nn nb sin nb cos nb
This coefficient then yields a final closed-form solution of the temperature field.
cos h x sin h x cos y
T x ,y T T sin b h
n n n n
(2.92)
k b sin b cos b
b f n 0
n 1 n n n n
n
44 Advanced Heat Transfer
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
T
q 2 k
0
y x 0
dy 2k A
n 1
n n sin nb
(2.93)
2.7 NONHOMOGENEOUS SYSTEMS
In the previous section, homogeneous equations and boundary conditions were considered. For a
homogeneous function, if all arguments are multiplied by a given factor, then the resulting value is
multiplied by some power of this factor. In contrast, a nonhomogeneous function or system does not
have this multiplicative scaling behavior.
Nonhomogeneous problems can be transformed into a homogeneous system through an appro-
priate change of variables (Jiji, 2009). For example, consider a nonhomogeneous problem of tran-
sient one-dimensional heat conduction with internal heat generation at a rate of q and specified
surface temperatures of TH and TL (see Figure 2.15).
The governing equations, boundary conditions, and initial conditions are given by
FIGURE 2.15 Schematic of heat conduction with internal heat generation in a plate.
Heat Conduction 45
1 T 2T q
(2.94)
t x 2 k
subject to:
Although the method of separation of variables cannot be applied directly in the usual manner
since the governing equation and boundary conditions are nonhomogeneous, an alternative form of
solution can be assumed as follows:
The function G(x) is included to account for the additional heat generation term. Substituting the
assumed profile into the heat conduction equation,
1 F 2 F d 2G q
2 2 (2.97)
t x dx k
For one equation and two variables, F and G, let the equation be separated and solved in two parts
(one for F and another for G).
1 F 2 F
2 (2.98)
t x
d 2G q
0 (2.99)
dx 2 k
q x
G x W x x TH TL W TL (2.100)
k
Using a product solution and method of separation of variables in the former equation for F(x,t),
assume a functional form as follows:
F x,t X x t (2.101)
Substituting into the heat conduction equation, separating variables, and equating the resulting
equations to a constant, ± λ2,
d2X
2X 0 (2.102)
dx 2
d
2 0 (2.103)
dt
n t Cne n t
2
(2.105)
The boundary condition of F(0,t) = 0 implies Bn = 0. The other condition of F(W,t) = 0 yields
eigenfunctions of λn = nπ/L, where n = 1, 2, 3, … Then substituting the values of F(x,t) and summing
all solutions,
F x,t a e
n 1
n
n2 t
sin n x
(2.106)
where the constants are determined from the remaining initial condition,
F x,0 Ti G x (2.107)
thereby yielding
an 2
1 1
n
Ti TH
2 1
2
T T
qL
2 2
2 n 2 1 2 qL
2 n
(2.108)
H L
n n n
3
2k k
TH T 2 x 2 x
n2 2 t / L2 n x
T
x qL an
1 e sin (2.109)
TH TL L 2k TH TL L L L TH L
n 1
At the steady state, t → ∞, the temperature approaches the expected solution with internal heat
generation and specified wall temperatures at x = 0 and x = W. If the initial temperature is not uni-
form, then Ti can be replaced by Ti(x) in the above solution procedure. This procedure allows non-
homogeneous problems to be solved in a similar manner as a homogeneous system.
2.8 CONFORMAL MAPPING
The method of conformal mapping is another useful method for solving the two-dimensional steady-
state heat equation, Laplace’s equation in Equation (2.47). A conformal mapping, also called a con-
formal transformation, or biholomorphic map, is a functional transformation between two different
coordinate systems which preserves local angles during the mapping process. A functional mapping
that preserves the magnitude of angles, but not their orientation, is called an isogonal mapping. The
method of conformal mapping will be briefly introduced in this section.
Conformal mappings are based on functional analysis of complex variables. A complex number
can be expressed in the form z = x + iy, where x and y are real numbers, and i is the “imaginary
number” (i2 = –1). The real part of the complex number is denoted by Re(z) = x and the imaginary
part is denoted by Im(z) = y.
Geometrically, complex numbers can be identified in a two-dimensional complex plane by using
the horizontal axis for the real part and the vertical axis for the imaginary part (see Figure 2.16).
The complex number, z, can be identified with the point (x, y) in the complex plane. For example, a
complex number whose real part is zero is purely imaginary and the point lies on the vertical axis
of the complex plane.
Heat Conduction 47
A complex number can also be represented in polar form. The distance of the point (x, y) from the
origin is denoted by r (magnitude or modulus) and the angle of the point with respect to the positive
real axis is called the argument of the complex number, θ. The complex number can be represented by
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, sub-
tracted, and multiplied, as polynomials in the variable i.
In the method of conformal mapping, the solution of the heat equation is obtained by transform-
ing 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 the mapping of w = g(z),
where g′(z) must be nonzero. Table 2.2 and Figure 2.17 illustrate a number of common functions of
conformal transformations.
In Table 2.2, 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 (2.111)
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 be shown that the
bilinear mapping can be rewritten again in terms of an implicit transformation as follows:
z z1 z2 z3 w w1 w2 w3
(2.112)
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 prescribed points in either
the z-plane or w-plane. For example, if w3 = ∞, then the bilinear mapping becomes:
w2 w3
1 (2.113)
w w3
48 Advanced Heat Transfer
TABLE 2.2
Conformal Mapping of Various Geometrical Configurations
Geometrical Function Conformal Transformation
(a) Linear translation function w = f(z) = z + b
(b) Linear magnification function w = f(z) = az
(c) Linear rotation function w = f(z) = zeiα
(d) Mapping of the vertical strip | x | < π/2 onto a w-plane slit along w = f(z) = sin (z)
the rays u ≤ −1, v = 0 and u ≥ 1, v = 0
(e) Mapping of the vertical strip | x | < π/4 onto the unit disk | w | < 1 w = f(z) = tan (z)
(f) Mapping of the angular strip − π ≤ y ≤ π in the z plane onto the w = f(z) = exp (z)
w plane with the point w = 0 deleted
(g) Bilinear (Mobius) mapping of a disk bounded by complex az b
w f z
points a, b, c and d onto a half-plane cz d
(h) Schwarz-Christoffel mapping of the upper half-plane with dw
A z x1 z x2 2 z xn1 n1
1 / / /
boundary points x1, x2, etc. to a closed polygon with internal dz
angles αj and corresponding vertices wj and exterior angles βj
(i) Joukowski mapping of an eccentric circle of semi-axis length k k2
from the origin to an ellipse or airfoil w f z z
z
z 2i 2i 0 w i / 2 (2.115)
z 0 2i 2i i / 2 i / 2
which yields an inverse mapping of w = 1/z from the z-plane to the w-plane. This result is a spe-
cial case of the bilinear mapping in Figure 2.17 with a = 0 = d and b = 1 = c.
The temperature distribution in the z-plane, exterior to two circles, can then be found by this
inverse mapping to a simpler geometrical domain - a horizontal strip in the w-plane (−1/2 < v <
1/2) - where an analytical solution of the heat equation is more readily available. The original
coordinates (x, y) in the z-plane are mapped onto new coordinates (u, v) in the w-plane. The
coordinate transformation is given by
1
w f z u iv (2.116)
z
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 (2.117)
x iy x iy
50 Advanced Heat Transfer
Then, equating the real and imaginary parts to u and v, respectively, in the w-plane,
x y
u ; v (2.118)
x2 y 2 x2 y 2
2 2
0 (2.119)
u 2
v 2
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 (2.121)
x2 y 2
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.
Conformal mappings are useful because they allow heat conduction problems with complicated
boundaries to be mapped into simpler geometrical domains where Laplace’s equation 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 corresponding 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 conformal mapping can be used to determine the temperature distribution
in a region exterior to two circles.
The method of conformal mapping can also be applied to the transformation of a standard coor-
dinate system to other orthogonal curvilinear coordinates, for example, elliptic cylindrical and
bipolar coordinates (see Figure 2.19). An orthogonal coordinate system is a 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 coor-
dinates 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) coordinates in the w-plane.
z f w a cosh i
2
e
a i
e
i
(2.122)
Heat Conduction 51
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
cos sin 1
2 2
(2.124)
a cosh a sinh
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
cosh sinh 1
2 2
(2.125)
a cos a sin
x a cosh cos u1
g1 a 2 cosh 2 cos2
y a sinh sin u2 g2 g1
z z u z g 1 (2.126)
3 3
Then the thermal resistance and shape factors for an elliptic cylindrical element can be deter-
mined accordingly. For example, consider a region in Figure 2.19 that represents a section bounded
by two isothermal confocal ellipses (η1 ≤ η ≤ η2; 0 ≤ ψ ≤ 2π; 0 ≤ z ≤ L). Using Equation (2.70), the
shape factor for the region is given by
52 Advanced Heat Transfer
L 2
1 du3du2 d dz 2 L
S
g / g du
b 2
2 1
(2.127)
kR
u2 u3 1 1 0 0 d
a 1
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 solution in
η-coordinates to the original (x, y) elliptic cylindrical domain. The one-dimensional heat equation
for the temperature field, Ψ, in η-coordinates is given by
d 2
0 (2.128)
d 2
1
(2.129)
2 1 2
Using this temperature distribution in Equation (2.58), the heat flux along on the boundary η = 0
is determined as
1 d k 1
1/ 2
q k 2 cosh cos
2 2
(2.130)
g d 0 a 2 1
Alternatively, mapping back to the original (x, y) coordinates, the temperature distribution can
be written as
T T1 cosh x /a
1
(2.131)
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 2.19):
i
z f w ia cot (2.132)
2
The metric coefficients for the transformation from Cartesian to bipolar coordinates are given by
Then the shape factor for the region bounded by two eccentric cylinders becomes:
L
1 du3du2 d dz 2 L
S
g / g du
b
2
2 1
(2.134)
kR
u2 u3 1 1 0 d
a 1
If the transformation of variables between z-and w-planes cannot be handled by analytical means
alone, then a point-by-point numerical matching and integration procedure can provide a numerical
solution based on the method of conformal mapping. For example, a numerical solution of a confor-
mal mapping between concentric polygonal regions and concentric cylinders was presented based
on the Schwarz-Christoffel transformation (Naterer, 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 (2.135)
Ts T f 1/h k
where Bi refers to the Biot number. The Biot number (Bi) is a dimensionless parameter that charac-
terizes 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 significant 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 2.21). 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
c pV hAs T T (2.136)
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 con-
stant throughout the solid object.
Define the temperature excess, θ, as the difference between the solid and fluid temperatures, so
that integrating both sides of the above energy balance, subject to an initial condition of θ(0) = θi
yields:
T t T
e t / (2.137)
i Ti T
cP L
(2.138)
h
The characteristic length, L, may be regarded as the ratio between the object’s volume to its sur-
face area. The time constant, τ, characterizes the object’s rate of temperature change with time. If τ is
small, the temperature change is rapid, and conversely, if the convection coefficient is small relative
to the thermal capacitance, then τ is large such that the rate of temperature change with time is slow.
Alternatively, the solution may be expressed as
ht
exp exp Bi Fo (2.139)
i cpL
where Fo refers to the Fourier number (Fo = αt/L2), 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.
2.9.2 Semi-Infinite Solid
For larger Biot numbers above 0.1, significant spatial variations of temperature occur throughout the
solid and hence the lumped capacitance approximation cannot be used. Both transient and spatial
derivatives must be included in the heat equation. In some cases, analytical solutions can be obtained
by separation of variables or other direct solutions of Equation (2.5).
Consider transient heat conduction in a one-dimensional semi-infinite domain, subject to three
types of boundary conditions in Figure 2.22. This configuration can also represent heat conduc-
tion in a finite one-dimensional domain at early stages of time before any effects of the opposite
end boundary are transmitted inward. The three boundary conditions in Figure 2.22 represent: (i) a
constant surface temperature; (ii) constant surface heat flux; and (iii) surface convection condition.
The governing heat equation is given by
1 T 2T
(2.140)
t x 2
x
(2.141)
2 t
56 Advanced Heat Transfer
FIGURE 2.22 Transient temperature distributions in a semi-infinite solid: (i) constant surface temperature;
(ii) constant surface heat flux; and (iii) surface convection condition.
Using this similarity variable, the heat equation can be reduced to an ordinary differential equa-
tion in terms of η alone.
d 2T dT
2 (2.142)
d 2 d
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
q0 t k (2.144)
x 0 t
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 pro-
files were illustrated in Figure 2.22. Temperatures within the solid approach the surface temperature,
Ts, over time and the wall heat flux decreases with t1/2.
T x,0 Ti (2.145)
T
k q0; T ,t Ti (2.146)
x 0
For this problem, it can be shown that the exact solution is given by
Heat Conduction 57
2q0 t x 2 xq0 x
T x,t Ti exp erfc (2.147)
k 4 t k 2 t
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. Referring to Figure 2.22 and
the above analytical solution, it can be observed that the wall temperature increases with t1/2.
T x,0 Ti (2.148)
T
k
x
h T T 0,t ; T ,t Ti (2.149)
0
T x,t Ti x hx h2
t x h
t
1 erf exp 2 1 erf (2.150)
T Ti 2
t k k 2
t 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 temperatures
approach T∞ with increasing time. The wall heat flux decreases with time due to the decreasing dif-
ference between the fluid and wall temperatures. Sample dimensionless temperature profiles for this
convection case (iii) are illustrated in Figure 2.23.
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 (2.150) in case (iii) becomes zero and
the analytical solution approaches case (i) for a constant wall temperature.
FIGURE 2.23 Temperature profiles in a semi-infinite domain with surface convection (adapted from
Schneider, 1957).
58 Advanced Heat Transfer
2.9.3 Finite Regions
Semi-analytical solutions for unidirectional transient heat conduction can also be obtained for other
geometrical configurations such as a finite thickness plane layer, long cylinder, and sphere. Solutions
are often illustrated in a graphical form through Heisler charts (1947). A Heisler chart typically
illustrates the variation of temperature at a particular position(s) at different times.
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 symmetrical 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 2.24).
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 temperature 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 qw L / k Ts Ti , with time for a semi-infinite solid, plane wall, infinite cylinder,
sphere, and exterior objects are illustrated in Figure 2.25.
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 prob-
lems are often separable, and as a result, can also be analyzed by the method of separation of vari-
ables. Using separation of variables, the temperature solution can be subdivided into a product of
functions of spatial coordinates and time. Further analysis and results of transient and multidimen-
sional heat conduction problems are presented by Bergman et al. (2011), Kakac and Yener (1988),
and Schneider (1957).
FIGURE 2.24 Temperature distributions in (a) a plane wall of thickness 2L and (b) long cylinder (adapted
from Heisler, 1947).
Heat Conduction 59
FIGURE 2.25 Transient wall heat flux for various geometrical configurations (adapted from Bergman et al.,
2011).
FIGURE 2.26 Schematic of (a) discrete and (b) continuous change of transient boundary conditions.
60 Advanced Heat Transfer
An auxiliary problem is defined wherein the boundary condition is constant. Writing T x,y as
the solution of the auxiliary problem, then the T1 solution is given as
This represents a solution to the original problem with the transient boundary condition. Similarly,
further solutions, T3, T4, etc. can be added if other step changes occur in temperature at t = τ2, τ3,
and so forth.
For the case of a continuously varying boundary condition, f(t) represents a general function such
as a time-dependent heat flux or surface temperature. The continuous distribution is approximated
by a superposition of solutions of consecutive discrete steps, Δf, corresponding to time steps, Δt.
Initially, the solution of the first problem, T0, starts at t = 0, where
f i
n
T x,t f 0 T x,t
i 1
i
T x,t i i
(2.157)
df
t
T x,t f 0 T x,t
0
d
T x,t d
(2.158)
Assuming an initial temperature of T x,0 0 in the plate, and using integration by parts,
T x,t
t
T x,t f 0 T x,t f
0
t
d
(2.159)
For a non-zero initial temperature, Ti, a new variable is defined as T – Ti so the same above pro-
cedure can be used. Although this discussion has focused on a time-varying boundary temperature,
the method can also be applied to other transient parameters in the boundary condition such as the
convective heat transfer coefficient or heat flux.
Heat Conduction 61
If there is a discontinuity in the boundary condition, for example with f(t) = f1(t) up to t < ta, but
afterwards f(t) = f2(t) for t > τa, then the previous solution must be modified by adding another solu-
tion, Ta(x,t), which accounts for the step change at t = ta. This additional solution is expressed as
T x,t T x,t
ta t
T x,t f2 t a f1 ta T x,t ta f1
0
t
d f2
ta
t
d
(2.161)
This approach can be extended similarly if there are multiple discontinuities. Using this method,
Duhamel’s superposition integrals can be used to extend solutions with constant boundary condi-
tions to problems involving time-dependent boundary conditions.
Vf Af W Af
=
P = = (2.162)
V AW A
As 1 P A (2.163)
E CV E in E out E g (2.164)
where individual inflow and outflow terms for the solid and fluid pores are illustrated in Figure 2.27.
Assembling these terms into the energy equation,
2 2 2
T
i 1
i ci fi
t
dx qcond,i fimh x,in qcond,i fimh xdx,out qAdx
i 1 i 1
(2.165)
where fi is the volumetric fraction of phase i = 1 (solid) and i = 2 (fluid) and qcond,i is the Fourier heat
flux in phase i,
dT
qcond ,i ki A (2.166)
dx
qcond ,i
qcond ,i qcond ,i dx (2.167)
x dx
x x
Also q is the volumetric heat generation rate. A volume-averaged property, φ , of the solid-fluid
matrix can be expressed as
1 P s P f (2.168)
f T q
1 T 2T mc
(2.169)
t x 2 Ak x k
Boundary conditions may involve a specified temperature or convection at the outlet boundary. If
an inlet supply reservoir at a temperature Tr supplies fluid to a porous wall,
T 0,t
f Tr T 0,t kA
mc (2.170)
x
This boundary condition represents an energy balance on the fluid between the inlet to the wall
(left side) and the outlet of the reservoir (right side).
Porous materials have been increasingly used in heat transfer applications due to their high sur-
face area-to-volume ratios. Various models have been proposed for the pore structures, such as
simplified channels earlier in this section, networks of capillaries, or arrays of solid particles (ran-
dom close packing of spheres). Subsequent chapters will consider additional transport phenomena
involving porous media flow with boiling (Chapter 5), solidification (Chapter 7) and chemical reac-
tions (Chapter 8).
network of arteries and heat is exchanged between arteries and veins in capillary bleed-off through
the vessel walls. In addition, metabolic heat generation occurs such as sweating through glands onto
the skin surface.
Figure 2.28 illustrates how blood is transported through vessels and arteries in living tissue.
When blood leaves the heart, it passes through the aorta (diameter ~5 μm), main supply arteries
and veins to muscles (300–1,000 μm diameter), secondary arteries (50–100 μm diameter), arteries
(20–40 μm diameter), and finally the smallest vessels, capillaries (5–15 μm diameter). Then blood
flow returns to the heart through veins, mostly through counter-flow channels parallel to the arteries.
Similarly, as the arteries, the veins proceed from a system of vessels including capillaries, venules,
secondary veins, and primary veins.
Modeling of heat transfer in living tissue, veins, and arteries was initially formulated by Pennes
(1948), extended afterward by other investigators (Chen and Holmes, 1980), and discussed by Jiji
(2009). Consider a blood-perfused tissue element of a capillary bed in Figure 2.28. It is assumed
that heat transfer between blood and tissue occurs within the capillary beds, arterioles, and venules.
Blood flow within the capillaries is assumed to be isotropic so directionality effects are neglected.
Under these assumptions, Pennes (1948) modeled the blood heat exchange as a heat source (heat
generation term) that is proportional to the difference between the local tissue temperature, T, and
body core temperature, T0, as well as the blood flow rate, m b. This net rate of energy added by the
blood per unit volume of tissue, qb, can be written as
where cb is the specific heat of blood. Another heat source is the rate of metabolic energy generation
per unit volume, q m .
Then the heat conduction equation can be expressed in its standard transient form including two
heat generation terms, qb and q m , as follows:
T 2T 2T
cp k 2 2 m bcb T0 T q m (2.172)
t x y
where ρ, cp and k, without a subscript, refer to tissue properties. Despite its simplifying assump-
tions, this Pennes equation has been adopted successfully in tissue regions with vessel diameters
more than 500 μm for applications such as blood perfusion, cryosurgery, and hyperthermia therapy.
The original analysis of Pennes (1948) applied this thermal model to predict arterial blood and
tissue temperatures in a forearm. The following case study examines this original configuration and
determines the resulting temperature distribution.
k d dT
r
r dr dr mbcb T0 T q m 0 (2.173)
dT dT
0; k h T T (2.174)
dr r 0 dr r R
r T T0 m c R2 q mR 2
; ; b b ; (2.175)
R T T0 k k T0 T
1 d d
0 (2.176)
d d
d d
0; Bi 1 1 (2.177)
d d
0 1
The Biot number is Bi = hR/k. Solving this second-order ordinary differential equation,
C1I0
C2K0
(2.178)
where I0 and K0 are Bessel functions and C1 and C2 are determined from the boundary conditions,
yielding:
Bi 1 /
I0
(2.179)
I1 BiI
0
It can be confirmed that the entire forearm reaches T0 under the limiting case of no convective
cooling (h = 0). The relative effects of blood perfusion and metabolic heat generation can be
determined by setting λ = 0 and q m = 0, respectively.
Heat Conduction 65
Further enhancements of the Pennes model were reported by several investigators. Chen and Holmes
(1980) considered the effects of blood flow directionality in the vascular geometry by including
terms that account for the perfusion rate within local vessel branching and directional convective
effects. Jiji et al. (1984) reported a three-temperature model to account for three elements of blood
perfused tissue, namely the arterial, venous, and tissue temperatures. These simplified models pro-
vide a useful basis for more detailed examination and extensions to heat and fluid flow in other
biomedical applications.
2.13 MICROSCALE CONDUCTION
Heat conduction at the micro-scale is significantly different than the macro-scale since both bulk
values of thermal conductivity and the classical Fourier’s Law are no longer applicable for time
scales or systems that are very small. Microscale systems typically have smaller values of thermal
conductivity than bulk systems. In some cases, this can be advantageous such as with insulation
coatings, however, it can become undesirable in other applications that aim to enhance the heat
transfer rates such as microelectronics cooling.
Fundamental concepts in microscale conduction are the wavepacket and phonon. A wavepacket
is a wave-like disturbance or energy carrier that propagates through a medium with both wave-like
and particle-like properties. A phonon is a collective excitation of atoms or molecules in solids and
some liquids, similar to the concept of photons or quantized light waves.
Figure 2.30 illustrates a range of length scales of wavepackets. If the characteristic length of the
medium, L, is comparable or smaller than the mean free path, Λ, between wavepacket collisions,
then the frequency of collisions is increased by scattering on the boundaries. As illustrated in Figure
2.30 (b), the wavepacket has particle-like properties in this case. If the characteristic length is further
reduced and comparable with the wavelength, λ, of the wavepacket, as shown in Figure 2.30 (c),
then a more wave-like analysis is required.
Various theoretical approaches can be used to analyze microscale conduction including classical
kinetic theory of gases, molecular dynamics, and the Boltzmann equation. For example, Fourier’s
Law and an expression for the thermal conductivity in microsystems can be derived based on kinetic
theory of gases. Collisions between molecules cause individual particles to undergo a “random
walk” with an average distance between collisions that is represented by the mean free path as
follows:
v0 (2.180)
where v0 and τ refer to the speed of particles and corresponding mean free time, respectively.
FIGURE 2.30 Wavepacket collisions with (a) bulk, (b) particle-like and (c) wave-like behavior.
66 Advanced Heat Transfer
Recall the illustration in Section 1.4 two layers (gas, liquid or solid) exchanging thermal energy
across an imaginary boundary at the midpoint between both layers Define two control volumes, each
of width, Λ, to the left and right of the midplane at x = x0. These boundaries of the control volumes
are located at x0 – Λ (left control volume) and x0 + Λ (right control volume). Heat flows from left to
right as particles are hotter in the left control volume.
Through translational motion of particles, approximately half of the particles exit through each
of the control volumes after a time of t = Λ/v0. Then the net energy exchange across the midplane
at x = x0 is given by
1 1 1
ELR EL E R A eL ,avg eR,avg (2.181)
2 2 2
where E, e, A, and the subscripts L, R, and avg refer to extensive and intensive (per unit volume)
internal energies, cross-sectional area, left, right, and average, respectively.
Using a Taylor series approximation of the internal energy difference,
de
eL ,avg eR,avg (2.182)
dx x0
Higher order terms are neglected. From a detailed analysis of three-dimensional locations and
collisions among particles, it can be shown that the above expression is more accurately represented
with a preceding coefficient of –2/3.
Assembling these results for the heat flux, and rewriting in terms of temperature using the spe-
cific heat,
ELR A 2 de 1 dT
q 3 dx 3 cv v0 dx (2.183)
A 2 A
This represents Fourier’s Law at the micro-scale with a corresponding thermal conductivity that
is expressed as
1
k cv v0 (2.184)
3
The mean free path, Λ, decreases for a smaller size of microstructure and therefore the thermal
conductivity decreases without affecting cv or v0.
In this analysis, the mean free path, Λ, and particle speed, v0, vary with the type of heat-conducting
medium, e.g., ideal gas, metal, insulator, fluid. For ideal gases, the speed distribution depends on
temperature and can be approximated by the speed of sound,
v0 RT (2.185)
where γ = cp/cv. For metals, free electrons carry thermal energy as the primary mechanism of heat
transfer. Typical values of the particle speed are 1 − 2 × 106 m/s. In electrical insulators and semicon-
ductors, thermal energy is transported by atomic vibrations within crystals rather than free electrons.
Sound waves represent an important class of atomic vibrations but have wavelengths much larger
than the interatomic spacing in the crystalline structure.
The processes of particle scattering within a medium are complex and influence the mean free
path. For metals at room temperature, electron scattering occurs primarily through collisions with
sound waves. The mean free path is typically inversely proportional to temperature and within the
Heat Conduction 67
order of tens of nm at room temperature. However, effects of scattering from defects, impurities,
and grain boundaries may also become significant. Further detailed treatment of these micro-scale
processes for ideal gases, metals, electrical insulators, thin films, and other common materials is
presented by Jiji (2009).
REFERENCES
Bergman, T.L., A.S. Lavine, F.P. Incropera and D.P. DeWitt. 2011. Fundamentals of Heat and Mass Transfer,
7th Edition, New York: John Wiley & Sons.
Carslaw, H.S. and J.C. Jaeger. 1959. Conduction of Heat in Solids, Oxford: Oxford University Press.
Cengel, Y. A., R. H. Turner, J. M. Cimbala, 2016. Fundamentals of Thermal-Fluid Sciences, New York:
McGraw-Hill.
Chen, M.M. and K.R. Holmes, 1980. “Microvascular Contributions in Tissue Heat Transfer”, Annals of the
New York Academy of Sciences, 335: 137–150.
Dames, C., 2009. Microscale Conduction, Chapter 11, in Heat Conduction, 3rd Edition, Ed. L.M. Jiji, pp.
347–401, Springer - Verlag, Berlin.
Han, J.C.. 2011. Analytical Heat Transfer, Boca Raton: CRC Press/Taylor & Francis Group.
Heisler, M.P.. 1947. “Temperature Charts for Induction and Constant Temperature Heating”, Transactions of
ASME, 69: 227–236.
Henrici, P.. 1986. Applied and Computational Complex Analysis, New York: John Wiley & Sons.
Holman, J.P.. 2010. Heat Transfer, 10th Edition, New York: McGraw-Hill.
Jiji, L.M., 2009. Heat Conduction, 3rd Edition, Springer-Verlag, Berlin.
Jiji, L.M., S. Weinbaum, and D.E. Lemmons, 1984. “Theory and Experiment for the Effect of Vascular
Microstructure on Surface Tissue Heat Transfer - Part II: Model Formulation and Solution”, ASME
Journal of Biomechanical Engineering, 106: 331–341.
Kakac, S. and Y. Yener. 1988. Heat Conduction, Washington, DC: Hemisphere.
Kern, D.Q. and A.D. Kraus. 1972. Extended Surface Heat Transfer, New York: McGraw-Hill.
Mathews, J.H.. 1982. Basic Complex Variables, Boston: Allyn and Bacon.
Myers, G.E.. 1971. Analytical Methods in Conduction Heat Transfer, New York: McGraw-Hill.
Naterer, G.F.. 1996. “Conduction Shape Factors of Long Polygonal Fibres in a Matrix”, Numerical Heat
Transfer A, 30: 721–738.
Pennes, H.H., 1948. “Analysis of Tissue and Arterial Blood Temperatures in the Resting Forearm”, Journal of
Applied Physiology, 1: 93–122.
Razavi, M., 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.
Schneider, P.J.. 1957. Conduction Heat Transfer, Cambridge, MA: Addison-Wesley.
Teerstra, P.M., 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.
Yener, Y. and S. Kakac. 2008. Heat Conduction, 4th Edition, Boca Raton: CRC Press/Taylor & Francis Group.
Yovanovich, M.M.. 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.
PROBLEMS
1. A system of mass m is initially maintained at a temperature of T0. Then the system is sud-
denly 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 system. 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. 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
68 Advanced Heat Transfer
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.
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 main-
tained at T(0, t) = T0.
a. Select an appropriate control volume and derive the differential heat equation 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 posi-
tion? Will a steady-state temperature be reached? Explain your response.
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).
4. A rectangular system of mass m involving one-dimensional heat transfer with constant
properties and no internal heat generation is initially at a temperature of Ti. Suddenly 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 radia-
tion was absorbed by the surface at x = L and the remainder was reflected from the
surface?
5. An experimental apparatus for the measurement of thermal conductivity uses an electri-
cally 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 contact 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 approxi-
mately as k(T) = k0(1 + cT2), 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 identical sam-
ples surrounding the heater, rather than a single heater-sample combination.
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 boundaries are isothermal. Is
your analysis a one-dimensional solution? Explain your response.
7. A thin film-type resistance heater is embedded on the inside surface of a wind tunnel obser-
vation window for defogging during testing of aircraft de-icing systems. The outside and
inside air temperatures are 23 and −12°C, respectively, and the outside and inside convec-
tive heat transfer coefficients are 12 and 70 W/m2K, 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.
Heat Conduction 69
16. A cylindrical fuel rod within a nuclear reactor generates heat due to fission according to
q = qo − qo (r/R), where R and qo refer to the radius of the fuel rod and an empirical coef-
ficient, respectively. The boundary surface at r = R (2.2 cm) is maintained at a uniform
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 conductivity is
k = 12 W/mK.
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 temperatures 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.
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 1400°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.
19. A team member in a race car competition has proposed to air-cool the cylinder of a com-
bustion 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 cylinder (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/m2K. 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 specific fin
configuration.
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 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/m2K. 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.
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 convection 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-direction along the collector
plate. Outline the boundary conditions.
Heat Conduction 71
b. Solve the governing differential equation to find the plate temperature halfway between
the copper tubes. The convection coefficient, thermal conductivity, plate thickness,
T∞, T0, qrad, and tube-to-tube spacing are 20 W/m2K, 200 W/mK, 8 mm, 20°C, 60°C,
700 W/m2, and 0.1 m, respectively.
22. A cylindrical copper fin is heated by an airstream at 600°C with a convective heat transfer
coefficient of 400 W/m2K. 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.
23. Derive the three-dimensional heat conduction equation (including heat generation) in the
spherical coordinate system.
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 = 20x2 − 40y2 + 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.
24. 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 adjoining outer
half-cylinders (thickness of Δ2, conductivities of k2 and k3). The contact resistances at the
interfaces between the cylindrical layers may be neglected.
25. The top and side boundaries of a long metal bar are maintained at 0°C, and the bottom side
is held at 100°C. If the square bar thickness is 1 cm, find the spatial temperature distribu-
tion inside the plate.
26. 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 insulated. 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 tem-
perature distribution within the rod.
27. 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 convection 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.
28. 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 conduction 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.
29. Consider an irregular region formed by the gap between concentric polygons. Use the
method of conformal mapping to establish an expression for the temperature distribution in
this region when Dirichlet (fixed temperature) boundary conditions are applied along the
surfaces of both polygons.
30. 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. Determine a conformal
mapping that allows the known temperature distribution within a horizontal strip of unit
width to be mapped to the crescent spaced region.
31. Find a conformal mapping that 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.
72 Advanced Heat Transfer
32. Material property issues have been observed after the quenching of steel ingots under con-
ditions when the surface cooling is not sufficiently fast to avoid the formation of soft pearl-
ite 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/m2K. 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.
33. 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) direction. Each plate has an initial tem-
perature of 25°C with ρ = 2,700 kg/m3, k = 230 W/mK, and cp = 1,030 J/kg K. Under typi-
cal operating conditions, the damper is thermally charged by passing a hot gas through the
hot spaces with a convection coefficient of h = 100 W/m2K and T∞ = 600°C.
a. How long will it take to reach 90% of the maximum (steady state) energy storage in each
central damper plate?
b. What is the plate temperature at this time?
34. A spherical lead bullet with a diameter of D = 5 mm travels through the air at a supersonic
velocity. A shock waveforms and heats the air to T∞ = 680 K around the bullet. The convec-
tion coefficient between the air and the bullet is h = 600 W/m2K. 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.
35. A pulverized coal particle flows through a cylindrical tube prior to injection and combus-
tion 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?
36. 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 coefficient of h = 80
W/m2K. An internal combustion process exchanges heat by radiation 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 practice, how would convective cooling be applied
to prevent melting of the cylinder?
37. 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 temperature 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 variations with time and
position.
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 indi-
vidual 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 significantly affects the rate of heat trans-
fer, detailed knowledge of the fluid velocity and pressure distributions are essential in convection
problems. Therefore, both fluid flow and energy equations must be considered in the analysis. In
addition to heat transfer, convection is also a major mode of mass transfer in fluids. These governing
equations and associated 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 thermophysical properties.
This heat transfer coefficient depends on various problem parameters such as the fluid thermo-
physical 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 common forms of convective heat trans-
fer is: (i) free convection with air (5 < h < 25 W/m2K); (ii) forced convection with air (10 < h < 500
W/m2K); (iii) forced convection with water (100 < h < 15,000 W/m2K); and (iv) condensation of
water (5,000 < h < 100,000 W/m2K). Several excellent textbooks have been written with a broad and
deep coverage of convective heat transfer processes, e.g., Bergman et al. (2011), Kakac et al. (2013).
Convection can be induced by a number of external forces leading to natural, forced, gravita-
tional, 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 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 presence of a temperature gradient
leads to a nonuniform magnetic body force and fluid movement. 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 acceler-
ates in the vertical direction.
DOI: 10.1201/9781003206125-3 73
74 Advanced Heat Transfer
A wide range of physical processes may induce convection. The above examples included buoy-
ancy, 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 sponta-
neously 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 inhomo-
geneous composition of the substances or temperature dependence of the surface tension forces
(called thermocapillary 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 correla-
tions for a range of common geometrical configurations. This chapter will also provide a brief
introduction to turbulent convection and the second law of thermodynamics, as it relates to convec-
tive heat transfer.
3.2 GOVERNING EQUATIONS
The convection governing equations include the conservation of mass (also called the continuity
equation), momentum, and energy, as well as the second law of thermodynamics. The conservation
equation for a general scalar quantity, B, is given by
B cv B in – B out B g (3.1)
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.
m x udy (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),
m x
m x dx m x dx (3.3)
x
Convection 75
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),
u v
dxdy udy vdx udy dxdy vdx dyddx (3.4)
t x y
By simplifying and re-arranging this result, the following mass conservation equation is obtained
for two-dimensional flow:
u v
0 (3.5)
t x y
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.
M x ,cv M x Fx M
in
x Fx out
(3.7)
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
M x
M x , x uudy; M x , x dx M x dx (3.8)
x
M x
M x , y vudx; M x , y dy M x dy (3.9)
y
76 Advanced Heat Transfer
FIGURE 3.1 Mass flow, momentum flux, and force components of a control volume balance.
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), consider
the pressure, p, shear stress, τij, and normal stress, σij, acting on the surfaces of the control 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.
Consider the normal and shear stress components acting in the x-direction along the surfaces at
locations x and y, respectively:
xx
Fn, x xx dy; Fn, x dx xx dx dy (3.10)
x
Fs, y yx dx; Fs, y dy yx yx 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 dxdy, 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 canceling terms using a similar procedure as the previous
continuity equation, yields the following x-momentum equation:
u uu vu p xx yx
Fbx (3.12)
t x y x x y
Alternatively, using the product rule on the left side and the continuity equation,
u u u p xx yx
u v Fbx (3.13)
t x y x x y
Convection 77
v v v p yy
u v xy Fby (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 proportional 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. Examples 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 rela-
tions describe the normal and shear stresses in terms of the strain rate (deformation rate, or velocity
gradients):
u 2 u v
xx 2 (3.15)
x 3 x y
v 2 u v
yy 2 (3.16)
y 3 x y
u v
yx xy (3.17)
y x
u u 2 uk
ij i j ij (3.18)
x j xi 3 xk
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.
78 Advanced Heat Transfer
Substituting the constitutive relations into the previous momentum equations leads to the follow-
ing two-dimensional incompressible form of the Navier–Stokes equations:
u u u p 2u 2u
u v 2 2 Fbx (3.19)
t x y x x y
v v v p 2v 2v
u v 2 2 Fby (3.20)
t x y x x y
In tensor form, the general three-dimensional incompressible form of the Navier–Stokes equa-
tions is given by
ui u p ui
u j i Fb,i (3.21)
t x j xi x j x j
ui u p
u j i (3.22)
t x j 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,
u ; v (3.23)
x y
Substituting this result into the continuity equation yields Laplace’s equation, as presented ear-
lier in Chapter 2, but in terms of the potential function. The continuity equation for inviscid flow
becomes:
2 2
0 (3.24)
x 2 y 2
Alternatively, for inviscid flows, the continuity equation can be written in terms of a stream func-
tion, ψ(x, y), defined as follows:
u ; v (3.25)
y x
Convection 79
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.
E cv E adv E cond W E
in
adv E cond W out
E g (3.26)
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 previous 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,
FIGURE 3.3 Heat flux and work terms for a control volume.
80 Advanced Heat Transfer
q
qcond , x dx qx x dx dy (3.27)
x
1 1
qadv, x dx u e V 2 u e V 2 dx dy (3.28)
2 x 2
where V = (u2 +v2)1/2 is the total velocity magnitude. The flow stream in the advection term carries
both internal (e) and kinetic energy (V2/2).
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 pvdx; Wp, y dy pv dy dx (3.29)
y
yy v
Wn, y yy vdx; Wn, y dy yy v dy dx (3.30)
y
yxu
Ws, y yxudx; Ws, y dy yxu 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 individual 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 q
t
e uiui u j
2
x j
e uiui i
2 xi xi
pui
xi
iju j ui Fb,i q (3.32)
where q 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.
1 1 p
uiui u j uiui ui ui ij ui Fb,i (3.33)
t 2 x j 2 xi x j
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 pres-
sure 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.
Convection 81
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 u
ui
xi xi
pui p i (3.34)
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 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 therefore subtracted because the pressure
work in the latter term is directed toward fluid compression 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:
ij
ui
x j x j
ijui ij xui (3.35)
j
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 function, Φ, can be
expressed as
u 2 v 2 u v
2
2 (3.37)
x y y x
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 continu-
ously 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 canceled each other.
82 Advanced Heat Transfer
e e q u u
u j i p i ij i q (3.38)
t x j xi xi x j
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 incompress-
ible 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
u j k ij q (3.39)
t x j x j x j 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 temperature
or enthalpy equations. Although temperature is not conserved like total energy, it is often more use-
ful to use a temperature form of the thermal energy equation because temperature 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
cv k q (3.40)
Dt x j x j
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 repre-
sent conduction of heat, viscous dissipation and source terms of thermal 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
cp T k q (3.41)
Dt Dt x j x j
1
(3.42)
T
1 p 1
(3.43)
RT 2 T
Convection 83
In addition to the previous conservation equations of mass, momentum and energy, an additional
thermodynamic relationship of the form e = e(T, p) (e.g., through the definition of the specific heat),
and an equation of state (e.g., ideal gas law), are required to solve the full system of governing equa-
tions for gas flows.
In this section, the mass, momentum and thermal energy equations have been derived in Cartesian
coordinates. The three-dimensional forms of these equations in cylindrical and 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
p y po y gdy (3.45)
yo
FIGURE 3.4 Isothermal object in a flow stream with forced and free convection.
84 Advanced Heat Transfer
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.
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
Du p 2u 2u
k 2 2 (3.47)
Dt x x y
Dv p 2v 2v
k g 2 2 g (3.48)
Dt x x y
DT 2T 2T
cv k 2 2 (3.49)
Dt x y
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 generat-
ing a body force term in Equation (3.48). The buoyancy force can be expressed in terms of tempera-
ture using Equation (3.42) and the thermal expansion coefficient, β,
1
(3.50)
T T
t x y
t ; x ; y (3.51)
L /U L L
u v pk
u ; v ; pk (3.52)
U U U 2 / 2
T T
(3.53)
Tw T
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 time scale of t = L2/α from conduction heat transfer, another possible reference
velocity is Uref = k/(ρLcp). The reference values for the dimensionless variables should have approxi-
mately 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
Dv p 1 2 v 2 v Gr
k (3.56)
Dt
y Re x 2 y2 Re2
DT 1 2T 2T Ec
(3.57)
Dt
Re Pr x 2 y2 Re
where Re, Gr, Ec, and Pr refer to the Reynolds, Grashof, Eckert and Prandtl numbers, respectively.
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
TABLE 3.1
Selected Nondimensional Parameters of Heat and Mass Transfer
Group Definition Interpretation
Biot number (Bi); Jean–Baptiste Biot (1774–1862) hL/ks Ratio of conduction resistance of the solid to the
boundary layer resistance of the fluid
Coefficient of friction (cf) τw/(ρV2/2) Nondimensional wall shear stress
Eckert number (Ec); Ernst Eckert (1904–2004) V /(cpΔT)
2 Ratio of kinetic energy of the flow to the boundary
layer enthalpy difference
Fourier number (Fo); Jean–Baptiste Fourier αt/L2 Nondimensional time, or the ratio of the rate of heat
(1768–1830) conduction to thermal energy storage in the solid
Grashof number (Gr); Franz Grashof (1826–1893) gβ(Ts – T∞)L3/ν 2 Ratio of buoyancy to viscous forces
Colburn j number (jh); Allan Colburn (1904–1955) St∙Pr2/3 Nondimensional heat transfer coefficient
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 (1882–1975) α/D Ratio of thermal and mass diffusivities
Nusselt number (Nu); Wilhelm Nusselt hL/kf Nondimensional temperature gradient at the surface
(1882–1957)
Peclet number (Pe); Jean Péclet (1793–1857) VL/α Heat transfer parameter characterizing the ratio of
convection and diffusion effects
Prandtl number (Pr); Ludwig Prandtl (1875–1953) ν/α Ratio of momentum and thermal diffusivities
Rayleigh number; John Rayleigh (1842–1919) gβ(Ts – T∞)L /να Ratio of buoyancy to viscous forces, multiplied the
3
nondimensional parameters as well as others which appear frequently in heat and mass transfer
problems.
Based on the form of these nondimensional equations, it can be observed that the momentum
and energy equations are coupled and the temperature solution has the following functional form:
x ,y ,t ,Re,Gr ,Ec,Pr (3.58)
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 num-
ber, Nu, as follows:
hL q /T L
NuL (3.59)
k k
L T T
NuL (3.60)
w T n w n w
where n is the coordinate normal to the surface, and w refers to the wall. Therefore, the Nusselt
number represents the nondimensional temperature gradient at the wall.
The average heat transfer coefficient is determined by integrating the local Nusselt number over
the object’s surface, S,
Nu
Nu dS Nu Re, Pr, Gr, Ec (3.61)
s
dS
s
Spatial integration and steady-state conditions allow the spatial and temporal dependencies to be
removed. Depending on the type of flow problem, certain dimensionless parameters 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 con-
vection in high-speed flows so the Grashof number dependence can be dropped.
Table 3.1 presents a summary of nondimensional variables for a range of physical processes such
as phase change and mass transfer. For example, in solidification and melting problems, the Stefan
number characterizes the ratio of sensible to latent heat during the phase change process. When
surface tension effects are significant, the Weber number characterizes the ratio of inertia to surface
tension forces.
3.3.2 Buckingham–Pi Theorem
Another useful method to determine the relevant nondimensional groups in a problem is called
the Buckingham–Pi theorem. This theorem states that a dimensional equation with k variables
can be reduced to a relationship involving k – r dimensionless products, where r refers to the
minimum number of reference dimensions of the variables. The final result of the analysis pro-
vides a functional relationship among the dimensionless variables. The procedure follows these
seven steps.
Convection 87
To demonstrate this procedure, consider the same example that was previously considered – con-
vective 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 temperature (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 c d
kg b kg kg m kg
1 a Lb c k d h 3 m
m
0 0 0 0
3 3 kg m s K (3.62)
ms s K s K
Solving this algebraic system of four equations for four unknown exponents yields a = 0, b = 1,
c = 0 and d = −1. Thus the first pi variable is π1 = hL/k (Nusselt number). Then Step 6 repeats this
process for each remaining repeating variable. For the second pi term,
a c d
kg b kg kg m m
2 a Lb c k d U 3 m
m
0 0 0 0
3 kg m s K (3.63)
ms s K s
hL U L c p
1 f 2 , 3 ; f , (3.65)
k k
As expected, this relationship is the same result that was obtained earlier with the prior method
of using dimensionless variables in the governing equations. In both methods, it was found that the
Nusselt number is a function of the Reynolds and Prandtl numbers for forced convection problems.
In problems involving more complex physical processes, the detailed form of a governing equa-
tion may be unknown. The Buckingham–Pi theorem is particularly useful in finding the appropri-
ate dimensionless parameters from given variables when the form of the governing equation is
88 Advanced Heat Transfer
unknown. Further details and examples using the Buckingham–Pi theorem are available in under-
graduate textbooks, e.g., White (2015).
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 turbulence effects are dominant.
To formulate the governing equations, a number of simplifying boundary layer assumptions will
be used. Two-dimensional incompressible flows under steady-state conditions with constant ther-
mophysical properties will be assumed. It is further assumed that the velocity component, v, perpen-
dicular to the wall (y-direction) is much smaller than the velocity component, u, in the streamwise
(x) direction. Also, since the boundary layer is normally 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 distribution
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 pressure 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
U (3.66)
x x
where U(x) refers to the freestream velocity distribution (assumed to be known for a specified sur-
face 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, momentum and
thermal energy equations become:
u v
0 (3.67)
x y
u u U 2u
u v U 2 (3.68)
x y x y
2
T T 2T u
c pu c pv k 2 (3.69)
x y y y
In nondimensional form, the boundary layer equations can be written as
u v
0 (3.70)
x y
u u
p 1 2u
u v (3.71)
x y x Re y2
T T
1 2T
u v (3.72)
x y RePr y2
These governing equations are subject to appropriate boundary conditions, such as a zero veloc-
ity 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
90 Advanced Heat Transfer
(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 constant freestream velocity.
The slope of the velocity profile in the boundary layer at the wall is related to the shear stress act-
ing 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
w x (3.75)
y 0
The local skin friction coefficient, cf, is determined based on this wall shear stress,
w x
cf x (3.76)
U 2 / 2
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 functional 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 integrating the
local skin friction coefficient over the surface.
Similarly, within the thermal boundary layer, the local heat flux based on an energy balance at
the wall is given by
T
qw x k f h x Tw T (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 con-
ductivity, 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)
k f y 0
Convection 91
FIGURE 3.6 Experimental setup for convective heat transfer coefficient measurements.
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 functional 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.
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.
hAs Ts T I 2 R (3.80)
The average convective heat transfer coefficient, h , can be determined based on the applied elec-
trical current, I, electrical resistance, R, and temperature difference between the surface and the air-
stream. Several experiments are performed at various incoming velocities 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 = CReLm 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 transfer
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, characterized 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 numbers on the vertical axis, the mul-
tiple 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.
92 Advanced Heat Transfer
FIGURE 3.7 Nondimensional correlations for heat and mass transfer. (adapted from Bergman et al., 2011)
c
Re f Nu Sh (3.82)
2
Thus, for boundary layer flows with a zero pressure gradient (flat plate) and Pr = 1, any ana-
lytical, 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= St m (3.83)
2
Nu
St = (3.84)
RePr
Sh
St m = (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
St Pr 1 n St m Sc1 n (3.86)
2
Convection 93
which is applicable for Pr ≠ 1 and Sc ≠ 1. A value of n = 1/3 is commonly assumed. The middle
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 convection
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 determined, experi-
mental data can be gathered and grouped based on this functional form to reduce the number of
independent parameters under investigation.
3.5 EVAPORATIVE COOLING
An important application of the heat and mass transfer analogy involves evaporative cooling.
Evaporative cooling occurs as a result of mass transfer due to concentration gradients across a
phase interface, for example, through the diffusive transfer of water molecules from the surface of
a liquid to the gas phase. This occurs only for water molecules with sufficient energy to overcome
the cohesive forces holding the molecules in the liquid phase along the surface. Since higher energy
molecules leave the liquid surface, the resulting liquid temperature will become lower (thus called
evaporative cooling). In an enclosed space, vapor can be added continually by this process up to a
point until the vapor reaches the saturation pressure (Figure 3.8).
The heat flux of evaporative cooling can be written as
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 condi-
tions, the latent energy lost by the liquid due to evaporation balances the rate of heat transfer from
the surrounding gas to the liquid. For example, if heat is supplied by convection, the heat gain bal-
ances the evaporative heat loss,
qevap h fg hm A Ts A, h T Ts (3.89)
where hm, ρA(Ts), ρA∞, T∞ and Ts refer to the convective mass transfer coefficient, vapor density
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,
h fg pA,sat Ts pA,
T Ts (3.90)
c p Le2 / 3 RTs RT
where Le refers to the Lewis number and the thermophysical properties (ρ and cp) are evaluated at
the mean temperature of the thermal boundary layer.
The gas constant, R, is evaluated for the particular constituent, e.g., water vapor. Alternatively, the
gas constant can be written in terms of the universal gas constant (Ru = 8.315 kJ/kmolK) 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
M Ah fg
T Ts
c p RuTav Le2 / 3
pA,sat Ts pA, (3.91)
where MA is the molecular weight of constituent A in the mixture. Thus, the evaporative heat flux
can be written as
qevap evap pA,sat Ts pA, (3.92)
where
M Ah h fg
e ap (3.93)
c p RuTav Le2 / 3
A less accurate (but more convenient) approximation is obtained by assuming that the partial pres-
sure (pA∞) is negligible relative to the saturation pressure (pA,sat).
Similarly to evaporative cooling, sublimation is a phase change process where a material is trans-
formed 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, liquid, and gas phases
all coexist in equilibrium). Heat transfer by sublimation can be calculated 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 refriger-
ant for transporting perishables), antiseptics, fungicides, and light-sensitive materials (silver iodide)
in photography, as well as the production of dyes. Iodine forms black crystals 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.
3.6.1 Scaling Analysis
A useful approximate method for analyzing fluid flow and heat transfer problems is the method
of scaling analysis (or an order-of-magnitude analysis). This method is a powerful tool for the
simplification of governing equations with many terms. The magnitudes of individual terms in
the equations are determined and then negligible small terms may be ignored. The objective of a
scale analysis is to use the governing equations to estimate the order of magnitude of key variables
of interest. Usually a scale analysis can be reliable within a factor of one order of magnitude.
Sometimes this general level of accuracy is sufficient and can significantly reduce the compu-
tational costs relative to an exact or numerical solution. Scale analysis is also useful for quickly
obtaining key trends although a more detailed method of analysis is required to determine accurate
numerical results.
In the scaling method, reference or characteristic values of each of the problem variables are
substituted directly in their respective places in the governing equations to represent the respective
order of magnitude of individual terms in the equations. The following general steps and guidelines
can be used in a discrete scaling analysis.
For example, consider external flow over a flat plate subject to the boundary layer equations 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 approxi-
mate order of the cross-stream velocity component becomes vo ≈ δU/L. Then substituting the scaled
quantities into the x-momentum equation, Equation (3.68), leads to:
U U U
U U 2 (3.94)
L L
L
(3.95)
Re
where Re = ρUL/μ is the Reynolds number. The symbol “≈” means an equivalent order of magni-
tude 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 coefficients and precise
numerical values.
Using these scaled quantities, the order of magnitude of the skin friction coefficient becomes:
w U / 2
cf (3.96)
U / 2 U 2 / 2
2
Re L
96 Advanced Heat Transfer
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 scaling analysis agrees
fully with the functional form and order of magnitude of the skin friction coefficient, but the precise
values of numerical coefficients require a more detailed analysis.
3.6.2 Integral Solution
In the integral solution method, the governing equations are integrated over the range of a given
independent variable (e.g., coordinate direction), thereby reducing the number of problem variables.
The procedure removes the variation of the dependent variable(s) in the integrated coordinate direc-
tion (or time) since those variations are effectively averaged by the integration procedure. However,
as a result, additional information is required for the profile variations in the integrated coordinate
direction. Normally an approximation of the dependent variable is introduced in a certain direction
(or time), where knowledge of its variation exists. Then the spatial variations in the other directions
are obtained from the integral solution.
The following Leibnitz rule is frequently used in the integral solution method.
b x b x
f x,y db da
x f x,y dy
a x a x
x
dy f b,y f a,y (3.97)
dx dx
In convection problems, the variables a(x) and b(x) typically refer to a spatial range of coordi-
nates, 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 direction(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 distribution 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),
x
0
udy U
x
v vo 0 (3.98)
where vo = 0 (no-slip condition at the wall), U is the freestream velocity at y = δ, and vδ refers to the
vertical velocity at the edge of the boundary layer.
Similarly, integrating the momentum equation, Equation (3.68), across the boundary layer, and
using Leibnitz’s rule, Equation (3.97),
u u
uudy U Uv Uvo pdy p
(3.99)
2
x x x x y y 0
0 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:
Convection 97
d dU
dx
u U u dy
0
dx
U u dy w (3.100)
0
At this stage, an assumed profile for the velocity, u(y), must be provided across the boundary
layer in the y-direction so that the above integration can be completed (Figure 3.9). The resulting
equation will be an ordinary differential equation to be solved in the x-direction subject to appropri-
ate boundary conditions.
When the momentum equation is integrated in the y-direction, an approximate velocity profile
must be provided back for the spatial variation in this direction. For example, a cubic profile is
assumed to represent the variation of velocity across the boundary layer:
2 3
u y y y
ao a1 a2 a3 (3.101)
U
Four boundary conditions are required to determine the four unknown coefficients. Assume a
zero velocity and zero second derivative of velocity at the wall. Also, u(δ) = U and the first deriva-
tive of velocity is zero at the edge of the boundary layer, y = δ. These conditions yield ao = 0, a1 =
3/2, a2 = 0, and a3 = −1/2.
Substituting this velocity profile into the momentum equation, Equation (3.100), and performing
the integration, leads to the following ordinary differential equation:
d 140
(3.102)
dx 13 u
This equation can be solved to give the boundary layer thickness, which is then substituted into
Equation (3.101) to establish the cubic velocity profile, and then differentiated at the wall to deter-
mine the wall shear stress. Differentiating the velocity profile and evaluating at y = 0 yields:
w 0.323
(3.103)
U 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 solution.
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.
98 Advanced Heat Transfer
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 exam-
ined 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 vol-
ume 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.
3.6.3 Similarity Solution
A classical analysis of external flow over a flat plate was first presented in the form of a similarity
solution by Heinrich Blasius (1883–1970) in 1908. A similarity solution may be available whenever
one independent coordinate in the solution domain exists and physical influences are carried only
in that one direction. In Figure 3.10, two examples of self-similar flows are illustrated (channel and
jet flows), as well as a non-similar flow (river flow). In the first case (channel flow), self-similarity
can be observed by the developing nature of the flow in the channel. If the y coordinate is stretched
at each location x by a factor g(x), then all solution curves over the y direction would collapse onto
a single curve. In other words, the solution 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 trans-
formations, where the same functional form of equations is retained through the transformation.
Examples of invariant transformations include stretching, rotation 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 → a2x, 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 direction 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 eventually indicate the incom-
patibility. Recall the steady incompressible boundary layer equations over a flat plate, Equations
(3.67)–(3.69), are given by
u v
0 (3.104)
x y
u u p 2u
u v 2 (3.105)
x y x y
2
T T 2T u
c pu c pv k 2 (3.106)
x y y y
u
f (3.107)
U
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 incorrect, 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:
u ; v (3.109)
y x
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,
Ug2 x f C (3.110)
where
y g1 x (3.111)
f Ug2 x
(3.112)
f f vg1 x
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 Equation (3.107) was incor-
rect, 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).
Equating both sides of Equation (3.112) with a constant, and solving subject to the boundary
conditions, yields the following Blasius equation:
1
f f f 0 (3.113)
2
where
U
y (3.114)
vx
subject to
f 1 (3.115)
f 0 0; f 0 0 (3.116)
This system can be readily solved by numerical integration methods, such as the Runge–Kutta
method. Sample results are shown in Table 3.2.
TABLE 3.2
Similarity Functions for Flat Plate Boundary Layer Flow
η f(η) f′(η) f″(η)
0 0.000 0.000 0.332
0.6 0.060 0.199 0.330
1.2 0.238 0.394 0.317
1.8 0.530 0.575 0.283
2.4 0.922 0.729 0.228
3.0 1.397 0.846 0.161
3.6 1.930 0.923 0.098
4.2 2.498 0.967 0.051
4.8 3.085 0.988 0.021
5.4 3.681 0.996 0.008
6.0 4.280 0.999 0.002
6.6 4.879 1.000 0.001
7.2 5.479 1.000 0.000
Convection 101
u
f
0
exp
f d ) /2 d (3.117)
0
U
exp
0 0
f d ) /2 d
Based on this result, the boundary layer thickness can be found where the velocity reaches 99%
of the freestream value, as follows:
5x
x (3.118)
Rex
Also, the wall shear stress and skin friction coefficient can be determined,
u f 0.332 U 2
w x U (3.119)
y y 0 y y 0 Rex
w x 0.664
cf x (3.120)
U 2 / 2 Re x
Similarly, for the energy equation, Equation (3.106) is also rewritten in terms of the similarity
variables, (x, η). Define the nondimensional temperature as
T Tw
(3.121)
T Tw
Following a similar procedure as the momentum equation, the boundary layer energy equation
becomes:
Pr
f 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 de-coupled
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:
0
exp Pr
f d /2 d (3.123)
0
exp Pr
0
0
f d /2 d
where Pr is the Prandtl number. It can be observed that θ rises to θ∞ faster as Pr → ∞ (compared
with Pr ≪ 1) since the thermal boundary thickness is small relative to the velocity boundary layer
thickness at large Prandtl numbers.
102 Advanced Heat Transfer
Then the thermal boundary layer thickness, δt, wall heat flux and local Nusselt number can be
determined as follows:
t x
Pr 1/ 3 (3.124)
x
T u
qw x k k T 0 (3.125)
y 0
vx
hx qw /T x
Nux 0 Re1x/ 2 (3.126)
k k
From the temperature results in Equation (3.123), it can be shown that the slope of the dimension-
less temperature profile varies with Pr according to θ′(0) ∼ 0.332 Pr1/3. Therefore,
where x refers to the position along the plate. This similarity solution agrees closely with experimen-
tally 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 transfer 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 posi-
tive 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 tem-
perature 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 cor-
relations over the appropriate range,
xc L
1
hL
L
0
xc
hlam x dx hturb x dx (3.128)
TABLE 3.3
Correlations for Laminar and Turbulent Flow over a Flat Plate
Skin Friction Coefficient Nusselt Number Correlation
Boundary Layer Thickness w hx
Conditions δ = δ(x) cf Nux =
U 2 / 2 k
Laminar δ(x) = 5x Rex−1/5 cf (x) = 0.664 Rex−1/2 Nu(x) = 0.332 Rex1/2 Pr1/3
(Rex < 5 × 105) (Pr > 0.6)
Turbulent δ(x) = 0.37x Rex−1/5 cf (x) = 0.059 Rex−1/5 Nu(x) = 0.0296 Rex4/5 Pr1/3
(Rex ≥ 5 × (0.6 < Pr < 60)
105)
Convection 103
TABLE 3.4
Average Heat Transfer Correlations for a Flat Plate Boundary Layer
Conditions Flow Heat Transfer
xc/L > 0.95 Laminar NuL = 0.664 ReL Pr1/3; 0.6 < Pr < 50
1/2
xc/L < 0.1 Turbulent NuL = 0.037 ReL4/5 Pr1/3; 0.6 < Pr < 60
0.1 ≤ xc/L ≤ 0.95 Mixed NuL = (0.037 ReL4/5 – 871) Pr1/3; (0.6 < Pr < 60)
where xc refers to the critical point (transition to turbulence). The average Nusselt number correla-
tions 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 correlations 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 deter-
mine a suitable function(s). In this method, a general similarity variable is defined based on a prod-
uct of all independent problem variables, such as x and y, each raised to an exponent (an unknown
coefficient). The specific values of the unknown coefficients 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.
U x
y ; u U x f (3.129)
x
2m
U x u1 x m ; (3.130)
m 1
Using the similarity solution method with these variables, the reduced form of the momentum
equation and boundary condition becomes:
m 1
f
2
ff m 1 f 0 (3.131)
2
f 0 f 0 f 1 0 (3.132)
A value of m < 0 corresponds to an adverse pressure gradient and thus potentially leads to bound-
ary layer separation, while m < 0 yields a favorable pressure gradient. Also, m = 0 reduces to the
Blasius flat plate boundary layer. Analytical solutions exist only in the range of −0.091 ≤ m ≤ 2
(corresponding to −0.199 ≤ β ≤ 4/3).
In terms of heat transfer, a dimensionless temperature, θ, and wall temperature, Tw(x), are defined
as follows:
T T
; T Tw x T T1 x n (3.133)
Tw T
For a constant wall temperature, n = 0, whereas for a constant wall heat flux, n = (1 – m)/2. Then
the energy boundary layer equation and temperature boundary conditions become:
m 1 2mn
Pr f n Pr f Pr Ec x f 2 (3.134)
2
0 1; 0 (3.135)
where Pr and Ec = u2/(cpT∞) are the Prandtl and Eckert numbers, respectively.
For an isothermal boundary condition (n = 0), if dissipative heating on the right side of Equation
(3.134) is neglected, then the energy equation can be solved analytically to yield:
1
Nux m 1
Re x
0
F m, Pr exp Pr
2
0
f d d (3.136)
hx U x x
Nux 0 Re x 0 (3.137)
k
Sample results in Figure 3.12 (a) show that the function F increases with Pr. For a flat plate (m =
0), the previous results for a flat plate can be recovered. In the case of flow stagnation flow (m = 1)
or flow separation (m = −0.091), other solution methods are required.
The thermal boundary layer flow for a three-dimensional boundary layer in a rectangular cor-
ner also has a self-similar profile with an external velocity distribution in the form of U(x) = Cxm.
Sample temperature results of this flow configuration as shown in Figure 3.12 (b) were presented by
Schlichting (1979). Lines of constant temperature are shown for Pr = 0.7 and Ec = 2.4.
Convection 105
FIGURE 3.12 (a) Nusselt number for wedge flow with a constant wall temperature and (b) temperature
distribution in a right-angled corner. (adapted from Schlichting, 1979)
FIGURE 3.13 Effects of separation and turbulence on external flow over a cylinder.
106 Advanced Heat Transfer
FIGURE 3.14 Drag coefficient for a smooth circular cylinder and sphere in cross-flow. (adapted from
Schlichting, 1979)
to flow separation. A narrower downstream wake and resulting change in pressure due to a smaller
recirculating low-pressure wake cause an abrupt decrease of the drag coefficient (cd ≈ 0.3 and
0.1 for the cylinder and sphere, respectively) following the transition to turbulence. Beyond this
point, the drag coefficient rises with ReD due to the increasing role of turbulent mixing on the flow
structure.
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 separa-
tion point, NuD increases with θ since flow separation causes local flow reversal, stronger mixing,
and thus enhanced heat transfer. In Figure 3.15, 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 tempera-
ture 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.
Based on experimental data for flow past a cylinder over a wide range of Reynolds numbers, the
following Zhukauskas correlation can be used (Zhukauskas, 1972; see Table 3.5).
FIGURE 3.15 Heat transfer from a cylinder and sphere in cross-flow. (adapted from Giedt, 1949; Hewitt
et al., 1997)
Convection 107
TABLE 3.5
Zhukauskas Correlation Coefficients for
External Flow Past a Circular Cylinder
ReD C m n (Pr ≤ 10) n (Pr > 10)
1–40 0.75 0.4 0.37 0.36
40–103 0.51 0.5 0.37 0.36
103–2 × 105 0.26 0.6 0.37 0.36
2 × 10 –10
5 6 0.076 0.7 0.37 0.36
1/ 4
Pr
Nu D CReDm Pr n ; 0.7 Pr 500 (3.138)
Prs
In Equation (3.138) and Table 3.5, all thermophysical properties are evaluated at the freestream
temperature, except the Prandtl number, Prs, which is evaluated at the surface temperature, Ts. Due
to the inherent instability and unsteadiness of vortex shedding from the cylinder, measurement
uncertainties and other factors, Equation (3.138) 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, e.g., Kakac et al. (2013),
Bergman et al. (2011).
3.9.1 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
Nu D 2 0.4 Re 1/ 2
D 0.06 Re 2 /3
D Pr 0.4
(3.139)
w
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 temperature 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 3.15 (Hewitt et al., 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 (decreas-
ing 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.
108 Advanced Heat Transfer
At sufficiently high velocities, the increasing pressure along the back side of the cylinder causes the
boundary 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 coefficient. 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
Re
Nu 2 D 3 10 4 Re1D.6 (3.140)
4
for 100 < ReD < 2 × 105. This correlation was determined experimentally for convective heat transfer
from isothermal sphere to air.
3.9.2 Tube Bundles
Another geometry of practical importance is a cross flow past a number of regularly spaced paral-
lel cylinders (see Figure 3.16). This geometry is commonly encountered in industrial systems such
as tube bundles in heat exchangers, tubes inside condensers, and air conditioner coils. In these
examples and others, an external flow past the tube bundles transfers heat to/from a fluid moving
inside the tubes.
The Zhukauskas correlation (1972) for this configuration is given by
1/ 4
Pr
Nu D CReDm,max Pr 0.36 (3.141)
Prw
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 20 or more rows
of tubes (NL ≥ 20), 0.7 < Pr < 500, and 1, 000 < ReD,max < 2 × 106 (Zhukauskas, 1972). The thermo-
physical 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 (mea-
sured 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 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 (1980) and Bergman et al. (2011).
3.10 INTERNAL FLOW
Internal flows with heat transfer occur in many industrial applications, such as fluid flows in pipes,
air flows in ventilating ducts, and shell-and-tube and concentric tube heat exchangers (to be ana-
lyzed in Chapter 9). In this section, the thermal behavior and analysis of internal flows will be exam-
ined, including the role of fluid mechanics and turbulence.
distance, called the entry length, xc. Here the flow is considered to be still 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 coordinates.
The general form of the three-dimensional Navier–Stokes and thermal energy equations in cylindri-
cal coordinates is shown in Appendix C. The boundary layer equations are equivalent to Equations
(3.67)–(3.69) except in cylindrical rather than Cartesian coordinates. 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 (3.142)
x r r
u u p u
u v r (3.143)
x r x r r
2
T T 2T k T u
c pu c pv k 2 r r (3.144)
x r x 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 coordinate in the flow direction).
Also, the viscous dissipation term in the energy equation may be neglected except for highly
viscous liquids.
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 u
0 r (3.145)
x r r r
T T
u r (3.146)
x r r r
where α is the thermal diffusivity (k/ρcp). On the right side of the x-momentum equation, the pres-
sure 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,
u r r 2
2 1
u ro
(3.147)
Convection 111
ro2 dp
u (3.148)
8 dx
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
1/ n
r
2 1 (3.149)
u ro
D dp / dx
f (3.150)
u 2 / 2
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 Figure 3.18), 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 condi-
tions. The friction factor abruptly increases for turbulent flow and then eventually reaches wholly
turbulent flow where a laminar sublayer within the boundary layer is so thin that the surface rough-
ness 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.
Using the previous laminar flow solution and Equation (3.147), the Moody friction factor can be
expressed as
64
f = (3.151)
ReD
FIGURE 3.18 Moody chart for the friction factor in cylindrical pipes. (adapted from Moody, 1944)
112 Advanced Heat Transfer
For turbulent flow in a smooth tube, the following friction factor correlations may be used:
f 0.316 ReD1/ 4 Re D
2 10 4 (3.152)
f 0.184 ReD1/ 5 Re
D 2 10 (3.153)
4
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
1 2 (3.154)
T x Ts T x Ts
1 2
where the subscripts 1, 2, and s refer to an upstream location, downstream location, and surface,
respectively. The mean temperature of the fluid is given by
1
T x
uAc
u r,x T r ,x dAc (3.155)
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 condition of
the velocity field, where the mean velocity remains constant, the mean temperature 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 (3.146) to determine the temperature
profile. Two different boundary conditions will be considered: (i) constant heat flux and (ii) constant
surface temperature. These two cases are illustrated in Figure 3.19. The case of a constant wall heat
flux leads to a constant difference between the mean fluid and wall temperatures in the fully devel-
oped 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 (3.146), can be solved to obtain:
FIGURE 3.19 Temperature profiles for constant heat flux and isothermal boundary conditions.
Convection 113
3
4 2
2uro2 dT 1 r 1 r
T r Ts (3.156)
dx 16 16 ro 4 ro
Substituting the temperature and velocity profiles into Equation (3.155) and integrating,
11 uro2 dT
T x Ts (3.157)
48 dx
Consider an energy balance over a control volume in the tube consisting of a disk of thickness 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
p
mc qs P (3.158)
dx
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,
qP
T x Tin s x (3.159)
p
mc
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 con-
stant, and so also q′′s /h is constant, leading to:
48 k
h (3.160)
11 D
Alternatively,
hD
=
Nu D = 4.36 (3.161)
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
p
mc h Ts T P (3.162)
dx
d T 1
Ph x dx (3.163)
T p
mc
where T Ts T x . Integrating this energy balance and exponentiation of both sides of the result-
ing expression involving ΔT leads to:
T Ts T x Pxh
exp (3.164)
Tin Ts Tin p
mc
114 Advanced Heat Transfer
where the average convection coefficient along the length of pipe is given by
x
1
h
xh x dx (3.165)
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,
p Tout Tin mc
qnet mc p Ts Tin Ts Tout (3.166)
Substituting ṁ from Equation (3.164),
r 2 T T dT d dT
2u 1 s
r (3.168)
r
o T Ts dx r dr dr
Solving this equation requires an iterative procedure with successive approximations of the tem-
perature profile. Once the temperature distribution is found, it can be differentiated with respect to
radial position, r, to give the radial heat flux, which is then equated with the convective transfer rate
from Newton’s law of cooling, to obtain the convection coefficient. It can be shown that the resulting
Nusselt number is given by
The previous analysis can be extended to other, more complicated flows, such as a combined
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
(3.170)
U hi ho
This result corresponds to convection resistances inside (subscript i) and outside (subscript o) of
the pipe in series.
These heat transfer results and correlations are summarized in Table 3.6. The names attributed to
the turbulent flow correlations are shown in parentheses. The subscript s implies evaluation of the
viscosity at the surface temperature. In these correlations, the diameter and length of the tube are D
and L, respectively.
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.
Convection 115
TABLE 3.6
Heat Transfer Correlations for Internal Flow in a Tube
Flow Conditions Heat Transfer (NuD = hD/k)
Laminar Constant wall heat flux; ReD < 104 NuD = 4.36
Laminar Constant wall temperature; ReD < 104 NuD = 3.66
Turbulent Constant wall heat flux or temperature; NuD = 0.027Re4/5Pr1/3(μ/μs)0.14
ReD ≥ 104; L/D > 10; 0.7 ≤ Pr ≤ 16 (Seider-Tate correlation)
Turbulent Constant wall heat flux or temperature; NuD = 0.023 Re4/5 Prn
ReD ≥ 104; 0.7 ≤ Pr ≤ 160; (Dittus-Boelter correlation)
n = 0.4 (heating); n = 0.3 (cooling)
VD 4m 4 1
ReD 54, 412 (3.171)
D 0.6 390 10 7
which represents turbulent flow. Using Table 3.6, the Seider-Tate correlation is selected, and the
convection coefficient becomes:
0.14
k
hi 0.027ReD4 / 5Pr1/ 3 15 W/m2K (3.172)
D s
For external cross-flow past the outside surface of the smokestack, ReD = 4.3 × 105 (turbulent
flow). The convection coefficient can be determined based on the Zhukauskas correlation,
1/ 4
k Pr
ho 0.26ReD0.6Pr 0.37 22.7 W/m2K (3.173)
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 coef-
ficient, U,
T Tout PLU
exp (3.174)
T Tin p
mc
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
(3.175)
U hi ho
Here the subscripts i and o refer to inner and outer, respectively. Substituting numerical
values,
0 .6 8 1
Tin 5 5 600 exp 1/15 1/ 22.7 (3.176)
1 1104
,
which yields a base mean temperature of 678.1°C. Then, constructing a thermal circuit with inter-
nal and external convection resistances in series,
Ti To Ti To
qs 5.37 kW/m2 (3.177)
Rtot 1/hi 1/ho
In order to refine this estimate, iterations are performed for the re-calculation of thermophysical
properties, based on the result obtained for the inlet temperature. From a practical 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 set-
tling of harmful flue gases in nearby farmland or residential areas.
3.10.2 Non-Circular Ducts
For non-circular ducts, the same correlations as the previous section for circular tubes may be used,
however, with the diameter in the friction factor, Reynolds number and Nusselt number replaced by
the equivalent hydraulic diameter, Dh = 4Ac/P, where Ac and P refer to the cross-sectional area and
perimeter of the non-circular duct, respectively.
Alternatively, Muzychka and Yovanovich (2016) have shown that a more effective length scale
than the hydraulic diameter for non-circular ducts is the square root of the flow area. Using this alter-
native approach, Figure 3.21 illustrates the resulting friction factor over a range of non-circular duct
geometries using a characteristic length of the square root of area. It can be observed that numerous
geometries, such as rectangular, elliptical, triangular and trapezoidal cross-sections, collapse onto
a single curve when the square root of area is used. Using a conventional hydraulic diameter, the
results are widely scattered and cannot be collapsed onto a single curve in the same manner. This
approach of re-formulating correlations in terms of the square root of area is a powerful method that
can be extended to other heat and fluid flow correlations.
Convection 117
FIGURE 3.21 Friction factor product, f ∙ Re, for: (a) rectangular, elliptic and (b) other duct shapes. (adapted
from Muzychka and Yovanovich, 2016)
For a rectangle of aspect ratio β = b/a, the various curves of friction factors at varying aspect
ratios can be normalized onto a single curve given by
12
f Re (3.178)
A
1 1 192 tanh / 2 / 5
Using the results of the previous circular tube analysis, the Nusselt numbers can be re-calculated
for noncircular cross-sections such as triangles, squares, and so on, based on the hydraulic diameter
and square root of the area. Table 3.7 summarizes the Nusselt numbers for fully developed Poiseuille
TABLE 3.7
Nusselt Numbers for Fully Developed Flow in Non-Circular Ducts
Isothermal Isoflux
flow using both the hydraulic diameter and the square root of the cross-sectional area as the charac-
teristic length scale.
For the rectangle cases, a range of different aspect ratios is presented. Two cases of isothermal
and isoflux boundary conditions are shown. It can be observed that the Nusselt numbers for the dif-
ferent 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 isother-
mal (superscript T) and isoflux boundary conditions (superscript H) can be correlated with respect
to the Reynolds number as follows. For the constant wall temperature boundary condition,
f Re A
NuT A 3.01
(3.179)
8
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,
f Re A
Nu HA 3.66 (3.180)
8
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. These
models, when based on a characteristic length of the cross-sectional area, agree well with experi-
mental 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.
3.11 FREE CONVECTION
Free convection (or natural convection) arises from buoyancy forces that occur from density differ-
ences due to temperature variations in the fluid. For example, warm air ascends above a hot surface
placed in a cool room due to buoyancy forces arising from lighter, warmer air near the hot surface.
The upward buoyant flow entrains ambient air into the thermal plume rising above the hot surface.
Mixed convection refers to problems involving both forced and free convection. In this section, the
governing equations, physical processes, and advanced solution methods for free convection prob-
lems will be presented.
layer – an inner region where diffusion and buoyancy are dominant; and an outer region where iner-
tial effects and buoyancy 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 convection were
presented earlier in Equations (3.46) through (3.49). Note that the y-direction is aligned with the grav-
ity 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 configuration in Figure 3.22. 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 1
(3.181)
T p T T
Therefore, using Equations (3.46)–(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 (3.182)
x y
u u 2u
u v 2 g T T (3.183)
x y y
T T 2T
c pu c pv k 2 (3.184)
x y y
The body force term of buoyancy in Equation (3.183) has been expressed in terms of temperature
differences through the Boussinesq approximation. This approximation assumes that density differ-
ences are negligible in all terms, except in buoyancy forces, where the density difference drives the
fluid motion. For free convection problems, it can be assumed that viscous dissipation in the thermal
energy equation can be neglected.
In nondimensional form, the free convection boundary layer equations become:
u v
0 (3.185)
x y
120 Advanced Heat Transfer
1 2
u
v 2 (3.187)
x y ReL Pr y
where the reference velocity in the dimensionless velocity is ν/L and the nondimensional tempera-
ture 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 nondimensionalization are differ-
ent. 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 simi-
larity variable, η, and stream function, ψ, as
1/ 4
y Grx
(3.188)
x 4
Gr 1/ 4
x,y f 4 x (3.189)
4
2 1/ 2
u Grx f (3.190)
x
The boundary layer equations are transformed to the following coupled ordinary differential
equations for momentum and thermal energy, respectively,
f 3 ff 2 f 0 (3.191)
2
3Pr f 0 (3.192)
0; f f 0; 1 (3.193)
; f 0; 0 (3.194)
Using a similar solution procedure as the Blasius solution, the ordinary differential equations can
be solved with a numerical integration method, such as the Runge–Kutta method. Sample results of
the velocity and temperature profiles at varying Prandtl numbers are shown in Figure 3.23. It can
be observed that the maximum velocity occurs inside the boundary layer as previously illustrated in
Figure 3.22. 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 temperature cases,
1/ 4
Gr
Nux x g Pr (3.195)
4
Convection 121
FIGURE 3.23 Dimensionless velocity and temperature profiles in a free convection flat plate boundary layer.
(adapted from Ostrach, 1953)
1/ 4
4 GrL
Nu L g Pr (3.196)
3 4
where
0.75Pr1/ 2
g Pr (3.197)
0.609 1.221Pr
1/ 4
1/ 2
1.238Pr
The following Churchill and Chu correlation (1975), based on experimental data, extends over
both laminar and turbulent regions of a free convection flat plate boundary layer.
2
8 / 27
Nu L 0.825 0.387 Ra1L/ 6 1 0.492 /Pr
9 /16
(3.198)
Boundary layer transition to turbulence occurs at approximately Rax ≈ 109 for flow along a verti-
cal flat plate.
From the similarity solution results for free convection along a vertical plate in Figure 3.23, the
edge of the boundary layer is located at:
1/ 5
Gr
H 5 (3.199)
H 5
1/ 5
3
0.015 9.8 1.320 45H
5 (3.200)
H 5 18 106
2
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 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 con-
straints 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
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 Hopt L Ts T s (3.202)
t2
q Hopt 3 4
1/ 4
0.609 1.221Pr1/ 2 1.238Pr
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 further 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.
Convection 123
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 previ-
ous flat plate correlations can be applied to large vertical cylinders of length L when surface curva-
ture effects are negligible and the boundary layer thickness, δ, is much less than the diameter; δ ≪
D and D > 35L /GrL1/ 4 .
2u
g x T T 0 (3.203)
y 2
2T
k 0 (3.204)
y 2
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 (3.204), subject to T(ym) = Tm, and substituting the result into the solution of
Equation (3.203),
g x y2 y2 y y
u y T Tw yym m (3.205)
2 2 6
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 follow-
ing mass flow rate, Γi, per unit width of surface,
ym
2 gx y3 5M
i udy
0
T Tw m 1
3 8
(3.206)
124 Advanced Heat Transfer
FIGURE 3.25 Energy balance and temperature profile in a free convection boundary layer.
where ym and M = ym/Δ are unknowns to be determined by matching conditions at the interface
between the inner and outer regions.
Consider a mass and energy balance for a differential control volume in the inner region (see
Figure 3.25), as follows:
d
mi i i i dx (3.207)
dx
h
qo mi hm i h
x
i
x dx
q (3.208)
where the subscripts i, o, m and x refer to inner region, outer region, location y = ym, and position
x. Also, mi′′, hm and h refer to the mass flow rate per unit width across the inner 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 3.25.
Combining the above mass and energy equations with the solution of Equation (3.204) and rear-
ranging, the wall heat flux becomes:
T T d i
q k m w
hm hi qo (3.209)
ym dx
Similarly, in the outer region, the mass and energy balances can be written as
do
mo mi (3.210)
dx
mo h ho qo mi hm ho (3.211)
where the subscript ∞ refers to ambient conditions. Combining these results and rearranging,
d i k
ym (3.212)
dx c p B x
where
T Ti d o /dx T To
B x (3.213)
Tm Tw d i /dx Tm Tw
Convection 125
Assume that B = B(M) due to self-similarity of velocity and temperature profiles in the boundary
layer. Using Equation (3.206) to eliminate ym and the boundary condition of Гi(0) = 0, the mass flow
rate in the inner region becomes:
3/ 4
1 5M /8
1/ 4
4k
3/ 4
2 g T Tw
1/ 4
x g 1/ 3
i
BM
3/ 4
3c p
3
x dx (3.214)
g
0
hx q x Mx
Nux (3.215)
k T Tw k ym
where
1/ 4
5 M B
f M M 1 (3.217)
8 4
Here the Rayleigh number, Ra = Pr ∙ Gr. Also, M can be correlated as a function of Prandtl num-
ber 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 (3.218)
0.861 Pr
Integrating the local Nusselt number over the surface, S, yields the following average Nusselt
number:
3/ 4
4 1 S g 1/ 3
Nu s f M Ra1s / 4
dx (3.219)
x
3 S g
0
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 (3.219) is called the body gravity
function. Lee et al. (1971) presented a generalized form of the body gravity function,
3/ 4
P sin
1/ 3
1
GL
A
A
A /L
dA
(3.220)
126 Advanced Heat Transfer
where P(θ) and L represent the local perimeter associated and characteristic length scale, respec-
tively. Also sin θ 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,
3/ 4
P sin
1/ 3
1
G A
A
˜
A
dA
(3.221)
A
Muzychka and Yovanovich (2016) presented a modified form of Equation (3.219) using this body
gravity function,
0.670
F Pr 4/9
(3.223)
1 0.5/ Pr 9 /16
This generalized approach with the body gravity function provides a useful surface geometry
factor to determine the Nusselt number over a wide range of geometrical configurations. The body
gravity function depends on the body shape, aspect ratio, orientation with respect to the gravity vec-
tor, and the characteristic length scale.
Consider a surface consisting of several segments with fluid flow sequentially over multiple
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 /A
4 /3 7/6
G A A i i (3.224)
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 differ-
ent surfaces with N different streams that are independent, then the body gravity function can be
obtained from the following parallel flow expression,
N
G A /A 7 /8
G A A i i (3.225)
i 1
An example of a parallel flow arrangement is a horizontal circular cylinder with cooling that
occurs simultaneously over the side and end surfaces. In this case, the body gravity functions for the
vertical end surfaces are different than the horizontal cylindrical surface but all three functions can
be combined through the above parallel flow expression.
Table 3.8 provides a summary of gravity body functions for several surface configurations. Using
these functions, the average Nusselt number can then be determined for each configuration by
Equation (3.222). Varying surface inclinations within enclosures have applications such as solar col-
lectors, channels within electronic enclosures and heat transfer through window panes in buildings.
Hollands et al. (1976) presented correlations for free convection in tilted rectangular enclosures.
Free convection correlations for other geometrical configurations can be assembled by this
method such as spheres, concentric spheres and enclosures. The reader is also referred to other
Convection 127
TABLE 3.8
Body Gravity Functions for Various Surface Configurations
Surface Configuration Body Gravity Function Comments
(a) Vertical surface with a 1/ 4 1/8 Constant perimeter, height H, area of A = PH,
P P
constant perimeter G A sin θ = 1
A H
(b) Vertical circular cylinder 1/8 Cylinder of height H, diameter D, area of A = PH
D
G A 1/8
H
(c) Vertical elliptical 1/8 Semi-major and semi-minor axes of a and b, and E
4a 2
cylinder G A E 1 b / a is the complete elliptic integral of the second kind
H
(d) Vertical rectangular 1/8 Plate of width W, both sides cooled, and P = 2W
W
plate G A 21/8
H
(e) Inclined rectangular 1/8 Plate inclination angle of θ, and bottom side
1/ 4 W
plate G A sin cooled
H
sources and books for additional geometries, for example, Kays and Crawford (1980), Kreith,
Maglik, and Bohn (2010).
3.11.3 Spherical Geometries
For free convection heat transfer from the external surface of an isothermal sphere of diameter D,
0.589 Ra1D/ 4
Nu D 1 4/9
(3.226)
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 rep-
resents an external flow since there are no obstructions by other surfaces in the flow field. The fol-
lowing configurations of concentric spheres and rectangular enclosures are internal flows that are
constrained by the physical presence of solid boundaries on all sides of the flow field.
Free convection heat transfer in the region between concentric spheres involves a combination
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
Pr
1/ 4
0.74 RaS 0.861 Pr (3.227)
k
where
L RaL
RaS (3.228)
5
D D D17 / 5 D27 / 5
4
1
4
2
128 Advanced Heat Transfer
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 RaS 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:
D1D2
q keff T1 T2 (3.229)
L
The effective thermal conductivity represents a thermal conductivity of a stationary fluid that
produces the same amount of heat transfer as the actual moving fluid due to free convection between
the concentric spheres.
3.12 INTRODUCTION TO TURBULENCE
When a flow becomes turbulent, fluid elements exhibit additional transverse motion which enhances
the rate of energy and momentum exchange in the flow stream, thereby also usually increasing
the heat transfer and friction coefficients. Turbulence is composed of fine-scale random velocity
fluctuations superimposed on a mean velocity. These random fluctuations significantly increase the
complexity of turbulent flow analysis. Also, turbulent flows are inherently three-dimensional. In this
section, the general nature of turbulence, definitions, spectrum, and modeling of turbulence will be
briefly introduced. Further detailed treatment of turbulent flows and heat transfer is provided in clas-
sic textbooks on turbulence by Hinze (1975) and Tennekes and Lumley (1972).
3.12.1 Turbulence Spectrum
Turbulence generally arises from instabilities as the fluid velocity increases in a laminar flow.
Consider the transition from laminar to turbulent flow in a boundary layer across a flat plate. Flow
near the leading edge exhibits a stable, laminar character. Further downstream, when the flow reaches
a transition Reynolds number (105 for a flat plate), a fluid instability occurs in the upper portion of
the boundary layer. Inertial transverse motion of fluid eddies overcomes viscous dampening near the
wall. The smooth parallel motion in the boundary layer breaks down into transverse vortices. These
individual vortices merge to create a “turbulent spot” (region of concentrated mixing) which further
leads to downstream wakes. As more vortices and spots grow and merge downstream, eventually
the flow becomes fully turbulent. Several stages of development occur during this transition from
laminar to turbulent flow.
Turbulence is a multi-scaled problem that often requires a statistical analysis. In most cases, a
single velocity or length scale cannot adequately describe the features and characteristics of the
turbulent flow. Also, turbulence is not a self-sustaining process. Energy must be continually added
to the flow in order to maintain a given level of turbulence. For example, when a projectile passes
through a fluid, a turbulent wake is formed behind the projectile, but the fluid returns to laminar flow
after the projectile passes away. The turbulence is not self-induced or self-sustained as it needs to be
supported by kinetic energy extracted from the mean flow by the moving projectile.
Turbulent flow properties consist of a mean flow component and fluctuations about this mean
value. A general scalar quantity, B, in turbulent flow can be expressed as
B B B (3.230)
where
t T
1
B
T Bdt (3.231)
t
Convection 129
Here T represents a time scale that spans a characteristic range of time over which the turbulence
occurs. An example of mean and fluctuating components of a general scalar quantity is illustrated
in Figure 3.26. 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.
For example, the velocity components can be expressed as a sum of mean and fluctuating
velocities:
Turbulence intensity is a scale that characterizes the magnitude of turbulence in terms of a per-
centage. For example, smooth air flow with no fluctuations in air speed or direction would have a
turbulence intensity of 0%. The components of turbulence intensity are defined as
u 2 v 2 w2
Ix ; Iy ; Iz (3.233)
u v w
Stationary turbulence occurs when the turbulent flow fluctuations become independent of time.
Also, homogeneous turbulence occurs when these fluctuations become independent of position.
Isotropic turbulence is a special case of homogeneous turbulence that is restricted to small-scale
motions. It occurs when the intensity of turbulence in each coordinate direction is identical. The
statistical features of the turbulence have no directional preference. Isotropic turbulence by its defi-
nition is always homogeneous.
A turbulent flow has a spectrum of different eddy sizes and motions. The eddy sizes are char-
acterized by the velocity, time and length scales. Large eddies are more unstable and eventually
break up into smaller eddies. The kinetic energy of the initial large eddies is subdivided into smaller
eddies which then undergo the same process, leading to even smaller eddies, and so on. The energy
is passed down from the large scales of motion to smaller and smaller scales until they eventually
reach such a small length scale that viscous effects dissipate the kinetic energy into internal energy.
This process of energy exchange among the various scales of eddy motion is represented by the
turbulence spectrum. This spectrum is also called the Kolmogorov spectrum, named after Andrey
Kolmogorov (1903–1987). General features of this spectrum are illustrated in Figure 3.27. The tur-
bulence spectrum depicts the relationship between the turbulent kinetic energy and the eddy size or
frequency, where the turbulent kinetic energy, k, is defined by
k
2
1 2 2 2
u v w (3.234)
In Figure 3.27, 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
130 Advanced Heat Transfer
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 spec-
trum, 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 turbu-
lence for length (Lko), velocity (Vko), and time (Tko):
1/ 4
v3
Lko (3.235)
Vko v (3.236)
1/ 4
1/ 2
v
Tko (3.237)
When the Reynolds number, based on these Kolmogorov scales, has an order of magnitude of
approximately 1, the Kolmogorov scales are applicable to the small-scale eddy motions in the vis-
cous dissipation region of the spectrum.
ui
0 (3.238)
xi
ui u p ui u j
u j i uiu j
(3.239)
t x j xi x j x j xi
Convection 131
T T T
cp c pu j k c p ujT (3.240)
t x j x j x j
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 (3.239) and (3.240). Recall that the first terms in brackets in Equations (3.239)
and (3.240) represent molecular diffusion of momentum and heat, respectively. Similarly, the sec-
ond terms in brackets can be interpreted as turbulent, rather than laminar (molecular), diffusion of
momentum and heat. Note that k in Equation (3.240), heat flux terms, and Fourier’s law, will refer to
the thermal conductivity and should not be confused with the turbulent kinetic energy.
The laminar and turbulent diffusion terms can be expressed as
u u
ijl i j ; ijt uiuj (3.241)
x j xi
T
q lj k ; qijt ujT (3.242)
x j
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 turbulent 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 Equations
(3.239) and (3.240), the concepts of turbulent viscosity, μt, and turbulent conductivity, kt, are intro-
duced as follows:
u u
ijt uiuj t i j (3.243)
x j xi
T
q tj ujT kt (3.244)
x j
In laminar flows, μ is a property of the fluid, whereas in turbulent flows, μt represents a property
of the flow. In turbulent flow modeling, often the turbulent conductivity is approximated by
t c p
kt (3.245)
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 turbu-
lent 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
132 Advanced Heat Transfer
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 infor-
mation. 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.
3.12.3 Eddy Viscosity
In the eddy viscosity method, the enhanced diffusion due to turbulence is modelled in terms of an
eddy viscosity, μt, and eddy conductivity, kt. Both laminar and turbulent components of diffusion in
the previous section are combined with a single total diffusion coefficient. The total stress and heat
flux terms become:
u u
ijtot ijl ijt t i j (3.246)
x j xi
T
j q j q j k kt
q tot l t
(3.247)
x j
where the superscript 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 tur-
bulent viscosity. For example,
t Vref L (3.248)
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 (3.249)
3.12.4 Mixing Length
The mixing length model was initially proposed by Ludwig Prandtl in 1925. The turbulent viscosity
is written in terms of a mixing length. Consider the turbulent fluctuations of a fluid element about an
imaginary plane (y = 0 plane). The fluid element arrives at the y-plane from a distance L away with
Convection 133
a velocity that is approximately L∙dū/dy different. The velocity fluctuation, u′, is proportional to this
difference. However, from the continuity equation, v′ is proportional to u′, and therefore:
du du
yxt uv L2m (3.250)
dy dy
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
du
yxt t (3.251)
dy
where
du
t L2m (3.252)
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 gener-
ally 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 non-symmetrical.
3.12.5 Near-Wall Flow
Near a wall, the boundary layer can be subdivided into three distinct regions (see Figure 3.5): the
inner viscous sublayer (where molecular diffusion is dominant); overlap or buffer layer; and an outer
layer (where turbulent stresses dominate). In the buffer layer, both molecular and turbulent stresses
are important. Molecular diffusion is dominant in the viscous wall sublayer since this sublayer is
very thin and any transverse turbulent eddy motion is suppressed by the wall. However, in the outer
region, eddy motion and the resulting turbulent stresses become more significant than molecular
diffusion (Figure 3.28).
FIGURE 3.28 Turbulent boundary layer (a) structure and (b) shear stress distribution.
134 Advanced Heat Transfer
In the inner viscous sublayer, the van Driest model can be used to determine the mixing length,
yu
Lm y 1 exp (3.253)
vA
where uτ = (τw/ρ)1/2 is the friction velocity, A+ = 26 and κ ≈ 0.41 (von Karman constant).
Outside the viscous sublayer, the mixing length is proportional to the distance from the wall. In
the overlap region, the velocity profile can be approximated by the law of the wall,
u 1 yu
ln C (3.254)
u v
where C ≈ 5.2. If the surface is rough, the roughness elements will break up and affect the wall
layer. In this case, the coefficients in the law of the wall are modified to κ = 0.4 and C ≈ 3.5 and the
fraction uτ/v in brackets is replaced by 1/yr, where yr refers to the average roughness element height.
With surface roughness, both pressure and molecular diffusion processes affect the wall shear stress.
In the outer layer, the mixing length can be determined from an exponential model,
y
n
Lm 1 1 (3.255)
n
where δ is the boundary layer thickness and n = 5 for boundary layer flow.
where cD ≈ 0.09 and σk ≈ 1 are empirical constants. After solving Equation (3.256) subject to appro-
priate boundary conditions, the spatial variation of turbulent kinetic energy and mixing length can
be obtained, after which the turbulent viscosity and conductivity are found.
From left to right in (3.256), 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 approximately 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 (3.249).
In the above one-equation model, a differential transport equation was used to determine the tur-
bulent kinetic energy and hence a velocity scale based on k using the Prandtl–Kolmogorov model. It
Convection 135
remains to find the other characteristic scales of the turbulent flow. A two-equation model includes
an additional transport equation for another key dependent variable, such as the rate of dissipation
of kinetic energy, ε. As more transport equations are introduced to improve the modeling of turbu-
lence, 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 k3/2/L. Using the latter product, the rate of dissipation of turbulent kinetic
energy, ε, is defined as
k 3/2
c (3.257)
L
where cμ = 0.09 is an empirical constant. Using this definition and Equation (3.249) for the reference
velocity, Vref, the turbulent viscosity becomes:
k2
t Vref L c (3.258)
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, turbulent, and pressure
terms), production, and dissipation. Consider two-dimensional incompressible 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 dif-
fusion plus production of ε, minus the destruction of ε, plus any additional terms, such as buoyancy
(denoted by E), as follows:
t 2
u v
1c P c E (3.259)
y y
y
2
x k k
where
2
u
P t (3.260)
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 (3.256) and (3.259) represent the well-known k – ε model of turbulence. This model
is a two-equation closure that solves Equations (3.256) and (3.259) 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 turbulence 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, empiri-
cal correlations, such as the law of the wall, are usually required in the near-wall region for numeri-
cal solutions. These correlations involve various limitations which constrain 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
136 Advanced Heat Transfer
for more advanced turbulence models, for example, Reynolds stress models (Rodi, 1984), alge-
braic stress models (Hanjalic and Launder, 1972), and renormalization group theory-based models
(Kirtley, 1992).
s s qj
u j Ps (3.261)
t x j x j 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 pro-
duction rate. This result in Equation (3.261) 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 D
T p (3.262)
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 (3.261) is needed to clearly indicate a positive
entropy production rate. To derive this alternative form, the Gibbs equation, Equation (3.262), is
combined with the continuity equation, Equation (3.5) and compared to Equation (3.261). On the
right side of Equation (3.261), the quotient rule of calculus and Fourier’s law are used to express the
Convection 137
heat flux in terms of temperature. Then by comparing the resulting terms with Equation (3.261), it
can be shown that the entropy production rate is a positive-definite sum of squared quantities:
k T T
Ps 2 0 (3.263)
T x j x j 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 this local entropy production
can be a powerful tool in engineering design because it gives a quantitative measure of irrevers-
ibilities 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 irrevers-
ibilities 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 power-
ful design tool that aims to reduce system irreversibilities 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.
UAtube (3.264)
m
where Atube is the cross-sectional area of the stream tube. The energy balance and second law can
be written as follows:
sout sin q As 0 (3.266)
Ps m
Ts
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 Ṗ s 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 (3.267)
1 1 m pout pin
Ps qAs (3.268)
T
Ts T
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 difference 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,
T T F U
Ps qAs s D (3.269)
TsT T
Assuming that ΔT ≪ T, the first term on the left side can be simplified, yielding:
2
q A F U
Ps s D (3.270)
T h T
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 (3.270) over a flat plate,
2 L L
q Wdx U
Ps
T 0
h
T
wWdx (3.271)
0
The following heat transfer and fluid friction correlations can then be used for a flat plate
boundary layer flow.
hx
=
Nu x = 0.458Pr1/ 3Re1x/ 2 (3.272)
k
w
cf , x 0.664Rex1/ 2 (3.273)
U 2 / 2
q2L2W U 2 kT 1/ 2
Ps 1/ 3 1/ 2
1.456Pr ReL 0.664 ReL (3.274)
2
kT q2L2
Convection 139
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 Ṗ s with respect to ReL equal to zero, yielding:
q2
ReL ,opt 2.193Pr 1/ 3 2 (3.275)
U kT
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 tem-
perature difference and thermal irreversibility occur, whereas above the optimal length, a larger
surface area creates a higher overall drag force and friction irreversibility. 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 useful
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).
Ds qj
Ps,te (3.276)
Dt x j T
k T T
Ps, pd 2 0 (3.277)
T x j x j T
140 Advanced Heat Transfer
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 (3.261).
Also, the heat flux can be expressed in terms of the temperature gradient through Fourier’s law.
Then, subtracting Equation (3.277) from Equation (3.276), and using the quotient rule of calculus
to combine and cancel terms with temperature derivatives, yields the following “apparent entropy
production difference”:
Ds k T
Ps Ps,te Ps, pd (3.278)
Dt T x j x j 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, approxi-
mation 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 (3.276) in terms of temperature as follows:
c p DT k T
Ps,T 0 (3.279)
T Dt x j T x j
where the right side must be equal or greater than zero, according to the second law of thermody-
namics. However, numerical approximation errors may violate the required positivity as a conse-
quence of the nonzero entropy production difference. Approximation errors leading to a negative
numerical entropy production are anticipated to lead to non-physical 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 validate the simulation results.
Postprocessing of the temperature results can be used to determine their physical plausibility
based on the sign computed in Equation (3.279) and magnitude of the computed 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 illustrated in the following example.
T 2T
cp k 2 (3.280)
t x
Convection 141
FIGURE 3.30 Heat conduction in a long bar: (a) approximate temperature profile and (b) entropy
compliant region.
Heat is conducted gradually into the bar and the temperature disturbances are propagated into
the positive x-direction. Define δ as the distance of thermal wave propagation. A linear approxi-
mation of the temperature profile is illustrated in Figure 3.30.
Using a scaling approximation to estimate the characteristic temperature, time and length
scales, the linearized derivatives of temperature in the heat equation can be approximated as
T T Ts To
cp s o k (3.281)
t 0
2
This yields a characteristic length of δ 2 = αt which leads to the following wall heat flux,
T T T Ts To
q k k s o ~k (3.282)
x 0 t
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 correct order
of magnitude of the heat flux except for the leading coefficient.
To determine the physical plausibility of the approximate temperature profile, it is useful to
determine whether or not it is consistent with the second law. Substituting the approximate profile
into the temperature based form of the entropy generation rate in Equation (3.279),
1x Ts To
Ts x Ts To 0 (3.283)
2t t t
az 2 bz c 0 (3.284)
where z = x / t and
a To Ts ; b Ts ; c 2 Ts To (3.285)
The solution of this quadratic equation yields a two-dimensional region in x-t space, defined by
x / t ≥ d (a constant), where d is the solution of a quadratic equation,
b b 2 4ac
d (3.286)
2a
Therefore the region satisfying this inequality is the entropy compliant region and consistent
with the second law. A schematic of the entropy compliant region is illustrated in Figure 3.30.
142 Advanced Heat Transfer
In the entropy compliant region, the approximate solution satisfies the second law of thermo-
dynamics. 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 rep-
resents a solution that is nonphysical and may yield misleading trends of variables, correlations or
processes based on the solution.
y u
y ; u (3.287)
L V
T T P
; Ps s 2 (3.288)
Ts T k /L
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
Re L c f (3.289)
2 y 0
hL
Nu (3.290)
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 veloc-
ity gradients. For two-dimensional incompressible flows in a boundary layer, the simplified and
reduced positive-definite form of the entropy production rate, Equation (3.263), becomes:
2 2
u k T
Ps 2 (3.291)
T y T y
1 2
1
4
Ps Pr Ec Re 2L c 2f 2
Nu (3.293)
where the Eckert number is given by
V2
Ec (3.294)
c p T
1 1 2
Ps Re 2L c 2f 2 Nu (3.295)
4 Pr Ec
Therefore the nondimensional entropy generation number can be expressed as the sum of fric-
tional 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 nondimensional entropy generation
number.
REFERENCES
Achenbach, E. 1978. “Heat Transfer from Spheres up to Re = 6 × 106”, Proceedings of the Sixth International
Heat Transfer Conference, Vol. 5, Washington, DC: Hemisphere.
Adeyinka, O.B. and G.F. Naterer. 2002. “Apparent Entropy Production Difference with Heat and Fluid Flow
Irreversibilities”, Numerical Heat Transfer B, 42: 411–436.
Bejan, A. 1996. Entropy Generation Minimization, Boca Raton: CRC Press/Taylor & Francis Group.
Bergman, T.L., A.S. Lavine, F.P. Incropera and D.P. DeWitt. 2011. Fundamentals of Heat and Mass Transfer,
7th Edition, New York: John Wiley & Sons.
Churchill, S.W.. 1983. “Free Convection around Immersed Bodies”, E.U. Schlunder, ed., in Heat Exchanger
Design Handbook, Sec. 2.5.7, New York: Hemisphere, 1–31.
Churchill, S.W. and H.H.S. Chu. 1975. “Correlating Equations for Laminar and Turbulent Free Convection
from a Vertical Plate,” International Journal of Heat and Mass Transfer 18: 1323–1329.
Dincer, I. and M.A. Rosen. 2013. Exergy: Energy, Environment and Sustainable Development, 2nd Edition,
Amsterdam: Elsevier.
Falkner, V.M. and S.W. Skan. 1930. “Some Approximate Solutions of the Boundary Layer Equations”, Great
Britain Aeronautical Research Committee Reports and Memoranda, 1314: 1–35.
Giedt, W.H.. 1949. “Investigation of Variation of Point Unit Heat-Transfer Coefficient Around a Cylinder
Normal to an Air Stream”, Transactions of the ASME, 71: 375–381.
Hanjalic, K. and B.E. Launder. 1972. “Reynolds Stress Model of Turbulence and its Application to Thin Shear
Flows”, Journal of Fluid Mechanics, 52: 609–638.
Hewitt, G.F., G.L. Shires and Y.V. Polezhaev, Eds., 1997. International Encyclopedia of Heat and Mass
Transfer, Boca Raton: CRC Press/Taylor & Francis Group.
Hinze, J.O. 1975. Turbulence, 2nd Edition, New York: McGraw-Hill.
Hollands, K.G.T., 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.
Kakac, S., Y. Yener and A. Pramuanjaroenkij. 2013. Convective Heat Transfer, 3rd Edition, Boca Raton: CRC
Press/Taylor & Francis Group.
Kays, W.M. and M.E. Crawford. 1980. Convective Heat and Mass Transfer, New York: McGraw-Hill.
Kirtley, K. 1992. “Renormalization Group Based Algebraic Turbulence Model for Three-Dimensional
Turbomachinery Flows”, AIAA Journal, 30: 1500–1506.
Kreith, F., R.M. Maglik and M.S. Bohn. 2010. Principles of Heat Transfer, 7th Edition, Pacific Grove: Brooks/
Cole Thomson Learning.
144 Advanced Heat Transfer
Lee, S., 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.
Moody, L.F. 1944. “Friction Factors for Pipe Flow”, Transactions of the ASME, 66: 671–684.
Muzychka, Y. and M.M. Yovanovich. 2016. “Convective Heat Transfer”, in Handbook of Fluid Dynamics, R.
Johnson, ed., Vol. 14, Boca Raton: CRC Press/Taylor & Francis Group, pp. 1–58.
Naterer, G.F. and J.A. Camberos. 2008. Entropy Based Design and Analysis of Fluids Engineering Systems,
Boca Raton: CRC Press/Taylor & Francis Group.
Ostrach, S. 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.
Raithby, G.D. 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.
Rodi, W. 1984. Turbulence Models and Their Application in Hydraulics, Brookfield, VT: Brookfield Publishing.
Schlichting, H. 1979. Boundary Layer Theory, New York: McGraw-Hill.
Tennekes, H. and J. L. Lumley. 1972. A First Course in Turbulence, Cambridge: MIT Press.
Whitaker, S. 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.
White, F. 2015. Fluid Mechanics, 8th Edition, New York: McGraw-Hill.
Zhukauskas, A. 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.
PROBLEMS
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 maintaining 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.
2. In this problem, Bernoulli’s equation is derived by integrating the mechanical energy equa-
tion (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 depth 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 integrated form of the internal
energy equation to obtain Bernoulli’s equation for compressible 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. 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 determine 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.
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 convective heat
transfer coefficient of 20 W/m2K. 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 convection coefficient for both
Convection 145
cases, what proportion of the total heat loss is due to evaporative cooling during the humid
day? Use a diffusion coefficient (air-H2O) of 2.6 × 10−5 m2/s.
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 measured 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.
6. Under what conditions of pressure and temperature will sublimation occur?
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.
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 of To which
suddenly has its surface temperature changed to Ts and maintained at that temperature. A
“thermal wave” propagates into the material and self-similarity of temperature profiles is
maintained over time. Find a similarity solution of this problem using the similarity vari-
able of θ = (T − Ts)/(To − Ts) as a function of η = xg(t).
9. Air flows at a velocity of 10 m/s above a flat surface of length 1.2 m. The air and surface
temperatures are 20°C and 75°C, respectively. Find the required plate width for a total heat
transfer rate of 800 W by convection from the surface to the air.
10. Consider frictional heating within a laminar boundary layer flow where the freestream flow
(at a temperature of T∞ and velocity of U) is parallel to an adiabatic wall. Assume that the
Blasius velocity solution may be adopted in the fluid flow problem.
a. Perform a scaling analysis to find the relevant boundary layer energy equation and
explain the meaning of each resulting scaling term. Show that the increase of wall tem-
perature scales as U2/cp and U2μ/k for cases where Pr ≫ 1 and Pr ≪ 1, respectively.
b. Find a similarity solution of this problem based on the following similarity variables:
U
y
vx
Uvx f
T T
U 2 / 2c p
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.
11. Consider a scaling analysis of transient heat conduction in a long bar. Assume one-dimen-
sional 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.
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?
146 Advanced Heat Transfer
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?
14. Air at 17°C flows past a flat plate at 77°C with a velocity of 15 m/s.
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?
15. A moist airstream with a velocity of U and a temperature of T∞ flows over a container hold-
ing 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 convective mass transfer
coefficient, hm, in terms of the measured recession rate of the liquid, dh/dt, in the
container.
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 and 60 cm, respectively.
a. What spacing between the plates should be used to reach a fully developed flow condi-
tion at the channel outlet?
b. At what position within the channel does the mean temperature fall 5°C below its inlet
value?
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 A2Ω). 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.
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 electri-
cal resistance of 4 × 10-4 Ω/m of cable length. Estimate the surface temperature of the cable
beneath the ice layer.
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)?
20. Metallic spheres (diameter of 2 cm) at 330°C are suddenly removed from a heat treatment
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 signifi-
cantly throughout the sphere? Neglect radiative heat exchange with the surroundings.
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.
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,
Convection 147
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 temperature of the refriger-
ant, Tw, to cool the oil to 316 K at the outlet of the tube.
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 elements pro-
vide 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 and 40°C, respectively, then find the inner wall tem-
perature at the outlet.
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 conditions and constant thermophysi-
cal 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 m, 1.2 m,
and 1.5 m, respectively.
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.
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 di and do, respectively. A uniform heat
flux, qwi, is applied at the inner surface and the outer surface is maintained 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
c pum r
x r r r
Solve this equation (subject to appropriate boundary conditions) to obtain the fluid tem-
perature distribution within the annulus.
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).
27. Air flows at 0.01 kg/s through a pipe with a diameter and length of 2.5 cm and 2 m, respec-
tively. 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.
148 Advanced Heat Transfer
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?
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 temperatures outside and
inside the wall are −20 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.
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 ambient temperature of
−25°C. The air temperatures in rooms A and B are 25 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.
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 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 radia-
tion 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 circuit for this
problem.
b. The transistor temperature should not exceed 70°C for reliable and effective perfor-
mance. 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?
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 and pipe wall resis-
tances are small in comparison to the thermal resistances associated with convection.
32. Perform a discrete scaling analysis for natural convection from a vertical plate. Consider a
flat plate maintained at T = Ts adjacent to a quiescent freestream at a temperature 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 available correlations for natural convection.
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?
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 separate 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, 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 pro-
files in the cross-stream (y) direction.
Convection 149
35. In a manufacturing process, a six-sided metal rod is heated by laminar free convection. 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, i.e.,
higher or lower NuL in comparison to part (a)? Explain your response.
36. Small indentations along a vertically oriented surface are proposed for more effective con-
vective 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 temperature is T∞ and the wall
temperature is Tw (where Tw > T∞).
a. Calculate the average Nusselt number for natural convection in this configuration. 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 mag-
nitudes of L1 and L2? Explain your response.
37. Liquid benzene is heated by free convection by a vertical plate of 10 cm height with a sur-
face 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?
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.
39. Thin metal plates are suspended in air at 25°C before processing in a manufacturing opera-
tion. What initial plate temperature is required to produce an initial rate of temperature
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.
40. Optimizing the parameters within a pipe flow has important implications in the design of
heat exchangers, underground oil pipelines carrying and many other applications. 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 diameter based on fixed values of ṁ
and q′ and the method of entropy generation minimization.
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 properties. Use drag
and heat transfer correlations of cD = 0.5 and NuD = C Re1/2 for external flow past the
sphere, where C = C(Pr).
42. A heat exchanger design involves turbulent flow of water at 310 K (average fluid tempera-
ture) through a tube while requiring a mass flow rate of 12 kg/s and a temperature rise of 6
K/m. Find the optimal diameter based on entropy generation minimization.
43. Air flows at 20 m/s and 400 K across a plate heated uniformly at 100 W/m. Find the opti-
mal plate length based on the method of entropy generation minimization.
150 Advanced Heat Transfer
44. Turbulent flow of air past a plate of length L and width W is encountered in an electronic
assembly. Under these conditions, the following correlations for heat transfer (Nusselt
number) and friction coefficient are adopted:
x
c f ,x 0.0576 Rex1/ 5
U 2 / 2
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.
45. A brass cylindrical fin is joined to the top surface of an electronic component to enhance
convective cooling of the component. Correlations for heat transfer and the drag coeffi-
cient, cd, in regard to external flow past the fin, are known as follows:
=NuD 0=
.0683 Pr1/ 3 Re 0D.466 ; cD 1.2
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 freestream 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 bet-
ter convective cooling of the microelectronic system.
46. Consider one-dimensional transient fully developed laminar flow between two parallel
plates spaced a distance of L apart (Couette flow). The upper plate temperature 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.
4 Thermal Radiation
4.1 INTRODUCTION
Radiative heat transfer is a form of energy emitted as electromagnetic waves or photons by all mat-
ter 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 situ-
ated 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 elec-
tromagnetic 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 surface’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), reflectivity (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 iso-
thermal 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, polished
surfaces often have a higher reflectivity and lower absorptivity than painted and metal surfaces.
Glass is an effective transmitter compared to polymers and metals. Nonconductors (such as ceram-
ics) generally have a higher surface emissivity than conductors (such as metals), except in a direction
perpendicular to the surface. This may be attributed to the smoother faceted surfaces of nonconduc-
tors. A faceted interface that is formed during a solidification process tends to reduce or close inter-
atomic gaps during the forming process of a nonconductor.
The spectral variations of radiative properties with wavelength of materials, fluids, and gases have
significance in our common everyday experiences. A well-known example is greenhouse gases. The
burning of fossil fuels produces carbon dioxide which has a high absorptivity and low transmissivity
in the long-wavelength region of the electromagnetic spectrum. Therefore surface emissions from
objects inside the atmosphere at relatively low temperatures and large wavelengths are essentially
trapped within the atmosphere by the increasingly higher absorptivity of atmospheric gases in the
4.2 FUNDAMENTALS OF RADIATION
4.2.1 Electromagnetic Spectrum
Two fundamental theories are core to the physics of radiation and electromagnetic waves: quan-
tum theory (Max Planck, 1858–1947) and electromagnetic theory (James Maxwell, 1831–1879).
In quantum theory, energy is transported by radiation in the form of photons. A photon is a type
of elementary particle or energy packet which has properties of both waves and particles. Photons
travel at the speed of light. The energy of a photon, e, is proportional to the frequency of radiation,
ν, according to:
e h (4.1)
c (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 oscillations 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 perpendicular 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 fre-
quencies (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 engineer-
ing applications occur in this range. From very small wavelengths, below 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 proper-
ties 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 generally
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. Further 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, particularly the skin and eyes.
Depletion of the ozone in the atmosphere, due to the release of chlorofluorocarbons (CFCs) from
refrigerants into the atmosphere, has a harmful effect since ozone absorbs UV radiation. The next
Thermal Radiation 153
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, yellow, 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 eventually 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.
4.2.2 Radiation Intensity
Radiation emitted by a surface is a directional quantity. The radiant emission from a surface propa-
gates in all directions and the manner in which a surface absorbs, reflects, and transmits radiation
depends on direction. These directional effects are characterized by a concept of radiation intensity.
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
d (4.3)
r2
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.
154 Advanced Heat Transfer
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 located at a distance L from
the first surface,
dA cos
d (4.5)
L2
where dA and θ refer to the surface area and directional angle of the incoming radiation measured
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 I e (4.6)
This result indicates that the total emissive power, E, and total intensity of emitted radiation, 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 irradiation on
a surface (incident radiation), G, and the total radiosity (the sum of emitted radiation and reflected
irradiation), J, as follows:
G Ii (4.7)
J I er (4.8)
where Ii refers to the total intensity of incident radiation. Also, Ie+r is the total intensity of both emit-
ted 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 illustrated in
Figure 4.3. Here “spectral” refers to a dependence on the wavelength, λ. For a transparent medium
Thermal Radiation 155
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 transmis-
sivity, respectively.
The spectral radiosity, Jλ, is the sum of emitted radiation from the surface and the reflected
irradiation,
J E G (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 irradia-
tion multiplied by each respective radiative property,
G G G G (4.12)
1 (4.13)
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.
156 Advanced Heat Transfer
4.2.3 Blackbody Radiation
A blackbody is an ideal surface that absorbs all incident radiation, irrespective of wavelength and
direction. For a given temperature and wavelength, no surface can emit more radiant energy than a
blackbody. Also, a blackbody is a diffuse emitter, as the emitted radiation is independent of direc-
tion. At room temperature, an object having these properties would appear to be perfectly black
(hence the term blackbody).
The spectral distribution of blackbody radiation was first determined by Planck (1959). Using
statistical methods, an exponential probabilistic distribution can be used to show that electrons are
more likely to occupy a particular configuration that has more sites at a certain quantum energy
level. This exponential probabilistic decay of the blackbody spectral emissive power, Eλ,b(λ,T), at a
temperature of T and wavelength of λ is given by Planck’s law:
C1
E ,T ,T I , b ,T (4.15)
5
exp C2 /T 1
where Iλ,b is the spectral radiation intensity. The subscripts b, λ and T denote a blackbody and depen-
dence on wavelength and temperature, respectively. The factor π is obtained from angular integra-
tion of the spectral distribution over the hemispherical range of radiative 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 temperatures above approxi-
mately 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 emis-
sive 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 wave-
length where the emissive power is maximized at a given surface temperature, Planck’s law can be
differentiated with respect to λ and the resulting expression is equated to zero, yielding:
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,
C T 4 x 3dx
0
Eb 1 4 x
C2 e 1
T 4
(4.18)
where σ = 5.67 × 10–8 W/m2K4 is the Stefan–Boltzmann constant. Here Eb represents the total
blackbody emissive power. Using this result, the radiation heat flux emitted from a surface becomes:
Ts4
qrad (4.19)
This result represents Stefan–Boltzmann’s law. The surface emissivity, ε, is the ratio of the radi-
ant 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 scattering of absorbed and reflected radiation,
thereby leading to abrupt or fluctuating changes with wavelength.
Define F(0→λ) as the fraction of total blackbody radiation in the wavelength range from 0→λ. This
fraction can be determined by integrating Eb,λ over the appropriate wavelength range,
E ,b d T
Eb ,
F 0 0
T 5
d T f T (4.20)
E
0
,b d 0
Here Eb,λ is a function of λT only (using Planck’s law) so the band emission is written as a func-
tion of this product alone.
158 Advanced Heat Transfer
F 1 F 0 (4.22)
Sample values of the blackbody function, F(0→λ), at various products of λT are shown in Table 4.1.
Using these blackbody functions, the fraction of emitted energy in any particular wavelength
region can be determined. A common example is finding the fraction of energy 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 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 tem-
perature, 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 sub-
tracting these two values gives the fraction of emitted energy in the specified range of wavelengths.
TABLE 4.1
Radiation Functions for a Blackbody
λT ( μm·K) F(0→λ) λT ( μm·K) F(0→λ)
1,000 0.0003 6,500 0.776
1,500 0.014 7,000 0.808
2,000 0.067 7,500 0.834
2,500 0.162 8,00 0.856
3,000 0.273 9,000 0.890
3,500 0.383 10,000 0.914
4,000 0.481 12,000 0.945
4,500 0.564 15,000 0.970
5,000 0.634 20,000 0.986
5,500 0.691 40,000 0.998
Thermal Radiation 159
,T
E ,T
cos sin d d
0 0
,
(4.23)
Eb, ,T 2 /2
cos sin d d
0 0
The integration is performed over a hemispherical range of angles θ and ϕ in spherical coordi-
nates. Each surface has a unique spectral distribution of emissivity. For example, the spectral emis-
sivity of alumina at 1,400 K is approximately constant in the visible range of the spectrum but then
increases, reaches a maximum value, and subsequently decreases at larger wavelengths.
The total emissivity, ε, is determined by integrating the energy emitted across all wavelengths in
the spectrum,
T
E d E T E T
0
b,
(4.24)
E T T 4
E d
0
b,
b
Similarly, the reflectivity (fraction of energy reflected by the object), ρ, absorptivity (fraction
of energy absorbed), α, and transmissivity (fraction transmitted), τ, can be determined by integra-
tion of the corresponding spectral distributions across all wavelengths. Figure 4.5 illustrates an
example of radiation exchange between two surfaces involving each of these processes. Based on
conservation of energy for the object in Figure 4.5, as shown earlier in Equation (4.14), the sum of
the reflectivity, absorptivity, and transmissivity of the object must equal unity. Radiant energy that
enters the object may be absorbed in the form of thermal energy or transmitted, partly or entirely,
through the surface.
Th reflectivity, absorptivity, and transmissivity are defined analogously to the emissivity, but in
terms of the incident radiation, Gλ,
T
G d
0
(4.25)
G d
0
T
G d
0
(4.26)
G d
0
T
G d
0
(4.27)
G d
0
The energy emitted normally depends more strongly on surface temperature than wavelength.
However, in general, the spectral radiative properties of reflectivity, absorptivity, and transmissiv-
ity 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 radiative properties of the emitting surface and source of incident
radiation. For example, the energy absorbed over a particular wavelength range depends on the sur-
face’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 various materi-
als, 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 reflec-
tor in all wavelength regions. Red paint alternates between good, moderate, and poor reflectivities
across various 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,
, , (4.28)
0
2
0
, cos sin d d
/2
I cos sin d d
0
2
0
/2
,
(4.29)
0 0
cos sin d d
I cos sin d d
0 0
Thermal Radiation 161
Furthermore, if the irradiation corresponds to emission from a blackbody at the surface tem-
perature, then Gλ = Eb,λ. Alternatively, if the surface is gray (ελ = αλ are independent of λ), then the
following further simplification can be applied,
0
E b , ,T d
G d
0
(4.30)
E 0
b , ,T d
G d
0
Therefore, either of the following two conditions can be satisfied to establish the equivalence
between the total emissivity and absorptivity.
1. Irradiation arrives from a blackbody at the same temperature as the incident surface. 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 surface, 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 occur-
ring 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 irradiation matches the emission
from the source of irradiation (sun) over the spectral range of the source.
Therefore, the surfaces possess the following characteristics:
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.
Radiative surface properties are often used in temperature measurement devices. For example,
radiation thermometry is a measurement technique for determining the surface temperature 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 differ-
ent wavelengths and the surface emissivity are used in a compensation algorithm to determine the
surface temperature. Wen and Lu (2010) used an emissivity model and multispectral radiation ther-
mometry (MRT) to measure surface temperatures in steel manufacturing processes. MRT uses radi-
ance measurements at three or more wavelengths and an emissivity model to determine an object’s
surface temperature.
qij
Fij = (4.31)
Ai J i
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 simulta-
neously 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 ele-
ment 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 subtended by surface j when viewed from surface i is given by
dA j cos j
d ji (4.32)
L2
Using Equation (4.6), the intensity of radiation, Ii, may be written as the radiosity, Ji (consisting
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 i cos j
qij J i
L2
dAi dA j Fij Ai J j (4.34)
Ai A j
1 cos i cos j
Fij
Ai L2
dAi dA j (4.35)
Ai A j
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 i cos j
Fji
Aj L2
dAi dA j (4.36)
Ai A j
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 rectangles with a common side are
FIGURE 4.7 View factors of (a) aligned parallel rectangles and (b) coaxial parallel disks. (adapted from
Howell, 1982)
164 Advanced Heat Transfer
FIGURE 4.8 View factors of perpendicular rectangles with a common side. (adapted from Howell, 1982)
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
B B 4 r2 /r1
2
F12 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.
A comprehensive list and analysis of radiation view factors for many geometrical configurations
was reported by Howell (1982).
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 exam-
ple). 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 A J
j 1 i i
(4.41)
where i = 1, 2, 3, … n. The lower limit in the summation represents the radiation leaving surface 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:
F 1
j 1
ij (4.42)
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 J i Gi (4.43)
where Ji, Gi, and qi refer to the radiosity, irradiation, and heat flow required to maintain surface i at
Ti, respectively. The energy balance requires that the energy needed to maintain surface 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 irra-
diation 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 surface is diffuse gray
(αi = εi) and opaque (τi = 0),
i i i i 1 (4.44)
The blackbody radiation emitted, Ebi, can be evaluated by Stefan–Boltzmann’s law based on a
surface temperature of Ti.
Substituting the expression for Gi into Equation (4.43) and rearranging terms,
Ebi J i
qi (4.46)
1 i / i 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.
Consider again the radiation exchange among individual surfaces in the enclosure. The radiosity
leaving a particular surface eventually becomes part of the irradiation on another surface within the
enclosure. Considering all irradiation terms arriving on a specific surface,
n
AiGi F A J
j 1
ji j j (4.47)
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 F J
j 1
ij j (4.48)
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
J i Ti 4 1 i F J
j 1
ij j (4.49)
This result indicates that the radiosity leaving surface i includes the emitted radiation and the
reflected irradiation.
Using Equations (4.43) and (4.48), the heat transfer to/from surface i can then be expressed in
the following form,
n
qi Ai J i A F J
j 1
i ij j (4.50)
Thus the heat required to maintain surface i at Ti balances the radiant energy leaving surface 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 J i Fij J j
Fij J i J j q ij (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
Ji J j
qi 1/ A F
j 1 i ij
(4.52)
168 Advanced Heat Transfer
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
non-ideal emissivities, based on Equation (4.46), and a spatial resistance(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 surface
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 surfaces are
blackbodies, then the same above equation may be used but with Ji replaced by Ebi and εi = αi = 1.
4.5.3 Two-Surface Enclosures
The previous results can be applied to a specific configuration of a two-surface enclosure. Consider
radiation exchange between two diffuse gray surfaces in an enclosure at temperatures of T1 and T2,
respectively (see Figure 4.11). The surface emissivities of surfaces 1 and 2 are ε1 and ε2, respec-
tively. The objective is to determine the net radiation exchange between each surface as a function
of the surface temperatures, emissivities, and view factor.
The heat exchange between surfaces 1 and 2 may be written as the potential difference between
the two surfaces divided by the sum of resistances,
Eb1 Eb 2
q1 q12 q2 (4.53)
1 1 / A11 1/ A1F12 1 2 / A2 2
The denominator includes the sum of both surface and spatial resistances. Using Stefan–
Boltzmann’s law and rearranging,
q12
1 A1 T14 T24 (4.54)
1 1 1 /F12 A11 1 2 / A2 2
Using this general result, other special cases can be obtained as follows for two diffuse gray sur-
faces 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),
q12
A1 T14 T24 (4.55)
1/1 1/ 2 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
q12
A1 T14 T24 (4.56)
1/1 r1 /r2 1 2 / 2
c. For concentric spheres, F12 = 1 and A1/A2 = (r1/r2)2 so the next heat exchange becomes:
q12
A1 T14 T24 (4.57)
1/1 r1 /r2 1 2 / 2
2
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.
170 Advanced Heat Transfer
d. For a small object (surface 1) in an enclosure (surface 2), A1 ≪ A2, F12 = 1, and therefore:
q12 A11 T14 T24 (4.58)
A radiation shield is a protective layer constructed from a low emissivity (high reflectivity) mate-
rial 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 differ-
ent 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 surface 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
q12
A1 T14 T24 (4.59)
1/1 1/ 2 1 s1 / s1 1 s 2 / s 2
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
J i i Ti 4 1 i F J
j 1
ij j (4.60)
qi A F J J
j 1
i ij i j (4.61)
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 radia-
tion 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 simultane-
ously. 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 use-
ful 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.
Thermal Radiation 171
dI a I ds (4.62)
where the proportionality constant, aλ, is called the extinction coefficient in units of reciprocal length
(m–1). It is usually tabulated as a function of wavelength, λ, gas temperature, Tg, partial pressure, and
concentration Ci of the i components in the gas mixture. The extinction coefficient, aλ, is composed
of two parts – an absorption coefficient, aλ,a, and a scattering coefficient, σλ,s:
a a ,a ,s (4.63)
S
I S I 0 exp a s ds
0
(4.64)
This result is known as Bouguer’s law, named after Pierre Bouguer (1698–1758). It states that
the intensity of radiation decays exponentially as it passes through an absorbing-scattering medium.
The exponent is the integral of the local extinction coefficient over the path length traversed by the
radiation. Bouguer’s law is sometimes also called Lambert’s law (Johann Lambert; 1728–1777).
Consider the multi-directional processes and angular dependencies of absorption and scattering
of radiation intensity (see Figure 4.13). The radiation intensity, Iλ, is scattered in the direction of
(θ, ϕ) from incident radiation within a solid angle of dωi. The scattering cross-section, σλ, represents
the apparent frontal area of a particle available to an incident beam which characterizes the particle’s
ability to deflect radiation from the beam. The apparent scattering area of a cloud of particles, d2Aλ,s,
is related to the scattering areas of individual particles as follows:
d 2 A ,s N s dV N s dsdA (4.65)
172 Advanced Heat Transfer
where Ns, σλ, dV and dA are the number density of particles (particles per unit volume), average
scattering cross-section of the particles, differential volume of the particle cloud, and cross-sectional
area of the incident beam, respectively. The ratio of the scattering cross-section to the geometric
projected area of a particle normal to the incident beam is called the scattering efficiency factor.
The scattering process depends on the size of particles or droplets immersed in the medium,
shape of the scattering region, wavelength, and polarization of the incident radiation. It can be
shown that the ratio of scattered radiation, dIλ,s, to the intensity of the incident radiation, Iλ, is equal
to the ratio of the projected scattering area occupied by the cloud of particles, d2As, to the cross-
sectional area of the incident beam, dA, i.e.,
dI ,s d 2 As
(4.66)
I dA
In the forward direction past the scattering cloud in Figure 4.13, the radiation intensity decreases
as a result of scattering,
I dI I dI ,s (4.67)
dI ,s dI
N s ds ,s ds (4.68)
I I
where σλ,s is the scattering coefficient as mentioned earlier. This coefficient can be interpreted as
the reciprocal of the mean free path that the beam of radiation will traverse before it is scattered. It
represents the scattering area per unit volume along the path of the radiation beam.
Integrating Equation (4.68) over a path from s = 0 to S, the radiation intensity at position S as a
result of scattering becomes:
S
0
I S I 0 exp ,s ds
(4.69)
If scattering is neglected, then a reduced form of Bouguer’s law which considers only absorption
through the gas layer is known as Beer’s law (August Beer; 1825–1863). Assume that aλ,a is inde-
pendent of position in Equation (4.64) as in the case of a gas of uniform temperature, pressure and
composition. Then integrating Equation (4.64) from s = 0, where I = Iλ(0), to s = S, where I = Iλ(S),
yields the following form of Beer’s law,
From this result, the intensity of radiation decays exponentially due to absorption of radiation in
the gas layer between the surfaces.
I ,
, 2
(4.71)
I
0 0
, sin d d
The phase function can be interpreted as a probability density function or probability that a beam
of radiation is scattered in a particular direction, (θ, ϕ). When integrated over a sphere, the phase
function satisfies the following normalization condition,
2
1
4 , sin d d 1
0 0
(4.72)
Often the azimuthal dependence is removed under the assumption of a spherical particle.
The phase function for isotropic scattering from spherical particles is Φ = 1. For Rayleigh scat-
tering (scattering by particles much smaller than the wavelength of the radiation), named after John
William Strutt, 3rd Baron Rayleigh (1842–1919), the phase function is given by
3
4
1 cos2
(4.73)
Consider scattering from all other directions by incident radiation, Iλ(θ, ϕ), from angles of (θ, ϕ),
passing through the volume dV (see Figure 4.13). The beam of radiation will pass through a path
length of ds/cos θ. Extending Equation (4.68) to include scattering from other directions, the inten-
sity scattered from Iλ(θ, ϕ) into the direction of Iλ(s) can be written as
ds
dI ,s ,s I , , (4.74)
4 cos s
The scattered intensity, Iλ,s, represents the incident energy from the scattered direction, per unit
area normal to the incident intensity. The area normal to Iλ(θ, ϕ) is dA∙cos θs. Then the spectral
174 Advanced Heat Transfer
energy scattered into the s direction of Iλ(s) as a result of Iλ(θ, ϕ), per unit dλ, per unit scattered solid
angle, per unit incident solid angle dωi, and per unit area normal to Iλ(θ, ϕ) is given by
Using Equation (4.74), the contribution of the scattered radiation to the intensity in the s direction
becomes:
d 5Q ,s ,
,s I , di ds (4.76)
dAd d 4
The left side represents the change of intensity, dIλ,s, in the direction of Iλ(s) by incoming scatter-
ing from other directions. Integrating over all solid angles to account for scattering from all direc-
tions leads to:
2
dI ,s ,s
ds
4 I
0 0
, , sin s d d
(4.77)
ds
4 I
0
d
(4.78)
This result assumes that scattering particles are randomly distributed so that the scattering cross-
section is independent of the incident radiation.
I ,1 I ,2 (4.79)
Alternatively, at a position of the subvolume, s, the spectral intensity is the same as the blackbody
intensity from the outer sphere surface, i.e.,
I s I ,b T (4.80)
The change of intensity as a result of the beam of radiation passing through the absorbing gas
subvolume is given by
I 0 I s I 0 1 exp a ,a ds I ,b T 1 exp a ,a ds (4.81)
Expanding the exponential term by a power series, and neglecting higher order terms when dV
and ds are sufficiently small,
dI I ,b T a ,a ds (4.82)
The energy absorbed by the subvolume ds∙dAs from the incident radiation becomes:
where dω and dA are the solid angle and projected area normal to Iλ(0).
Performing an energy balance within the enclosure, the subvolume dV must emit energy equal
to the radiation absorbed to maintain steady-state conditions and thermal equilibrium. Equating
the above absorbed energy with the emitted radiation from the subvolume, and integrating over
the surface area of the subvolume, solid angles, and wavelengths, the emitted intensity of radiation
becomes:
d 5Q ,a
dI ,e a ,a I ,b T ds (4.84)
dAs d d
Therefore, the rate of increase of intensity in the s-direction due to local emission as the beam of
radiation passes through the subvolume is given by
dI ,e
a ,a I ,b (4.85)
ds
This result is valid for a non-participating medium with an index of refraction of n = 1. The
refractive index refers to the speed of light in a vacuum divided by the speed of light in the medium.
If n ≠ 1, then a pre-multiplicative factor, n2, is combined with the absorption coefficient. Then the
rate of increase of intensity can be written as
dI ,e T 4
a ,a n2 (4.86)
ds
The pre-multiplicative factor accommodates the effects of light traveling slower through the
medium than in a vacuum. The refractive index, n, is a key parameter in analyzing radiation through
participating media.
176 Advanced Heat Transfer
From left to right, the terms represent the loss of radiation intensity by absorption, loss by scat-
tering, gain by scattering from other directions, and gain by emission. These individual processes
are illustrated in Figure 4.15.
Alternatively, in terms of the solid angle, ω, at a given position, r, in the direction, s,
4
dI r,s
ds
a ,a I ,s I ,s
4 I
0
r,s s d a ,a I ,b
(4.88)
Define the albedo for scattering, Ωλ, as the ratio of the scattering coefficient to the extinction
coefficient,
,s ,s
(4.89)
a a ,a ,s
For absorption alone, Ωλ → 0, while for scattering alone, Ωλ → 1. Also, define an intensity source
function, Iλ,s, as the source of intensity along the optical path from both emission and incoming scat-
tering as follows:
4
I ,s 1 I ,b
4 I
0
r,s s d
(4.90)
dI
I I , s (4.91)
d
where dκλ = aλ(s)∙ds is the optical differential thickness. The equation of transfer is a first-order
linear integro-differential equation since Iλ is both within the integral of the source function and the
primary dependent variable in the differential equation.
An integrated form of the equation of transfer is obtained by multiplying by an integrating factor,
exp(κλ), yielding:
dI d
exp I exp I exp I ,s exp (4.92)
d d
I exp I 0
exp
I ,s d (4.93)
0
From left to right in this result, the intensity at an optical depth of κλ is the attenuated incident
radiation at κλ plus incoming scattering in the s-direction along the path length, reduced by exponen-
tial attenuation due to scattering.
In order to solve this equation of transfer, the temperature and scattering distributions in the
source function are first needed, as well as the absorption and scattering coefficients which also
depend on temperature. The temperature is obtained based on the conservation of energy which
requires the total absorbed radiation in each volume element and hence the intensity at each loca-
tion, integrated over all solid angles and wavelengths. These coupled equations generally require a
numerical solution, although in some simplified geometries and conditions, an analytical solution
may be obtained.
qr,n I r,s cos d
0
(4.94)
The components of this flux in the x-, y-, and z-directions are denoted by qr′′, x , qr′′, y , and qr′′, z, respec-
tively. For example, qr′′, x is obtained by setting θ as the angle between the x-axis and the direction of
beam of radiation intensity, I.
The net radiant energy supplied to a differential control volume, dV, can be determined from an
energy balance on the volume element (see Figure 4.16). The radiation outflows in two dimensions
Dividing by dx∙dy and forming a similar expression in the y-direction yields the following net
radiant energy supplied per unit volume,
It remains to determine an explicit expression for the radiation flux vector in terms of the inten-
sity field, Iλ. Consider the energy absorbed by a volume element dV from the incident radiation, Iλ,
which arrives within a solid angle, dω. This absorbed energy can be written as
I
4 I
0
, d
(4.98)
d Qa
a ,
dVd d 4 dV a ,a I d
2
,a I (4.99)
0 0 0
Also, the radiant energy emitted from the surface dAs of the subvolume dV over a solid angle dω
is given by
The differential subvolume dV is very small so that all energy emitted by dV leaves the subvol-
ume before it is reabsorbed within dV.
Using Equation (4.85) and integrating over all wavelengths and solid angles,
4
d 2Qe
a 0 0
, a I ,b dVd d
(4.101)
By integrating the intensity over the hemispherical range of angular directions, it was shown
earlier in this chapter that the blackbody emissive power, Eλ,b, is the intensity of radiation, Iλ(λ),
multiplied by π, therefore,
d Qe 4dV a ,a E ,b ,T d
2
(4.102)
0
Thermal Radiation 179
Subtracting the energy absorbed from the radiant energy emitted by dV leads to:
qr, j
x j
4 a ,a ,T E ,b ,T I d
0
(4.103)
x j
aa
E T I s d
0
b (4.104)
where aa is the total absorption coefficient. This result represents the net outflow of radiant energy
per unit volume.
4.7.3 Conservation of Energy
Recall the net conduction into a volume element in the energy conservation equation (Chapter 3)
was written as the divergence of the Fourier heat flux vector. The radiant energy term in the previous
section was also expressed in divergence form. Therefore, combining the radiation and conduction
flux vectors together in the energy equation yields:
T T
cv
t
cvu j
x j x j
qc, j qr, j q (4.105)
qc, j T
k (4.106)
x j x j
qr, j
x j
4 a ,a I E ,b d
0
(4.107)
Unlike radiation problems earlier in this chapter, the processes of absorption, emission and scat-
tering within participating media occur not only at the system boundaries, but also at all locations
within the medium. Therefore, a complete solution of the above energy equation requires the tem-
perature, radiation intensity, and physical properties (such as the absorption and scattering coef-
ficients) throughout the medium. This significantly increases the complexity of solving coupled
radiation and convection problems.
Alternatively, in integrated form over the range of solid angles, by substituting the radiative flux
vector, Equation (4.105), into Equation (4.105), yields the following alternate form of the energy
conservation equation:
4
T T T
cv
t
cvu j k aa
x j x j x j
I s E T d q
0
b (4.108)
where q contains other non-radiative source terms. Although there is no scattering coefficient in the
expression for the radiation flux vector and energy equation, the radiation intensity is a function of
180 Advanced Heat Transfer
scattering since it is obtained by integration of the equation of transfer for an absorbing, emitting,
and scattering medium.
S S S
I s I 0 exp a , a , s ds
,a ,s
a I
s
exp
a , a , s ds ds (4.109)
0 0 S
where s* is an integration variable. As discussed earlier, since the absorption and scattering coef-
ficients depend on temperature, a simultaneous solution of the equation of transfer and energy
equation is required. The energy equation involves the integration of radiation intensities and thus
it is closely coupled with the equation of transfer. Approximate solutions for simplified limiting
cases may be obtained, for example, an optically thin medium, or weak scattering relative to
absorption.
I s I 0
a ,a
,s I s ds (4.110)
0
For a transparent gas, there is no attenuation of intensity emitted or scattered along the path after
it enters at s = 0.
Furthermore, if the optical depth is small, then the integral in Equation (4.109) becomes small
relative to the intensity at s = 0, yielding the following strong transparent approximation,
I s I 0 (4.111)
The medium has a low extinction coefficient so that the incident intensity remains unchanged by
absorption, emission, or scattering along a path in the medium.
4.8.2 Emission Approximation
In the previous strong transparent approximation, the optical depth was considered small, and the
local intensity was strongly influenced by incident intensities from the gas boundaries. In the emis-
sion approximation, it was assumed that negligible incoming energy arrives from the gas boundar-
ies. Considering only emission within the medium, without absorption or scattering, the equation of
transfer, Equation (4.109), becomes:
S
I s a ,a s I ,b s ds (4.112)
0
Thermal Radiation 181
Integrating over all wavelengths yields the following total intensity, as a result of all emissions
along a path through the gas without attenuation,
S
I s
a s I s d ds (4.113)
0 0
,a
,b
Earlier it was shown that the blackbody emissive power, Eλ,b, was the intensity of radiation, Iλ(λ),
multiplied by π, so therefore,
S
1
I s aP ,a s T 4 s ds (4.114)
0
where aP,a is the Planck mean absorption coefficient. This coefficient is applicable for the emission
term in a gas of any optical thickness.
2 /2
q I cos d 2
0
I cos sin d (4.115)
0
Integrating Equation (4.115) for an isothermal gas over a path of H/cos (θ) through the gas layer
yields:
1 H
I aP ,a T 4 (4.116)
cos
/2
q 2 aP ,a T 4H sin d 2aP ,a T 4H (4.117)
0
Substituting numerical values yields the following Planck mean absorption coefficient,
q 230
aP ,a 0.012 cm1 (4.118)
2 T 4H 2 5.67 10 8 0.02
4.8.3 Rosseland Approximation
Assume that material properties such as the temperature and absorption coefficient depend only on
the depth of the medium (called a plane–parallel assumption). The direction normal to a plane, z,
182 Advanced Heat Transfer
is related to the direction of a beam of radiation, s, by dz = s∙cos θ. Then the reduced form of the
equation of transfer in Equation (4.109) can be expressed as
dI z
cos a ,a ,s I I ,b (4.119)
dz
Alternatively,
cos dI
I I ,b (4.120)
a ,a ,s dz
Under conditions of local thermodynamic equilibrium, the intensity changes slowly on the scale
of the mean free path. The derivative term in the above equation is assumed to be small relative to
other terms. Then the “zeroth” approximation becomes:
I I ,b (4.121)
0
Replacing Iλ by Iλ,b in the transfer equation yields the following “first-order” approximation,
dI ,b
I I ,b
1
(4.122)
,a ,s
a dz
where β = cos θ. Substituting this expression into the general radiation flux vector, Equation (4.94),
4 1
qr z I cos d 2 I d (4.123)
1 1
0 1
The component of the intensity which is independent of angle, Iλ,b, does not contribute to the flux.
Substituting Equation (4.122) and using the chain rule of calculus yields the following monochro-
matic heat flux,
I ,b T T
1
2 I ,b 4
qr z
a ,a ,s z 1
2d
3 a ,a ,s T z
(4.124)
I ,b T
4 T 1
qr z qr , d
0
3 z a
0
,a
,s T
d (4.125)
Thermal Radiation 183
I ,b T I ,b T 4 T 3
0
T
d
T
I ,b T d
0
T
(4.126)
1 I ,b T
1
0 a , a , s T
d
(4.127)
,b T
I
aR,a
0 T
d
16 T 3 T
qr z (4.128)
3aR,a z
Generalizing this result to multiple dimensions, the Rosseland approximation (Rosseland, 1931)
of the radiation flux in tensor form for an absorbing–emitting gas can be written as
16 T 3 T
qr, j z (4.129)
3aR,a x j
where j = 1, 2 and 3.
The Rosseland approximation can be used when the medium is optically thick. The optical thick-
ness is defined as the natural logarithm of the ratio of the incident to transmitted radiation through
the medium. At large optical depths, the mean free path of photons is small compared to the length
scale of spatial temperature gradients. The above result is valid for an index of refraction, n, of unity.
If n ≠ 1, then a pre-multiplicative factor, n2, is used and the Rosseland approximation becomes:
16n2 T 3 T
qr, j z (4.130)
3aR,a x j
This result (also called the radiation–diffusion approximation) is valid when the optical thickness
is larger than 3. It is generally applicable in optically dense media at large optical distances from
boundaries and regions with high-temperature gradients such as boundary layers.
Consider steady-state laminar boundary layer flow of an ideal absorbing–emitting gas over an
inclined surface (see Figure 4.18). Viskanta and Grosh (1962) presented a similarity solution of the
momentum and energy boundary layer equations using the Rosseland approximation.
The steady-state two-dimensional form of the boundary layer equations can be expressed as
u u p u
u v (4.131)
x y x x x
2
T T T u
c pu c pv keff (4.132)
x y y y y
where the Rosseland approximation in Equation (4.129) has been used for the radiation heat flux.
The effective thermal conductivity, keff, consists of the conduction thermal conductivity, k, and the
radiation conductivity, kr, defined as follows based on Equation (4.129),
16n2 T 3
keff k kr k (4.133)
3aR,a
m 1 U 2m
y ; (4.134)
2 0x m 1
0
u Uf ; U x u1 x m (4.135)
T
(4.136)
T0
aR,a
N (4.137)
4n2 T03
where the subscript 0 designates a variable evaluated at a reference temperature.
Then using the similarity solution method, similarly as the case presented in Chapter 3 for “wedge
flow”, the reduced form of the momentum and energy equations becomes:
0 f 0 f
f 0 f 0 f 0 f 2 0 (4.138)
0
0
FIGURE 4.18 Boundary layer flow with radiation over an inclined surface.
Thermal Radiation 185
FIGURE 4.19 (a) Temperature profiles for Pr = 1.0; β = 0; θw = 1.0 and θ0 = 0.1; and (b) temperature
gradients across the boundary layer. (adapted from Viskanta and Grosh, 1962)
keff
Pr f 0 (4.139)
k
keff 4 3
1 (4.140)
k 3N
The parameter N characterizes the relative roles of the conduction and radiation terms. For values
of η less than 4% of the thermal boundary layer thickness, a coefficient of 2/3 is used instead of 4/3
in Equation (4.140).
Sample results in Figure 4.19(a) show that the dimensionless temperature decreases across the
boundary layer from the wall value to the freestream value. As N increases, the temperature profile
approaches the limit for pure conduction. The effect of radiation is to increase the boundary layer
thickens similar to the effect of reducing Pr. Also, the result indicates the departure of the tempera-
ture profile from the conduction limit increases for a higher wall temperature.
The variation of the heat flux across the boundary layer, in terms of the temperature gradient,
is shown in Figure 4.19(b). For a cool wall, the heat flux is maximum at the wall, while for a hot
wall, the maximum occurs at some point away from the wall. The region near the wall where θʹ is
essentially constant represents the conduction dominant zone with nearly zero velocities. For a hot
wall, the local radiation heat flux is maximum at the wall and the local conduction flux is maximum
at a point away from the wall.
4.10 SOLAR RADIATION
4.10.1 Components of Solar Radiation
Solar radiation is an essential source of energy for all life on Earth. In outer space (beyond the
earth’s atmosphere), the intensity of emitted radiation, Iλ, from the sun remains constant along a
particular path of travel since no scattering or absorption of radiation occurs in a vacuum. However,
186 Advanced Heat Transfer
once the radiation passes through the top the Earth’s atmosphere, the absorption and scattering of
radiation by dust particles, moisture, and so on, reduces the radiant intensity with distance traveled.
Consider 100% of the incident radiation arriving at the top of the atmosphere. Then this incom-
ing radiant energy is subdivided into four components (see Figure 4.20): 1%–6% scattered back to
space; 11%–23% absorbed by atmospheric gases such as O2, H2O, CO2, and dust; 5%–15% scat-
tered throughout the sky and ultimately arrives back on the Earth’s surface; and 56%–83% arrives
directly on the Earth’s surface.
The magnitude of solar radiation arriving at the top of the Earth’s atmosphere is called the solar
constant, Gs = 1,353 W/m2. This solar energy flux is based on an energy balance over the region
between concentric spheres of the outer edge of the sun and a sphere encompassing the Earth’s orbit.
The energy emitted across the sun’s spherical surface (approximately) balances the energy passing
across a much larger spherical surface area formed around the orbit of the Earth. The spectral dis-
tribution of this extraterrestrial irradiation, from outside of the Earth’s atmosphere, can be closely
approximated by the blackbody emissive power of an object at a temperature of 5,800 K.
The component of solar radiation actually passing through a unit surface area at the top of the
atmosphere, Go, is the component normal to the surface. Thus, Gs is multiplied by cos θ, where θ is
the angle between the incident radiation and the tangent to the surface. The surface 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 4.20 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 predomi-
nantly 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 T 4 (4.141)
where typical values of the emissivity, ε, are 0.97 for water and 0.93–0.96 for soil. For the sky radia-
tion, Gsky,
Gsky Tsky
4
(4.142)
Ts 4 qconv
sGs atmGatm qelec (4.143)
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 irradiation, electrical
heat supplied, surface emission, and convective heat losses.
The absorptivity of solar radiation, αs, is determined by
s
0
G ,sund
E
0
b , 5,800K d
(4.144)
G 0
, sund
E 0
b , 5,800K d
2
s
E d E d 0.8F
0
b ,
2
b ,
0 2 m
0.1 1 F0 2 m (4.145)
E d E d
0
b ,
0
b ,
Using the blackbody function table (Table 4.1) with λT = 2 × 5800 = 11,600 μK, it is observed
that F(0 → λT) = 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 similarly. For
atmospheric irradiation,
atm
G 0
, atmd
0.8F0 2 m 0.1 1 F0 2 m (4.146)
G 0
, atmd
In this case, the product λT in the blackbody function is determined as λT = 2 × 285 = 570 μmK,
yielding F(0 →λT) = 0.0 and αatm = 0.1.
Assuming a diffuse surface to calculate the plate emissivity,
E 345K d E 345K d 0.88F
0
b ,
0
b ,
0 2 m
0.1 1 F0 1 m (4.147)
E d 0
b ,
E d 0
b ,
From the blackbody table (Table 4.1), using λT = 2 × 345 = 690 μmK, it is found that F(0 → λT) =
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 10 4 (4.148)
19.9 10 6
hL
=
Nu L = 0.664 Re1L/ 2 Pr1/ 3 (4.149)
k
Substituting numerical values, the Nusselt number and convection coefficient become 99 and
27.7 W/m2K, respectively.
Thus, the convective heat flux from the plate surface is estimated as
Using this result and substituting the above values into Equation (4.143),
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 irradiation
assumption led to the results that ελ = αλ and ε = αatm.
4.10.2 Solar Angles
The Earth’s position in its orbit influences the solar radiation at a particular location and the resulting
energy balances. Define the following angles with reference to a point P on the Earth’s surface, and
in particular, an inclined surface at that location (see Figure 4.22):
• 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.
A trigonometric analysis can be performed to find relationships between these solar angles. The
declination angle can be written in terms of the day of the year, n, as follows:
360 n 10
sin sin 23.45 cos (4.152)
365.25
360
t sol 12hr (4.154)
24hr
where the solar time, tsol, is based on the local civic time, tloc,civ, as follows:
Et
t sol tloc,civ (4.155)
60 min/hr
where
n 1
B 360 (4.157)
364
for day n of the year. Equation (4.156) 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 (4.155).
The local civic time in Equation (4.155) 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 sin
sin s (4.158)
sin s
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 equi-
nox (March 23) and autumnal equinox (September 21), 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.
As a result, the local solar time becomes tsol = 10:01 – 0.08 = 9:53 a.m. which yields the fol-
lowing 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 sin 23.45 cos 0.37 (4.162)
365.25
50
sin sin 49 sin 21.7 cos 49.83 cos 31.5 cos 21..7 (4.163)
60
192 Advanced Heat Transfer
which yields a solar altitude angle of 13.45° at this time. Furthermore, the azimuth angle is
obtained by
cos i cos 30 cos 13.45 sin 50 sin 13.45 cos 50 0.795 (4.165)
It can be observed that the times of sunset and sunrise in the previous example could be deter-
mined by setting α = 0 and solving the resulting hour angles. Substituting these hour angles into
Equation (4.154) would then give the corresponding times of sunset and sunrise. The total daily
extraterrestrial radiation is obtained by integrating cos θs from sunrise to sunset and multiplying by
the solar constant (I0).
Some incoming radiation from the sun is scattered within the atmosphere, both scattered 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 (scat-
tered) 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
0.095 m z , 0.65 m z ,
atm 0.5 e e (4.168)
Thermal Radiation 193
patm z
m z, m 0, (4.169)
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 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 scatter-
ing and absorption of radiation occurs over the greater distance traveled by the incoming beam of
radiation through the atmosphere. 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 hori-
zontal radiation, Ih, and extraterrestrial radiative flux on a horizontal surface, I0,h, as follows:
1 cos p 1 cos p
I dif , p I dif ,h I h I dir ,h (4.171)
2 2
where
Ih
0.8302 0.03847m z, 0.04407 CC 0.011013 CC 0.01109 CC (4.172)
2 3
I 0,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 (4.171) 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 hori-
zontally 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 (4.171) is the reflected component of solar radiation:
1
I ref , p
2
g I h 1 cos p (4.173)
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.
The accurate calculation of solar angles and the incident radiation heat flux is an important ele-
ment 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.
194 Advanced Heat Transfer
360 38 10
sin sin 23.45 cos 0.27 (4.174)
365.25
sin sin 49.8 sin 15.7 cos 49.8 cos 15.7 cos 40 (4.175)
cos i cos 48.6 0 cos 15.7 sin 60 sin 15.7 cos 60 (4.177)
cos i cos 48.6 0 cos 15.7 sin 0 sin 15.7 cos 0 0.27 (4.178)
1/ 2
m 0, 614 sin 15.7 1229
2
, 614 sin 15.7 3.66 (4.179)
atm 0.5 exp 0.095 3.65 exp 0.65 3.65 0.4 (4.180)
Idir , p Idir cos i I0 atm cos i 1353 0.4 cos 46.7 371.2 W//m2 (4.181)
For the second component (diffuse), under partial cloud cover (CC = 0.2),
Ih
0.8302 0.03847 3.65 0.04407 0.2
I0,h (4.182)
0.011013 0.2 0.01109 0.2 0.68
2 3
where
As a result,
Also,
Idir ,h I0 atm cos i ,h 1, 353 0.4 0.27 146.1 W/m2 (4.185)
1 cos 60
Idif , p 102.3 2
76.7 W/m (4.187)
2
1 cos 60
Iref , p 0.68 2
248.4 42.2 W/m (4.188)
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.
4.11 SOLAR COLLECTORS
4.11.1 Collector Efficiency and Heat Losses
Solar energy is a rapidly growing form of renewable energy that is being increasingly adopted
worldwide as a replacement of hydrocarbon-based power systems. Examples of solar energy tech-
nologies include photovoltaic cells, solar thermal energy, solar water heating systems, and solar-
based production of hydrogen as a clean energy carrier. As the cost of solar energy systems has
fallen significantly over the years, the number of grid-connected solar photovoltaic systems and
utility-scale solar power stations has grown to become an inexpensive, low-carbon technology.
A key component of various types of solar energy systems is the solar collector. A solar collector
is a heat exchanger that absorbs incident solar radiation and transfers it to a working fluid, such as
an ethylene glycol-water solution,. This process increases the fluid temperature for purposes such as
space or water heating in buildings.
The main components of a solar collector include the top glass cover(s) above an absorptive col-
lector plate and tubes carrying the working fluid. The tubes are bonded and joined to an absorber
plate. The back and bottom sides of the solar collector are insulated to reduce heat losses so that
most of the incoming solar energy is transferred to the working fluid. By adding a top glass cover(s),
convective heat losses to the environment can be reduced, thereby increasing the net amount of heat
absorbed by the working fluid. A single-glazed collector refers to a single glass cover, whereas a
double-glazed collector refers to two glass covers above the absorber plate.
Based on an overall energy balance for a solar collector,
From left to right in this energy balance, the energy absorbed by the absorber plate balances
the undesired heat losses to the surroundings due to reflected radiation and convective 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 transmitted 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
c (4.190)
Ac I p
Integrating over a characteristic time period, T, such as an entire day, yields the following average
collector efficiency:
T
c
q dt (4.191)
T
0
u
A I dt
0
c p
Consider a solar collector with two glass covers (top and lower covers), absorber plate, collec-
tor tubes beneath the absorber plate, and insulation at the bottom side of the solar collector (see
Figure 4.23). 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 col-
lector plate, a series sum of resistances, includes R1 (between Tc and the base of the 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)
FIGURE 4.23 Flat plate solar collector with two glass covers.
Thermal Radiation 197
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 resis-
tance includes heat transfer from both the top (convection to the air, and reflected and emitted radia-
tion) 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 (4.192)
Rtot R1 R2 R3 R4 R5
In the first fraction on the right side, heat losses from the insulated bottom side of the collector
will be neglected. The heat losses to the surroundings are written as follows:
qloss U c Ac Tc Ta (4.193)
Based on the portion of the thermal network between the absorber plate and bottom glass cover,
qtop,loss Achc 2 Tc Tg 2
Ac Tc4 Tg42
Tc Tg 2
(4.194)
1/ p,i 1/ g 2,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,
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,
Tg1 Tair
qtop,loss Achc,a Tg1 Tair g1,i Ac Tg41 Tsky
4
R4
(4.196)
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 q p 0.00013 q 2p (4.197)
f 1 – 0.4 hc,a 0.0005 hc2,a 1 0.091 N (4.198)
Then the heat loss from the top cover of the solar collector is estimated by
qtop,loss
Ac Tc Ta
Ac Tc4 Ta4 (4.200)
1/ p 0.05N 1 p 2 N f 1 / g N
0.33
NTc /C Tc Ta / N f 1/hc
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.
4.11.2 Temperature Distribution
In order to determine the plate temperature distribution in the x-direction, consider an energy balance
across a differential section of the absorber plate (thickness t) between the tubes (see Figure 4.24).
Under steady-state conditions,
dTc
qx k (4.202)
dx
Expanding the heat flux at position x + dx about the value at x using a Taylor series expansion,
dqx
qx dx qx dx (4.203)
dx
where higher order terms can be neglected for a differential thickness of the control volume, dx.
Substituting the heat fluxes into Equation (4.201) and rearranging,
d 2Tc U c I
2
Tc s s Ta (4.204)
dx kt Uc
Solving this equation, subject to Tc = Tb at x = W (above the tube) and dTc/dx = 0 at x = 0 (midway
between the tubes), yields:
sIs I cosh mx
Tc x Ta Tb Ta s s cosh mW (4.205)
U c Uc
where cos h(mW) = (emW + e–mW)/2 is the hyperbolic cosine function and m2 = Uc/(kt). Differentiating
with respect to x, and evaluating at x = W, yields the following heat flux per unit depth,
dTc I
qfin kt
dx
ktm Tb Ta s s
Uc tanh mW (4.206)
W
where tan h(mW) refers to the hyperbolic tangent function. Alternatively, based on the definition of
the fin efficiency (Chapter 2),
where the factor 2 is introduced for heat transfer to a tube from two sides. The term in brackets 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 difference (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
fin (4.208)
mW
Thus the fin efficiency increases with a decreasing total conductance, larger plate conductivity,
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 thermal conductivity (lower m).
Combining the heat flows from the fin, Equation (4.207), 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,
dT f
qu mc
p
dy
Dhc Tb T f y (4.210)
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.
eliminated, the following alternative expression for the heat gain, per unit depth, of the working fluid
is obtained:
Typical values of the total conductance, Uc, are 4 W/m2K (two glass covers) and 8 W/m2K (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/U c 1 1
Fc (4.212)
D 2W Uc D 2W fin hc Di
On the right side, the terms within square brackets refer to the fin resistance and convection resis-
tance to fluid flow in the tubes, respectively. This collector efficiency factor represents 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 (4.211) varies from Tf,in at the tube inlet to Tf,out at the tube out-
let. 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 4.24 of differential thickness dy in the y-direction (along the
tube axis). Performing a energy balance for this control volume,
dT f
mc
p qu D 2W Fc s I s U c T f Ta (4.214)
dy
T f ,out Ta s I s /U c U FG
exp c c (4.215)
T f ,in Ta s I s /U c cp
m
G (4.216)
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 (4.213),
Gc p U c Fc
Fr 1 exp (4.218)
Uc Gc p
Thermal Radiation 201
The heat removal factor characterizes the ability of the working fluid to remove heat through the
fin and collector.
Using Equations (4.190) and (4.217), an alternative expression for the solar collector efficiency
becomes:
qu T T
c Fr s s FrU c f ,in a (4.219)
Ac I p Ip
The solar collector efficiency is typically plotted in terms of an independent variable, XT = (Tf,in –
Ta)/Ip. The efficiency becomes nearly linear with respect to XT since αsτs, Fr, and Uc are usually
approximately constant for a given solar collector design. The intercept of the curve is Frαsτs at XT
= 0 and the slope (negative) is FrUc. A high intercept with a shallow slope leads to a high collec-
tor efficiency over a wide range of fluid and incident radiation conditions. Although a curve with a
steep slope may provide a higher efficiency at lower values of Ip, it would drop off to lower values at
higher Ip. Since XT varies with the time of day, surface orientation, and other factors (due to Ip in the
denominator), a trend of a shallow slope of the ηc – XT curve is usually preferable.
The objectives of a solar collector design include simultaneously achieving a high αsτs, high Fr ,
and low Uc. This design would lead to a higher collector efficiency and energy gain by the working
fluid for a given level of incident solar radiation. An example is provided below to illustrate the vari-
ous components of a system design.
The previous sections have described selected elements of solar energy systems. A further com-
prehensive treatment of solar energy systems is provided in other excellent sources such as Duffie
and Beckman (1974), Hsieh (1986), Garg (1982), and Howell and Bereny (1979).
Uc 8
m 9.19 (4.220)
kt 237 0.0004
Then the collector efficiency factor and heat removal factors are given by
1
1 1 1
Fc 0.934 (4.222)
8 0.1 8 0.01 0.09 0.95 1100
, 0.01
4, 200
m 8 0.934
Fr 0.74 1 exp (4.223)
2.2 3.1 8 4, 200m / 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 T T
c Fr s s Uc f ,in a (4.224)
AcIp Ip
320 288
0.13 0.74 s 0.92 8 (4.225)
400
which gives a required collector plate absorptivity of 0.89.
REFERENCES
Baker, T.M. 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.
Bergman, T.L., A.S. Lavine, F.P. Incropera, and D.P. DeWitt. 2011. Fundamentals of Heat and Mass Transfer,
7th Edition, New York: John Wiley & Sons.
Cengel, Y.A. 2008. Introduction to Thermodynamics and Heat Transfer, 2nd Edition, McGraw Hill.
Duffie, J.A. and W.A. Beckman. 1974. Solar Energy Thermal Processes, New York: John Wiley & Sons.
Garg, H.P. 1982. Treatise on Solar Energy, New York: John Wiley & Sons.
Hottel, H.C. 1954. “Radiant Heat Transmission”, in W.H. McAdams, Ed., Heat Transmission, 3rd Edition, New
York: McGraw–Hill.
Howell, J.R., M.P. Menguc, and R. Siegel. 2015. Thermal Radiation Heat Transfer, 6th Edition, Boca Raton:
CRC Press/Taylor & Francis Group.
Howell, J.R. 1982. Catalog of Radiation Configuration Factors, New York: McGraw–Hill.
Howell, J.R., R.B. Bannerot and G.C. Vliet. 1982. Solar–Thermal Energy Systems: Analysis and Design, New
York: McGraw–Hill.
Howell, Y., and J.A. Bereny. 1979. Engineer’s Guide to Solar Energy, San Mateo: Solar Energy Information
Services.
Hsieh, J.S. 1986. Solar Energy Engineering, Upper Saddle River, NJ: Prentice Hall.
Klinzing, G.E. 1981. Gas–Solid Transport, New York: McGraw–Hill.
Kreith, F., R.M. Manglik and M.S. Bohn. 2010. Principles of Heat Transfer, 7th Edition, Stamford: Cengage
Learning.
Lienhard IV, J.H., and J.H. Lienhard V. 2000. A Heat Transfer Textbook, 3rd Edition, Cambridge: J.H. Lienhard
IV and J.H. Lienhard V.
Lynch, P., 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.
Planck, M. 1959. The Theory of Heat Radiation, New York: Dover Publications.
Rosseland, S. 1931. Astrophysik auf Atom–Theoretischer Grundlage, pp. 42, Berlin: Springer–Verlag.
Wen, C.D. 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.
Thermal Radiation 203
Viskanta, R. and R.J. Grosh, 1962. “Boundary Lyaer in Thermal Radiation Absorbing and Emitting Media”,
International Journal of Heat and Mass Transfer, 5: 795–806.
PROBLEMS
1. The tungsten filament in a light bulb reaches a temperature of 2,800 K.
a. Find the wavelength corresponding to the maximum amount of radiative emission at this
temperature.
b. What range of wavelengths contains 90% of the emitted radiation? Find the range con-
taining the largest portion of the visible range.
2. A cube with sides of 0.5 m is exposed to air at a temperature of 300°C. The cube is assumed
to act as a blackbody. Determine the rate at which the cube emits radiation energy and its
spectral emissive power at a wavelength of 7 μm.
3. Find the amount of radiation emitted by a surface at 1,600 K with the following spectral
emissivities: (i) ε = 0.4 for 0 < λ < 1 μm; (ii) ε = 0.8 for 1 ≤ λ < 4 μm, and (iii) ε = 0.3 for
λ = 4 μm. Express your answer in units of W/m2.
4. The reflectivity of chromium is 0.2 for radiation at a wavelength of less than 5 μm and 0.99
for radiation above 5 μm. If the temperature is 295 K, determine the average reflectivity of
the chromium surface arriving from surfaces at room temperature. Also, find the absorptiv-
ity and emissivity of this surface at room temperature.
5. What wavelength range contains 80% of emitted solar energy? Explain a general procedure
for determining the wavelength range that contains a specified percentage of emitted
energy from a blackbody at a prescribed temperature.
6. A flat exterior spacecraft surface consists of a 12 cm thick layer of material with a thermal
conductivity of 0.04 W/mK. The Biot number corresponding to convection on the inner
surface is 90 (based on the wall thickness). The exterior wall emissivity is 0.1. Find the air
temperature inside the spacecraft which leads to net heat losses of 20 W/m2 from the exte-
rior wall into space at 0 K.
7. The energy flux from solar radiation falls on the outer surface of the Venus atmosphere at
rate of 1,915 W/m2. The sun and Venus have a diameter of about 1.39 × 106 km and 12,104
km respectively and the distance between Venus and the sun is 1.0894 × 108 km.
a. Using this information, find the emissive power of the sun.
b. Determine the temperature of the sun’s surface while assuming the surface of the sun as
a blackbody.
c. At what wavelength is the spectral emissive power of the sun a maximum?
d. Estimate the surface temperature of Venus. Assume its surface to be a blackbody and
that the sun is the only source of energy to Venus.
8. A spherical satellite orbiting around the earth is exposed to solar radiation of 1,400 W/m2.
To maintain a prescribed temperature, a fraction F of the satellite surface is covered with
an aluminum film (ε = 0.05, αs = 0.11) while the remaining fraction (1-F) is coated with
white zinc oxide paint (ε = 0.9, αs = 0.27). The satellite is assumed to be isothermal with
no internal power dissipation. Find the fraction, F, of the satellite surface to maintain a
temperature of 350 K.
9. The top side of a horizontal plate is exposed to solar and atmospheric irradiation. The back
side of the opaque plate is electrically heated. A layer of insulation covers this back side
below the resistance heating elements. The ambient air and effective sky temperatures are
10°C and 0°C, respectively, and the convection coefficient is 20 W/m2K. Assume that the
plate is diffuse with a hemispherical reflectivity of 0.1 below wavelengths of 1 μm, and ρλ
= 0.8 for λ > 1 μm. Find the solar irradiation when 60 W/m2 of electrical power is required
to maintain the plate surface temperature at 40°C.
204 Advanced Heat Transfer
10. Consider a small object in a vacuum chamber. The hot object of mass m is cooled by radia-
tion exchange with walls of an evacuated chamber at temperatures of Tw. Find the govern-
ing equation describing heat transfer from the object. It may be assumed that the Biot
number is small (i.e., Bi < 0.1; lumped capacitance approximation is valid) and the cavity
can be represented as a blackbody.
11. A furnace heats a diffuse, cylindrical iron ingot with a diameter and length of 0.9 m and 3
m, respectively. The furnace wall temperature is 1,700 K and hot combustion gases at
1,400 K flow across the ingot at 4 m/s from a burner inlet to a flue outlet. The ingot is ele-
vated so that the gas flow approximates a symmetric cross–flow past a cylinder. The iron
surface emissivity is 0.3 for wavelengths below 2 μm and ελ = 0.15 for λ > 2 μm. Find the
steady–state ingot temperature. Assume that combustion gas properties may be approxi-
mated by air properties and the furnace walls are large, compared to the ingot.
12. A steel sheet emerges from a hot–roll section of a steel mill at a temperature of 1,100 K
with a thickness of 4 mm and uniform properties (ρ = 7,900 kg/m3, cp = 640 J/kgK, and k =
28 W/mK). In the spectral range below a wavelength of 1 μm, the sheet has an emissivity
of ε = 0.55. The spectral emissivity values are ε = 0.35 and ε = 0.25 in the ranges 1 < λ < 6
μm and λ ≥ 6 μm, respectively. The convection coefficient between air and the strip is h =
10 W/m2K with T∞ = 300 K. Neglect conduction effects in the longitudinal (x) direction.
a. Determine whether significant temperature gradients exist in the strip in the transverse
(y) direction.
b. Calculate the initial cooling rate (rate of temperature change with time) of the strip as it
emerges from the hot–roll section.
13. A horizontal black plate (plate 1 area of 7 cm2) is located beneath and between two vertical
plates (plate 2 of area of 8 cm2 and plate 3). Both vertical plates emit 33 kW/m2sr in the
normal direction. The angles between each surface normal (first subscript) and the beam of
radiation connecting it with an adjacent surface (second subscript) are given as follows:
θ12 = 60°, θ13 = 50°, θ21 = 40° and θ31 = 70°.
a. Find the solid angle subtended by plate 1 and the required distance between plates 1 and
2 if surface 1 intercepts 0.03 W of irradiation from surface 2. Assume that the surfaces
are sufficiently small and that they can be approximated as differential surface areas.
b. Find the required surface area of plate 3 to yield the same irradiation (0.03 W) from that
surface onto plate 1. Plate 3 is located at the same distance from plate 1 as determined
for plate 2.
14. A small object is cooled by radiation exchange with the walls of an enclosure around the
object. Find the view factors between two surfaces represented by the enclosure walls (sur-
face 1) and the small object (surface 2).
15. A cubical cavity with side lengths of 6 cm has a 3 cm diameter hole in a side wall. Find the
view factor representing the total radiation emitted from the walls to (through) the single
hole.
16. A radiation shield is used in a spectrometer instrument for controlling temperature and
measuring the composition of certain gas molecules in the Earth’s troposphere. The radi-
ation shield encloses a cube–shaped inner metal component. The configuration consists
of an upper half cube (shield, 4 × 4 × 2 cm) with a base that intersects the midplane of
the enclosed inner metal cube (2 × 2 × 2 cm). Find the following radiation view factors:
F12, F21, F11, and F13, where surfaces 1, 2, and 3 refer to the outer shield, inner metal
piece, and surroundings formed by the gap between the shield and the metal piece,
respectively.
17. In a circular cylindrical enclosure, the height is twice the diameter of its base. Find the view
factors of all surfaces in this enclosure.
18. Thermal control of a spacecraft requires no more than 25 W/m2 of heat loss from the exte-
rior surface into space to maintain internal operating conditions. An exterior shield of 10
Thermal Radiation 205
cm thickness is used with a thermal conductivity of 0.04 W/mK, exterior wall emissivity of
0.1, and Biot number of convection for the inner surface of 70 (based on the wall thick-
ness). Find the internal temperature within the spacecraft at a design point of 25 W/m2 of
heat loss.
19. A black spherical object (diameter of 4 cm) is heated in a furnace oven containing air at a
temperature of 160°C. Find the required oven wall temperature to produce a sphere surface
temperature of 60°C with a net heat flow rate of 60 W to the sphere. The convection coef-
ficient between the sphere and surrounding airflow is 50 W/m2 K. Assume that the wall
reflectivity and view factor (wall–wall) are 0.1 and 0.9, respectively.
20. A cubical shaped furnace has sides of 6 m. The ceiling and floor are painted black, while
the surfaces on the sides are reradiating. The temperature of the floor and ceiling is 400 K
and 1000 K, respectively. Find the net rate of radiation heat transfer between the ceiling
and the floor.
21. Consider a long duct of semi-cylindrical shape with a diameter of 3 m. Heat is emitted from
the base (black) surface at a rate of 1,500 W/m2. The side surface has an emissivity of 0.6
and temperature of 700 K. Neglect end effects. Find the base surface temperature.
22. A special coating on a disk surface is heated and cured by a heater located 30 cm above the
lower disk. The two parallel disks are placed in a large room at 300 K. Their radii are 30
cm (absorber surface) and 40 cm (heater), respectively. The emissivity and heat loss from
the absorber surface are 0.42 and 3 kW, respectively. The heater 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.
23. 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.
24. 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 emis-
sivity 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 sur-
face 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.
25. How can Beer’s law be generalized to an equation of transfer for a participating gas–par-
ticle 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 asymptotic trends in free space,
without emission or absorption?
26. Generalize the result from the previous question to a vector form of the equation of transfer
for a participating medium.
27. 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 radia-
tive heat transfer between the gas and walls.
28. 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 expres-
sion that shows the appropriate range of solid angles.
29. 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 vector, qr, and alternatively,
in terms of the intensity of radiation, Iλ.
206 Advanced Heat Transfer
30. Solar collectors are used to deliver heated water in a building. Consider a single solar col-
lector 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°)?
31. 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.
32. 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.
33. The copper tubes of a single–glazed solar collector are connected by a 2 mm thick copper
plate with a conductivity of 400 W/m2K. Water flows through the tubes (0.5 mm thick wall)
with a heat transfer coefficient of 440 W/m2K. The total conductance is Uc = 10 W/m2K.
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?
34. 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 temperature, Ta,
and incoming fluid temperature (within the collector), Tf,in:
T T
c 0.82 7.0 f ,in a
Ip
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 and 4°C, respec-
tively, calculate the incoming fluid temperature, Tf,in, into the solar collector.
c. Find the collector efficiency factor when the mass flow rate of water is 8 kg/min through
the collector tubes.
35. 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, respec-
tively. 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?
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 gas–liquid phase change, per unit mass, is rep-
resented by the latent heat of vaporization. This chapter examines the physical processes of boiling,
condensation, and two-phase flows. Phase change in gas–liquid 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 sev-
eral 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. Transition 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 eventually the bubble
detaches from the surface cavity. Then additional liquid comes into contact with the heating sur-
face 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 conductivity. 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 downward-facing orienta-
tion. 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 micro-
layer forming 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 contact 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 condensation or film condensation.
The liquid does not fully cover or wet the surface in dropwise condensation, 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°C to 20°C below the saturation point, the average heat trans-
fer coefficient typically ranges between 11,000 and 23,000 W/m2K for horizontal tubes and between
5,700 and 11,000 W/m2K for vertical surfaces. For condensation of ethanol along a horizontal tube,
the heat transfer coefficient typically varies between 1,700 and 2,600 W/m2K.
DOI: 10.1201/9781003206125-5 207
208 Advanced Heat Transfer
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.
5.2 POOL BOILING
5.2.1 Physical Processes
Pool boiling refers to boiling along a heated surface submerged in a large body of quiescent liquid.
Liquid motion arises from free convection and mixing due to bubble growth and detachment from
the heated surface. Pool boiling occurs under two possible modes of operation (see Figure 5.1) –
either power-controlled heating or thermal heating. In the first case, power-controlled heating (or
electrical heating) allows the heat flux to be determined and controlled based on the applied current
and voltage. The power setting and heat flux are independently controlled variables while tempera-
ture becomes the dependent variable. 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. Subsequent 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 saturation temperatures 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 convection
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 differences 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. Intermittent 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 addition 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.
Differences in thermophysical properties lead to various boiling curves for different fluids.
Selected thermophysical properties of different solids, gases, and liquids are shown in Appendices
D–F, respectively. Thermophysical properties vary significantly with temperature and pressure.
Such variations lead to changes in the boiling curve at different saturation pressures.
g l v L3 c p T c p c g l v L2
a b d
qw /T
C (5.1)
k 2
h fg k
210 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, surface
length, surface tension, and coefficients to be determined from the Buckingham–Pi analysis, respec-
tively. 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 cor-
relation for the wall heat flux:
g
l
v
1/ 2 3
c p,l T
qw l h fg n
(5.2)
cs. f h fg Pr
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 experimental
data over a range of operating conditions. Typical values of the coefficients for various liquid–sur-
face 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 stainless 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 (isopro-
pyl alcohol–copper); and cs,f = 0.015, n = 1.7 (n-pentane-chromium).
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 investigated in past
studies, such as Vachon, Nix, and Tanger (1968), including modified empirical coefficients and
exponents which account for various surface modifications. Naterer et al. (1998) used similarity and
conformal mapping methods to predict the bubble formation and other nucleate boiling processes
on inclined surface configurations.
The addition of other constituents into the working fluid also significantly affects the heat transfer
characteristics of the liquid–surface combination. For example, at atmospheric 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).
FIGURE 5.3 Rohsenow’s correlation of pool boiling heat transfer for water. (adapted from Rohsenow, 1952)
Gas–Liquid Two-Phase Flows 211
The maximum heat flux (or critical heat flux; CHF) occurs at the transition between nucleate
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 temperature can lead to system damage,
such as melting of components in a nuclear reactor, or overheating of a chemical reactor. Even oper-
ating near the CHF has safety risks due to boiling 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
g l v
1/ 4
0.149h fg v
qmax (5.3)
v2
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 tem-
perature. 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:
l / v
3/ 4
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 similari-
ties in approaching the region of transition boiling from the side of cooling to the minimum 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 superheat 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
g l / v
Ch fg v
qmin (5.6)
l v
2
where C ≈ 0.09 (Berenson, 1961) was obtained from experimental data for horizontal surfaces. The
result is accurate within approximately ±50% at moderate pressures for most fluids.
1/ 4
kv3h fg v l v g 0.4c p T
h 0.62 1 (5.7)
D
v T h fg
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 super-
heating level, the effective heat transfer coefficient, including both convection and radiation, can be
expressed as
h 4 / 3 hconv
4 /3
hrad h1/ 3 (5.10)
For the case of hrad < hconv, the heat transfer coefficients can be approximated by
T 4 Tsat
4
hrad s w (5.12)
Tw Tsat
where εs 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.
The boiling Reynolds number is expressed in terms of the heat flux, q″, and latent heat of vapor-
ization, hfg, as follows:
qD
Re D (5.14)
f h fg
The coefficients A and F are functions of the critical pressure ratio, pr = p/pc, as follows:
where pc is the critical pressure of the fluid in bars (10–5 N/m2). The correlation was developed for
a wide range of common fluids (water, organics, and refrigerants) but excludes liquid metals and
cryogenic fluids.
Typical trends of the heat transfer coefficient for boiling of R–113 at a pressure of 1 atm at vari-
ous 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.
FIGURE 5.4 Heat transfer coefficient for boiling with forced convection at q″ = 25 kW/m2. (adapted from
Cornwell and Houston, 1994)
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 temperatures, 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 (5.18)
hc
It can be observed that the overall heat transfer coefficient approaches the single-phase convec-
tion limit as the boiling coefficient diminishes to zero.
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 concentration 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 coalescence. 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 pressure 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 gradient
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 pulsating 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.
5.4.2 Formation of Bubbles
Within the bubble flow regime, there are several further classifications that characterize how the
moving and deformable bubbles are dispersed or suspended in the bulk liquid flow. In ideally sepa-
rated bubble flow, the bubbles do not interact with each other and behave like single bubbles. As
the bubble number density increases, the bubbles interact with each other via collisions or wakes
216 Advanced Heat Transfer
from other bubbles (called interacting bubble flow). With a further increase in bubble numbers, the
bubbles begin to coalesce and form cap bubbles which are highly agitated by turbulence and create
a churn turbulent bubble flow. Occasionally a clustering of bubbles occurs and the cluster behaves
like a single gas slug (clustered bubble flow). Typically the void fraction varies from 0.01 to 0.06 as
the transition occurs from interacting bubble flow to churn turbulent bubble flow. Two significant
processes in these flow regimes involve the growth and coalescence of bubbles.
Consider the growth of a vapor bubble of radius rb(t) in the radial direction, r, in a superheated
liquid (see Figure 5.6). Two constraints exist on the rate of growth of a bubble – first the inertia of
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
4 rb2 constant 4 r 2u (5.19)
dt
The kinetic energy of the moving liquid, supplied by the expanding bubble, can be written as
2
u2 dr
rb
4 r 2 l
2
dr 2 rb3 l b
dt
(5.20)
The corresponding vapor pressure of the bubble as it expands against the external pressure is
given by
4
pdV p p 3 r 3
g b (5.21)
V
rb = At (5.23)
The constant A can be determined by replacing the pressures with temperature using the
Clausius–Clapeyron equation,
2 T T
A2 h fg g sat (5.24)
3 l 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 solu-
tion (Plesset and Zwick, 1954) is obtained for the growth of the bubble radius with time,
rb = Bt 1/ 2 (5.25)
c p,l Tw Tsat
1/ 2
12 Ja 2 kl
B ; Ja (5.26)
l c p ,l h fg
Alternatively, the following model can be used for either the inertia-controlled or thermal
diffusion-controlled regions,
2
1
3/2 3/2
rb t 1 t (5.27)
3
where
Arb A2
rb ; t t (5.28)
B2 B2
l ug1 ug 2 req
We 0.01 (5.29)
where
1 1 1
(5.30)
req 2rb1 2rb 2
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
218 Advanced Heat Transfer
covered by a thin liquid-rich layer through which thin stems of vapor are connected to a cloud of
coalescing bubbles. Correlations at the onset of fully developed boiling were presented by Saha
and Zuber (1974). 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 controlled through the wall heat flux, or hydrody-
namically controlled, through the mass flow rate of the bulk liquid flow.
dp P P
lul 1 g lul2 1 g 1 g g l 1 g sin w i i (5.31)
t z dz A A
P
t
g gug z g gug2 g dp
dz
g g g sin i i
A
(5.32)
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, interfacial shear stress (rate of
momentum transfer from the gas to liquid phase per unit interfacial area), tube perimeter, interfacial
perimeter, and the fraction of the tube volume that is occupied by the gas phase.
For steady-state flow in a vertical tube, the gas phase equation reduces to:
P
z
g gug2 g
dp
dz
g g g i i
A
(5.33)
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
2
c (5.34)
ro2
m le m g
c (5.35)
m le /l m g / g
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 occu-
pied by droplets, ζd, where
m / m le /l
c g d g g g (5.36)
m g / g
Then the momentum balance in the gas phase, Equation (5.33) can be rewritten as
P
z
c cuc2 c
dp
dz
g c c i i
A
(5.37)
m
uc
c c
g e 1 g (5.38)
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 estimated 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:
dp
2 2
2 rz 2 r i z l gz ri2 r 2 p p
z r ri
dz
(5.39)
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 ri2 r 2
i l g (5.40)
r 2 dz r 2
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 r 1 dp
u
l
iri l g ri2 ln o
2 dz ri 4 l l g ro2 ri2
dz
(5.41)
220 Advanced Heat Transfer
Further integration of this velocity profile across the film yields the following film flow rate:
2
l 1 dp 2 1 2 2 1 2 ro
l dp 2 2 2
m f
l
iri 2
l g dz ri 4 ro ri 2 ri lnn r 8
i
l g dz ro ri (5.42)
l
where the subscripts f, i, and o refer to the liquid film, interface (or inner), and outer, respectively.
For turbulent flow, the effective viscosity can be calculated from the sum of the laminar and tur-
bulent viscosities (Chapter 3). Further details on turbulent viscosity modeling 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 determined 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 rela-
tionship 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 generated on the film surface
which are carried upward by the gas stream. This transition point is called “flooding”. Eventually
as the gas velocity increases, the liquid film reverses direction 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 boiling
coefficients, hc and hb, respectively. Define the mass velocity, G, and density-weighted phase quality
(or Martinelli parameter), X, as follows:
m
G (5.43)
D2 / 4
0.9 0.5 0.1
1 g v l
X (5.44)
g l v
where χg is the quality (vapor fraction). Also, the dynamic transition factors, F and S, are defined as
follows:
0.1 if ReT 70
S 1/ 1 0.42 ReT0.78 if 32.5 ReT 70 (5.46)
1/ 1 0.12 Re1T.14 if ReT 32.5
where
GD 1 g
ReT 10 4 F 1.25 (5.47)
l
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.
GD 1 g
0.8
hD
Nu D,c c 0.023F Pr 0.4 (5.48)
kl l
Also, the boiling coefficient, hb, can be estimated based on the form of previously developed
boiling correlations in this chapter:
hb D c 0.45 0.49
Nu D,b 0.00122 D S 0.5 0.21p,l 0.29l 0.24 0.24 T 0.24 psat
0.75
(5.49)
kl kl l h fg v
Using these two correlations, the overall heat transfer coefficient, htot, for the combined
convection-boiling flow is approximated by
htot F hc S hb (5.50)
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).
layer. Then slug flow appears when the waves contact the top of the tube to form a liquid slug that
moves rapidly through the tube. Lastly, annular flow conditions arise when the vapor fraction further
increases. This regime is similar to vertical flows although there is an asymmetry in the film thick-
ness due to gravitational effects.
The prediction and control of transition patterns in horizontal two-phase flow have significant
importance in many industry applications such as oil transport in offshore pipelines. A transition
occurs from dispersed to stratified flow as the vapor fraction increases and bubbles are grouped to
form larger gas bubble regions that can no longer be suspended in the liquid phase. These processes
of bubble entrainment are complex and involve several factors including the geometrical configura-
tion 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 transi-
tion of slug to annular flow, classical 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 chan-
nel 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 5.8. The graphs are normally plotted in terms of the velocity, mass flux, and/or phase frac-
tion. In the two-phase flow map for horizontal tubes in Figure 5.8, the map is plotted in terms of the
pressure gradient parameter (Xp) and superficial velocity parameter (F) as follows:
FIGURE 5.8 Two-phase flow maps for (a) vertical and (b) horizontal tubes.
Gas–Liquid Two-Phase Flows 223
dp /dx l
X p2 (5.51)
dp /dx g
g ug
F (5.52)
l g Dg cos
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 accel-
eration 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 (x) direction, the indi-
vidual components can be approximated by
dp d 1
m 2 (5.54)
dx a dx m
dp
m g sin (5.55)
dx h
dp C m 2 g
1 ftp g 1 (5.56)
dx f D 2 l
Here ρm, θ, χg and D refer to the mixture density of the two-phase flow, angle of inclination 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 fac-
tors are functions of the Reynolds number, determined 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 homoge-
neous 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 dispersed 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.
distributed across the cross-section of a tube and negligible relative motion occurs between the
phases. Then define the total superficial velocity as
U Ul U g (5.57)
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
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 fraction of the
plane occupied by each phase,
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 maxi-
mum 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 u g 1 g
n
(5.62)
2 lurb g l4
Re b ; Gal (5.63)
l l 3
where rb and σ refer to the bubble radius and surface tension, respectively.
Gas–Liquid Two-Phase Flows 225
TABLE 5.1
Expressions and Coefficients for the Bubble Rise Velocity
Range Bubble Rise Velocity, u∞ Coefficient, n
Reb < 2 0.22r b
2
l g g / l 2.0
1.75
2 Re b 4Gal2.2 0.33g 0.76 l /l
0.52
rb1.28
In the bubble flow regime, Rohsenow and Griffith (1955) estimated the total wall heat flux as
follows:
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 s Equation (5.2), can be used in this
model. This approximation can be applied to forced 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
1 g
0.4
Y 1 0.1 l 1 (5.66)
g
mixuD
Re mix g g 1 g (5.67)
g l
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:
gDb2 l g
Eo (5.68)
g 4 l g
Mo (5.69)
l2 3
226 Advanced Heat Transfer
where hm and Dlg refer to the mass transfer coefficient and diffusion coefficient of mass transfer from
the liquid to gas phase, respectively. This correlation predicts the convective mass transfer from the
liquid to the gas phase of moving bubbles in a tube.
dp
Al 0l Pl i Pi l Al g sin 0 (5.71)
dx
dp
Ag 0 g Pg i Pi g Ag g sin 0 (5.72)
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 P 1 1
0g 0l l i Pi l g g sin 0 (5.73)
Ag Al Al Ag
1 1
0l fl lul2 ; 0g fg gug2 (5.74)
2 2
1
fi g ug ul
2
i (5.75)
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
Du Du
fl Cl l l l ; fg Cg g g g (5.76)
l g
Gas–Liquid Two-Phase Flows 227
4 Al 4 Ag
Dl ; Dg (5.77)
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.
Define the following nondimensional variables,
Dl Dg Al Ag
Dl ; Dg ; Al ; Ag (5.78)
D D D2 D2
Pl Pg Pi
Pl ; Pg ; Pi (5.79)
D D D
ul ug
ul ; ug (5.80)
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 (5.73), the momentum balance
becomes:
Pl P P P
n m
X 2 ul Dl ul2 ug Dg ug2 l g i 4Y 0 (5.81)
Al Al Ag Ag
where X is the Martinelli parameter. The pressure gradient parameter, Y, represents the effect of tube
inclination as follows:
Y
g l g sin (5.82)
dp f /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, hl hl /D , as follows (refer to
Figure 5.9):
1 2
Al
Al
cos1 2hl 1 2hl 1
D 2 4
1 2hl 1 Ag
4
(5.83)
Pi
Pi
2
Pi 1 2hl 1 ; Pl cos1 2hl 1 Pg (5.84)
D D
Also, the dimensionless equivalent diameter for the liquid phase and superficial velocities can be
expressed as
4 Al 4 Al 4 Ag 4 A
Dl ; Dg g (5.85)
Pl D Pl D Pg Pl Pg Pi
ug A Ag ul A Ag
ug l ; ul l (5.86)
Ug Ag Ul Al
When these relationships are substituted into Equation (5.81), for given values of X and Y cor-
responding to the known flow conditions, the nondimensional liquid height, hl∗, can be determined.
Then the corresponding gas fraction can also be determined based on the above geometrical relation-
ships between the liquid area, gas area, and total cross-sectional area of both liquid and gas phases.
ug C0U us (5.87)
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
g (5.88)
ug
The rise velocity can be expressed in terms of the Viscosity number, Vi, and Eötvos number, Eö,
as follows:
D 3 g l g l
Vi (5.90)
l
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).
Gas–Liquid Two-Phase Flows 229
For low fluid viscosities and large tube diameters, the expression for the rise velocity can be
simplified as follows:
us = 0.345 gD (5.91)
This expression for the slug rise velocity can also be obtained from a potential flow solution
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 configurations which
may occur, universal relations for all annular flow conditions are not available, 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 details on the modeling of laminar and turbulent annular
flows are described by Hetsroni (1982).
hD
b
Nu D c ReD2 F (5.92)
kf
in the range of 109 < Re2 F < 0.7 × 1012 where the Reynolds number, Re = GD/μl, is based on the
tube diameter, D, and mass velocity, G, in Equation (5.43). 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
h fg g
F 102 (5.93)
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
g 1 (5.94)
5 b / 1 8 / 7
where
0.5
m g / g
b 4a Re 0L.125 g ; (5.95)
l m g / g m lf m le /l
230 Advanced Heat Transfer
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 4 21.9 Frl (5.96)
m l2
Frl (5.97)
l2 gD
The mass flow rate of liquid, ṁl, is the sum of the flow rates of liquid in the wall film and entrain-
ment 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 flow coefficient, hl, is given by
htp
C1C0C2 25Frl C3 B0C4 Ffl
c5
(5.98)
hl
where the convection number, C0, and boiling number, B0, are defined as follows:
0.8 0.5
1 g v
C0 (5.99)
g l
qw
B0 = (5.100)
AGh fg
The single-phase convection coefficient, hl, can be determined based on the Dittus–Boelter cor-
relation (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 different working
fluids including refrigerants and water.
TABLE 5.2
Coefficients for the Kandlikar Correlation
C0 < 0.65 C0 > 0.65
(Convective Region) (Boiling Region)
C1 = 1.1360 0.6683
C2 = –0.9 –0.2
C3 = 667.2 1058.0
C4 = 0.7 0.7
C5 = 0.3 0.3
Gas–Liquid Two-Phase Flows 231
2 r lf 2 ro w ro2 r 2 dp
dz
g sin
l
(5.101)
where τfl and τw refer to the liquid film and wall shear stresses, respectively. In the axial z-direction,
the momentum equation for the total flow can be represented as
gG 2 d g dp 2 w
m g sin (5.102)
g dz dz ro
where χg, G, ζg, ρm, and υg refer to the vapor mass fraction, mass velocity, void fraction (volume frac-
tion of gas), mixture density, and gas specific volume, respectively. Expanding dυg/dz by the chain
rule of calculus as a product of dυg/dp and dp/dz yields:
dp 2 w gG 2 d g
1 dz
ro
m g sin ;
g dp
(5.103)
FIGURE 5.10 Annular flow in an inclined tube – (a) phase distribution; and (b) flow configuration.
232 Advanced Heat Transfer
Substituting Equation (5.103) into Equation (5.101) eliminates the pressure gradient so that:
1 r ro ro2 r 2 m
lf r r r w 2r l 1 g sin (5.104)
1 o
Changing from the radial coordinate, r, to the distance from the tube wall, y, yields:
1 y ro y 2ro y m
lf y 1 w l g sin (5.105)
1 ro ro y 2 ro y 1
1 y 1 m
gc y 1 r w 2 ro y 1 c g sin (5.106)
1 o
where ρc is the droplet-laden gas core density. Drag forces are exerted by the gas carrier phase on the
entrained liquid droplets. The droplets are sheared off the top of the disturbance waves and eventu-
ally redeposited back onto the liquid film. A net exchange of momentum occurs between the gas
core and liquid film during these processes.
Using the turbulent eddy viscosity method, the total shear stresses in the liquid film, τlf, and gas
core, τgc, can be expressed as
dVlf
lf y l l ,t ; 0 y ylf (5.107)
dy
dVgc
gc y gc gc,t ; ylf y ro (5.108)
dy
where μl, μgc, Vlf, Vgc, and ylf refer to the liquid and droplet–laden gas core viscosities; liquid film
and gas core velocities; and average liquid film thickness, respectively. Also, the subscript t refers
to turbulent.
Substituting Equations (5.107) and (5.108) for the liquid film and gas core into Equations (5.105)
and (5.106),
dVlf 1 y ro
lf 1
clf y ; 0 y ylf (5.109)
dy 1 ro ro y
dVgc y
gc
1
dy 1
1
r
cgc ro y ;
ygc y ro (5.110)
o
where
l , t Vlf y ro ylf
lf 1 ; Vlf ; y ; ro ; ylf (5.111)
l V l y l y l y l
The nondimensional velocities, shear length scales, and parameters clf and cgc are defined as
w w l gc
Vlf ; Vgc ; y l ; y g (5.113)
l gc lVlf gcVgc
y2l m y2g m
clf l g sin ; cgc gc g sin (5.114)
lV l 1 2 gcV g 1
The gas core viscosity, μgc, and mixture density, ρm, are determined based on mass fraction
weighted sums of their respective pure liquid and gas property values as follows:
gc 1 gc l gc g (5.115)
m 1 gc l gc g (5.116)
Equations (5.109) and (5.110) are solved for the liquid film and gas core velocities subject to
boundary conditions of no-slip and continuity of the velocity profile at the liquid–gas interface,
Vlf 0 0 (5.117)
gc
Vgc ylf
l
Vlf ylf (5.118)
In order to solve (5.109) and (5.110), two algebraic turbulence models are required for the tur-
bulent viscosities as functions of the turbulent length scales. The turbulence structure in the liquid
film is affected by the interaction of the liquid film with the gas flow. As a result, the turbulence and
liquid film eddy viscosity are characterized by the liquid film thickness. Through comparisons with
experimental data (Cioncolini et al., 2009),
2
lf f ylf 1 9 10 4 ylf (5.119)
The turbulent structure of the gas core is mostly affected by the presence of the wall, as typically
experienced in wall-bounded flows. The characteristic length is assumed to be the distance from the
wall. Through curve fits with experimental data,
gc
f y
1
y
agc
(5.120)
where agc is an empirical parameter. Substituting these turbulent viscosities, Equations (5.119) and
(5.120), into the momentum equations, Equations (5.109) and (5.110), yields the following velocity
profiles in the liquid film and gas core,
2
y y
1 y 1
2
V
ro ln 1
clf y ; 0 y ylf (5.121)
lf 1
lf
ro ro 2 lf
234 Advanced Heat Transfer
gc 1 y y ygc
Vgc
l
Vlf ylf agc
1
cgcro ln
ygc
ro
;
ygc y ro (5.122)
The algebraic turbulence models in Equations (5.119) and (5.120) are calibrated by integrating
the above velocity profiles and comparing the resulting mass flow rates with experimental data of
the liquid film and gas core flow rates. Through these comparisons, the empirical parameter of the
gas core turbulent viscosity can be obtained (agc = 4.2).
Sample results of the turbulent eddy viscosity in the liquid film and dimensionless gas core veloc-
ity are illustrated in Figure 5.11 (Cioncolini et al., 2009). Close agreement with experimental data
of CISE researchers (Adorni et al., 1963; Casagrande et al., 1963) is observed. The eddy viscosity
increases approximately linearly with the dimensionless liquid film thickness in Figure 5.11(a). In
the limit of thin films, the eddy viscosity approaches unity, indicating that turbulence is damped as
the liquid film becomes thinner.
In Figure 5.11 (b), the core flow velocity overpredicts the experimental data close to the liquid
film interface and under–predicts close to the center of the channel. In the simplifying assumptions
of the turbulence model, a homogeneous core flow is assumed and entrained liquid droplets are
distributed uniformly throughout the core flow. However, experimental data indicate that higher
droplet concentrations occur near the liquid film, which reduces the local velocity. Improvements
to the turbulence model can be achieved by including spatial variations of the droplet distribution
across the tube cross-sectional area.
FIGURE 5.11 Dimensionless (a) eddy viscosity in the liquid film and (b) gas core velocity. (adapted from
Cioncolini et al., 2009)
Gas–Liquid Two-Phase Flows 235
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 values of the condensation heat
transfer coefficient, h, for steam range between 10,000 and 30,000 W/m2K for horizontal tubes and
between 5,000 and 15,000 W/m2K 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 conden-
sation is not usually sustained over long periods of time.
5.7.1 Axisymmetric Bodies
Consider a quiescent vapor at Tsat in contact with a vertical plate of length L at a uniform temperature
of Tw (see Figure 5.12). A scale analysis (Chapter 3) can be used to determine the significant and
relevant terms of the governing conservation equations. As an example, consider a window inside a
humid room exposed to cold outdoor conditions. Typical measurements might include a thin 0.2 mm
film thickness (δl) over a window length of L = 5 m with ΔT = Tw – Tsat = 15°C. By also measuring
the runoff flow rate of condensate from the plate, an average film velocity can be determined, for
example, 0.08 m/s. Then the characteristic time and velocity scales could be estimated from these
measurements, so t ≈ L/ul = 63 s and vl ≈ δ/t = 3 × 10–6 m/s.
Performing a scale analysis by using these characteristic time, length, and velocity scales in the
governing momentum and energy equations, it can be shown that viscous drag and the body force
(buoyancy) are the two dominant terms in the y-direction momentum equation. Also, conduction
across the film layer in the y-direction is much larger than other terms in the energy equation. The
reduced form of the mass, momentum, and energy equations becomes:
ul vl
0 (5.123)
x y
2ul
l g x l v (5.124)
y 2
2T
0 (5.125)
y 2
FIGURE 5.12 Film condensation on a vertical plate – velocity and temperature profiles.
236 Advanced Heat Transfer
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 condi-
tions at the interface (y = δ) are T = Tsat, a zero velocity gradient in the liquid film, and the following
interfacial heat balance:
T
kl m v h fg (5.126)
y
y
v 1 2
l y l gx y y (5.127)
l 2
T T
T y Tw sat w y (5.128)
Define the liquid mass flow rate along the surface, per width W, as Г = ṁl /W. From a mass bal-
ance 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,
d kl Tsat Tw
m v (5.129)
dx h fg
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
(5.127), across the film from y = 0 → δ, as follows:
l l v g x 3
0
lul dy
3 l
(5.130)
3h fg
3 l
0
x dx
g
(5.131)
The film thickness can be determined from Equation (5.130) based on the result in Equation
(5.131). A correction to Equation (5.131) 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:
Gas–Liquid Two-Phase Flows 237
3/ 4
hS 4 g l l v hfg S
3
1/ 4
1 s g 1/ 3
Nu s
dx
x
(5.133)
kl 3 4kl Tsat Tw l S g
0
The effects of the Prandtl number on the heat transfer from a laminar film of condensate are
illustrated in Figure 5.13. Additional restraining effects of vapor drag on the downward 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 5.13).
Good agreement is achieved between this model and experimental data for problems involving
laminar film condensation. Discrepancies arise due to factors such as wall roughness, 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.
5.7.2 Other Configurations
Film condensation outside horizontal tubes arises in many engineering systems. For example, shell–
and–tube condensers with banks of horizontal tubes are used in thermal power plants and processing
industries. Using the body gravity function method and integrating the film condensation result in
Equation (5.133) around the circumference of a cylinder, the average heat transfer coefficient becomes:
1/ 4
g l l v hfg kl3
h 0.725 (5.135)
l D Tsat Tw
FIGURE 5.13 Heat transfer of laminar film condensation on a vertical plate. (adapted from Chen, 1961)
238 Advanced Heat Transfer
This correlation is also illustrated in Figure 5.13 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 (5.135). 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):
hN N 1/6h (5.136)
For surfaces such as plates or cylinders inclined at an angle of θ with respect to the horizontal
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 surface. This body grav-
ity 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),
1/ 4
g l l v hfg D 3
h 0.815 (5.137)
l 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.
4m
Re (5.138)
l P
where the liquid viscosity is evaluated at the film temperature. Also, P is the wetted perimeter. For a
vertical plate, P is the plate width, whereas P = πD for a vertical tube, and P is twice the tube length
for a horizontal tube. During condensation, the following three regions can be identified along the
surface: (i) laminar (wave–free) for Reδ < 30; (ii) wavy–laminar (transition) for 30 ≤ Reδ ≤ 1800;
and (iii) turbulent (Reδ > 1800). Each region will be examined separately in the following set of
correlations.
In the laminar region, the previous film condensation results can be written in the following form:
1/ 3
hx vl2 /g 3
1/ 3
This result provides a modified Nusselt number with a characteristic length based on the fluid
viscosity (rather than the plate length or width).
Gas–Liquid Two-Phase Flows 239
Alternatively, using Equation (5.134) in the laminar region for the average heat transfer coefficient,
1/ 3
hL L l2
1.468 Re L1/ 3 (5.140)
kl g
1/ 3
hL vl2 /g Re
NuL (5.141)
kl 1.08 Re1.22 52
1/ 3
hL vl2 /g Re
NuL (5.142)
kl
8, 750 58 Pr 0.5 Re0.75 253
Chen, Gerner and Tien (1987) presented the following correlation of the average heat transfer
coefficient through both the wavy and turbulent regions.
1/ 3
hL l2
1/ 2
Re L0.44 5.82 10 6 Re 0L.8 Prl1/ 3 (5.143)
kl g
This correlation as well as the laminar flow correlation in Equation (5.134) are illustrated in
Figure 5.14. The heat transfer coefficient reaches a minimum and then increases with both Prandtl
and Reynolds numbers in the turbulent region.
FIGURE 5.14 Average heat transfer coefficient for laminar, wavy and turbulent film condensation. (adapted
from Bejan, 2013)
240 Advanced Heat Transfer
In general, these correlations agree within ±10% of experimental data when the vapor is motion-
less 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).
5.8.2 Outside a Sphere
Consider downward flow of a pure vapor at the saturation temperature, Tsat, moving at a uniform
velocity, U, past a sphere at a uniform wall temperature of Tw (see Figure 5.15), where Tw < Tsat. A
condensation film occurs and flows down along the surface under the effects of gravity, wall resis-
tance and shear forces by the external flow on the vapor. The condensation film becomes partially
turbulent at high vapor flow velocities. An integral solution of the turbulent film condensation was
presented by Hu and Chen (2005).
The reduced form of the steady-state momentum and energy equations within the surface film
are given by
u v 1 dp u
u
x
v
y
dx y
t y gx (5.144)
T T T
u
x
v
y y t
y
(5.145)
where νt and αt represent the turbulent eddy viscosity and eddy diffusivity, respectively.
Integrating the energy equation across the condensate layer leads to:
l d kl dT
sin d sin udy r h
0
o
fg dy y 0
(5.146)
w g l v sin 0 (5.147)
This result indicates a balance among the wall shear stress, gravitational force on the condensate
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:
f 2C Re v 0.2 (5.148)
vuv2 / 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 (5.149)
2
9
w C vU 2 Re v 0.2 sin 2 g l v sin 0 (5.150)
4
yu u
y ; (5.151)
l l
w u T Tw
u ; u ; T (5.152)
u Tsat Tw
Re uro
Re ; Re (5.153)
Gr1/ 3 l
U2 gro3 l v c p Tsat Tw
Fr ; Gr ; Ja (5.154)
gro l2 l h fg
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 2
Re3 0.9 2
Re Fr sin sin (5.155)
8
Rearranging this result yields the following expression for the local dimensionless thickness of
the condensate layer:
In terms of nondimensional variables, the energy equation in the condensate layer becomes:
d t dT
1 Pr 0 (5.158)
dy l dy
t
l
0. y 1 exp 0.0017 y 2
(5.159)
dT
1 0.4 y 1 exp 0.0017 y 2 Pr
(5.160)
dy
0
dy / 1 0.4 y 1 exp 0.0017 y 2 Pr
Then the Nusselt number is determined based on this dimensionless temperature gradient. The
heat balance at the surface is given by
T
kl h Tw Tsat (5.161)
y
y 0
dT
Nu Re (5.162)
dy y 0
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 models 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 velocities, the influence of the shear parameter on
the heat transfer coefficient is less significant.
FIGURE 5.16 Flow regimes for condensation with forced convection in a horizontal tube.
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 5.16. 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 convection 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 0.45 f
St (5.163)
RePr 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
0.046 g2.5 X 2
f (5.164)
m v D /v
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 volume fraction can be
expressed in terms of the mass fraction of vapor, χg, as follows:
1
g (5.165)
1 g / l 1 g / g
1/ 2
Then the total wall heat flux, including the effects of condensation and forced convection, is
given by
2r
qw qi m ch fg i (5.166)
D
244 Advanced Heat Transfer
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 (5.163), 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 expres-
sion for the wall heat flux excluding the interfacial heat flux, qi.
Jaster and Kosky (1976) reported various criteria for transition between annular and stratified
flows based on the following transition factor, F:
flul2
F (5.167)
2 g
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 condensate velocity
can be approximated by
Vl
ul (5.168)
1 g
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
5.8). Mandhane (1974) and Travis and Rohsenow (1973) presented flow maps for forced convec-
tion 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 veloc-
ity, Vsg, for classification of the various regions,
Vsl
1 g G ; Vsg
gG
(5.169)
l v
1/ 4
g l v kl3hfg
h 0.555 (5.171)
vl D Tsat Tw
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 5.16, 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,
l
h Pr1/ 2 0.046c p,lGm f (5.173)
where
Gm G 2
1
G1G2 G22 / 3
(5.174)
f 0.046 Re v 1/ 5 (5.175)
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 = χ1 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 luv D
Nu D C Re1D/ 2 ; ReD (5.177)
kl l
246 Advanced Heat Transfer
where D is the tube diameter. Here C = 0.9 for ReD < 106, whereas C = 0.59 is recommended when
ReD > 106.
For vertical vapor downflow over a tube with a condensate surface film, the Nusselt number can
be approximated by
0.9 0.728F 1/ 2
NuD Re1D/ 2 (5.178)
1/ 4
1 3.44 F 1/ 2
F
where
Grl l h fg
F (5.179)
Re D kl T Tsat
2
l l v D 3
Grl (5.180)
l2
The subscripts l and v refer to liquid condensate and vapor, respectively. The results agree reason-
ably 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 turbulence 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
h
NuD 0.728F 1/ 4 Re1D/ 2 0.728 l fg Grl (5.181)
k
l T
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
Gas–Liquid Two-Phase Flows 247
layer separation. The shear stress exerted by the vapor on the surface film affects the rate of conden-
sation 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 sat-
urated vapor past a number of tube bundles. The vapor velocity, uv, in Equation (5.177), 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 (5.136).
5.9.3 Finned Tubes
Another common configuration is condensation on horizontally finned tubes. Correlations for the
average heat transfer coefficient were presented by Beatty and Katz (1948),
1/ 4 Ab Af
h 0.689 g l hfg kl3 l l v Tw Tsat AD1/ 4 1.3 AL1/ 4
(5.183)
i
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 A f (5.184)
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,
Do2 Di2
L (5.185)
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 (5.183) 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 (5.183) 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 comparison
to an unfinned tube. The last term in brackets in Equation (5.183) 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 con-
denser equipment, including fins of a sawtooth shape, as well as wires loosely attached to the tubes.
condenses and releases latent heat along the upper walls, and falls under gravity as a condensate film
back to the lower base section where the cycle continues. The device uses gravity to return liquid so
the evaporator must be located below the condenser. Closed, gravity-assisted, two-phase thermosy-
phons (GATPTs) are used in 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 evaporator. 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 condensation, and liquid return flow by capillary
action in the wick back to the evaporator (see Figure 5.19). 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 operating 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.
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 separator 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, microelec-
tronics 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 telecommunications 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 control 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 ther-
mal 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, work-
ing 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 circu-
lation 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 (capil-
lary forces) within the wick must exceed the sum of frictional 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 process,
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 from a force balance on
the fluid, yielding:
2 cos
p (5.186)
rp
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
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 1023 K 181 at 1023 K
773–1,173 Sodium Nickel, stainless steel 9.3 at 1123 K 224 at 1033 K
250 Advanced Heat Transfer
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:
l K w Aw pL
m (5.187)
l 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,
lVl2 Leff 64
pL (5.188)
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.
TABLE 5.4
Typical Wick Parameters of Heat Pipes
Material Pore Radius (cm) Permeability (m2)
Copper foam 0.021 1.9 × 10–9
Copper powder (45 μm) 0.0009 1.74 × 10–12
Copper powder (100 μm) 0.0021 1.74 × 10–12
Felt metal 0.004 1.55 × 10–10
Nickel felt 0.017 6.0 × 10–10
Nickel fiber 0.001 0.015 × 10–11
Nickel powder (200 μm) 0.038 0.027 × 10–10
Nickel powder (500 μm) 0.004 0.081 × 10–11
Nickel 50 0.0005 6.635 × 10–10
Nickel 100 0.0131 1.523 × 10–10
Nickel 200 0.004 0.62 × 10–10
Phosphorus/bronze 0.0021 0.296 × 10–10
Gas–Liquid Two-Phase Flows 251
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:
2 cos L m
l eff l gLeff sin (5.190)
rp l K w Aw
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.
h 2 l gLeff sin
max hfg l fg Aw Kw
qmax m (5.191)
l Leff rp
The maximum mass flow rate can be determined based on given parameters and thermophysical
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 correspond-
ing 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.
5.10.2 Operational Limitations
In addition to the equilibrium design condition in Equation (5.190), four operational limits on the
heat flux affect the performance of a heat pipe: (i) wicking; (ii) entrainment; (iii) sonic; and (iv) boil-
ing limits. These four limitations are illustrated in Figure 5.20 as well as a key performance indicator
called the merit factor, M.
The wicking limitation on the axial heat flux is obtained at the point of maximum flow through
the wick at the maximum capillary pressure rise. In this case, the equilibrium design condition is
rearranged to give:
252 Advanced Heat Transfer
FIGURE 5.20 (a) Heat pipe limitations and (b) figure of merit.
A K 2 gL sin
qmax m max h fg M w w l eff (5.192)
Leff rp
l h fg
M (5.193)
l
Another key design parameter is the heat transfer factor, HTF, defined as follows:
2 MAw K w
HTF qmax Leff 0
(5.194)
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 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 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 param-
eter 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:
vVv2 Lh
We 1 (5.195)
where Lh refers to the hydraulic diameter of the wick surface pores. The actual vapor velocity should
remain lower than the above value in the entrainment limit. Otherwise, entrainment 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
Gas–Liquid Two-Phase Flows 253
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
fg
mh
qaxial vVvh fg (5.196)
Av
The sonic limit, as well as the other limits, are often shown graphically in terms of the fluid oper-
ating temperature (see Figure 5.20). Each limitation spans across a given temperature range and the
maximum heat flux occurs in the range of the wicking limit.
The heat flux limits generally increase with evaporator exit temperature due to the effect of tem-
perature 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 containment. As a result, this situation places an additional
limit on the design of the heat pipe.
In addition to standalone devices, heat pipes can also enhance the rates of heat transfer within
other conventional heat exchange devices. For example, embedded heat pipes can significantly
improve the efficiencies of fins and extended surfaces. Bowman et al. (2000) compared the heat pipe
performance, temperature distributions, and efficiencies of fins with and without embedded heat
pipes. The results showed the promising effectiveness of heat pipes to enhance the overall thermal
performance. 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).
REFERENCES
Adorni, N., I. Casagrande, L. Cravarolo, A. Hassid, E. Pedrocchi, and M. Silvestri. 1963. “Further Investigation
in Adiabatic Dispersed Flow: Pressure Drop and Film Thickness Measurements with Different Channel
Geometries – Analysis of the Influence of Geometrical and Physical Parameters”, CISE Report R–53,
Advisory Committee of the Computer and Information Science and Engineering, National Science
Foundation, Alexandria, VA.
Altman, M., 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.
Beatty, K.O. and D.L. Katz. 1948. “Condensation of Vapors on the Outside of Finned Tubes”, Chemical
Engineering Progress, 44: 55–70.
Bejan, A. 2013. Heat Transfer, 4th Edition, New York: John Wiley & Sons.
Berenson, P.J. 1961. “Film Boiling Heat Transfer from a Horizontal Surface”, ASME Journal of Heat Transfer,
83: 351–358.
Bowman, W.J., T.W. Moss, D. Maynes and K.A. Paulson. 2000. “Efficiency of a Constant–Area, Adiabatic Tip,
Heat Pipe Fin”, AIAA Journal of Thermophysics and Heat Transfer, 14: 112–115.
Bromley, L.A. 1950. “Heat Transfer in Stable Film Boiling”, Chemical Engineering Progress, 46: 221–227.
Carey, V.P. 2007. Liquid–Vapor Phase Change Phenomena, 2nd Edition, Boca Raton: CRC Press/Taylor &
Francis Group.
Casagrande, I., L. Cravarolo, A. Hassid, and E. Pedrocchi. 1963. “Adiabatic Dispersed Two–Phase Flow:
Further Results on the Influence of Physical Properties on Pressure Drop and Film Thickness”, CISE
Report R–73, National Science Foundation, Alexandria, VA.
Chen, M.M. 1961. “An Analytical Study of Laminar Film Condensation: Part 1 – Flat Plates”, ASME Journal
of Heat Transfer, 83: 48–54.
254 Advanced Heat Transfer
Chen, J.C. 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.
Chesters, A.K. and G. Hofman. 1982. “Bubble Coalescence in Pure Liquids”, Applied Scientific Research, 38:
353–361.
Chen, S.L., F.M. Gerner and C.L. Tien. 1987. “General Film Condensation Correlations”, Experimental Heat
Transfer, 1: 93–107.
Chi, S.W. 1976. Heat Pipe Theory and Practice, Washington, DC: Hemisphere.
Chin, Y.S., 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.
Cioncolini, A., J.R. Thome, and C. Lombardi. 2009. “Algebraic Turbulence Modeling in Adiabatic Gas–Liquid
Annular Two–Phase Flow”, International Journal of Multiphase Flow, 35: 580–596.
Collier, J.G. and J.R. Thorne. 1999. Convective Boiling and Condensation, 3rd Edition, Oxford: Oxford
University Press.
Cornwell, K., S.D. Houston. 1994. “Nucleate Pool Boiling on Horizontal Tubes: A Convection Based
Correlation”, International Journal of Heat and Mass Transfer, 37: 303–309.
Drolen, B.L. 2017. “Performance Limits of Oscillating Heat Pipes: Theory and Validation”, AIAA Journal of
Thermophysics and Heat Transfer, 31: 920–936.
Duignam, M.R., 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.
Dunn, P.D. and D.A. Reay. 1994. Heat Pipes, 4th Edition, New York: Pergamon.
Fujii, T., H. Honda, S. Nozu and S. Kawakaml. 1977. “Condensation of Superheated Vapor inside a Horizontal
Tube”, Refrigeration, 52: 553–575.
Hetsroni, G. Ed., 1982. Handbook of Multiphase Systems, New York: McGraw–Hill.
Hu, H.P. and C.K. Chen. 2005. “Turbulent Film Condensation on an Isothermal Sphere”, AIAA Journal of
Thermophysics and Heat Transfer, 19: 81–86.
Jaster, H. and P.G. Kosky. 1976. “Condensation Heat Transfer in a Mixed Flow Region”, International Journal
of Heat and Mass Transfer, 19: 95–99.
Jianu, O.A., 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.
Kandlikar, S.G. 1990. “General Correlation for Saturated Two–Phase Flow Boiling Heat Transfer Inside
Horizontal and Vertical Tubes”, ASME Journal of Heat Transfer, 112: 219–228.
Kern, D.Q. 1958. “Mathematical Development of Tube Loading in Horizontal Condensers”, AIChE Journal,
4: 157–160.
Kutateladze, S.S. 1948. “On the Transition to Film Boiling under Natural Convection”, Kotloturbostroenie, 3:
10–12.
Kutateladze, S.S. 1963. Fundamentals of Heat Transfer, New York: Academic Press.
Labunstov, D. 1957. “Heat Transfer in Film Condensation of Pure Steam on Vertical Surfaces and Horizontal
Tubes”, Teploeneroetika, 4: 72–80.
Lienhard, J.H., V.H. Dhir and D.M. Riherd. 1973. “Peak Pool Boiling Heat Flux Measurements on Horizontal
Finite Horizontal Flat Plates”, ASME Journal of Heat Transfer, 95: 477–482.
Mandhane, J.M., C.A. Gregory and K. Aziz. 1974. “Flow Pattern Map for Gas–Liquid Flow in Horizontal
Pipes”, International Journal of Multiphase Flow, 1(4): 537–554.
Miropolski, Z.L. 1963. “Heat Transfer in Film Boiling of a Steam–Water Mixture in Steam Generating Tubes”,
Teploenergetika, 10: 49–52.
Naterer, G.F., 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.
Plesset, M.S. and S.A. Zwick. 1954. “Growth of Vapor Bubbles in Superheated Liquids”, Journal of Applied
Physics, 25: 693–700.
Rohsenow, W.M. 1952. “Method for Correlating Heat Transfer Data for Surface Boiling of Liquids”,
Transactions of ASME, 74: 969–976.
Rohsenow, W.M. 1956. “Heat Transfer and Temperature Distribution in Laminar Film Condensation”,
Transactions of ASME, 78: 1645–1648.
Rohsenow, W.M. and P. Griffith. 1955. “Correlation of Maximum Heat Flux Data for Boiling of Saturated
Liquids”, AIChE–ASME Heat Transfer Symposium, Louisville, KY.
Rose, J.W. 1988. “Fundamentals of Condensation Heat Transfer: Laminar Film Condensation”, JSME
International Journal, Series II, 31: 357–375.
Gas–Liquid Two-Phase Flows 255
Saha, P., N. Zuber. 1974. “Point of Net Vapor Generation and Vapor Void Fraction in Subcooled Boiling”, 5th
International Heat Transfer Conference, 4: 175–179.
Shekriladze, I.G. and V.I. Gomelauri. 1966. “Theoretical Study of Laminar Film Condensation of Flowing
Vapor”, International Journal of Heat and Mass Transfer, 9: 581–591.
Takeyama, T. and S. Shimizu. 1974. “On the Transition of Dropwise–Film Condensation”, Proceedings of the
5th International Heat Transfer Conference, Tokyo, 3: 274.
Tong, L.S. and Y.S. Tang. 1997. Boiling Heat Transfer and Two–Phase Flow, 2nd Edition, Boca Raton: CRC
Press/Taylor & Francis Group.
Travis, D.P. and W.M. Rohsenow. 1973. “Flow Regimes in Horizontal Two–Phase Flow with Condensation”,
ASHRAE Transactions, 7: 31–34.
Vachon, R., G. Nix and G. Tanger. 1968. “Evaluation of Constants for the Rohsenow Pool–Boiling Correlation”,
ASME Journal of Heat Transfer, 90: 239–247.
Young, E.H. and D.J. Ward. 1957. “How to Design Finned Tube and Shell–and–Tube Heat Exchangers”, The
Refining Engineer, 29: 32–36.
Whalley, P.B.. 1987. Boiling, Condensation and Gas–Liquid Flow, Oxford: Clarendon Press.
Zuber, N. 1958. “On the Stability of Boiling Heat Transfer”, Transactions of ASME, 80: 711–720.
PROBLEMS
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.
2. Consider a phase change process with boiling and bubble detachment from a heated hori-
zontal 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, 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 Chapter 2. Neglect the effects of convective motion and
the region of influence encompassing the vortices and wakes of departing bubbles.
3. During a boiling process, bubbles grow outward from small indentations or cavities along
a heated horizontal wall and detach from the surface once their size becomes sufficiently
large. Beneath a bubble surrounding the cavity, a thin microlayer 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 transient 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.
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 surround-
ing the bubble (note: due to bubble expansion alone, not including advection).
b. Perform an energy balance on an expanding bubble to find the pressure difference
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. 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/m2K,
256 Advanced Heat Transfer
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/m2K. 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?
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 variation of wall tempera-
ture with time. How would thermocapillary effects influence this variation with time?
7. During phase change with boiling, a bubble grows spherically outward from a small inden-
tation 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–dimen-
sional form of these equations to find the radial component of velocity outside the
boundary layer. Assume that the phase interface movement is inversely proportional to
the square root of time, V ≈ A t–1/2 where A depends on thermophysical properties and
other factors.
8. Liquid methanol at its saturation temperature (65°C) is suddenly exposed to a heated verti-
cal 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 forma-
tion. Can this data provide sufficient information to determine the relative significance of
viscous forces acting on the liquid, associated with bubble expansion? The following prop-
erties may be used for methanol: ρl = 751 kg/m3, ρv = 1.2 kg/m and μl = 3.3 × 10–4 kg/ms.
9. Surface roughness can have a substantial effect on various forces, including surface ten-
sion, which affect the processes of phase change at a wall. How could the surface rough-
ness 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.
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 bubble. This region
may be transformed to a planar channel region by the following Mobius transformation:
z z1 z2 z3 w w1
z z z z w w
3 2 1 2 1
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 cor-
responding 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.
11. In the previous problem, conformal mapping was used to transform the temperature solu-
tion 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.
Gas–Liquid Two-Phase Flows 257
Express your answer in terms of Tw, Tsat, δ0 (initial microlayer thickness) and thermophysi-
cal 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 temperature 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.
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 suddenly 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 position 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.
13. A balance of pressure and surface tension forces on a droplet can be used to find the pres-
sure difference across the phase interface. An elongated droplet of saturated liquid with
radii of curvature of r1 and r2 in the xz and yz planes, respectively is in thermal and mechan-
ical 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
r1 r2
where σ, pv, and psat refer to the surface tension, pressure of subcooled vapor outside 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.
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.
15. Repeat the previous problem while using an integral method to account for time–dependent
conduction in the vapor region. Derive the governing differential equation 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.
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 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 (position), t (time), and thermophysical
properties.
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 pressure. 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
258 Advanced Heat Transfer
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 prop-
erties of copper at standard temperature and pressure (STP) conditions.
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?
19. Pool boiling occurs along the outer surface of a copper pipe (2 cm outer diameter) sub-
merged 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 coefficient when ΔT = 12°C.
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 maxi-
mum 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?
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 different regimes of heat
transfer and dominant modes of heat transfer in each regime.
b. How would different initial water temperatures and different surface roughness affect
the results in part (a)?
c. How would agitation (fluid circulation) within the bath affect the maximum cooling
rate? Explain your response.
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 coef-
ficient 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.
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 pres-
sure. 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?
24. Forced convection with boiling occurs in an 8 cm heated vertical pipe. The water tempera-
ture 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.
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 horizontal plane. A baffle is a plate
that connects the base of a duct with the top corner 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 baffles. 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, respectively. Explain the
functional dependence on N. In other words, explain why the ratio increases (or
decreases) with N.
Gas–Liquid Two-Phase Flows 259
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 tempera-
ture increases abruptly to Tn downstream of x = 0 and a thermal boundary layer, δT, devel-
ops and grows in thickness within the film.
a. Start with the relevant film momentum equation and derive the following result for the
2
u y y
velocity distribution in the film: 2
U
where U refers to the velocity at the film–air interface. Assume that the terminal 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 approxi-
mated linearly, 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.
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 = r2/(4αt) to solve the following heat
conduction equation for the temperature in the condensate layer (r > ro):
2T 1 T 1 T
r 2 r r t
The boundary and initial conditions are T = Tw at r = ro and T = Ti at t = 0. This analysis can
be used to evaluate the effects of slight wall temperature perturbations on the rate of heat
transfer.
28. A vertical plate with a height of 3 m is maintained at 40°C and exposed to saturated steam
at atmospheric pressure.
a. Determine the heat transfer and condensation flow rate per unit width of the plate.
b. Would it be more effective to use a shorter and wider plate instead of the high and nar-
row plate? Explain your response in terms of total heat transfer enhancement. Assume
that an equal surface area is maintained in both cases.
29. A heat exchange apparatus contains a 2 m high vertical pipe exposed to saturated steam at
atmospheric pressure. Cooling water flows internally within the pipe to maintain a pipe
surface temperature of 50°C.
a. What pipe diameter is required to produce 0.1 kg/s of condensate on its outer surface?
What heat transfer coefficient is obtained at this diameter?
b. What effects on the heat transfer coefficient and heat transfer rate would be observed if
the operating pressure increases? Explain your response.
30. A horizontal arrangement of 10 tubes (each of length 2 m and diameter of 4 cm) is used
within a section of an industrial heat exchanger to condense saturated steam at atmospheric
pressure.
a. What tube surface temperature is required to produce a condensate mass flow rate of 700
kg/h from the 10–tube set?
b. How would the presence of a non–condensable gas in the vapor (such as air) affect the
results in part (a)?
31. The outer surface temperature of a metal sphere (diameter of 4 cm) is maintained at 90°C.
Find the average heat transfer coefficient when saturated steam at atmospheric pressure is
condensing on the outer surface of the sphere.
260 Advanced Heat Transfer
32. A cylindrical thermosyphon is oriented vertically with the heater surface at 120°C along
the bottom boundary. The copper thermosyphon of 6 cm in height is equally divided into a
lower insulated section (3 cm) and upper condensing section (3 cm). Saturated water is
boiled at atmospheric pressure above the heater. The top surface (above the condensation
section) is well insulated.
a. What is the heat flux during nucleate boiling?
b. Find the thermosyphon diameter required to produce a total condensation flow rate of
0.03 g/s at a steady state.
33. A suitable heat pipe is required to transport 20 kW over a distance of 24 cm in a high–tem-
perature application. For these conditions, find and compare the required pipe diameters
for the following two types of heat pipes: (i) potassium–nickel and (ii) sodium–stainless
steel.
34. Compare the axial heat flux in a heat pipe, using water as the working fluid at 200°C, with
the heat flux in a copper bar (10 cm long) experiencing a maximum temperature difference
of 80°C.
35. A water heat pipe is tested in a ground–based facility before it is applied to spacecraft
thermal control. The ground–based heat pipe produces a maximum liquid flow rate of 0.02
kg/h at an inclination angle of 5°. The internal wick consists of 200 μm nickel powder. The
heat pipe operates at 100°C at atmospheric pressure. Its inner diameter and length are 2 and
10 cm, respectively.
a. Is the same heat pipe capable of producing a maximum heat transfer rate of 60 kW/m2
under microgravity conditions (where g = 0 is assumed)?
b. How can the heat transfer capability of the heat pipe be enhanced to further exceed its
present capacity?
36. An ammonia heat pipe is constructed from a stainless steel tube with an effective length of
1.6 m and an outer diameter of 0.075 in. The aluminum fibrous slab wick has a permeability
of 16 × 10–10 m2. A set of performance tests was conducted whereby the burnout heat load
was determined as a function of heat pipe inclination (note: adiabatic section maintained at
22 ± 1°C). The results for the burnout heat load and heat pipe angle are listed as follows:
qmax (W) at ϕ (inclination angle); 83 W at 0°; 69 W at 0.25°; 50 W at 0.5°; 31 W at 0.75°;
and 12 W at 1.0°. Estimate the effective pore radius and cross–sectional area of the wick.
37. Compare the axial heat flux that can be achieved by a heat pipe of 1.2 cm diameter inclined
at 10° with the heat flux through a copper rod. In both cases, the length is 0.3 m and the
temperature difference for the rod (end to end) is 60°C. Also, the heat pipe operates with
water at atmospheric pressure. Its wick consists of the following characteristics: five layers
of wire screen, rp = 1 × 10–5 m, and Kw = 0.1 × 10–9 m2 (wire diameter = 0.01 mm). Estimate
the vapor velocity for the heat pipe in this example.
38. A heat pipe uses water as the working fluid and copper–nickel as the vessel material. The
35 cm long heat pipe must have an energy transport capability of at least 130 W at 200°C.
If the measured axial heat flux at this temperature is 0.67 kW/cm2, estimate the required
cross–sectional area. How can the axial heat flux at higher temperatures be obtained?
39. A water heat pipe provides a maximum heat transfer rate of 60 W in the horizontal orienta-
tion. The copper foam wick has a cross–sectional area of 2 × 10–6 m2.
a. What is the effective heat pipe length?
b. What heat transfer rate is expected at a heat pipe inclination of 10°?
40. Component temperatures in a satellite are passively controlled by heat pipes of 28 cm length
and 1.1 cm diameter using water as the working fluid. The wick material is nickel felt with
a characteristic pore size of 0.34 mm and thickness of 1 mm. What minimum number of
heat pipes is required to ensure that the component temperatures do not exceed 110°C?
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 others. These systems involve the transport
of suspended droplets or particles in a gas flow stream. Often heat transfer by convection, radia-
tion, 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, particularly 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–particle 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 accurately 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 concentration is sufficiently low. This critical concentra-
tion of transition between one-way 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 pipeline, 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 solids 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 transport, the gas velocity must be sufficiently larger
than the settling velocity of the solid particles to retain the particle transport. In horizontal pneu-
matic 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 alter-
nating regions between suspended particles and slugs, similarly to slugs arising in vapor–liquid
flows with forced convection boiling. Packed bed heat transfer 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 trajectory mod-
els, 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 trajectory 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 trajectory and two-fluid models, a lumped capacitance approximation is
often used for a thermal analysis of particles that are small enough to be assumed isothermal.
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 formation, melting of particles, radiation
exchange involving particles, and liquid–particle and slurry flows will be presented and discussed.
For a single particle, the equation of motion in the xi direction can be written as
du p,i 1 D p2
u p,i ug,i mgi
2
m Fd ,i Fg,i g cD (6.1)
dt 2 4
where i = 1, 2 and 3, and up,i and ug,i refer to the particle and gas velocities, respectively. Assuming
a spherical particle of mass m and dividing the equation of motion by m yields:
du p,i cD 18
u p,i ug,i gi (6.2)
dt 24 / Re p p D p2
where the particle Reynolds number, Rep, is based on the relative velocity difference between the gas
stream and particle. This form of equation is convenient because it contains a factor of 24/Rep in the
denominator which represents the Stokes drag coefficient.
The reciprocal of the second factor in brackets in Equation (6.2), p D p2 / 18 , 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 interpretation 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 state, whereas when the actual time far exceeds the characteristic response time, the particle
approaches the state of the surroundings. In between these limits, the particle’s state is in transition
between the initial and surrounding conditions. Response times, such as the velocity and thermal
response times, are effective ways of characterizing the various transport mechanisms associated
with particle interactions with the surrounding gas stream.
For laminar “creeping” flow over a stationary sphere, where advective inertial forces are small
compared to viscous forces, the Navier–Stokes equations can be solved to determine the drag coef-
ficient for a spherical particle at low Reynolds numbers, Rep < 1,
24
cD = (6.3)
Re p
This result is known as the Stokes law for low Reynolds number flow. This flow configuration
is equivalent to a spherical particle moving at a small relative velocity (ug − up) with respect to the
gas stream. At higher particle Reynolds numbers, the Schiller–Naumann correlation can be used at
Reynolds numbers up to Rep = 800:
cD
24
Re p
1 0.15Re0p.687 (6.4)
The drag force is proportional to the square of the relative velocity and the drag coefficient
decreases with particle Reynolds number. For 800 < Rep × 105, Clift et al. (1978) recommended:
cD
24
Re p
1 0.15Re0p.687 0.42
1 4.25 10 4 Re p1.16
(6.5)
For very small particles, when the particle diameter approaches the mean free path of the gas
molecules, slip effects increase in importance between the molecular and particle motions. The drag
force, Fd, on the particle is modified via a correction factor as follows:
264 Advanced Heat Transfer
3
Fd up Dp (6.6)
C
where C is a slip correction factor (Cunningham coefficient). It is defined as the ratio between the
drag force in a continuum flow at the same Reynolds number to the drag force on the particle in the
presence of slip effects.
For spherical particles, the Davies correlation (Davies, 1945) can be used to relate C with the
Knudsen number (Kn) as follows:
0.55
C 1 Kn 2.514 0.8 exp (6.7)
Kn
where Kn = λ/Dp and λ is the mean free path of the gas molecules. Extensions of this correlation to
account for non–spherical particles were formulated by Clift, Grace, and Weber (1978).
Other forces on particles may arise depending on the flow conditions and additional transport
phenomena. Cohesive forces between particles, such as electrostatic forces and van der Waals
forces, can have significant effects on the gas–particle interactions. Particles may exhibit an attrac-
tive positive–negative charge combination for grouping and collection. For example, in electrostatic
precipitators, fly ash from coal combustion has cohesive and charged particles. Van der Waals forces
arise from the attraction between dipoles at the atomic and molecular levels. Electrostatic devices
are commonly used to separate particles from a gas stream, for example, in sawmills, smokestacks,
oil refineries, cement plants, among others. Furnaces use precipitators to remove dust, dirt, and other
particles from the air prior to ventilation of return air through a house or building.
Also, capillary forces may affect forces on the particles. If a liquid film resides on the surface of
a particle or joined particles in a gas stream, then capillary forces may become increasingly signifi-
cant. Capillary forces are typically several orders of magnitude larger than the van der Waals forces.
Surface tension in the film between two coalesced particles creates an attractive force. Correlations
for capillary forces between particles due to a surface liquid film have been reported by Hewitt et
al. (1997).
dTd NuD k
mcd Dd2 Tg Td mh
fg (6.8)
dt Dd
where NuD is the Nusselt number of flow past a sphere and Tg and Td refer to the gas and droplet
temperatures, respectively. The last term refers to evaporative heat transfer where hfg is the latent
heat of vaporization, respectively.
To determine the Nusselt number, the following Ranz–Marshall correlation (1952) for convec-
tive heat transfer from a sphere can be used:
where Rer is the relative Reynolds number based on the relative velocity difference between the
gas and droplets. The Ranz–Marshall correlation yields good agreement with experimental data
Multiphase Flows with Droplets and Particles 265
up to approximately Rer ≈ 5 × 104. Other correlations for forced convection past the sphere were
presented in Chapter 3.
For turbulent flows, Raithby and Eckert (1968) reported that the Nusselt number of heat transfer
to spheres increases rapidly at turbulent intensities up to 1%, but thereafter varies approximately
linearly with turbulence intensity (see Figure 6.2). Clift et al. (1978) presented the following depen-
dence of the Nusselt number of turbulence intensity:
Nu I 0.57
1 4.8 10 4 r Rer (6.10)
Nu0 I rc
where Ir is the turbulence intensity based on the relative velocity, and Irc is the critical turbulent
intensity at the critical Reynolds number of transition to turbulence of the sphere boundary layer.
Flows of gases below standard atmospheric pressure are called rarefied flow, or also sometimes
low-pressure gas flow. Examples include flows confined to a pipe between a chamber or vessel to be
evacuated and a pump, or exhaust gases emitted from a rocket launched into the upper atmosphere.
For rarefied flows, the Kavanau correlation (1955) can be used for the Nusselt number,
Nu0
NuD (6.11)
1 3.42 Nu0 Ma / Rer Pr
where Nu0 and Ma are the Nusselt number for incompressible flows and Mach number, respectively.
This correlation represents a rarefaction correction to the continuum solution for heat transfer from
spherical droplets or particles. It agrees with experimental data within ±10%.
For the evaporative component of the energy balance, the rate of change of droplet mass, ṁ, due
to mass efflux across the droplet surface can be expressed in terms of Fick’s law (Chapter 1). At the
surface of an evaporating or condensing droplet in a binary mixture,
A
m d Ad m d Dd2 DA (6.12)
n
FIGURE 6.2 Dependence of Nu on the turbulence intensity for (a) Re = 3,600 and (b) Re = 14,000 (adapted
from Raithby and Eckert, 1968)
266 Advanced Heat Transfer
where Ad, Dd, DA, ρA, and n refer to the droplet surface area, droplet diameter, diffusion coefficient
of species A in the mixture, mass density of species A, and coordinate normal to the surface, respec-
tively. Alternatively, in terms of the concentration of species A in the mixture, CA,
A / m C C A, s
m d Dd2 DA m Dd2 DA m A, (6.13)
n Dd
Here ρm, CA,∞ and CA,s are the mixture density and concentration of species A in the freestream
and droplet surface, respectively. Similar to convective heat transfer where the Nusselt number is
the dimensionless temperature gradient at a surface (Chapter 3), also the Sherwood number, Sh,
is the dimensionless concentration gradient at the surface. So the above constant of proportionality
is the Sherwood number.
Assembling these terms, then the energy equation becomes:
The Sherwood number is 2 for an evaporating droplet if forced or free convection effects are
negligible. At higher relative velocities between the droplet and carrier phase, the evaporation or
condensation rate increases and the Ranz–Marshall correlation can again be used:
k
Vk
k dV
Vk
(6.16)
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 fractions of
mass, m, and volume, V, of phase k within a multiphase control volume,
mk
k (6.17)
m
Multiphase Flows with Droplets and Particles 267
Vk
k (6.18)
V
The volume fraction can be written in terms of the mass fraction after multiplying by the density.
Consider the conservation of a general scalar quantity, ϕk, for a multiphase mixture in a two-
dimensional control volume. Figure 6.3 illustrates a multiphase flow with droplets and particles in a
differential control volume. 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 conservation
equations for mass, momentum, and energy, it can be shown that a general conservation balance for
the general scalar quantity, ϕk, in tensor form, can be written as
t
kk
x j
kk uk , j
x j
jk , j S k (6.19)
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 (6.19), the terms represent the transient accumu-
lation of ϕk in the control volume occupied by phase k; net advection of ϕ across the phase k portion
of the control surface; diffusion into phase k across the interfacial 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 (6.19) can be volume averaged by inte-
gration of the governing equation over Vk, leading to:
m
1
t
k kk x k kkuk, j jk, j V
k k jk ,m nk ,m dS k Sk (6.20)
j
S
where the subscripts are j = 1, 2, 3, and m = 1, 2, 3, and nk,m refers to the component of the unit out-
ward normal vector to phase k in the m coordinate direction. In Equation (6.20), the m ′′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 integra-
tion on the right side of Equation (6.20) refers to integration over the total surface area, including
the interfacial area, per unit volume (Si along the bundary 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.
Detailed modeling and analysis of these volume averaged formulations have been presented by
Crowe et al. (1998) for multiphase flows with droplets and particles. In the following subsections,
the above-generalized formulation will be used to obtain the conservation equations of mass,
momentum, and energy for multiphase flows.
6.3.2 Conservation of Mass
For conservation of mass, set ϕk = 1, jk = 0, and Sk = 0 in Equation (6.20) to obtain the following
volume-averaged continuity equation:
t
k k x
j
k k uk , j V1 m k dS (6.21)
S
For turbulent flows, the velocity component, uk,j, in the xj direction for phase k can be expressed
as a sum of a mean component, uk , j, and fluctuating component, u′k , j, as follows:
uk , j uk , j uk , j (6.22)
where u′k , j refers to the turbulent fluctuating deviation from the volume-averaged mean velocity
component in the xj direction.
In Equation (6.21), the integrand of the last term represents the interfacial mass flux. It refers to
mass transfer at the phase interface, such as phase change by evaporation, which leads to accumula-
tion or destruction of mass in one phase at the expense of the other. Evaporation of droplets will
increase the volume fraction in the vapor phase at the expense of liquid of the droplet. The interfacial
mass flux can be approximated as
m
1 1
V
m k dS
S
V m
m n m d
(6.23)
where ṁm and n refer to the rate of change of mass of droplet m due to evaporation, and the local
number density of droplets (in units of m–3), respectively. The summation is performed over all drop-
lets in the control volume including boundary droplets. If the droplets evaporate at the same rate, the
source term becomes n∙ṁd.
Then the mass conservation equations for the carrier (gas; subscript g) and dispersed (droplet;
subscript d) phases become:
t
g g x g
g ug, j nm d (6.24)
j
t
d d x d d ud , j nm d (6.25)
j
where ṁd refers to the droplet evaporation rate. Note that evaporation is a sink of mass in the dis-
persed droplet phase and a source term in the carrier gas phase.
Alternatively, Equation (6.21) can be written in terms of the mass of liquid droplets per unit vol-
ume of the gas–liquid mixture, called the liquid water content, as follows:
md
d d d (6.26)
V
Multiphase Flows with Droplets and Particles 269
This definition is analogous to the concept of specific humidity, except that it includes discrete
droplets rather than a continuous mixture of moisture in the air. Neglecting evaporation of droplets,
the mass conservation equation for the droplet phase becomes:
t
d
x j
dud , j 0 (6.27)
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 can be solved to determine the volume fraction of droplets, ζk, occupying the droplet–
gas control volume.
6.3.3 Momentum Equations
For the momentum equations, set ϕk,i = uk,i, jk,ij = pkδij – τk,ij and Sk,i = Fk,i,b in Equation (6.20) where
δij, τk,ij and Fk,i,b refer to, respectively, the Kronecker delta function, shear stress tensor including
molecular and turbulent components, and body forces, e.g., Fk,i,b = ρkgi (see Figure 6.4). In tensor
form, the volume-averaged momentum equation in the xi direction for phase k becomes:
k k uk ,i k k uk ,iuk , j k jk ,ij 1
t
x j
x j
V m u
S
k k ,i jk ,ij nk , j dS k Fk ,i b
(6.28)
x j
k k uk ,iuk , j
x j
k k uk ,iuk , j
x j
k k uk ,iuk , j (6.29)
The last term on the right side is the averaged Reynolds turbulent stress. It is analogous to the
Reynolds stress in single-phase flow (Chapter 3). However, in multiphase flow, the presence of
droplets (or particles) in the gas stream alters the length scales of turbulence due to gas–droplet
FIGURE 6.4 Momentum flux and force components of a control volume balance.
270 Advanced Heat Transfer
interactions. Similarly to the eddy viscosity method in single-phase flow (Chapter 3), the concept
of turbulent stress, τ kt ,ij , and turbulent viscosity, μk,t, can be defined as follows for multiphase flows,
u u
ktot,ij kl ,ij kt ,ij ; kt ,ij k k uk ,iuk , j k ,t k ,i k , j (6.30)
x j xi
where the superscripts tot, l, and t, refer to total, laminar, and turbulent, respectively. In subsequent
equations, these superscripts will be dropped for brevity. Unlike single-phase flow, the turbulent
stress can be caused by flow around the droplets irrespective of whether the bulk carrier flow is
turbulent.
The fourth surface integral term in Equation (6.28) represents interfacial momentum exchange
and forces arising from coalescence and/or evaporation of droplets, pressure effects and molecular
and turbulent shear stresses at the interface of the dispersed phase, respectively. The interfacial mass
transfer term can be approximated as
m u
1 1
V
m k uk ,i dS
S
V m
m d ,i d ,i
nmu
(6.31)
It will be assumed that all droplets have the same mass and size; evaporate at the same rate; and
move at the same local velocity. The cross-phase interactions due to the interfacial pressure and
shear stress components of the surface integral in Equation (6.28) are proportional to the relative
velocity difference between the gas and liquid (droplet) streams:
3 g
F D
1 1
V p
S
ij k ,ij nk , j dS
V m
m ,i
V m
m m f ug,i um,i V ug,i ud ,i
(6.32)
where the drag factor, fm, of droplet m represents a ratio of the drag coefficient, cd, to the Stokes
drag, 24/Red, as developed earlier in this chapter from the particle equation of motion. The empirical
coefficient, βV, is given by
cd 18 g
V (6.33)
24 / Re d Dd2
The subscripts g, d, and m refer to the gas (carrier phase), droplet (dispersed phase), and droplet
number, respectively. Also, Red and Dd are the droplet Reynolds number (based on the relative veloc-
ity) and mean droplet diameter. Assuming a spherical droplet, the drag coefficient can be related to
the Reynolds number as follows:
24 4.73
cd 0.00624 Re 0d.38 (6.34)
Re d Re 0d.37
g g ug , i g g ug,iug, j jg,ij
nm d ud ,i V ug,i ud ,i g g gi (6.35)
g
t x j x j
Multiphase Flows with Droplets and Particles 271
where i = 1, 2, 3, j = 1, 2, 3, and the total shear stress, τg,ij, in the above jg,ij term includes both laminar
and turbulent components using the eddy viscosity approach. The cross-phase interaction terms on
the right side represent the net averaged force on the carrier (gas) flow due to interactions with the
droplets (or particles).
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 turbulent
stress term can be formulated in terms of an eddy viscosity including a 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
modeling of turbulent gas–particle flows. In this approach, an additional equation is introduced
for the kinetic energy associated with the turbulent fluctuations of droplet velocities. The resulting
velocity is then related to the bulk viscosity of the gas/droplet mixture. The model is primarily suit-
able for vertical gas–particle flows with negligible particle–particle and particle–wall interactions.
Further detailed formulations of turbulent and transient gas–solid flows with particles were pre-
sented by Gidaspow (1994) and Martin (2011).
k kk k k uk , jk k jk , j 1
t
x j
x j
V m j
S
k k n
k, j k, j dS k k uk ,i gi
(6.36)
From left to right, the terms represent the transient energy change in phase k; convection of ther-
mal and kinetic energy; heat transfer; work rate due to shear stresses; interfacial flux of energy, heat
and work; and work due to body forces. The internal energy convection term on the left side and
flow work term (pressure gradient term) on the right side can be combined into a single convection
term with the specific enthalpy, hk, in phase k, rather than the current form in terms of the specific
internal energy, ek.
FIGURE 6.5 Energy, work, and heat transfer terms for a control volume.
272 Advanced Heat Transfer
Expanding the velocity, uk,i, into a mean component, uk ,i, and turbulent fluctuating deviation, u′k ,i,
the transient and convection terms for the carrier (gas; subscript g) phase can be approximated as
k kk 1
g g eg ug,iug,i (6.37)
t t 2
k k uk , jk 1
x j
x j g g ug, j eg 2 ug,iug,i (6.38)
where the turbulent fluctuating components are assumed negligible for low Mach number flows
where kinetic energy changes are small compared to thermal energy changes.
Combining and approximating the heat flux terms in Equation (6.36),
, j 1
Q
qeff
x j
k qk , j V1 q n k, j k, j dS
x j V
m (6.39)
S m
where the summation is performed over all droplets in the volume and boundary droplets. Also,
the effective mixture heat flux through the carrier phase and the droplets, q′′eff , j , and convective heat
transfer between the carrier phase and droplets, Q m, can be expressed as
Tg
qeff , j g qg, j d qd , j keff (6.40)
x j
Q m hAd ,m Td ,m Tg k Nu g D
Dm Td ,m Tg T
T d Tg
(6.41)
u
1 1
V pu
S
k, j uk ,i k ,ij nk , j dS
V m
F
m ,i m ,i ud ,i V ug,i ud ,i
(6.42)
where
Here Dm, fm, and the subscript m refer to the droplet diameter, drag factor, and droplet number,
respectively. As discussed previously, the drag factor is the ratio of the drag coefficient to the Stokes
drag, cD /(24/Re).
Therefore, assembling all terms yields the following total energy equation for the carrier phase,
D g g g pug, j Tg 1
gug,i ij
E
g
keff m g g ug,i gi (6.44)
Dt x j x j x j V x j
m
where D( )/Dt refers to the total (substantial) derivative, as defined in Chapter 1, including both
transient and convective components, and the subscripts are i = 1, 2, 3 and j = 1, 2, 3. The interfacial
work and heat transfer, Ėm, between the carrier and dispersed phases, is given by
Multiphase Flows with Droplets and Particles 273
Also, the subscripts s and m refer to evaluation at the surface, s, of droplet m, where the summa-
tion over m refers to all droplets in the control volume and boundary droplets. The convection and
flow work terms in Equation (6.44) can be combined to form an enthalpy term for the carrier gas
phase, or else retained separately as a pressure gradient term as shown above. Effects of contraction
or dilation of droplets have been neglected in the above analysis.
In cases where the volume fraction of the dispersed (droplet) phase is small (ζg ≈ 1), the total
energy equation reduces to:
D g g ug, j p 2T j ug,i ij
kg T Td Tg nm s, m ud ,i V ug,i ud ,i g ug,i gi (6.46)
Dt x j x j2
x j
ug,i
D g g ug,i u g
p
ug,i g
ij
d ,iug,i V ug,i ug,i ud ,i g g ug,i gi
nmu (6.47)
g, j
Dt x j x j
Using the continuity equation (6.24) and the product rule of calculus, the transient and convec-
tion terms on the left side can be rewritten as
D g g ug,i nmu u g g
u g , i ug , i ug,iug, j
ug,i
Dt
g ,i g ,i
t 2 g g ug,i x 2 (6.48)
j
Also, using the product rule, the pressure and viscous forces can be rewritten as
p ug,i
ug,i
xi
xi
ug,i p p xi
(6.49)
ij ug,i
ug,i
x j
x j
ug,i ij ij
x j
(6.50)
Then the thermal energy equation for the carrier (gas) phase is obtained by subtracting the
mechanical energy equation (6.47) from the total energy equation (6.44), yielding:
D g g eg p
ug, j
Tg
kg
T Td Tg g ij
ug,i
s V ud ,i ug,i (6.51)
nme
2
g
Dt x j x j x j x j
where ẽs refers to the combined internal specific energy at the surface, s, of a droplet, and its kinetic
energy based on the relative velocity difference,
274 Advanced Heat Transfer
ud ,i ug,i
2
es es (6.52)
2
From left to right in Equation (6.51), the terms represent the transient change of thermal energy
in the carrier phase; convective flux; pressure flow work to increase thermal energy; heat transfer
through the mixture; viscous flow work; energy influx from the dispersed phase; energy dissipation
(always positive) from droplet–gas interactions; and turbulent dissipation.
If the volume of the dispersed phase is small (ζg ≈ 1), pressure flow work effects on thermal
energy are neglected, and the specific heat, cv, is introduced, then the thermal energy equation can
be rewritten in terms of temperature as follows:
ud ,i ug,i
u u 2 (6.53)
2
DT Tg ug,i
g cv k
g T d g
T T ij
nm es V d ,i g ,i
t x j x j x j 2
If there are no droplets or particles in the flow field, it can be shown that this equation reduces
to the thermal energy equation for single-phase flow. This result has not included the effects of
radiation or multiple constituents in the carrier phase. If absorption, emission, or scattering effects
from droplet/particles become significant, or the gas contains a mixture of chemical species, then
additional terms must be included in the above thermal energy equation.
dud ,i 1 D p2 2
md Fd ,i Fb,i Fg,i g cD ur ,i md gi g 1 (6.54)
dt 2 4
d
where i = 1, 2, 3, and md, ud,i, cD, Dd, ur,i and gi represent the droplet mass, droplet velocity in the xi
direction, drag coefficient, droplet diameter, relative velocity difference between the droplet and gas,
and component of gravitational acceleration in the xi direction, respectively. The subscripts d and g
refer to the droplet and ambient gas, respectively.
Solving the equation of motion in the y-direction yields the following droplet velocity component
in the vertical direction:
t
vd t vd ,t tanh g 1 g (6.55)
d vd ,t
where vd,t is the terminal velocity at which the droplet weight is balanced by upward buoyancy and
drag forces,
4 gDd d g
vd ,t (6.56)
3cD g
Integrating the expression for the droplet velocity over time yields the following vertical position
of the droplet as a function of time,
vd2,t t
y t ln cosh g 1 g (6.57)
g 1 g / d d vd ,t
Heat transfer from the droplet is determined by an energy balance involving the rate of heat
transfer by convection, mass transfer, and radiation. Under a lumped capacitance assumption for the
droplet (Bi < 0.1), the temperature distribution within a droplet is assumed to be spatially uniform.
The energy balance between the droplet and surrounding gas can be expressed as
dTd
md c p , d Q c Q m Q r (6.58)
dt
Q c hc Ad Td ,s Tg, (6.59)
where the convective heat transfer coefficient, hc, can be determined from the Whitaker correlation
(1972):
1/ 4
NuD
hc kg
Dd
2 0.4 Re1D/ 2 0.06 Re 2D/ 3 Pr 0.4
s
(6.60)
Here ReD is the Reynolds number of the relative velocity between the droplet and gas stream.
This correlation is valid for 3.5 ≤ ReD ≤ 80,000 and 0.7 ≤ Pr ≤ 380. The fluid properties are evalu-
ated at the freestream temperature, Tg,∞, except for μs, which is evaluated at the surface temperature
of the droplet, Td,s. Alternatively, Hilpert’s correlation (Hilpert, 1933) may also be used to predict
the heat transfer coefficient.
Wang et al. (2008) developed correlations that account for the effects of co–flowing droplets
in the gas stream on the convective heat transfer coefficient. A multiphase Reynolds parameter,
Rem = Re (1 + w), is defined, where w is a nondimensional liquid droplet content in the freestream.
The Nusselt number correlation is expressed as
NuD c Re 1 w
m
Pr1/ 3 (6.61)
W
w= (6.62)
W0
276 Advanced Heat Transfer
Here W0 is a reference value of the liquid droplet content in the gas stream and c and m are
empirical coefficients. The resulting Nusselt number can then be used for both single-phase (with-
out droplets) and two-phase flows (with droplets), provided the Reynolds number is replaced by the
multiphase Reynolds number. Using this modified Reynolds number, Wang et al. (2008) collapsed
a range of data onto a single curve for cases with and without droplets in an airstream. The defini-
tion of Re(1 + w) provides a normalization of measured data and incorporates the effects of both
the single-phase Reynolds number and the droplet content in the gas stream on the heat transfer
coefficient.
For the present case, the rate of evaporative heat transfer from the droplet can be expressed as
where hfg, hm, ρv,s, and ρv,∞ refer to the latent heat of vaporization, convective mass transfer coef-
ficient, and vapor density at the droplet surface and freestream, respectively. The mass transfer
coefficient can be determined from the following Ranz–Marshall correlation (1952):
hm Dvg
ShD 2 0.6 Re1D/ 2 Sc1/ 3 (6.64)
Dd
where Dvg and Sc are the vapor–gas diffusivity and Schmidt number, respectively.
Also, the radiation heat transfer from the droplet surface can be written as
Q r Ad Td4,s Tg4, (6.65)
where ε and σ refer to the emissivity of the droplet surface and Stefan–Boltzmann constant, respec-
tively. Assembling these above expressions for the convection, mass transfer, and radiation into the
energy balance, the droplet temperature can then be obtained.
This lumped capacitance model was examined for high–temperature molten salt droplets in vari-
ous gases by Ghandehariun et al. (2017). Molten copper chloride droplets with a diameter of 4 mm,
zero initial velocity, and initial temperature of 504°C were considered. Sample results for the varia-
tions of droplet acceleration and temperature with time are shown in Figure 6.7.
FIGURE 6.7 (a) Acceleration and (b) temperature of a freely falling CuCl droplet in various gases. (adapted
from Ghandehariun et al., 2017)
Multiphase Flows with Droplets and Particles 277
Figure 6.7 (a) shows the downward acceleration of the falling droplet, initially at the gravitational
acceleration, but then decreasing in time due to the rising drag force. Three gases are considered –
argon; air; and helium. The kinematic viscosity of helium is about ten times larger than argon so
the Reynolds number is lower. Also, its density is an order of magnitude smaller so the drag force
is smaller. As a result, the acceleration of a droplet in helium is higher. Figure 6.7 (b) shows the
surface temperature variation of the droplet with time in the three gases. The change of temperature
is highest in helium. The thermal conductivity and heat transfer coefficient of helium are higher than
argon. As a result, the rate of heat transfer is also higher which leads to a larger temperature differ-
ence with time for helium.
T T 2kd T
d c p, d kd (6.66)
t r r r r
T r,0 Ti (6.67)
T T
kd Ad Q c Q m Q r ; 0; (6.68)
r rd r 0
The heat transfer rates for convection, mass transfer, and radiation in Equation (6.68) were
obtained earlier in the previous section.
The nondimensional radial coordinate, r*, and temperature, T*, are defined as follows:
r T T
r ; T (6.69)
rd Ti T
where rd, Ti, and T∞ represent the droplet radius, initial temperature, and ambient temperature,
respectively.
Using the method of separation of variables (Chapter 2), it can be shown that the solution of
Equation (6.66) can be written as
sin nr
T
n 1
Cn exp n2 Fo nr
(6.70)
Here Fo = αt/rd2 is the Fourier number and ζn are positive roots of the following transcendental
equation:
1 n cot n Bi (6.71)
278 Advanced Heat Transfer
where Bi = hrd/kd is the Biot number, h is the convective heat transfer coefficient, and kd is the ther-
mal conductivity of the droplet. Also, the values of the coefficients, Cn, in Equation (6.70) are given
by
4 sin n n cos n
Cn (6.72)
2 n sin 2 n
A six-term approximation of the series solution in Equation (6.70) for the nondimensional tem-
perature is accurate within ±7% for Fo = 0. The accuracy improves with an increase of the Fourier
number or number of terms in the series solution.
6.4.3 Solidification of Droplets
If the droplet is further exposed to a gas stream below its freezing point, solidification will occur
within the droplet. Consider solidification of molten droplets released from the top of a spray column
(see Figure 6.8). The droplets fall under gravity, release heat to a counter-flowing gas stream, and
change phase from liquid to solid when the droplet temperature falls below the solidification point.
Assume one-dimensional and symmetrical counter-current flow in a cylindrical heat exchanger.
Droplets enter from the top of the spray column at an initially zero velocity and fall under the influ-
ence of the drag force, Fd, buoyancy force, Fb, and gravitational weight, Fg. The relative velocity, vr,
is the difference between the gas velocity, vg, and droplet velocity, vd.
From Equation (6.54), Newton’s second law for the equation of droplet motion is given by
md ad Fd Fb Fg (6.73)
Substituting the prior expressions for the drag, buoyancy, and gravitational forces, the resultant
acceleration of a droplet can be determined as
cD g Ad vr2
ad g (6.74)
2 md
Ad N d 4 rd2 (6.75)
where Ṅd is the number of droplets per unit time (droplets per second).
FIGURE 6.8 Vertical spray column with upward gas flow and falling molten droplets.
Multiphase Flows with Droplets and Particles 279
The droplet velocity is obtained by integrating the droplet’s downward acceleration over the
height of the heat exchanger. Discretizing this height over a number of incremental distances, Δy,
the velocity can be found by summing the incremental increases of velocity through each element,
thereby yielding:
vd ,e vd2,i 2a y
m 0
d (6.76)
where M, vd,e, and vd,i refer to the number of elements of height Δy that sum to the total height of
the spray column, droplet velocity at the exit, and inlet velocity of the droplet at the top of the spray
column.
The cooling gas stream enters from the bottom of the spray column. Its inlet velocity, vg,i, can be
expressed in terms of the incoming mass flow rate of gas, ṁg, as follows:
m g
vg ,i (6.77)
g,i Ag
where ρg,i and Ag refer to the gas density at the inlet conditions and cross-sectional area of the cylin-
drical heat exchanger, respectively.
As the gas flow upwards, heat transfer from the droplets causes the gas temperature to increase,
and therefore also expansion of the gas due to the temperature increase. From conservation of mass,
this implies that the gas velocity also increases to maintain a uniform mass flow rate at a steady
state. This increase in the gas velocity due to expansion effects is called the gas expansion velocity,
vg,exp. From a mass balance over the heat exchanger, the net expansion flow rate is the change of the
volume flow rates between the inlet and exit of the spray column, therefore:
vg,exp
g,e g,i m g (6.78)
Ag
where υg,e and υg,i are the specific volumes of the gas at the exit and inlet conditions, respectively.
The relative velocity between the gas and droplet becomes:
These velocities vary with position and temperature. The heat transfer analysis will require this
relative velocity in order to determine the Reynolds and Nusselt numbers.
Assuming a lumped capacitance approximation for the droplet (Bi < 0.1), the temperature dis-
tribution within the droplet is spatially uniform. The energy balance for convective heat exchange
between the upward flowing gas and a droplet can be written as
dTg
m g c p, g hc Ad Td Tg (6.80)
dy
Similar to the previous equation of motion, the energy balance is solved by discretizing the height
over a number of incremental distances, Δy, and approximating the temperature difference by a
central-difference method. The rate of convective heat transfer on the right side of Equation (6.80)
becomes:
280 Advanced Heat Transfer
T T T T
Q c hc Ad d ,i d ,e g,i g,e (6.81)
2 2
where the subscripts i and e refer to the inlet and exit of each element in the discretization, respectively.
The rate of heat transfer from the droplet during sensible cooling is given by
Q d m d c p,d Td ,i Td ,e (6.82)
where ṁd is the droplet mass flow rate. Also, the rate of heat transfer from the droplet to the gas in
an element can be written as
Then a steady-state energy balance requires that the convective heat transfer from the droplets to
the gas must balance the energy loss of the droplets, as well as the energy gain by the gas stream,
c Q=
Q= Q g (6.84)
d
Substituting the individual expressions into the latter equality, and rearranging in terms of the
droplet inlet temperature, yields the following result:
m g c p, g
Td ,i
m d c p,d
Tg,e Tg,i Td ,e (6.85)
This equation is only valid for the sensible cooling portion of the droplet trajectory as it does not
include the latent heat released by the droplets upon solidification.
When the droplet temperature falls below the solidification point, the latent heat transfer process
must also be included. Equating Equations (6.81) and (6.83) and rearranging in terms of the exit gas
temperature,
Td ,i Td ,e 1 Tg,i hc Ad
Tg,e ; (6.86)
1 2m g c p, g
This equation is valid for both sensible and latent heat sections of the heat exchanger. During the
phase change process, the droplet inlet temperature, Td,i, and exit temperature, Td,e, in an element
remain constant in the above expression. The point at which the change of enthalpy in the gas stream
matches the latent heat released by the molten droplet represents the final point in the phase change
process.
Before/after the phase change process is completed, sensible cooling occurs and the temperature
of the droplet changes between the inlet and exit of an element. Substituting Equation (6.85) into
(6.86) yields:
Td ,e 1 Tg,i hc Ad
Tg,e ; (6.87)
1 2m d c p,d
This result can be used to determine the exit temperature of the gas across each element during
the sensible heat transfer processes.
Multiphase Flows with Droplets and Particles 281
For the initial sensible heat sections of the spray column, Equation (6.85) is used to find the
droplet inlet temperature for an element. When the droplet temperature reaches the phase change
temperature, Equation (6.86) is used for the latent heat transfer processes. Finally, once the change
in air enthalpy equals the latent heat of the molten droplet, the phase change process is completed,
and subsequent heat transfer occurs between solid particles and the gas stream.
Predictions of heat recovery from molten salt droplets in air through a counter–flow spray col-
umn were presented by Jaber et al. (2010). The phase change temperature of molten copper chloride
is 412°C. Solidification occurs when molten CuCl droplets are initially sprayed at a temperature of
530°C and cooled in their downward trajectory by the counter-flowing air stream. Droplets of 1 mm
diameter are sprayed from the top of the spray column at an initial velocity of 0.02 m/s. Figure 6.9
illustrates the droplet, relative, and expansion velocities, as well as the temperatures of the molten
CuCl and air through the spray column. Note that CuCl droplets move from right (top) to left (bot-
tom of the spray column) in the figure.
A droplet enters the heat exchanger and accelerates as it moves downward to reach a final exit
velocity of about 3.2 m/s. The air inlet velocity is 0.2 m/s. In Figure 6.9 (a), the rate of increase of
the droplet velocity falls with distance due to the increasing drag force. The air expansion velocity
is zero as it enters and increases to 0.31 m/s at the top of the heat exchanger. As the air temperature
increases in the vertical direction due to heat recovery from the droplets, it expands and therefore its
velocity increases with height.
Figure 6.9 (b) illustrates the temperature profiles of air and molten CuCl in the spray column.
Sensible heat transfer occurs initially between the molten CuCl droplets and air. Further cooling and
solidification occur when the droplets reach 412°C. During the phase transition, the temperature
remains nearly constant. The air temperature increases as it absorbs latent heat of the droplets. The
temperature difference between both streams decreases and the heat transfer rate also decreases.
When all of the latent heat released by the CuCl droplets is absorbed by the air stream, a final pro-
cess of sensible heat transfer occurs between the droplets and air stream. At the top of the spray
column, the exiting air reaches a temperature of about 493°C.
Solidification of droplets occurs in a range of applications, such as spray coatings, manufactur-
ing processes 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 (see Figure 6.10). The dis-
integrated droplets come into contact with a cross-stream fluid to initiate secondary breakup into
FIGURE 6.9 (a) Velocity and (b) temperature of CuCl droplets in a vertical spray column. (adapted from
Jaber et al., 2010)
282 Advanced Heat Transfer
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. De-stabilization 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. Heat transfer correla-
tions of the heat transfer coefficient for melt particularization processes have been presented by past
investigators such as Mehrotra (1981) and Sahm and Hansen (1984).
B
i c FWV (6.88)
t
where ρi, B, ηc, F, W, and V, refer to the ice density, ice thickness, collection efficiency, view fac-
tor, 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 surface
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 direction 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 (6.88) subject to an initial condition of B = 0 at t = 0, yields:
FWV
B c t (6.89)
i
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 condi-
tions. It was developed particularly for applications to unheated cylindrical conductors (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 6.11).
where ξ, ca, hsl, Tw, and Ta refer to the recovery factor (an empirical coefficient accounting for vis-
cous 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 curvature are assumed to be negligible for
the thin water layer.
When sufficient energy is imparted into the ice surface during freezing to sustain an unfrozen
water layer, including the release of latent heat from the impinging droplets, then transition to glaze
ice occurs. Performing an energy balance in the unfrozen water layer on a glaze ice surface,
B hV 2 1
qw i hsl cV 3W h (Tw | B b Ta cWVcw Tw Ta (6.90)
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 sensible 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 mecha-
nism of heat transfer in the thermal energy equation is lateral conduction across the layer in the
y-direction, perpendicular to the surface, yielding,
284 Advanced Heat Transfer
2Tw
0 (6.91)
y 2
subject to T(B) = Tf (phase change temperature). At the liquid/air interface (y = B + b), the energy
balance in Equation (6.90) provides the remaining boundary condition for the solution of the ther-
mal energy equation.
Solving the heat equation and differentiating the resulting temperature profile to find the tem-
perature gradient,
Tw C2 C1 T f Ta
(6.92)
y 1 C1b
where
h cVWcw hV 2 kwc1 T f Ta
C1 ; C2 (6.93)
kw 2ca i hsl aT
Then substituting the water temperature into Equation (6.90) and rearranging,
B q CC b
w 1 2 C3 C2 (6.94)
t i hsl 1 C1b
where
cWV 3
C3 (6.95)
2 i hsl
Thus, the ice growth initially follows Equation (6.89) during rime ice growth and then Equation
(6.94) after the transition to glaze (wet) ice.
In Equation (6.94), the thickness of the unfrozen water layer, b(t), is required. From a mass bal-
ance at y = B (ice/water interface),
dB db
i w c FWV (6.96)
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 (6.94), equating this result with Equation (6.89) 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
(6.96),
FWV i
b c t tw B Bw (6.97)
w w
The nonlinear first-order ordinary differential equations in Equations (6.94) and (6.96) can be
solved numerically to yield solutions for B(t) and b(t).
Multiphase Flows with Droplets and Particles 285
Sample model results of ice accretion on a circular conductor of radius 1.05 cm due to freezing
rain precipitation are shown in Figure 6.12. 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 6.12, 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 measurements of ice growth due to freezing pre-
cipitation 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).
FIGURE 6.12 Thickness of an ice layer during freezing precipitation on a heated circular conductor for (a)
rime ice and (b) glaze ice conditions. (adapted from Naterer, 2011)
286 Advanced Heat Transfer
through a nozzle into a heated gas stream and vaporized. Solids form as moisture leaves the droplets.
Detailed heat and mass transfer models of these evaporation and particle 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 surrounding 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.
During the shrinkage period, the solvent evaporates from the droplet, leaving a higher concen-
tration of solute in the droplet. As more volatiles evaporate from the droplet, saturation 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 equilibrium
saturation (ES). The solute begins to precipitate from nucleation sites and crystals through the drop-
let (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 saturation. In this
case, solution precipitation occurs where the local solute concentration is higher than the equilib-
rium 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 that 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.
1
t r 2 r
r 2 vr 0
(6.98)
where vr refers to the velocity component in the radial direction. Within the droplet, solvent migrates
outward at a rate proportional to the diffusion coefficient of liquid solute within 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
l 1 2 l
l lr Dls 0 (6.99)
t r 2 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 transfer 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 surface 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,
l
w (6.100)
1 l
4 r dr
2
l s
m
y s 0
ro
(6.101)
4 r dr
m0 2
l s
0
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:
w 16 2 lr 2 w
2
Dls (6.102)
t mo2 y 1 w y
l 0
w
1 l 0
at t 0 (6.103)
w
y
0 at y 0 (6.104)
288 Advanced Heat Transfer
2
dmd 16 2 r 2 w
Dls l o at y 1 (6.105)
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:
1
3mo 1 w
ro3
4
0
l
dy
(6.106)
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 pre-
cipitated to create a thin outer shell on the particle (called an induction period). After this induc-
tion 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 evaporation 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 written as
4
ms
3
l rp3,o rp3,i lVp
(6.107)
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 6.14. The sol-
vent 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 carrier gas flow rate of 2 L/
min. The gas and droplet temperatures at the reactor inlet are 25°C. The results show different drop-
let evolution stages in the shrinkage, induction, and constant diameter periods. In Figure 6.14, 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
FIGURE 6.14 (a) Droplet diameters and (b) evaporation rates during different drying periods. (adapted from
Eslamian et al., 2009)
Multiphase Flows with Droplets and Particles 289
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 temperature
of a droplet increases, the thermal energy of molecules also increases at the surface of the droplet,
therefore increasing the rate of evaporation. Hoffman and Ross (1972) presented the following cor-
relation for evaporating droplets and 1 < ReD ≤ 400:
NuD 1 1 Re D Pr
1/ 3
Re 0D.077 (6.108)
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 atomi-
zation 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 We Oh (6.109)
D0
The Sauter mean diameter, D32, represents an average droplet size defined by the diameter of a
sphere with the same volume to surface area ratio as a droplet of interest. The Ohnesorge number,
Oh, is a ratio of the square root of the Weber number, We (defined in Chapter 3), to the Reynolds
number, Re,
We
Oh (6.110)
L 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 rotational 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.
The sphericity, Φ, of the particle is the ratio of the surface area of a sphere (with an equivalent
volume as the particle) to the actual surface area of the particle,
1/ 3 6V
2 /3
(6.112)
Ap
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 D p2
m pc p h Ta Tp,s (6.113)
dt
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 tem-
perature of the particle, and Ta is the air temperature. The heat transfer coefficient can be determined
by the Frossling correlation,
hD p
Nu p 2 0.552 Re1p/ 2 Pr1/ 3 (6.114)
ka
hD p
Nu p 2 0.552 Re1p/ 2 Pr1/ 3 1/ 4 (6.115)
ka
hm D p
Shp 2 0.552 Re1p/ 2 Sc1/ 31/ 4 (6.116)
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 identical 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.
Multiphase Flows with Droplets and Particles 291
In the second stage during the phase transition process, the heat balance can be written as
Nu p
Dp ka Ta Tmelt m evhv m f hsf (6.117)
where ṁev is the evaporation rate, ṁf is the melting rate and hv and hsf refer to the latent heat of vapor-
ization and fusion of ice, respectively.
The mass of the particle ice core surrounded by liquid water, mp,i, can be determined by
dm p Sh
m ev D p p a Dv,a v,s v,a (6.119)
dt
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
m p , w m p m p ,i (6.120)
Also, the mass of the particle can be found based on the mass of the liquid layer,
m p m p ,i w
6
D 3p D 3p,i
(6.121)
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).
dI a I dx (6.123)
where the proportionality constant, aλ, is called the absorption coefficient. It is usually tabulated
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 radiation equation of transfer (as presented earlier in Chapter 4) for a
homogeneous, non–scattering gas layer:
dl
a dx (6.124)
I
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 radiation 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,
I 0 IL
1 exp a L (6.126)
I 0
Based on Kirchkoff’s law for a diffuse gas (recall αλ = ελ; Chapter 4),
E d 1 E 1 exp a L d
b,
T
g 0
4 b, (6.127)
E d
0
b,
g
0
For participating gases (water vapor and carbon dioxide) in radiation exchange at high tempera-
tures, the following correlations can be used:
0.45
T
w cw w g (6.128)
Ts
0.65
T
c cc c g (6.129)
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 surface 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 emissivity 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,
g w H 2O w CO2 (6.130)
g w H 2O w CO2 (6.131)
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 concentrations 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 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 hemi-
spherical domain are shown in Figure 6.17. These emissivities increase at higher beam lengths and
partial pressures of the gas. Also, emissivity correction factors at temperatures of 400 K and 1,200
K are shown in Figure 6.18. In a gas mixture of water vapor and carbon dioxide, the emissivity and
absorptivity correction factors are the same, Δα = Δε.
FIGURE 6.17 Emissivities of (a) water vapor and (b) carbon dioxide in a gas mixture at 1 atm total pressure
in hemispherical domain. (adapted from Hottel, 1954)
294 Advanced Heat Transfer
FIGURE 6.18 Emissivity correction factors of mixtures of water vapor and carbon dioxide at temperatures of
(a) 400 K and (b) 1,200 K. (adapted from Hottel, 1954)
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).
qrad A gTg4 gTs4 (6.132)
When soot particles are not present in the gas region, these emissivity and absorptivity coeffi-
cients are readily available.
Consider a gas–particle mixture within gray walls instead of a blackbody enclosure, while still
assuming no soot is present. The incident radiation on the surface (irradiation, Gλ) and radiosity
(reflected and emitted radiation from the surface, Jλ) may be written as
G g J gEbg (6.133)
s gEbg g sEbs
J sG sEbs (6.134)
1 s g
The latter equality in Equation (6.134) is obtained through substitution of Equation (6.133) into
the first equality of Equation (6.134). It was assumed that εs = αs (gray wall) and αs = 1 – ρs.
From an energy balance on the surface of the enclosure, the net radiation exchange is given by
G J g J gEbg J
qrad (6.135)
s A
qrad
1 s g
gTg4 gTs4 (6.137)
where ρs = 1 – εs and τg = 1 – αg. This result assumed gray walls, but still without soot particles 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 oper-
ating pressure, quality of fuel atomization, particle distribution within the chamber, air/fuel ratio,
and other factors.
Using Beer’s law and a functional form of the absorption coefficient based on experimental
data,
s
a (6.138)
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:
,mix 1 exp ag L 1 exp asL 1 exp ag L 1 exp asL
(6.140)
Treating the soot particles as a gray substance (εs = αs) leads to:
mix ~ g s g s (6.142)
This result can be used to account for soot particles in the radiation exchange during gas–
particle flows with combustion.
296 Advanced Heat Transfer
The previous example showed that solid particles can significantly affect the process of radiation
exchange in gas–particle flows. Radiation through participating media of combustion gases with
particles was reported by Baker and Miller (2013). Further details on radiative heat transfer in gas–
droplet flows were presented in books by Klinzing (1981) and Kreith, Manglik, and Bohn (2010).
6.8.2 Particulate Radiation
Particulate matter is a complex mixture of solid particles and liquid droplets suspended in a gas in
the form of soot, dust, volatile matter, dust, sulfates, etc. Particulates can emit and absorb radiation
directly, for example in fires, or indirectly through the reaction of gases. When particulate matter
is released to the environment and ingested into lungs, it may cause serious health issues such as
cardiovascular disease, respiratory irritation, asthma, among other health conditions. Fine particles,
in particular, smaller than 10 microns in diameter, can pose significant health risks if ingested into
lungs and the bloodstream.
Detailed effects of particulate matter on radiation heat transfer can be analyzed via the radiation
equation of transfer (Chapter 4). The intensity field, Iλ, propagates along a given direction, s, and
crosses a solid angle, ω, at a given position, r, according to the following equation of transfer:
4
dI r,s T 4 ,s
ds
a ,a ,s I r,s a ,a n2
4 I
0
r,s s d
(6.143)
where aλ,a, σl,s, Φ, and n represent the absorption coefficient, scattering coefficient, scattering phase
function and refractive index, respectively. Adding the contribution of a particulate phase as the scat-
tering mechanism in the transfer equation,
4
dI r,s T 4 , p
ds
a ,a a , p , p I r,s a ,a n2
4 I
0
r,s s d E p
(6.144)
where aλ,p, σλ,p and Ep are the equivalent absorption coefficient due to the presence of particulates,
equivalent scattering factor, and equivalent emission by particulates, respectively.
The equivalent emission of particles, Ep, and equivalent absorption coefficient, aλ,p, are defined as
N
Tp4,n
E p lim
V 0
n 1
p, n Ap , n
V
(6.145)
Ap , n
a , p lim p, n (6.146)
V 0 V
n 1
where εp,n, Ap,n and Tp,n represent the emissivity, projected area, and temperature of particle n within
a volume, V, consisting of N particles. Also, the projected area, Ap,n, of particle n is given by
D p2,n
Ap , n (6.147)
4
where Dp,n is the diameter of particle n. The equivalent particle scattering factor, σλ,p, is defined as
N
1 f
Ap , n
, p lim
V 0
p, n 1 p,n V
(6.148)
n 1
Multiphase Flows with Droplets and Particles 297
where fp,n is the scattering factor of particle n. Using these particulate functions, the radiation equa-
tion of transfer can be solved based on methods presented earlier in Chapter 4 to predict the effects
of particulate matter on radiation heat transfer.
du
o b (6.149)
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:
mum,c D D 2 1/ 2
Re crit 1000 1 1 o m
2
(6.150)
b 3000 b
where um,c is the critical (transition) velocity. The subscripts b and m refer to Binghamian and mix-
ture, 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 sufficiently small diameter, less
than Ds,c, will remain in stable laminar flow (Dedegil, 1986),
3 o
Ds,c (6.151)
2 s l g
The pressure drop, Δp, over a section of length ΔL in a pipe, for laminar Binghamian slurry flows
can be estimated based on the following Buckingham equation (Wasp et al., 1977):
p 16 o 32u
B 2m (6.152)
L 3 D D
The pressure drop is an important parameter in the design of pumping systems for slurries and
liquid–solid flows. If particles exceed a critical diameter of Ds,c, then increasing the flow rate and
velocity through the pipe to turbulent flow conditions may prevent settling out of particles. However,
mixing due to turbulence and pumping power is also increased.
us uso s (6.153)
4 D p,50 g s l
uso (6.154)
3 cd l
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 cumulative distribution.
Also, cd is the drag coefficient, which can be correlated as
4 24
cd 0.4 (6.155)
Res Res
where
uso D p,50
Re p = (6.156)
vl
These results can be derived based on approximate solutions of the settling and inertial effects of
particles during transport by the liquid–solid flow.
In vertical liquid–solid flows, there is a low tendency for segregation of solid particles due to the
magnitude and direction of forces acting on the solid particles. In general, a nearly uniform solid
Multiphase Flows with Droplets and Particles 299
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 profile, which describes the spatial distribution of
particle density across the pipe, becomes an important design factor in horizontal slurry flows.
1/ 6 1/ 2
D s l
uc 3.525 s0.234 p,50 2 gD (6.157)
D l
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. Goharzadeh et al. (2010)
developed alternative correlations based on a balance of energy dissipated by turbulence eddies
against gravity forces acting on entrained particles, which serve to settle out the particles along the
deposition layer.
At the critical velocity, sand particles are entrained from the bottom deposition layer and trans-
ported by the core flow to generate a bed–load pattern. The energy required to maintain particles
in suspension will balance the fraction of turbulent energy dissipated in suspending the particles.
Beyond the critical velocity, solid particles are gradually lifted off the bed interface. But since
the flow rate and fluid forces are not sufficient to hold the particles in suspension, the particles
fall back down onto the stationary bed. This process of saltation generally occurs with larger
particles due to gravitational effects. The saltation mode is characterized by the formation of
dunes as illustrated in Figure 6.20. This flow pattern and sand transport processes depend on the
dune wavelength, height, and length. The characteristics of the particles (density and diameter)
and Reynolds number also affect the development of the saltation flow patterns. As the flow rate
increases, the proportion of particles in suspension also increases and leads to a fully developed
slurry mixture.
Similar to two-phase flow maps in the previous chapter, phase diagrams are also useful to deter-
mine various characteristics and features of the particle 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,
FIGURE 6.20 Flow distributions of (a) bed–load; (b) saltation; (c) suspension flows.
300 Advanced Heat Transfer
in terms of a friction velocity, uo. The laminar sublayer near the wall of the pipe corresponds to a
low value of uo. 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 (consisting 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.
FD V
k1 k2 V 2 (6.158)
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 (6.159)
v
The area is represented by the product of the number of particles, Np, and the surface area of a
single particle, sp,
As = N p s p (6.160)
The number of particles can be calculated from the ratio of the total volume of solids to the vol-
ume of a single particle, Vp, as follows:
S0 L 1 v
Np (6.161)
Vp
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:
vVp D
rh v p (6.162)
1 v s p 6 1 v
where Dp and Φ refer to the particle diameter and sphericity.
Substituting these equations into Equation (6.158), and using experimentally determined coef-
ficients for k1 and k2 of 150/36 and 1.75/6, respectively, the following Ergun equation (subscript E)
for the friction factor is obtained,
150
f p, E 1.75 (6.163)
Re p
which is valid for 1 < Red < 1,000, where the Reynolds number for a packed bed, Red, is defined by
D pVs
d (6.164)
v 1 v
At low Reynolds numbers below 1 (Stokes flow; or “creeping flow”), viscous effects are domi-
nant and the Navier–Stokes equations can be solved to obtain the following Stokes law:
FD /As 6 v
cd (6.166)
V 2 / 2 D pVs
Rewriting this result in terms of the friction factor for Stokes flow (subscript S),
24 v
f p,S 4cd (6.167)
D pVs
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
Re Re p
f p,T l ,T f p,S f p, E (6.168)
Re p Reu,T
302 Advanced Heat Transfer
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 Stokes formula-
tions becomes:
1/ n
1 n 1 n
f p,C f p, S (6.169)
f p,T f p, E
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 condi-
tions, including the transition region between low and high Reynolds numbers.
The pressure drop through the packed bed can be described by the Ergun equation (1952):
The pressure drop through the packed bed depends on the length, fluid velocity, packing charac-
teristics including the void fraction, and fluid properties of dynamic viscosity and density.
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 super-
ficial 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:
hD p 1 v
NuD
k
v
0.5Re1D/ 2 0.2 ReD2 / 3 Pr 1/ 3
(6.171)
in the range of 20 < ReD < 104 and 0.34 < ζv < 0.78, where
D pVs
ReD (6.172)
v 1 v
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.
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 fluids 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).
6.10.1 Transport Phenomena
Brownian diffusion and thermophoresis are significant transport mechanisms in nanofluids. Brownian
diffusion involves random drifting of suspended nanoparticles in a base fluid as a result of inter-
molecular collisions between the liquid molecules and nanoparticles in a flow stream. This random
Brownian motion of particles enhances the rate of heat transfer. Thermophoresis (or a Ludwig–Soret
effect) is a thermodiffusion process where nanoparticles exhibit an unusual migration against the
temperature gradient from warmer to colder regions. This process leads to a nonuniform nanopar-
ticle volume fraction distribution.
Since the length scales of nanofluids approach the molecular length scales, further nanoscale
transport processes occur which affect the motion and thermophysical properties of the fluid. For
example, the fluid’s viscosity and thermal conductivity increase near the pores of the wall as well
as potentially the chemical reactivity of species at the surface. Electrolyte solutions confined in
nanopores of a wall contain surface charges that induce a charge distribution near the surface
called an electrical double layer. This may affect the near-wall fluid motion as counter–ions (ions
charged oppositely to the static wall charges) interact with co-ions in the fluid (same charge as the
wall charges). These processes can be used for the 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/par-
ticle interface; and nanoparticle clustering. Liquid molecules may form a nanolayer around the par-
ticles and hence increase the ordering of the atomic structure around the particle. Also, nanoparticles
can form clusters that become more conductive due to the changes of packing fraction and effective
volume as a result of clustering.
Nanofluids can be treated as two-phase multicomponent fluids consisting of a base fluid and
nanoparticles with volume fraction weighted properties as follows:
2
m
k 1
k k 1 p f p p
(6.174)
c p, m c
k 1
k p, k 1 p c p, f pc p, p
(6.175)
c p m k c p k 1 p c p f p c p p (6.176)
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 & Passman,
1999) as follows:
m f 1 2.5 p (6.177)
where ζp is the particle volume fraction. Using these properties, the energy equation for fully devel-
oped flow of a nanofluid in a heated tube becomes:
2
c p m u Tx km kd xT2 1r r r Tr m ur
2
(6.178)
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 thermal 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
k
vdr , p v p v f vm
k 1
k
m
v f ,k
(6.179)
where the bold font refers to a vector quantity. The slip velocity, vfp, can be determined from:
Multiphase Flows with Droplets and Particles 305
p p m
v pf am (6.180)
fdrag p
where the particle relaxation time, τp, drag factor, fdrag, and acceleration, am, are given by
p D p2
p (6.181)
18 f
am g vm vm (6.183)
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 direc-
tion of the net bulk motion of the fluid.
6.10.3 Thermal Conductivity
The previous calculations of the mixture density and specific heat of a nanofluid were based on
the volume fraction-weighted sum of the base fluid and nanoparticle properties. The mixture ther-
mal conductivity is more complex and cannot be readily determined by a simple volume fraction-
weighted conductivity. In general, universal theories or relations for the mixture thermal conductivity
are not available for all types of nanofluids. From experimental studies, it is known that the thermal
conductivity of nanofluids depends on many factors such as the thermal conductivities of the base
fluid and nanoparticles, volume fraction, surface area, shape of the nanoparticles, temperature, and
intermolecular interactions.
Although there are no universal relations to predict the thermal conductivity of nanofluids, vari-
ous semi-empirical correlations have been developed based on the following definition of the mix-
ture thermal conductivity of a two-component mixture:
k p p T /x p k f f T /x f
km (6.184)
p T /x p f T /x f
Various correlations have been developed based on 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.
k p 2 k p k f p 2k f
km kf (6.185)
k p k p k f p 2k f
Maxwell’s model predicts that the mixture thermal conductivity depends on the thermal conduc-
tivity of the spherical particle and base fluid, as well as the volume fraction of the solid particles.
More recent developments have attempted to include other effects such as intermolecular inter-
actions among randomly distributed particles, nonspherical particles, clustering of particles, the
306 Advanced Heat Transfer
nanolayer between the nanoparticles and base fluid, Brownian motion, and thermophoresis. 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) k 2 k k 1 3 2k
km kf
p p f p f
k p k p k f 1 3 p 2k f
(β is the ratio of the nanolayer thickness, or average thickness of a liquid molecular layer
around the nanoparticles, to the particle radius)
Putnam et al. model k p 2 k p k f p 2k f 1 18 T
(2006)
km 1 A Re m Pr 0.333 p k ; Re
k p k p k f p 2k f f p D p
(includes the effects of convection near the nanoparticles due to Brownian movement, where
A and m are empirical coefficients, and κ is Boltzmann’s constant)
Koo and Kleinstreuer k p 2 k p k f p 2k f T
model (2005) km k f 5 10 4 p p c p f T , p
k p k p k f p 2k f p Dp
f(T, ζp) = (−6.04ζp + 0.4705)T + (1722.3ζp − 134.63)
(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)
k p k f k p k f p 2k p p ln k p k f / 2k f
km
k
k p k f k p k f p 2k f p ln k p k f / 2k f
f
(includes the effects of an interfacial shell, axial ratio of nanoparticle shapes, and the spatial
distribution of carbon nanotubes in the nanofluid)
Udawattha et al. model 3 p 1 2h / D p k p k f
3
0.0009T 0.25c p k p D p T
(2018) km 1 f p kf
k p 2k f p 1 2h /D p k p k f 200 f p Dp
3 3
(includes the effects of the nanolayer thickness, h, particle radius, Dp, particle volume
fraction, ζp, and temperature, T, in K)
Multiphase Flows with Droplets and Particles 307
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 0.4328 1.0 11.285 0p.754 PeD0.218 Re 0D.333 Pr 0.4 laminar
NuD 0.0059 1.0 7.6286 0.6886
p Pe 0.001
D Re 0.9238
D Pr 0.4
turbuleent
(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 NuD = 0.021 Re 0D.8 Pr 0.5
(1998) (turbulent flow of Al2O3–water and TiO2–water nanofluids in a tube)
Das et al. correlation NuD = c Re mD Pr 0.4
(2003) (correlation for pool boiling of Al2O3–water nanofluids, where c and m are empirical
coefficients that depend on the particle volume fraction)
Yang et al.
NuD c Re mD Pr 0.333 1/3 f /
0.14
correlation (2005)
(laminar flow of a graphite–synthetic oil mixture in a horizontal tube heat exchanger, where
α, c, and m represent the nanoparticle aspect ratio, and nanofluid and temperature-dependent
empirical parameters, respectively)
Buongiorno
NuD
f /8 Re D 1000 Pr
correlation (2006)
1 f /8
1/ 2
Pr 2 /3
1
(turbulent flow of nanofluids in a tube, where f and δϕ are the friction factor and
dimensionless thickness of the laminar sublayer, respectively)
REFERENCES
Baker, T.M. 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.
Bergman, T.L., A.S. Lavine, F.P. Incropera, and D.P. DeWitt. 2011. Fundamentals of Heat and Mass Transfer,
7th Edition, New York: John Wiley & Sons.
Buongiorno, J. 2006. “Convective transport in Nanofluids”, ASME Journal of Heat Transfer, 128: 240–250.
Crowe, C. 1999. “Modeling Fluid–Particle Flows: Current Status and Future Directions”, AIAA Paper 99–3690,
30th AIAA Fluid Dynamics Conference, Norfolk, VA.
Crowe, C., M. Sommerfeld, and Y. Tsuji, 1998. Multiphase Flow with Droplets and Particles, Boca Raton:
CRC Press/Taylor & Francis Group.
Clift, R., J.R. Grace, and M.E. Weber. 1978. Bubbles, Drops and Particles, New York: Academic Press.
Das, S.K., N. Putra, and W. Roetzel. 2003. “Pool Boiling Characteristics of Nanofluids”, International Journal
of Heat and Mass Transfer, 46(5): 851–862.
Davies, C.N. 1945. “Definitive Equations for the Fluid Resistance of Sphere”, Proceedings of the Royal Society,
A83(57): 259–270.
Dedegil, M.Y. 1986. “Drag Coefficient and Settling Velocity of Particles”, International Symposium on Slurry
Flows, ASME, FED, Anaheim, CA.
Drew, D.A. and S.L. Passman. 1999. Theory of Multicomponent Fluids, Berlin: Springer.
Ergun, S. 1952. “Fluid Flow through Packed Columns”, Chemical Engineering Progress, 48: 89–94.
Eslamian, M., 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.
Ghandehariun, S., M.A. Rosen, M. Agelin–Chaab, and G.F. Naterer. 2017. “Heat Transfer from Molten Salt
Droplets in Various Gases”, International Journal of Heat and Mass Transfer, 105: 140–146, 2017.
Gidaspow, D. 1994. Multiphase Flow and Fluidization: Continuum and Kinetic Theory Descriptions,
Cambridge: Academic Press.
308 Advanced Heat Transfer
Goharzadeh, A., P. Rodgers, and C. Touati. 2010. “Influence of Gas–Liquid Two-phase Intermittent Flow on
Hydraulic Sand Dune Migration in Horizontal Pipelines”, ASME Journal of Fluids Engineering, 132(7):
071301 (7 pages).
Goodwin, E.J., 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, pp. 267–273, Hanover, NH.
Hauk, T., E. Bonaccurso, P. Villedieu, and P. Trontin. 2016. “Theoretical and Experimental Investigation of
the Melting Process of Ice Particles”, AIAA Journal of Thermophysics and Heat Transfer, 30: 946–954.
Hewitt, G.F., G.L. Shires, and Y.V. Polezhaev, Eds. 1997. International Encyclopedia of Heat and Mass
Transfer, Boca Raton: CRC Press/Taylor & Francis Group.
Hilpert, R. 1933. “Heat Transfer from Cylinders”, Forschung im Ingenieurwesen, 4: 215.
Hoffman, T.W. 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.
Hottel, H.C. 1954. “Radiant Heat Transmission”, in W.H. McAdams, Ed., Heat Transmission, 3rd Edition, New
York: McGraw–Hill.
Jaber, O., G.F. Naterer, and I. Dincer, 2010. “Convective Heat Transfer from Molten Salt Droplet in a Direct
Contact Heat Exchanger”, Heat and Mass Transfer, 46(8): 999–1012.
Kakaç, S. and A. Pramuanjaroenkij. 2009. “Review of Convective Heat Transfer Enhancement with Nanofluids”,
International Journal of Heat and Mass Transfer, 52: 3187–3196.
Kavanau, L. 1955. “Heat Transfer from Spheres to a Rarefied Gas in Subsonic Flow”, ASME Journal of Heat
Transfer, 77: 617–623.
Klinzing, G.E. 1981. Gas–Solid Transport, New York: McGraw–Hill.
Koo, J. and C. Kleinstreuer. 2005. “Laminar Nanofluid Flow in Micro Heat Sinks”, International Journal of
Heat and Mass Transfer, 48(13): 2652–2661.
Kreith, F., R.M. Manglik, and M.S. Bohn. 2010. Principles of Heat Transfer, 7th Edition, Pacific Grove:
Brooks/Cole Thomson Learning.
Lynch, P., 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.
Lu, M.L., 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.
Martin, M.J. 2011. “Scaling of Transient Particle–Fluid Heat Transfer in Brownian Motion”, AIAA Journal of
Thermophysics and Heat Transfer, 25: 177–180.
Maude, A.D. and R.L. Whitmore. 1958. “A Generalized Theory of Sedimentation”, British Journal of Applied
Physics, 9: 477–482.
Maxwell, J.C. 1881. A Treatise on Electricity and Magnetism, 2nd Edition, Oxford: Clarendon Press.
Mehrotra, S.P. 1981. “Mathematical Modeling of Gas Atomization Process for Metal Powder Production”,
Powder Metallurgy International, 13(3): 132–135.
Myers, T.G. and D.W. Hammond, 1999. “Ice and Water Film Growth from Incoming Supercooled Droplets”,
International Journal of Heat and Mass Transfer, 42: 2233–2242.
Naterer, G.F. 2011. “Multiphase Transport Processes of Droplet Impact and Ice Accretion on Surfaces,” Cold
Regions Science and Technology, 65: 5–12.
Pak, B. and Y. Cho. 1998. “Hydrodynamic and Heat Transfer Study of Dispersed Fluids with Submicron
Metallic Oxide Particles”, Experimental Heat Transfer, 11(2): 151–170.
Poots, G. 1996. Ice and Snow Accretion on Structures, New York: John Wiley & Sons.
Pope, K., 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.
Putnam, S.A., D.G. Cahill, P.V. Braun, Z. Ge, and R.G. Shimmin. 2006. “Thermal Conductivity of Nanoparticle
Suspensions”, Journal of Applied Physics, 99(8): 084308.
Raithby, G.D. and E.R.G. Eckert, 1968. “The Effect of Turbulence Parameters and Support Position on the Heat
Transfer from Spheres”, International Journal of Heat and Mass Transfer, 11(8): 1233–1252.
Ranz, W.E. and W.R. Marshall, 1952. “Evaporation from Drops – I and II”, Chemical Engineering Progress,
48(4): 173–180.
Sahm, P.R. 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.
Udawattha, D.S., S. Dilan, and M. Narayana. 2018. “Development of a Model for Predicting the Effective
Thermal Conductivity of Nanofluids: A Reliable Approach for Nanofluids Containing Spherical
Nanoparticles”, Journal of Nanofluids, 7(1): 129–140.
Multiphase Flows with Droplets and Particles 309
Wang, X.Q. and A.S.M.S. Mujumdar. 2008. “A Review of Nanofluids – Part I. Theoretical and Numerical
Investigations”, Brazilian Journal of Chemical Engineering, 25(4): 613–630.
Wang, X., 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.
Wasp, E.J., J.P. Kenne, and R. Gandhi. 1977. “Solid–Liquid Flow, Slurry Pipeline Transportation”, Trans Tech
Publications, pp. 9–32, Zurich, Switzerland.
Whitaker, S. 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.
Xuan, Y. and Q. Li. 2002. “Investigation of Convective Heat Transfer and Flow Features of Nanofluids”,
Journal of Heat Transfer, 125: 151–155.
Xue, Q. and W.M. Xu. 2005. “A Model of Thermal Conductivity of Nanofluids with Interfacial Shells”,
Materials Chemistry and Physics, 90(2): 298–301.
Yang, Y., 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.
Yu, W. 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.
PROBLEMS
1. Impinging supercooled droplets on an overhead power transmission line freeze immedi-
ately upon impact on the surface. The wind speed is 8 m/s and the liquid water content of
droplets in the air is 0.9 g/m3. Estimate the equivalent ice thickness covering the cable after
1 hour of precipitation.
2. During spray cooling of an industrial component, forced convection occurs with a thermal
boundary layer and impinging droplets on a flat plate. The plate length is L and the tem-
peratures of the wall and ambient air are Tw and T∞, respectively (ΔT = Tw – T∞). The
freestream air velocity is V. In the freestream, the droplet and air temperatures are equal to
each other.
a. Explain the physical meaning of each term in the following representation of the wall
heat flux: qw q1 q2 q3
T 1
Where q1 k ; q2 WVcw Tw T ; q3 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 specific heat, respec-
tively. 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 ∆T so that only q1 and q2 contribute
appreciably to the wall cooling.
3. The quality of solidified layers obtained in a plasma spray deposition process is signifi-
cantly affected by solidification of the droplets impinging on a cold surface. Initially, 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 according to:
db a2
a1
dt B a3 b
310 Advanced Heat Transfer
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 thermophysi-
cal properties.
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 constant. 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 computed under these
conditions.
5. Supercooled droplets fall on high voltage electric cables. The droplets freeze immediately
on impact on the surface. The wind speed is 12 m/s. After two hours of this freezing pre-
cipitation, the ice thickness is 20 mm. Determine the water content of droplets in the air
stream.
6. 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 internally by electrical
resistive elements so that the wall temperature beneath the solidified layer on the plate var-
ies with time. The ambient temperature and convective heat transfer coefficient of the sur-
rounding airflow, above the solidified layer, are Ta and h, respectively. Find the variation of
plate temperature with time in terms of thermophysical properties, h, Ta, plate dimensions,
and the electrical parameters (current, I, and resistance, R).
7. An airplane wing has an internal heating system to prevent ice formation. The plane flies at
a maximum airspeed of 125 knots. Wings have a 1 m cord length and cross-sectional
perimeter of 2.5 m. If the plane flies through a cloud with a moisture content of 0.2 and
temperature of –30°C for an extended period, what is the minimum amount of heat output
for the heater per unit length of the wing to prevent icing of the surface?
8. Large individual hailstones form from ice particles carried up and down through the super-
cooled water of a thunderstorm by strong air currents. As the particle moves upward, one
layer is formed and then during the downward travel, another distinctive layer is formed. If
a hailstone was found to have a diameter of 3 cm and 10 total layers, how far would it have
traveled in total through the cloud and what would be the distance between the top and
bottom of travel? Assume the cloud has a moisture content of 0.5 g/m3 and the collection
efficiency is 1.
9. Ice particles with an initial temperature of −50°C enter a jet engine’s compressor at 300 km/
hr which has an average air temperature of 300°C. The ice particles are spherical with a
mass, diameter, and specific heat of 31 μg, 150 μm, and 1.8 kJ/kgK, respectively. Assume
the particles have a constant speed and size. Sublimation and vapor loss effects are neglected.
How long will it take for the particles to reach 0°C and for all of the ice to then melt?
10. Metal powders and components with fine dimensional tolerances, such as gears, are pro-
duced in a multistage melt particularization process. Atomization of the liquid alloy through
an impinging cool air jet separates individual liquid droplets. Subsequent solidification of
droplets creates a temperature fluctuation, ∆Tu. 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?
Multiphase Flows with Droplets and Particles 311
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 convective cooling coefficient, hmin, required to ensure that
the thermal fluctuation is avoided. Give your result in terms of ρ, hsl, air–melt tempera-
ture difference, and the solidification rate, dχl/dt.
11. 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 liquid metal with
an initially uniform temperature that is suddenly subjected to convective and radiative
cooling on its outer surface. The extent of partial solidification prior to impact on the sub-
strate 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 tem-
perature of the droplet, Ts.
12. 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 question, a closed-
form (analytic) solution is sought for the transient temperature distribution 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, ambient gas temperature, and thermo-
physical properties.
13. Molten steel is formed into solid 4 mm steel balls through the use of a 300 K water stream.
What relative velocity would ensure the balls are completely disconnected from each other
while also minimizing the heat transfer coefficient of the balls? Find the heat transfer coef-
ficient. Assume the surface tension of steel is 1.8 N/m.
14. During a melt particularization process, a molten metal stream is injected into nitrogen gas
at 120°C. Assume a Weber number of We = 3. If the metal stream disintegrates 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.
15. 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 temperature of T∞.
16. 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 sand-
stone 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 tempera-
ture is 100°C. What is the initial rate of heating (K/s) of the particles by the wall and air-
stream as they enter the furnace? Assume that the particles do not appreciably participate
in the radiation exchange with the furnace wall at 600°C.
17. In a particulate removal process, solid particles migrate at 4 m/s relative to an airstream at
500 K. Find the drag force of the gas on an individual particle when the particle diameter
is 0.2 mm.
18. 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 particles is ΔT0.
The particles are heated by convection and radiation, at a heat transfer coefficient of h, dur-
ing their trajectory over a specified distance, L. If the particle velocity, V, is assumed to
remain constant throughout the trajectory, then find the increase of temperature of a parti-
cle in terms of ΔT0, L, V, h, D (particle diameter) and thermophysical properties.
312 Advanced Heat Transfer
19. 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 transfer coefficient
when the superficial fluid velocity is 12 m/s.
20. A building has an activated carbon air filtration system with a revisor of carbon of 1 m
length and 0.5 m in diameter. Air at 17°C flows through the filter at a rate of 1.2 m3/s. The
carbon particles in the filter have a diameter of 3 mm, sphericity of 0.6, and void fraction
of 0.35. What pressure drop is expected within the filter?
21. Spherical-shaped particles of diameter 4 × 10−7 m move through water at 343 K with a
superficial velocity of 0.5 m/s in a single pass unit packed bed filter with a surface area of
8 m2 and depth of 0.6 m. Estimate the force on a particle and total surface area of particles
in the flow stream.
22. A packed bed with a 33% void fraction contains spherical-shaped particles that have a
diameter of 0.003 m and superficial velocity of 8 m/s. The particles flow through air at 400
K. Estimate the drag force on a particle cloud if the bed has a cross-sectional area of 10 m2
and a depth of 3 m.
23. 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 2T
c c pu c c pv 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 predicted by
u ; v
y x
where
1/ 2
U
Uvx f ; y
1/ 2
vx
The freestream velocity is U = Cxm, where the exponent, m, is related to the surface inclina-
tion 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.
24. 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 properties, D, Tw, and T0 (initial
particle temperature).
25. Radiative heat exchange involves a participating gas at 1,600 K containing finely dispersed
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.
26. A gaseous mixture containing pulverized coal particles at 1,200°C has a radiative absorp-
tion coefficient of κλ = 4 – exp(–λ/6) where λ refers to wavelength. A monochromatic 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.
Multiphase Flows with Droplets and Particles 313
uT vT 2T
cp cp k 2 q
x y y
7.1 INTRODUCTION
Solidification and melting occur in various engineering systems such as materials processing (cast-
ing solidification, extrusion, and injection molding), phase change materials (PCMs), and molten
salts for thermal energy storage, among others. Two methods are commonly used in the analysis
of solid–liquid phase change problems – either an interface tracking method or a mixture formula-
tion. In the interface tracking method, the moving phase interface is explicitly tracked by analytical
methods or a computational grid. 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 the 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 inter-
facial transport processes. As a result, further supplementary relations, typically in the form of alge-
braic equations, are added back into the formulation to resolve the interfacial transport processes.
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 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 liq-
uid alloy is cooled, it first solidifies at the liquidus temperature corresponding to the concentration of
constituent A in the mixture of A and B. The solid has a composition 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 liq-
uid is 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, num-
ber of noncompositional variables (normally 2; p and T) and number of degrees of freedom (or the
number of independent intensive variables), respectively.
Effective control of a moving solidification or melting process has importance in several materi-
als processing technologies, such as wire casting, continuous casting, and zone refining. Figure 7.2
illustrates various types of casting processes where control of the interface velocity is a critical fac-
tor in the resulting thermomechanical properties of the solidified material.
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 cast-
ing (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).
Zone refining is another important solidification technology for crystal growth and material puri-
fication (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 solu-
bilities in the solid and liquid phases, solute is transferred into the molten region at the phase inter-
face 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.
Another application is energy storage via PCMs (phase change materials). A PCM is a substance
with a high heat of fusion that absorbs or releases heat when then material changes from solid to
liquid and vice versa. Spherical PCM particles are used in a range of applications such as thermal
energy storage, cooling of foods, medical applications (transportation of blood), waste heat recov-
ery, spacecraft thermal systems, and thermal management of electronic systems. Small PCM pellets
may be added to drywall mixtures for subsequent use in walls and roofs of houses and buildings
as an innovative method of solar-based heating in houses and buildings. PCM microcapsules have
also been used in clothing fibers. The body generates heat and initiates melting of the microcapsules
during outdoor activities. Then the PCM releases heat when it freezes, thereby maintaining a con-
stant temperature while a person sits and relaxes. The PCMs provide a form of temperature regula-
tion. Thermal analysis and applications of PCMs have been presented by Lane (1996), Jackson and
Fisher (2016), and Darkwa et al. (2015).
In this chapter, phase change heat transfer in solidification and melting processes will be exam-
ined. Topics include the formulation of the governing equations, one-and two-dimensional prob-
lems in various geometries, and advanced solution methods based on integral, similarity, and
quasi-stationary models.
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.
A system is in thermodynamic equilibrium at its most stable state when it has the lowest Gibbs
free energy, i.e., dG = 0. Figure 7.3 (a) illustrates the Gibbs free energy for stable equilibrium and
metastable (apparently stable but capable of change if disturbed) states. Any change resulting in a
decrease of the Gibbs free energy is possible.
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)
318 Advanced Heat Transfer
Gs H s TSs (7.4)
where the subscripts l and s refer to liquid and solid, respectively. As illustrated in Figure 7.3(b),
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
G 0 H Tm S (7.6)
hsl
s f (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.
Solidification is first initiated by a nucleation process. Nucleation occurs when the probability of
atoms arranging themselves on a crystal lattice is high enough to form a solid crystal from the liquid.
Homogeneous nucleation refers to solidification of liquid initiated by undercooling alone, without
particle impurities in the liquid to assist in the crystal formation. On the other hand, heterogeneous
nucleation is a more common case that occurs when the walls of a container or particle impurities
pre-exist to assist in generating nucleation sites for the solidifying crystals. This chapter will focus
on heterogeneous nucleation as the onset of solidification processes.
To start the solidification process, the liquid must be undercooled to a temperature below the
freezing point. A small crystal forms in the liquid and grows to a minimum critical size that remains
stable. The probability of atoms arranging themselves on a solid crystal lattice becomes high enough
to form a solid crystal from the melt. The critical condition for the onset of homogeneous nucleation
occurs when the decrease in the Gibbs free energy of the crystal becomes (Kurz & Fisher, 1984):
G A gV (7.8)
FIGURE 7.3 Schematic of Gibbs free energy for (a) a system in equilibrium and (b) phase change.
Solidification and Melting 319
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, characterized 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 in a lattice structure from melting. It can be expressed in terms of the Gibbs–Thomson
coefficient, Гgt, as follows:
gt s f (7.9)
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 at the phase change temperature,
h
g h T s sf Tm T (7.10)
Tm
where ΔT = Tm – T is the undercooling level below the phase change temperature. Combining these
results for a characteristic spherical volume of radius ro which encompasses the nucleating crystal,
4
G 4 ro2 ro2 s f T (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.
The onset of nucleation occurs when the driving force for solidification 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 liquid.
d G
8 ro 4 ro2 s f T (7.12)
dr
2
Tr (7.13)
ro s f
This leads to very low undercooling, i.e., ΔTr = 250°C for liquid nickel before solidifying. In
practice, heterogeneous nucleation is more common than homogeneous nucleation, whereby the
formation of new phase nuclei occur in contact with heterogeneities in the generating phase on the
surface.
Another important concept in solidification processes involving multiple constituents is the
chemical potential. The chemical potential of species A in a multicomponent 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,
320 Advanced Heat Transfer
g A
A (7.14)
N A
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.
A typical process of crystal formation and movement of a solid–liquid interface is illustrated in
Figure 7.4. The structure of the phase interface can be characterized by the nature of the surface area
bonding at the interface. Figure 7.4 illustrates a simplified process of attachment of cube-shaped
crystals onto a lattice structure at an interface. The interface growth is determined by the probability
that a molecule will reach the interface and remains 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 7.4, 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 unstable 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 7.5). 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 7.5. 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 perturbation 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 radi-
ally 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 perturba-
tion 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 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 7.6.
Figure 7.6 depicts the variation of structures in terms of the interface velocity, V (mm/s), and liq-
uid temperature gradient at the phase interface, G = ∂T/∂x (K/mm). For low-temperature gradients,
cellular, dendritic, and equiaxed dendritic structures are observed. However at high liquid tempera-
ture gradients, planar, cellular, dendritic, and equiaxed dendritic structures are obtained. Transitions
between each type of interface structure usually 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 analysis of various interface structures was presented by Chalmers (1964)
and Sahm and Hansen (1984).
Unlike solidification, which typically requires undercooling for the onset of nucleation, melt-
ing processes do not require superheating. Phase transition from a solid to a liquid phase typi-
cally 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.
7.3 GOVERNING EQUATIONS
7.3.1 General Scalar Transport Equation
Define a general scalar quantity, ϕ, as the mass fraction-weighted sum of individual phase compo-
nents of ϕ, in the solid and liquid phases, as follows:
2
k 1
k k (7.15)
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
l l s s (7.16)
Consider the transport of the general scalar quantity for phase k, ϕk, in a differential control
volume comprising the two- phase mixture (see Figure 7.7). For multicomponent mixtures, a
phase equilibrium diagram will also be required. In Figure 7.7, a binary eutectic phase diagram is
illustrated including the liquidus and solidus lines, which are functions of temperature and solute
concentration.
The general conservation equation for ϕk may be written as
t
k kk x k kuk, jk x k jk, j k Sk (7.18)
j j
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 (7.18), the terms represent the transient accu-
mulation 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
FIGURE 7.7 (a) Multiphase control volume and (b) phase diagram.
Solidification and Melting 323
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 control 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 balances will also be developed in
upcoming sections.
t x j
u j 0 (7.19)
dn
sVs lVl s l (7.20)
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 distance 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 significant 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 (7.18) 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 momen-
tum equation:
p
t
ui x k kuk,iuk, j x x k x kuk,i Fk,i,b Fk,i,int (7.21)
j i j j
where i = 1, 2, 3 and j = 1, 2, 3. The body forces and momentum exchange due to inter-phase inter-
actions require additional supplementary relations. For example, a significant body force in metal-
lurgical solidification problems is the combined thermal and solutal buoyancy in the y-direction,
Fb g T T T0 g C C C0 (7.22)
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.
324 Advanced Heat Transfer
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 boundary.
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
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:
3
K K0 l
(7.24)
1 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 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 inter-phase interaction forces and single-phase
flow equations are obtained, again as expected. Other investigators have extended the Blake–Kozeny
model to account for inter-dendritic directionality effects (Naterer and Schneider, 1995).
7.3.3 Energy Equation
For the thermal energy equation during solid–liquid phase change, substitute ϕk = ek, jk,j = –kk
(∂T/∂xj) (Fourier’s law) and Sk = Ėk into Equation (7.18). Then adding both individual phase equa-
tions, the following mixture energy equation is obtained:
T
t
e
x j
k k uk , j ek k Ek
x j x j
(7.25)
where the ranges of subscripts are i = 1, 2, 3 and j = 1, 2, 3. It will be assumed that viscous dissipa-
tion 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
T0
(7.26)
Solidification and Melting 325
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 (7.26) can be written in terms of tem-
perature alone. Using a piecewise linear approximation,
ek er ,k cr ,k T Tr ,k (7.27)
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
l el ses dt kl
dn l
ks
dn s
lVl el sVs es (7.28)
where n, Vs, Vl, and dn/dt denote the normal to the interface, individual phase velocities normal to
the interface, and interface velocity, respectively. Alternatively, using Equation (7.20),
dT dT
kl ks sVs hsl sVi hsl (7.29)
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.
T /x j C /x
jk , j k k k c,k Dk c,k j (7.30)
T T
where μc,k is the chemical potential of constituent c in phase k. For a multicomponent mixture of N
components, c = 1, 2 … N. In tensor form, Equation (7.18) becomes:
k T c Dc Cc
t
s
x j
k k uk , j sk
x j T x j x j T
Ps,k
x j
(7.31)
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 variables as
follows:
N
Tdsk dek pd c 1
c,k dCc,k
(7.32)
where υ represents the specific volume. Differentiating the Gibbs equation for an incompressible
substance in each phase,
D k k sk D k k ek D k k Cc , k
N
T
Dt
Dt
c 1
c,k
Dt
(7.33)
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 pro-
duction rate, by comparing the differentiated Gibbs equation with the transport form of the second
law in Equation (7.30), it can be shown that the mixture entropy production rate in phase k can be
written as
k T c Dc Cc
Ps 2 (7.34)
T x j x j T T 2 x j x j
where Ф 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 second law of
thermodynamics.
At the solid–liquid interface, entropy is produced due to irreversibilities of heat and mass trans-
fer 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 k dT dn
l sl s ss dt s
Tl dn l Ts dn s
lVl sl sVs ss l Ps,i
dt
(7.35)
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 (7.20), the interfacial entropy production rate becomes:
s h k dT 1 1
Ps,i s f sf l (7.36)
l T lVi 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).
Solidification and Melting 327
Ts 2T
s c p, s ks 2s (7.37)
t x
Tl 2T
l c p ,l kl 2l (7.38)
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 (7.39)
Ts X ,t T f Tl X ,t (7.40)
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 T dX
kl ks s hsl (7.41)
x x dt
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 tem-
perature profiles are self-similar so a similarity solution can be constructed. Define a similarity
variable, η, as follows:
x
(7.42)
2 st
where αs is the thermal diffusivity of the solid. Using this similarity variable, all temperature pro-
files 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 (7.37) become derivatives in terms of η through the chain rule, thereby yielding:
d 2Ts dT
2 s 0 (7.43)
d d
Ts Tw C1erf (7.44)
where C1 is a constant of integration and the error function, erf (η), is given by
2
erf
2
e s ds (7.45)
0
Based on a similar procedure in the liquid region, Equation (7.38) is solved to yield:
Tl Ti C2 1 erf (7.46)
At the solid–liquid interface (x = X), this liquid temperature must balance the solid temperature
in Equation (7.44). Equating both temperatures, it can be observed that the numerators in the error
function arguments must be proportional to the square root of time, so as to eliminate the appear-
ance 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 2 st (7.47)
The constant β is obtained by substituting x = X and matching the above temperature profiles in
the solid and liquid phases, yielding:
Tw C1erf T f Ti C2 1 erf s / l
(7.48)
After substituting the derivatives of temperature with respect to x into Equation (7.41), 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 (7.44) through (7.46),
T f Tw x
Ts Tw erf
2 s t
(7.49)
erf
Solidification and Melting 329
Ti T f x
Tl Ti 1 (7.50)
1 erf s / l 2 l t
hsl
exp 2
kl s / l Ti T f exp s 2 / l
0
(7.51)
cs T f Tw erf
ks T f Tw 1 erf s / l
2
erf (7.52)
cs T f Tw
(7.53)
2hsl
2ks T f Tw t
X 2 Ste t (7.54)
s hsl
cs T f Tw
Ste (7.55)
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
TABLE 7.1
Values of β for Water Where ΔTi = Ti – Tf
(Tf – Tw) °C ΔT = 0°C ΔT = 1°C ΔT = 2°C ΔT = 3°C ΔT = 4°C
1.0 0.056 0.054 0.053 0.051 0.050
2.0 0.079 0.077 0.076 0.074 0.073
3.0 0.097 0.095 0.093 0.091 0.090
4.0 0.111 0.110 0.108 0.106 0.104
5.0 0.124 0.123 0.121 0.119 0.117
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 (7.52) was truncated after
one term, such as freezing or thawing in the ground, the approximate solution in Equation (7.54)
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.
7.4.2 Integral Solution
Consider an approximate integral analysis of the same problem of freezing of a liquid initially at a
temperature of Ti in a one-dimensional domain. Equations (7.37) and (7.38) can be integrated from
the wall to the solid–liquid interface at x = X, as well as to the edge of the penetration depth of ther-
mal waves into the liquid, δ, yielding:
X X
Ts 2Ts
0
t
dx s
x 2
0
dx
(7.56)
Tl 2Tl
X
t
dx l
x 2
X
dx
(7.57)
dtTs x,t dx T f
0
dt
s s
x X
s
x 0
0
(7.58)
d d dX T Tl
dt
Tl x,t dx Ti
X
dt
Tf
dt
l l
x
s
x X
0
(7.59)
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 quadrati-
cally with x, respectively, where x refers to the distance from the wall,
Ts a1 a2 x (7.60)
Tl b1 b2 x b3 x 2 (7.61)
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. The
unknown coefficients, from a1 to b3, can be determined from the interfacial and boundary condi-
tions, where Ts = Tw at x = 0, Ts = Tf = Tl at x = X, and Tl = Ti at x = δ. Finding these unknown coef-
ficients and substituting back into the temperature profiles,
x
Ts Tw T f Tw (7.62)
X
Solidification and Melting 331
2
xX xX
Tl T f 2 Ti T f Ti T f (7.63)
X X
T f 2Ti dX dX 2 l Ti T f dX
3 3 p 1 dt Ti p dt p 1 X T f dt 0 (7.64)
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 2 st (7.65)
2 l t (7.66)
where β and ψ are unknown constants. Substituting this functional form of the interface position into
Equation (7.64), it can be shown that:
1 1 dX s 2kl Ti T f
Ste 2 dt X 1 k T T p 1 (7.67)
s f w
Solving this equation subject to X = 0 at t = 0 yields the same functional form of the interface
position as Equation (7.65), which is a useful verification of the initial assumption of Equation
(7.65). The parameter β is given by
p22 4 p1 Ste
2
p
2
2
(7.68)
2 p1 2 p1
and
2kl s Ti T f Ste
p1 2 Ste 2 Ste (7.69)
ks l T f Tw
Once the parameter β is known, then the interface position, X, can be determined, followed by the
temperature profiles in both the liquid and solid regions.
This approximate integral solution agrees well with the exact solution using the β parameter in
Equation (7.51). 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
332 Advanced Heat Transfer
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.
d 2T U dT
0 (7.71)
dx 2 dx
where α is the thermal diffusivity. In the liquid region, α represents the liquid thermal diffusivity,
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 Uhsl (7.74)
dx
xm dx
xm
where hsl is the latent heat of fusion and the notations of xm+ and xm− refer to the liquid and solid sides
of the interface, respectively. The frozen temperature approximation is applicable since temperature
gradients at the phase interface are constant with respect to time.
Differentiating the temperature profiles and substituting the results into Equation (7.74),
T2 Tm Tm T1 h
fg (7.75)
1 exp U x2 xm /
1 exp U xm x1 / cp
Solving this equation for the unknown interface position, xm, leads to:
T2 T1 1 2 1 T2 T1 exp Pe
T T Ste Ste 1 exp Pe exp Pe T T 1 Ste 0 (7.76)
m 1 m 1
where
U xm x1
exp (7.77)
U x2 x1 c p Tm T1
Pe ; Ste (7.78)
hsl
Here Pe refers to the Peclet number. The root solution of Equation (7.76) 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.
2T 2T
0 (7.79)
x 2 y 2
2T 2T
f T 0 (7.80)
x 2 y 2
334 Advanced Heat Transfer
where f(T) is a nonlinear function of temperature and 0 ≤ ε ≤ 1. It can be shown that solutions of
Equation (7.80) may be obtained by adding small perturbations to the solution of Equation (7.79)
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 leading 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 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 2T
2 (7.82)
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 T T f hsl (7.83)
x X dt
k T f Tw
Xs (7.84)
h T T f
x X h T T f
x ; X ;t t (7.85)
Xs Xs hsl X s
T Tw c p T f Tw
; Ste (7.86)
T f Tw hsl
Rewriting Equation (7.82) and the boundary and interface conditions in terms of these dimen-
sionless variables,
2 dX
2
Ste (7.87)
x X dt
dX
1 (7.88)
dt x X
0,t 0; X ,t 1
(7.89)
Using the perturbation solution method, the temperature and nondimensional interface velocity
are expanded in the form of a power series about the perturbation parameter, Ste, as follows:
0 Ste 1 Ste2 2
(7.90)
dX dX 0 dX dX X X X
Ste 1 Ste2 2 0 Ste 1 Ste2 2 1 (7.91)
dt dt dt dt x x x
2 0 21 2 2
2
2 Ste 2 Ste 2
Ste 0 Ste 1 Ste2 2
x x x X X X
(7.92)
1 0 Ste 1 Ste2 2
x x x
Setting the terms with the same powers of Ste equal to each other, the following sequence of
equations is obtained:
2 0
0 (7.93)
x 2
21 0 0
1 (7.94)
x 2 X dx
2 2 1 0
1 0 1 (7.95)
x 2 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
(7.93) and its boundary conditions are solved to give θ0. Then, substituting the differentiated θ0 into
336 Advanced Heat Transfer
Equation (7.94), the next solution of θ1 is obtained. The differentiated θ0 and θ1 expressions are then
substituted into Equation (7.95) to yield the solution for θ2, and so forth.
Also, individual terms in the power series in Equation (7.91) involving temperature are equated
to corresponding terms in the power series of the interface position terms,
dX 0 0
1 (7.96)
dt x X
dX1 1
1 (7.97)
dt x X
x
0 (7.98)
X
X 1 2
1
6 X 3
x X 2 x
(7.99)
1 X 3 2 X x 5 x 3 X 2 19
2 X x (7.100)
12 X 3 X
2
10 3 5 6
t t0 Ste t1 Ste2 t2
(7.101)
The following results are obtained for the first few cases:
t0 X ln 1 X (7.102)
1
t1 t0 (7.103)
3
t2
1
90
3 X 2 2 X 2 ln 1 X
(7.104)
Once individual terms are substituted and assembled into Equation (7.101), a final resulting equa-
tion 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 perturbation solution and oth-
ers involving forced convection freezing are presented by Seeniraj and Bose (1982) and Lunardini
(1988).
Solidification and Melting 337
7.6.2 Quasi-Stationary Solution
Another useful and powerful tool is the quasi-stationary solution method. This approach assumes
that heat conducts throughout the solid rapidly in comparison to the time scale associated with
movement of the solid–liquid phase interface. As a result, the transient term can be neglected in
comparison to the spatial temperature derivatives in the governing heat equation. Transient effects
are retained through the movement of the phase interface in the interfacial heat balance.
Consider a quasi-stationary solution for freezing of a liquid due to convective cooling at a
freestream temperature of T∞ from one side of the liquid (see Figure 7.11). For example, freezing of
a lake starts along the top surface of the lake and proceeds downward into the water. There are two
key thermal resistances: (i) a conduction resistance in the solid, which increases with time due to the
growing ice thickness; and (ii) a convective resistance which remains constant unless the convec-
tion coefficient, h, or T∞ vary with time. As time elapses, the relative significance of the convection
resistance diminishes relative to conduction through the solid. Since heat conducts through the ice
layer much more rapidly than a time scale associated with movement of the freezing ice interface,
then the quasi-stationary approximation can be used.
Using the quasi-stationary approximation, the one-dimensional governing, boundary, and inter-
facial equations for the solid are given by
2Ts
0 (7.105)
x 2
Ts dX
k hsl (7.106)
x X dt
Ts
k
x
h Ts 0,t T (7.107)
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, Equation (7.106), but
not the heat equation in the solid, Equation (7.105).
Solving Equation (7.105) subject to Equations (7.106) and (7.107),
x k /h
Ts T T f T (7.108)
X k /h
dX k T f T
X k /h (7.109)
dt hsl
X t
2k
hsl
T f T t hk hk (7.110)
The temperature in the solid can then be obtained by substituting this result into Equation (7.108).
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 resis-
tance that remains constant; and a conduction resistance that 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 resistance 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, relative to sen-
sible cooling, increases the interface velocity, thereby allowing less time for thermal equilibrium in
the solid, compared to cases with low Stefan numbers. Further detailed analysis and results of freez-
ing of liquids by convective cooling were reported by London and Seban (1943).
FIGURE 7.12 (a) Surface temperature and (b) depth of solidification in a semi-infinite domain with a
convection boundary condition. (adapted from Kreith and Romie, 1955)
Solidification and Melting 339
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 convective
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 fro-
zen 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
2Ts
0 (7.111)
x 2
Ts
ks
x
h Ts 0,t T (7.112)
0
Ts Tl dX
ks kl hsl (7.113)
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 approximation, the second
term is assumed to be constant in Equation (7.113),
Tl
qw kl (7.114)
x X
T hT X /ks hx h
Ts f 1 k k T x (7.115)
1 hX /ks s s
Differentiating this expression with respect to x and substituting the result along with the constant
interfacial heat flux into Equation (7.113) yields:
dX h T f T q
w (7.116)
dt hsl 1 hX /ks hsl
Solving this equation subject to the initial condition, X(0) = 0, yields (Lunardini, 1988):
hX hqw h T f T h T f T qw
t ln (7.117)
ks ks hsl qw h T f T qw 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 tempera-
tures 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 (7.115).
Setting the left side of Equation (7.116) to zero and solving for X on the right side yields the fol-
lowing steady-state position of the solid–liquid interface:
340 Advanced Heat Transfer
FIGURE 7.13 Depth of solidification for (a) a constant and (b) sinusoidal ambient temperature. (adapted
from Foss and Fan, 1972)
T T ks
X max ks f (7.118)
qw 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.
7.6.4 Multicomponent Systems
Unlike pure materials, phase change occurs over a range of temperatures for multicomponent mix-
tures between the solidus temperature, Tsol, and liquidus temperature, Tliq (see Figure 7.1). Consider
solidification in the one-dimensional x-direction of a binary mixture, initially at a temperature of Ti
and suddenly exposed to a wall temperature of Tw. The positions of the resulting liquidus and solidus
interfaces are x = Xsol and x = Xliq, respectively.
Define the solid fraction, χs, to vary linearly between the solidus and liquidus interface,
x Xliq
s e (7.119)
X sol Xliq
where Xsol < x < Xliq and χe refers to the solid fraction at the eutectic composition. The liquid frac-
tion, χl, is 100% minus the solid fraction. The linear approximation assumes that the liquid 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
(7.37) and (7.38). However, an additional heat equation is required in the two-phase (mushy) region:
Tsl 2Tsl df
sl csl ksl sl hsl s (7.120)
t x 2 dt
Solidification and Melting 341
where the subscript sl refers to the two-phase region (i.e., solid and liquid co-existing simultane-
ously in equilibrium). This heat equation will be solved subject to the interfacial conditions of Ts =
Tsol = Tsl at x = Xsol and Tsl = Tliq = Tl at x = Xliq (see Figure 7.14).
Also, the interfacial heat balance, Equation (7.41), is modified to the following two heat balances
at the solidus and liquidus interfaces, respectively:
Ts dX T
ks s hsl 1 e sol ksl sl (7.121)
x dt x
Tsl T
ksl kl l (7.122)
x x
The latent heat absorbed at the liquidus interface is modeled within the two-phase region by
Equation (7.120) and thus does not appear in the interfacial heat balance in Equation (7.122).
Unlike a pure material with a single-phase interface, a binary mixture requires tracking of both
the solidus and liquidus interfaces. Extending Equation (7.65) to two interfaces,
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 s t (7.125)
Tsol Tw erf sol
Tsl Tw
1
hsl e sol x / 2 s t T Tsol hsl e /cl
liq
erf x / 2 l t erf sol s / l
(7.126)
Tsol Tw cl Tsol Tw liq sol Tsol Tw
erf / erf
liq s l
sol s / l
Tl Ti
1 erf x / 2 l t (7.127)
Tliq Ti 1 erf liq s / l
The parameters βsol and βliq in are given implicitly by
Tliq Ti
kl m / l exp liq
2
s / l Tliq
Tsol exp liq
2
m / s hsl e /cm
hsl e m / l (7.128)
km 1 erf liq s / l
erf liq s / l erf sol s / l 2cm liq sol
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 composed of more than one
solid phase. These phases can exhibit a range of geometrical arrangements.
7.7 CYLINDRICAL GEOMETRY
Solidification and melting problems in cylindrical geometries occur in various scientific and engi-
neering systems. In this section, solid–liquid phase changes problems in cylindrical geometries are
examined.
1 T 1 T
r (7.129)
t r r r
T dR
k hsl (7.130)
r R dt
r2
(7.131)
4 t
d 2T dT
1 0 (7.132)
d 2
d
T Ti CE1 (7.133)
where the constant of integration, C, is determined from boundary conditions. The exponential inte-
gral, E1(ξ), is defined by
e t dt
E1
t
(7.134)
Using the expected functional form of the moving phase interface as determined previously in
Equation (7.47),
R 2 t (7.135)
Differentiating Equations (7.135) and (7.133), and substituting the results into the interfacial heat
balance, Equation (7.130), leads to:
2hsl E1 2 exp 2 c p T f Ti 0
(7.136)
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 rais-
ing the temperature of the subcooled liquid to Tf. This heat transfer into the subcooled liquid sustains
the movement of the phase interface into the remainder of the subcooled liquid.
For one-dimensional heat condition in the solid in the radial direction outside the tube (r > ro),
1 T 1 T
r (7.137)
t r r r
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 L (7.138)
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 (7.139)
L r R
dt
rTdr RT f
ro
dt
R
r
R
ro
r
ro
(7.140)
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
(7.141)
T f 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 (7.140) and solving for the interface position, R
(Lunardini, 1988),
n
2 n ln R
1 R2 ln R 1 R2 1 t
Ste 2
R 1 (7.142)
4 n 1
nn ! 2 4 4
where R* = R/ro and t Ste t /ro2 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 equa-
tions are again given by Equations (7.137) and (7.138) but the interface moves in the negative r-
direction (inward) and the temperature gradient in Equation (7.138) becomes a negative 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.
Solidification and Melting 345
k T
R ro t (7.143)
hsl r R
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 resulting solidifica-
tion time in terms of the interfacial temperature gradient and thermophysical properties,
hsl ro
tf (7.144)
k dT /dr R
As expected, the solidification time increases for materials with a larger latent heat of fusion or
lower thermal conductivity.
At small Stefan numbers, the quasi-stationary approximation may be used to neglect the tran-
sient term in Equation (7.137). Solving the reduced heat conduction equation subject to the
boundary conditions,
Tf Tw r
Ts Tw ln (7.145)
ln R/ro ro
Substituting this differentiated temperature profile into Equation (7.143) yields an explicit
expression for the interface position in terms of the thermophysical properties. Then the time
required for complete solidification can be determined. The expression for the interface position
is equated to R = ro and then the time required for the interface to reach the center of the tube
becomes:
hsl ro2
tf (7.146)
4k Tf Tw
It can be observed that this result is similar to Equation (7.144) but with a specific expression
for the temperature gradient based on the quasi-stationary approximation of the heat conduction
equation.
q r2
Ts Tm
4 ks
Ei Ei
4
s t
2
(7.147)
Ti Tm r2
Tl Ti Ei (7.148)
Ei 2 s / l
4 l t
Ei x
t
(7.149)
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:
kl Ti Tm 2 q 2
e s /l
e 2 s hsl (7.150)
Ei s / l
2
4
Tl
q 2 rkl (7.151)
r
The expected functional form of the moving phase interface, based on earlier examples, is given
by
R t 2 s t (7.152)
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 bat-
tery 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.
Using a quasi-stationary approximation, the reduced form of the one-dimensional governing heat
equation is given by
d dT
r
dr dr 0 (7.153)
subject to Ts = Tf = Tl at the phase interface (r = R); Ts = Tw at the wall (r = ro); and Tl = Ti (at r = δ).
Solving Equation (7.153) in the solid and liquid phases subject to the boundary and interfacial
conditions,
ln r /ro
Ts Tw
ln R /ro
T f Tw (7.154)
ln r /
Tl Ti Tf (7.155)
ln R /
where the subscripts s, l, f, w, and i refer to solid, liquid, fusion (phase change temperature), wall,
and initial, respectively.
The total heat flow through the wall, per unit length of pipe, consists of two components – sen-
sible cooling due to temperature changes between the liquid and wall temperatures; and latent heat
released at the interface to freeze the material.
R
Also, the rate at which heat is released by the freezing front balances the rate of heat loss through
the wall,
dTs
ks 2 ro
Q tot (7.157)
dr ro
The differentiated temperature from Equation (7.154) is substituted into Equation (7.157). Also,
using the above heat balances, the following implicit solution can be obtained for the interface posi-
tion (Lunardini, 1988),
R 2 R R R 2 R R 4 Ste
2
Ste 1 ln ln Ei 2 ln M 2 ln 1 2 t (7.158)
ro ro ro ro ro ro ro
where
c T f Ti 2 R 2 2 R 2 ln /R
M 1 Ste l T T 1 (7.159)
cs w f 2 R 2 ln /R
Then the resulting temperature distributions can be determined from Equations (7.154) and
(7.155). The current analysis was based on the quasi-steady approximation in Equation (7.153)
for low Stefan numbers. Better accuracy at higher Stefan numbers can be obtained through a heat
348 Advanced Heat Transfer
balance integral solution method but at the cost of additional complexity of equations that may no
longer sallow a closed-form solution.
7.8 SPHERICAL GEOMETRY
Another common geometrical configuration is a spherical system. Consider freezing of a subcooled
spherical droplet initially below the phase change temperature, Tf. A nucleation site forms at the
center of the droplet (r = 0) after which freezing proceeds outwards in the r-direction. Also consider
early stages of time in the freezing process where the effects of the outer surface of the droplet
have not propagated inwards to influence the freezing process from the initial nucleation site. This
assumption is analogous to an outward moving freezing front in a semi-infinite spherical domain
using the frozen temperature approximation.
The one-dimensional heat equation in spherical coordinates in the radial direction (see Appendix
C) is given by
1 T 1 2 T
r (7.160)
t r 2 r r
Define the following change of variables, similarly as Equation (7.131) in cylindrical coordinates,
r
(7.161)
2 t
d 2T 1 dT
2 0 (7.162)
d 2
d
T A BF (7.163)
where A and B are constants of integration, determined from the boundary and initial conditions.
Also, F(ξ) is the spherical function defined by
F
1
2
exp 2
2
1 exp
(7.164)
Specific forms of Equation (7.163) for the temperature field can be obtained once appropriate
boundary and initial conditions are applied.
Assume the center nucleation site and solid phase remain at Tf. The interfacial and initial condi-
tions are given by T(R, t) = Tf; T(r, 0) = Ti and
T dR
kl hsl (7.165)
r R dt
The constants of integration in Equation (7.163), A and B, can be determined based on the initial
and interfacial temperatures, Ti and Tf, respectively, thereby yielding the temperature profile.
Solidification and Melting 349
Also, using the expected functional form of the interface position as obtained previously in
Equation (7.47),
R 2 t (7.166)
The coefficient β can be determined by substituting the interface position and temperature into
the interfacial heat balance, Equation (7.166), yielding:
F
T Ti
F
T f Ti (7.167)
where
2
1
2 exp 2 exp 2 erf Ste
(7.168)
REFERENCES
Carslaw, H.W. and J.C. Jaeger. 1959. Conduction of Heat in Solids, 2nd Edition, Oxford: Clarendon Press.
Chalmers, B.. 1964. Principles of Solidification, New York: John Wiley & Sons.
Cho, S.H. and J.E. Sunderland. 1969. “Heat Conduction Problems with Melting or Freezing”, ASME Journal
of Heat Transfer 91: 421–426.
Darkwa, J., 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.
Duan, X. 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.
Foss, S.D. 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.
Jackson, G.R. 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.
Kreith, F. 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”, Proceedings
of the Physical Society B, 68: 277–291.
Kurz, W. and D.J. Fisher. 1984. Fundamentals of Solidification, Switzerland: Trans Tech Publications.
Lane, G.A. 1996. Solar Heat Storage: Latent Heat Materials, Boca Raton: CRC Press/Taylor & Francis Group.
London, A.L. and R.A. Seban. 1943. “Rate of Ice Formation”, Transactions of ASME, 65: 771–779.
Lunardini, V.J., 1988. “Heat Conduction with Freezing or Thawing”, CRREL Monograph 88–1, U.S. Army
Corps of Engineers.
Naterer, G.F. and G.E. Schneider. 1995. “PHASES Model of Binary Constituent Solid–Liquid Phase Transition,
Part 2: Applications”, Numerical Heat Transfer B, 28: 127–137.
Niyama, E., 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.
Ozisik, M.N., 1993. Heat Conduction, 2nd Edition, New York: John Wiley & Sons.
Pardo, E. 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.
Pfann, W.G.. 1958. Zone Melting, New York: John Wiley & Sons.
Sahm, P.R. 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.
350 Advanced Heat Transfer
Seeniraj, R.V. and T.K. Bose. 1982. “Planar Solidification of a Warm Flowing Liquid under Different Boundary
Conditions”, Warme Stoffubertragung 16: 105–111.
Stefan, J.. 1891. “Uber die Theorie des Eisbildung, Insbesonder uber die Eisbildung im Polarmere”, Annual
Review of Physical Chemistry, 42: 269–286.
Tien, R.H. and G.E. Geiger. 1967. “Heat Transfer Analysis of the Solidification of a Binary Eutectic System”,
ASME Journal of Heat Transfer 89: 230–234.
PROBLEMS
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 indepen-
dent thermodynamic variables in the pure phase and two-phase regions.
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 homoge-
neous nucleation in the solidifying liquid.
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 inter-
faces 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.
4. Find the governing equation for solute diffusion and redistribution at the phase interface
during solidification in a binary component mixture. Consider diffusion in the liquid and
solute redistribution at the phase interface according to the phase equilibrium diagram.
Solid diffusion and convection may be neglected. Express the result in terms of a uniform
interface velocity, U, and the liquid mass diffusivity.
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 tempera-
ture of ocean water during the voyage is 13°C. Consider only friction and heat transfer
along the flat base in the analysis. State the assumptions used in the solution.
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 Equation (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 probability of quantum state i and κ is
Boltzmann’s constant.
7. What differences are encountered between the solidification of metals as compared with
polymers?
8. In the liquid region of a binary component mixture ahead of an advancing solid–liquid inter-
face 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 solute concentration
decreases exponentially with distance ahead of the phase interface. 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?
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 equation of
mass conservation for a control volume occupied by two phases (solid and liquid) by per-
forming a mass balance in each phase individually and then summing over both phases. In
the derivation, use an averaged interfacial velocity, Vi, and a volume occupied by phase k,
Vk, based on the phase fraction, χk, multiplied by the total volume, V.
Solidification and Melting 351
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.
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 thermophysical properties, Tf, Tw and the
ratio of slopes of temperatures in the liquid and solid.
12. A large tank contains liquid at a temperature of Ti, above the phase change temperature, 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 transforma-
tion such as Equation (7.67) be used to find the interface velocity in terms of thermophysi-
cal properties and the above temperatures? Explain your response by reference to a
one-dimensional transient solution of the problem.
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 speci-
fied 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 mate-
rial. Express your answer in terms of β, L, qw, and the effective conductivity of the mate-
rial, k.
c. What values of β are required to maintain a unidirectional flow of heat at the left
boundary?
14. Consider a region of liquid initially at the phase change temperature (Tf) that suddenly
begins freezing outwards from a wall due to cooling at a uniform rate through the wall at x
= 0.
a. Using the quasi-stationary approximation for a one-dimensional semi-infinite domain,
estimate the position within the solid where the temperature reaches a specified fraction
of Tf (in Kelvin units) at a given time. Express your answer in terms of the wall cooling
rate, time, thermophysical properties and Tf.
b. For a wall cooling rate of 19 kW/m2, estimate where the temperature of tin reaches 96%
of Tf after 100 minutes.
c. What minimum cooling rate is required to ensure that the specified temperature in part
(b) can be obtained anywhere in the domain?
15. A layer of solid with a given thickness, W, is initially at a temperature of Tf (phase change
temperature). Suddenly both outer edges of the solid are heated and held at a specified
temperature of Tw, where Tw > Tf. The outer parts of the solid are melted over time. The
melted solid is self-contained within the same region so that the boundary temperature of
Tw is applied subsequently at the outer edges of the liquid.
a. Using the one-dimensional quasi-stationary approximation, find the temperature distri-
bution in the liquid.
b. Estimate how much time is required to melt the entire solid layer.
16. Repeat the previous problem and compare the results for the following three specific mate-
rials: aluminum, lead, and tin. Use an initial solid width of 10 cm and a wall temperature
of 10°C above the fusion temperature for each material.
352 Advanced Heat Transfer
17. A long metal bar with a specified thickness, W, is initially at a temperature of Tf (phase
change temperature). The right boundary (x = W) is suddenly heated and maintained at a
temperature of Tw, where Tw > Tf. Other surfaces, except the left boundary at x = 0, are well
insulated such that a one-dimensional approximation may be adopted. What uniform heat
flux, qw, must be applied at the left boundary (x = 0) to melt the entire bar within a specified
amount of time, ts? Derive an expression for qw in terms of thermophysical properties, time
and temperatures. Use a quasi-stationary analysis and assume that liquid is retained within
the same rigid container after it melts.
18. A liquid is initially held at a uniform temperature of Ti, above the phase change tempera-
ture, Tf. Solidification begins when the surface is cooled to a wall temperature of Tf. In this
directional solidification process, the cooling process is controlled such that the tempera-
ture gradient at the liquid side of the phase interface is nearly constant. Using a one-
dimensional, quasi-stationary approximation, find expressions for the phase interface
position and temperature distribution in the solid.
19. A solid layer of thickness W is initially kept at a temperature of Ti, below the phase change
temperature, Tf, in a large rigid container. Suddenly, two external surfaces of the container
are heated and held at a temperature of Tw, where Tw > Tf, so that melting of the solid begins
from both sides. Other surfaces are insulated. The heat flux into the solid is controlled such
that it declines approximately with inverse proportionality to the interface position, X(t), so
the heat flux decreases when X increases.
a. Estimate the time required to melt all of the solid. Use a one-dimensional, quasi-
stationary approximation in the analysis.
b. What range of constants of proportionality are required to ensure that the solid is melted
entirely?
20. A tube is submerged in liquid at an initial temperature of Ti, above the phase change tem-
perature, 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 solidification 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 temperature distribu-
tion in the liquid. Express your answer in terms of Tf, R, thermophysical properties, and
the interface velocity.
b. At what position does the liquid temperature exceed Tf by 10°C?
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 tem-
peratures in the solid and liquid phases.
22. Inward phase change from a cylindrical surface occurs during freezing of water in an annu-
lar 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 uniform 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, thermophysical prop-
erties, outer radius ro, and the constant of proportionality.
Solidification and Melting 353
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 thermophysical properties,
pipe radius, and the interface velocity. Assume that the interface moves inward at a con-
stant 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).
24. A solid material at a temperature of Tf (phase change temperature) in a spherical container
is heated externally. Assume that the melting front moves inward at an approximately con-
stant 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 uniform 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 the interfacial heat
flux, thermophysical properties, and the initial (unmelted) sphere radius.
b. Determine the rate of change of wall temperature with time.
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 tem-
perature). 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.
26. By imposing a prescribed pressure gradient, a non-Newtonian slurry of a liquid–sand mix-
ture 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. Nondimensionalize these equations to derive the parameters affect-
ing the temperature and Nusselt number distributions. Assume that the thermophysical
properties, such as cp of the mixture, remain constant, and the inlet temperature is Ti.
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 temperature 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 analy-
sis 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:
2
x x
T a0 a1 a2
to the liquid fraction. Using a velocity of v = – c2 y into the phase interface, where c2 is a
constant, and Darcy’s law, estimate the pressure distribution in the liquid based on gravita-
tional effects only. Express your result for pressure in terms of y, po, c1, c2, L and thermo-
physical properties.
29. The inter-phase momentum exchange and porosity of a solid matrix in a solidifying mate-
rial have been characterized by the Blake–Kozeny equation. Derive this equation for a solid
matrix by treating the interdendritic flow as a viscous flow through a tube (Hagen–Poiseuille
flow) of a hydraulic diameter equivalent to the pores of a solid matrix. Assume that a con-
stant pressure gradient is experienced by the flow through the solid matrix and each fluid
element travels an equal distance through the porous matrix.
30. A compartment of PCM is used for thermal control of a satellite. The satellite generates
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 satel-
lite 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.
31. Consider the use of PCMs in a solar thermal power generation system. A solar concentrator
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 subsequent power generation.
Define U, Tc, Tm, and To as the total thermal conductance of the PCM heat exchanger, sur-
face temperature of the solar collector, PCM melt temperature 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 optimal phase change temperature, Tm, based on the method of
entropy generation minimization (yielding a maximum power output) of the power
station.
8 Chemically Reacting Flows
8.1 INTRODUCTION
Heat transfer is a major element of the design and analysis of chemical reaction engineering sys-
tems. 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 characteristic reaction rate, tem-
perature, 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 spontane-
ous 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 noncatalytic and
catalytic reactors. A distinguishing feature between homogeneous and heterogeneous systems is the
number of phases. A homogeneous reaction involves one phase only, while a heterogeneous reac-
tion involves at least two phases. Alternatively, a reaction is considered homogeneous if a chemical
change occurs across the entire volume of the fluid, whereas it is heterogeneous if the chemical
change forms within a restricted region such as the surface of a catalyst. Catalytic reactions have
foreign 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 noncatalytic while liquid phase
reactions are usually catalytic.
In chemical reaction engineering, two common types of reactors are batch and continuous reac-
tors. 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 continu-
ous reactors. It characterizes 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, respec-
tively. 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, and scatter radia-
tion 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). 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).
8.2 MIXTURE PROPERTIES
Consider a mixture involving multiple constituents in a chemically reacting system. Define mi
and Ni as the mass and number of moles, respectively, of constituent i in the mixture. Also define
the total mixture mass, mass fraction, total number of moles, and mole fraction, respectively, as
follows.
• 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 Mi (8.1)
m N M
i 1
i i (8.2)
The units of the molecular weight (or atomic weight) are kg/kmol. The molecular weights of ele-
ments are presented in Appendix H.
The mixture molecular weight is given by
n
y M
m
M i i (8.3)
N
i 1
i mi /m mi N Mi
(8.4)
yi N i /N N i m M
A molar analysis involves the number of moles of each component in a mixture, whereas a gravi-
metric 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
y=
i = (8.5)
V p
where Vi refers to the volume occupied by component i, or the volume occupied if the individual
component was isolated from the other components. Also, pi and p refer to the partial pressure of
Chemically Reacting Flows 357
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.
Dalton’s Law of Partial Pressures states that the sum of partial pressures of the individual non–
reacting gases is equal to the total pressure.
n
p p
i 1
i (8.6)
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 typically 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
c p Ac p, A F c p, F (8.7)
c p y A c p , A yF c p , F (8.8)
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 balance equations can
be solved in the regular manner as previous chapters in terms of these mixture properties.
8.3 REACTION RATES
A chemical compound is considered fully reacted when its chemical identity has completely
changed. In general, there are three ways for a species to lose its chemical identity:
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
358 Advanced Heat Transfer
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 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 concentration, for example, –rA = kCA, where
k is a reaction rate coefficient and CA is the concentration of species A in the mixture.
The reaction rate represents the speed of the reaction or how quickly a reactant is converted to a
product. Consider a typical reaction of the following form:
aA g bB s dD g eE s (8.9)
where the lower case letters (a, b, d, e) represent stoichiometric coefficients and the capital letters
represent the reactants (A, B) and products (D, E).
Assuming the stoichiometric amount of reactants are converted to products, without excess reac-
tants 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
1 dC A 1 dC B 1 dC D 1 dCE
r (8.10)
a dt b dt d dt e dt
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 stoichiometric 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 concentra-
tion of each reactant. It is typically expressed as follows:
r = kC AnC Bm (8.11)
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 pres-
sures of each constituent. The rate constant includes the effects of all parameters (except concentra-
tion) 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 stoi-
chiometric 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 mol-
ecules 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 that influence the rate of reaction. For example, the nature of the
reaction, such as the number of reacting constituents and physical state, affects the complexity of
the reaction. The reaction rate normally increases with concentration and pressure since the fre-
quency of collisions among reactants increases at higher concentrations and pressures. Also, tem-
perature increases the reaction rate due to more energetic colliding particles with higher activation
energies to initiate the chemical reactions. The presence of a catalyst increases the rate of reaction
as it provides an alternative pathway for the chemical reaction at a lower activation energy. As men-
tioned earlier, the activation energy is the minimum energy that must be available in the reactants
Chemically Reacting Flows 359
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 depen-
dence is described by the Arrhenius equation,
E
k A exp a (8.12)
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 characterizes 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, multiplied by the probability that any given collision results in
a reaction (exponential term in brackets).
The rate of reaction can be expressed in terms of the mass, volume, and/or number of moles con-
sumed or released in the reaction. For example, in the gas–solid reaction of Equation (8.9), 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,
dms dN D
dt
dt
dM D bM B (8.13)
where the subscript s refers to solid. The variables MB and MD represent the molar masses of con-
stituents 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
1 dN i moles of i formed
ri (8.14)
VB dt volume of B time
The volume of the reactant solid, VB, as a function of time, can be written as
MB
VB t N B (8.15)
B
where ρB and NB represent the density and molar quantity of constituent B. The molar quantity of B
is calculated by
dN B
N B N0 (8.16)
dt
where N0 represents the initial quantity of B in the reactor when the reaction is initiated, or alterna-
tively, 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 consumption of
solid reactant, dNB/dt, balances the rate of formation of product gas, dNC/dt. Then, combining the
previous two equations allows the volume of reactant solid to be determined by
M B m0, B dNC
VB (8.17)
B M B dt
360 Advanced Heat Transfer
This expression can be used to determine the reaction rate based on the volume of the reactant
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 inter–dependencies, for example, as presented
by Sinha and Reddy (2011).
dN A
FA0 FA Z A (8.18)
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 volume, V, the total molar generation
of species A is given by
Z A rAdV
V
(8.19)
For spatially uniform molar generation of species A, the molar generation rate becomes ZA = rAV.
8.4.2 Energy Balance
Consider a reaction system involving two reactant inflows, R1 and R2, and two product outflows, P1
and P2, through the control volume in Figure 8.2. Under steady-state conditions, the conservation
of mass can be written as
m R1 m R 2 m P1 m P 2 (8.20)
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 super-
script o),
h T ,P h fo h T ,P h o h fo h (8.22)
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 reac-
tion, 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 composition. Tables of enthalpies of formation and at varying
temperatures and pressures are shown in Appendix I.
A standard convention is that the enthalpy of every element in its natural state at STP is zero.
For example, the enthalpy of formation at STP is zero for O2 (gas), H2 (gas), and C (solid). Also, an
enthalpy scale is defined with reference to STP. For a compound designated by subscript j,
T,p
h j p,T h
dh
o
f,j j (8.23)
STP
where the first term on the right side is the chemical (formation) enthalpy and the second term is sen-
sible 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:
h j T h f , j c p, j T 298 (8.24)
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 8.3,
2 2
i 1
N R,i h fo h h o R ,i
N h
j 1
P, j
o
f h ho P, j
Q 0
(8.25)
or alternatively,
2 2
i 1
N R,i h fo h R ,i
N h
j 1
P, j
o
f h P, j
Q 0
(8.26)
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 balance 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 proce-
dure can be used.
1. Estimate the outlet temperature and then calculate the enthalpies of the products of the reac-
tion 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 only the enthalpy dif-
ferences 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 available if
the reactants tend to produce other products. For example, consider the following reaction:
1
Cs O2 g CO g (8.27)
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:
C s O2 g CO2 g (8.28)
1
CO s O2 g CO2 g (8.29)
2
with an enthalpy of formation of –283 MJ/kmol. Subtracting Equation (8.29) from the prior equa-
tion, together with their respective enthalpies of formation,
1
Cs O2 g CO g (8.30)
2
The enthalpy of formation of CO becomes –110 MJ/kmol. It should be noted that CO is not an
element in Equation (8.29). Thus –283 MJ/kmol does not represent the enthalpy of formation 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, s o , and specific heat, c po , of one
mole of substance at the standard state; enthalpy and entropy of formation, h fo and s fo , respectively;
and the enthalpy of combustion, ∆hco .
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 shown that the entropy
of component i in the mixture can be written in terms of the temperature, total pressure, and mole
fraction, as follows:
T p
si si,0 c p,i ln Ri ln p R ln yi (8.31)
T0 0
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 standard temperature and pres-
sure (STP; 298 K and 1 atm).
8.5.1 Batch Reactor
For a batch reactor, FA0 = 0 = FA, which is well mixed. Then Equation (8.18) yields:
dN A
= rAV (8.32)
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,
N Af
dN A
tf
N A0
rAV
(8.33)
364 Advanced Heat Transfer
FIGURE 8.4 Schematic of batch, stirred, plug flow, and pack bed reactors.
Rearranging this result yields the following CSTR volume necessary to reduce the molar flow
rate from FA0 to FA,
FA FA0
V (8.35)
rA
where the terms represent the molar inflow, molar outflow, and generation of species Α over the
volume ΔV. Rearranging this result and taking the limit as the volume becomes small, ΔV → 0,
FA | V V FA | V
lim rA (8.37)
V 0 V
which implies
dFA
= rA (8.38)
dV
Chemically Reacting Flows 365
V
FA 0
rA
(8.39)
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 constant
throughout the volume), for this case of a plug flow reactor, the reaction rate varies throughout the
volume. Therefore, the mole balance in Equation (8.18) under steady-state conditions becomes:
FA0 FA rAdV 0
(8.40)
Once the functional dependence of rA on species concentration, CA, is obtained throughout the
volume, the last term on the right side can be integrated over the volume, V.
N A = k0 (8.41)
where ṄA = dNA/dt is the molar accumulation rate of species A. A first-order reaction can be written
as
N A = k1C A (8.42)
Similar expressions are written for higher order reactions. In Equations (8.41) and (8.42), k0 and
k1 are the reaction rate constants. The units of k0 and k1 are kmol/s m3 and 1/s, respectively. The
chemical reaction may occur at a constant rate (e.g., Equation (8.41) is zero-order) or a rate that is
proportional to the local concentration (e.g., Equation (8.42) is first-order). If Ṅ̇A is positive, then
the reaction leads to the production of constituent A. On the other hand, if ṄA is negative, then the
reaction is characterized by a consumption of constituent A.
FIGURE 8.5 Schematic of (a) catalytic reactor and (b) concentration distribution in a gas film.
1
N Bz N Az (8.43)
2
Performing a steady-state mass balance over a differential element, Δz, the amount of species A
at plane z balances the amount crossing at z + Δz,
SN Az |z SN Az |z z 0 (8.44)
where S is the cross-sectional area of the element. Dividing by SΔz and taking the limit as Δz → 0,
dN Az
=0 (8.45)
dz
From Chapter 1, the diffusion flux of species A can be determined from Fick’s Law. It can be
shown (Bird et al., 2002) that the convective flux is proportional to NAz + NBz and the combined flux
of diffusion and convection parts can be expressed as
dC A
N Az DAB C A N Az N Bz (8.46)
dz
DAB dC A
N Az (8.47)
1 C A / 2 dz
Then by differentiating both sides with respect to z, combining with the prior mass balance, and
solving for CA, subject to the boundary conditions of CA(0) = CA0 and CA(δ) = 0,
1 CA /2
1 (8.48)
1 C A0 / 2
z / 1
Chemically Reacting Flows 367
2 DAB 1
N Az ln (8.49)
1 C A / 2
This result represents the local rate of reaction per unit area of catalyst surface. The process is
“diffusion controlled” because the conversion of A to B proceeds at a finite rate due to diffusion even
though the chemical reaction occurs instantaneously on the catalyst surface.
8.6.2 Homogeneous Reaction
In this section, a homogeneous chemical reaction is considered with absorption of a species A (gas)
by a liquid solution B according to the chemical reaction: A + B → AB. An example of this reaction
is a carbon capture process of absorption of CO2 by an aqueous soluton of NaOH. Figure 8.6 illus-
trates gas A dissolving in liquid B and diffusing into the liquid phase. Bird et al. (2002) reported how
this type of chemical reaction affects the rate of absorption and diffusion in homogeneous systems.
Similarly as the previous section, by performing a mass balance on species A for a control vol-
ume of thickness Δz,
where k1 is a first-order rate constant for the chemical decomposition of A. Dividing by SΔz and
letting Δz → 0,
dN Az
k1C A 0 (8.51)
dz
dC A
N Az DAB (8.52)
dz
Combining,
d 2C A
DAB k1C A 0 (8.53)
dz
2
C A cosh k1L /DAB 1 z /L (8.54)
C A0 cosh k1L2 /DAB
FIGURE 8.6 Schematic of diffusion of gas A into B with a homogeneous reaction in the liquid.
368 Advanced Heat Transfer
If the second boundary condition has a prescribed concentration rather than a zero flux condition
at z = L, then the solution will be modified. For example, consider a gas A that enters an agitated
tank with liquid B (see Figure 8.7). The dissolved gas A undergoes a first-order reaction and forms a
liquid film of thickness δ around the gas bubble. Liquid outside the film has a concentration of CAδ
which varies slowly in time and therefore may be assumed constant.
Assuming the gas is not significantly soluble in the liquid, and a quasi-steady concentration
distribution is formed around the bubble, then the problem formulation remains the same as the
previous case, except with a different boundary condition of CA = CAδ at z = δ. Solving the species
concentration equation subject to this boundary condition,
CA 1 C A Bz Bz
cosh coth B sinh (8.55)
C A0 sinh B C A0
where B2 = k1δ2/DAB. As shown in Figure 8.7, the concentration decreases linearly to CAδ at z = δ
without a chemical reaction, but more rapidly in the case of a homogeneous reaction in the gas film.
Using this result, the molar flux can be determined at the liquid–gas interface, or alternatively, the
total rate of absorption during the chemical reaction,
(N Az |z 0 B 1
cosh B (8.56)
C A0 DAB sinh B cosh B V / S B sinh B
where S and V refer to the bubble surface area and volume, respectively.
Typical results of the molar flux are illustrated in Figure 8.8. At a given value of V/Sδ, the figure
shows that the dimensionless absorption rate at the liquid–gas interface increases with B. As B
approaches zero, for very slow reactions, the liquid is nearly saturated with dissolved gas. At large
values of B, the reaction is rapid and the dimensionless molar flux increases rapidly with B. Between
these limiting cases at intermediate values of B, the bulk of the solution is nearly solute free but the
reaction is slow enough to have nominal effects on solute diffusion in the film.
4 r 2 N Ar | r 4 r r N Ar | r r 4 r 2 r RA 0
2
(8.57)
Chemically Reacting Flows 369
where 4πr2Δr∙RA represents a source term and molar rate of production of species A due to the
chemical reaction in the spherical shell.
Dividing by 4πΔr and letting Δr → 0 yields:
d 2
dr
r N Ar r 2 RA
(8.58)
Here the quantities are interpreted as averaged values since the porous catalyst is granular rather
than continuous. An effective diffusivity, DA, is used to characterize diffusion of species A through
Fick’s Law as follows:
dC A
N AR DA (8.59)
dr
Assume that the molar rate of production of species A can be expressed through a first-order reac-
tion on the catalytic surface of the particle,
RA k1aC A (8.60)
where a is the available catalytic surface per unit volume including solids and voids. Then combin-
ing the above equations and re-arranging,
370 Advanced Heat Transfer
1 d 2 dC A
DA r k1aC A (8.61)
r 2 dr dr
where CA(R) = CAR and CA is finite at r = 0. Solving the second-order differential equation subject to
the boundary conditions,
C A R sinh
k1a /DA r
(8.62)
C AR r sinh ka /D R
1 A
This molar flux indicates the rate of conversion of A to B in the catalyst particle. Further detailed
analysis of concentration distributions in gas–solid reactions has been presented by Bird et al. (2002).
A B; r k2C A (8.65)
It is assumed that only reactant A exists in the external flow. The reaction rate in the far-field bulk
fluid (beyond the outer edge of the boundary layer) is zero. The governing mass, momentum, energy,
and species concentration equations can be expressed as
u v
0 (8.66)
x y
u u p 2u 2u
u v 2 2 (8.67)
x y x x y
T T 2T 2T H1
c pu c pv k 2 2 k1C AC B2 (8.68)
x y x y A
Chemically Reacting Flows 371
C A C 2C A 2C A
u v A DA 2
k1C AC B2 (8.69)
x y x 2
y
C B C 2C B 2C B
u v B DB k1C AC B
2
(8.70)
x y x
2
y 2
subject to:
T H 2 C A C
y 0; u 0 v; kT k2C A ; DA DB B k2C A (8.71)
y A y y
y ; u ax; v 0; T T ; C A C ; C B 0 (8.72)
Here ΔH1, ΔH2, δA, and kT, denote the homogeneous and heterogeneous heats of reaction, stoi-
chiometric coefficient for the heterogeneous reaction of species A, and effective thermal conductiv-
ity due to the chemical reaction, respectively.
It can be shown that a similarity solution exists when the following nondimensional similarity
variables are defined,
a
x,y a xf ; y (8.73)
T T C C a
; A ; B ; y (8.74)
T C C
Assuming that the diffusion coefficients of reactants A and B are comparable, then the species
concentration equations can be combined and summed to unity, i.e., ϕ + Φ = 1. Then only one func-
tion (ϕ or Φ) needs to be calculated after which the other concentration can be obtained from the
summation relation.
Using these similarity variables, the reduced form of the governing equations becomes:
f ff f 2 1 0 (8.75)
1
f RH 1 0
2
(8.76)
Pr
1
f K 1 0
2
(8.77)
Sc
subject to:
; f 1; 0; 1 (8.79)
where
1 H1 C3
Pr ; Sc ; RH k1 (8.80)
DA c p A aT
372 Advanced Heat Transfer
FIGURE 8.10 Dimensionless (a) temperature and (b) species concentration (Pr = KT = Sc = RH = 1).
k1C3 k2 C H 2
K ; KT (8.81)
a kT T A a
Here Pr, Sc, RH, a, K and KT refer to the Prandtl and Schmidt numbers, homogeneous reaction
heat parameter, similarity constant, strength coefficient of the homogeneous reaction, and effective
thermal conductivity due to the homogeneous reaction, respectively.
Predicted results of the nondimensional temperature, θ, and species concentration, ϕ, at the wall,
η = 0, are illustrated in Figure 8.10. A numerical solution by a fourth-order shooting technique was
presented by Xu (2017). In Figure 8.10, both the temperature and concentration decrease monotoni-
cally as K increases. For a prescribed value of KS = 0.2, solutions exist over a wide range of K values.
The results indicate that multiple solutions may exist at certain values of KS and K.
In Figure 8.10, for a fixed value of K, the nondimensional profiles also decrease monotonically
with KS. Solutions with KS < 0 represent a purely homogeneous reaction such as autoignition of a
system where the autocatalytic reaction is strong enough to sustain itself. It would not require an
autocatalyst to be continuously supplied into the system. Note that predicted solutions with F > 1 are
non–physical since the concentrations are bounded by 0 < F < 1 so only those solutions with F < 1
are shown. Numerical solutions are possible for a limited range of parameters as multiple solutions
begin to appear if K becomes sufficiently large.
The chemical equation for the reaction of a hydrocarbon fuel in air can be written as
b b b
Ca H b 4.76 a 0.21O2 0.79N 2 aCO2 H 2O 3.76 a N 2 (8.83)
4 2 4
Chemically Reacting Flows 373
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 stoi-
chiometric 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. Otherwise, with incom-
plete 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 (8.83).
Consider the combustion of methane in air. Using a = 1 and b = 4 in Equation (8.83) 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 balance the total
mass of products.
The mass of reactants or products, m, can be determined by the number of moles, N, multiplied
by the molecular weight, M. Since air consists mainly of oxygen and nitrogen, the molecular weight
is based on the mole fraction-weighted sum of individual molecular weights,
2
M air y M y
i 1
i i O2 M O2 yN2 M N2
(8.85)
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 (8.83) to chemical
reactions with excess air,
b b
Ca H b 1 EA 4.76 a 0.21O2 0.79N 2 aCO2 H 2O
4 2
(8.87)
b b
1 EA 4.76 a 0.79N 2 EA 4.76 a 0.21O2
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 H bOc a O2 aCO2 H 2O l (8.88)
4 2 2
374 Advanced Heat Transfer
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, ∆hco , is the negative of the
enthalpy change for the reaction, which is given by
b o
hco ahof CO2 ,g h f H 2O,l hof Ca H b ,Oc (8.89)
2
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 nitro-
gen 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 calculate 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 chamber at Tad.
Thus all heat generated by the combustion reaction is transferred to thermal energy of the combus-
tion 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 non-reacting 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 stoichio-
metric 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 products is in a gaseous phase. Further
detailed analysis of combustion reactions is available in thermodynamic textbooks, for example,
Moran et al. (2014).
dT d dT
p
mc kg r 2 (8.91)
dr dr dr
where m, kg, and cp are the evaporation rate of the droplet, thermal conductivity, and specific heat of
the gas, respectively. The difference between the temperature at the droplet radius, T(rd), and flame
radius, T(rf), is designated by ΔT.
Then solving the energy equation between r = rd and r = rf yields:
4 kg ln 1 c p T /h fg
m (8.92)
c p 1/rd 1/rf
where hfg is the latent of vaporization of the fuel droplet. Alternatively, by defining a standoff dis-
tance between the flame and droplet surface, δ = rf – rd, the burning rate can be expressed as
8kg c p T
m D2; ln 1 (8.93)
8 cp h fg
Here λ is a linear proportionality called the burning rate constant and D is the droplet diameter.
Values of the standoff distance, δ, and burning rate constant were reported by Strehlow (1984).
Sample values are presented in Table 8.1.
The following Ranz–Marshall correlation may be adopted to predict how the burning rate
increases with a relative velocity between the droplet and oxidizer.
0 1 0.24 Re1r / 2 Sc1/ 3 (8.94)
The burning rate with no convection effects, and Reynolds number based on the relative velocity,
are designated by λ0 and Rer, respectively. For multiple droplets in a spray, combustion occurs via
flames around a group of moving droplets rather than individual droplets.
Smoot and Smith (1985) examined the process of combustion of coal particles. The particle con-
sists of volatiles, moisture, char, and ash. Unlike the previous fuel droplet analysis, the coal particle
combustion process involves an initial step with volatiles, with moisture driven off the surface, after
which the volatiles then contribute to the gas-phase flame.
8.8.3 Radiation Exchange
Radiative heat transfer significantly affects the combustion process. The Weighted–Sum of Gray
Gases Model (WSGGM) is commonly used to predict and analyze radiation exchange in combus-
tion processes (for example, see Ansys Fluent User’s Guide, 2006). WSGGM assumes that the total
TABLE 8.1
Burning Rate Constants for Fuel
Droplets in Air at 100 kPa and 20°C
Fuel λ × 107 m2/s
Benzene 9.9
Kerosene 9.6
Diesel Oil 7.9
Iso–octane 11.4
376 Advanced Heat Transfer
emissivity and absorptivity of a participating gas can be expressed as a sum of a gray gas emissivity
and weighted by a temperature-dependent factor. The weighting factors are written in terms of poly-
nomial functions of gas temperature. The model assumes that the total emissivity of a gas, ε, over a
distance, s, can be written as
M
c
i 0
,i T 1 e a ps i
(8.95)
where M is the number of gas constituents, cε,i are weighting factors for the i–th component, ai is
the absorption coefficient of the ith gray gas, p is the sum of partial pressures of all gas constituents,
and s is the path length. Typical values of cε,i and ai were reported by Coppalle and Vervisch (1983)
and Smith et al. (1982).
The absorption coefficient for i = 0 is assumed to be a0 = 0 to account for zones in the spectrum
between high absorption regions. The weighting factor for i = 0 is expressed as
M
c ,0 1 c
i 1
,i
(8.96)
c ,i b j 1
,i , j T
j 1
(8.97)
where the polynomial coefficients bε,i,j and values of ai are determined based on curve fits with
experimental data (Coppalle and Vervisch, 1983). Values of the coefficients for 600 ≤ T ≤ 2,400 K
and 0.001 ≤ ps ≤ 10.0 atm–m were reported by Smith et al. (1982), and for values for T > 2,400 K,
values were reported by Coppalle and Vervisch (1983).
If aips ≪ 1, Equation (8.95) reduces to:
M
c
i 0
,i ai ps
(8.98)
It can be assumed that α = ε, where α is the absorptivity of radiation from the wall. This assump-
tion is commonly used, except if the medium is optically thin or the wall temperature differs signifi-
cantly from the gas temperature.
The above model has assumed that the total (static) gas pressure, ptot, is 1 atm. If patm ≠ 1 atm,
for example in combustion at high temperatures, correction factors are introduced based on scaling
rules. For patm > 1.1 atm or patm < 0.9 atm, an absorption coefficient is rescaled as
ai → ai ptot
m
(8.99)
where m is an empirical coefficient that depends on the temperature and partial pressures of the gas
constituents.
With soot formation in combusting flows, the absorption coefficient also depends on the soot
concentration. The effective absorption coefficient of a mixture of soot and absorbing gas, am, can be
expressed as the sum of the absorption coefficients of pure gas (without soot), ag, and pure soot, as,
where ρs is the soot density (kg/m3), b1 = 1.232 m2/kg, and bT = 4.8 × 10–4 K–1. Further details of
empirical correlations and coefficients over a range of flow conditions were reported by Smith et al.
(1982).
which is followed by
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.
Another common example of a gas–liquid reaction system is steel production. Steel was first
produced in the mid-nineteenth century. Its application is a vital part of many industries such as
automobile manufacturing, machinery production, buildings, and others. The raw materials in steel
include reduced iron ore, scrap steel, and iron. Impurities in these materials are removed by oxidiz-
ing them with blasts of air or oxygen. Most of the impurities (such as phosphorus and sulfur) are
converted to their respective oxides, which are later combined with other waste materials to produce
slag. After all impurities are removed, extra elements are added in careful proportions to modify or
enhance the mechanical properties of the steel.
The gas–liquid reactions and removal of impurities during steel production are often performed
in a basic oxygen furnace (BOF). High–velocity jets of oxygen are blown onto the molten iron in
the BOF vessel. Then carbon and other impurities in the molten iron react with oxygen to form car-
bon monoxide, which escapes from the vessel, and 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 (8.103)
Si l O2 g SiO2 l s (8.104)
In the former reaction, dissolved carbon within the liquid is transported to the gas–liquid inter-
face and subsequently removed as carbon monoxide through its chemical reaction 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 overheating 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.
Multiphase reactions can be analyzed based on two well-known models: a shrinking core model
and a progressive conversion model. The progressive conversion model is used when the diffusion
of gaseous reactant into a particle or droplet is much faster than the chemical reaction. The solid or
liquid reactant is consumed nearly uniformly throughout the particle. In contrast, a shrinking core
model is used if the diffusion of gaseous reactant is much slower so that it restricts the reaction zone
to a thin layer that advances from the outer surface into the particle or droplet. The reaction condi-
tions are assumed to be isothermal with a constant particle size. A pseudo–steady-state approxima-
tion is commonly adopted. The following sections will examine both shrinking core and progressive
conversion models for the analysis of reacting gas–solid flows.
diffusion of gaseous reactant occurs slowly and restricts the reaction zone to a thin layer. The decom-
position of the particle can be analyzed by diffusion through the product layer. The total gas pressure
is assumed to be constant and the particle is assumed spherical.
Consider further the following general form of a gas–solid reaction:
A g bB s cC s dD g (8.105)
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 consumed nearly
uniformly throughout the particle. On the other hand, a shrinking core model can be used if the dif-
fusion of gaseous reactant, A, is much slower. As illustrated in Figure 8.13, this restricts the reaction
zone to a thin layer that moves inward into the particle over time during the reaction.
In the shrinking core model, several simplifications and assumptions are made to determine 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 condi-
tions, equimolar counter-diffusion of gases/reactants and products (opposite direction concentration
gradients leading to diffusion in counter directions), 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 (subscript 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 (see Figure 8.14).
1. Diffusion of gaseous reactant, A, through a gas film surrounding the solid particle.
rA 4 ri2 g C Ab C As (8.106)
4 rr
i c Ds
rA
ri rc
C As C Ai (8.107)
rA 4 rc2 sC Ai (8.108)
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 r r 1
rA i c C Ab (8.109)
4 ri g 4 rr 4 ri2 s
2
i c Ds
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 B
rc brA (8.110)
dt 3 M B
Substituting for rA and integrating Equation (8.110), then rewriting in terms of the solid conver-
sion rate,
ri
3 g
r2
6 Ds s
r
s c 1 3 1 s 2 1 s i 1 1 s
2 /3 1/ 3 bC Ab M B
B
t
(8.112)
This represents the conversion of solid with time, including the processes of the gas film, diffu-
sion through the product layer, and the chemical reaction. The time required for complete conver-
sion of the reactant particle to the product can be obtained by setting χs = 1, yielding:
r r2 r B
t i i i (8.113)
3
g 6 D s s bC Ab M B
Chemically Reacting Flows 381
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 reaction of cupric chloride
particles with steam to produce copper oxychloride solid and hydrogen chloride gas in a process of
thermochemical hydrogen production.
The previous reaction in Equation (8.105) involved a combination reaction with two reactants
(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:
bB s E g fF s (8.114)
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 written similarly to the previ-
ous analysis,
1
1 r r
rE i c CEi CEb (8.115)
4 ri g 4 rr
2
i c Ds
d 4 3 B
rc brE (8.116)
dt 3 M B
Substituting for rE in this equation, integrating, and following a similar procedure as the combi-
nation reaction, the following equation for the conversion of solid, χs, is obtained:
ri r2 K bM B
s i 1 3 1 s 2 1 s
2 /3
CEb
t (8.117)
3 g 6 Ds RgTb B
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 decomposition of solids for
isothermal systems without bulk diffusion of gases.
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
k1AC (8.120)
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 decomposition of the
pellet,
dL3
C
dt
k1 6L2 C
(8.121)
dL
2k1 (8.122)
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),
Chemically Reacting Flows 383
L0 = 2k1tf (8.123)
This result represents the initial side length of the cube-shaped pellet. This length characterizes
the size of a pellet that is completely decomposed by the chemical reaction with the CO2 gas over
a time duration of tf.
8.10 FLUIDIZED BEDS
8.10.1 Hydrodynamics
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 transported 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 transfer 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 fluid-
ized state occurs when a sufficiently high velocity of the fluid penetrates 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 fluidized 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
heterogeneous 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 solution meth-
ods are applicable to homogeneous fluidization.
A fluidized bed typically consists of a vertical cylinder or tank loaded with solid particles and sup-
plied with a gas through a perforated distributor plate at the bottom of the cylinder (see Figure 8.16).
Various engineering systems, such as coal gasification, polyethylene manufacturing, granulation,
combustion of fuels, disposal of organic, toxic, and biological wastes, and particulate processing,
use fluidized beds (Howard, 1983). A detailed understanding of chemically reacting gas–solid flows
is important for the effective design and operation of fluidized beds.
A wide range of flow conditions may occur in gas fluidization (Geldart, 1973). These conditions
vary from dilute to dense particle packing. At conditions with low incoming gas velocities, no signif-
icant particle motion occurs and the system resembles flow through a porous medium (packed bed).
As the incoming gas velocity increases, a condition is reached whereby the particle weight balances
the upward momentum of the gas (called particle fluidization). Further increases in the gas velocity
lead to bubble formation in the gas phase due to regions of low particle concentration. These bubbles
move upward and enhance the mixing processes. The bubbles grow and eventually fill the tube in a
slug flow as the gas velocity increases. Subsequently, gas slugs further separate the solid particles
into clusters with periodic and nonuniform mixing. At very high gas velocities, sufficient momen-
tum may drive the particle clusters up and out of the tank, with some resulting backflow down along
the walls. This case is called fast fluidization. As a result, more particles must be injected into the
cylinder to maintain a steady-state operation.
Consider the following general gas–solid reaction within a heterogeneous fluidized bed,
A g bB s cC s dD g (8.124)
A schematic of a fluidized bed and diffusion resistances are illustrated in Figure 8.17. 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 emulsion resistance, exter-
nal film resistance around a solid particle, and an inter-particle resistance.
Define the concentration efficiency for constituent A as follows:
C A,i C A, o
c, A (8.125)
C A, i C A, 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
c, A 1 exp (8.126)
where
K be b
NTU (8.127)
U o /H b
U o U mf
(8.128)
Uo
Here ζb, Kbe, Uo, Umf, β, and Hb refer to the bed void fraction, overall mass transfer coefficient
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:
c ,i c , b
xc , b (8.129)
c,i 1 c,b
where χc,i and χc,b represent the mass fraction of solid reactant, C, in the feed (inlet) and bed, respec-
tively. 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):
gU mf D p
33.7 0.0408 Ar 33.7
1/ 2
Remf (8.130)
g
gD p2 g p g
Ar (8.131)
g2
Here the subscripts g and p refer to the gas and particle, respectively. The Reynolds number at
the minimum fluidization velocity corresponds to the condition where all particles in the fixed bed
begin floating.
The pressure drop across the bed, Δpb, can be determined based on the overall particle weight
(Hewitt, Shires, and Polezhaev, 1997),
pb 1 mf p g gH b (8.132)
where ζmf is the porosity of the particulate (emulsion) phase at the state of minimum fluidization.
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 surface. As a result,
386 Advanced Heat Transfer
the free surface is often wavy and chaotic rather than smooth and horizontal. A large particle con-
centration 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 container
during pool boiling. The bubble rise velocity, Vb, can be approximated in terms of the bubble diam-
eter, Db, as follows (Hewitt et al., 1997):
Unlike bubbles that rise in pool boiling, upward gas flow in a fluidized bed keeps the particles
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 particle 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 fraction
occupied by the bubbles and the fluidized bed porosity. The ratio of the bed height, Hb, to the mini-
mum fluidization height, Hmf, can be expressed as
H b 1 mf (8.134)
H mf 1b
where the bed porosity, ζb, can be determined from the following Todes correlation,
0.21
Re p 0.02 Re2p
b mf 2
(8.135)
Remf 0.02 Remf
Since the gas velocity at the free surface may exceed the mean fluidization velocity, larger par-
ticles 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 individual particles that
later descend back down into the fluidized bed surface. Only small 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
TDH 1200 H mf 1p.1 (8.136)
Ar
This correlation is generally applicable for particle diameters between 0.075 and 2 mm.
Chemically Reacting Flows 387
The gas–to–particle heat transfer coefficient increases with Rep / ζb (as expected) since the inten-
sity 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 transfer occurs when
the reacting gas diffuses toward the surface of the active particle. Both diffusion and convection pro-
cesses contribute to mass transfer. A widely used correlation for estimating the mass transfer coeffi-
cient, hm, in a fluidized bed is the following La Nauze–Jung correlation (La Nauze and Jung, 1982):
1/ 2
Re
Shp 2 mf 0.69 p Sc 0.33 (8.139)
b
where Shp and Sc refer to the Sherwood and Schmidt numbers (based on the particle diameter) for
mass transfer, respectively. This correlation incorporates the movement of clusters of particles car-
rying fresh gas from the fixed bed toward the active particles.
Although the above correlations are useful, their limitations should not be overlooked, consider-
ing the complexity of the transport processes during the heterogeneous fluidization. The processes
involve complicated interactions involving chemical reactions, turbulence, radiation, phase change,
and others. As a result, experimental data is often scattered and the accuracy of the previous correla-
tions generally lies within a broad range of ±50%. In addition to the flow complexities, this wide
scatter arises because the correlations usually do not include contributing parameters such as bubble
sizes, tube arrangements, geometrical factors, among others. Further detailed analysis of transport
processes in fluidized bed systems was presented by Doraiswamy and Majumdar (1989).
C A, o
xg 1 (8.140)
C A,i
388 Advanced Heat Transfer
n
C
p A, e (8.141)
C A,i
where the subscripts i, e and o refer to inlet, emulsion phase, and outlet, respectively. The gas con-
version can be determined by
xg 1 1p/ n c
(8.142)
where n denotes the order of the reaction and ηc is the concentration efficiency in Equation (8.125).
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
balance of the number of particles of a particular state. Monte Carlo methods, discretization meth-
ods, and moment methods have been used to solve population balance equations. Further detailed
analysis of population balances was presented by Ramkrishna (2000).
It can be shown that a population balance over the reactor yields the following distribution of
solid conversion within the fluidized bed,
1 1 1 c ,i xc xc
pb xc exp (8.143)
Das F xc 1 c,i xc,i
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 chemical 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.
F xc p Fp xc (8.147)
Chemically Reacting Flows 389
In order to determine the distributions of Θ(xc) and F(xc) in Equations (8.144) and (8.147), 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 8.2 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 tempera-
ture in the emulsion. It is calculated by
Mp
kr ,e bkiC An ,e (8.148)
p
where ki is the kinetic coefficient based on the particle volume, determined at the inlet conditions,
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.
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
W R s p s ds
xc 0
b b (8.149)
k
R xc c , i r , e F xc (8.150)
1 c ,i xc
Das
rc,bed 1 Fi (8.151)
Also, performing a mass balance of nonreacted material of solid particles, S, and combining it
with the expression for the conversion distribution of solids, pb(xc), it can be shown that:
where
s
1
1 s
f1 xc,i ,
F s exp
xc 0
ds
(8.153)
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
390 Advanced Heat Transfer
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 (8.154)
bU o Ac,bedC A,i M p
(8.155)
Fi
where Ac,bed is the bed cross-sectional area. Taking into account the average conversion of solid par-
ticles within the bed, xc,b, and solving the distribution function, pb(xc), over the bed, leads to:
f1 xc,i ,
xc , b 1 (8.156)
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 (8.157)
1/ c,i xc,i
Das r
1 c,bed (8.158)
Fi
The overall rate of reaction, rc,bed, is determined from Equations (8.143), (8.149), and (8.150),
yielding:
1
rc,bed
W R s p s ds
b b (8.159)
xc ,i
c ,i
rc,bed Fi f2 xc,i , (8.160)
1 c ,i xc ,i
Chemically Reacting Flows 391
where
s
1
f 2 x c , i ,
exp
ds
(8.161)
xc ,i
Furthermore, by combining this result with Equation (8.160), and substituting χc,i = 1 and xc,i = 0,
the Damköhler number can be expressed as a function of λ as follows:
s
1
Das
1 f2 1 exp
0
ds
(8.162)
From Table 8.2, the kinetic functions for the progressive conversion model are given by
Fp xc 1 xc (8.163)
xc ln 1 xc (8.164)
ln 1 xc
1
0
f2 exp
ds
1
(8.165)
Hence, the Damköhler number is obtained by substituting this result into Equation (8.162),
Das 1
(8.166)
1
The overall rate of the reaction can be determined by combining this result with Equation (8.158),
yielding:
rc,bed Fi (8.167)
1
In summary, the following solution procedure is used to evaluate the conversion of solid particles
and gas based on the progressive conversion model.
TABLE 8.2
Limiting Cases of Two Kinetic Models
Progressive Conversion Model Shrinking Core Model
Fp(xc) 1 – xc (1 – xc)2/3
392 Advanced Heat Transfer
4. Check the convergence status by calculating whether xg,new – xg,old is less than a specified
tolerance.
5. Return to step 2 and repeat these steps until convergence is achieved.
6. Calculate xc = α xg.
Through this trial–and–error procedure, the conversion of reactants and spatial distribution 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 decomposition follows this process,
the kinetic functions in Table 8.2 are used as follows:
Fp xc 1 xc
2 /3
(8.168)
xc 3 1 1 xc
1/ 3
(8.169)
3
1/ 3
f2 exp 1 1 s ds
(8.170)
0
Evaluating this integral and combining the result with Equations (8.162) and (8.145) yields:
2 2
2
3
rc,bed Fi 1 2 exp (8.171)
3 3 3
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 non-catalytic reaction of cupric
chloride particles with superheated steam in a fluidized bed are shown in Figure 8.18. The bench-scale
Chemically Reacting Flows 393
FIGURE 8.18 (a) Solid/gas conversion and (b) reaction rates for a non-catalytic reaction of cupric chloride
particles with superheated steam in a fluidized bed. (adapted from Haseli et al., 2008)
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 predic-
tions of two kinetic models (shrinking core model and progressive conversion model) are compared
against each other under identical process conditions. Figure 8.18 indicates that the predicted con-
version of particles and fluidizing steam is 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 con-
verted to products so there are no unreacted particles or gas at the exit. The bed effectiveness can be
represented by a ratio of the total number of moles of reactants that are converted to products to the
total reactant moles at the inlet stream,
bed
N
i
R ,i inlet N
i
R ,i outlet
(8.172)
i N R,i inlet
N p,o C p,o
1 xc , b (8.173)
N p ,i C p ,i
N g ,o C g ,o
1 xg (8.174)
N g ,i C g ,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
394 Advanced Heat Transfer
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 float-
ing, 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 particle motion in the outer regions. The reader is referred to other sources for
a more detailed analysis of heat transfer in fluidized bed reactors such as Grace, Knowlton, and
Avidan (2012), Chen (2003), and Kunii and Levenspiel (1991; Levenspiel 1998).
REFERENCES
Ansys Fluent User’s Guide. 2006. Fluent Inc., Lebanon, NH.
Bird, R.B., W.E. Stewart, and E.N. Lightfoot. 2002. Transport Phenomena, 2nd Edition, New York: John Wiley
and Sons.
Chen, J.C. 2003. “Heat Transfer”, Chapter 10, pp. 257–286, In: Yang, W.C., ed., Handbook of Fluidization and
Fluid Particle Systems, Boca Raton: CRC Press/Taylor & Francis Group.
Coppalle, A. and P. Vervisch. 1983. “The Total Emissivities of High–Temperature Flames”, Combustion and
Flame, 49: 101–108.
Crowe, C.T., J.D. Schwarzkopf, M. Sommerfeld, and Y. Tsuji. 2012. Multiphase Flows with Droplets and
Particles, 2nd Edition, Boca Raton, FL: CRC Press.
Daggupati, V.N., G.F. Naterer and I. Dincer. 2011. “Convective Heat Transfer and Solid Conversion of
Reacting Particles in a Copper(II) Chloride Fluidized Bed”, Chemical Engineering Science, 66:
460–468.
Doraiswamy, L.K. and A.S. Majumdar. 1989. Transport in Fluidized Particle Systems, Amsterdam: Elsevier
Science Publishers.
Fogler, H.S. 2016. Elements of Chemical Reaction Engineering, 5th Edition, Upper Saddle River, NJ: Prentice
Hall.
Geldart, D. 1973. “Types of Gas Fluidization”, Powder Technology, 7: 285–292.
Gelperin, H.I. and V.G. Einstein. 1971. “Heat Transfer in Fluidized Beds”, pp. 450–471, in Fluidization (J.F.
Davidson, D. Harrison, eds.), Cambridge: Academic Press.
Gómez–Barea, A., B. Leckner, D. Santana and P. Ollero. 2008. “Gas–Solid Conversion in Fluidized Bed
Reactors”, Chemical Engineering Journal, 141: 151–168.
Grace, J.R., T.M. Knowlton and A.A. Avidan (eds.), 2012. Circulating Fluidized Beds, Berlin: Springer Science
and Business Media.
Guthrie, R.I.L. 1993. Engineering in Process Metallurgy, Oxford: Oxford University Press.
Haseli, Y., 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.
Hedley, A.B., A.S.M. Nuruzzaman, and G.F. Martin, 1971. “Combustion of Single Droplets and Simplified
Spray Systems”, Journal of the Institute of Fuel, 44: 38.
Hewitt, G.F., G.L. Shires, and Y.V. Polezhaev, eds., 1997. International Encyclopedia of Heat and Mass
Transfer, Boca Raton: CRC Press/Taylor & Francis Group.
Howard, J.R. ed., 1983. Fluidized Beds Combustion and Applications. London: Applied Science Publishers.
Kunii, D. and O. Levenspiel. 1991. Fluidization Engineering, 2nd Edition, London: Butterworth–Heinemann.
La Nauze, R.D. and K. Jung. 1982. “The Kinetics of Combustion of Petroleum Coke Particles in a Fluidized
Bed Combustor”, 19th International Symposium on Combustion, Haifa, Israel, pp. 1087–1092.
Levenspiel, O. 1998. Engineering Flow and Heat Exchange, 2nd Edition, Plenum Chemical Engineering
Series, Berlin, Germany.
Levenspiel, O. 1999. Chemical Reaction Engineering, 3rd Edition, New York: John Wiley & Sons.
Moran, M.J., H.N. Shapiro, D.D. Boettner and M.B. Bailey. 2014. Fundamentals of Engineering
Thermodynamics, 8th Edition, New York: John Wiley & Sons.
Ramkrishna, D. 2000. Population Balances: Theory and Applications to Particulate Systems in Engineering,
Cambridge: Academic Press.
Sinha, K., 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.
Chemically Reacting Flows 395
Smith, T.F., Z.F. Shen, and J.N. Friedman. 1982. “Evaluation of Coefficients for the Weighted Sum of Gray
Gases Model”, ASME Journal of Heat Transfer, 104: 602–608.
Smoluchowski, M. 1916. “Drei Vorträge über Diffusion, Brownsche Molekularbewegung und Koagulation von
Kolloidteilchen” (in German), Zeitschrift für Physik, 17: 557–571.
Smoot, L.D. and P.J. Smith, 1985. Coal Combustion and Gasification, Plenum Press, New York.
Strehlow, R.A. 1984. Combustion Fundamentals, McGraw–Hill, New York.
Wen, C.Y. and Y.H. Yu. 1966. “A Generalized Method for Predicting the Minimum Fluidization Velocity”,
AIChE Journal, 12: 610–612.
Xu, H. 2017. “Homogeneous – Heterogeneous Reaction Model for Heat and Fluid Flow in the Stagnation
Region of a Plane Surface”, International Communications in Heat and Mass Transfer, 87: 112–117.
PROBLEMS
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 frac-
tion (Dalton’s law of additive pressures).
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.
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.
4. In a container, 3 moles of gas A and 2 moles of gas B are separated by a barrier. The tem-
perature and pressure on both sides are T1 and P1. The barrier is removed and the gases are
mixed. Find the entropy increase when the ideal gasses are mixed. Assume the gases are
ideal.
5. A mixture of 60% nitrogen and 40% propane (by volume) flows through a well-insulated
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, non–reacting 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.
6. For the previous question, find the specific entropy change of each gas component sepa-
rately (i.e., propane and nitrogen separately) between the inlet and outlet of the compres-
sor. Which gas component in the mixture undergoes a larger specific entropy change? Give
a physical explanation of this difference of computed specific entropy values.
7. A rigid tank is initially divided by a partition into two sides. The left side contains methane
(component A, mass of mA and NA moles) and the right side contains nitrogen (component
B, mass of mB and 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 mixing process. Verify that this entropy
change complies with the second law of thermodynamics.
8. A chemical reaction is represented by 2A + B → 3R. Find the order of reaction. Also, find
the order of reaction when the rate of reaction is increased by 4 times and when the con-
centration of reactants is increased by a factor of 1.5.
9. Explain how the adiabatic flame temperature, Tad, can be calculated for the combustion of
methane (CH4) at STP in air. Show the relevant chemical reaction and energy balance equa-
tions. Describe the steps required to calculate Tad.
10. 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 reactants and
products of combustion.
11. Perform an energy balance where the required heat input is supplied by a gas stream whose
inlet temperature is higher than the reactor temperature. The general form of the chemical
reaction is A (g) + b B (s) → c C (s) + d D (g). Perform molar and energy balances for this
396 Advanced Heat Transfer
system to determine molar flow rates and enthalpies of the products and unconverted
reactants.
12. Droplets of liquid propane (C3H8) are burned at atmospheric pressure with 150% excess air
in a combustion chamber. Assume the fuel burns completely with excess air under steady-
state conditions and the fuel inflow rate is ṁ. The fuel and air enter at STP and a tempera-
ture of TA, respectively. Find the heat released within the combustion chamber if the
products of combustion exit the chamber at a temperature of TP.
13. Significant variations can be observed between adiabatic flame temperatures obtained from
actual experiments, tabulated values, and calculations based on a theoretical analysis in this
chapter. Explain why these variations may occur and how more accurate calculations can
be carried out in the previous question.
14. A stoichiometric mixture of liquid hydrogen and oxygen react in a low-temperature com-
bustion chamber. Exhaust gases are released at a rate of 60 kg/s from the chamber. The
cylindrical reactor has a length of 60 cm and diameter of 50 cm. Find the reaction rates of
hydrogen and oxygen for the complete stoichiometric combustion reaction.
15. The average minimum daily requirement of food energy intake is about 1,800 kilocalories
(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 chemical reaction
of C6H12O6 + 6O2 → 6CO2 + 6H2O and an enthalpy of reaction of 2,816 kJ. Find the meta-
bolic rate of oxygen used for an 80 kg person.
16. An engineering student weighs 50 kg and eats around 4,000 kJ of carbohydrates per day.
The reaction in food digestion is expressed as follows.
Determine the student’s rate of metabolism in terms of moles of oxygen used per m3 of
body volume per second.
17. Large chemical processing units (called fluid catalytic crackers) in the petroleum industry
are used to convert long-chained hydrocarbons into shorter molecules. The process con-
verts the high-molecular-weight hydrocarbon fractions of petroleum crude oils into more
valuable hydrocarbons including gasoline, olefinic gases, and other products. The feed-
stock is a portion of the crude oil that has an initial boiling point of 340°C or higher and an
average molecular weight from about 200 to 600. Consider 70% of a feedstock of C18H38
which is heated and cracked by bringing it into contact with a catalyst. The chemical reac-
tion is given by C18H38 + 55/2 O2 → 18CO2 + 19H2O. The reactor contains 40 tons of
porous catalyst with a density of 900 kg/m3. The catalyst breaks the long-chain molecule
hydrocarbons into much shorter molecules which are collected as a vapor. The reactor is
fed by 5,000 m3/day of crude oil with a density of 920 kg/m3. Find the rate of reaction of
the hydrocarbon feedstock.
18. In a first-order isothermal liquid polymerization, monomer A (3,000 kg) and monomer
B (6,000 kg) are copolymerized in a 30 m3 batch reactor. The reaction is A + 3.2 B →
polymer. Assume a constant density and k = 0.047 kmol–1 h–1. The molecular weight of
A is 105 kg/kmol and B is 54 kg/kmol. What is the concentration of A and B after 12
hours?
19. The biochemical oxygen demand (BOD) represents the amount of dissolved oxygen
needed by biological organisms to break down organic materials in a water sample.
Consider a BOD of 100 mg O2/liter for a water treatment process involving microbes that
break down organic materials in a tank. Waste water flows through the tank at a rate of
2,000 m3/day and the average length of time that the water molecules remain in the tank
(mean residence time) is 6 hr. Find the rate of reaction of organic material in the treatment
tank.
Chemically Reacting Flows 397
20. For a gas–solid reaction at 360 K, the rate of reaction of reactant can be expressed in terms
of the partial pressure, pA, as dpA /dt 4.1 p2A (atm/hr). The reaction rate is also related to
the gas concentration according to rA kC A2 . Determine the rate constant, k, for this
reaction.
21. In a chemically reacting flow in a combustion chamber, the rate of reaction within the reac-
tor increases by a factor of 2.5 when the incoming concentration of reactant gas is tripled.
Find the order of the reaction.
22. Consider the combustion of octane fuel (C8H18). Determine the air–fuel ratio for complete
combustion with 140% theoretical air (40% excess air).
23. Complete combustion occurs when ethane (C2H6) is burned in excess air. The products
formed from this reaction are 85% nitrogen, 9% carbon dioxide, 4% oxygen, and 2% car-
bon monoxide. Find the combustion equation, percentage of excess air, and the air-fuel
ratio for this reaction.
24. A hydrocarbon gas has the following molar fractions: CH4, 75.6%; C2H6, 6.6%; C3H8,
2.6%; C4H10, 2.1%; and N2, 13.1%. After burning the gas with dry air, the products have the
following molar fractions: CO2, 8.2%; CO, 0.4%; O2, 8.1%; and N2, 83.3%.
a. Find the air–fuel ratio on a molar basis.
b. Assuming ideal gas behavior of the fuel mixture, determine the volume of fuel mixture
that is required to produce 5 kmol of products at 310 K and 1 bar.
25. Methane enters a combustion chamber at 400 K and 1 atm and reacts with 140% theoretical
air (40% excess air) at 500 K and 1 atm. The products of combustion leave the combustion
chamber at 1,800 K and 1 atm. Under steady-state conditions, find the rate of heat transfer
from the combustion chamber. Assume the average specific heat for the reacting methane
is 38 kJ/kmolK.
26. An expression for a controlled pellet size was obtained in Equation (8.23) for the thermal
decomposition of calcium carbonate (CaCO3) into CaO (solid phase) and CO2 (gas phase).
Alternatively, the rate of thermal decomposition can be controlled based on how CO2 gas
can diffuse through the gas boundary layer surrounding the pellet. Use Fick’s law to esti-
mate the required species concentration of CO2 in the bulk gas phase (called C∞) to provide
a specified steady-state diffusive flux, Ṅs, at the surface of a spherical pellet (at r = R).
27. 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 decom-
posed to one-half of its initial radius of 1 cm.
28. Heat is released in a process of desulfurization of a liquid iron alloy. Sulfur is separated out
and removed from a molten iron alloy during the magnesium injection operation. During
the process, magnesium gas bubbles react with sulfur in the molten alloy to form particles
of magnesium sulfide which rise to the surface of the container. 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.
29. 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 exam-
ple of a basic procedure to remove water from a substance by phase transformations to
solid and gas phases.
30. Spherical particles of zinc sulfide (ρs = 5 g/cm3) and radius of 2.5 mm react with an 8%
oxygen stream at 1,000°C and 1 atm. The reaction is given as follows:
Using a shrinking core model, find the relative resistance of the ash layer and the time
required to completely convert the particle. Neglect the gas film resistance. Also, assume
398 Advanced Heat Transfer
ks = 3.5 cm/s (reaction rate constant) and De = 0.05 cm2/s (gas diffusion coefficient in the
ZnO layer).
31. 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 fur-
nace enclosure with internal heating elements above the packed bed; a packed bed of pol-
ished 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 follow-
ing 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 mini-
mize these errors.
32. 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?
33. What thermophysical properties of solid particles affect the characteristics and parameters
of fluidized beds? Describe how these properties affect the relevant transport processes.
34. 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.
35. A reactor bed has a height of 3 m, inner diameter of 0.7 m, εm = 0.5, and εmf = 0.62. Ιt is
made of limestone particles with a density of 3,100 kg/m3. Air with a temperature of 25°C
flows upward in the reactor bed. Determine the pressure of the inlet air that is required to
fluidize the solid particles. Assume the outlet pressure is 1 atm.
36. A bed of solids of density 3,500 kg/m3, diameter 1.6 mm and εm = 0.42 is resting on a
screen in a fluidized bed reactor. Inlet air flows upward through the solid particles. Will the
solids fluidize when air is passed through the reactor? Inlet and outlet pressures are 140 kPa
and 115 kPa, respectively, while T = 110°C. The bed height is 3.5 m and inner diameter is
0.55 m.
37. What similarities exist between particle movement in fluidized beds and two-phase flows
of liquid–gas or solid–liquid mixtures?
38. 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?
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 separated by walls to
prevent mixing or can have direct contact with each other. Heat exchangers are used in space heat-
ing of buildings, 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 exchang-
ers. In some limited cases, semi-analytical methods can be used to predict the temperature distribu-
tions, heat transfer rates, and thermal effectiveness of heat exchangers. This chapter will provide an
introduction to the various methods of analysis of heat exchangers, including both steady-state and
transient operating conditions.
Heat exchangers have various flow and geometrical configurations. The selection of an appropri-
ate type of heat exchanger involves many design considerations, such as the fluid types; operating
pressure; flow rates; desired heat transfer rate; target outlet temperature; among others. The basic
component of most heat exchangers is a fluid flowing through one or more tubes with another fluid
flowing on the outside of the tubes. In this process, the main processes of heat transfer include
convection on both sides of the tube wall and conduction through the tube wall. The overall heat
transfer coefficient characterizes the combined thermal resistance of these multiple modes of heat
transfer.
The driving temperature difference across a heat transfer surface is called the log mean tempera-
ture difference. In order to improve the thermal efficiency, heat exchangers are typically designed
to increase this temperature difference, minimize the resistance to fluid flow through the exchanger,
and maximize the surface area of the wall between the two fluids. The exchanger’s performance can
also be affected by the addition of fins or corrugations in one or both directions, which increase sur-
face area and may channel fluid flow or induce turbulence. Also, finned and other extended surfaces
are often used in heat exchangers for heat transfer enhancement. Finned surfaces can be represented
by using an overall surface efficiency of the finned surface in the overall heat transfer coefficient
(Chapter 2).
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 non–cir-
cular 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 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.
Active research is currently underway in the field of heat exchanger design and analysis. One of
the active fields of research involves nanofluids. Nanofluids have received increasingly more atten-
tion in recent years due to their promising potential of major enhancement of heat transfer and ther-
mal efficiency in heat exchangers. As discussed in Chapter 6, a nanofluid is a fluid with suspended
nanometer-sized particles, called nanoparticles, typically made of metals, oxides, carbides, or car-
bon nanotubes. The fluids have colloidal suspensions of nanoparticles in a base fluid. Buongiorno
(2006) and Vermahmoudi et al. (2013) discussed the enhanced thermal properties of nanofluids as
coolants in heat exchangers.
In this chapter, the design and analysis of various types of heat exchangers will be investigated.
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; effec-
tiveness–NTU method; thermal response to transient temperature changes; and condensers and
evaporators.
(typically consisting of a tube bundle, fan, and support structure), heat pipes, and thermosyphons
(Chapter 5).
Cross-flow heat exchangers typically consist of an outer flow across tubes carrying fluid that
flows in a direction perpendicular to the cross-flow. Often 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, such as by fins separating the fluid streams, then
the configuration is unmixed. Otherwise a mixed configuration permits complete mixing of the fluid
streams in the external cross-flow.
A plate heat exchanger is another example of indirect contact which separates the fluids by
plates. It is commonly used in cryogenic (low temperature) and food processing industries. The
plates can readily be taken apart for cleaning. This configuration normally has a high surface area
to volume ratio which is advantageous. A plate-fin heat exchanger has fins embedded between the
parallel plates in any combination of crossflow or parallel flow configuration. They are commonly
used in liquefaction processes.
A shell-and-tube heat exchanger is one of the most common types of heat exchangers due to its
versatility and effectiveness of operations over a range of pressures and temperatures. It consists of
an outer shell where fluid enters through one end, passes across internal tubes carrying a fluid at a
different temperature, and exits through the other end. One fluid flow through the tubes, while the
other fluid flows over the outside of the tubes. The fluids can flow in parallel, or a cross/counter flow
configuration, in either a single or two-phase flow. An example of a shell-and-tube heat exchanger
with one shell and one tube pass is illustrated in Figure 9.2.
The main sections of the heat exchanger are the front header where fluid enters the tube side; rear
header where fluid leaves or where it is returned to the front header in multiple passes; tube bundles;
baffles; and outer shell that contains the tube bundle. 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 remain-
ing 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.
Three common configurations exist for the shell and tube geometry – fixed tube sheet; U–tube;
and floating header. In a fixed tube exchanger, the tube sheet is rigidly attached and welded to the
shell. This yields an economical and relatively simple design although the outer surfaces of the tubes
may be inaccessible. In a U–tube exchanger, the tubes are bent in the shape of a U. This configu-
ration has better flexibility of thermal expansion and the tube bundle can be readily removed for
FIGURE 9.3 (a) Single and (b) double segmental baffle types.
cleaning. The third common configuration is a floating head exchanger, where the tube sheet at the
rear header end is allowed to move or float. The tube sheet has a slightly smaller diameter than the
shell so as to allow the bundle to be pulled through the shell.
Tubes within the heat exchanger are typically aligned in triangular or square patters in the tube
sheets (see Figure 9.3). The ratio of the tube pitch (shortest center-to-center distance between tubes)
to the tube diameter is normally between 1.25 and 1.33. A triangular layout allows more tubes in a
given volume but also creates more difficulty in reaching the tube surfaces for mechanical cleaning.
Baffles are used on the shell side to enhance mixing, turbulence, and heat transfer, as well as to
support the tubes and reduce structural vibrations. A single segmental baffle (most common) forces
the fluid or gas across the entire tube bundle. A double segmental exchanger lowers the shell side
velocity and pressure drop by allowing the fluid to flow between alternating sections of the tube
in a straighter overall direction (Figure 9.3). Additional heat transfer enhancements such as tube
inserts or extended surfaces are also commonly used to enhance the heat transfer rate. Other forms
of special surfaces can be used to promote nucleate boiling or condensation when the temperature
differences are small.
Regular maintenance and safety are important considerations for shell-and-tube heat exchangers.
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 atmo-
spheric pressure, leakage can occur. Pressurizing a system 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 a leakage point.
Several criteria and decisions influence the optimal selection and design of a shell-and-tube heat
exchanger. These decisions include the following considerations:
• allocations of fluids to the shell and tube sides – safety and reliability factors, cleaning issues,
cost;
• shell selection – shell side pressure drop needs, unit size limitations, type of applications;
• header selection – accessibility and cleaning of tubes, servicing, high-pressure applications,
thermal expansion;
• selection of heat exchanger geometry and size – mechanical cleaning, weight and cost limita-
tions, and materials of construction.
The mechanical design of a heat exchanger varies significantly depending on the operating condi-
tions such as pressure, temperature, and flow rate. Any heat exchanger that operates at above the
atmospheric pressure must be designed in accordance with a pressure vessel design code such as the
Boiler and Pressure Vessel Code (BPVC) of the American Society of Mechanical Engineers (ASME).
The design code specifies the requirements and standards for a pressure vessel, such as key struc-
tural features like the shell and flange thicknesses. The BPVC was first created in response to public
anger after a major explosion of a fire–tube boiler at a shoe factory in Brockton, Massachusetts, in
1905. The explosion resulted in the death of 58 people and 150 seriously injured.
Heat Exchangers 403
In addition, the thermal design is another iterative process that further considers numerous fac-
tors such as the flow distributions, geometrical configuration (tube diameter, layout, areas), number
of tube passes, shell type, among others. The thermal design is closely linked with the mechanical
design. 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).
1 1 R ln ro /ri R f ,c 1
f ,h (9.1)
UA hh A A 2 kL 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), con-
duction through the inner pipe wall, fouling (inner surface), and convection (inner cold fluid stream).
Periodic cleaning of the heat exchanger surfaces 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)
m hch m ccc m hch m ccc
d T 1 1
U dA (9.4)
T
mhch mccc
By rearranging, the total rate of heat transfer from the hot fluid stream to the cold stream between
the inlet (1) and outlet (2) can be rewritten as
q UATlm (9.6)
T2 T1
Tlm (9.7)
ln T2 /T1
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 counterflow con-
figuration, except that the variables are replaced by ∆T1 = Th,1 – Tc,1 and ∆T2 = Th,2 – Tc,2.
In the previous energy balances, thermal resistances were required, including the convection and
fouling coefficients. If these are unknown and instead the mass flow rates and inlet/outlet tempera-
tures are given, an alternative approach can be used by performing heat balances on the entire tube.
For the inner tube,
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 direction. If the same inlet and outlet temperatures are used, then
the log mean temperature difference 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.
q FUATlm (9.10)
The correction factor, F, is usually based on experimental data to account for baffles and other
geometrical parameters. The value of F depends on the type of heat exchanger. For a one–shell–
pass, one–tube–pass heat exchanger, the correction factor is F = 1.
The correction factor, F, can be graphically depicted based on the following parameters (see
Figure 9.5):
Ti To
R (9.11)
t o ti
t o ti
P (9.12)
Ti ti
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 temperature 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) presented 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
F
R 2 1 / R 1 log 1 P / 1 PR
(9.13)
log a R 2 1 / a R 2 1
For two shell passes and any multiple of four tube passes,
F
R 2 1 / 2 R 2 log 1 P / 1 PR
(9.14)
2
log a b R 1 / a b R 2 1
where
2 2
a
P
1 R; b
P
1 P 1 PR (9.15)
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 transfer rate. The following
example demonstrates how the correction factors are applied.
ccc Tc ,o Tc ,i
qm (9.17)
1.58 106
Tc ,o 20 145.5C (9.18)
3 4,195
The outlet temperature can then be used to compute the heat exchange factors as follows:
145.5 20
P 0.39 (9.19)
340 20
340 180
R 1.27 (9.20)
145.5 20
which together yield a correction factor of F ≈ 0.98. Then using Equation (9.10), the surface area
can be obtained as
Since the total surface area of tubes is A = 2πDL and L = 60m, the required tube diameter 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 temperature difference.
Depending on the axial temperature gradient relative to the radial conductance, this effect may
lead to a difference of up to ±5% between the hot and cold fluid streams.
9.3.3 Pressure Drop
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:
1 G 2 A ff i
2
A
p 1 1 f i (9.22)
2 i A fr o
2
m A ff
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 maximum veloc-
ity), 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 docu-
mented extensively by Kays and London (1984) for a variety of heat exchangers, including finned
and various tubular configurations.
408 Advanced Heat Transfer
q
(9.23)
qmax
As outlined earlier, the maximum heat transfer rate is achieved in a counterflow arrangement,
when one of the fluids experiences the maximum possible temperature difference,
qmax mc
min Th,i Tc,i (9.24)
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 mc
c (9.25)
Ch mc
h (9.26)
where the subscripts c and h refer to the cold and hot fluid streams, respectively. Then combining
Equations (9.8), (9.9), and (9.24),
Tc,o Tc,i
c (9.28)
Th,i Tc,i
Th,i Th,o
h (9.29)
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 = (9.30)
Cmin
where U is the overall heat transfer coefficient. The effectiveness of a particular heat exchanger, ε,
is often tabulated and graphically illustrated as a function of Cmin/Cmax (denoted by Cr) and NTU.
Heat Exchangers 409
FIGURE 9.6 Effectiveness curves of various heat exchangers. (adapted from Bergman et al., 2011)
Sample curves of the effectiveness for a number of different heat exchanger configurations are
illustrated in Figure 9.6. Corresponding algebraic expressions involving ε, NTU, and Cr have been
developed and summarized for a variety of heat exchangers by Kays and London (1984).
The number of transfer units, NTU, can be interpreted as the ratio of the total heat conductance
(or reciprocal of the total thermal resistance, R) to the minimum heat capacity rate. Also, a time scale
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 represents a mass (or fluid) residence time, tres, in the
heat exchanger. Based on this interpretation, Equation (9.30) can be rewritten as
410 Advanced Heat Transfer
UA UA 1 tres
NTU (9.31)
Cmin mc 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 indi-
cates 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 9.6(a)–(f) have been presented by Kays and London (1984). The Cr is given by Cmin/Cmax.
(a) Concentric tube (parallel flow): Figure 9.6(a)
1 exp NTU 1 Cr (9.32)
1 Cr
1 exp NTU 1 Cr (9.33)
1 Cr exp NTU 1 Cr
(c) Shell-and-tube (one shell pass; two, four, … tube passes): Figure 9.6(c)
1
1 exp NTU 1 Cr2
2 1 Cr 1 Cr
2
(9.34)
1 exp NTU 1 Cr2
NTU 0.22
1 exp
Cr
exp Cr NTU 0.78 1
(9.35)
1
Cr
1 exp Cr 1 exp NTU
(9.36)
1
1 exp 1 exp Cr NTU (9.37)
Cr
Fins can substantially improve the heat transfer rates from heat exchanger surfaces. The follow-
ing example demonstrates how fins can be incorporated into the previous analysis.
Heat Exchangers 411
a
m 1
=
G = = 20 kg/sm2 (9.38)
Ac 0.05
DhG 0.01 20
Re 11173
, (9.39)
1.79 10 5
Then based on the specified Colburn factor and Chilton–Colburn analogy (Chapter 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 (9.41)
o 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 = 4209 J/kgK, ρ = 963 kg/m3, k = 0.677 W/mK, Pr = 1.91, and μ = 3.1 × 10–4 kg/msec.
Using these values,
which suggests the following correlation for internal flow within a pipe (Chapter 3):
k 0.677
1.91
0 .8 0.33
h Nu 0.023 6.2 104 6, 577 W/m2K (9.44)
D 0.02
Then the thermal resistance based on water flow within the pipes is given by
1 1
Rw 0.0003 K/W (9.45)
hw A 6, 577 0.02 0.2 40
412 Advanced Heat Transfer
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/m2K (9.47)
ARtot 1.6 0.0013 0.0003 106
aca 11007
m , 1, 007 W/K Cmin (9.48)
which yields a ratio of Cmin /Cmax = 0.02. Then the number of thermal units becomes:
Based on these Cmin/Cmax and NTU values, the effectiveness charts yield ε ≈ 0.5,
Ta,o Ta,i
0 .5 (9.51)
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.
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).
FIGURE 9.7 (a) Flow configuration and (b) coordinate system of a honeycomb heat exchanger.
The effective thermal conductivity of a cellular array in the y-direction can be written as (Gu
et al., 2001):
ke cky r ks 1 r k f (9.52)
where the subscripts e, s, and f of the thermal conductivity, k, denote effective, solid, and fluid,
respectively. Also, ρr is the relative density of the honeycomb structure with respect to the fluid
and cky is a tortuosity coefficient that accounts for the tortuous shape of the cell walls. For a regular
hexagonal cell, cky = 0.5. Setting kf = 0 for the effective thermal conductivity of the solid yields kse =
0.5ρr ks. Similarly for the fluid, kfe = 1 – ρr kf.
The surface area density, γa, and Nusselt number, Nu, of the honeycomb structure are defined as
1 r
a ca (9.53)
l
hla
Nu = (9.54)
kf
where h, l, la, and ca denote the local heat transfer coefficient, cell wall length, characteristic
length scale, and a geometric shape factor, respectively. For a hexagonal honeycomb structure,
la l 3 1 r and ca = 4 / 3 (Gu et al., 2001).
Consider a cooling fluid at a temperature of T0 and velocity u0 entering at the inlet of a honey-
comb heat exchanger. The heat exchanger module is insulated on the left and right sides while a
constant heat flux is applied along the top and bottom surfaces. Under steady-state and fully devel-
oped flow conditions, the energy equations in the solid matrix and fluid phases (subscripts s and f,
respectively) can be expressed as
2Ts
kse h a Ts T f 0 (9.55)
y 2
T f 2T f
f c pu0 k fe h a Ts T f (9.56)
x y 2
From an energy balance at an axial position, x, the heat inflow balances the energy absorbed by
the fluid, i.e.,
x
f c pu0 H T f x T0 2 qw d
0
(9.57)
414 Advanced Heat Transfer
T f 2qw x
(9.58)
x f c pu0 H
This temperature gradient can be substitute back into fluid energy equation to yield an ordinary
differential equation in terms of the lateral coordinate, y, alone.
Two possible boundary conditions are an isothermal boundary with a uniform wall temperature,
Tw, and a uniform wall heat flux, qw. Lu (1999) reported that the only difference between both con-
ditions is that Nu = 3.35 for an isothermal surface and Nu = 4.021 for the constant wall heat flux
surface. Then the boundary conditions become:
Ts T f
0; y 0 (9.59)
y y
H
Ts T f Tw x ; y (9.60)
2
where Tw(x) is the temperature of the substrate wall and constant for isothermal conditions. For the
isoflux condition, when the heat flux is applied at the outer surface, the wall temperature is likely
to be uniform due to the high thermal conductivity of the substrate. Then the same isothermal wall
conditions can be applied to the isoflux condition except that Tw(x) is unknown and must be obtained
as part of the solution.
To nondimensionalize the governing equations, define the following dimensionless variables:
h a 2 1 r k f
e2 ; ke (9.61)
kse r ks
s Ts Tw x ; f T f Tw x (9.62)
2 s
e2 s f 0 (9.63)
y 2
2 f 2qw x
ke e2 s f (9.64)
y 2
kse H
s y
Ake cosh 1 1/ke e y 2qw x
qw x
y2 B (9.65)
1 1/ke
2
e Hkse 1 ke Hkse 1 ke
2
e
f y
A cosh 1 1/ke e y qw x
y2 B (9.66)
1 1/ke
2
e Hkse 1 ke
Heat Exchangers 415
where
2qw x
A (9.67)
Hkse cosh
1 1/ke e H / 2 ke ke 1
qw x H 2qw x
B (9.68)
4kse 1 ke Hksee2 1 ke 2
Then the surface-averaged temperature difference (between the solid substrate and fluid) and
wall heat flux can be obtained as follows:
Tw x T f x
qw x
2 1
2 tanh 1 1/ke e H / 2 H
(9.69)
kse 1 ke e2 1 ke H e H 1 1/ke 6
qs x
qw x ks 2 tanh 1 1/ke e H / 2
1
(9.70)
kse 1 ke e H 1 ke
Sample results of the effects of fluid conductivity and relative density on the temperature profiles
of the fluid and solid are shown in Figure 9.8. Problem parameters include: L = 50 mm; ṁ = 1.2 ×
10–5 kg/s; T0 = 283.15 K; qw″ = 5,000 W/m2; l = 1 mm; H = 10 3l ; ρf = 1.226 kg/m3; cp = 1,006.43
J/kgK; μf = 1.7894 × 10–5 Pa ⋅ s; kf = 0.026 W/mK; and ks = 200 W/mK.
In Figure 9.8(a), the relative density is maintained constant (ρr = 0.02) and the temperatures of
the solid and fluid increase in the y-direction. The temperature difference increases at the lower
fluid conductivity, kf = 0.026 W/mK, due to the larger thermal resistance. As the fluid conductivity
decreases, both the local heat transfer coefficient and effective thermal conductivity also decrease,
thereby increasing the temperature difference between the solid and fluid. In Figure 9.8(b), the fluid
conductivity is held constant (kf = 0.026 W/mK) and the temperature profiles of the solid and fluid
FIGURE 9.8 Effects of (a) fluid conductivity and (b) relative density on temperature profiles. (adapted from
Liu et al., 2016)
416 Advanced Heat Transfer
increase in the y-direction and for higher relative densities (denoted as rd in the caption). The excess
temperatures of the solid and fluid decrease from the heated surface to the center as the relative den-
sity increases. Further results of the heat transfer coefficient and thermal performance index have
been reported by Liu et al. (2016).
Ts 2T 2T
s c psus k y 2s k x 2s (9.71)
x y x
Ts Ts
0; k y h Ts x,w T f (9.72)
y y
y 0 yw
Ts
Ts 0,y Tsi y ; 0 (9.73)
x x
The condition at x → ∞ ensures that the temperature profile remains bounded as the heat
exchanger increases in size. The boundary condition at y = 0 reflects the symmetry of the tempera-
ture profile across the centerline of the channel. At y = w, heat conduction from the solids across the
separator plate balances the convective heat transfer from the fluid.
FIGURE 9.9 Schematic of a parallel plate channel in a moving bed heat exchanger.
Heat Exchangers 417
y x Ts T f
y ; x ; s (9.74)
w H Tsi T f
2
hW hH kx w
Bi ; NTU ; R H (9.75)
ky sus wc ps ky
where h, kx, and ky are axial and lateral solids effective thermal conductivities and overall heat trans-
fer coefficient, respectively. Then the energy equation and boundary conditions become:
Bi s 2 s 2 s
R (9.76)
NTU x x 2 y2
s s
y
0;
y
Bi s x ,1 0 (9.77)
y 0 y 1
s
s 0,y si y ; x
0 (9.78)
x
It can be shown that the problem is separable, so in other words, the method of separation of
variables can be applied to solve the problem in the following form:
s x ,y X x Y y (9.79)
Substituting this expression into the energy equation and solving the resulting ordinary differen-
tial equation for the Y(y*) problem, subject to the boundary conditions,
Yn y C1,n cos n y (9.80)
where Y(y*), C1,n, and λn are orthogonal eigenfunctions, integration constants, and eigenvalues as
defined by
n tan n Bi (9.81)
where n = 1, 2, 3, … Similarly, eigenfunctions are determined for the Xn(x*) problem as follows:
X n x C2er1n x (9.82)
where
2
1 Bi Bi
r1n 4 Rn2 (9.83)
2 R NTU NTU
Then by constructing the full solution from superposition of the eigenfunctions, and applying the
boundary conditions, the solids temperature becomes:
418 Advanced Heat Transfer
FIGURE 9.10 Predicted temperatures for Bi = 10, NTU = 10 and R = 10 for (a) a constant solids entrance
temperature and (b) an entrance temperature profile of θs = cos (πy*/3).
s
F y cos y dy
0
n
exp
1 Bi Bi
2
4 R 2
x
(9.84)
2 R NTU NTU
n
1
cos y dy
2
n 1 n
0
1
F y
F y cos y dy
0
1
n
cos n y (9.85)
cos y dy
2
n 1 n
0
FIGURE 9.11 Schematic of symmetrical channel with metal foam in a plate heat exchanger.
non-equilibrium model (considers fluid and metal foam separately, thus a local temperature differ-
ence). Although the latter approach is generally more accurate, it is also more difficult to apply since
it requires the varying interfacial heat transfer coefficient. In this section, a Brinkman–extended form
of the momentum equations will be used to determine the velocity, temperature, Nusselt number, and
friction factor for fluid motion through heat exchanger channels with metal foam.
Consider a plate heat exchanger partially filled by metal foam. Several metal-foam-enhanced
plates are joined together to form the heat exchanger. As each of the interior channels are the same
as each other, they can be represented by a series of symmetrical filled channels (see Figure 9.11).
Assume fully developed flow through a channel of height, a, with a metal foam layer in a′ < x < a.
The width (x-direction) and length (z-direction) are relatively large compared with the height (x-
direction). An adiabatic condition is applied along the symmetry boundary (x = 0), while the wall
enhanced by metal foam is uniformly heated. Lu et al. (2016) presented a detailed analysis of this
flow configuration.
Assume constant thermophysical properties and a homogeneous, isotropic porous medium (metal
foam). Within the interior foam-free region (0 < x < a′), for steady one-dimensional, fully developed
flow conditions, the reduced forms of the z-momentum and energy equations become:
p 2u
0 f 2 (9.86)
z x
T f 2T f
f cfu kf (9.87)
z x 2
In the metal-foam-filled region (a′ < x < a), a Brinkman–extended Darcy model (Brinkman,
1949) and two-equation non-equilibrium model are used in reduced forms as follows in the fluid
phase.
p f 2u f
0 u (9.88)
z x 2 K
T f 2T f
f cfu kf hsf asf Ts T f (9.89)
z x 2
2Ts
0 ks hsf asf Ts T f (9.90)
x 2
where ε, K, h, and the subscripts f, s and sf refer to the porosity of the metal foam, permeability, heat
transfer coefficient, fluid, solid, and solid–fluid interface, respectively.
420 Advanced Heat Transfer
The boundary conditions are specified as follows. Along the lone of symmetry (x = 0), bottom
wall (x = a) and flui/metal foam interface (x = a′):
u T f
x 0; 0; 0 (9.91)
x x
=x a=
; u 0; T=
f T=
s Tw (9.92)
f u u
x a : u|a u|a ; f (9.93)
x a x a
T f T T T
x a : T f |a T f |a ; k f ks s k f ; ks hsf Ts T f (9.94)
x x a x a x a
a
X
x
; U
u
; s
Ts Tw ks ; f
T f Tw ks (9.95)
a um qw a qw a
a K K dp hsf a kf
H ; Da ; P ; A ; B (9.96)
a a2 f um dz ks ks
k fe hsf asf a 2 D C 1
C ; D ; S ; R (9.97)
ks ks Da C
Solving the momentum equation in the foam-free region (0 < X < H), subject to the boundary
conditions, yields the following fluid velocity distribution,
1
U P X 2 M1 (9.98)
2 Da
U P M 2e SX M3e SX 1 (9.99)
where
H2
M1 M 2 e SH e2 S SH e S SH 1
2 Da
(9.100)
e S SH SH
M2 ; M3 e S M 2e2 S (9.101)
e SH e2 S SH
H3
P 1/
M1H SDa M 2e SH M3e SH SH (9.102)
6 Da
Heat Exchangers 421
Integrating the velocity profile over the cross-section of the channel yields the volume-averaged
velocity, um, designated as u in the foam-free region, and εu in the metal-foam region. Then the fric-
tion factor becomes:
dp /dz 4a 32 P
f (9.103)
1/ 2 f um
2
Da Re
The temperature distribution can then be determined from the solution of the energy equations in
the fluid and solid. Performing an energy balance in the interior foam-free region over a differential
control volume of width, dz,
f c f um a dT f qw dz (9.104)
Similarly in the solid region for thermally fully developed flow conditions, it can be shown that:
dT f dTs dTw qw
(9.105)
dz dz dz f c f um a
Substituting into the fluid energy equation (foam-free region, (0 < X < H) and solving,
P 1 1
f X 4 M1 X 2 N1 (9.106)
B 24 Da 2
P 1
s
C 1
1
2
D1e RX D2e RX R3 M 2e SX M3e SX X 2 N 2 X N 3 2
R
(9.107)
P D1e RX D2e RX S 2 1
f
C 1 C
1
D
R3 M 2e M3e
SX SX 1
X 2 N2 X N3
2
(9.108)
CR 2
The average temperature difference of fluid along the cross-section of channel, θf,b, is determined
by integration of the above temperature profile across the channel.
a
u dx U dX S Da U dX
f f
H 1
f ,b 0
a f
2
f (9.109)
udx
0
f 0
H
B D1e RH D2e RH S 2 M H2 1
N1 1 R3 M 2e SH SH3 2 N 2 H N 3 CR 2
C 1 C D e
4 2
H MH (9.110)
1
24 Da 2
M 2e SH M3e SH H3 M e S M 3e S 1
N2 M1 1 H ; N 3 2 N2 (9.111)
S 6 Da S 2
422 Advanced Heat Transfer
D1
R1e R R AR 2 /D R2e RH ; D2 R1e R D1e2 R (9.112)
e 2R
R AR /D e R AR /D
2 2 RH 2
R1 R3 M 2e S M3e S N 2 N 3
1 1
R2 2
(9.113)
A 1 A 1 A
R2 S 2 R3 M 2e SH M3e SH N 2 H (9.114)
D S D S D
R2 1 S2
R3 ; R4 N 3 1 R3 (9.115)
S 2
R2 S 2 2
CR D
Then the Nusselt number, Nu, for a symmetrical channel can be determined from the temperature
difference as follows:
h 4a 4a qw 4
Nu (9.116)
kf k f Tw T f ,b B f ,b
The dimensionless velocity and temperature distributions are shown in Figure 9.12. As expected,
the fluid velocity is maximum in the center of the channel and diminished in the metal–foam-filled
region. When the porosity ε increases or nominal pore density (PPI) decreases, the flow resistance in
the metal foam decreases, and therefore the resulting fluid velocity increases. Also, as the thickness
of the metal layer increases, also the maximum velocity increase in the foam-free region.
In the foam-filled region, the solid temperature of the metal foam decrease nearly linearly from
the heated wall due to conduction-dominated effects. The fluid temperature approaches the solid
temperature near the wall. The solid temperature is relatively independent of porosity. But the fluid–
solid temperature difference is higher for the 10 PPI foam because the higher pore density has a
higher heat transfer area between the solid and fluid. The flow resistance also impacts the rate of
flow through the foam-filled region. In the foam-free region, the temperature drops more rapidly at
10 PPI because of the higher flow rate.
FIGURE 9.12 (a) Fluid velocity and (b) temperature distributions. (adapted from Lu et al., 2016)
Heat Exchangers 423
FIGURE 9.13 Effects of dimensionless height on the friction factor and Nusselt number. (adapted from Lu
et al., 2016)
Figure 9.13 shows the resulting Nusselt number and friction factor. The friction factor increases
with dimensionless height of the metal foam, Hm. The friction factor asymptote of f = 0.1067 is illus-
trated for a fully foam-free channel for comparison and validation purposes. The Nusselt number
also increases with Hm. The heat transfer effectiveness also improves by increasing the thickness of
the metal-foam layer. The result approaches the correct asymptote of Nu = 8.24 when the dimen-
sionless height becomes Hm = 0 (rectangular channel without metal foam). For Hm < 0.6, the Nusselt
number increase is slower since the metal-foam layer is thinner and heat transfer enhancement only
applies to a relatively small amount of fluid. At Hm > 0.8, Nu increase more dramatically since the
rate of flow through the metal foam increase more significantly.
d h
C Rh NTU h c H hC Rh NTU h a,h 0 (9.117)
dX
d c
C Rc NTU h c H cC Rc NTU c a,c 0 (9.118)
dX
h 0 1; c 0 0 (9.119)
where
Here H represents a ratio of the thermal resistance between the fluids to that between the ambient
condition and corresponding fluid. In the above energy equations for the hot (subscript h) and cold
(c) fluid streams, the terms from left to right designate the streamwise convection, and heat transfer
between the fluids, and between the ambient and individual fluids, respectively.
Both energy equations are coupled and solved simultaneously to yield:
where
n1,2 1 C Rh 1 H h C Rc 1 H c C Rh 1 H h C Rc 1 H c 4C RhC Rc
2
(9.125)
2
kh k2 k2 kc k1 kc kh k1
C1 ; C2 (9.126)
k1 k2 k1 k2
Heat Exchangers 425
FIGURE 9.15 ε–NTU relationships for (a) varying heat capacity ratios (Hh = 0 = Hc, θa,h = 1 = θa,c) and (b)
varying thermal resistance ratios (CRh = 1 = CRc, θa,h = 1.5 = θa,c)
n1 C Rh 1 H h n2 C Rh 1 H h
k1 ; k2 (9.127)
C Rh C Rh
Once the temperatures are determined, the effectiveness of the heat exchanger with respect to the
heat and cold fluids can be expressed as
Predicted results of the effectiveness of the MCHX were reported by Mathew and Hegab (2011).
Sample results are illustrated in Figure 9.15. In Figure 9.15(a), the effectiveness decreases with the
heat capacity ratio. In this case, when the ambient temperature is the same as the inlet temperature of
the fluid, the predicted results correspond to the conventional ε–NTU relationship (without external
heat transfer to the ambient).
In Figure 9.15(b), the ambient temperatures are higher than the fluid inlet temperatures so exter-
nal heat transfer occurs from each of the surrounding ambients to the respective fluids. As a result,
the hot and cold stream effectiveness, decreases and increases, respectively, in comparison with the
prior case without external heat transfer. Peak values of the effectiveness occur for the hot fluid.
The effectiveness initially rises because heat transfer between the fluids exceeds the heat transfer
between the ambient and hot fluid. When NTU exceeds the peak NTU value, this net heat transfer
decreases below the earlier peak value of NTU.
step changes of temperature within a heat exchanger was analyzed by Naterer and Lam (2006),
including varying convection coefficients and multiple-step changes in temperature. Assume that
fluid temperatures are functions of time (t) and position (x) and that longitudinal and transverse heat
conduction within the wall and fluids are negligible, compared to convection heat transfer.
Energy balances will be formulated for differential control volumes of thickness dx in Figure 9.16.
From a heat balance in the wall, the transient change of wall thermal energy equals the difference of
convective heat transfer rates from the single-phase fluid and phase-change fluid,
Tw
mwc pw hs x Ac Ts Tw hc x Ac Tw Tc (9.130)
t
where the subscripts w, s, and c refer to wall, single-phase fluid, and constant temperature (two-
phase) fluid, respectively.
Similarly, energy balances within differential control volumes in the single-phase fluid and
phase–change fluid, respectively, yield:
Ts T
ms c ps m s c ps L s hs x As Tw Ts (9.131)
t x
g
m c Lh fg hc x Ac Tw Tc (9.132)
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 temperature. 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 m T Ts,i
x ; t s t; T T (9.133)
L ms c, s ,i
mwc pw hc x Ac hs x As m c
N1 ; N2 x ; N3 x ; N4 (9.134)
ms c ps m s c ps m s c ps m s Ja
N 2 N3 c ps Tw Ts
NTU ; Ja (9.135)
N2 N3 h fg
Heat Exchangers 427
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:
w
N1 N 2 x w 1 N 3 x w s 0 (9.136)
t
s s
N 3 x w s (9.137)
t x
g
N4 N 2 x w 1 (9.138)
x
Integrating over the length of the heat exchanger (from x* = 0 to 1), these equations become:
1 1
1 1
N1 w dx N 2 w 1 dx N 3 w dx s dx 0 (9.139)
t
0 0 0 0
1
1 1
s dx s 1 s 0 N 3 w dx s dx (9.140)
t
0 0 0
1
N 4 g 1 g 0 N 2 w 1 dx (9.141)
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 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 Tc is applied. Using the steady-state and initial temperature
profiles, it can be shown that the fluid and wall temperatures become:
s 1 e NTUx
f t
(9.142)
NTU NTUx
w 1
N2
e g t
(9.143)
where
f t f t c0 1 f t (9.144)
g t g t c0 1 g t (9.145)
Substituting these profiles into the integral equations and integrating from x* = 0 → 1 leads to:
C f
df t
dt
1 t C g t
1
(9.146)
428 Advanced Heat Transfer
D f
dg t
dt
1 t D g t D
2
3 (9.147)
where
C1
N 3 NTU e NTU 1 NTU 1 e NTU (9.148)
NTU
NTU e 1
N2 N3 N2
D2 ; D3 (9.150)
N1 N1 N1 / N 2 e NTU N1 / N 2
Solving these coupled equations, subject to the initial conditions of f t 0 c0 g t 0 ,
yields:
D2 1t D2 2 t
f t a 1
D1
e b 2
D1
e
C1D2 D3 D
3
D1 C2 D1 C1D2 D1
(9.151)
C1D3
g t ae 1t be 2 t (9.152)
C2 D1 C1D2
where
D D2 2 0 D3 C1 2 C1D2 C2 D1
a 1 c (9.153)
1 2 1 2 C1D2 C2 D1
D D2 1 0 D3 C11 C1D2 C2 D1
b 1 c (9.154)
1 2 1 2 C1D2 C2 D1
1
C1 D2 4 C2 D1 C1D2
2
1,2 C1 D2 (9.155)
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 temperature) 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 con-
vection 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.
Heat Exchangers 429
FIGURE 9.17 Changes of fluid and wall temperatures with (a) time at varying N1 values; and (b) position at
varying times. (adapted from Naterer and Lam, 2006)
Sample results are presented in Figure 9.17 for the transient response to step-wise temperature
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 9.17 – (1) N2 = 1, N3 = 1 (NTU = 0.5), x* = 1 (outlet)
and co 0; and (2) N1 = 300, N2 = N3 = 1 (NTU = 0.5) and co 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 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 distur-
bances to propagate downstream.
In the spatial variations of temperatures, both fluid and wall temperatures rise nearly linearly 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 values of the latent heat of vaporization. These results indi-
cate 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.
A typical tube configuration for Case P1 is shown in Figure 9.19. It can be observed that each fluid
is exchanging heat with the other two streams in the heat exchanger. Assume steady-state operation,
constant mass flow rates, and thermophysical properties, negligible axial conduction in the pipe
walls, and a uniform fluid temperature in the transverse direction, equal to the mean temperature.
In Figure 9.18 for Case P1, three control volumes of differential width, dx, are illustrated – CV1,
CV2, and CV3, for energy balances on fluid streams 1, 2, and 3, respectively. Note that fluid stream
1 has thermal communication with both streams 2 and 3 as illustrated in Figure 9.19. An energy bal-
ance on fluid 1 within control volume CV1 yields:
where U is the overall heat transfer coefficient, L is the length of the heat exchanger, A is the contact
area of heat transfer, and 1/(UA)ij is the thermal resistance between fluid streams i and j. As shown
in Figure 9.18, fluid streams 1, 2, and 3 are cold, hot, and intermediate temperature streams, respec-
tively. If fluid 1 is flowing in the opposite (negative) x-direction, a negative sign is placed in front of
the square bracket term on the left side. Alternatively, using the variable κ1, either of the two equa-
tions, in either flow direction, can be reduced to:
Similar energy balances can be written for fluids 2 (CV2) and 3 (CV3), as well as other flow
directions of each fluid stream.
Heat Exchangers 431
x T T1,in
x ; (9.158)
L T2,in T1,in
where the subscript in refers to the inlet value. Then the energy balances for fluid streams 1 to 3
become:
dT1
1 NTU1 2 1 R2 3 1 (9.159)
dx
dT2
2 NTU1C12 1 2 R1 3 2 (9.160)
dx
dT3 C
3 NTU1 12 R1 2 3 R2 1 3 (9.161)
dx C32
where R1 and R2 refer to the ratio of the thermal resistance between streams 1 and 2 and streams 2
and 3 (R1), and the ratio between streams 1 and 2 and streams 1 and 3 (R2), respectively. Also, NTU1
and Cij denote the number of transfer units of fluid 1, i.e., (UA)12/(ṁcp)1, and the capacitance ratio
for fluids i and j, i.e., (ṁcp)i/(ṁcp)j, respectively.
Boundary conditions are also required to solve the above energy equations. For Case P1, the
boundary conditions for the three fluid streams are given by θ1(0) = 0; θ2(0) = 1; and θ3(0) = θ3,in.
For Case P2, θ1(0) = 0; θ2(1) = 1; and θ3(0) = θ3,in. For Case P3, θ1(0) = 0; θ2(1) = 1; and θ3(1) = θ3,in.
Lastly, for Case P4, θ1(0) = 0; θ2(0) = 1; and θ3(1) = θ3,in.
Then solving the three energy equations for the three unknowns, θ1, θ2, and θ3 yields:
3
i a exp D x
j 1
ij j
(9.162)
B B2 4 AC B B2 4 AC
D1 0; D2 ; D2 (9.163)
2C 2C
where
1 1 R2
B 1 2 R1 R2 1 3C32 1 R1 2 3C32 (9.165)
NTU1 C12
1 2 3C32
C (9.166)
NTU12C12
R1 R2 R1R2 2 D j R2 / NTU1C12
bj (9.167)
R1 R2 R1R2 1D j R1 /NTU1
432 Advanced Heat Transfer
1 1 D j
cj 1 R2 b j 1 (9.168)
R2 NTU1
where j = 1, 2, 3. Also, define the variables dj and ej as follows: dj = cj (cases P1 and P2); dj = cj
exp(Dj) (cases P3 and P4); ej = 1 (cases P1 and P4); and ej = exp(Dj) (cases P2 and P3). Then the
coefficients aij can be expressed as
=a1 j a=
2 jbj ; a3 j a2 j c j (9.169)
Numerical simulations for a range of flow configurations were presented by Shrivastava and
Ameel (2004). Sample results for Case P2 are illustrated in Figure 9.20. The effects of varying R2
are shown for fixed parameter values of R1 = 2.0, C12 = 0.5, C32 = 0.8, NTU1 = 1.0 and θ3,in = 0.5 in
Figure 9.20 (a). Recall R2 is the ratio of thermal resistances between fluid streams 1 and 2 and fluid
streams 1 and 3.
For Case P2, two streams (1 and 3) are flowing in one direction and another (2) flows in the
opposite direction. All of the streams interact with each other and affect each other’s temperature
distribution. As the value of R2 increases, the difference between the temperatures of fluids 1 and
3 decreases as the thermal resistance between streams 1 and 3 is reduced. Thus the temperature of
fluid 1 approaches fluid 3. Since R2 > 1 in Figure 9.20(a), the thermal resistance between fluids 2 and
3 is less than the overall resistance between 1 and 2, so the temperature of fluid 2 follows a similar
trend as steam 3.
In Figure 9.20(b), recall that θ3,in was defined as the ratio of the temperature difference between
the inlet temperatures of fluids 1 and 3 to fluids 1 and 2. Parameter values in Figure 9.20(b) are speci-
fied as R1 = 2.0, R2 = 1.5, C12 = 0.8, C32 = 0.5, and NTU1 = 1.0. A higher value of θ3,in implies more
FIGURE 9.20 Effects of (a) R2 and (b) θ3,in on the temperature distribution for Case P2. (adapted from
Shrivastava and Ameel, 2004)
Heat Exchangers 433
heat transfer to fluid 1 and a lower temperature difference between fluids 1 and 2. As the value of θ3,in
increases, there is a temperature rise in the other two fluid streams, 1 and 2. Conversely, when θ3,in
decreases, fluids 1 and 2 cool down further and therefore the temperature of fluid 1 is reduced. Various
other physical trends can be inferred from the model with other varying parameters of NTU1, C12, and
C23. As NTU1 increases, the temperature difference increases between the inlet and outlet tempera-
tures of the three fluids. Also, the axial temperature change in fluid 1 decreases at larger values of C12.
FIGURE 9.21 (a) Multichannel and (b) finned passages of a compact heat exchanger.
434 Advanced Heat Transfer
TABLE 9.1
Correlations of Convective Boiling in Heat Exchangers
Authors Correlation Flow Conditions and Parameters
Robertson (1983) F 7X 0.36
tt
Annular flow; vertical perforated fins; 60 ≤ G
≤ 115 kg/m2s; 0.03 ≤ χg ≤ 0.8; 1.45 ≤ Prl ≤ 2
Mandrusiak and Carey (1989) Turbulent annular flow; vertical fin geometry; 1
0.372
F 1 28 /X tt2 Prl0.296
≤ 1/Xtt ≤ 60; 1.8 ≤ Prl ≤ 8.6; constants A and n
Xtt = (μl/μg)n/2(ρg/ρl)1/2(1/χg − 1)1 − n/2
selected by curve fits to single-phase data
Nul hl Dh /kl A Prl1/3 Re1l n
Troniewski and Witczak (1988) F = 2.62(D/Dh)1.03K0.06(1/Xtt)0.66 Turbulent annular flow; rectangular channels;
Xtt = (μl/μg)0.1(ρg/ρl)0.5(1/χg − 1)0.9 Pr = 1.7; H = ratio of sides in a rectangular
Nul = 0.023 Prl0.33 Rel0.8 duct; K = H (horizontal duct) or 1 (vertical)
D/Dh = H−0.5(H + 1/H + 2)0.5
Kenning and Cooper (1989) F 1 1.8 X tt0.79 Turbulent annular flow; circular tube
Nul = 0.023 Prl0.4 Rel0.8
Xtt = (μl/μg)0.1(ρg/ρl)0.5(1/χg − 1)0.9
Wattelet et al. (1992) F = 3.37(1/Xtt)0.686 Turbulent annular flow; horizontal circular
Xtt = (μl/μg)0.1(ρg/ρl)0.5(1/χg − 1)0.9 tube; 2 ≤ 1/Xtt ≤ 13; 3 ≤ Prl ≤ 4
For flow boiling in compact heat exchangers, typically both modes of nucleate and convective boil-
ing occur in the channels or tubes. Heat transfer can be predicted by individual correlations in each of
these flow regimes, or else combined correlations that account for both heat transfer mechanisms. Table
9.2 summarizes past correlations which have been developed for these combined modes of heat trans-
fer, including the heat transfer coefficients for both nucleate boiling, hnb, and convective boiling, hc.
TABLE 9.2
Combined Nucleate and Convective Boiling Correlations
Authors Correlation Flow Conditions and Parameters
Liu and Winterton (1991) 1/ m Turbulent vertical or horizontal liquid flow;
h Shnb hcm
m
The variables pred, M, q″, χg, Fr, h, hnb, and hc refer to the reduced pressure (p/pcr where pcr is
the critical pressure), molecular weight, wall heat flux, vapor mass fraction, Froude number (= ṁ/
ρ2gDh)), total heat transfer coefficient, nucleate boiling heat transfer coefficient, and convective boil-
ing heat transfer coefficient, respectively.
The trends of the heat transfer coefficient are significantly different between the convective and
nucleate boiling regions. In the convective boiling region, the flow pattern is either intermittent with
slug or churn flow patterns at low vapor fractions, or else steady annular flow at higher mass fluxes.
The parameters with the strongest influence on the flow pattern are the geometry, mass flux, and
vapor fraction. In the nucleate boiling region, the flow pattern of isolated or confined bubbles in the
tube is most significantly affected by the heat flux and channel dimensions.
TABLE 9.3
Flow Boiling Correlations for Helically Coiled Tube Heat Exchangers
Authors Correlation Flow Conditions and Parameters
Crain and Bell (1973) htp/hg = 1/Xtt Turbulent flow boiling; steam/water; 35 < q″ < 85
Xtt = (μl/μg)0.1(ρg/ρl)0.5(1/χg − 1)0.9 kW/m2; 0.137 < p < 0.234 MPa; 29 < G < 34 kg/m2s;
Nug 0.023 Re 0D.85 Pr 0.4 Dt /Dc
0.1 0.7 < χg < 1
Kozeki et al. (1970) htp/hg = 2.5(1/Xtt)0.75 Vertical flow; steam/water; 0.032 < δ < 0.035; 0.5 <
p < 2.1 MPa; 151 < q″ < 348 kW/m2; 161 < G < 486
kg/m2s; 0 < χg < 1; 628 < Dc < 682 mm; Dt = 21.7 mm
Yi et al. (2003) 1.737 10 6 Re3.5939 Bo 2.235 Laminar vertical flow; steam/water; 0.04 < δ < 0.062;
Nutp 5 < q″ < 39 kW/m2; 977 < Wo < 1,322; 0.5691 < Bo
Pr 0.8809 Wo 4.1044 FR 9.4881
q″ < CHF < 1.5584; 0.89 < Pr < 2.71; 221 < Re < 1,749; 0.11 <
FR < 0.18; 64.2 < Dc < 100.2 mm; Dt = 4 mm
Yi et al. (2003) 15.3188 Bo1.6674 FR 2.2973 Laminar and turbulent vertical flow; steam/water;
Nutp = 0.04 < δ < 0.062; 10 < q″ < 62 kW/m2; 616 < Wo
Re 0.0373 Pr 2.968 Wo3.7535
q″ > CHF < 2,044; 0.5691 < Bo < 1.5584; 0.89 < Pr < 3.103;
3,393 < Re < 4,694; 0.11 < FR < 0.18; 64.2 < Dc <
100.2 mm; Dt = 4 mm
Bai and Guo (1997) htp/hs = 1 + 2.21(1/Xtt)0.33; 1/Xtt < 1.2 Turbulent horizontal flow; steam/water; δ = 0.043;
htp/hs = 3.06(1/Xtt)0.47; 1/Xtt ≥ 1.2 0.5 < p < 3 MPa; 230 < q″ < 500 kW/m2; 200 < G <
2,500 kg/m2s; 0.7 < χg < 1; Dt = 11 mm; Dc = 255 mm
Wongwises and 23259.27 Deq 0.432
Bo 0.132 Vertical tube flow; R–134a; δ = 0.025;
Polsongkram (2006) Nutp = 5.055 0.0238 5 < q″ < 10 kW/m2; 400 < G < 800 kg/m2s;
Prl X tt
Deq = [Rel + Reg(μg/μl)(ρl/ρg)0.5](Dt/Dc)0.5 10oC < T < 20oC; 0 < χg < 1; Dc = 305 mm;
Rel = G(1 − χg)D/μl Dt = 7.2 mm
Reg = GχgD/μg
Chen et al. (2011)
htp /hs 2.84 X tt0.27 4,616 Bo1.15 0.88 Horizontal flow; R–134a; δ = 0.025; 0.2 < p < 0.75
MPa; 0.115 < q″ < 2.1 kW/m2; 50 < G < 260 kg/m2s;
Xtt = (μl/μg)0.1(ρg/ρl)0.5(1/χg − 1)0.9 0.18 < χg < 0.4; Dc = 300 mm; Dt = 7.6 mm
Nug 0.023 Re 0D.85 Pr 0.4 Dt /Dc
0.1
Kang et al. (2000) Nutp = 2.3(Re⋅)0.94Pr0.4 Laminar and turbulent vertical flow; condensation;
Re G g Dt g /l /l R–134a; δ = 0.075; 100 < G < 400 kg/m2s; T = 33oC;
1,500 < Re < 9,000; 0 < χg < 1; Dc = 177.8 mm;
Dt = 12.7 mm
in slug flow as a result of the pulsation effects. These effects are expressed through a dependence on
the Wo number in the flow boiling correlations in Table 9.3.
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 analysis of two-phase heat
exchanger design, operation, and safety are presented by Fraas (1989), Hewitt (1998), Kakac, Liu,
and Pramuanjaroenkij (2012), Penoncello (2015).
Heat Exchangers 437
Tds dh dp (9.172)
where s and h refer to the specific entropy and enthalpy, respectively. By substituting dh = cp dT
(Chapter 1) and υ = RT/p (ideal gas law), the entropy change for a process can be obtained by
integrating the Gibbs equation between states 1 (initial or inlet) and 2 (final or exit state), yielding:
T p
s2 s1 c p ln 2 R ln 2 (9.173)
T1 p1
where R is the gas constant. Also, the second law of thermodynamics for an open system (control
volume) can be written as
S Q
t T ms ms P
in out
s (9.174)
FIGURE 9.23 Schematic of (a) counterflow streams and (b) temperature distributions.
438 Advanced Heat Transfer
m,
where S, s, Q, and Ps refer to entropy (J/K), specific entropy (J/kgK), heat transfer rate (W), mass
flow rate (kg/s), and rate of entropy production (W/K). Under steady-state conditions, for an exter-
nally insulated system, the second law for the counterflow heat exchanger becomes:
T T p p
Ps m 1c p,1 ln 1,e m 2c p,2 ln 2,e m 1R1 ln 1,e m 2 R2 ln 2,e (9.175)
T1 T2 p1 p2
where the subscript e refers to exit or outlet. The first two terms on the right side refer to the thermal
irreversibility, while the last two terms account for fluid friction irreversibilities.
Define the effectiveness ratio, ε, and number of transfer units, NTU, as follows:
T1 T1,e T2,e T2
(9.176)
T1 T2 T1 T2
UA
NTU = (9.177)
p
mc
where the overall heat transfer coefficient, U, neglecting conduction wall resistances between the
two gas streams, is given by
1 1 1
(9.178)
UA UA 1 UA 2
Assuming a “balanced” heat exchanger with the same thermal capacity flow rate on each side of
the heat exchanger (ṁ1cp,1 = ṁ2cp,2 = ṁcp), it can be shown that:
NTU
(9.179)
1 NTU
Also, assuming (1 – ε) ≪ 1 and (Δp/p)1,2 ≪ 1, the entropy generation rate can be simplified and
nondimensionalized as a sum of irreversibilities associated with each side of the heat exchanger as
follows:
T2 T1
2
Ps
2
R p
Ns
p
mc
i 1
N s ,i ; N s,i N st ,i N sf ,i
NTUiT1T2
c p p i
(9.180)
where the subscripts i, t, and f refer to the index number for each side of the heat exchanger, ther-
mal, and friction, respectively. From this result, it can be observed that the thermal irreversibility
approaches zero when the area becomes large (NTU → ∞). As expected, the temperature difference
required to transfer a specified rate of heat flow, as well as the corresponding thermal irreversibility,
becomes smaller when NTU increases since there is a larger surface area available to transfer the
fixed amount of heat flow. The friction irreversibility also decreases, as expected, when the pressure
drop approaches zero.
Define the mass velocity, Gi, on side i, and its corresponding nondimensional value as
m i Gi
Gi ; Gi (9.181)
A 2 pi
Also, recall the Stanton number, St, and friction factor, f, as defined previously in Chapter 3,
Heat Exchangers 439
NuD
St = (9.182)
Re D Pr
p U2
f (9.183)
L 2 Dh
where L, U, and Dh refer to the tube length, mean flow velocity, and hydraulic diameter, respectively.
Then the nondimensional entropy generation rate, Ns,i, on side i, can be expressed as
T 2 Dh R 4L
N s ,i Gi2 fi (9.184)
Sti 4 L c p Dh
where
T2 T1
2
T 2 (9.185)
T1T2
Note that the Stanton number is a function of the Reynolds and Prandtl numbers, St = St (ReD,
Pr), while the friction factor is a function of the Reynolds number, f = f (Re).
The entropy generation is a sum of thermal and friction irreversibilities on each side of the heat
transfer surface. It varies with several parameters that depend on the working fluids (Pr, R/cp), inlet
conditions (T1, T2), operating conditions (ReD, Gi), and geometrical parameters of the heat exchanger
(L, D, aspect ratio 4L/D). The entropy generation rate can be minimized with respect to each of
these parameters by differentiating Ns with respect to the prescribed parameter, holding the other
parameters fixed, setting the resulting expression to zero (for minimization), and finally determining
the optimal design points.
Sample results of the optimization are summarized in Table 9.4 for four different cases: 1) opti-
mal flow path length, Lopt (or tube aspect ratio 4L/Dh) at a fixed mass velocity; 2) optimal aspect ratio
at a fixed NTU; 3) optimal mass velocity; and 4) optimal Reynolds number subject to both area and
volume constraints of the heat exchanger. When the subscript i is removed (for brevity), it refers to
TABLE 9.4
Entropy Generation Number for Four Optimization Cases
Constraint Optimized Parameter Entropy Generation Number
1) Mass velocity constraint 4L T 1/ 2
R f
N s,min 2TG
Dh opt G R /c p f St
1/ 2
c p St
FIGURE 9.24 Entropy generation number for compact crossflow heat exchanger with T1/T2 = 0.8. (adapted
from Bejan, 1987)
either side of the heat exchanger. Sample results for Case 2 (NTU constraint) are also illustrated in
Figure 9.24.
In Table 9.4, the nondimensional area, A*, and volume, V*, are defined as
2 p 8p
A A; V V (9.186)
m m
For a fixed number of heat transfer units, the results in Figure 9.24 show that the entropy genera-
tion number decreases to a minimum and then increases thereafter with aspect ratio (or slenderness
ratio). As NTU increases, the optimal aspect ratio also increases. The total entropy production in the
heat exchanger varies with both thermal and friction irreversibilities, which are coupled between
both sides of the heat exchanger. At a fixed value of NTU, there exists an optimal pair of aspect ratios
for which the entropy generation number is minimized.
C
mc
p
max
Cmax
(9.187)
mc
p
min
Cmin
T1 T1,e T T
C 2,e 2 (9.188)
T1 T2 T1 T2
From an energy balance, the heat transfer between the fluid streams is given by
FIGURE 9.25 Entropy generation number versus effectiveness for (a) T2/T1 = 1.2 and (b) T2/T1 = 1.5. (adapted
from Hesselgreaves, 2000)
By a similar procedure as the previous section, while neglecting the pressure drop contribution in
the total entropy production, it can be shown that:
T1Ps T1 T2 T1
N s,1 ln 1 1 C ln 1 1
(9.190)
Q T2 T1 T1 C T2
Hesselgreaves (2000) analyzed several flow configurations and sample results for ratios of T2/
T1 = 1.2 and 1.5 are presented in Figure 9.25. The highest capacity flow rate, ṁcp, is in the hot fluid
stream. At a given effectiveness, the entropy generation number increases with C*. The results indi-
cate that flow imbalance increases the entropy generation.
The value of Cs,1 approaches zero in the limit of ε → 1 only for the balanced flow case of C* = 1.
Also, it can be observed that the entropy generation number increases at higher temperature ratios
of heat exchange. Note that the results are general and independent of any heat exchanger flow
arrangement since the results are expressed in terms of the effectiveness, ε. Results for specific heat
exchanger configurations can be obtained by substituting the appropriate ε–NTU relationship.
Substituting this expression into Equation (9.190), the entropy generation number becomes:
N s,1
ln 1 1 exp NTU T2 /T1 1 T1
(9.192)
1 exp NTU T2 /T1 1 T2
Similarly, in the case of an evaporating liquid on one side, C* → 0. The ε–NTU relation is the
same as the above case of condensation but the entropy generation number is obtained as
442 Advanced Heat Transfer
FIGURE 9.26 Entropy generation number for (a) condensation and (b) evaporation on one side. (adapted
from Hesselgreaves, 2000)
N s,1
ln 1 1 exp NTU 1 T1 /T2
1 (9.193)
1 exp NTU T2 /T1 1
Sample results of the entropy generation number at varying T2/T1 ratios are illustrated in
Figure 9.26. A significant difference is observed between the condensation and evaporation cases.
As the ratio T2/T1 increases at a given value of NTU, the value of Ns,1 increases more rapidly for the
evaporation case. In both cases, Ns,1 decreases with NTU as a result of a higher overall heat transfer
coefficient, U, thereby reducing the temperature differences and mass flow rate, which leads to a
lower friction irreversibility.
FIGURE 9.27 (a) Finned and (b) peripheral finned tube configurations in a heat exchanger.
Heat Exchangers 443
Consider a crossflow configuration with a hexagonal arrangement of open-pore cells and six radial
fins with bases attached to the tubes. The tips are connected to the six peripheral fins. Combining
an energy balance, the second law of thermodynamics, and the Gibbs equation, as developed previ-
ously in Chapter 3, the entropy production rate per unit length of the heat exchanger becomes:
q Tw Ta m a dp
Ps Ps,t Ps, f (9.194)
Ta2 aTa dx
where the subscripts t, f, w, and a refer to thermal, friction, wall, and air, respectively. Two cases of
boundary conditions in the interior fins will be considered: 1) constant wall temperature; and 2) constant
wall heat flux. First, for a constant wall temperature, the fin heat transfer rate, q′, is rewritten in terms of
an effective, or equivalent, fin efficiency, ηf (Chapter 2), such that the thermal irreversibility becomes:
2
T T dA
Ps,t f h w a (9.195)
Ta dx
where A is the fin surface area. The heat transfer coefficient, h, is rewritten in terms of the Stanton
number as follows:
NuD h hAc
St (9.196)
Re D Pr Gc p,a m a c p,a
where ζ and Ac are the cell porosity and cross-sectional area, respectively. The model treats the
airflow paths between the open cells as a porous medium with a porosity of ζ and equivalent par-
ticle diameter of Dp. The particle diameter represents an average characteristic length of the porous
medium, as defined by the interstitial area, A, per unit solid volume, Vs, where:
6Vs
Dp = (9.197)
A
The porosity of the packed structure of finned tubes is defined as the ratio of the fluid volume to
the total volume,
V Vs V
1 s (9.198)
V V
Also, it can be shown that the rate of change of surface area with distance in the flow (x) direction
can be simplified as
1 dA 6
Ac dx D p
1 (9.199)
Then using Darcy’s law of porous media flow and the Blake–Kozeny equation (Chapter 7), the
pressure gradient in the porous medium can be expressed as
dp U D2 1
f (9.200)
dx Dp 3
where UD is called the Darcian superficial velocity, or alternatively, the flow rate per unit cross-
sectional area of the porous medium.
444 Advanced Heat Transfer
Substituting these expressions into Equation (9.194) and integrating over the length, L, of the
heat exchanger, yields:
L 3 L
Ps
2
6St 1 Tw Ta Re 2D a2 1 dx
Ns
m a c p,a
f
D p Ta
0
dx f
D 3pc p,a
T
0
a
(9.201)
where the Reynolds number and air-side temperature distribution are given by
U D Dp
Re D (9.202)
1 a
Tw Ta hA x
exp f (9.203)
Tw Ta,i m a c p,a L
The second case is a fin boundary condition of a constant wall heat flux, q″. The thermal irrevers-
ibility per unit length of heat exchanger can be written as
q2 dA
Ps,t (9.204)
f hTa2 dx
Ns
f Pr ka St 1 Re 2D
2 2
0
Ta2
f 3D a
D p c p, a
T
0
a
(9.205)
where the air-side temperature distribution for this constant flux case is given by
6q
Ta Ta,i x (9.206)
Pr ka Re D
Sample results for the constant wall temperature (CWT) and constant wall heat flux (CWF) cases
are illustrated in Figure 9.28. In both cases, q = 300 W, ηf = 0.8, Ac = 0.008 m2, ζ = 0.85, Ta,in = 0 oC,
and the heat exchanger length is L = 0.1123 m. Also, a fixed geometry was maintained in both cases,
whereby the face area and heat exchanger length are held fixed during the optimization.
In Figure 9.28(a) for the CWT case, the critical value of ReD at the minimum Ns is observed to
increase with the particle diameter, Dp. An increase of Dp requires a reduction of the air-side sur-
face area, for example, by reducing the length, L, or increasing the fin thickness, while retaining a
constant solid volume. At a given ReD, a larger Dp implies a smaller superficial air velocity. Also, at
a prescribed face area, if the mass flow rate decreases, then the outlet air temperature will increase.
When the total heat transfer rate is fixed, and the air temperature increases, the overall thermal irre-
versibility will increase also.
Similar trends are observed in Figure 9.28 for the CWF case. As Dp increases, the surface area
decreases to maintain a fixed porosity and heat transfer rate. Also, the superficial velocity decreases
at a higher Dp for a given ReD. Therefore, similarly as the CWT case, a larger Dp leads to an increase
in Ns due to heat transfer.
As these previous examples have shown, the method of entropy generation minimization is a ver-
satile tool for the optimization of geometric and flow parameters in heat exchangers. Optimization of
various other configurations such as cross-flow (cases of streams mixed and unmixed), counterflow,
Heat Exchangers 445
FIGURE 9.28 Entropy generation number for (a) constant wall temperature and (b) constant wall heat flux
boundary conditions. (adapted from Pussoli et al., 2012)
parallel flow, and two-phase flows have been presented by Bejan (1987, 1996). Extensions of the
method to include CFD (computational fluid dynamics) were presented by Naterer and Camberos
(2008).
REFERENCES
Bai, B.F. and L.J. Guo. 1997. “Study on Convective Boiling Heat Transfer in Horizontal Helically Coiled
Tubes”, Chinese Journal of Nuclear Science and Engineering, 17: 302–308.
Bejan, A. 2013. Heat Transfer, 4th Edition, New York: John Wiley & Sons.
Bejan, A. 1987. “The Thermodynamic Design of Heat and Mass Transfer Processes and Devices”, International
Journal of Heat and Fluid Flow, 8(4): 258–276.
Bejan, A. 1996. Entropy Generation Minimization: The Method of Thermodynamic Optimization of Finite–Size
Systems and Finite–Time Processes, Boca Raton: CRC Press/Taylor & Francis Group.
Bergman, T.L., A.S. Lavine, F.P. Incropera and D.P. DeWitt. 2011. Fundamentals of Heat and Mass Transfer,
7th Edition, New York: John Wiley & Sons.
Bowman, R.A., A.C. Mueller and W.M. Nagle. 1940. “Mean Temperature Difference in Design”, Transactions
of ASME, 62: 283–294.
Brinkman, H.C. 1949. “A Calculation of the Viscous Force Exerted by a Flowing Fluid on a Dense Swarm of
Particles”, Applied Scientific Research, 1: 27–34.
Brogan, R.J. 2011. Thermopedia: Heat Exchangers, Oxfordshire, UK: Harwell Laboratory.
Buongiorno, J. 2006. “Convective Transport in Nanofluids”, ASME Journal of Heat Transfer, 128(3): 240–250.
Chen, C.N., J.T. Han, T.C. Jen, and L. Shao. 2011. “Thermo–chemical Characteristics of R134a Flow Boiling
in Helically Coiled Tubes at Low Mass Flux and Low Pressure”, Thermochimica Acta, 512: 163–169.
Crain, B. and K.J. Bell. 1973. “Forced Convection Heat Transfer to a Two-phase Mixture of Water and Steam
in a Helical Coil”, American Institute of Chemical Engineers Symposium Series, 69: 30–36.
Fraas, A.P. 1989. Heat Exchanger Design, 2nd Edition, New York: John Wiley & Sons.
Fsadni, A.M. and J.P. Whitty. 2016. “A Review on the Two-phase Heat Transfer Characteristics in Helically
Coiled Tube Heat Exchangers”, International Journal of Heat and Mass Transfer, 95: 551–565.
Gu, S., T.J. Lu and A.G. Evans. 2001. “On the Design of Two-dimensional Cellular Metals for Combined Heat
Dissipation and Structural Load Capacity”, International Journal of Heat and Mass Transfer, 44(11):
2163–2175.
Harrison, J. ed., 1999. Standards of the Tubular Exchange Manufacturers Association, 8th Edition, Tarrytown,
NY: Tubular Exchange Manufacturers Association.
Hesselgreaves, J.E. 2000. “Rationalisation of Second Law Analysis of Heat Exchangers”, International Journal
of Heat and Mass Transfer, 43: 4189–4204.
446 Advanced Heat Transfer
Hewitt, G.F. ed., 1998. Heat Exchanger Design Handbook, New York: Begell House.
Isaza, P.A., A. O’Brien, W.D. Warnica, M. Bussmann. 2017. “Assessing Axial Heat Conduction in Moving Bed
Heat Exchangers”, International Journal ot Thermal Sciences, 120: 303–313.
Kakac, S., H. Liu and A. Pramuanjaroenkij. 2012. Heat Exchangers, 3rd Edition, Boca Raton: CRC Press/
Taylor & Francis.
Kang, H.J., C.X. Lin and M.A. Ebadian. 2000. “Condensation of R134a Flowing inside a Helicoidal Pipe”,
International Journal of Heat and Mass Transfer, 43: 2553–2664.
Kays, W.M. and A.L. London. 1984. Compact Heat Exchangers, 3rd Edition, New York: McGraw–Hill.
Kenning, D.B. and M.G. Cooper. 1989. “Saturated Flow Boiling of Water in Vertical Tubes”, International
Journal of Heat and Mass Transfer, 32(3): 445–458.
Kozeki, M., H. Nariai, T. Furukawa, and K. Kurosa, 1970. “A Study of Helically Coiled Tube Once–Through
Steam Generator”, Bulletin of JSME, 66: 1485–1494.
Liu, H., Q.N. Yu, Z.C. Zhang, Z.G. Qu and C.Z. Wang. 2016. “Two-equation Method for Heat Transfer
Efficiency in Metal Honeycombs: An Analytical Solution”, International Journal of Heat and Mass
Transfer, 97: 201–210.
Liu, S., Y. Zhang and P. Liu. 2008. “New Analytical Model for Heat Transfer Efficiency of Metallic Honeycomb
Structures”, International Journal of Heat and Mass Transfer, 51(25): 6254–6258.
Liu, Z. and R.H.S. Winterton. 1991. “A General Correlation for Saturated and Subcooled Flow Boiling in
Tubes and Annuli Based on a Nucleate Pool Boiling Equation”, International Journal of Heat and Mass
Transfer, 34(11): 2759–2766.
Lu, T.J. 1999. “Heat Transfer Efficiency of Metallic Honeycombs”, International Journal of Heat and Mass
Transfer, 42(11): 2031–2040.
Lu, W., T. Zhang and M. Yang. 2016. “Analytical Solution of Forced Convective Heat Transfer in Parallel Plate
Channel Partially Filled with Metallic Foams”, International Journal of Heat and Mass Transfer, 100:
718–2016.
Mandrusiak, G.D. and V.P. Carey. 1989. “Convective Boiling in Vertical Channels with Different Strip Fin
Geometries”, ASME Journal of Heat Transfer, 111: 156–165.
Mathew, B. and H. Hegab. 2011. “Modeling Non–adiabatic Parallel Flow Microchannel Heat Exchangers”,
International Journal of Thermal Sciences, 50: 350–360.
Nabity, J.A. 2014. “Modeling a Freezable Water–Based Heat Exchanger for Use in Spacecraft Thermal
Control”, AIAA Journal of Thermophysics and Heat Transfer, 28: 708–716.
Naterer, G.F. and C.H. Lam. 2006. “Transient Response of Two–Phase Heat Exchanger with Varying Convection
Coefficients”, ASME Journal of Heat Transfer, 128: 953–962.
Naterer, G.F. and J.A. Camberos. 2008. Entropy–Based Design and Analysis of Fluids Engineering Systems,
Boca Raton: CRC Press/Taylor & Francis Group.
Penoncello, S.G. 2015. Thermal Energy Systems: Design and Analysis, Boca Raton: CRC Press/Taylor &
Francis Group.
Pussoli, B.F., J.R. Barbosa, L.W. da Silva, and M. Kaviany. 2012. “Optimization of Peripheral Finned Tube
Evaporators using Entropy Generation Minimization”, International Journal of Heat and Mass Transfer,
55: 7838–7846.
Qu, Z.G., W.Q. Li and W.Q. Tao. 2014. “Numerical Model of the Passive Thermal Management System for
High Power Lithium Ion Battery by using Porous Metal Foam Saturated with Phase Change Material”,
International Journal of Hydrogen Energy, 39: 3904–3913.
Robertson, J.M. 1983. “The Boiling Characteristics of Perforated Plate–Fin Channels with Liquid Nitrogen
in Upflow”, Heat Exchangers for Two–Phase Applications, Proceedings of the ASME Heat Transfer
Division, pp. 35–40.
Shrivastava, D. and T.A. Ameel. 2004. “Three–Fluid Heat Exchangers with Three Thermal Communications,
Part A: A General Mathematical Model”, International Journal of Heat and Mass Transfer, 47:
3855–3865.
Troniewski, L. and S. Witczak. 1988. “Heat Transfer Boiling through Rectangular Ducts”, in: T.N. Veziroglu
(Ed.), Particulate Phenomena and Multiphase Transport, pp. 89–97, New York: Hemisphere Publishing
Corp.
Vermahmoudi, Y., S.M. Peyghambarzadeh, M. Naraki and S.H. Hashemabadi. 2013. “Statistical Analysis
of Nanofluid Heat Transfer in a Heat Exchange System”, AIAA Journal of Thermophysics and Heat
Transfer, 27: 320–325.
Watel, B. 2003. “Review of Saturated Flow Boiling in Small Passages of Compact Heat Exchangers”,
International Journal of Thermal Sciences, 42: 107–140.
Wattelet, J.P., J.C. Chato, A.L. Souza, and B.R. Christoffersen. 1994. “Evaporative Characteristics of R12,
R134a, and a Mixture at Low Mass Fluxes”, ASHRAE Transactions, 100: 603–615.
Heat Exchangers 447
Wattelet, J.P., J. M. Saiz Jabardo, J. C. Chato, J. S. Panel, and A.L. Souza. 1992. “Experimental Investigation
of Convective Boiling of Refrigerants HFC 134a and CFC 12”, Two–Phase Flow and Heat Transfer,
Proceedings of ASME Heat Transfer Division, 197: 121–127.
Wongwises, S. and M. Polsongkram. 2006. “Evaporation Heat Transfer and Pressure Drop of HFC–134a in a
Helically Coiled Concentric Tube–in–Tube Heat Exchanger”, International Journal of Heat and Mass
Transfer, 49: 658–670.
Xu, H.J., L. Gong, S.B. Huang and M.H. Xu. 2015. “Flow and Heat Transfer Characteristics of Nanofluid
Flowing Through Metal Foams”, International Journal of Heat and Mass Transfer, 83: 399–407.
Yi, J., Z.H. Liu and J. Wang. 2003. “Heat Transfer Characteristics of the Evaporator Section using Small Helical
Coiled Pipes in a Looped Heat Pipe”, Applied Thermal Engineering, 23(1): 89–99.Placeholder Text
PROBLEMS
1. In a cross-flow heat exchanger, water flows over a copper pipe (1.9 cm inner diameter; 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 overall heat transfer
coefficient of 140 W/m2K per unit length of pipe (based on the inner tube area).
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?
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 coeffi-
cients for the air and water sides of the pipe are 120 and 2,400 W/m2K, respectively. 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%?
4. Water flows across aluminum pipes in a cross-flow heat exchanger. The pipe has a 2.5 cm
inner diameter and 3 cm outer diameter. Oil flows inside the pipe. The convection coeffi-
cients for water and oil flows are 230 W/m2K and 310 W/m2K, respectively. Determine the
overall heat transfer coefficient.
5. A tubular heat exchanger operates in a counterflow arrangement. A design modification
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 convec-
tion coefficient of the external flow in the outer annulus. Assume that the external convec-
tion coefficient remains approximately identical under both flow conditions.
6. 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 temperature of 25°C
through the inner pipe. The thermal resistances due to steam convection 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.
7. In a concentric counter-flow heat exchanger with a 20 mm inner diameter and 40 mm outer
diameter, water enters at 15°C with a mass flow rate of 0.45 kg/s. It exits with a tempera-
ture of 55°C. The water is heated by oil that flows at 120°C and exits with a temperature of
65°C at a 0.75 kg/s mass flow rate. It is known that the specific heat of oil is 2.2 kJ/kgK,
water convection coefficient is 180 W/m2K, and the oil convection coefficient is 300 W/
m2K. Estimate the length of the tubular heat exchanger.
8. In a concentric tube heat exchanger, with a total surface area of 30 m2, water flows at 50°C
and exits at 160°C.The mass flow rate of the water is 2 kg/s. Hot gases are heating the
water. The gases enter at 375°C and exit at 180°C. The specific heat of the hot gas is 1,000
J/kgK and the mass flow rate is 3 kg/s. Estimate the overall heat transfer coefficient.
448 Advanced Heat Transfer
9. A set of finned tubes is arranged uniformly in rows and placed normal to an incoming air-
flow 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 diam-
eter 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).
10. A cross-flow heat exchanger in a power plant is used for intercooling between two com-
pressor 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 pressure 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 resis-
tance due to tube fouling is 10–5 °C/W.
11. A cross-flow heat exchanger is used to cool electronic equipment. The air passing through
the heat exchanger is heated from 20°C to 60°C (unmixed stream) by another stream
(mixed stream) entering at 95°C.The mass flow rates of cold and hot streams are 2 kg/s and
3 kg/s, respectively. Estimate the raised temperature of the cold stream when the percent-
age increase of the overall heat transfer coefficient is 40%.
12. Derive the expression in Equation (9.45) for the effectiveness of a concentric tube (parallel
flow) heat exchanger.
13. Water flows through tubes of a single-shell and two-tube pass heat exchanger in a chemical
refinery for cooling hydraulic oil. The pipe outer diameter, wall thickness, thermal conduc-
tivity, and length are 2.5 cm, 3 mm, 32 W/mK, and 5 m, respectively. The water mass flow
rate is 0.5 kg/s on the shell side. The inlet temperature is 290 K. Oil flows at 0.3 kg/s in the
tubes with a convection heat of 4,600 W/m2K. If the inlet and outlet temperatures of the oil
are 350 K and 330 K, respectively, calculate the convection coefficient of water.
14. 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.
15. Oil is cooled from 55°C to 35°C as it flows at a rate of 0.4 kg/s through a tube within a
single-shell-pass and four-tube-pass heat exchanger. On the shell side of the heat exchanger,
water enters at 10°C at a rate of 1 kg/s. Under modified operating conditions, the oil flow
rate is reduced to 0.3 kg/s. Up to what maximum temperature of oil can be introduced at
the inlet of the tube under the same water operating conditions, without exceeding an oil
outlet temperature of 25°C? Constant thermophysical properties (cp = 2,100 J/kgK for oil)
may be assumed.
16. 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.
Heat Exchangers 449
17. 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 con-
ductivity, 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 tempera-
ture 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.
18. 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 2 kg/s and 18 m2,
respectively. What overall heat transfer coefficient is required to yield a water outlet tem-
perature of 70°C on the shell side?
19. 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?
20. 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.
21. 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 condens-
ing at a temperature of Tsat. Find the number of tubes required to condense 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.
22. Discuss the main design features of importance when selecting a shell-and-tube heat
exchanger for condensing refrigerant fluid.
23. What methods can be used for heat transfer enhancement in heat exchangers? Explain the
methods and their effects on the heat exchanger performance.
24. 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 compactness of various
types of heat exchangers such as automotive radiators and plate heat exchangers.
25. 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?
26. 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 condensation of 60
g/s?
b. What temperature rise of the cooling water occurs in this tube configuration? The flow
rate is 0.1 kg/s per tube.
10 Computational Heat Transfer
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).
T 2T 2T
cp k 2 2 (10.1)
t x y
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 and
row numbers, respectively,
T 2T x 2 3T x 3
Ti 1, j Ti, j x 2 3 (10.2)
x i x i 2 x i 6
T 2T x 2 3T x 3
Ti 1, j Ti, j x 2 3 (10.3)
x i x i 2 x 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(Δx2) refers to the higher order terms with an order of magnitude proportional to Δx2.
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 approxi-
mation 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
4
Ti 1, j Ti 1, j Ti, j 1 Ti, j 1 (10.6)
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
Qn Qw Qs Qe 0 (10.7)
where Qn, Qw, Qs, and Qe denote the heat flows across the north, west, south, and east edges, respec-
tively, 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, Equation (10.7)
becomes:
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).
Computational Heat Transfer 453
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.
T=
1 T=
2 T3 = 240 (10.9)
T=
7 T=
8 T=
9 120 (10.10)
Q2 5 Q4 5 Q6 5 Q8 5 0 (10.11)
T T T T T5 T6 T5 T8
k 0.1 5 2 k 0.1 5 4 k 0.1 0.1 k 0.1 0.1 0 (10.12)
0 .1 0 .1
At boundary node 4, the convective cooling condition is applied through Newton’s law of cool-
ing at the edge of the control volume,
T T T T T4 T5
60 0.1 40 T4 k 0.05 4 1 k 0.05 4 7 k 0.1 0.1 0 (10.15)
0 .1 0 .1
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 solvers, such
as Gaussian elimination, iterative, or matrix inversion methods (Cheney and Kincaid, 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 corner), is given
454 Advanced Heat Transfer
by Q123 = 813.3 W/m. Similarly, the heat flows across the bottom and left boundaries are calcu-
lated 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 overall 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 depen-
dent 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.
10.1.2 Transient Solutions
For transient problems, the temperature varies with spatial coordinates as well as time, t. Consider
one-dimensional transient heat conduction,
T 2T
cp k 2 (10.19)
t x
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),
Ti n 1 Ti n Ti n1 Ti n Ti n Ti n1
Ax c p t
kA kA (10.20)
x x
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),
Ti n 1 1 2 Fo Ti n Fo Ti 1n Ti 1n (10.21)
where Fo = αΔt/Δx2 is the Fourier modulus (nondimensional time) and α = k/ρcp is the thermal dif-
fusivity. Based on the first term on the right side of Equation (10.20), a time step stability 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.
Computational Heat Transfer 455
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 that
evaluates the diffusion flux terms at the current time level (n + 1) in Equation (10.19),
Ti n 1 Ti n Ti 1n 1 Ti n 1 Ti n 1 Ti 1n 1
Ax c p t
kA
x
kA
x
(10.22)
This formulation is implicit because the transient and diffusion flux terms involve temperatures
that are currently sought, rather than known from a previous time level. Rearranging this equation in
terms of the nondimensional Fourier number,
1 2 Fo Ti n 1 Fo Ti 1n 1 Ti 1n 1 Ti n (10.23)
In this case, there is not a stability limit involving the time step size, Δt. However, a set of simul-
taneous algebraic equations needs to be solved at each time step with coefficients in non-diagonal
entries of the solution matrix. Thus, there is a trade-off between more time steps (an explicit formu-
lation) or more computational effort required at each time step (implicit method). In most problems,
implicit solutions generally have more attractive benefits, robustness, and accuracy than explicit
methods.
v S 0 (10.24)
where ℒ, v, Γ and S refer to an operator (acting on its argument, such as ϕ), velocity vector, diffusion
coefficient, and source term, respectively. In general, a numerical solution seeks to find an approxi-
mate solution, φˆ , of the conservation equation. In the weighted residual method, a solution residual,
R(x), is introduced such that ℒ(ϕ) = R(x). The residual is an indicator of how well the governing
partial differential equation is satisfied.
The residual function, R(x), represents a difference obtained after substituting the exact solution,
ϕ, into the governing equation, in comparison to substitution of the approximate solution, φˆ . Using
a linear combination of appropriate basis functions, ϕn(x), to represent the approximate solution,
n
ˆ a11 x ann x a x
i 1
i i (10.25)
456 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 approximate
solution will precisely match the exact solution. As a result, the coefficients (a1, a2,…, an) are deter-
mined by minimizing the residual over the domain or setting the integrated and weighted value of
the residual to zero. The weighted residual method poses n constraints on the coefficients a1,…, an
through the following formulation:
W ˆ dV W RdV 0
V
i
V
i (10.26)
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 equa-
tions for the unknown coefficients a1,…, an. In practice, the unknown coefficients represent problem
variables such as temperature (energy equation) or velocity (momentum equations). The solution
procedure yields n integral equations for these n coefficients, leading 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 govern-
ing equation is satisfied exactly at the given point where the weight 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 element method will be presented in the next section.
10.2.2 Solution Procedure
The finite element method is a powerful numerical method that is used in many branches of sci-
ence and engineering including heat transfer, fluid mechanics, structural analysis, mass transfer, and
electromagnetics. The method subdivides a problem domain into smaller subregions called finite
elements. The discretized equations for each element are then assembled into a larger system of
equations for the entire problem. The finite element equations are determined from other methods,
such as the weighted residual method, to approximate a solution by minimizing the associated error
functions.
Unlike the finite difference method, the finite element method yields discretized equations that
are entirely local to an element. As a result, the algebraic equations are developed in isolation of
other elements and independent of the mesh configuration. In this way, finite elements can readily
accommodate unstructured and complex grids, unlike finite difference methods which require a
structured (i, j) grid format. As a result, the finite element method provides excellent geometric flex-
ibility. After the local element equations are formed, assembly rules are then required to reconstruct
the entire domain from its parts (individual elements).
The overall steps in the finite element method can be summarized as follows.
1. Discretize the problem domain by specifying the number of elements, shape of elements, 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 tem-
perature equations for a heat transfer analysis.
Computational Heat Transfer 457
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, 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
n 1
cmn Tn C
n 1
cmnTn C c
n 1
mn 0
(10.27)
where nnp refers to the number of nodal points. The first term on the right side is a summation
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.
a1 a1 x a3 y (10.28)
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 (10.28) in terms of the unknown
coefficients, it can be shown that
1
a1 x2 y3 y2 x3 1 x3 y1 x1 y3 2 x1 y2 x2 y1 3 (10.29)
2A
1
a2 y2 y3 1 y3 y1 2 y1 y2 3 (10.30)
2A
1
a3 x3 x2 1 y3 y1 2 x2 x1 3 (10.31)
2A
1
A
2
x2 y3 x3 y2 x1y3 x1y2 y1x3 y1x2 (10.32)
Rearranging terms, the interpolation in Equation (10.28) can be rewritten in terms of the shape
functions, N1, N2, and N3, as follows:
where
1
N1
2A
a1 b1x c1y (10.34)
1
N2
2A
a2 b2 x c2 y (10.35)
1
N3
2A
a3 b3 x c3 y (10.36)
and
a1 x2 y3 x3 y2 ; a2 x3 y1 x1 y3; a3 x1 y2 x2 y1 (10.37)
b1 y2 y3 ; b2 y3 y1; b3 y1 y2 (10.38)
c1 x3 x2 ; c2 x1 x3 ; c3 x2 x1 (10.39)
N1 N N
1 2 2 3 3 (10.40)
x x x x
N1 N N
1 2 2 3 3 (10.41)
y y y y
Computational Heat Transfer 459
N j b N j c
j ; j (10.42)
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 coor-
dinates (ξ1, ξ2) in the following fashion:
x x3 1 x1 x3 2 x2 x3 (10.43)
y y3 1 y1 y3 2 y2 y3 (10.44)
These equations can be inverted or isolated in terms of the local coordinates, yielding
a1 b1 x c1 y 1
1
a1 a2 a3
2A
a1 b1x c1y N1 (10.45)
1
2
2A
a2 b2 x c2 y N 2 (10.46)
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 subarea to the total area of the element.
For example, ξ1 = A1/A, where A1 refers to the subarea consisting of side 2–3 joined with the inte-
rior 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:
3 1 1 2 N 3 (10.47)
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 geo-
metrical interpolation,
For example, a superparametric quadratic triangular element consists of three nodes along each
side of a triangle (midpoint and endpoints) to permit a second-order, or quadratic, interpolation
along each edge. Then both geometrical (x and y) and scalar interpolations are based on six shape
functions,
6
N , ,
i 1
i 1 2 3 i (10.49)
460 Advanced Heat Transfer
where
N i 2i 1 i (10.50)
Using higher order interpolation can increase the accuracy of interpolation within an element, but
generally at the expense of increased computational time and storage.
Local and global coordinates must be related to each other for differentiation of the scalar vari-
ables involving triangular elements. Using the chain rule of calculus,
x y
(10.52)
1 x 1 y 1
x y
(10.53)
2 x 2 y 2
Rearranging and inverting these equations by writing the derivatives in global coordinates with
respect to the local derivatives,
y y
x 1 2 1 1
(10.54)
J x x
y 2 1 2
where
x y y x
J (10.55)
1 2 1 2
dA dx dy (10.56)
where
x x
dx d1ˆ1 d 2ˆ2 (10.57)
1 2
y y
dy d1ˆ1 d 2ˆ2 (10.58)
1 2
Here, the overheat notation designates the unit vector in the direction of interest.
Combining these above three equations,
Computational Heat Transfer 461
x y x y ˆ
dA d1d 23 (10.59)
1 2 2 1
where ξ3 is the unit vector normal to the ξ1 and ξ2 planes. The magnitude may be written as
A
f 1, 2 ,3 dA
f , , Det J d d
0 0
1 2 3 2 1 (10.61)
For example, if an exponential heat source appears in the energy equation, then a spatial integra-
tion 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 integrands,
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 11 n
0 0
f 1, 2 ,3 d 2 d1 W f , ,
i 1
i
i
1
i
2
i
3 (10.62)
where the weights, Wi, and abscissae, ξ ji , 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 characteristic
length of the triangular element, such as a side length. For example, for a linear triangle, the inte-
gration is second-order accurate since the error is approximately reduced by a factor of 4 when the
element size is reduced in half. A quadratic element yields third-order accuracy due to a reduction
by a factor of 8, or proportional to h3 rather than h2.
10.3.2 Quadrilateral Elements
Consider a four-noded quadrilateral element with bilinear interpolation. Here it is required that Ni =
1 at local node i and 0 at the other nodes. Using a local orthogonal coordinate system, denoted by
TABLE 10.1
Gaussian Quadrature for Triangular Elements
n i ξ1i ξ2i ξ3i Wi Order of Accuracy
1 (linear) 1 ⅓ ⅓ ⅓ ½ O(h2)
coordinates s and t (see Figure 10.2; −1 ≤ s ≤ 1, −1 ≤ t ≤ 1), the shape functions can be constructed
individually and combined in an appropriate manner. Along side 1–4, where s = 1, the shape func-
tion N4 must satisfy the following relation:
1
N 4 s 1 1 t (10.63)
2
The interpolation is linear in the coordinate t along that edge. Similarly, along side 3–4, where
t = –1,
1
N 4 t 1 1 s (10.64)
2
Since N4 must satisfy both above equations, the product describes the appropriate bilinear inter-
polation in terms of both local coordinates,
1
N 4 s,t 1 s 1 t (10.65)
4
1
N1 s,t 1 s 1 t (10.66)
4
1
N 2 s,t 1 s 1 t (10.67)
4
1
N 3 s,t 1 s 1 t (10.68)
4
For linear quadrilateral elements, spatial derivatives of ϕ can also be obtained based on the shape
functions as follows:
4
N i
x s,t
x
i 1 s ,t
i
(10.69)
4
N i
y s,t
y
i 1 s ,t
i
(10.70)
Using the chain rule of calculus, the shape function derivatives can be expressed as
N i y y
x 1 t N i
s s
(10.71)
N i J x x N i
y t s t
Computational Heat Transfer 463
x y y x
J (10.72)
s t s t
Also,
4 4
x N i x N i
s
i 1
s
xi ;
t
t
i 1
xi
(10.73)
4 4
y N i y N i
s
i 1
s
yi ;
t
ti 1
yi
(10.74)
The shape function derivatives with respect to the local coordinates can be readily evaluated from
the expressions for Ni (s, t).
Functional expressions can be integrated based on the following area transformation:
1 1
f x,y dxdy f s,t Det J dsdt
A 11
(10.75)
11
f s,t Det J dsdt f s ,t Det J W W
j 1 i 1
i j i j (10.76)
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 imprac-
tical to evaluate, then these quadrature rules can be used to find accurate approximations of the
integrations.
10.3.3 Time-dependent Problems
For time-dependent problems, the general form of the transport equation for a general scalar quan-
tity, ϕ, can be written as
TABLE 10.2
Gauss–Legendre Quadrature for Quadrilateral Elements
n Number of Gauss Points Gauss Points Weights Error
1 1×1=1 s, t = 0 4 (at center) O(h2)
2 2×2=4 s, t = ±1/ 3 1 (at each point) O(h4)
2 2
2 2 Q 0 (10.77)
t x y
where Γ, λ and Q refer to the diffusion coefficient, capacitance and source term, respectively. 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
N
j 1
j
t
j (10.78)
where the overdot denotes a time derivative, the superscript t refers to transient and nnpe is the num-
ber 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:
kij j cij ri
(10.79)
nel
N i N j N i N j e e
kij
e 1 x
x
y
dV N i h N j dS
y
e
(10.80)
V e Se
nel
ri
N iQ e dV e N iC e dS e
(10.81)
e 1 Ve Se
nel
cij N i N j dV e
(10.82)
e 1 Ve
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 (10.80) and (10.81) 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 subelement
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 integrations in Equation
(10.82), 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 (10.82), yielding
a value of zero in the left half–element (x* ≤ 1/2 in one-dimensional elements) and 1 in the right
Computational Heat Transfer 465
half–element. Then after performing the resulting integrations in Equation (10.82), 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 (10.79) can be replaced by a backward difference
in time, as follows:
t t
jt
j j (10.83)
t
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 (10.79) can also be evaluated at any intermediate point in time, for
example, at t + βΔt,
j t
t
t
kij j ri
t
t
cij j (10.84)
t
jt t 1 jt jt t (10.85)
1 1
kij t cij j 1
kij cij j ri
t t t t t
(10.86)
t
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 (10.86). For
example, β = 0 represents an explicit formulation, where future ϕj values are based on calcula-
tions 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.
Substituting β = 0 into Equation (10.86) for a consistent explicit formulation yields:
1 1
cij j kij cij j ri
t t t t
(10.87)
t t
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 computa-
tional effort in an explicit formulation are generally not realized. As a result, a consistent formula-
tion 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 Equation (10.87). For small
time steps, if all conduction heat flows are directed into a control volume, the right side of Equation
(10.87) 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 (10.87) 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.
466 Advanced Heat Transfer
A positive diagonal matrix to ensure numerical stability requires positive right–side values in
Equation (10.87). For the first term on the right side of Equation (10.87) to stay positive in two-
dimensional problems,
x 2 y 2
t (10.88)
2 min
where the subscript min refers to the minimum value in the domain. These time step restrictions 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 diffusion and
source terms are evaluated at the new time level. Substituting β = 1 in Equation (10.86),
1 1
kij t cij j cij j ri
t t t t t
(10.89)
t
Increased temporal accuracy can be achieved by setting β = 1/2 in Equation (10.86). This cor-
responds to a Cranck–Nicolson scheme, whereby diffusion and source terms are evaluated at the
intermediate time level (t + Δt/2). Substituting β = 1/2 into Equation (10.86),
1 1 1 1
2 kij t cij j kij cij j ri
t t t t t /2
(10.90)
2 t
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 intermediate 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.
1. The first step requires discretization of the domain into eight uniformly distributed trian-
gular elements (see Figure 10.3). There are different possible methods of subdividing the
Computational Heat Transfer 467
FIGURE 10.3 Problem domain and finite element discretization of heat conduction in a plane wall.
domain into eight elements. For example, the discretization could involve four subs-
quares, 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 there-
fore the temperature becomes:
T T
N x k x y k y dx dy
Ae
i
e e
WRi e
(10.93)
where WRie refers to the weighted residual and the superscript e refers to elemental
(local within the element). Using integration by parts for the first part of the integral,
T T T Ni T
N x k x dx dy
e e
N k x x k x dx dy (10.94)
e
i Ni k x dy
i dy e
e e
y x y x
Ae
y x
y x
y e xe
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 definitions,
468 Advanced Heat Transfer
dy | dS | cos (10.95)
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 considerations of
the component of the heat flux vector normal to the element’s edge, it can be shown that
T T
k k l1 (10.97)
x n
T e e T Ni T
l12 dS iˆ
x k x dx dy e e
Ni k dx dy Ni k (10.98)
x x n bndy
Ae bndy y exe
where iˆ is the unit x-direction vector and bndy refers to the boundary. A similar result is
obtained for the second term in Equation (10.93), except that l1 is replaced by l2 (direc-
tion cosine with respect to the y–axis). Then, combining this result with the analogous
y-direction result into Equation (10.93),
T Ni T Ni T
N k n dS x k x y k y dx dy
i
e
e e
e e
WRi e (10.99)
bndy y x
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 heT C e (10.100)
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 Equation (10.100) into Equation (10.99),
Ni T Ni T
x k x y k y dx dy N h TdS N C dS
e e e e e e
i i WRi e (10.101)
y exe Se Se
Computational Heat Transfer 469
The first term represents contributions from all elements in the domain. The second and
third terms (surface integral terms) represent only the boundary element contributions,
since all internal adjacent elements have common sides where the surface integrals
mutually self–cancel.
The temperature is approximated based on interpolation via the shape functions,
T N T j j
e
(10.102)
j 1
k e T e R e
(10.103)
where
Ni N j Ni N j e e
kij e kd e kg e
x k
e e
k dx dy Ni h N j dS (10.104)
e e
x y y
y x Se
ri e rg e rr e NiC edS e WRi e (10.105)
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 (10.33) to (10.41), the interior coefficients are obtained as
follows:
bi bj ci cj bi b j ci c j
kij ,d e
2A k 2A 2A k 2A dx dy
y exe
e e
k
4A
k
4A
(10.106)
For the specific domain discretization in this example problem (see Figure 10.3), the
coefficients can be calculated so that the stiffness matrix becomes:
2 1 1
1
kd e 1 1 0
(10.107)
2
1 0 1
It can be observed that the stiffness matrix is symmetric and the sum of entries along a
e
column or row is zero (as expected). Also, rr in Equation (10.105) is not included since
it becomes zero after summation over all elements, as expected by the weighted resid-
ual 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 10.3. The values
within the table represent the ie(e, j) array, where the top row refers to the local node
numbers (j).
470 Advanced Heat Transfer
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 stiff-
ness matrix.
Performing a loop over all rows, i, and columns, j,
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 numbers of the
local stiffness matrix, respectively. The updates in the loop involving Equation (10.108)
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 T2 0
0 1 2 0 0 1 0 0 0 T3 0
1 0 0 4 2 0 1 0 0 T4 0
1
0 2 0 2 8 2 0 2 0 T5 0 (10.109)
2
0 0 1 0 2 4 0 0 1 T6 0
0
0 0 1 0 0 2 1 0 T7 0
0
0 0 0 2 0 1 4 1 T8 0
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 (10.99) 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
Computational Heat Transfer 471
a counter-clockwise ordering convention for an element with local nodes 1 and 2 on the
e
domain boundary, the following boundary integral is added to kij :
kije, g Ni heN j dS e
S e
(10.110)
where dSe = ΔSedξ1 along side 1–2. Also, N1 = ξ1, N2 = 1 – ξ1, and N3 = 0 on Se along
side 1–2 of the element.
As a result, the integrations yield:
2 1 0
he S e
kge 1 2 0 (10.111)
6
0 0 0
Similarly,
e e
1
r C 2S
e
g
1 (10.112)
0
which is added to the right-side vector in the global system on the domain boundaries.
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,
h2S e
kije, g C (10.113)
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,
C e S e
rie, g (10.114)
2
e
when local node i lies on the current boundary and ri , 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
472 Advanced Heat Transfer
Using this boundary node mapping to add the boundary stiffness matrix, Equation
(10.113), and right side, Equation (10.114), 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 T2 10
0 1 2 0 0 1 0 0 0 T3 5
1 0 0 4 2 0 1 0 0 T4 0
1 (10.115)
0 2 0 2 8 2 0 2 0 T5 0
2
0 0 1 0 2 4 0 0 1 T6 0
0 0 0 1 0 0 3, 335.3 4, 999 0 T7 0
0
0 0 0 2 0 1, 665.7 3, 337.3 1, 665.7 T8 0
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 Gaussian
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 bound-
ary. 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 specified 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.
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 tolerance, 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 variables should not change from one mesh to another mesh.
u u p 2u 2u
1 u,v,p uo vo 2 2 0 (10.116)
x y x x y
Computational Heat Transfer 473
v v p 2v 2v
2 u,v,p uo vo 2 2 0 (10.117)
x y y x y
u v
3 u,v,p 0 (10.118)
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 velocity 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 previous 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 (10.116) to (10.118),
u v
W x y dV 0
(10.121)
p
i
V
where the superscripts u, v, and p refer to the x-momentum, y-momentum, and continuity equations,
respectively.
Define the approximate local velocities and pressure as a linear combination of nodal values
weighted by the shape functions at their corresponding nodes,
nnpe nnpe nnpe
u N u ;
i 1
u
i i v N v ;
i 1
v
i i p N
i 1
p
p
i i
(10.122)
where nnpe refers to the number of nodal points per element (e.g., nnpe = 2 for linear elements and
nnpe = 4 for quadrilateral elements). Using Galerkin’s method, the weight functions are selected as
equal to the shape functions so that Wiu = N iu , Wi v = N iv , and Wi p = N ip.
Assembling all finite elements and using integration by parts in the x–momentum equation, anal-
ogously to the method presented earlier for heat conduction, leads to:
nel nel
u u p N i u N i u e u
N u u
i
o
x
vo
dV
y x x x y y
N n u
i
S e
dS e 0 (10.123)
e 1 V e e 1 S e
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.
474 Advanced Heat Transfer
k uu ui k up pi ri u
(10.124)
k vv vi k vp pi ri v
(10.125)
k pu ui k pv vi ri p
(10.126)
ri u
N dS u e
i u
e
(10.129)
e 1 S e
nel
o N vj N vj N iv N vj N iv N vj e
kijvv N i
v
u
x
vo
y x
x
y
dV
y
(10.130)
e 1 V e
nel
N jp
kijvp N iv
y
dV e (10.131)
e 1 V e
nel
ri v N dS v e
i v
e
(10.132)
e 1 S e
nel
N uj
k pu
ij N ip
x
dV e (10.133)
e 1 V e
nel
N vj
kijpv N i
p
x
dV e (10.134)
e 1 V e
ri p = 0 (10.135)
where τ ue and τ ve refer the net elemental shear stress components in the x and y–directions, respec-
tively. The surface integral terms, riu and riv are evaluated only at external surfaces. The above expres-
sions 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.
Computational Heat Transfer 475
Boundary conditions must be specified for closure of the problem definition. Examples of com-
mon types of CFD boundary conditions are listed below. A typical example of boundary conditions
for flow past a two-dimensional circular cylinder is illustrated in Figure 10.4.
As discussed earlier, an iterative solution procedure is required for the nonlinear convection 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 previ-
ous 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).
Overrelaxation (1 ≤ ω ≤ 2) is used if convergence is proceeding but slowly and it is desirable
to speed up the iterations. On the other hand, underrelaxation (0 ≤ w ≤ 1) is used if the results
are diverging or oscillating and the iterations need to proceed more slowly.
4. Check the convergence status by calculating whether max|(u – uo)/uref| (or some other residual)
is less than a specified tolerance.
5. Return to the first step and repeat the steps until convergence is achieved.
The performance of the CFD solver is closely related to the structure of the matrix of coefficients
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 per-
formance. The placement of coefficients in the stiffness matrices and orientation of elements are key
factors that influence the bandwidth. Rather than ordering of variables in the usual fashion (u fol-
lowed by v and p), a rearrangement of the ordering sequence can drastically reduce the overall band-
width. For instance, at each node, the continuity and y-momentum equations can be interchanged,
thereby ensuring nonzero entries on the diagonal and reducing the matrix bandwidth.
u v
S (10.136)
t x y x x y y
v S (10.137)
t
This equation describes the equations of conservation of mass, x-momentum, y-momentum, and
energy by selecting ϕ = 1, u, v, and e, respectively.
Assuming steady-state conditions, and integrating this conservation equation over a volume at
node i, denoted as Vi, and from a time of t to t + Δt, yields:
t t t t t t t t
t Vi
t
dVdt
t Vi
v dVdt
t Vi
dVdt
S dVdt
t Vi
(10.138)
Using the Gauss divergence theorem of calculus to transform the volume integral to a surface
integral over the surface Si encompassing Vi, and taking the volume average of the last term over
the volume,
t t t t t t t t
t
dVdt
v nddSdt
ndSdt S dVdt (10.139)
t Vi
t Si
t Si
t Vi
where Si represents the total surface area of the control volume and n is a unit vector normal to the
surface and pointing outward.
t t t t
t Vi
t
dVdt
t
t P
VP dt VP Pt t VP Pt
(10.140)
1 1
e sign m e e P sign m e e E (10.141)
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 approximation 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 = ½
= 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 equation as follows.
2
u 2 (10.142)
x x
Solving this equation subject to boundary conditions, ϕ(0) = ϕP and ϕ(Δx) = ϕΕ, yields:
P exp Pe x /x
(10.143)
E P exp Pe 1
1 exp Pe / 2 1
e (10.144)
2 exp Pe 1
At approximately Pe = 10, the weighting factor gives the UDS convection limit. An alternative
approximation of the weighting factor over the entire range is given by
Pe2
e (10.145)
10 2 Pe2
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 writ-
ten as
A A
ndS
Si
n e
Ae
e E
e P
x x
(10.146)
where a diffusion weighting function, β, is introduced as a correction factor because the temperature
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:
Computational Heat Transfer 479
Pe e Pe /2
(10.147)
e Pe 1
or alternatively by approximation,
1 0.005Pe2
(10.148)
1 0.05Pe2
S dVdt S V t
t Vi
P (10.149)
aPP a S
i
i i
P (10.150)
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.
480 Advanced Heat Transfer
aeue anb
u
unb bu pP pE Ae (10.152)
ae vn anb
v
vnb b v pP pN An (10.153)
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:
aeue anb
u
unb bu pP pE Ae
(10.154)
an vn anb
v
vnb b v pP pN An
(10.155)
To improve the updated velocities, u* and v*, such that they also satisfy the mass conservation
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 represented by
u u u; v v v (10.156)
p p p (10.157)
Subtracting the u* and v* momentum equations from the u and v equations yields the following
momentum equations for the velocity and pressure correction terms.
aeue anb
u
unb pP pE Ae (10.158)
an vn anb
v
vnb pP pN An (10.159)
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 pressure correction dif-
ference 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 Equations
(10.158) and (10.159). The velocity correction equations are simplified as follows:
The velocity corrections are determined and the updated velocities in Equation (10.156) are then
used as solutions at the next iteration level. The updated velocities are then substituted into the mass
conservation equation, Equation (10.151), yielding the following pressure correction equation:
where
aP anb (10.164)
bP uy u y v x v x
w
e
s
n
(10.165)
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.
In the SIMPLE algorithm, the nearby velocity corrections were neglected in Equations (10.158)
and (10.159). In contrast, the SIMPLEC algorithm uses a “consistent” approximation wherein the
term Σ anb unb is subtracted from both sides of Equation (10.158), rather than neglecting the nearby
velocity corrections. Therefore, instead of the simplified result of Equation (10.160) in SIMPLE, the
velocity correction equations become:
In SIMPLEC, the first term on each right side is neglected, yielding the following expressions
for d:
Ae An
de ; dn (10.168)
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 Equations
(10.158) and (10.159). Also, no under-relaxation is needed for the pressure correction equation in
the SIMPLEC algorithm.
482 Advanced Heat Transfer
In SIMPLEC, the velocity correction equations in Equations (10.158) and (10.159) are extended
to neighboring velocity corrections by assuming:
This assumption extrapolates the expression of the velocity correction to adjacent locations so
that the velocity correction equations become:
Further comparisons between SIMPLE and its other variants for several test cases on various
benchmark problems were presented by Zeng and Tao (2003).
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 real-
izable 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, experimentally
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 Reynolds 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, and derivative of the dissipation
rate, ε, perpendicular to the wall, are set to zero at 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.
boundaries, or subsurfaces (SS), are coincident with the element exterior boundaries and the local
coordinate surfaces defined by s = 0 and t = 0. An integration point (ip) is defined at the midpoint
of each subsurface.
In a manner similar to earlier examples for one-dimensional and triangular elements, a general
scalar variable, ϕ, can be related to local node values, Φi, as follows:
4
s,t N s,t
i 1
i i (10.171)
where the shape functions and their derivatives for quadrilateral elements were defined earlier in
Equations (10.65) to (10.68). 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 (10.136) and (10.139),
u v
S (10.172)
t x y x x y y
t t t t t t t t
t Vi
t
dVdt
v ndSdt ndSdt S dVdt
t Si t Si
t Vi
(10.173)
t t
t
dVdt J1 1t t 1t (10.174)
t SCV 1
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 (10.72). 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 convec-
tion fluxes. Alternatively, the method of integration point equations is a robust method to determine
the convection fluxes along the SCV boundaries of a control volume. Discretized approximations of
the transport equation, Equation (10.172), 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 (10.172) at each integration point again involves tran-
sient, convection, diffusion, and source terms. The transient term at an integration point is approxi-
mated by a backward difference time, for example, at integration point 1 (subscript ip1),
ipt1t ipt1
(10.175)
t ip1 t
Computational Heat Transfer 485
An upstream difference approximation is used to represent the convection term at the integration
point,
u v V ip1 u (10.176)
x y ℒc
where V u2 v 2 represents the fluid velocity magnitude, Lc is the convection length scale in the
streamwise direction, and ϕu represents the upwind value of ϕ. The method of skewed upwinding
identifies the direction of the line segment between ϕip1 and ϕu. The upstream value, ϕu, is calculated
by an interpolation upstream of the subvolume edge where the local streamline through the integra-
tion point intersects that edge. For example, if the line constructed in the upwind direction intersects
the quadrant edge between local nodes 2 and 3,
a a
u 2 1 3 (10.177)
b b
where a and b refer to coefficients corresponding to linear interpolation for ϕu in terms of Φ2 and Φ3
along the intersected edge. This skewed upwinding aims to retain both the directional and strength
influences of convection at the integration point.
Shape function interpolation is used to approximate the pressure gradient at the integration point,
for example, at ip1,
4
p N i
x ip1
x
i 1
pi
(10.178)
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,
1
4
2u 2u
2 2
x y
2
ℒd
N j j ip1
(10.179)
ip1 j 1
where Ld is a diffusion length scale. For integration point 1, it can be shown that:
1
2 8
ℒd 2
2 (10.180)
x 3y 2
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 interpola-
tion of nodal values using the shape functions.
After all integration point operators are assembled back into Equations (10.172) at the integration
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 integra-
tion point and nodal point values in order to calculate the diffusion fluxes for the control volume
486 Advanced Heat Transfer
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 (10.173) must be evaluated at both SS1 and SS4 surfaces cor-
responding to SCV1. At SS1,
t t
ndSdt
t Si
n ip1
t y1
n ip1
t x1
(10.181)
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 (10.173), a lumped approximation is used,
t t
S dVdt J t(S|
t SCV 1
1 1/ 2,1/ 2
(10.182)
where the (½, ½) subscript refers to the local coordinate position in the center of SCV1.
T
c
V
p
t
dV Q2,1 Q4,1 Qe1,1 Qe 2,1
(10.183)
where Q refers to the rate of heat flow across the sub–surface 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 computed as
follows:
1
T T
QSS1
SS 1
kT ndS k x dy y dx
0 s 0
dt
(10.184)
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 N y N x
Q2,1
i 1
k
0
i
x t
k i dt Ti
y t s 0
(10.185)
Computational Heat Transfer 487
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 T t t T1t
cp c p J1 1 (10.186)
t t
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
k
j 1
e
T r1e
1, j j
e
(10.187)
where the local (elemental) stiffness matrix coefficients and right–side vector for SCV1 can be writ-
ten as
1 1
N j y N j x N j y N j x 1
k1e, j k
0
x t
k
y t
dt k
s 0 0
x s
k
y s
ds
t 0
t
c p J1
(10.188)
1
r1e c p J1T1t (10.189)
t
Similar terms are constructed for the other SCVs such that the entire 4 × 4 stiffness coefficient
matrix can be assembled and completed. The previous equations have been constructed locally
within the element so that the subscripts refer to local nodes. For example, T1 refers to the tempera-
ture 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 double
loop over i = 1, 2, 3, 4 and j = 1, 2, 3, 4. The first loop is indicated by the range of the subscript j in
Equation (10.188). 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 (10.183), are not required in the
stiffness coefficient in Equation (10.188), 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 bal-
ances 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 temperatures must be specified through appropriate boundary conditions.
Once the elemental stiffness matrix is completed, a standard finite element assembly procedure
is performed for all finite elements in the mesh. The assembly process ensures that energy conser-
vation 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.
and solved for the velocity, pressure, and temperature fields. However, the effects of radiation were
neglected. Accurate prediction of radiation exchange requires the further consideration of absorp-
tion, emission, and scattering of radiant energy. Radiation may be attenuated by absorbing gases
such as water vapor, carbon dioxide, and suspended particles like coal, ash, and soot in combustion
products.
As described in Chapter 4 for participating media, the analysis of radiative heat transfer considers
the transfer equation for the radiation intensity field, Iλ. The radiation intensity was defined as the
radiant energy carried by photons that propagate along a given direction, s, and crosses a solid angle,
ω. In Chapter 4, the following form of the radiation transfer equation at a given at a given position,
r, in the direction, s, was presented:
4
dI r,s
ds
a ,a I ,s I ,s
4 I
0
r,s s d a ,a I ,b
(10.190)
where aλ,a, σl,s, Φ and Iλ,b, represent the absorption coefficient, scattering coefficient, scattering
phase function and blackbody intensity, respectively. Modeling of the intensity field is a key feature
of radiation numerical methods.
This section will provide a brief overview of three common numerical methods for radiation heat
transfer: (1) discrete transfer radiation model (DTRM); (2) discrete ordinates (DO) method; and (3)
the finite volume method (FVM). Commercial software packages, such as ANSYS Fluent, typically
provide capabilities for one or more of these radiation models).
dI a T 4
a ,a I ,a (10.191)
ds
which assumes a refractive index of n = 1. Assuming aλ,a is constant and then integrating along a ray
emanating from a boundary surface,
T 4
I s I 0e a ,a s
1 e a ,a s
(10.192)
where I0 is the radiation intensity at the start of the ray’s path. Using this formulation, the radiation
effects can then be included through the change of intensity along the path of each ray which is
traced through a control volume. The accuracy is limited by the number of rays and control volumes.
At each radiating surface, ray paths are traced over the full range of polar (θ) and azimuthal (ϕ)
angles to cover the entire hemispherical range of angles. These paths are determined and stored in a
file, as well as the intercepts and corresponding lengths within each control volume. The solution of
the radiative transfer equation then uses this information in the solution procedure.
Computational Heat Transfer 489
With an increasing number of surfaces to trace rays, and volumes crossed by rays, the computa-
tional time of DTRM also increases rapidly. Clustering of surfaces and cells into surface and volume
“clusters” can effectively reduce the computational time. A volume cluster is formed by adding a
specified number of neighboring cells and their neighbors to a given cell. Similarly, a surface cluster
is formed by adding neighboring surface faces.
The incident radiation flux and radiation source term are calculated for each surface and volume
cluster, respectively, and then distributed to faces and cells in the cluster to establish the surface and
cell temperatures. Area and volume averaging of the surface and volume clusters then yields:
1/ 4
M
A f T f4
f 1
Tsc (10.193)
M
Af
f 1
1/ 4
N
VcTc4
Tvc c 1
(10.194)
N
Vc
c 1
where M and N are the numbers of faces of a surface cluster and cells of a volume cluster, respec-
tively. Also, Tsc, Tvc, Tf, Tc, Af, and Vc refer to the temperatures of the surface cluster, volume cluster,
face, and cell; face area; and cell volume, respectively.
Along the walls, the net radiation heat flux is computed as the sum of reflected portions of the
incident radiation and the emissive power of the surface. At flow inlets and outlets, the net radiation
flux is calculated in the same manner as the walls. Normally the emissivity of flow inlets and outlets
is assumed to be 1.0.
a0 ds
I r,s 1 0 I ,b 0
4 I
0
r,s s d
(10.195)
where a0 = aλ,a + σλ,s is the extinction coefficient, Ω0 = σλ,s/a0 is the scattering albedo (discussed in
Chapter 4), and Φ is the scattering phase function (equals unity for isotropic scattering).
For a diffusely emitting and reflecting wall, the boundary condition on the surface, S, can be
expressed as
1 w
I w s w I w,b
I w s s nw d
S
(10.196)
where the subscript w denotes the location of the wall and nw is the unit normal vector at the wall
toward the medium.
Figure 10.7 illustrates a sample discretization with unstructured triangular cells, including both
internal and near-wall elements. Each vertex is numbered in a consistent counter-clockwise direction.
There are two common types of control angle discretizations in the finite volume method. The
first approach, called an SN DO discretization, uses the direction cosine and corresponding angular
weights spanning the range of the 4π solid angle. Alternatively, the total solid angle, 4π steradians
is subdivided into Nθ × Nϕ directions of the polar and azimuthal angles from 0 to π and 0 to 2π,
FIGURE 10.7 Typical triangular elements in (a) the interior and (b) near the domain boundary.
Computational Heat Transfer 491
FIGURE 10.8 Schematic of angular discretization in the (a) θ and (b) ϕ directions.
respectively (see Figure 10.8). For two-dimensional problems, only 2π steradians are considered due
to angular symmetry, i.e., 0 ≤ θ ≤ 0.5π and 0 ≤ ϕ ≤ 2π. This section will present the latter approach.
Integrating Equation (10.195) over a control volume, ΔV, and control angle, Δωm, in Figure 10.7,
yields the following form of discretized finite volume equation about nodal point P,
Ii
i
m
Ai Dcim a0 I m Srm P
VP m
(10.197)
The summation is taken over neighboring volumes, the subscripts i and m refer to surface i and
direction m, respectively, and
Dcim
m
s n d i m (10.198)
4
S 1 0 I ,b 0
I ,m m d m
m
r (10.199)
4
0
m 1 n1 m 1
m
d sin d d
m
n m
(10.200)
m
where ni is the outward unit normal vector at face i. The discretized transfer equation indicates that
the net outgoing radiation across the surface of control volume P is balanced by the net effects of
absorption, emission, and scattering within the control volume and control angle. The directional
weighting factor, Dcim , determines whether an inflow or outflow of radiation occurs across the surface.
Various schemes have been used to relate the control volume face intensity, I im, to the nodal inten-
sity, I Pm, such as the diamond scheme, step scheme, and exponential scheme (Carlson and Lathrop,
1968). Here a step method will be used whereby the downstream face intensity is set equal to the
upstream nodal value. This method ensures a positive intensity and has a relatively simple imple-
mentation as it does not consider complex geometric or directional information.
Using a step scheme, the control volume face intensity, I im, can be approximated as
I im Dcim I Pm max Dcim ,0 I Im max Dcim ,0 (10.201)
where the subscript I represents the neighboring nodes and i refers to its corresponding face. Then
Equation (10.197) can be rewritten in the following general standard form of a finite volume equation,
492 Advanced Heat Transfer
aPm I Pm a II
m m
I I bPm
(10.202)
where
aIm max Ai Dcim ,0 (10.203)
aPm max A D ,0 a
i
i
m
ci 0, P VP m
(10.204)
bPm a0 Srm P
VP m
(10.205)
The boundary condition in Equation (10.196) for a diffusely emitting and reflecting wall becomes:
1 w
I wm w I w,b
I m
w
m
Dcw (10.206)
m
Dcw 0
The full set of algebraic equations is obtained by assembling all control volumes in the domain.
An iterative solution proceeds until convergence of the intensity field is achieved.
I Pm I Pm,old
max (10.207)
I Pm
where ε is a tolerance (typically 10–6) and I Pm,old is the previous iteration value of I Pm. Once the inten-
sity is obtained, the radiative wall heat flux is estimated by
4 M
r ,w
q I w s s nw d
0
I m 1
m
w
m
M cw
(10.208)
m
Where Dcm is the directional weighting at the wall defined by the dot product of the unit direction
vector and unit normal vector at the wall toward the medium (s ∙ nw).
Sample results from this finite volume method for radiation exchange in two-dimensional enclo-
sures were presented by Kim et al. (2008). Figure 10.9 illustrates the results for two cases of a
triangular enclosure and a semi-circular enclosure with a circular hole. The solid angle of 2π is
subdivided into Nθ × Nϕ directions with uniform intervals of Δθm = 0.5 π/Nθ and Δϕ m = 2π/Nϕ. The
spatial domain was discretized into triangular cells.
For the triangular enclosure, the length of each side is 1 m and the walls are held cold (Tw = 0 K)
and black (εw = 1), while the gas is held at a constant temperature (Tg = 1,000 K). The mesh consists
of 400 elements and the angular discretization consists of Nθ = 4 and Nϕ = 16. For a large absorption
coefficient of 10 m–1, the radiant energy arriving on the wall approaches the blackbody intensity.
The optically thick medium absorbs nearly all of the radiation so the incident intensity on the wall is
influenced only by emission near the enclosing walls. When the absorption coefficient is small, the
emission of the gas is weaker, so the radiation flux is reduced. The results indicate close agreement
between the predicted and exact solutions.
For the hemispherical enclosure, again the gas temperature is Tg = 1,000 K and the walls are
Tw = 0 K and black (εw = 1). The mesh has 1,356 triangular elements with Nθ = 4 and Nϕ = 16.
Computational Heat Transfer 493
FIGURE 10.9 Dimensionless wall heat flux at varying absorption coefficients in an (a) equilateral triangular
enclosure and (b) semi-circular enclosure with a circular hole. (adapted from Kim et al., 2008)
A circular hole of 0.2 m radius is centered at y = 0.4 m from the bottom wall of the semi-circular
enclosure with a 1.0 radius. From the results, the heat flux near the center is slightly reduced due
to the existence of the cold inner circle. The results for varying absorption coefficients show good
agreement with the exact solutions. As the absorption coefficient decreases, the radiative heat flux
at the bottom wall also decreases since the emission by the medium is lower at a smaller absorption
coefficient.
10.8 TWO-PHASE FLOWS
Unlike numerical methods of single-phase flows, the analysis of two-phase flow involves the track-
ing of a moving phase interface, for example, a gas/liquid interface in boiling or condensation
problems, or a solid/liquid interface in soldification and melting problems. A brief introduction to
two-phase flow modeling will be presented in this section. The reader is referred to further books by
Gross and Reusken (2011) and Spelt et al. (2010) for a more detailed analysis of numerical methods
for two-phase flows.
c dH Ui Th Tc dAi
dq m (10.209)
where the subscripts h and c refer to hot and cold streams, respectively, and ṁc denotes the mass
flow rate of the cold stream. Also, Ai refers to the inner area of the tube based on the inner diameter,
Di. The energy balances will primarily involve enthalpy, rather than temperature, in order to include
both sensible heat (relating to a temperature change) and latent heat (associated with phase change)
portions of the heat exchange.
494 Advanced Heat Transfer
The overall heat transfer coefficient, U, can be expressed in terms of the heat transfer coefficients
for the inner and outer tube surfaces, Ui and Uo, respectively, as follows:
1 1 1
= = (10.210)
UA Ui Ai U o Ao
where hc and hh refer to the convection coefficients for the hot and cold streams, respectively, along
the surface, dx, separating both fluid streams.
The fluid may pass through several phase-change regimes (as described in Chapter 5). Therefore
the tube length is subdivided into discrete elements. Consider a uniform subdivision of the tube
length into N elements of width Δx. An energy balance is applied within each element. For the jth
element, Equation (10.209) may be discretized as follows:
Ui, j Di x
H j 1
m c
Th, j Tc, j H j (10.213)
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 be used to determine the tempera-
ture 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 ele-
ments can be used to find the corresponding temperature rise as follows:
c p T j 1 T j H j 1 H j (10.214)
This yields Tj + 1 in terms of Tj (computed in the previous upstream element) and the computed
enthalpy change from Equation (10.213). 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 super-
heated vapor, the vapor-specific heat at the mean temperature (between Tj and Tj + 1) can be used.
Computational Heat Transfer 495
In the two-phase regime, if the enthalpy in Equation (10.213) is less than the saturated vapor
enthalpy, then the mass fraction of vapor (or quality), χg, in element j + 1 becomes:
h j 1 h f
g, j 1 (10.215)
hg h f
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). Then an updated
overall heat transfer coefficient is computed based on Equation (10.211),
1
1 Di
Ui, j 1 (10.216)
hc, j 1 hh, j 1Do
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 variables
are computed in the first element, j = 1, after which variables in successive elements are calcu-
lated based on the previous element’s values. For example, after Ui,1 is calculated from Equation
(10.216), the enthalpy, temperature, and quality for the following element, j = 2, can be calculated
from Equations (10.213) to (10.215). 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 solution procedure can be summa-
rized as follows.
A similar procedure can be adopted for both condensation and boiling problems. A numerical solu-
tion 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 (Chapter 5). This would distinguish between the various flow regimes, such as wavy,
annular, and slug flow regimes. For example, when saturated liquid enters a heated horizontal pipe,
boiling occurs initially 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 (10.209) 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.
496 Advanced Heat Transfer
This section has discussed a numerical model of liquid–gas phase change in condensers and boil-
ers. Similar 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
exchanger 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.
10.9 MACHINE LEARNING
10.9.1 Introduction
Machine learning is a method of data analysis that collects, interprets and learns from external data
to accomplish specific goals and tasks. It is a branch of a broader field of artificial intelligence (AI).
Unlike the natural intelligence of humans and animals, machine learning is an artificial form of
intelligence. AI has numerous applications in today’s society, such as financial forecasting, medical
diagnosis, driverless vehicles, online retail and advertising, among others. With rapid advances in
computing power the past few decades, the capabilities of AI have also risen dramatically.
In general, there are two types of machine learning – unsupervised and supervised learning.
Unsupervised learning analyzes information based on input data alone. This approach finds patterns
in the data, for example, through clustering patterns or groupings of data. Examples of unsuper-
vised learning algorithms include Hierarchical, Neural Networks, Hidden Markov Model, K-Means,
M-Medoids, and Gaussian Mixture methods.
Computational Heat Transfer 497
In supervised learning, future outputs are predicted based on known input and output data, either
by classification or regression techniques. A classification model predicts a discrete response and
classifies the input data into categories, for instance, in medical imaging to predict whether a tumor
is cancerous or benign. Common classification methods are Support Vector Machines, Discriminant
Analysis, Naive Bayes and Nearest Neighbour. In contrast, regression techniques predict continuous
responses, such as changes in temperature over time. Common regression algorithms include Neural
Networks, Decision Trees, Linear Regression and Ensemble methods.
There are six main steps as follows in the machine learning process.
1. Enter the data. All pertinent data are loaded into a database. The machine learning algorithm
cannot distinguish between noise and pertinent information.
2. Pre-process the data. Decide whether or not outliers are included, check for missing values,
substitute approximations for any missing values, and subdivide the data into testing and
training data sets.
3. Derive significant features from the data. Extract key features of data to improve accuracy,
performance of the model, and interpretability of the data.
4. Build and train the model. This involves the creation of a decision tree and matrix that com-
pares classifications made by the model with class labels created in Step 1.
5. Iterate to improve the model. Make the model simpler by reducing the number of features
using a correlation matrix, principal component analysis (PCA) or sequential feature reduc-
tion. Alternatively, add complexity by combining simpler models or adding more data sources.
6. Integrate the model into an actual application.
Commercial software tools such as MATLAB (2019) have extensive capabilities to design and
implement machine learning algorithms using the above procedures.
are particularly effective for highly nonlinear systems, unexpected changes in the input data and
cases where data is available incrementally so the model is continuously improved.
In Figure 10.11, an ANN is an interconnected group of nodes, similar to the network of neurons
in a human brain. Each circular node designates an artificial neuron. The arrows represent connec-
tions from the output of one neuron to the input of another. Signals move from the first layer (called
the input layer) to the last layer (output layer), potentially traversing the layers back and forth many
times to improve the accuracy of the model.
A deep neural network consists of an input layer, multiple processing layers and an output layer,
all connected by nodes or neurons. Each processing layer, or hidden layer, uses the output of the
previous layer as its input. These represent processes operating in parallel analogously to the human
nervous system.
Consider an ANN with n layers, each with a number of nodes, mi. The first and last layers are
the input and output layers, respectively. Of the M sets of data, M1 are used for training and M2 for
testing. Define (i, j) as the jth node in the ith layer; xi,j as its input; yi,j as its output; θi,j as its bias
,j
(a parameter which adjusts the output to better fit the given data); and wii−1, k as the synaptic weight
between nodes (i–1,k) and (i, j). All variables are usually normalized to fit within a defined range
such as [0, 1].
The input to node (i,j), xi,j, is related to outputs, yi–1,k, from the previous layer, i–1, though a func-
tional relationship of the following form:
mi 1
xi, j i, j w
k 1
i, j
i 1, k yi 1,k
(10.217)
The functional relationship between the output, yi,j, from node (i,j) and its input at the same
node, xi,j, is called an activation function and represented by yi,j = g(xi,j). Examples of some common
activation functions are identity (linear), sigmoid, Gaussian and inverse square root unit (ISRU)
functions.
The Backpropagation Method is commonly used for training of an artificial neural network
through a back propagation of errors. The error at the output layer, n, is estimated from:
where ζi,j is the target output for node (n, j). After calculating δn,j, the procedure moves back to the
(n–1)th layer. Then errors in the (n–2)th layer are computed based on layer (n–1) and so forth back
to the second layer, i = 2.
To train the network, the weights are changed repeatedly until output values become sufficiently
close to the known or desired target values. Weights and biases are assumed to be proportional to
the errors as determined by
i, j a i, j (10.221)
where a is the learning rate. It represents a factor used to scale the degree of change between con-
nections in layers. A higher value leads to faster learning but also more likely for over–shooting or
under-shooting the desired target output. After a full cycle of back propagation is completed with the
above biases and corrections, an updated set of weights is determined. Then iterations are continued
until the outputs differ little from the desired target outputs.
The backpropagation algorithm is repeated with the set of training data as many times as neces-
sary until a specific convergence criterion is achieved. A specified problem variable predicted by the
ANN for run i, FANN,i, is compared with the value supplied by the training data set, F. The standard
deviation of the ratio, fi = F/FANN,i, is given by
fi f
M1 2
i 1
M1
(10.222)
where f is the average of fi over the number of training runs, i = 1, 2 … M1. When the ANN is initially
formed, the starting conditions for the weights and biases can be set randomly. With a sufficiently
large number of cycles, the starting conditions usually do not make a significant difference on the
desired target output.
kA
Q
x
T1 T2 (10.223)
where T1 and T2 are the left and right wall temperatures corresponding to the heat inflow and outflow
surfaces, respectively. Using a basic 2–1–1 ANN, the input layer (i = 1) has 2 inputs, T1 and T2. The
output is the heat flow rate at the left wall, Q2,1, which serves as the input to the next layer (i = 3;
output layer). The final output is the heat outflow from the right wall, Q3,1.
From the general functional relation, Eq. (10.217), starting at the output layer (i = 3),
On the left side, the input to the third layer, x3,1, was related to its output, y3,1, through an identity
activation function. A linear activation function can be used, i.e., yi,j = g (xi,j) = xi,j, because 1–D
steady-state heat conduction is a linear problem and thus the functional relationship of Q (T1, T2) is
linear. If the relationship is nonlinear, for example, if thermal conductivity varies with temperature,
then a linear activation function would not accurately represent the relationship. For nonlinear prob-
lems, the weight factors do not have a physical significance so a more generally applicable activation
function can be used, such as the sigmoid function.
Moving back from the output layer (i = 3) to the hidden layer (i = 2), again using the identity
activation function,
Q3,1 3,1 w23,,11 2,1 w12,1,1 T1 w12,,21 T2 (10.226)
Comparing with the analytical expression of the functional relation, Eq. (10.223),
kA kA
3,1 w23,,11 Q2,1 0; w23,,11 w12,1,1 ; w23,,11 w12,,21 (10.227)
x x
The weight factors, wii,,kj are not unique since other values may be used but which must satisfy
the above relations. For example, w23,,11 = 1 can be selected and then w12,1,1 and w12,,21 are determined from
Eq. (10.227).
Consider another heat transfer example of forced convection in a duct of perimeter P, length L,
mass flow rate of ṁ, and constant wall temperature, Tw. From Chapter 3, the rate of heat transfer
along the length of the duct is given by
hPℒ
p Tin Tw exp
Q mc 1 (10.228)
mc p
FIGURE 10.12 Comparison of ANN and actual heat flows for laminar and turbulent convection in a duct.
(adapted from Diaz et al., 1999)
Computational Heat Transfer 501
For laminar flow, the differences between QANN and Q are generally less than 2%. A similar close
agreement between QANN and Q is obtained for turbulent flow. However, when both flow regimes are
combined in the ANN, larger errors are observed. Separating the flow data into different regimes
yields more accurate heat transfer predictions.
10.9.4 Linear Regression
Linear regression is another common type of supervised machine learning. It assumes that the rela-
tionship between input(s) or independent variable(s) with an output or dependent variable can be
approximated by a linear equation. Different forms of linear regression include simple linear regres-
sion and multiple linear regression.
Simple linear regression is the most basic approach with only one input and a single output.
An example of simple linear regression is an equation to predict the steady-state volume flow rate
through a fixed diameter pipe based on a single input of the mean flow velocity. Multiple linear
regression advances the relationship beyond a single input variable to multiple inputs with one out-
put. This could be predicting the flow rate not only from the fluid velocity but also on the varying
cross-sectional area of the pipe.
A linear regression assigns random weights, wn, typically in the range of 0 – 1, to each input in
an individual data set. A random bias, θ, also generally in the range 0 – 1, is created. The sole input
is multiplied by a corresponding weight. A bias is then added to the product of the input and weight
to give a predicted output. In a linear regression model, z = mx+b, the weight represents the slope,
m, and the bias represents the axis intercept, b. More generally for multiple linear regression, the
predicted output, ŷ, for a set of n inputs can be written as
ŷ w1 x1 w2 x2 wn xn (10.229)
The goal of linear regression is to find the combination of weights and biases that yield the
best linear fit of the input data. This is accomplished by assigning initial random weights and then
iterating between cycles of predicting outputs, calculating errors in the predictions, and updating
the weights and biases accordingly. Iterations are repeated until the error between the predictions
and answers falls within a specified range. This process is achieved by a training data set, which
contains a set of several unique inputs and corresponding outputs. The error between the predicted
outputs and answers is found by either using a loss function or a cost function.
A loss function computes the error between the predicted and correct output from a single set of
inputs. A commonly used loss function is the Squared Error function. The loss, L, between a predic-
tion and corresponding output, y, using the Squared Error function is given by
(10.230)
ℒ yˆ y
2
A cost function is the average of all loss functions for an entire training set. It yields a single
number that represents the error between all predictions and the corresponding correct answers.
Standard cost functions are the Mean Squared Error (MSE) and Root Mean Squared Error (RMSE)
functions. The MSE is the average of the sum of the squared errors for an entire data set. The RMSE
takes the square root of the MSE so that the dimensions of the error are the same as the inputs. The
value of RMSE gives the average distance from a data point to the fitted line as predicted by the
model. Then the Root Mean Squared Error is written as
yˆ y
1 2
RMSE (10.231)
N
i 1
502 Advanced Heat Transfer
To reduce the error as much as possible, the minimum value of one of these functions must be
obtained. The cost function can be minimized by one of several optimization methods. The most
common method is Gradient Descent.
Gradient Descent updates the weights and biases of a model so that the cost function, which is
dependent on the weights and biases, reaches the lowest possible value. This is achieved by taking
the gradient of the cost function, multiplying each weight and bias partial derivative (with respect to
the cost function) by a learning rate, a, and subtracting it from the corresponding weight/bias from
the previous iteration. The learning rate characterizes the rate of speed where the gradient moves
during gradient descent. A positive gradient indicates that the cost function is increasing so it is
subtracted from the current weight. Then the weights are updated to reach the global minimum of
the cost function rather than a global maximum. The weights are updated according to the gradient.
This method has a benefit of making larger changes to the weights when they are further from the
global minima, since the gradient is larger.
One iteration of Gradient Descent occurs in predicting the outputs for all sets of training inputs,
calculating the cost and its gradient, and then updating all weights and biases accordingly. Updated
weights and biases using Gradient Descent can be written as
C C
wi 1 wi a ; i 1 i a (10.232)
wi i
where i, i+1 and C refer to the iteration number, updated value for the next iteration, and cost
function, respectively. Since Gradient Descent uses errors for all inputs in a training set at once,
for a single iteration, the partial derivative of the cost function with respect to each of the weights/
biases must be completed for every set of inputs in a training set. This can become computationally
intensive especially for large data sets with many weights. Therefore, in large datasets, Stochastic
Gradient Descent is a preferred alternative.
Stochastic Gradient Descent functions are similar to Gradient Descent but instead of calculating
the gradient of the cost function for each iteration, the method calculates the gradient of the loss
function for every set of inputs in a training data set. The weights and biases are then updated after
calculating the gradient for a single set of inputs. This yields weights that are updated more fre-
quently than the standard Gradient Descent. In most cases, Stochastic Gradient Descent uses more
iterations per training set to converge. However, due to less computation required per iteration, it is
faster to converge than Gradient Descent. For Stochastic Gradient Descent, one iteration is defined
by predicting the output for one set of inputs, calculating the current loss and its gradient, and then
updating all weights/biases accordingly.
Updates to the weights and biases using Stochastic Gradient Descent are calculated based on the
loss at each iteration rather than the cost function:
ℒ ℒ
wi 1 wi a ; i 1 i a (10.233)
wi i
Once the specified convergence of either the loss function or the cost function is achieved, then
the iterations are concluded. The most recent set of updated weights and biases give the final param-
eters for the linear relationship. These weights and biases are then applied to new inputs to give the
predicted outputs.
T
q k (10.234)
x
For a set of varying temperature gradients in a flat plate with a thermal conductivity, k, the pre-
dicted heat flux can be written as
T
ŷ w1 1 (10.235)
x
Using Stochastic Gradient Descent to update the weight and bias, a linear regression model is
created to predict heat flux values based on the temperature gradients. After the loss function has
converged to a sufficiently low value, the following conditions exist:
w1 k ; 1 0 (10.236)
After a predicted output of the heat flux, the loss is calculated based on the squared error loss
function, after which the weight and bias must be updated. To update these values from Stochastic
Gradient Descent, their partial derivatives with respect to the loss function must be calculated.
This is achieved by computing the partial derivatives and applying the chain rule. The partial
derivatives of the model are obtained as follows:
L T L
2 yˆ y ; 2 yˆ y (10.238)
w1 x 1
T
w 2 w1 2a
x
yˆ y ; 2 1 2a yˆ y (10.239)
After the weight and bias have been updated, they are used in the next temperature gradient
calculation to make the predicted output. The loss is calculated and the weights are updated
again. This procedure is repeated for all temperature gradients in the set until the specified con-
vergence of the loss function is met.
Sample results from the linear regression model are shown in the figure. In the calculations,
300 sets of input data with randomly generated temperature gradients are used in the range [–1,
1] and a random value of thermal conductivity in the range (0,1]. From these 300 data sets, 80%
were used to train the model, leaving 60 sets of inputs to test the final equation. The model used
a convergence criterion of the loss function of 1 × 10−11 which resulted in an average 99.9%
accuracy for all predicted outputs. These predicted outputs are shown in the figure on the y–axis
with their corresponding calculated outputs on the x-axis (Figure 10.13).
For a linear regression model, the relationship between the predicted output, y, and corre-
sponding correct output, x, is ideally y = x. This indicates that the predicted output is the exact
same as the expected output. The linear regression model in the figure achieved a final relation-
ship between the predicted output and correct output of y = 1.001x – 6 × 10−7 which is very
accurate. The corresponding Root Mean Square Error is 8.5275 × 10−6. This could be further
improved by lowering the value specified for the convergence criterion of the loss function.
504 Advanced Heat Transfer
FIGURE 10.13 Calculated heat flux and corresponding flux from a linear regression model.
To summarize, random weights are first assigned to each input and a random bias is created.
Based on the size of the training data (number of data sets and inputs per data set), either Gradient
Descent or Stochastic Gradient Descent is used. For Gradient Descent, iterations are performed by
predicting the outputs for all sets of inputs, calculating the cost, and updating the weights until a
specified convergence of the cost function is achieved. If Stochastic Gradient Descent is used, itera-
tions occur by predicting the output for a single set of inputs, calculating the loss, and then updating
the weights until a specified convergence of the loss function is met. Once convergence is reached,
the final weights are then applied to new inputs in order to predict the outputs.
10.10 OTHER METHODS
This chapter has described the most widely used methods of computational heat transfer, including
the finite difference, finite element, and finite volume methods. In addition, further methods have
been developed for specific physical processes and types of problems. This last section will briefly
highlight other methods based on concepts for free surface flows, spectral elements, boundary ele-
ments, Monte Carlo methods, discrete ordinates, and high-resolution schemes.
The volume of fluid method (VOF method; Hirt and Nichols, 1981) is a common method for pre-
dicting and analyzing free surface flows. In this method, a fluid fraction, F, is defined within each
discrete cell (finite volume or element) of the mesh. Values of F = 1, F = 0, and 0 < F < 1 represent
a full cell, empty cell and cell that contains a free surface, respectively. Marker particles are defined
as cell intersection points that identify where the fluid is located in a cell. The perpendicular direc-
tion to the free surface lies in the direction of the gradient of F. Once both the normal direction and
value of F in a boundary cell are known, a line through the cell can be constructed to approximate
the free surface. The motion of F (void movement) can be predicted based on the solution of a tran-
sient advection equation wherein the transient accumulation of F in a control volume balances its
net inflow across the cell surface. Boundary conditions, including surface tension forces, can then
be applied at the free surface. Since the position of the moving interface is generally unknown, itera-
tive or matching procedures from both sides of the free surface are required to track the free surface
movement.
Computational Heat Transfer 505
The spectral element method is a type of weighted residual finite element method that approxi-
mates the governing equations by a weak formulation. Similar to the weighted residual method, the
differential equation is multiplied by an arbitrary basis function and integrated over the domain.
The interpolation and basis functions are normally high-order polynomials up to 10th order. The
high order accuracy leads to a rapid convergence of the method. Since there are a large number of
integrations due to the high order accuracy, a very efficient integration procedure must be used. Sun
and Li (2010) applied a spectral element method to coupled heat conduction and radiation problems.
The solution domain for the radiation equation of transfer and energy equation was discretized by
spectral elements. 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 subdivided 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, Rakheja, 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 several advan-
tages over other methods. Only the boundary of the domain needs to be discretized, 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 solu-
tion of the differential equation, which is generally only available for linear partial differential equa-
tions 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 inte-
gral 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 equa-
tion describing the statistical behavior of a system not in a state of equilibrium). Molecules 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 re-entry aerodynam-
ics and micro-and nano-electromechanical systems. It is a powerful method for the computation
of complex, nonequilibrium gas flows, such as hypersonic flows where nonequilibrium conditions
occur at high altitudes 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–differential radiation equation
of transfer. However, analytical solutions are usually not available even in simple geometrical con-
figurations. The method of discrete ordinates provides an approximate solution by discretizing both
the xyz-domain and angular velocities representing the direction, e.g., 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 discontinui-
ties. 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
506 Advanced Heat Transfer
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 simi-
lar accuracy. Shu (2009) presented a review of high-resolution schemes for convection-dominated
flows and applications in computational fluid dynamics.
REFERENCES
Boyd, I.D. 2015. “Computation of Hypersonic Flows Using the Direct Simulation Monte Carlo Method”, AIAA
Journal of Spacecraft and Rockets, 52, 38–53.
Carlson, B.G. and K.D. Lathrop. 1968. “Transport Theory: The Method of Discrete Ordinates”, in H. Greenspan,
C.N. Kelber, and D. Okrent (Eds), Computing Methods in Reactor Physics, pp. 167–265, New York:
Gordon and Breech.
Chai, J.C., H.S. Lee and S.V. Patankar. 1994. “Finite Volume Method for Radiation Heat Transfer”, AIAA
Journal of Thermophysics and Heat Transfer, 8(3): 419–425.
Cheney, W. and D. Kincaid. 1985. Numerical Mathematics and Computing, 2nd Edition, Brooks/Cole
Publishing Company.
Diaz, G., H. Sen, K.T. Yang and R.L. McClain. 1999. “Simulation of Heat Exchanger Performance by Artificial
Neural Networks”, HVAC&R Research, 5(3): 195–208.
Gross, S. and A. Reusken, 2011. Numerical Methods for Two- phase Incompressible Flows, Berlin:
Springer–Verlag.
Hirt, C.W. and B.D. Nichols, 1981. “Volume of Fluid (VOF) Method for the Dynamics of Free Boundaries”,
Journal of Computational Physics, 39: 201–225.
Ho, C.J. and R. Viskanta, 1984. “Heat Transfer during Inward Melting in a Horizontal Tube”, International
Journal of Heat and Mass Transfer, 27: 709–716.
Kamden, H.T. 2015. “Ray Effects Elimination in Discrete Ordinates and Finite Volume Methods”, AIAA
Journal of Thermophysics and Heat Transfer, 29: 306–318.
Kim, M.Y., S.W. Baek and S. Park. 2008. “Evaluation of the Finite Volume Solutions of Radiative Heat
Transfer in a Complex Two-dimensional Enclosure with Unstructured Polygonal Meshes”, Numerical
Heat Transfer: Part B, 54: 116–137.
Kolaei, A., 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: 43–58.
Mathur, S.R. and J.Y. Murthy. 1999. “Coupled Ordinates Method for Multigrid Acceleration of Radiation
Calculations”, AIAA Journal of Thermophysics and Heat Transfer, 13(4): 467–473.
Minkowycz, W.J., E.M. Sparrow and J.Y. Murthy. 1988. Handbook of Numerical Heat Transfer, New York:
John Wiley & Sons.
Nabity, J.A. 2014. “Modeling a Freezable Water–Based Heat Exchanger for Use in Spacecraft Thermal
Control”, AIAA Journal of Thermophysics and Heat Transfer, 28: 708–716.
Naterer, G.F. and G.E. Schneider, 1995. “PHASES Model of Binary Constituent Solid – Liquid Phase Transition
– Part 1. Numerical Method”, Numerical Heat Transfer B, 28: 111–126.
Ozisik, M.N., 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 Group.
Patankar, S.V. 1980. Numerical Heat Transfer, Washington, DC: Hemisphere.
Reddy, J. N. and D.K. Gartling. 2010. Finite Element Method for Heat Transfer and Fluid Dynamics, 3rd
Edition, Boca Raton: CRC Press/Taylor & Francis Group.
Schneider, G.E. 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.
Schneider, G.E. 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: 363–390.
Shu, C.W. 2009. “High Order Weighted Essentially Non–oscillatory Schemes for Convection Dominated
Problems”, SIAM Review, 51(1): 82–126.
Spelt, P.D., H. Ding, and S.J. Shaw, 2010. Computational Methods for Two-phase Flows, Singapore: World
Scientific Publishing Co Pte Ltd.
Sun, Y.S. and B.W. Li. 2010. “Spectral Collocation Method for Transient Conduction–Radiation Heat Transfer”,
AIAA Journal of Thermophysics and Heat Transfer, 24(4): 823–832.
Van Doormaal, J.P. and G.D. Raithby. 1984. “Enhancements of the SIMPLE Method for Predicting
Incompressible Fluid Flows”, Numerical Heat Transfer, 7: 147–163.
Computational Heat Transfer 507
Wrobel, L.C. and M.H. Aliabadi. 2002. The Boundary Element Method, New York: John Wiley & Sons.
Zeng, M. 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: 320–340.
Zhang, C., 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.
PROBLEMS
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 coef-
ficient 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 varia-
tions of thermal conductivity with temperature.
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/m2K. The left boundary is maintained at 160°C and the
initial wall temperature is 60°C. The thermal conductivity and diffusivity are 50 W/mK 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?
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/m2K. A two-dimensional explicit finite differ-
ence method (Δx = 6 mm, Δy = 5 mm) is used to predict the transient temperature variation
within the chip (ρcp = 1,659 kJ/m3K, 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 boundary condi-
tion? Explain your response.
4. Consider a turbulent flow and decay of turbulent kinetic energy as described by the follow-
ing k – ε equations:
k
t
2
c 2
t k
Both the turbulent kinetic energy (k) and dissipation rate (ε) are positive definite quantities.
Develop an explicit finite difference formulation that ensures that solutions of these equa-
tions remain positive.
5. A weighted residual method uses Wi = P at the nodal points (where P is a large number)
and Wi = 0 away from the nodal points. Describe how this selection influences the residual
and error distributions throughout the problem domain.
6. State three significant advantages of the finite element method over other types of numeri-
cal methods for structured grids.
508 Advanced Heat Transfer
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 element does the ϕ = 150 contour intersect
that side?
8. Are the shape functions linear along the edge of a four-node quadrilateral element? Explain
your response.
9. Describe the differences between a residual solution error and the order of accuracy in the
context of finite element analysis.
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 prob-
lems? Explain your response.
11. Explain how over-relaxation and under–relaxation models affect the solution performance
and accuracy in finite element analysis of heat and fluid flow.
12. Explain how unstructured grid capabilities give a key advantage of finite element methods
in comparison to other conventional numerical methods for structured grids.
13. Under what circumstances should Gaussian quadrature with quadrilateral elements be used
in the weighted residual method?
14. A specified velocity profile is provided at the inlet of a diffuser. The outflow conditions are
unknown. What boundary location and conditions at the outlet would be suitable for inter-
nal flow simulations?
15. Determine the requirement of the shape functions to ensure that a scalar, ϕ, is constant
throughout a finite element.
16. Consider the following differential equation subject to two boundary conditions:
d 2u
4 0
dx 2
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.
18. One-dimensional heat conduction in the x-direction occurs with internal heat generation,
Ṗ, in a solid of length L. The boundary temperatures are maintained at 0°C. Assume steady-
state conditions and constant thermophysical properties. Find the temperature distribution
in the solid using a finite element method with a 4–node grid.
Computational Heat Transfer 509
19. 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, respec-
tively. Use a one-dimensional finite element method to determine the interface tempera-
tures and heat flux through the third (right) section.
20. Heat transfer in a circular fin is governed by the following fin equation:
d 2T
kA hP T T f 0
dx 2
where h = 1 W/cm2K 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.
21. 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/m2K. The right surface is
well insulated. The thermal conductivities 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 temperatures at the external boundaries, steel–fuel interface and fuel element
centerline.
22. 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/m2K.
A uniform internal heat source generates 2 kW/m3 within the element. Use a single trian-
gular element to estimate the nodal temperature distribution in terms of the positions of the
nodal points.
23. 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.
24. 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 tempera-
ture 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 along side 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.
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.
25. 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:
d 2u
g sin 0
dy 2
510 Advanced Heat Transfer
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 linear (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?
26. 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.
27. 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.
2T 2T
k k 2 0
x 2
y
28. The purpose of this question is to write three computer subroutines (SHAPE, STIFF, and
BNDRY) to solve the following two-dimensional field equation with a CVFEM and quad-
rilateral finite elements.
2 2
k 2 2 G Q 0
x y
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 sample ele-
ment 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 (counter-clockwise 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 properties 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 following
boundary conditions: (i) temperature-specified and (ii) flux–specified conditions. Write
a subroutine (BNDRY) to specify the boundary conditions by defining appropriate coef-
ficients, A, B, and C, at the boundary.
A B 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 distribu-
tion. A banded solver can be used to solve the resulting system of algebraic equations. A
grid refinement should be performed to demonstrate grid independence of the results.
512 Advanced Heat Transfer
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 pro-
gram, input files, and numerical results for both validation and application problems. For
the application problem, provide additional background information, problem parameters,
boundary conditions, and a discussion 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 stiffness matrix; c(nnp, nnp) =
global stiffness matrix; rq(nnpe) = element right side; r(nnp) = global right side; dn = shape
functions and derivatives; isemi = semi–bandwidth; cond = conductivity; s, t = local coor-
dinates; dx, dy = Δx and Δy; jac = Jacobian; dqdx, dqdy = scalar derivatives; lc = integra-
tion 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).
29. 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 concentration of a pol-
lutant in an airstream, is transported through a flow field by diffusion and convection pro-
cesses. The purpose of this project is to write two computer subroutines (IPOINT and
STIFF) to solve the following governing differential and finite volume equations with a
CVFEM for two-dimensional steady flows.
u v
S
x y x x y y
v ndS ndS S 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 + ½, 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 following 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 ~ Pe2/(5 + Pe2) 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 + ½.
The relative influences of upstream and downstream nodal values depend on the Peclet
number, Pe, and influence coefficients which pre–multiply each nodal value. Write a com-
puter 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 subroutine as
Computational Heat Transfer 513
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 convection influences are dominant in
the results. Conversely, for low Pe values (including the limit of zero flow), check that dif-
fusion causes uniform transport of ϕ in both upstream and downstream directions.
The presentation should include a brief description of the numerical model, problem
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.
30. 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 convection. 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.
31. Consider heat conduction through a one-dimensional layer with a constant thickness, Δx,
thermal conductivity, k, and cross-sectional area, A. Create a Python program using mul-
tiple linear regression to predict the resultant heat flow rate, Q, for varying temperatures at
both edges of the surface, T1 and T2, respectively. Include functions to make datasets for
training and testing, split the data into testing and training sets, and initialize the random
weights and biases. In the solution method, calculate the loss function using a squared error
function, and update the weights and biases accordingly. Use Stochastic Gradient Descent
to update the weights and biases with a specified convergence of the loss function of 1 ×
10–9. Use a learning rate of 0.001 and use 80% of the 100 created datasets to train the
model.
Appendices
APPENDIX A: VECTOR AND TENSOR NOTATIONS
A vector is a quantity with a prescribed magnitude and direction. Vectors are denoted by boldface
letters in this book. A unit vector is a vector of unit magnitude. For example, i and j refer to the
unit vectors in the x-and y-directions, (1, 0) and (0, 1), respectively. The symbol | a | designates the
magnitude of the indicated vector.
The dot product between two vectors, a∙b, can be represented by a summation of their respective
individual components. For example, if two vectors are defined by a = axi + ayj and b = bxi + byj,
then their dot product is given by
a b a x bx a yby (A.1)
In an analogous way, matrices are contracted when their individual entries are multiplied with
each other and added together. For example, define:
Tensor notation (or indicial notation) provides a concise method of representing lengthy or com-
plicated 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 vec-
tor, a, and matrix, A, 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 by
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 non-
repeating indices and therefore it is a tensor of rank 2 (matrix). Further details describing the opera-
tions of vectors and tensors can be found in advanced calculus or continuum mechanics textbooks.
515
516Appendices
TABLE B.1
Conversion of Units and Constants
Conversion Factors SI Imperial
Acceleration 1 m/s2 = 4.252 × 107 ft/hr2 = 3.2808 ft/s2
Area 1m 2 = 1550.0 in. 2 = 10.764 ft2
Density 1 kg/m3 = 0.06243 lbm/ft
Dynamic viscosity 1 kg/m·s = 2419.1 lbm/ft·h
Energy 1 kJ = 0.9478 Btu = 737.56 ft·lbf
Force 1N = 0.22481 lbf
Heat flux 1 W/m2 = 0.3171 Btu/h·ft2
Heat transfer coefficient 1 W/m2K = 0.1761 Btu/ft2 h °F
Heat transfer rate 1W = 3.4123 Btu/h = 1.341 × 10–3 hp
Kinematic viscosity 1 m2/s = 10.7636 ft2/s
Latent heat 1 kJ/kg = 0.4299 Btu/lbm
Length 1m = 39.37 in. = 3.2808 ft
Mass 1 kg = 2.2046 lbm = 1.1023 × 10–3 U.S. tons
Mass diffusivity 1 m2/s = 10.7636 ft2/s = 3.875 × 104 ft2/h
Mass flow rate 1 kg/s = 7936.6 lbm/h
Mass transfer coefficient 1 m/s = 1.181 × 104 ft/h
Pressure, stress 1 Pa = 1.4504 × 10–4 lbf/in.2
Specific heat 1 kJ/kgK = 0.2388 Btu/lbm·°F = 0.2389 cal/g·°C
Temperature K = °C + 273.15 = (5/9)(°F + /459.67)
R = °F + 459.67 = (9/5)(°K)
Temperature difference 1K = 1 °C = (9/5)°F
Thermal conductivity 1 W/mK = 0.57782 Btu/h·ft·°F
Thermal diffusivity 1 m2/s = 10.7636 ft2/s = 3.875 × 104 ft2/h
Thermal resistance 1 K/W = 0.5275°F/h·Btu
Velocity 1 m/s = 3.2808 ft/s = 3.6 km/h
Volume 1m 3 = 264.17 gal (U.S.) = 1000 L
Volume flow rate 1 m /s3 = 1.585 × 10 gal/min
4 = 2118.9 ft3/min
SI Unit Conversions
Prefix (Symbol) Multiplier
Tera (T) 1012
Giga (G) 109
Mega (M) 106
Kilo (k) 103
Milli (m) 10–3
(Continued)
Appendices 517
TABLE C.1
Governing Equations in Cartesian, Cylindrical, and Spherical Coordinates
Cartesian Coordinates: (x, y, z)
Mass Conservation
(Continued)
518Appendices
u u u u p 2u 2u 2u
u v w 2 2 2 Fx (C.2)
t x y z x x y z
T T T T 2T 2T 2T
cp u v w k 2 2 2 q (C.5)
t x y z x y z
u 2 v 2 w 2 u v 2 v w 2 w u 2
2 (C.6)
x y z y x z y x z
Cylindrical Coordinates: (r, θ, z)
Mass Conservation
1
t r r
rvr 1r v z vz 0 (C.7)
v v v v v2 v p 2 v 1 vr vr 1 2 vr 2 v 2 vr
r vr r r vz r 2r 2 2 2 Fr (C.8)
t r r r z r r r r r r 2 r 2 z
(Continued)
Appendices 519
v v v v v p 2 v 1 vz 1 2 vz 2 vz
z vr z z vz z 2z 2 Fz (C.10)
t r r z z r r r r 2 2 z
Energy Equation
T T v T T 1 T 1 2T 2T
cp vr vz k r r r r 2 2 q
t r r z r
2
z (C.11)
v 1 v
2
v v
2 2
v v
2 2
1 vr 1 vz v vr vz
2
2 r r z (C.12)
r r r z r r r r z z r
1
t r 2 r
r 2 vr
1
r sin
1
v sin r sin
v 0
(C.13)
Dv v2 v2 p 2v 2 v 2v cot 2 v
r 2 vr 2r 2 2 2 Fr (C.14)
Dt r r r r r r r sin
where
D v v
vr (C.17)
Dt t r r r sin
1 2 1 1 2
2 r 2 sin 2 2 (C.18)
r r r r sin
2
r sin 2
(Continued)
520Appendices
T T v T v T
cp vr
t r r r sin
(C.19)
1 2 T 1 T 1 2T
k 2 r 2 sin 2 2 q
r r r r sin r sin 2
v 2 1 v v 1 v vr v cot v
2 2
1 vr
2
(C.20)
2 r r r
r r r r sin r r r r r
2 2
sin v 1 v 1 vr v
r
r r sin r sin r sin r r
TABLE D.1
Properties of Metals at STP (101 kPa, 25°C)
Thermal Specific Coefficient Heat of
Melting Boiling Conductivity, Heat, cp of Expansion, Density, Fusion, hsl
Metal Point (°C) Point (°C) k (W/mK) (kJ/kgK) β (×106/K) ρ (kg/m3) (kJ/kg)
Aluminum 660 2441 237.0 0.900 25 2700 397.8
Antimony 630 1440 18.5 0.209 9
Beryllium 1285 2475 218 1.825 12
Bismuth 271.4 1660 8.4 0.126 13
Cadmium 321 767 93 0.230 30
Chromium 1860 2670 91 0.460 6 7150 330.8
Cobalt 1495 2925 69 0.419 12 8860 276.4
Copper 1084 2575 398 0.385 16.6 8960 205.2
Gold 1063 2800 315 0.130 14.2 19300 62.8
Iridium 2450 4390 147 0.130 6
Iron 1536 2870 80.3 0.452 12 7870 272.2
Lead 327.5 1750 34.6 0.130 29 11300 23.0
Magnesium 650 1090 159 1.017 25
Manganese 1244 2060 7.8 0.477 22
Mercury –38.86 356.55 8.4 0.138
Molybdenum 2620 4651 140 0.251 5 10200 288.9
Nickel 1453 2800 89.9 0.444 13 8900 297.3
Niobium 2470 4740 52 0.268 7
Osmium 3025 4225 61 0.130 5
Platinum 1770 3825 73 0.134 9
Plutonium 640 3230 8 0.134 54
Potassium 63.3 760 99 0.753 83
(Continued)
Appendices 521
Source: Data reprinted with permission from Weast, R.C., ed., CRC Handbook of Tables for Applied Engineering Science,
Table 1–60, Boca Raton: CRC Press/Taylor & Francis Group, 1970.
TABLE D.2
Properties of Nonmetals and Phase Change Materials
Material Density, ρ (kg/m3) Thermal Conductivity, k (W/mK) Specific Heat, cp (kJ/kgK)
Asbestos millboard 1400 0.14 0.837
Asphalt 1100 1.67
Brick, common 1750 0.71 0.920
Brick, hard 2000 1.3 1.00
Chalk 2000 0.84 0.900
Charcoal, wood 400 0.088 1.00
Coal, anthracite 1500 0.26 1.26
Concrete, light 1400 0.42 0.962
Concrete, stone 2200 1.7 0.753
Corkboard 200 0.04 1.88
Earth, dry 1400 1.5 1.26
Fiberboard, light 240 0.058 2.51
Fiberboard, hard 1100 0.2 2.09
Firebrick 2100 1.4 1.05
Glass, window 2500 0.96 0.837
Gypsum board 800 0.17 1.09
Ice (0°C) 900 2.2 2.09
Leather, dry 900 0.2 1.51
(Continued)
522Appendices
Source: Kenisarin, M., Mahkamov, K., “Salt Hydrates as Latent Heat Storage Materials: Thermophysical Properties and
Costs,” Solar Energy Materials and Solar Cells, Tables 1, 22, vol. 145, pp. 255–286, 2016; Data reprinted with
permission from Weast, R.C., ed., CRC Handbook of Tables for Applied Engineering Science, Table 1–110, Boca
Raton: CRC Press/Taylor & Francis Group, 1970.
Appendices 523
TABLE E.1
Properties of Air at Atmospheric Pressure
Temperature, Density, Specific Heat, Viscosity, Thermal Conductivity,
T (K) ρ (kg/m3) cp (kJ/kgK) μ (kg/ms) k (W/mK) Pr
150 2.367 1.010 10. 28 × 10 –6 0.014 0.758
200 1.769 1.006 13.28 × 10–6 0.018 0.739
250 1.413 1.005 15.99 × 10–6 0.022 0.722
260 1.359 1.005 16.50 × 10–6 0.023 0.719
270 1.308 1.006 17.00 × 10–6 0.024 0.716
275 1.285 1.006 17.26 × 10–6 0.024 0.715
280 1.261 1.006 17.50 × 10–6 0.025 0.713
290 1.218 1.006 17.98 × 10–6 0.025 0.710
300 1.177 1.006 18.46 × 10–6 0.026 0.708
310 1.139 1.007 18.93 × 10–6 0.027 0.705
320 1.103 1.007 19.39 × 10–6 0.028 0.703
330 1.070 1.008 19.85 × 10–6 0.029 0.701
340 1.038 1.008 20.30 × 10–6 0.029 0.699
350 1.008 1.009 20.75 × 10–6 0.030 0.697
400 0.882 1.014 22.86 × 10–6 0.034 0.689
450 0.784 1.021 24.85 × 10–6 0.037 0.684
500 0.706 1.030 26.70 × 10–6 0.040 0.680
550 0.642 1.040 28.48 × 10–6 0.044 0.680
600 0.588 1.051 30.17 × 10–6 0.047 0.680
700 0.504 1.075 33.32 × 10–6 0.052 0.684
800 0.441 1.099 36.24 × 10–6 0.058 0.689
900 0.392 1.121 38.97 × 10–6 0.063 0.696
1000 0.353 1.142 41.53 × 10–6 0.068 0.702
Source: Data reprinted with permission from Weast, R.C., ed., CRC Handbook of Tables for Applied Engineering Science,
Table 1–2, Boca Raton: CRC Press/Taylor & Francis Group, 1970.
TABLE E.2
Properties of Other Gases at STP (110 kPa, 25°C)
Density, Specific Heat, Gas Constant Thermal Conductivity, Dynamic Viscosity,
Gas ρ (kg/m3) cp (kJ/kgK) (J/kg°C) k (W/mK) μ (kg/ms)
Acetylene, C2H2 1.075 1.674 319 0.024 1.0 × 10–5
Ammonia, NH3 0.699 2.175 488 0.026 1.0 × 10–5
Argon, Ar 1.608 0.523 208 0.0172 2.0 × 10–5
(Continued)
524Appendices
Source: Data reprinted with permission from Weast, R.C., ed., CRC Handbook of Tables for Applied Engineering Science,
Table 1–14, Boca Raton: CRC Press/Taylor & Francis Group, 1970.
TABLE E.3
Properties of Other Gases (Effects of Temperature)
Temperature, Density, Specific Heat, Thermal Conductivity, Dynamic Viscosity,
Gas T (°C) ρ (kg/m3) cp (kJ/kgK) k (W/mK) μ (kg/ms)
Ammonia, NH3 0 0.956 2.176 0.022 9.18 × 10–6
20 0.894 2.176 0.024 9.82 × 10–6
50 0.811 2.176 0.027 1.09 × 10–5
Argon,Ar –13 1.87 0.523 0.016 2.04 × 10–5
–3 1.81 0.523 0.016 2.11 × 10–5
7 1.74 0.523 0.017 2.17 × 10–5
27 1.62 0.523 0.018 2.30 × 10–5
77 1.39 0.519 0.020 2.59 × 10–5
227 0.974 0.519 0.026 3.37 × 10–5
727 0.487 0.519 0.043 5.42 × 10–5
1227 0.325 0.519 0.055 7.08 × 10–5
Butane, C4H10 0 2.59 1.591 0.013 6.84 × 10–6
100 1.90 2.026 0.023 9.26 × 10–6
200 1.50 2.454 0.036 1.17 × 10–5
300 1.24 2.812 0.052 1.40 × 10–5
400 1.05 3.127 0.069 1.64 × 10–5
500 0.916 3.402 0.090 1.87 × 10–5
600 0.812 3.642 0.113 2.11 × 10–5
(Continued)
526Appendices
Source: Data reprinted with permission from Weast, R.C., ed., CRC Handbook of Tables for Applied Engineering Science,
Table 1–19, Boca Raton: CRC Press/Taylor & Francis Group, 1970.
TABLE E.4
Ideal Gas Properties of Selected Gases (Effects of Temperature)
Carbon Dioxide, CO2 Oxygen, O2 Nitrogen, N2
h of = –393,520 kJ/kmol h of = 0 kJ/kmol h of = 0 kJ/kmol
(Continued)
528Appendices
Adapted from D.R. Stull, 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.
Appendices 529
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
Hydrogen 6.11 × 10–5 7.12 × 10–5 0.217 0.216
Ammonia 1.98 × 10 –5 2.29 × 10 –5 0.669 0.673
Nitrogen 1.78 × 10 –5 0.744
Oxygen 1.78 × 10 –5 2.06 × 10 –5 0.744 0.748
Carbon dioxide 1.42 × 10 –5 1.64 × 10 –5 0.933 0.940
Methyl alcohol 1.32 × 10–5 1.59 × 10–5 1.00 0.969
Formic acid 1.31 × 10–5 1.59 × 10–5 1.01 0.969
Acetic acid 1.06 × 10–5 1.33 × 10–5 1.25 1.16
Ethyl alcohol 1.02 × 10 –5 1.19 × 10 –5 1.30 1.29
Chloroform 9.1 × 10–6 1.46
Diethylamine 8.84 × 10 –6 1.05 × 10 –5 1.50 1.47
n–Propyl alcohol 8.5 × 10–6 1.00 × 10 –5 1.56 1.54
Propionic acid 8.46 × 10–6 9.9 × 10–6 1.57 1.56
Methyl acetate 8.40 × 10–6 1.00 × 10–5 1.58 1.54
Butylamine 8.21 × 10–6 1.01 × 10–5 1.61 1.53
Ethyl Ether 7.86 × 10 –6 9.3 × 10–6 1.69 1.66
Benzene 7.51 × 10 –6 8.8 × 10–6 1.76 1.75
Ethyl acetate 7.15 × 10 –6 8.5 × 10–6 1.85 1.81
Toluene 7.09 × 10 –6 8.4 × 10–6 1.87 1.83
n–Butyl alcohol 7.03 × 10 –6 9.0 × 10–6 1.88 1.71
i–Butyric acid 6.79 × 10–6 8.1 × 10–6 1.95 1.90
Chlorobenzene 7.3 × 10–6 2.11
Aniline 6.10 × 10–6 7.2 × 10–6 2.17 2.14
Xylene 5.9 × 10–6 7.1 × 10–6 2.25 2.17
Amyl alcohol 5.89 × 10 –6 7.0 × 10–6 2.25 2.20
n–octane 5.05 × 10 –6 6.0 × 10–6 2.62 2.57
Naphthalene 5.13 × 10 –6 5.2 × 10–6 2.58 2.96
Source: Data reprinted with permission from Weast, R.C., ed., CRC Handbook of Tables for Applied Engineering Science,
Table 5–47, Boca Raton: CRC Press/Taylor & Francis Group, 1970.
530Appendices
TABLE F.1
Properties of Liquids (300 K, 1 atm, 297 K)
Thermal
Density, Specific Heat, Dynamic Viscosity, Conductivity,
Liquid ρ (kg/m3) cp (kJ/kgK) μ (kg/ms) k (W/mK) Freezing Point (K)
Acetic acid 1049 2.18 0.001155 0.171 290
Acetone 784.6 2.15 0.000316 0.161 179.0
Alcohol, ethyl 785.1 2.44 0.001095 0.171 158.6
Alcohol, methyl 786.5 2.54 0.00056 0.202 175.5
Alcohol, propyl 800.0 2.37 0.00192 0.161 146
Ammonia 823.5 4.38 0.353
Benzene 873.8 1.73 0.000601 0.144 278.68
Bromine 0.473 0.00095 245.84
Carbon disulfide 1261 0.992 0.00036 0.161 161.2
Carbon tetrachloride 1584 0.866 0.00091 0.104 250.35
Castor oil 956.1 1.97 0.650 0.180 263.2
Chloroform 1465 1.05 0.00053 0.118 209.6
Decane 726.3 2.21 0.000859 0.147 243.5
Dodecane 754.6 2.21 0.001374 0.140 247.18
Ether 713.5 2.21 0.000223 0.130 157
Ethylene glycol 1097 2.36 0.0162 0.258 260.2
Fluorine, R–11 1476 0.870 0.00042 0.093a 162
Fluorine, R–12 1311 0.971 0.071a 115
Fluorine, R–22 1194 1.26 0.086a 113
Glycerine 1259 2.62 0.950 0.287 264.8
Heptane 679.5 2.24 0.000376 0.128 182.54
Hexane 654.8 2.26 0.000297 0.124 178.0
Iodine 2.15 386.6
Kerosene 820.1 2.09 0.00164 0.145
Linseed oil 929.1 1.84 0.0331 253
Mercury 0.139 0.00153 234.3
Octane 698.6 2.15 0.00051 0.131 216.4
Phenol 1072 1.43 0.0080 0.190 316.2
Propane 493.5 2.41 0.00011 85.5
Propylene 514.4 2.85 0.00009 87.9
Propylene glycol 965.3 2.50 0.042 213
Sea water 1025 3.76–4.10 270.6
Toluene 862.3 1.72 0.000550 0.133 178
Turpentine 868.2 1.78 0.001375 0.121 214
Water 997.1 4.18 0.00089 0.609 273
(Continued)
Appendices 531
Source: Data reprinted with permission from Weast, R.C., ed., CRC Handbook of Tables for Applied Engineering Science,
Table 1–46, Boca Raton: CRC Press/Taylor & Francis Group, 1970.
532Appendices
TABLE F.2
Properties of Saturated Water
T (oC) P (kPa) ρf (kg/m3) ρv (kg/m3) hfg (kJ/kg) cρ,f (J/kgK) μf · 106 (kg/ms) kf (W/mK) Prf σf (N/m)
0.01 0.612 999.8 0.005 2501 4229 1791 0.561 13.50 0.0757
10 1.228 999.7 0.009 2477 4188 1308 0.580 9.444 0.0742
20 2.339 998.2 0.017 2453 4182 1003 0.598 7.010 0.0727
30 4.246 995.6 0.030 2430 4182 798 0.615 5.423 0.0712
40 7.381 992.2 0.051 2406 4183 653 0.631 4.332 0.0696
50 12.34 988.0 0.083 2382 4181 547.1 0.644 3.555 0.0680
60 19.93 983.2 0.130 2358 4183 466.8 0.654 2.984 0.0662
70 31.18 977.8 0.198 2333 4187 404.5 0.663 2.554 0.0645
80 47.37 971.8 0.293 2308 4196 355.0 0.670 2.223 0.0627
90 70.12 965.3 0.423 2283 4205 315.1 0.675 1.962 0.0608
100 101.3 958.4 0.597 2257 4217 282.3 0.679 1.753 0.0589
110 143.2 951.0 0.826 2230 4233 255.1 0.682 1.584 0.0570
120 198.5 943.2 1.121 2202 4249 232.2 0.683 1.444 0.0550
130 270.0 934.9 1.495 2174 4267 212.8 0.684 1.328 0.0529
140 361.2 926.2 1.965 2145 4288 196.3 0.683 1.232 0.0509
150 475.7 917.1 2.545 2114 4314 182.0 0.682 1.151 0.0488
160 617.7 907.5 3.256 2082 4338 169.6 0.680 1.082 0.0466
170 791.5 897.5 4.118 2049 4368 158.9 0.677 1.025 0.0444
180 1001.9 887.1 5.154 2015 4404 149.4 0.673 0.977 0.0422
190 1254.2 876.2 6.390 1978 4444 141.0 0.669 0.937 0.0400
200 1553.7 864.7 7.854 1940 4489 133.6 0.663 0.904 0.0377
220 2317.8 840.3 11.61 1858 4602 121.0 0.650 0.857 0.0331
240 3344.7 813.5 16.74 1766 4759 110.5 0.632 0.832 0.0284
260 4689.5 783.8 23.70 1662 4971 101.5 0.609 0.828 0.0237
280 6413.2 750.5 33.15 1543 5279 93.4 0.581 0.848 0.0190
300 8583.8 712.4 46.15 1405 5751 85.8 0.548 0.901 0.0144
320 11279 667.4 64.6 1239 6536 78.4 0.509 1.006 0.0099
340 14594 610.8 92.7 1028 8241 70.3 0.469 1.236 0.0056
360 18655 528.1 143.7 721 14686 60.2 0.428 2.068 0.0019
373 21799 402.4 242.7 276 21828 46.7 0.545 18.69 0.0001
Source: Data reprinted with permission from Weast, R.C., ed., CRC Handbook of Tables for Applied Engineering Science,
Tables 1–7, 1–8, Boca Raton: CRC Press/Taylor & Francis Group, 1970.
Appendices 533
TABLE F.3
Properties of Liquid Metals (101 kPa)
Density, Specific Heat, Thermal Conductivity, Dynamic Viscosity,
Metal Melting Point, (°C) T (°F) T (°C) ρ (kg/m3) cv (kJ/kgK) k (W/mK) μ (kg/ms)
Aluminum (1220) 1,300 704 2,370 1.084 104.2 2.8 × 10–3
1,350 732 2,360 1.084 109.7 2.4 × 10–3
1,400 760 2,350 1.084 111.3 2.0 × 10–3
1,450 788 2,340 1.084 121.0 1.6 × 10–3
Bismuth (520) 600 316 10,000 0.144 16.4 1.62 × 10–3
800 427 9,870 0.149 15.6 1.34 × 10–3
1,000 538 9,740 0.154 15.6 1.10 × 10–3
1,200 649 9,610 0.159 15.6 0.923 × 10–3
Lead (621) 700 371 10,500 0.159 16.1 2.39 × 10–3
850 454 10,400 0.155 15.6 2.05 × 10–3
1,000 538 10,400 0.155 15.4 1.74 × 10–3
1,150 621 10,200 0.155 15.1 1.52 × 10–3
Lithium (355) 400 204 506 4.184 41.5 0.595 × 10–3
600 316 497 4.184 39.8 0.506 × 10–3
800 427 489 4.184 38.1 0.551 × 10–3
Mercury (–38) 50 10 13,600 0.138 8.3 1.59 × 10–3
200 93 13,400 0.138 10.4 1.25 × 10–3
300 149 13,200 0.138 11.6 1.10 × 10–3
400 204 13,100 0.134 12.5 0.997 × 10–3
600 316 12,800 0.134 14.0 0.863 × 10–3
Tin (449) 500 260 6,940 0.243 32.9 1.82 × 10–3
700 371 6,860 0.251 33.6 1.46 × 10–3
850 454 6,810 0.259 32.9 1.26 × 10–3
1,000 538 6,740 0.268 32.9 1.13 × 10–3
1,200 649 6,680 0.276 32.9 0.997 × 10–3
Zinc (787) 850 454 6,900 0.498 58.3 3.12 × 10–3
1,000 538 6,860 0.485 57.5 2.56 × 10–3
1,200 649 6,760 0.473 56.8 2.07 × 10–3
1,500 816 6,740 0.448 56.4 1.46 × 10–3
Source: Data reprinted with permission from Weast, R.C., ed., CRC Handbook of Tables for Applied Engineering Science,
Table 1–53, Boca Raton: CRC Press/Taylor & Francis Group, 1970.
534Appendices
TABLE G.1
Radiative Properties of Selected Materials
Total Radiation Emissivities 0°C–38°C 260°C–538°C
Metallic Materials (Clean, Dry)
Polished aluminum, silver, gold, brass, tin 0.02–0.024 0.03–0.10
Polished brass, copper, steel, nickel 0.03–0.08 0.06–0.2
Polished chromium, platinum, mercury 0.03–0.08 0.06–0.2
Dull, smooth, clean aluminum and alloys 0.08–0.20 0.15–0.45
Dull, smooth, clean copper, brass, nickel, iron 0.08–0.20 0.15–0.45
Dull, smooth, clean stainless steel, lead, zinc 0.08–0.20 0.15–0.45
Rough-ground, smooth-machined castings 0.15–0.25 0.3–0.65
Steel mill products, sprayed metal, molten metal 0.15–0.25 0.3–0.65
Smooth, slightly oxidized aluminum, copper 0.2–0.4 0.3–0.7
Smooth, slightly oxidized brass, lead, zinc 0.2–0.4 0.3–0.7
Bright aluminum, gilt, bronze paints 0.3–0.55 0.4–0.7
Heavily oxidized and rough iron, steel 0.6–0.85 0.7–0.9
Heavily oxidized and rough copper, aluminum 0.6–0.85 0.7–0.9
Nonmetallic Materials
White or light-colored paint, plaster, brick, porcelain 0.80–0.95 0.6–0.85
White or light-colored tile, paper, plastics, asbestos 0.80–0.95 0.6–0.85
Medium red, brown, green, and other colors of paint, brick, 0.85–0.95 0.70–0.85
tile, inks, clays, stone, concrete, wood, water
Glass and translucent plastics, oil, varnish, ice 0.85–0.95 0.75–0.95
Carbon black, tar, asphalt, matte–black paints 0.90–0.97 0.90–0.97
Normal Emissivities of Glass 50°C 250°C
Thickness of 1/4 in. (6.35 mm)
Borosilicate, low-expansion 0.89 0.90
96% silica 0.87 0.81
Soda-lime plate 0.91 0.91
Thickness of 1/2 in. (12.7 mm)
Borosilicate, low expansion 0.89 0.90
96/% silica 0.87 0.83
Soda-lime plate 0.91 0.92
Total Solar Absorptivities 0.3–2.5 μm
Surface Material 0.1–0.3
White surfaces: paint, paper, plaster, plastics, fresh snow
Light-colored surfaces: paint, paper, textiles, stone, dry grass 0.25–0.5
(Continued)
Appendices 535
Source: Data reprinted with permission from Weast, R.C., ed., 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 Group, 1970.
536Appendices
TABLE G.2
Total Emissivity of Carbon Dioxide and Water Vapor
P × L (ft × atm) 500 R 1000 R 1500 R 2000 R 2500 R 3000 R
(total pressure 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 Weast, R.C., ed., CRC Handbook of Tables for Applied Engineering Science,
Table 2–11, Boca Raton: CRC Press/Taylor & Francis Group, 1970.
TABLE G.3
Total Emissivities of Gases (Dependence on Partial Pressure)
P×L P×L P×L P×L P×L
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
H2O 1000 538 0.02 0.07 0.11 0.24 0.32
1500 816 0.01 0.06 0.08 0.20 0.27
2000 1093 — 0.04 0.06 0.16 0.23
2500 1371 — 0.03 0.05 0.14 0.19
CO2 1000 538 0.03 0.06 0.07 0.12 0.14
1500 816 0.03 0.06 0.07 0.12 0.14
2000 1093 0.03 0.06 0.07 0.12 0.14
2500 1371 0.02 0.05 0.06 0.11 0.13
(Continued)
Appendices 537
Source: Data reprinted with permission from Weast, R.C., ed., CRC Handbook of Tables for Applied Engineering Science,
Table 2–12, Boca Raton: CRC Press/Taylor & Francis Group, 1970.
TABLE H.1
Atomic Weights of Elements
Element kg/kmol Element kg/kmol Element kg/kmol
Hydrogen, H 1.01 Bromine, Br 79.90 Thulium, Tm 168.93
Helium, He 4.00 Krypton, Kr 83.80 Ytterbium, Yb 173.04
(Continued)
538Appendices
Source: Data reprinted with permission from Hewitt, G.F., et al., eds., International Encyclopedia of Heat and Mass
Transfer, Table 3–1, Boca Raton: CRC Press/Taylor & Francis Group, 1997.
TABLE I.1
Selected Thermodynamic Properties at STP (101 kPa, 25°C)
h fo g of so c po
Compound (kJ/mol) (kJ/mol) (J/K mol) (J/K mol)
AgCl(s) –127.07 –109.80 96.2 50.79
Ar(g) 0 0 154.734 20.786
Br(g) 111.88 82.429 174.912 20.786
Br2(g) 30.907 3.14 245.35 36.0
Br2(liq) 0 0 152.23 75.688
CO(g) –110.52 –137.15 197.56 29.12
CO2(g) –393.51 –394.36 213.6 37.1
(Continued)
Appendices 539
Adapted from M. Kaufman, Principles of Thermodynamics, Appendix C, Boca Raton: CRC Press/Taylor & Francis Group,
2002.
540Appendices
TABLE I.2
Heats of Combustion
Molecular Formula Name ∆hco (J/K mol) Molecular Formula Name ∆hco (J/K mol)
Inorganic Substances C3H8O 1–Propanol (l) 2021.3
C Carbon (graphite) 393.5 C3H8O3 Glycerol (l) 1655.4
CO Carbon monoxide (g) 283.0 C4H10O Diethyl ether (l) 2723.9
H2 Hydrogen (g) 285.8 C5H12O 1–Pentanol (l) 3330.9
H3N Ammonia (g) 382.8 C6H6O Phenol (s) 3053.5
H4N2 Hydrazine (g) 667.1 Carbonyl Compounds
N2O Nitrous oxide (g) 82.1 CH2O Formaldehyde (g) 570.7
Hydrocarbons C2H2O Ketene (g) 1025.4
CH4 Methane (g) 890.8 C2H4O Acetaldehyde (l) 1166.9
C2H2 Acetylene (g) 1301.1 C3H6O Acetone (l) 1789.9
C2H4 Ethylene (g) 1411.2 C3H6O Propanal (l) 1822.7
C2H6 Ethane (g) 1560.7 C4H8O 2-butanone (l) 2444.1
C3H6 Propylene (g) 2058.0 Acids and Esters
C3H6 Cyclopropane (g) 2091.3 CH2O2 Formic acid (l) 254.6
C3H8 Propoane (g) 2219.2 C2H4O2 Acetic acid (l) 874.2
C4H6 1,3–Butadiene (g) 2541.5 C2H4O2 Methyl formate (l) 972.6
C4H10 Butane (g) 2877.6 C3H6O2 Methyl acetate (l) 1592.2
C5H12 Pentane (l) 3509.0 C4H8O2 Ethyl acetate (l) 2238.1
C6H6 Benzene (l) 3267.6 C7H6O2 Benzoic acid (s) 3226.9
C6H12 Cyclohexane (l) 3919.6 Nitrogen Compounds
C6H14 Hexane (l) 4163.2 CHN Hydrogen cyanide (g) 671.5
C7H8 Toluene (l) 3910.3 CH3NO2 Nitromethane (l) 709.2
C7H16 Heptane (l) 4817.0 CH5N Methylamine (g) 1085.6
C10H8 Naphthalene (s) 5156.3 C2H3N Acetonitrile (l) 1247.2
Alcohols and Ethers C2H5NO Acetamide (s) 1184.6
CH4O Methanol (l) 726.1 C3H9N Trimethylamine (g) 2443.1
C2H6O Ethanol (l) 1366.8 C5H5N Pyridine (l) 2782.3
C2H6O Dimethyl ether (g) 1460.4 C6H7N Aniline (l) 3392.8
C2H6O2 Thylene glycol (l) 1189.2
Adapted from M. Kaufman, Principles of Thermodynamics, Appendix C, Boca Raton: CRC Press/Taylor & Francis Group,
2002.
Appendices 541
REFERENCES
Hewitt, G.F., G.L. Shires, and Y.V. Polezhaev, eds., 1997. International Encyclopedia of Heat and Mass
Transfer, Boca Raton: CRC Press/Taylor & Francis Group.
Kaufman, M., 2002. Principles of Thermodynamics, Boca Raton: CRC Press/Taylor & Francis Group.
Kenisarin, M., K. Mahkamov. 2016. “Salt Hydrates as Latent Heat Storage Materials: Thermophysical
Properties and Costs”, Solar Energy Materials and Solar Cells, Tables 1, 22, vol.145, pp. 255–286.
Stull, D.R., H. Prophet, 1971. JANAF Thermochemical Tables, 2nd Edition, Washington, DC: U.S. National
Bureau of Standards, Office of Standard Reference Data.
Weast, R.C., ed., 1970. CRC Handbook of Tables for Applied Engineering Science, Boca Raton: CRC Press/
Taylor & Francis Group.
Index
Page numbers in italics denote figures and those in bold denote tables
A Boltzmann, L., 13
Boltzmann equation, 65, 505
absorption coefficient, 171, 174, 175, 179, 292, 376, 492, Bose, T.K., 336
493 Bouguer’s law, 171–173
absorptivity, 151, 158–161, 159, 188, 292 boundary conditions
acceleration, 3–4, 75, 262, 276–278, 305 computational fluid dynamics, 475, 475
Achenbach, E., 108 constant wall heat flux, 118, 414, 443, 444
activation function, 498 surface tension forces, 504
Adeyinka, O.B., 140 time-dependent, 59–61, 59
Ahmed, M., 286 two-dimensional formulation, 470–472
air–fuel mass ratio (AF), 373 types, 23, 23, 24
albedo for scattering, 176 boundary element method, 505
algebraic stress models, 136 boundary layer, 12
Aliabadi, M.H., 505 assumptions, 89
Altman, M., 229 Chilton–Colburn analogy, 92–93
Ameel, T.A., 429, 432 energy balance, 90–91
annular-dispersed flow, 218, 223 experimental setup, 91, 91
annular flow, 214, 215, 218–221 flat plate, 94
apparent entropy production difference, 139–142, 141 integral solution method, 96–98, 97
Armand’s correlation, 229 scaling analysis, 95–96
Arrhenius equation, 359 similarity solution, 98, 98–102, 100–103
artificial neural networks (ANN), 497, 497–501 free convection, see free convection
Ashgriz, N., 286 laminar flow, 88–89
atmospheric irradiation, 187, 187–189 nondimensional correlations, 91, 92
atomic weights of elements, 537–538 pressure gradient, 89–90
Reynolds analogy, 92
B temperature layer thickness, 88, 88, 90
turbulence, 88–89
Backpropagation method, 498 velocity layer thickness, 88, 88, 90
Baker, T.M., 296 wedge flow, 103, 103–104
Bannerot, R.B., 193 boundary value problems, 44
basic oxygen furnace (BOF), 378 Boussinesq approximation, 119
basis functions, 456, 466, 505 Bowman, R.A., 406
batch reactors, 355, 363, 364 Boyd, I.D., 505
Beatty, K.O., 247 Brinkman–extended Darcy model, 419
Beckman, W.A., 201 Bromley, L.A., 211
Beer’s law, 173, 292, 295 Brownian diffusion, 303, 304
Bejan, A., 136, 406, 437, 445 bubble flow, 214, 225
Berenson, P.J., 212 Buckingham equation, 298
Bereny, J.A., 201 Buckingham–Pi theorem, 86–88, 209
Bergman, T.L., 58, 73, 107, 109, 293 Buongiorno, J., 399
biholomorphic map, see conformal mapping
Binghamian slurries, 297, 298
Biot, J.P., 85 C
Biot number (Bi), 54, 55, 58, 85, 264, 277, 278 Camberos, J.A., 136, 445
bipolar coordinates, 51, 52–53 capacitance matrix, 464
Bird, R.B., 17, 366, 367, 370 capillary action, 74, 248–250
blackbody, 13, 14, 151, 156, 156–158, 158, 161, 166, 186, Carey, V.P., 212
187 Carpenter–Colburn correlation, 245
Blake–Kozeny equation, 324, 443 Carslaw, H.S., 21, 41, 343
Blasius, H., 98 Cartesian coordinates, 12, 25, 35–36, 36, 37, 39, 40, 83,
Blasius boundary layer, 100, 103, 103, 104 110, 517–520
blood perfusion, 63 Central Difference Scheme (CDS), 478, 512
body gravity function method, 123–127, 124, 127, 237, 238 ceramics, 2, 151
Bohn, M.S., 127 Chalmers, B., 321
Boiler and Pressure Vessel Code (BPVC), 402 chemically reacting flows
543
544Index
liquid water content, 268, 283, 285 moving bed heat exchangers (MBHEs), 416, 416–418, 418
Liu, H., 403, 416, 436 Mueller, A.C., 406
local convection coefficient, 12, 90 Mujumdar, A.S.M.S., 303
log mean temperature difference, 114, 399, 403–405, 403 multidimensional conduction
London, A.L., 338, 409, 410, 412 Cartesian coordinates, 35–36, 36
loss function, 501–504 curvilinear coordinates, 36–39, 37, 38
Lu, C.T., 162 cylindrical and spherical coordinates, 39–41
Lu, M.L., 285 multiphase flows with droplets and particles
Lu, T.J., 414 carrier phase equations
Lu, W., 419 conservation of energy (total energy equation), 271,
Ludwig–Soret effect, 303 271–273
Lumley, J.L., 128 conservation of mass, 268–269
lumped capacitance method, 53–55, 54 momentum equations, 269, 269–271
lumped thermal capacitance, 54, 54 thermal energy equation, 273–274
lumped transient model, 464–465 volume averaging method, 266–268, 267
Lunardini, V.J., 336, 339, 349 collision–dominated flows, 261
Lynch, P., 294 contact-dominated flows, 261
dispersed phase equations
M convective heat and mass transfer, 264–266, 265
particle equation of motion, 262–264
machine learning droplet to particle transition
artificial neural networks, 497, 497–499 droplet shrinkage, 286–289, 288
conduction and forced convection, 499–501, 500 physical processes, 285–286
linear regression, 501–504, 504 solvent evaporation, 286–289, 288
types, 496–497 dune flow, 262
Maglik, R.M., 127 forced convection melting of particles, 289–291, 290
Majumdar, A.S., 387 heat transfer from droplets
Mandhane, J.M., 244 internal temperature distribution, 277–278
Marangoni effect, 74 lumped capacitance solution, 274, 274–277, 276
marker particles, 504 solidification, 278, 278–282, 281, 282
Martin, M.J., 271 homogeneous flow, 262
mass diffusivity, 15–17, 16 impinging droplets on freezing surface, 282, 282–285,
mass transfer, 15–17, 16, 85, 93, 225, 264–266, 265, 276, 285
277, 387 internal flows with particles
Mathew, B., 423, 425 horizontal transport of solid particles, 299, 299–300
Mathews, J.H., 53 packed bed flow, 300, 300–302
Mathur, S.R., 489 slurries, 297–298
matrix bandwidth, 475 vertical flows in pipelines, 298–299
Maxwell, J.C., 305 nanofluids and nanoparticles, 302–303
mean free path, 3, 65–67, 172, 182, 183, 263, 264 governing transport equations, 304–305
mean intensity, 178 heat transfer correlations, 306, 307
Mean Squared Error (MSE), 501 thermal conductivity, 305–306, 306
melt particularization process, 281, 282 transport phenomena, 303
Menguc, M.P., 171 one-way coupling, 261
metal properties, 520–521 particle–gas interactions, 261
method of pixelation, 489 particle–particle interactions, 261
method of separation of variables, 41–44, 42 pneumatic transport, 261
method of unknown coefficients, 103 radiative heat transfer from particles
metric coefficients, 38 absorption and emission in gas layer, 291–296, 292,
microchannel heat exchangers (MCHX), 423–425, 424, 425 293, 294
microscale conduction, 65, 65–67 particulate radiation, 296–297
Miller, T.F., 296 saltation velocity, 262
Miropolski correlation, 225 two-way coupling, 261
mixed convection, 12 multiphase heat transfer, 1
mixing length model, 132–133 multiphase reacting mixtures
molecular dynamics, 65 physical processes, 377, 377–378
molten aluminum refinement, 377, 377 progressive conversion model, 381, 382
momentum equation, 77–80, 96, 269, 269–271, 323–324, shrinking core model, 378–381, 379, 380
480 multiple linear regression, 501
Moody, L.F., 111 multi–regime Nusselt number correlations, 229–230, 230
Moody chart, 111, 111–112 multispectral radiation thermometry (MRT), 162
Moran, M.J., 374 Murthy, J.Y., 489
Morton number (Mo), 225 Muzychka, Y.S., 35, 116, 118, 123, 126
Index 551
P R
packed bed reactor (PBR), 363, 365 radiation, 12–14, 13, 487
parabolic process, 477 blackbody, 156–158, 156, 158, 166
parallel adiabats, 28, 28 coupled, 183–185, 184, 185
parallel isotherms, 28, 28 DO model, 489, 490, 491, 493
particle fluidization, 384 DTRM, 488–489
Peclet, J., 85 electromagnetic spectrum, 152–153, 153
Peclet number (Pe), 85 finite volume method, 490, 490–493, 491, 493
Pennes, H.H., 63 intensity, 153–155, 154, 155, 171–173, 172, 488
Pennes model, 63–65 participating medium, 171–175, 172, 174
Penoncello, S.G., 403, 436 shield, 170
perturbation solution method, 333–336, 334 solar, 161, 185–195, 186, 187, 189
phase change, 12, 14–15, 15, 317–322, 318, 320, 321, 345, surface properties, 158–162, 159
441–442, 442, 493–496, 494 radiation–diffusion approximation, 183
phase change materials (PCMs), 317, 346, 521–522 radiation exchange, 162–168, 162–165, 168
phase diagram, 14–15, 15, 299, 315, 322 radiation thermometry, 162
phenomenological law, 9, 486 radiative flux vector, 177–179, 178
phonon, 65 radiative properties, 534–537
photons, 13, 65, 152, 183, 488 radio wave propagation, 153
552Index
W White, F., 88
Wrobel, L.C., 505
Wang, X., 275, 276
Wang, X.Q., 303
Wang, Z., 300 X
Wasp, E.J., 299 X-rays, 152, 153
Watel, B., 433 Xu, H., 370, 372
wavepacket, 65, 65, 66
weak formulation, 505
Weber, M.E., 264
Y
Weber, Moritz, 85 Yener, Y., 21, 41, 58
Weber number, 85, 86, 217 Yi, J., 435
wedge flow, 103, 103–105, 105, 184, 184 Yovanovich, M.M., 31, 40, 116, 118, 123, 126
weighted residuals, 455–456
weighted–sum of gray gases model (WSGGM), 375–376
weighting function, 478, 479
Z
Wein’s displacement law, 157 Zedan, M., 487
Wen, C.D., 162 Zeng, M., 482
Wen–Yu correlation, 385 zero-equation model, 132, 134
Whalley, P.B., 212 Zhang, C., 495
Whitaker, S., 302 Zhukauskas correlation, 107, 107, 108
Whitaker correlation, 275 Zuber, N., 218