Book2c With Cover Page v2

Accelerat ing t he world's research.
Optimisation Analyses of Fluid-

Thermal Systems using Excel
Mohamed El-Awad
Related papers Download a PDF Pack of t he best relat ed papers 
Comput er-Aided T hermal Design Opt imisat ion

Mohamed El-Awad
Use of Excel' s 'Goal Seek' feat ure for T hermal-Fluid Calculat ions
Mohamed El-Awad
Modelling t he Regenerat or for Mult i-Object ive Opt imisat ion of t he Air-Bot t oming Cycle wit h Excel - Sp…
Mohamed El-Awad
Design Analyses of Fluid-Thermal
Systems using Excel
Mohamed M. El-Awad
Mohamed M. El-Awad (UTAS)
Design Analyses of Fluid-Thermal Systems

using Excel
Design Analyses of Fluid-Thermal Systems Using Excel
Design Analyses of Fluid-Thermal Systems

Using Excel
Mohamed M. El-Awad
Associate Professor,
University of Technology and Applied Sciences
College of Applied Sciences, Sohar
Ghada This book is dedicated to my beloved family:

February, 2021
The Thermax add-in can be downloaded from:
https://www.mendeley.com/library/fileproxy?id=04a975bf-59cf-cde5-
7abe54d5b118d1e6
Preface
The traditional approach of using hand-calculations with property tables and charts is
an effective method for introducing the principles of thermodynamics, fluid mechanics,
and heat-transfer to engineering students. However, the application of these principles
in design analyses of fluid-thermal systems requires computer-aided methods for
various reasons. A common reason is the need for iterative solutions because of the
nonlinear equations involved or because of the inter-dependence between fluid
properties and the imposed system forces, pressures, and temperatures. The analyses
that deal with complex fluid-thermal systems, such as pipe-networks and heat-
exchanger networks, are best performed with the aid of computers which are also
needed for design optimisation analyses of even the simple fluid-thermal systems like
thermal insulation. Although the use of computer-software for fluid-thermal analyses
nowadays has become common place in the industry, the disadvantage of these
dedicated applications for educational purposes is that they do not allow the student to
develop their own analytical models from the basic principles apart from being
expensive to acquire and require proper training to use.
Microsoft Excel is a general-purpose application that is taught to engineering students

in the introductory courses on computer applications. The powerful tools of Excel for
data analysis and visualisation, together with its wide availability and simplicity,
encouraged its use as an educational tool in various engineering subjects. With respect
fluid-thermal analyses, a number of articles and research papers have been presented in
this regard at various conferences and specialised publications over the past two
decades. The lack of built-in functions for fluid properties also motivated a number of
academic institutions and individual researchers to develop relevant Excel add-ins.
However, a textbook that presents a general pedagogical approach for using Excel as a
platform for computer-aided fluid-thermal analyses and provides the lecturers and
students with sufficiently detailed examples and exercises has, so far, been missing. The
aim of this book is to complement the previous efforts by gathering the information
needed for using Excel as an effective modelling platform for various types of
computer-aided fluid-thermal analyses.
The Excel-based modelling platform described in the book has four elements; (i) Excel
with its user-interface and built-in functions, (ii) the integrated programming language
Visual Basic for Applications (VBA), (iii) the Solver add-in that comes with Excel, and
(iv) an Excel add-in for fluid properties called Thermax. While Excel and Solver are
adequate for most fluid mecanics and heat-transfer analyses, the Thermax add-in helps
the students to perform thermodynamic analyses with Excel in an effective and accurate
manner. VBA is needed for the development of custom functions when the analytical
model cannot be executed by only using Excel‟s built-in functions and Thermax
functions. Proper use of the Excel-based modelling platform minimises the effort of
developing the analytical model so that more attention can be paid to the application of
the physical and economic principles in fluid-thermal analyses.
The book is organised in two main parts. The first four chapters of the book review the
basic principles of thermofluid analyses, describe the four components of the Excel-
based modelling platform, and show how the platform can be used for performing basic
optimisation analyses of fluid-thermal systems. The following six chapters of the book
show how the Excel-based platform can be utilised for design analyses of fluid-thermal
systems that include analyses of thermal insulation, determining the economic diameter
for a pump-pipe system, least-cost optimisation analyses of pipe-networks, design
analyses of pump-pipe systems and heat-exchangers, and optimisation analyses of the
air-bottoming gas-turbine cycle and the gas-turbine cycle with cogeneration and inlet-
air cooling options.
The book adopts a learning-by-example approach and most of its examples are based on
relevant cases or examples given in popular thermofluid textbooks so that the students
can look for additional information if needed and verify their Excel solutions. Exercises
are given at the end of Chapters 2, 3, and 4 that help students to sharpen their skills
related to these chapters. The appendices present more challenging cases of analyses
and provide comprehensive exercises and projects for using the Excel-based platform in
the analyses of fluid and thermal systems.
The material covered in the book, which is adequate for a stand-alone course on design
analyses of fluid-thermal analyses, can be used to supplement the existing courses on
the design of fluid-thermal systems. It is also hoped that the book can be a useful
reference for practicing engineers in the area of fluid-thermal systems.
Acknowledgements
The writing of this book would not have been possible without benefitting from the
efforts of many colleagues who have made their information and software available in
the open literature or their websites. In this respect, a special gratitude goes to the
Mechanical Engineering Department at the University of Alabama (USA) whose
initiatives both inspired and helped me throughout the preparation of the book. I am
also indebted to my colleagues and students at the College of Applied Sciences in Sohar
and hope that they find this book a worthy token of appreciation and gratitude. Last, but
not least, I am grateful to my beloved family, Ghada, Ahmed, and Ula, for their
tolerance and unfailing support that I desperately needed to complete this work.
CONTENTS
Nomenclature xi
1. Introduction 1
1.1. A review of thermofluid principles 2
1.1.1. Thermodynamics 2
1.1.2. Fluid dynamics 7
1.1.3. Heat transfer 10
1.2. Advantages of computer-aided thermofluid analyses 13
1.3. The Excel-based modelling platform for thermofluid analyses 16
1.4. Closure 18
References 19
2. Excel 20
2.1. Elements of Excel‟s user-interface 21
2.2. Excel‟s formulae 22
2.2.1. Cell reference by location 22
2.2.2. Use of cell labels 24
2.3. Excel‟s built-in functions 27
2.3.1. Logical functions 28
2.3.2. Functions for matrix operations 29
2.4. Solution of linear systems of equations 33
2.5. Iterative solutions with Excel 35
2.5.1. Iterative solutions with Goal-Seek 35
2.5.2. Iterative solutions with circular calculations 37
2.6. Excel‟s graphical tools for data presentation and analysis 40
2.7. Closure 43
References 43
Exercises 44
3. Solver , VBA, and Thermax 47

3.1. Solver 48
3.1.1. Activation of Solver 48
3.1.2. The GRG Nonlinear method50
3.1.3. The Evolutionary method 52
3.1.4. The Simplex LP method 54
3.1.5. The default settings of Solver options 56
3.2. VBA and the development of user-defined functions 58
3.2.1. Activation of VBA 59
3.2.2. Development of UDFs 61
3.3. Thermax installation and use 63
3.4. Closure 67
References 67
Exercises 68
4. Optimisation analyses 71
4.1. Analytical vs. Excel-aided optimisation of fluid-thermal systems 72
4.1.1. The approximate analytical optimisation method 73
4.1.2. The exact Excel-aided optimisation method 75
4.2. Optimisation with a single variable 79
4.2.1. Optimum thickness of thermal insulation 79
4.2.2. Optimisation of a rectangular fin 82
4.2.3. Optimum outlet temperature from a heat-exchanger 85
4.2.4. Optimisation of a hot-water generation system with two steam
heaters 88
4.3. Multi-variable optimisation 94
4.4. Optimisation analyses with the Evolutionary method 97
4.5. Closure 99
References 100
Exercises 100
5. Thermal insulation 105

5.1 The analytical model for thermal insulation around a pipe 106
5.2 Determining the insulation thickness from safety considerations 107
5.3 Determining the insulation thickness from economic considerations 110
5.4 Determining the optimum thickness of insulation 112
5.5 Multivariable optimisation of thermal insulation 119
5.6 Closure 123
References 123
6. Determination of the economic pipe diameter 124

6.1. The objective function for pipe-diameter optimisation 12
6.2. The analytical method for pipe-diameter optimisation 128
6.2.1. Optimisation with the Hazen-Williams equation 130
6.2.2. Optimisation with the Darcy-Weisbach equation
(laminar flow) 132
6.2.3. Optimisation with the Darcy-Weisbach equation
(turbulent flow) 134
6.3. An Excel-Solver method for pipe-diameter optimisation 139
6.4. Advantages of the Excel-Solver method over standard programming 142
6.5. Closure 144
References 144
7. Excel-based design procedures for pump-pipe systems and double-pipe heat-

exchangers
7.1. Pump-pipe systems
7.1.1. The design case
7.1.2. The analytical model

7.1.3. The Excel-based design procedure
7.2. Double-pipe heat exchangers
7.3. Closure
References
8. Pipe-network optimisation
8.1. The pipe-network optimisation problem
8.2. The analytical model for pipe-network optimisation
8.3. Optimisation of the two-loop gravity-fed pipe network
8.3.1. Darcy-Weisbach equation with the Swamee-Jain formula
8.3.2. Darcy-Weisbach equation with the Colebrook formula
8.3.3. Optimisation with the Hazen-Williams equation
8.3.4. Effect of Solver options on the optimised solutions
8.4. Optimisation of a two-loop pump-driven pipe network
8.4.2. The optimisation method
8.4.3. Optimisation analyses with alternative friction formulae
8.5. Closure
References
9. Thermodynamic and economic optimisation of the air-botomming cycle

9.1. The analytical model for the air-bottoming gas-turbine system
9.2. Modelling the gas-to-air heat-exchanger
9.2.1. The effectiveness model
9.2.2. The pinch-analysis model
9.3. Thermodynamic optimisation of the AB cycle
9.4. Economic optimisation of the AB cycle
9.5. Solver time
9.6. Closure
References
10. Analyses of the gas-turbine cycle with cogeneration and inlet-air cooling
10.1. Cogeneration
10.1.1. The thermodynamic model
10.1.2. The objective function for optimization
10.2. Otimisation analysis of the cogeneration system
10.3. The analytical model for gas-turbine with inlet-air cooling
10.3.1. The approximate analytical model
10.3.2. The exact analytical model
10.4. Optimisation analyses of the gas-turbine with inlet-air cooling

10.4.1. Optimisation analysis with ideal cycle
10.4.2. Optimisation analysis with actual cycle
10.5. Closure
References
Appendices
Appendix A: Properties of air and liquid water at atmospheric pressure
Appendix B: Methods of heat-transfer analyses for heat-exchangers and fins
Appendix C: The numerical tools provided by Thermax
Appendix D: Optimisation analysis of an air-conditioning duct
Appendix E: Optimisation analysis of a pump-pipe system
Appendix F: Optimisation analyses of a pipe network with eight loops
Appendix G: Comprehensive exercises and projects
Subject index
Nomenclature
A Area
C Friction coefficient in Hazen-Williams equation, Equation (1.28)
cp The cost per linear meter for a pipe as defined by Equation (11.2), $/m
cp Specific heat at constant pressure, kJ/kg· o C
cv Specific heat at constant volume, kJ/kg·o C
D Diameter
E Total energy of the system
e Energy per unit mass of the system
EG Total energy generated in the system
eG Energy generated per unit volume
f Darcy friction factor, defined by Equation (1.21)
f Manning friction factor, defined by Equation (1.27)
F Correction factor, defined by Equation (B.5)
FD Drag force, defined by Equation (C.2)
g Acceleration of gravity
h Average heat-transfer coefficient
h Enthalpy, kJ/kg
hf Major friction in a pipe system
k Thermal conductivity, W/m· o C
k Isentropic exponent, dimensionless (k=cp /cv)
K Minor losses friction coefficient in a pipe system, defined by Equation (1.29)
L Length
m Mass ˙
m Mass rate of flow
N Lifetime of a fluid-thermal system, years
P Pressure, usually in kPa
Pr Relative pressure (for an ideal gas)
q Heat-transfer per kg of the working fluid, usually kJ
Q Heat, usually in kJ
Q Volume flow rate
Q Rate of heat transfer, W or kW
r Radius or radial distance
R Gas constant, kJ/kg.K
Rth Thermal resistance, usually o C/W
Ru Universal gas constant kJ/kmol.K
s Entropy
T Temperature
u Internal energy per unit mass, kJ/kg
U Overall heat-transfer coefficient of a heat-exchanger
v Specific volume, m3 /kg
V Velocity, usually m/s
w Work-done per kg of the working fluid, usually kJ

W Power, W or kW
Z Elevation
x,y,z Space coordinates in the Cartesian system
Greek Characters
α Amortisation rate, defined by Equation (11.4)
γ Specific weight of a fluid
δ Thickness (e.g. of insulation)
∆ Difference (e.g. temperature)
ε Roughness of surface material
ε Heat-exchanger effectiveness
η Efficiency
μ Dynamic viscosity, kg/m.s
ν Kinematic viscosity, m2 /s
ρ Density, kg/m3
 Time, annual operating hours of a system
 Mesh Fourier number in the finite-difference method, defined by Equation (9.6)
Dimensionless Numbers and Groups

Bi Biot number
Cf Friction coefficient, defined by Equation (C.3)
Fi Fourier number
Nu Nusselt number
Pr Prandtl number
Re Reynolds number
Subscripts
f Saturated liquid condition
fg Difference in property between saturated liquid and saturated vapour
g Saturated vapour condition
lm Log-mean
s Saturation temperature or pressure
s Evaluated at the surface
∞ Evaluation at free-stream ambient conditions
Introduction
The energy that operates our cars, refrigerators, and air-conditioners mainly comes from
burning fossil fuels in power-generation plants. Apart from being non-renewable
sources of energy, fossil fuels are the main cause for global warming and its
increasingly devastating effects witnessed nowadays at different parts of the world.
Therefore, proper design and operation of energy-conversion systems are required for
minimising the effects of large-scale utilisation of fossil fuels. The design methods of
fluid-thermal systems, which are heart of any energy-conversion system, are mainly
based on the principles of thermodynamics, fluid mechanics, and heat transfer. This
chapter reviews the main principles of these three thermofluid subjects with the
objective of showing how they can be used to reduce the losses and minimise energy
consumption. For a number of reasons the equations involved in thermofluid analyses
are difficult to solve without using computer-aided methods. Therefore, these analyses
introduce many simplifications that reduce their accuracy. In this respect, the chapter
highlights the advantages of computer-aided methods for thermofluid analyses and
describes the Excel-based modelling platform used in this book for these analyses.
1.1. A review of thermofluid principles

The two main principles that form the framework for thermofluid analyses are the
conservation of mass (the continuity equation) and the conservation of energy (the first-
law of thermodynamics). These principles take different mathematical forms depending
on whether the system under consideration is open or closed and on whether the flow is
steady or unsteady, compressible or incompressible, laminar or turbulent, etc.
Numerous auxiliary relationships are needed in order to quantify the various parameters
involved in the resulting equations such as pressure-variations, friction losses, and rates
of heat-transfer. In what follows, the main concepts of thermodynamics, fluid
dynamics, and heat-transfer are reviewed by considering typical applications.
1.1.1. Thermodynamics
The principles of engineering thermodynamics allow us to determine the amount of
energy transfer in the form of work or heat between any system and its surroundings.
They also allow us to determine the efficiency and effectiveness of the system if energy
conversion is involved. There are four basic laws for thermodynamics the most
important of which are the first and the second laws of thermodynamics. To apply these
two basic laws, thermodynamic analyses use many property relationships, tables, and
charts that determine the properties of the particular fluid involved and its phase (a
liquid, a liquid-vapour mixture, a gas, or a gaseous mixture). To illustrate the
application of thermodynamic laws and relationships in a typical analysis, consider the
air-compression system shown in Figure 1.1.a. Air enters the system that has two stages
of compression separated by an intercooler at a temperature T1 and pressure P1 . The
first-stage compressor, C1 , compresses the air adiabatically to state 2, after which it
enters the intercooler where its temperature is reduced to T3 . The second-stage
compressor, C2 , then increases the air pressure to P4 at which the temperature increases
to T4 . Figure 1.1.b shows the compression process on a temperature-entropy diagram.
T
4 P4
P2
Air 4 2
C1 C2 P1
1
T 3 1
2 Intercooler 3 Intercooling
Cooling
water s
(a) (b)
Figure 1.1. Schematic and T-s diagrams of a two-stage air compressor with inter-stage
intercooling
How the total compression work is divided between the two compressor stages depends
on their compression ratios and there is a certain value of the intermediate pressure (Pi )
that minimises the total compression work. The principles of thermodynamics help us
to determine this optimum pressure as shown below.
Treating the two compressor stages as steady-flow processes and neglecting changes in
kinetic and potential energy, the first-law of thermodynamics is expressed as [1]:
q  w  hout  hin  (1.1)
Where q and w are the amounts of heat transfer and work transfer per unit mass flow of
air, respectively, and (h out –h in ) is the resulting enthalpy change over the stage. Equation
(1.1) adopts the usual sign convention that heat into the system is positive, while work
into the system is negative. Assuming the compression processes in both stages to be
isentropic as shown in Figure 1.1.b means that they are adiabatic (q=0) and reversible.
Therefore, using an average specific heat for air at constant pressure (cp ), the
compression work per unit mass flow of air in stage 1 (w1 ) and in stage 2 (w2 ) can be
determined from Equation (1.1) as follows:
w1  h2  h1   c p T2  T1  (1.2)
w2  h4  h3   c p T4  T3  (1.3)
The total compression work (wtotal ) is then given by:
wtotal  w1  w2  c p T2  T1   T4  T3  (1.4)

Assuming perfect intercooling, i.e., T3 = T1 , Equation (1.4) can be rearranged as:

 T   T    T   T 
wtotal  c p T1 1  2   1  4   c p T1 2   2    4  (1.5)
 T1   T3    T1   T3 
Since the two compression processes are assumed to be isentropic and the specific heat
cp for air to be constant, the temperature ratios in Equation (1.5) can be converted into
pressure ratios by using the following approximate relationships:
k 1
T2  P2  k
  (1.6)
T1  P1 
k 1
T4  P4  k
  (1.7)
T3  P3 
Where k is the ratio of specific heats (k=cp /cv ; cv is the specific heat for air at constant
volume). With the assumption that there is no pressure loss in the intercooler, P3 = P2 =
Pi . Substituting from Equations (1.6) and (1.7), Equation (1.5) becomes:
 k 1 k 1

   
 c p T1 2       
 Pi
k P4
k
wtotal (1.8)
  1   pi  
P

To see how the total compression work varies with the intermediate pressure Pi , a
specific case was considered in which T1 = 300K, P1 =100 kPa, and P4 = 900 kPa. Using
Equation (1.8), the total compression work in the system was calculated for different
values of Pi and Figure 1.2 shows the result. The figure shows that the value of Pi at
which the total compression work is minimal is around 300 kPa. Increasing and
decreasing Pi from this value both increase the compression work.
310
Total compression work (kJ)
300
290
280
270
260
0 200 400 600 800 1000
Intermediate pressure (kPa)
Figure 1.2. Variation of the total compression work with the intermediate pressure
The principles of thermodynamics are particularly useful for performance evaluation
and optimisation of power-generation and refrigeration systems. For example, consider
the regenerative steam-turbine power plant shown in Figure 1.3. This plant consists of a
boiler house for producing superheated steam, a high-pressure steam turbine (HPT), a
low-pressure steam turbine (LPT), a condenser, an open feed-water heater (FWH) and
two feed-water pumps. A fraction of the steam (y) is extracted after the HPT for
preheating the feed-water before going back to the boiler house.
1 2
HPT LPT
Boiler
House y
3
1-y
7 FWH
Condenser
5
6
4
Pump 2 Pump 1
Figure 1.3. Schematic diagram of a regenerative steam-turbine power plant
Although extracted steam reduces the work output from plant, it reduces the amount of
heat added in the boiler and its net effect is to increase the thermal efficiency of the
plant. There is also a certain extraction pressure for the steam at which the plant‟s
thermal efficiency attains a maximum value. As shown below, the principles of
thermodynamics can be used to determine this optimum steam-extraction pressure.
The total specific work output from the two turbines (wout ) and the total work input to
the two pumps (win ) are given by:
wout  wHPT  wLPT (1.9)
win  wP1  wP 2 (1.10)
Where wHPT and wLPT are the specific work output from the high-pressure turbine and
the low-pressure turbine, respectively, and wP1 and wP2 are the specific work inputs in
pump 1 and pump 2, respectively. Assuming the two turbines and the two pumps to be
adiabatic and neglecting the changes in kinetic and potential energies, the work output
or input for each device can be determined from the enthalpy difference across the
device. Per each kg of steam generated in the boiler, these are given by:
wHPT  h1  h2  (1.11)

wLPT  1  y h2  h3  (1.12)
wP1  1  y h5  h4  (1.13)
wP 2  h7  h6  (1.14)
Mass and energy balance over the open feed-water heater gives:
yh2  1  y h5  1  h6 (1.15)
The net specific work output from the plant (wnet ) is then given by:
wnet  wout  win (1.16)
Similarly, the specific heat input to the boiler (q in ) is determined by the relevant
enthalpy change as follows:
qin  h1  h7  (1.17)
Finally, the thermal efficiency of the plant (η) can be calculated from:
  wnet / qin (1.18)
Both wnet and η depend on the fraction of steam extracted for regeneration (y); which in
turn depends on the extraction pressure (P2 ). Figure 1.4 shows the variation of y and η
with P2 for an ideal cycle in which P1 = 15 MPa, T1 = 600o C, and P4 = 10 kPa. The
figure shows that the cycle‟s efficiency attains a maximum value of 45.55% when P2 is
in the range of 1000 kPa.
It should be mentioned that the working fluid in the above power plant, which is water,
changes its phase from a liquid to superheated steam in the boiler, to a saturated
mixture of water and steam in the low-pressure turbine, and returns to liquid water in
the condenser. Therefore, appropriate property tables or charts are needed for
determining the thermodynamic properties of water at the different states. In general,
thermodynamic analyses use many tables and charts for various working fluids. The
principles of thermodynamics are also needed for the analyses of air-conditioning
systems and processes and for the analyses of the processes that involve combustion
and other chemical reactions. For such analyses, thermodynamics provides the basic
relationships needed to quantify the effects of fluid mixing and chemical reactions on
the properties of the working fluids and to determine the transfer of energy and
effluents to or from the system.
0.456 0.3
Fraction of extracted steam (y)

Thermal efficiency (η)
0.455 0.25
0.454 0.2
0.453 0.15
0.452 0.1
Efficiency
0.451 0.05
y
0.45 0
0 1000 2000 3000 4000
Extraction pressure (kPa)
Figure 1.4. The effect of intermediate pressure (P2 ) on the fraction of extracted steam
(y) and thermal efficiency (η) of a regenerative steam-turbine power plant
1.1.2. Fluid dynamics

In addition to pipes and ducts, fluid-transporting systems require various equipment
such as pumps, compressors, control valves, and flow-measuring devices. The
principles of fluid dynamics help us to estimate the power needed for overcoming
friction in these equipment and to determine suitable types and sizes for them. To
illustrate the application of these principles, consider the pump-pipe system shown in
Figure 1.5 that conveys a liquid between two non-pressurised tanks A and B through a
pipe of known length L, diameter D, and roughness ε. Suppose that we want to
determine the needed pump power for transporting the liquid at a certain flow rate Q.
L1 L3
B
L2
D
Q
A
Figure 1.5. Schematic diagram of a simple pump-pipe system
The required pump power ( W ) can be determined from the following “power
equation”:
W    Q  h p /  [W] (1.19)
Where γ is the specific weight of the transported liquid (N/m3 ), Q is the volume flow
rate of the liquid (m3 /s), h p is the pump head (m) needed to circulate the fluid through
the pipe from A to B, and η is the combined efficiency of the pump and the electric
motor. For a steady flow of an incompressible fluid, h p can be determined from the
following “energy equation”:
VB2  V A2
h p  h f ,total  Z B  Z A   [m] (1.20)
2g
Where h f,total is the total head loss through the system due to friction (m), ZA and ZB are
the elevations (m) at points A and B, respectively, and VA and VB are the corresponding
fluid velocities (m/s). If the two tanks are not open to the atmosphere, the energy
equation should also include a term for the pressure difference between the tanks.
The total friction head loss h f,total consists of two parts: the major friction loss (h f ),
which is the part lost in the pipe itself, and the minor friction head loss (h c), which is
the part lost in other components of the system like nozzles, elbows, valves, etc. The
major friction loss can be determined from the following Darcy-Weisbach equation [2]:
L V2
hf  f [m] (1.21)
D 2g
Where f is the dimensionless Dracy friction factor, V the fluid velocity (m/s), L the total
length of the pipe (m), and D the internal diameter of the pipe (m). The value of the
friction factor, which depends on the roughness of the pipe surface and on whether the
flow is laminar or turbulent, can be obtained from a Moody diagram [2] or calculated
from a relevant formula. For laminar flows, f can be calculated from:
f  64 / Re Re < 2300 (1.22)
Where Re is the Reynolds number defined as:
Re  VD /  (1.23)
Where ν is the kinematic viscosity of the flowing fluid (m2 /s). For a turbulent flow in
rough pipes, f can be obtained from the following Swamee-Jain formula [2]:
2
   5.74 
f  0.25 / log10   0.9  Re > 4000 (1.24)
  3.7 D Re 
For more accuracy, the friction factor for a turbulent flow can be determined by using
the following Colebrook-White formula (frequently referred to as the Colebrook
equation):
1  / D 2.51 
 2.0 log10   (1.25)
f  3.7 Re f 

The Colebrook equation is an example of the implicit equations met in thermofluid

analyses that need to be solved iteratively. For turbulent flows in smooth tubes, f can be
determined from the first Petukhov formula [2]:
f  0.790 ln Re 1.64

2
104 < Re < 106 (1.26)
Chemical engineers usually determine the pipe friction by using the following Chezy-
Manning equation instead of the Darcy-Weisbach equation:
L V2
hf  2 f [m] (1.27)
D g
Where f is the Fanning friction factor. Comparison with Equation (1.21) reveals that the
value of the Fanning friction factor is 4 times the corresponding value of the Darcy
friction factor. Civil engineers determine the friction head loss in water-transporting
pipes by using the following Hazen-Williams equation:
10.67 LQ1.852
hf  [m] (1.28)
C 1.852 D 4.8704
Where C is a coefficient that depends on the roughness of the pipe. Unlike Equations
(1.21) and (1.27), Equation (1.28) is applicable for both laminar and turbulent flows.
The minor friction losses, h c, can be determined from the following equation:
n
V2
hc  
1
K
2g
[m] (1.29)
Where n is the total number of components in the fluid system and K is a coefficient the
value of which can be found for each component in relevant tables.
Given the values of the pipe length and diameter, flow rate, fluid viscosity, and pipe
material or roughness, the equations described above can be used to determine the
required pump power. The equations can also be used to determine the maximum flow
rate of the fluid to be delivered via a pipe of a certain diameter such that the friction
loss in the system or the needed pump power does not exceed a specified limit.
Moreover, by taking into consideration the initial cost of the pump-pipe system that
increases with D, and the cost of electrical energy needed by the pump that decreases
with D, the equations can be used to determine the economic pipe diameter Dopt that
gives the lowest total owning cost for the system over its entire life-time. In general, the
equations are also applicable for analysing and optimising pipe-networks.
The principles of fluid dynamics also enable us to select the appropriate type and size of
the pump for a given pump-pipe system by matching the “pump curve” with the
“system curve”. This is achieved with the help of pump characteristic curves usually
provided by the manufacturers. In many situations a single pump or compressor may
not be adequate for the required flow rate or delivery pressure and more than one pump
or compressor have to be used. In this situation, the principles of fluid dynamics help us
to decide when to arrange the pumps/compressors in parallel or in series.
1.1.3. Heat transfer

The design practices of energy-conversion equipment that deal with the transfer of
thermal energy such as boilers, condensers, and heat exchangers are mainly based on
the principles of heat transfer. Three independent physical laws are used in heat-
transfer analyses to quantify the rate of heat transfer between an object and its
surroundings depending on whether the heat transfers by conduction (Fourier‟s Law),
convection (Newton‟s law of cooling), or radiation (Stefan-Boltzmann law). The
physical properties that determine the rate of heat transfer by conduction, radiation, and
convection are the thermal conductivity (k), the surface emissivity (ε) and absorptivity
(α), and the heat-transfer coefficient (h), respectively.
While k, ε, and α are material or surface-specific, h depends on both the fluid and the
flow. Numerous analytically-obtained relationships and empirical formulae are used for
determining h depending on whether the flow is forced or natural and whether the flow
is internal or external to the system being considered. These formulae usually give the
Nusselt number (Nu) which is related to h as follows:
k
h Nu [W/m2 .K] (1.30)
D
Where D is the pipe diameter and k is the thermal conductivity of the transported fluid.
Many analytical or empirical formulae are used for determining the Nusselt number for
forced or natural flows over single tubes, bank of tubes, plates, etc. For example, the
following Dittus-Boelter equation is used for determining Nu inside a fluid-transporting
pipe due to forced convection [3]:
Nu  0.023 Re 0.8 Pr n (1.31)

Where Re is the Reynolds number, Pr the Prandtl number, and n is a constant that takes
a value of 0.4 when the pipe is being heated and 0.3 when it is being cooled.
The subject also describes the methods that can be used to minimise or maximise the
rate of heat-transfer between the system‟s components or between the system and its
surroundings by means of thermal insulation, fins, heat-pipes, etc. To illustrate the use
of heat-transfer concepts in thermal-insulation analyses, consider the metal pipe shown
in Figure 1.6 that has an internal radius r1 and external radius r2 . The pipe carries a fluid
at a temperature Ti , while the surrounding air is at a different temperature T∞.
r3
T∞
r1
r2 δ
Ti
Figure 1.6. Schematic for an insulated metal pipe
The temperature difference between the pipe and the surroundings will cause heat gain
or heat loss to/from the pipe and, in order to reduce this heat gain or heat loss, the pipe
has to be covered by an insulating material. The principles of heat transfer help us to
account for the effect of thermal insulation on the rate of heat-transfer to/from the pipe.
The rate of heat transfer ( Q ) can be calculated from [3]:
Q  Ti  T  / Rth [W] (1.32)
Where Rth is the combined thermal resistance to heat-transfer by conduction,

convection, and radiation, which is given by [3]:
1 lnr2 / r1  lnr3 / r2  1
Rth     [K/W] (1.33)
hi A1 2Lk1 2Lk 2 ho A3
Where h i and A1 are the heat-transfer coefficient and surface area inside the pipe,
respectively, h o and A3 are the heat-transfer coefficient and surface area outside the
insulated pipe, respectively, L is the length of the pipe, and k 1 and k 2 are the thermal
conductivities of the pipe and the insulation, respectively. To simplify the analysis, h o
in Equation (1.33) takes into account the heat-transfer by both convection and radiation
to/from the insulation surface. The thickness of the metal pipe is usually small
compared to its diameter, while its thermal conductivity is much higher than that of the
insulation material. Therefore, the equation can be simplified further by neglecting the
term that represents the thermal resistance due to conduction through the pipe.
Equations (1.32) and (1.33) can be used to determine the required thickness of
insulation (δ) for reducing the rate of heat transfer to the required tolerance or for
controlling the surface temperature within a range that is dictated by safety or other
practical considerations. Although the thicker the insulation the lower will be the rate
heat transfer, the cost of insulation increases with its thickness and, therefore, adding
more insulation will not be economically profitable beyond a certain thickness. By
extending the above heat-transfer model so that the cost of insulation and the value of
the saved thermal energy can be calculated and compared, the above equations can also
be used to determine the economically optimal thickness of insulation (δopt ).
Figure 1.7 shows a metal pipe with circular fins attached to its surface so as to boost the
rate of heat-transfer between the fluid being transported with the pipe and the
surrounding medium, usually air. The principles of heat transfer can be used to develop
the required mathematical equations that determine the the rate of heat transfer from the
pipe and, therefore, to evaluate the effectiveness and efficiency of the fin. This type of
heat-transfer analyses is discussed in more details in Appendix B.
Figure 1.9. Circular fins attached to a metal pipe
Another important application of heat-transfer principles is that related to the design

and selection heat-exchangers. A heat-exchanger is any device that allows the transfer
of thermal energy between two fluids through a separating surface usually a pipe, a
duct, a tube, or a plate. Figures 1.8 and 1.9 show two types of heat-exchangers
commonly used in industries and power-plants. Figure 1.8 shows a shell-and-tube heat-
exchanger while Figure 1.9 shows a cross-flow heat-exchanger. Heat-exchanger
analyses either aim at determining the required size (i.e., surface area) for a specified
heat-transfer duty or determining the exit temperatures of the two streams from a
specified heat-exchanger type and size. As shown in Appendix B, two methods are used
for these two types of analyses which are the log-mean temperature difference (LMTD)
method and the effectiveness-number of transfer units (ε-NTU) method. Complex

thermal systems use heat-exchanger networks [HENs] and finding the configuration
that minimises the annual cost of the network is based on the principles of heat transfer.
Hot Cold
out out
Cold Hot
in in
Figure 1.8. A parallel-flow shell-and-tube exchanger
Cross-flow
(unmixed)
Tube-flow
(unmixed)
Figure 1.9. A cross-flow exchanger with both streams unmixed
1.2. Advantages of computer-aided thermofluid analyses

Apart from saving time and eliminating possible human errors, computer-aided
thermofluid of analyses offer a number of advantages over traditional methods that use
property tables and charts. An important advantage of computer-aided methods is their
ability to give more realistic results by avoiding unnecessary simplification of the
models and by using more accurate estimations of fluid properties. Moreover, they offer
reliable techniques for iterative solutions, optimisation analyses, and the analyses of
complex thermofluid systems. In what follows, these advantages are illustrated by
means of relevant examples.
A. Avoiding excessive simplification of the model

In many situations, traditional analytical methods adopt excessive simplifications of the
analytical models; which makes their results grossly deviate from the behaviour of real
systems. A good example of this situation is given by the models of internal-
combustion (IC) engines. Traditional air-standard models of IC engines, such as the

Otto cycle and the Diesel cycle, neglect heat-transfer and friction losses, treat the
combustion process as heat-addition from an external source, and use constant specific
heats for the working fluids. These assumptions enable the engine processes to be
represented by simple closed-form relations for calculating the amount of heat added to
the engine and net work from the engine [4]. However, air-standard models usually
overestimate the engine‟s output and thermal efficiency.
By comparison, computer-aided models of IC engines closely mimic the behaviour of

actual IC engines by taking into consideration the geometrical as well as the
thermodynamic characteristics of the engines. Therefore, these models can be used to
investigate the effect of important design and operation factors such the ignition or
injection timing on the engine performance or the effect of engine speed on the specific
fuel consumption. However, the formulation of these models leads to a set of ordinary
differential equations that need to be solved simultaneously by using a numerical
method such as the Newton-Raphson method [5].
B. Accurate representation of fluid properties and processes

The ideal-gas law can be used with reasonable accuracy for determining the specific
volume of a superheated vapour, but when the temperature approaches the saturation
line, the value of the specific volume determined by the ideal-gas law departs
significantly from the actual volume. More accurate estimates can be obtained by using
the following Soave-Redlich-Kwong (SRK) equation of state [1]:
RT a
P~u ~ ~ (1.34)
v  b v ( v  b)
~
Where P is the absolute pressure of the gas, v is the molar specific volume, Ru is the
universal gas constant, T is the absolute temperature of the gas, and the constants a, b
and  are fluid-dependent. Figure 1.10 shows the deviations from the tabulated values
by those obtained from the ideal-gas law and the SRK equation of state for refrigerant
R134a at 0.2 MPa. The figure shows that the error of the ideal-gas law is more that 2%
even at high temperatures and increases as the temperature approaches the saturation
value, but the accuracy of the SRK equation remained higher than 99% even close to
~
the saturation line. However, since the SRK equation is implicit in v it cannot be used
directly to determine the specific volume. A number of standard iterative procedures
(e.g. Newton-Raphson method) can be used to solve the equation, but they are more
suitable for computer-aided analyses than hand calculations.
Another important implicit equation for thermofluid analyses is the Colebrook-White

equation, Equation (1.25), that determines the friction factor (f) for turbulent pipe-
flows. Since the equation involves f on both sides and needs to be solved iteratively, the
explicit relationships such as the Swamee-Jain formula are preferred even though the
Colebrook-White equation is more accurate. Many other nonlinear equations like the
SRK equation and the Colebrook-White equation give advantage to computer-aided
thermofluid analyses by enabling more realistic and accurate estimations.
7
Error in specific volume (%)

6 Ideal-gas model
SRK model
5
4
3
2
1
0
250 300 350 400
Temperature K
Figure 1.10. Errors in the specific volume of R134a by the ideal-gas law
and the SRK equation of state
C. Dealing with iterative solutions and optimisation analyses

A good example of thermofluid analyses that require iterative solutions is found in
pipe-flow analyses. Pipe flow problems that require the friction head loss to be
determined when both the diameter and flow rate are known can be solved in a
straightforward manner by using Equation (1.21). However, in design analyses of
pump-pipe systems we may need to find the flow rate in a given pipe that gives a
specified head loss or to find a suitable pipe diameter for specified head loss, flow rate,
and pipe length. In these two cases, the friction factor f cannot be determined in
advance because it depends on the Reynolds number. Therefore, these two types of
pipe-flow problems, referred to as type-2 and type-3 problems, need to be solved by
iteration. It is much easier to carry out the iterative process to the required level of
accuracy by using a computer-aided method than by doing it manually. Other types of
thermofluid analyses that also require iterative solutions include rating analyses of heat
exchanger (discussed in Appendix B) and the determination of the adiabatic flame
temperature by first-law analysis of the combustion process.
Optimisation analyses are needed for determining the best design for a thermofluid
system such as the optimum intermediate pressure for a two-stage air-compression
system, the optimum steam-extraction pressure for a regenerative Rankine cycle, and
the economic thickness of insulation for a pipe. While simple optimisation analyses that
involve a single design parameter can be performed by means of calculus techniques
and graphic tools, optimisation analyses of complex systems that involve multiple
design variables require the use of computer-aided techniques.
D. Analyses involving complex models

The complexity of modelling certain thermofluid systems makes their analyses only
possible with the help of computer-aided methods. The model complexity can be either
due to the complexity of the physical structure of the system itself or the complexity of
its mathematical representation. An example of physically complex systems is the pipe
network shown in Figure 1.11 that consists of four pipe loops and four consumption
points and fed by two water tanks; tank A and tank B. Suppose that the flow rates from
the two supply tanks are specified together with the pipe diameters and lengths and it is
required to determine the discharges at the four consumption points.
Q1
Q2
QA
A
QB
B
Q3
Q4
Figure 1.11. A looped pipe network supplied by two tanks
Although the solution is mainly based on the two well-known principles of

conservation of mass and conservation of energy, it is difficult to solve the problem by
using manual analytical methods especially when a minimum or a maximum pressure
level is to be met at the discharge points. In this case, a computer-aided method, such as
the Hardy-Cross method, has to be used [6, 7]. The optimisation analyses of heat-
exchanger networks give another example of the models that deal with physically
complex systems [8].
Mathematically complex thermofluid models that need computer-aided numerical

methods are found in multi-dimensional fluid-flow and heat transfer analyses. This type
of analyses involves coupled and nonlinear partial differential equations that have to be
solved by using computational fluid dynamics (CFD) methods such as the finite-volume
method or the finite-difference method. Many commercial CFD applications are
available nowadays that offer great flexibility and user-friendliness.
1.3. The Excel-based modelling platform for thermofluid analyses

Microsoft Excel is commonly used for data visulation and for dealing with simple
computer-based operations like matrix inversion and matrix multiplications [3,9].
However, Excel is equipped with other features that make it a capable modelling
platform for a wide range of engineering “What-if” analyses [10-12]. In addition to its
Goal Seek command and the Solver add-in, the “Developer” ribbon in Excel provides a
programming language called Visual Basic for Applications (VBA) that can be used for
developing customised user-defined functions (UDFs) not provided by Excel. The
Developer ribbon also allows the use of macros to remove the tedium of parametric
studies and repetitive calculations. The main limitation of Excel with respect to
thermofluid analyses, which is the lack of built-in functions for fluid properties, could
be solved by developing suitable add-ins for this purpose [13-15].
This book uses an Excel-based modelling platform for thermofluid analyses that
includes in addition to Excel, Solver, and VBA, an educational Excel add-in called
Thermax. Thermax provides seven groups of property functions for ideal gases,
saturated water and superheated steam, synthetic and natural refrigerants, atmospheric
humid air for psychrometric analyses, two aqua solutions for vapour-absorption
refrigeration, chemically-reacting substances, and air at standard atmospheric pressure.
Thermax also provides two interpolation functions and a Newton-Raphson solver for
nonlinear equations that enhance the usefulness of the Excel-based modelling platform.
Table 1.1 summarises the roles of the four components of the Excel-based modelling
platform as used in this book.
Table 1.1. Roles of the four components of the Excel-based modelling platform
Component Role
 Provides the basic functions needed for thermofluid analyses
including the general mathematical functions and the matrix-
operation functions
 Provides the Goal Seek command needed for performing
Excel unconstrained iterative solutions involving a single parameter
 Allows circular calculations which can be a convenient method for
dealing with parameter dependency in certain analyses
 Provides graphical tools needed for data visualisation and analyses
 Allows macros to be recorded for repetitive calculations
 Allows constrained iterative solutions involving multiple parameters
 Allows optimisation analyses with single and multiple design
Solver
variables
 Offers three solution options that suit different types of analyses
 Provides the physical properties of various fluids
Thermax  Provides two interpolation functions for tabulated data and a Newton-
add-in Raphson solver for non-linear equations such as the Colebrook-White
equation and the SRK equation
 Needed for developing custom functions for the non-linear equations
involved in iterative solutions and optimisation analyses
VBA  Needed for developing numerical solvers for large systems of linear
equations
 Can be used to develop additional fluid property functions or other
functions not provided by Excel or the Thermax add-in
1.4. Closure
The following four chapters describe the Excel-based modelling platform in more
details and illustrate its use for optimisation analyses of fluid-thermal systems. Chapter
2 focuses on the features of Excel that are mostly needed for these analyses such as its
matrix functions and its Goal Seek command. Chapter 3 that introduces the other three
components of the modelling platform gives examples of using the three solution
methods offered by Solver, describes the development of user-defined functions with
VBA, and shows how the property functions provided by Thermax can be used in Excel
formulae. Chapter 4 that deals with optimisation analyses of fluid-thermal systems
shows how Solver can be used for such analyses that involve a single design variable
and multiple design variables.
The Excel-based platform is then used in Chapters 5 to 10 as a modelling platform for

design analyses of various types of fluid-thermal systems. Chapter 5 focuses on thermal
insulation and Chapter 6 on determining the optimum pipe diameter for a pump-pipe
system. Chapter 7 illustrates the use of the platform for design analyses of pump-pipe
systems and double-pipe heat-exchangers while Chapter 8 deals with pipe-network
optimisation. Chapters 9 and 10 consider three options for utilising the exhaust-gas
energy in gas-turbine systems which are: the air-bottoming cycle, cogeneration, and the
generation of steam for a vapour-absorption system that cools the inlet-air. Thermax
property functions for ideal gases enable the exact variable specific heat method to be
used in the thermodynamic models of the gas-turbine cycle considered in these two
chapters.
References
[1] Y.A. Cengel, and M.A. Boles, Thermodynamics an Engineering Approach,
McGraw-Hill, 7th Edition, 2007
[2] C. T. Crowe, D. F. Elger, B. C. Wiliams, and J. A. Roberson, Engineering Fluid
Mechanics, 9th edition, John Wiley & Sons, Inc., 2009.
[3] Y.A. Cengel and A.J. Ghajar, Heat and Mass Transfer: Fundamentals and
Applications. 4th edition, McGraw Hill, 2011.
[4] W.W. Pulkrabek, Engineering Fundamentals of the Internal Combustion Engine,
2nd Edition, Pearson Education International, 2004.
[5] C.R. Ferguson, Internal Combustion Engines, John Wiley & Sons, 1986.
[6] A. Rivas, T. Gómez-Acebo, and J. C. Ramos. The application of spreadsheets to
the analysis and optimization of systems and processes in the teaching of
hydraulic and thermal engineering, Computer Applications in Engineering
Education, Vol. 14, Issue 4, 2006, pp. 256-268.
[7] D. Brkic, Spreadsheet-Based Pipe Networks Analysis for Teaching and Learning
Purpose, Spreadsheets in Education (eJSiE): Vol. 9: Iss. 2, Article 4, 2016.
Available at: http://epublications.bond.edu.au/ejsie/vol9/iss2/4
[8] A. Bejan, G. Tsatsaronis, M. Moran, Thermal Design & Optimisation, John

Wiley & Sons, 1996.
[9] J. P. Holman, Heat Transfer, 10th edition, McGraw-Hill. 2010.
[10] A. Karimi, Using Excel for the thermodynamic analyses of air-standard cycles
and combustion processes, ASME 2009 Lake Buena Vista, Florida, USA.
[11] Z. Ahmadi-Brooghani, Using Spreadsheets as a Computational Tool in Teaching
Mechanical Engineering, Proceedings of the 10th WSEAS International
Conference on computers, Vouliagmeni, Athens, Greece, July 1315, 2006, 305-
310
[12] S.A. Oke, Spreadsheet Applications in Engineering Education: A Review, Int. J.
Engng Ed. Vol. 20, 2004, No. 6, 893-901
[13] The University of Alabama, Mechanical Engineering, Excel for Mechanical
Engineering project, Internet: http://www.me.ua.edu/excelinme/index.htm (Last
accessed July 11, 2019).
[14] J. Huguet, K. Woodbury, R. Taylor, Development of Excel add-in modules for
use in thermodynamics curriculum: steam and ideal gas properties, American
Society for Engineering Education, 2008, AC 2008-1751.
[15] K. Mahan, J. Huguet, K. Woodbury, R. Taylor, Excel in ME: Extending and
refining ubiquitous software tools, American Society for Engineering Education,
2009, AC 2009-2295.
Excel
With its user-interface, built-in functions, iterative tools, and graphical tools, Excel
forms the backbone of the modelling platform used in this book for thermofluid
analyses. This chapter is not intended to review all the features and built-in functions
provided by Excel, but to focus on those needed for building analytical models for
thermofluid analyses. In this respect, the chapter highlights the use of cell-labelling in
writing formulae instead of the usual referencing by location. The chapter also
illustrates the use of Excel‟s matrix functions for the solution of linear systems of
equations and the use of Goal Seek and circular calculations for the solution of
nonlinear equations. The final section on Excel‟s graphical tools demonstrates the use
of the trendline feature for curve-fitting of tabulated data.
2.1. Elements of Excel’s user-interface

Excel‟s user-interface allows us to manipulate the stored data by providing a large set
of built-in functions and a number of analytical tools. It also provides numerous
commands for adjusting the appearance of the workspace and presenting the primary
data and the results in various forms. Figure 2.1 shows a screenshot of an Excel sheet
that stores the scores obtained by a group of students in one semester.
2
3
Figure 2.1. The main elements of Excel‟s user-interface
To allow easy access to the large number of functions, tools, and commands provided
by Excel, its interface is divided into a number of elements with different purposes.
Figure 2.1 shows four of these elements which are:
1. The ribbon
2. The name box
3. The formula bar
4. The workspace
The Ribbon, which occupies the upper part of the sheet, organises the numerous
commands provided by Excel into nine “tabs”, e.g., the File, Home, and Insert tabs.
Each tab consists of a number of command-groups that have a common purpose. The
Workspace, which is the main part of the sheet, is divided into a grid of columns and
rows that form separate “cells” at their intersections. A cell is referred to by a letter and
a number, e.g., A1, B3, H2, etc. The lettert represents the cell‟s column while the
number represents its row. The Name box shows the location of the current cell.
As Figure 2.1 shows, a cell can simply contain a character data, such as “Saeed” and
“Salim”, or a numerical data, such as 62.5 and 70. A cell can also contain a formula for
data manipulation using the numerous built-in functions provided by Excel. The
formula bar in Figure 2.1 reveals the formula typed in cell H2 that uses the built-in
function “AVERAGE” to determine the average score of the first student in the list
(Saeed) in the five subjects as 64.0. Note that, unlike simple numerical or character
cells, a cell that includes a formula must include the equal sign “=”. The role of the
Formula bar will be explained in more details in the following section.
2.2. Excel’s formulae

In general, Excel‟s formulae include mathematical or logical operators, built-in
functions, and cell references. Excel provides two ways to refer a particular cell; either
by its location in the sheet, e.g., A2, C10, etc., or by giving it a relevant name, e.g.,
efficiency, diameter, etc. The two methods will be illustrated below.
2.2.1. Cell reference by location

To illustrate this method, let us write a formula that calculates the area of a circle from
its radius. To do this, open a new Excel sheet and type the number 5, which is the radius
of the circle, in cell A1 as shown in Figure 2.2.
Formula bar
Figure 2.2. Writing an Excel formula to determine the area of a circle
Now, go to cell A2 and type the following formula:
=PI()*A1^2/4
The function “PI()” is a built-in function that returns the value of Archimedes‟ constant
π. The formula also contains a reference to cell A1 that stores the value of the circle‟s
radius, the multiplication operator *, the division operator /, the power operator ^, and
the constants 2 and 4. Note that the formula is shown in the formula bar which can also
be used to edit the formula. Pressing the Enter key after typing the formula, you will
obtain the result shown in Figure 2.3; which is 19.63495 square metres.
Figure 2.3. The Excel sheet with a formula that determines the area of a circle
The following example shows how Excel‟s formulae and built-in functions can be used
in a simple thermodynamic analysis.
Example 2-1. Determining the specific volume of R134a by using the ideal-gas law
Develop an Excel sheet that calculates the specific volume (v) of refrigerant R134a
from the ideal-gas law at a pressure of 200 kPa (Tsat = -10.09o C) and temperatures in the
range 0o C to 100o C (273 to 373 K). Compare your results with the tabulated data.
Solution
Figure 2.4 shows the Excel sheet prepared for this example. The pressure (P), the gas
constant (R), and the temperature (T) are stored in columns A, B, and C, respectively.
Column D stores the values of v obtained from relevant property tables and column E
stores the corresponding values obtained from the ideal-gas law:
v  RT / P (2.1)
Where, P and T are the absolute pressure and temperature, respectively, and R is the gas
constant for R134a (R = 0.08149 kJ/kg.K). The percentage error of the ideal-gas law in
estimating the specific volume is given by:
v Ideal  vTable
Error   100 (2.2)
vTable
To determine the percentage error at 273K, go to cell F2 and type the following formula
which is equivalent to Equation (2.2):
=(E2 – D2)/D2*100
Figure 2.4.The sheet developed for determining error in the ideal-gas law for R134a
Note that the formula bar in Figure 2.4 reveals the above formula. When you press the
Enter key, the number 6.566 will appear in cell F2 as shown in the figure. To find the
percentage errors at other temperatures you can simply copy the formula in cell F2 and
paste it on cells F3 to F12. The calculated values of the errors show that the maximum
error occurs at the lowest temperature, which is 273K. The error decreases gradually as
the temperature increases.
2.2.2. Use of cell labels

The usual reference to cells by their columns and rows suits perfectly statistical
analyses in which the same formula is applied to a large body of data that is stored
column-wise or row-wise. For example, we may want to determine the average value,
maximum value, or minimum value of the tabulated data. Example 2-1 illustrated this
situation. However, thermofluid analyses usually involve numerous formulae but a
small set of data, e.g. the diameter of a pipe, the density or viscosity of a fluid, the
effectiveness of a heat exchanger, etc. For such analyses, it is more convenient to give
the cell a meaningful name or “label” that matches its content. The cell can then be
referred to by its label instead of its relative location. This method makes it easier to
recognise the quantities involved in the Excel formulae.
For the purpose of illustration, let us develop an Excel sheet to compare the density of
air before and after an isentropic compression process from an initial condition of P1 =
100 kPa and T1 =300K to a final pressure of P2 = 800 kPA. The two air densities
involved can be calculated from the ideal-gas law as follows:
1  P1 / RT1 (2.3)
 2  P2 / RT2 (2.4)
Where R is the gas constant for air (R = 0.287 kJ/kg.K). For an isentropic process, T2 is
related to T1 according to the following approximate relationship:
k 1
T2  T1  P2 / P1  k (2.5)
Where k is the ratio of specific heat at constant pressure (cp ) and at constant volume
(cv). For air, k can be taken as 1.4. Note that in Appendix A, Table A.1, k is used for the
thermal conductivity.
Figure 2.5 shows the sheet prepared for this analysis in which the respective cell labels
are typed in the column to the left of the different pressures and temperatures, while the
corresponding units are written in the column to the right of each quantity. This is also
done to the other quantities in the calculations. The sheet also shows the units of the
different properties involved for more clarification.
Figure 2.5. Excel sheet for calculating the air densities before and after compression
Placing the cursor on cell F4 makes the formula bar reveal the formula used in the
calculation of the temperature T2 according to Equation (2.5) which is:
=B4*B7^((B8-1)/B8)
The above formula can be made more understandable by using familiar labels to refer
to the different cells involved. To do that, select the cells in columns A and B as shown
in Figure 2.6, then go to Formulas and, at the Name Manager, select Create from
Selection. When the form shown in Figure 2.6 appears to you, tick the “Left column”
option. Pressing the “OK” button will make Excel create names for the different values
in the selection box according to the labels written on the left column. The cell F3 that
stores the value of P2 can also be associated with its corresponding label in cell E3.
Figure 2.6. Creating names for a selected group of cells
Now, retype the formula in cell F4 that determines T2 as follows:
=T_1*P_r^((k_-1)/k_)
The formula bar in the sheet shown in Figure 2.7 reveals the formula with the
corresponding labels instead of location references.
Figure 2.7. Formulae using cells labels instead of locations
Labelled formulae are easier to edit than those using location referencing particularly
when intricate formulas are involved. Another advantage of cell-labelling is that if you
copy a labelled formula and paste it in any other cell you will get the same answer, but
if you copy a formula that uses the usual referencing by location in another cell you will
get a different answer. To reveal or hide all the formulae in the sheet, press the control
key (ctrl) with the tilde key (~). When naming your cells, choose suitable representative
names for the variables involved, e.g. P_1 and T_1 for P1 and T1 . Note that Excel does
not accept “P1” or “T1” as labels since these can be confused with usual cell references
by locations. Therefore, Excel automatically changes these labels to “P1_” and “T1_”.
More information about Excel formulae can be found in Walkenbach [1].
2.3. Excel’s built-in functions

Excel provides a large library of built-in functions for data manipulation, like the
AVERAGE function, and other functions commonly used in engineering analyses like
the PI, SIN, and COS functions. To view the full range of Excel‟s functions, type “=”
in any Excel cell as shown in Figure 2.8 and then place the cursor on the “Insert
Function” fx button in the formula bar and click it. The dialog box shown in Figure 2.9
will appear to you. List all the categories via the “Select a category” slot.
Figure 2.8. Exploring Excel‟s built-in functions
Figure 2.9. The various categories of Excel‟s built-in functions

The Math & Trig group includes the mathematical and trigonometric functions used in
different types of engineering analyses. Figure 2.10 shows some of the numerous
functions in this group. Note that the dialog box gives a brief explanation of each
function. For example, the explanation given to the ABS function is that it returns the
absolute value of a number. The functions ACOS, ASIN, and ATAN apply the familiar
inverse trigonometric functions: cos -1 , sin-1 , and tan-1 , respectively. By scrolling down
the list, you can find many other familiar functions. The following discussion focuses
on two types of functions that are needed for the development of analytical models in
subsequent chapters of the book, which are (a) the logical functions and (b) the
functions for matrix operations.
Figure 2.10. Common mathematical functions supported by Excel
2.3.1. Logical functions

To determine the major friction loss (h f ) in a pipe by using the Darcy-Weisbach
equation we have to establish whether the flow is laminar or turbulent so as to select the
relevant formula for the friction factor (f). The flow remains laminar before the
Reynolds number (Re) reaches a certain value, which can be taken as 2,000, but the
flow can only be considered fully turbulent beyond Re = 3000. There is a transitional
region between laminar and turbulent flows when 2000 < Re < 3,000. Suppose that we
want to write an Excel formula that determines the type of flow from the given value of
the Reynolds number. Using a simple IF function, we can write the following formula:
=IF(Re<=2000, “Laminar”, “Turbulent or transitional”)

Note that the above IF-formula does not differentiate between a turbulent and a
transitional flow. Therefore, we need to use a second logical test inside the first logical
test. This can be done by using the following nested IF function:
=IF(Re<=2000, “Laminar”, IF(Re>=3000, “Turbulent”, “Transitional”))
Figure 2.11 shows an Excel sheet containing the above formula (shown in the formula
bar) and the response of the formula when Re = 500, which is “Laminar”. Depending
on the value of Re stored in cell C2, the result of the If-formula can be “laminar”,
“Turbulent”, or “Mixed”. Excel supports six other logical functions; AND, FALSE,
IFERROR, NOT, OR and TRUE that can be combined in the same formula so as to
handle more intricate choices.
Figure 2.11. A formula using the nested IF function to determine the type of flow
2.3.2. Functions for matrix operations

Adjacent cells can be treated as a matrix or a vector and a group of Excel‟s formulae
allow for the addition, subtraction, and multiplication of these matrices and vectors
according to the established rules of matrix operations. Figure 2.12 shows a 3x3 matrix
(A) stored in the cells B3:D5 and a vector (b) stored in cells F3:F5.
Figure 2.12. Step 1 of using the matrix multiplication function
Matrix (A) and vector (b) can be multiplied and the result stored in a third vector (c) by
using the matrix function MMULT. The procedure is as follows:
1. After keying in the data of matrix (A) and vector (b) as shown in Figure 2.12,
position the cursor at cell H3 and type the formula: =MMULT(B3:D5;F3:F5).
2. Now press ENTER key and cell H3 will take the value 14, which the result of
multiplying the first row of the matrix with the vector (b) as Figure 2.13 shows.
The other two elements of the result vector will not appear automatically. To view the
complete solution vector, do as follows:
1. Select the cells H3:H5 as shown in Figure 2.14,

2. Press the function key F2 once and then simultaneously hold the (SHIFT +
CONTROL) keys together and press ENTER. The complete solution vector (c)
will now appear as shown in Figure 2.15.
An important matrix-operation function provided by Excel is the matrix-inversion

function MINVERSE which is needed for the solution of linear systems of equations.
The following example illustrates the use of this function.
Example 2-2. Using the matrix inversion function

By using the MINVERSE function, find the inverse of matrix [A] given by:
1 0 3
[A] = 0 5 6
 
7 0 5
Solution
The first step is to enter the elements of the matrix as shown in Figure 2.16. After
entering the data, go to cell F2 and type the formula “=MINVERSE(B2:D4)”. When
you press ENTER, this cell will have the value -0.3125, which is the first element of
the inverse matrix [A]-1 shown in Figure 2.17.
Figure 2.16. Step 1 of using the MINVERSE function
Figure 2.17. Step 2 of using the MINVERSE function
Starting with the formula in cell F2, select the range F2 to H4 as shown in Figure 2.17.
Press and release the function key F2 and then simultaneously hold the CTRL+SHIFT
keys and press ENTER. Other elements of the inverse matrix [A] -1 will then appear as
shown in Figure 2.18. You can check the solution by multiplying matrix [A] with its
inverse by using the MMULT functions. The procedure is illustrated by Figures 2.19 to
2.21. As should be expected, Figure 2.21 shows that the resultant matrix is the identity
matrix.
Figure 2.18. The complete inverse matrix [A]-1
Figure 2.19. Multiplying matrix [A] by its inverse [A]-1
Figure 2.20. The first element of the product (identity) matrix
Figure 2.21. The complete solution which is the identity matrix

2.4. Solution of linear system of equations

An example of thermofluid analyses that involve systems of linear equations is the
solution of the heat conduction equation with the finite-difference method. Linear
systems of equations can be solved with Excel by applying the matrix-inversion
method. For illustration, consider the following linear system written in matrix notation:
Ax  y (2.6)
Where [A] is the coefficient matrix, {x} the vector of unknowns, and {y} the right-side
or “load” vector. By applying the matrix-inversion method, the solution vector {x} can
be obtained as follows:
x  A1 y (2.7)
Where [A]-1 is the inverse of matrix [A]. The following example illustrates the
procedure of applying the method by using Excel‟s matrix functions.
Example 2-3. Solution of a system of linear equations

Find the values of xi in the following system of linear equations:
 14 14  9 3  5   x1    15 
 14 52  15 2  32  x   100
   2   
 9  15 36  5 16   x3   = 106 
   x   329 
3 2  5 47 49 
  5  32 16 49 79   4  
 x 5   463 
(2.8)
Solution
Note that the system is symmetric; which is typically the case with linear systems that
arise in the solution of heat-conduction problems by the finite-difference method. For
larger systems of equations, the symmetry of the system can be utilised for reducing the
required computer memory by storing only one half of the coefficient matrix. However,
this requires a complicated computer programming. For small systems like the one
considered here, it is more convenient to use Excel‟s matrix inversion and
multiplication functions. Figure 2.22 shows the Excel sheet that stores both the
coefficient matrix [A] and the load vector {y}. The inverse matrix [A]-1 , which is
obtained by following the procedure described in the previous section, is stored below
the coefficient matrix as shown in the figure. The inverse matrix [A] -1 is then multiplied
with the load vector {y} and the result stored below the load vector as shown in Figure
2.23. The complete solution is shown in Figure 2.24. The first element is practically
zero and, therefore, the solution vector is{x} = (0, 1, 2, 3, 4).
Figure 2.22.The coefficient matrix [A], the load vector {y}, and the inverse matrix [A] -1
Figure 2.23. Multiplying the inverse matrix [A]-1 with the load vector {y}
Figure 2.24. The complete solution vector {x}

It should be mentioned that the procedure described above by using Excel‟s functions
suits best one-dimensional fluid-flow and heat-transfer analyses. This is because the
linear systems generated in multi-dimensional analyses are usually too large to be
solved efficiently by using the matrix-inversion method.
2.5. Itearative solutions with Excel

Excel offers its user two methods to perform iterative solutions: (i) by using the Goal
Seek command and (ii) by using circular calculations. In what follows the two
methods will be illustrated with the help of simple examples.
2.5.1. Itearative solutions with Goal Seek

The Goal Seek command is used for finding the value of an independent variable (x)
that yields a specified value of a dependent variable (y). It is a simple, yet very useful
tool for “What-if” analyses. The following example illustrates how this command can
be used to solve a nonlinear equation.
Example 2-4. Solution of a nonlinear equation by Goal Seek

A centrifugal pump is used for lifting
water from the utility network at the
ground level to a tank at the top of a
building that is 30-m high as shown in
Figure 2.25. The pump‟s characteristic
curve can be represented by the
following formula:
h p  h0  aQ  bQ 2  cQ 3 30 m
(2.9)
Where h p and Q are the pump‟s head (m)

and discharge (m3 /s), respectively, and
h 0 , a, b, and c are constants the values of
which are 47.22, 2.985x103 , 1.549x105 ,
and 2.348x108 , respectively.
Pump
Neglecting friction losses in the pipe,
determine the water flow rate (m3 /s) that
Figure 2.25. Schematic for Example 2-4
can be delivered by the pump.
Solution
Figure 2.25 shows the Excel sheet prepared for this example in which the values of the
four constants in Equation (2.9) are stored at the top of the sheet. The pump‟s discharge
is calculated at various values of the discharge and plotted as shown in the figure. We
can see from the plot that the value of Q that yield h p = 30 m is approximately 0.003
m3 /s.
Figure 2.25. Excel sheet for Example 2-4
To solve the problem by using Goal Seek, enter an initial guess for Q in cell B7, say 0,
and then enter the following formula that uses Equation (2.9) to calculate h p in cell B9:
=$B$2-$B$3*B7-$B$4*B7^2-$B$5*B7^3
Note the dollar sign ($) that has been added to the references of the four constants, e.g.
B2 has become $B$2. The formula bar in Figure 2.25 reveals the above formula by
placing the cursor at cell B9.
To activate the Goal Seek command, go to the Data tab, select the What-If-Analysis
option in the Data Tools group and then select Goal Seek, as shown in Figure 2.26.
The Goal Seek dialog box shown in Figure 2.27.a will then appear to you. The dialog
box asks you to select the “Set cell”, i.e. the cell that contains the dependent variable,
which is B9 in this case. You also have to specify the value sought for this cell and the
adjustable cell that stores the parameter to be changed. In this case, we seek the value in
the cell B9 to be 30 by changing the value of the cell B7. The completed form is shown
in Figure 2.27.b.
Figure 2.26. Activation of the Goal Seek command

(a) (b)
Figure 2.27. Goal Seek Set-up for Example 2-4: (a) before completion (b) the
completed box
By pressing the “OK” button after completing the Goal Seek form, Excel will change
the value in the adjustable cell (B7) until the Set cell (B9) acquires the required value.
As shown in Figure 2.28, the answer obtained is Q = 0.003 m3 /s which agrees with the
estimated value from the plot in Figure 2.25.
Figure 2.28. Goal Seek solution for Example 2-4
2.5.2. Iterative solution with circular calculations

A circular reference occurs when an Excel formula tries to refer to its own cell in a
direct or indirect manner. In standard programming this is not allowed, but Excel gives
its user the option to use “circular calculation” if intended. In this case, Excel will
itearate until all the formulae involved are satisfied. The following example illustrates
this special feature which is useful for thermofluid analyses.
Example 2-5. Determining the final temperature of heated air

Heat is added to a piston-cylinder device that contains one kg of air initially at 300K. If
100 kJ of heat is added to the air at constant pressure, determine the final temperature of
~
air taking into consideration that its molar specific-heat ( c p ) varies with temperature
according to the following formula:
c~p  a  bT  cT 2  dT 3 [kJ/kmol] (2.10)
Where a = 28.11, b =1.97x10-03 , c = 4.80x10-06 , and d = -1.97x10-09 .
Solution
From the defenition of specific heat, the final temperature (T2 ) is given by:

T2  T1  Q / c~p / M  (2.11)
Where T1 is the initial temperature, Q is the amount of heat added, and M is the molar
~
mass for air (M=29). If the variation of c p with temperature is ignored and its value at
T1 alone is used, Equation (2.11) determines T2 as 399.73K. However, we can be more
~
accurate by using Equation (2.10) to determine c p at the average temperature, Tavr =
(T1 +T2 )/2. Figure 2.29 shows the Excel sheet developed for this method which reveals
the formulae inserted in cells F2, F4, and F6.
Figure 2.29. Excel sheet developed for Example 2-5
As soon as we type Equation (2.11) in cell F6, Excel will make the warning message
that there is a circulare refernce as shown in Figure 2.30.
Figure 2.30. The circular-reference prompt

~
The circular reference occurs because T2 depends on c p according to Equation (2.11)
~
while c p itself depends on T2 according to Equation (2.10). If we press the “OK” button
shown in Figure 2.30, the cells involved in the circular reference whill be identified as
shown in Figure 2.31. In this case, three cells are involved in the circular reference,
which are F2, F4, and F6.
Figure 2.31. The cells involved in the circular reference
~
Excel can iterate to determine the values of both T2 and c p that satisfy the relevant
equations if permitted because the iterative-calculation option is not allowed by default.
To allow it, go to File and select Options. The Backstage View form shown in Figure
2.32 will appear to you. Select Formulas, then the form will appear as shown in Figure
2.33. Enable iterative calculations by ticking (√) the box indicated in the figure and
press the “OK” button. Excel can now iterate to find the values of T2 and cp that
simultaneously satisfy Equations (2.10) and (2.11). Figure 2.34 shows the solution
found by this method, which is T2 =398.976K.
Figure 2.32. Selecting Excel's option (Formulas)

Figure 2.33. Enabling iterative calculations from Excel's Formulas option
Figure 2.34. Solution of Example 2-5 by circular calculations
This example can also be solved by using the Goal Seek command. In this case, we
have to start the iterative solution by providing Excel with a guessed value for T2 , call it
T2o , based on which a new value for T2 is calculated, called it T2c. Since the guessed
value T2c is unlikely to be correct, it will be different from T2o . Goal Seek is then used to
adjust the value of T2o until the difference (Diff = T2c - T2o ) vanishes.
Most iterative solutions in this book are obtained by using the Goal Seek command.
However, the circular-calculation option is more useful in certain situations as
demonstrated in Chapters 8 and 9 that deal with the numerical solution of the heat-
conduction equation with the finite-difference method.
2.6. Excel’s graphical tools for data presentation and analysis

Excel has numerous graphical tools that can be used to present the stored data in a
variety of charts. Figure 2.35 shows one type of Excel charts that displays the annual
variation of temperature and relative humidity at one location in a certain day. The
figure shows a line chart in which the temperature is scaled on the primary y-axis (on
the left) while the humidity is scaled on the secondary y-axis (on the right). This
arrangement is useful for displaying two or more types of data that differ significantly
in magnitude such as the net specific work and thermal efficiency of a power cycle.
45 60
40
50
35
30 40
Temperature (oC)
Humidity (%)
25
30
20
15 20
10 temperature
10
5 Humidity
0 0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Month
Figure 2.35. An example of line charts
Excel supports other types of charts that allow the user to select the most appropriate
way to display his/her data in the form of bar, area, or scatter charts. For more
information about the different types of Excel‟s charts, the reader can refer to
specialised references such as Walkenbach [2]. A number of tutorials and videos that
show how to create different types of charts can also be found in the internet.
Excel‟s charts provide a curve-fitting capability of numerical data by using the

Trendline feature. This capability is particularly useful for computer-aided thermofluid
analyses because it can be used to convert tabulated fluid-properties and other data into
analytical equations that make the data more suitable for iterative solutions and
optimisation analyses. To illustrate the use of this feature, consider Table 2.1 that shows
properties of saturated water in the range 0.001o C – 60o C. These values of the saturation
pressure (Psat ) and saturated liquid enthalpy (h f ) are used in psychrometric analyses of
air-conditioning applications. For computer-aided analyses, it is useful to convert these
data into relevant equations.
The trendline feature provides a number of options, which include exponential, linear,
logarithmic, polynomial, and power equations as shown in Figure 2.36. To fit a
trendline to the data, we have to create line charts for the two properties as shown in
Figures 2.37.a and 2.39.b. Trendlines can then be added on the line charts. Figures
2.37.a and 2.39.b also show the corresponding trendline equations of the tabulated data
as determined by using polynomial equations. As Figure 2.37.b shows, a linear equation
is adequate for the h f data since its variation over the given temperature range is mild.
However, a third-order polynomial is required to represent the variation of Psat with
temperature as shown in Figure 2.37.a.
Table 2.1. Properties of saturated water at temperatures in the range 0o C- 60o C taken
from Cengel and Boles [3]
ToC Psat vf vg uf ug hf hg sf sg
[kPa] [m3/kg] [m3/kg] [kJ/kg] [kJ/kg] [kJ/kg] [kJ/kg] [kJ/kg.K] [kJ/kg.K]
0.01 0.6117 0.001000 206.00 0.000 2374.9 0.001 2500.9 0.0000 9.1556
5 0.8725 0.001000 147.03 21.019 2381.8 21.020 2510.1 0.0763 9.0249
10 1.2281 0.001000 106.32 42.020 2388.7 42.022 2519.2 0.1511 8.8999
15 1.7057 0.001001 77.885 62.980 2395.5 62.982 2528.3 0.2245 8.7803
20 2.3392 0.001002 57.762 83.913 2402.3 83.915 2537.4 0.2965 8.6661
25 3.1698 0.001003 43.340 104.83 2409.1 104.83 2546.5 0.3672 8.5567
30 4.2469 0.001004 32.879 125.73 2415.9 125.74 2555.6 0.4368 8.4520
35 5.6291 0.001006 25.205 146.63 2422.7 146.64 2564.6 0.5051 8.3517
40 7.3851 0.001008 19.515 167.53 2429.4 167.53 2573.5 0.5724 8.2556
45 9.5953 0.001010 15.251 188.43 2436.1 188.44 2582.4 0.6386 8.1633
50 12.352 0.001012 12.026 209.33 2442.7 209.34 2591.3 0.7038 8.0748
55 15.763 0.001015 9.5639 230.24 2449.3 230.26 2600.1 0.7680 7.9898
60 19.947 0.001017 7.6670 251.16 2455.9 251.18 2608.8 0.8313 7.9082
Figure 2.37. The Format Trendline window

Saturation pressure (kPa) 25 300

y = 9E-05x3 - 0.0012x 2 + 0.0788x
250
Enthalpy (kJ/kg)
20 + 0.5418
y = 4.184x + 0.146
200
15
150
10
100
5
50
0 0
0 25 50 75 0 25 50 75
Temperature (oC) Temperature (oC)
(a) (b)
Figure 2.37. Fitting trendlines on tabulated data for water of
(a) saturation pressure and (b) saturated liquid enthalpy
2.7. Closure
This chapter described the main features of Excel needed for thermofluid analyses. The
chapter highlighted the importance of using cell labelling with Excel‟s formulae and
illustrated the use of Excel‟s general mathematical functions, logical functions, and the
functions for matrix operations. The chapter also demonstrated the use of Excel‟s two
iterative tools: Goal Seek and circular calculations. In spite of its simplicity, the Goal
Seek command is very useful for iterative thermofluid analyses as shown in later
chapters of this book. Finally, the chapter illustrated the usefulness of Excel‟s charting
tools for the presentation of tabulated data particularly the trendline feature.
It should be mentioned that the Developer tab in Excel‟s user-interface provides a

number of other useful features that have not been mentioned in the chapter but needed
for enhancing the effectiveness of Excel as modelling platform for thermofluid
analyses. One of these features is the ability to record macros for conducting repetitive
calculations and parametric analyses.
References
[1] J. Walkenbach, Excel 2010 Formulas, Wiley Publishing Inc., 2010
[2] J. Walkenbach, Excel 2007 Charts, Wiley Publishing Inc., 2007
[3] Y. A. Cengel and M. A. Boles. Thermodynamics an Engineering Approach,
[4] S. C. Chapra and R. P. Canale, Numerical Methods for Engineers, 6th Edition,
McGraw Hill, 2010
Exercises
1. The following table shows measured values of the temperature by two different
methods compared to the correct corresponding values. Find the average error for
each method.
Correct T (o C) Method 1 Method 2

0 0.1044 0.1112
10 10.1092 10.1153
20 20.1139 20.1194
30 30.1186 30.1235
40 40.1231 40.1275
50 50.1276 50.1316
60 60.1320 60.1357
70 70.1364 70.1397
80 80.1407 80.1438
90 90.1450 90.1479
100 100.1493 100.1520
2. The following table shows the data for the saturation pressure of a certain fluid. Use
a nested IF statement to develop an interpolation formula that determines the
saturation pressure of the fluid at any temperature in the range 5o C ≤ T ≤ 30o C.
T(o C) Psat (kPa)

5 0.872
10 1.228
15 1.705
20 2.339
25 3.169
30 4.246
3. A system of algebraic equations can be expressed in matrix form as follows:
70 1 0  a  636 
60  1 1  b   518 
    
  
40 0  1  c  307 
Solve the system of equations to determine the values of the three unknowns a, b,
and c. This exercise is based on Example 9.11 in Chapra and Canale [4]. The
answer is: a = 8.5941, b=34.4118, and c = 36.7647.
4. Figure 2.P4 shows a triangular fin attached to the surface of a wall. Solution of the
conduction heat transfer equation with the finite-difference method resulted in the
following system of linear equations the solution of which gives the temperatures
in o C at different distances from the fin base as shown in the figure:
0 1 2 3 4 5
Figure 2.P4. Trianangular fin
-8.008 T1 + 3.5 T2 = -900.209
3.5 T1 -6.008 T2 + 2.5 T3 = -0.209
a. T2 -4.008 T3 + 1.5 T4 = -0.209
1.5 T3 -2.008 T4 + 0.5 T5 = -0.209
T4 - 1.008 T5 = -0.209
Use Excel functions to solve the above system of linear equations.

5. Adopting suitable names in your formulae, prepare an Excel sheet for calculating
the frictional loss (h f ) in a circular pipe of diameter D, length L, and roughness k s .
Use your sheet to determine h f in the following cases:
(a) D = 25 cm, L = 150 m, V = 2 m/s, k s = 0.045 mm, carrying water at 20o C.

(b) D = 25 cm, L = 150 m, V = 0.2 m/s, k s = 0.045 mm, carrying oil at 20o C.
(c) D = 25 cm, L = 150 m, V = 7 m/s, k s = 0.045 mm, carrying air at 20o C.
Use the Dracy-Weisbach equations and determine the values of the kinematic
viscosity from relevant property tables.
6. Using a line chart, plot the variation of sine θ for -180 ≤ θ ≤ 180 in steps of 10o then
add cosine θ on the same chart.
7. Using the data shown in Table 2.1, make a line chart for vf and vg . Add polynomial
trendlines for both and comment on the trendlines equations.
8. The table below shows some of the thermo-physical properties of air at
atmospheric pressure and different temperatures. Use Excel charts to show the
variation of the properties ρ, β, cp , k, α, μ, ν, and Pr with temperature and use

trendline to obtain suitable equations for these properties.

  103 c p   106   106
T k
(K) (kg/m3 ) (J/kg.K) (W/m.K) (m2 /s) 2 Pr
(1/K) (N S/m ) (m2 /s)
273 1.252 3.66 1011 0.0237 19.2 17.456 13.9 0.71
293 1.164 3.41 1012 0.0251 22.0 18.240 15.7 0.71
313 1.092 3.19 1014 0.0265 24.8 19.123 17.6 0.71
333 1.025 3.00 1017 0.0279 27.6 19.907 19.4 0.71
353 0.968 2.83 1019 0.0293 30.6 20.790 21.5 0.71
373 0.916 2.68 1022 0.0307 33.6 21.673 23.6 0.71
473 0.723 2.11 1035 0.0370 49.7 25.693 35.5 0.71
573 0.596 1.75 1047 0.0429 68.9 29.322 49.2 0.71
673 0.508 1.49 1059 0.0485 89.4 32.754 64.6 0.72
773 0.442 1.29 1076 0.0540 113.2 35.794 81.0 0.72
9. The volume V of liquid in a spherical tank of radius r is related to the depth h of the
liquid by:
V = πh 2 (3r −h)/3
Using the Goal Seek command, determine the value of h for the tank with r=1 m
and V = 0.5 m3 .
This exercise is based on Problem 8.9 in Chapra and Canale [4]. Answer: h = 0.431
m.
Solver, VBA, and Thermax

As the main component of the present modelling platform for thermofluid analyses,
Excel provides the user-interface and built-in functions, two iterative tools, and graphic
tools. However, what make Excel an effective platform for thermofluid analyses are the
three auxiliary components; Solver, VBA, and Thermax. This chapter focuses on these
three components and illustrates their use by means of relevant examples. With respect
to Solver, the chapter shows how its three solution methods can be used for performing
optimisation analyses and solving systems of linear and nonlinear equations. The
chapter also shows how VBA can be used for developing custom functions and how
Thermax functions can be used in Excel formulae.
3.1. Solver
Solver is a multi-purpose iterative tool developed by Frontline Systems [1] as an Excel
add-in for “What-if” analyses. Compared to Excel‟s own iterative tool, which is the
Goal Seek command described in Chapter 2, Solver offers the following advantages:
1. While Goal-Seek can only be used for simple problems that involve only one
decision variable, Solver can deal with more difficult problems in which the
objective cell is affected by numerous decision variables.
2. Goal Seek allows only a required value of the objective cell to be achieved, but
Solver also enables Excel to perform an optimisation analysis by finding the
maximum or minimum value for the formula in the objective cell.
3. Solver allows constraints to be applied on the solution, which is not possible
with Goal Seek. The ability to impose constraints is needed for both
optimisation analyses and iterative solutions.
4. Solver offers three solution methods that suit different types of problems and
gives the user a number of numerical options for applying the same method.
3.1.1. Activation of Solver

Like the Goal Seek command, Solver is found in the Data tab as shown in Figure 3.1.
If it doesn‟t appear in your Data tab, then go to File and then click Options and select
Add-Ins. From the Manage option at the bottom of the menu select Excel Add-ins and
then klick “Go”. The Add-Ins dialog box shown in Figure 3.2 will be shown. To add
Solver to the add-ins menu, tick (√) on the “Solver” option and return to the Data tab.
When you click the Solver button from the Data tab, the Solver Parameters dialog
box shown in Figure 3.3 will be shown.
Figure 3.1. The Solver add-in in the Data tab

Figure 3.2. Activating Solver from the menu of Excel add-ins
Figure 3.3. Solver Parameters dialog box

Solver Parameters dialog box helps the user to select a formula in one cell, called the
objective cell, and a group of other cells, called decision variables or variable cells,
which are directly or indirectly related to the formula in the objective cell. By adjusting
the decision variables, the objective cell can be maximised, minimised, or made to
acquire a certain value. As shown in Figure 3.3, constraints can be applied on the
decision variables. To suit different types of problems, Solver offers three solution
methods which are:
1. The GRG Nonlinear method,

2. The Evolutionary method,
3. The Simplex LP method.
The GRG Nonlinear method involves the determination of the function‟s gradient like
the Steepest Descent method [2, 3]. The Evolutionary method adopts a variety of
genetic algorithms and local search methods [4, 5]. While both the GRG Nonlinear
and the Evolutionary methods are suitable for non-linear problems, the Simplex LP
method, which is a linear-programming method, is suitable for linear problems. As
shown in Figure 3.3, Solver uses the GRG Nonlinear method by default. The
following sections give examples of using the three solution methods.
3.1.2. The GRG Nonlinear method

In an optimisation analysis we may require the objective function to be maximised or
minimised depending on the situation at hand. For example, the optimisation of pipe
insulation requires its total cost to be minimised, while the optimisation of a thermal
power plant requires its thermal efficiency to be maximised. The following example
illustrates the use of the GRG Nonlinear method in optimisation analyses.
Example 3-1. Finding the minimum value of a quadratic function

Find the minimum value of the following quadratic function in the specified range.
f (x) =x2 −2x – 1; −2 ≤ x ≤3 (3.1)
Solution
Figure 3.4 shows the Excel sheet developed for this example. The line chart inserted in
the figure shows the variation of f with x from which we can see that the minimum
value of f is -2 and occurs at x =1. An initial value for x is entered in cell B3 based on
which the function f(x) is calculated in cell B6 according to Equation (3.1). Note the
curser is placed on cell B6 to reveal the following formula typed in the cel:
= B3^2-2*B3−1
We can now use Solver to determine the minimum value of the function. To do so,
select Solver from the Data tab and fill its parameters dialog-box as shown in Figure
3.5 that shows the top part of the completed box.
Figure 3.4. Excel sheet for determining the local minimum of the quadratic function
Figure 3.5. The completed Solver dialog box for Example 3-1
The dialog box in Figure 3.5 has been filled as follows:
Set Objective: B6 and Min have been selected for this option since
want the value of the function in cell B6 is to be
minimised
By Changing Variable Cells: B3 which is the cell that stores the value of the
independent variable x
Subject to the Constraints: Two constraints have been added that specify the
minimum and maximum values of x, e.g. x ≥ -2 and
x≤3
Select a Solving Method: The GRG Nonlinear method (the default option)
Pressing the “Solve” button of the completed parameters box sparks Solver to find the
required solution. As shown Figure 3.6, shows, the answer found by Solver, which is x
= 1, f = -2, agrees with the graphical solution shown in Figure 3.4.
Figure 3.6. Solver solution for Example 3-1
3.1.3. The Evolutionary method

When the function to be optimised has more than one point of inflection, the solution
found by the GRG Nonlinear method may only be a local minimum or maximum. The
following example illustrates the capability of the Evolutionary method to find the
global optimum solution for a simple function.
Example 3-2. Finding the global minimum of a function

Determine the global minimum value for the following function:
f x   x cos(x) 3 ≤ x ≤ 14 (3.2)
Solution
Figure 3.7 shows the Excel sheet developed for solving this example. The insert shows
that the function has two minima in the specified range of x; one at x ≈ 5 and another at
x ≈ 11. Let us first try to solve the problem with the GRG Nonlinear method. An initial
value for x has been entered in cell B2 based on which the function f is calculated in
cell B4. The formula bar in Figure 3.7 reveals the formula entered in the cell B4. Figure
3.8 shows the completed Solver parameters dialog-box with two constraints that specify
the upper and lower limits for x. From Figure 3.9 that shows the solution found by
Solver it is clear that it found the local minimum (y = − 4.81) which is nearer to the
initially specified value and not the global minimum.
In order to locate the global minimum by the GRG Nonlinear method, the solution has
to be started with an initial guess that is closer to the global minimum, e.g., x = 9. The
advantage of the Evolutionary method is that such an arrangement is not required. The
set-up shown in Figure 3.8 only needs the solution method to be changed to
“Evolutionary”. Figure 3.10 that shows the solution obtained by this method confirms
that the method produced the global minimum (y = − 11.041).
Figure 3.7. The Excel sheet for Example 3-2
Figure 3.8. Solver set-up for Example 3-2 with GRG Nonlinear method
Figure 3.9. Solver solution for Example 3-2 with the GRG Nonlinear method
Figure 3.10. Solver solution for Example 3-2 with the Evolutionary method
The Evolutionary method is useful for optimisation analyses that involve non-smooth
and discontinuous functions, which are difficult to solve with the GRG Nonlinear
method. However, its disadvantage is that it takes long computer times. While the GRG
Nonlinear method took less than a second to solve the above problem, the
Evolutionary method took one minute and 35 seconds.
3.1.4. The Simplex LP method

This option provides a method for solving small systems of linear equations that can be
used instead of the method described in the Chapter 2 by using Excel‟s matrix
functions. The method will be illustrated by reconsidering the problem of Example 2-3.
Figure 3.11 shows a new Excel sheet for solving the problem with the present method.
Figure 3.11. Excel sheet for solving Example 2-3 with Solver
The top part of the sheet stores the coefficient matrix [A] and the right-hand vector {y}
of the system of linear equations to be solved. The procedure starts with a guessed
solution which is stored as vector {x0} in cells F9:F13. All the elements of this vector
are given a value of 1 as shown in Figure 3.11. The coefficient matric [A] is then
multiplied by the guessed vector {x0} using Excel‟s ”MMULT” function and the result
stored in cells H9:H13. If this initial guess is the correct answer, the multiplication
[A]{x0} will be the same as the true right-hand side vector, i.e,:
[A]{x0} = {y} (3.3)
However, Figure 3.11 shows that the vector [A]{x0} is different form the true right-
hand side vector {y} stored in cells H2:H6. Solver can now be used to adjust the
variable cells D9:D13 so that all elements of the vector [A]{x0} become equal to their
counterparts in vector {y}, i.e:
H9 = H2
H10 = H3
H11 = H4
H12 = H5
H13 = H6
Solver set-up for this task is shown in Figure 3.12.
Figure 3.12. Solver set-up for Example 2-3 with the Simplex LP method
Note that the objective cell is left blank and the Simplex LP method is selected as the
solution option. In this case, Solver will iterate to find the values of the decision
variables that satisfy all the imposed constraints. The solution found by Solver using the
above set-up is shown in Figure 3.13.
Figure 3.13. Solution of Example 2-3 with the Simplex LP method
All the elements of the [A]{x0} are now equal to their corresponding elements in the
vector {y}. The first element of the solution vector, which is -6.6x10-16 , is practically
zero. Therefore, the solution is [0,1,2,3,4], which is the same as that obtained in
Example 2-3 by using the matrix-inversion method. The advantage of Solver compared
to the matrix-inversion method is that it can be used for solving systems of nonlinear
equations by following a similar procedure (Refer to Problem 3.5 in the Exercises).
3.1.5. The default settings of Solver options

Solver gives it user the chance to improve the performance of its three solution methods
by allowing alternative options for applying these methods. While some options are
common to all three solution methods, others are particular to the GRG Nonlinear or
the Evolutionary method. By clicking the “Options” button in Solver‟s parameters
dialog box shown in Figure 3.14, the dialog box shown in Figure 3.15 will be shown.
Figure 3.14. Solver options in the Properties dialog box

Figure 3.15. Default Solver options adopted in the analyses for all solution methods
Figure 3.15 shows the default settings of the options that are common to all three
solution methods. Some of these default settings may have to be changed in order to
reduce the computation time or increase the precision of the solution in certain
situations. Sometimes, Solver fails altogether to find the solution if the default options
are used. For example, certain analyses require the automatic-scaling (AS) option to be
used while others require it not to be used. AS enables Solver to handle a poorly-scaled
model, i.e. a model in which the values of the objective and constraint functions differ
by several orders of magnitude. By using AS, the values of the objective and constraint
functions are scaled internally in order to minimise the differences between them.
Figures 3.16.a and 3.16.b show the default settings which are particular to the GRG
Nonlinear method and to the Evolutionary method, respectively. The GRG Nonlinear
method uses by default the forward difference (FD) approximation of derivatives. Most
of the analyses presented in later chapters of the book used the GRG Nonlinear
method for which the default FD approximation is used. The effect of Solver options on
Solver‟s results is analysed in Chapter 12 and the results show that FD is preferred over
CD because it saves computer time and does not cause numerical problems.
(a) (b)
Figure 3.16. The default Solver options specific to: (a) the GRG Nonlinear method and
(b) the Evolutionary method
Figure 3.16.b shows the default settings used by the Evolutionary method. Note that
this method has more adjustable parameters than the GRG Nonlinear method.
According to the default set-up, the population size is 100 and the maximum allowable
time without improvement is 30 seconds. The few cases in this book solved with the
Evolutionary method show that, with this set-up, the method needs long computer
times. Although the time required by the Evolutionary method can be reduced by
reducing the population size, number of iterations, or the maximum allowable time, the
optimsation analyses presented in later chapters to do not show a clear advantage to this
method over the GRG Nonlinear method which is easier to apply.
3.2. VBA and the development of user-defined functions

Although Solver and the Goal Seek command enable Excel to deal with a wide range of
thermofluid analyses, there are situations where the analytical model requires the
development of a customised user-defined function (UDF) that is not provided by
Excel. This arises, for example, in thermodynamic analyses that require functions that
determine the properties of fluids at various pressures and/or temperatures. This section
illustrates the process of activating VBA and using it to develop simple UDFs.
3.2.1. Activation of VBA

As shown in Figure 3.17, VBA is found on the left side of the Developer tab. This tab
gives many other development tools. If the Developer tab is not shown in the ribbon of
your Excel sheet, then go to File, select Options, and then select to Customise Ribbon
from the Backstage View shown in Figure 3.18.
Figure 3.17. Selection of VBA from the Developer tab
Figure 3.18. Adding VBA to the Developer tab

From the Main Tabs, select the Developer check box and then click “OK”. The
Developer tab will now be shown in the ribbon of your Excel sheet. To start writing
the UDF, go to Developer tab menu and select Visual Basic. The Visual Basic editor
will appear to you as shown in Figure 3.19. Select Insert → Module and the blank
page shown in Figure 3.20 will be open for you to type the VBA code.
Figure 3.19. Inserting a new module
Figure 3.20. A new VBA module

3.2.2. Development of UDFs

As a first example let us write a VBA function for determining the area (A) of a circle
given its diameter (D) using the following mathematical equation:
A  D 2 / 4 (3.4)
The following UDF determines the circle‟s area according to Equation (3.4):
Function Circ_area(Dia)
Pi = 3.141593
Circ_area = Pi * Dia ^2 / 4
End Function
Note that the first line in the code starts with the word “Function” followed by the
name given to the function, which is “Circ_area”. The required input parameters are
specified between two brackets after the function name. The present function has only
one input parameter, which is the diameter (Dia). As soon as you type the first line of
the code and press the “Enter” key, the editor will automatically add the End line of
the function. Now, type the rest of the code as shown in Figure 3.21.
Figure 3.21. A UDF for caclcuating the area of a circle with a given diameter
After typing the code correctly, the function can be used via Excel UI just like any
built-in function as shown in Figure 3.22. Note that the formula bar in Figure 3.22
reveals the formula in cell B2 as:
= Circ_area(10)
Where the number 10 refers to the diameter of the circle, which is the only input to the
UDF. You can now check the answer of your user-defined function by calculating the
circle‟s area with a normal Excel formula by typing in any cell “=pi()*10^2/4”.
Figure 3.22. Using the “Circ_area” function in Excel
For thermodynamic analyses, VBA is useful for developing UDFs for fluid properties.
As an example, let us develop a UDF that determines the molar specific-heat at constant
pressure ( c~p ) for air. For an ideal gas, c~p is given by the following polynomial [6]:
c~p  a0  a1T  a 2T 2  a3T 3 [kJ/kmol.K] (3.5)
Where T is the absolute temperature and a 0 , a 1 , a 2 , and a 3 are constants that have
different values for different gases. For air, the constants‟ values are 28.11, 0.1967x10-2 ,
0.4802 x10-5 , and -1.966 x10-9 in this order.
The following VBA function, called “cp_air”, determines c~p for air based on Equation
(3.5). The only input for the function is the absolute temperature (TempK).
Function cp_air(TempK)
a0 = 28.11
a1 = 0.00196
a2 = 0.000004802
a3 = -0.000000001966
M = 28.97
cpbar = a0 + a1 * TempK + a2 * TempK ^ 2 + a3 * TempK ^ 3
cp_air = cpbar / M
End Function
Figure 3.23 shows the VBA function and the formula bar in Figure 3.24 shows how the
~
function can be used in an Excel formula to determine c p for air at 300K. The value
returned by the function is 29.0771 kJ/kmol.K. By making suitable extensions, the
~
function can easily be used to determine the values of c p for other ideal gases in
addition to air (Refer to Exercise 3.8). The various functions provided by Thermax for
the thermo-physical properties of water, refrigerants, etc., have been developed by
writing similar functions with VBA.
Figure 3.23. A UDF for caclcuating the molar specific-heat for air
Figure 3.24. Using the cp_air function in Excel
In writing the UDF for the circle‟s diameter shown in Figure 3.21 we assigned a value
for the constant π because VBA does not provide a built-in function for it like Excel.
However, this is not necessary since it is possible to use the built-in functions PI
provided by Excel within the VBA function. This is very useful when built-in functions,
like MIN and MAX, can be used to minimise the programming effort needed for
developing the required UDF. More information about this and other features of the
VBA language can be found in specialised references [7-9].
3.3. Thermax installation and use

By providing custom functions for fluid properties, Thermax enables Excel to be used
as an educational platform for a wide range of thermofluid analyses [10]. Before
Thermax can be recognised by Excel you have to install it in your computer. To do that,
open the Thermax.xla file and then save it as an “Excel Add-in”. Recent Excel
versions locate all add-ins in a certain folder in the computer and automatically direct
you to that location when you want to save a new add-in. Save the Thermax add-in in
the specified location and restart Excel in order to activate it. Open a new Excel sheet
and then do the following:
1. Go to File and then click Options.

2. Select Add-Ins. From the Manage ribbon at the bottom of the menu select
Excel Add-ins and then press Go. The pull-down menu shown in Figure 3.25
will appear to you.
3. To add Thermax to the add-ins menu, tick (√) the corresponding box.
4. If for any reason you saved the add-in in a location that is different from the
default folder, then click on Browse and search for it in the destination folder
and select it.
Figure 3.25. Adding Thermax to the menu of Excel add-ins
Once installed, Thermax functions can be used in Excel's formulae just like the built-in
functions. For illustration, let us start a formula by entering the equal sign (=) in any
cell (say cell B2). If you now press the fx button in the formula ribbon, the Function
Wizard shown in Figure 3.26 will be shown. The Function Wizard first lists the various
categories of built-in functions provided by Excel. Scroll down the list of function
categories and select the User-defined functions. Then, all the functions provided by
Thermax will be listed alphabetically as shown in Figure 3.27.
Figure 3.26. Finding the add-in user-defined functions in the Function Wizard
Figure 3.27. Thermax functions listed alphabetically in the User Defined category
The first function in the list, Air_Data, is the auxiliary function that stores the data for
the thermo-physical properties of air at standard atmospheric pressure. This function is
called by other functions in the same to obtain the values of these properties at the
required temperature.
To start using the add-in functions, scrol down the list and select the function Airk_T
that determines the thermal conductivity (k) of air at a given temperature. Upon
pressing the OK button, the Function Arguments box shown in Figure 3.28 will be
shown.
Figure 3.28. The Function Arguments box for the “Airk_T” function
Figure 3.28 shows that this function has one input parameter, which is the temperature
in o C “TempC”, and gives a brief description of its intended use. Let us use the function
to determine the thermal conductivity for air at 25o C. Fill the form by entering the value
of the temperature, 25, as shown in Figure 3.29.
Figure 3.29. Using the function “Airk_T” to determine the thermal conductivity of
atmospheric air at 25o C
Note that the formula ribbon now shows the formula in cell B2, which is
“=Airk_T(25)”. The form also shows the calculated value of k, which is 0.02551
W/m.o C. When you press the “OK” button, this value will appear in the cell B2. You
can check this value with that given in Table A.1 and try other Thermax functions.
In certain situations the confinement of Excel‟s formula to one cell becomes too
restrictive for developing the analytical model. This situation arises, for example, when
an iterative process involves a non-linear equation such as the Colebrook-White
equation or the Soave-Redlich-Kwong equation of state. In this case, a UDF is needed
solve the nonlinear equation and return the result to Excel‟s formula like any built-in
function. For such a case, Thermax provides a Newton-Raphson solver for nonlinear
equations in addition to its property functions. Appendix C shows how to use this tool.
Appendix C also describes two interpolation functions provided by Thermax for
tabulated data which are useful for including additional fluid properties or other
tabulated data needed in a thermofluid analysis.
3.4. Closure
This chapter introduced the three auxiliary components of the Excel-based modelling
platform, Solver, VBA, and Thermax, that make it an effective platform for thermofluid
analyses. The chapter gave Examples of using the three solution methods provided by
Solver to deal with different types of problems. It also showed how VBA can be used
for developing user-defined functions not provided by Excel and explained the
procedure for installing Thermax and using its functions in Excel formulae. Appendix C
demonstrates the use of the Newton-Raphson solver provided by Thermax for solving
nonlinear equations and its two interpolation functions for tabulated data.
References
[1] Frontline Systems, internet: http://www.Solver.com/ (Last accessed November
23, 2015).
[2] L.S. Lasdon, R.L. Fox, M.W. Ratner, Nonlinear optimization using the
generalized reduced gradient method, Revue Française d'Automatique,
Informatique et Recherche Opérationnelle, tome 8, V3 (1974), p. 73-103.
Available at: http://www.numdam.org/article/RO_1974__8_3_ 73_0.pdf
[3] Wikkipedia, https://en.wikipedia.org/wiki/Gradient_method
[4] https://en.wikipedia.org/wiki/Evolutionary_algorithm
[5] Wikkipedia, https://en.wikipedia.org/wiki/Evolutionary_algorithm
[6] Y. A. Cengel and M. A. Boles. Thermodynamics an Engineering Approach,
[7] J. Walkenbach, Excel 2013 Power Programming with VBA, Wiley Publishing,
Inc. 2013.
[8] J. L. Latham, Programming in Microsoft Excel VBA, An Introduction, 2008.
Internet: http://ies.fsv.cuni.cz/default/file/download/id/21101 (Last accessed
November 29, 2015).
[9] http://www.fontstuff.com/vba/vbatut01.htm (Last accessed November 26, 2017).
[10] Thermax: An Excel add-in for Thermofluid Analyses. Available at:

https://www.mendeley.com/library/fileproxy?id=04a975bf-59cf-cde5-
7abe54d5b118d1e6
[11] S.C. Chapra and R.P. Canale, Numerical Methods for Engineers, 6th Edition,
McGraw Hill, 2010.
Exercises
1. A system of algebraic equations can be expressed in matrix form as follows:
70 1 0  a  636 
60  1 1  b   518 
    
  
40 0  1  c  307 
Solve the system of equations by using Solver to determine the values of the three
unknowns a, b, and c. This exercise is based on Example 9.11 in Chapra and
Canale [11]. The answer is: a = 8.5941, b=34.4118, and c = 36.7647.
2. Draw a line chart with Excel to show the variation of the following function in the
range 0 ≤ x ≤4:
f(x) = 2 sin x − x2 /10
Use Solver to find the maximum of the function in the same range. Based on
Example 13.1 in Chapra and Canale [11]. The answer is: f(x) = 1.7757 at x
=1.4276.
3. The curve shown in Figure 3.P3 is a plot of the function:
f  e sin2 
4
2
0
-2 0 100 200 300 400
Function
-4
-6
-8
-10
-12
-14
Angle
Figure 3.3.P A composite function
Use Solver to find:
a) The minimum value of the function and the corresponding angle

b) The maximum value of the function and the corresponding angle
c) The angle at which value of the function equals 4.0
4. Using the Excel sheet developed to solve Example 3-1 by the GRG Nonlinear
method, study the effect of using central-difference approximation of derivatives
instead of the default forward-difference approximation on the solution.
5. Consider the following set of simultaneous nonlinear equations:
x2 + xy = 10 (A)
y +3xy2 = 57 (B)
To solve the system with Solver, rearrange the equations as follows:
u(x, y) = x2 + xy−10 = 0 (C)
v(x, y) = y + 3xy2 −57 = 0 (D)
Create two cells (B1 and B2) to hold initial guesses for x and y. Enter the function
values themselves, u(x, y) and v(x, y) into two other cells (B3 and B4). The initial
guesses may result in function values of u and v that are far from zero. Determine
the sum of the function squares, i.e. u 2 + v2 , and store it in cell B5.
Use Solver to find the values of x and y in cells B1 and B2 (the Changing cells)
that make the value in cell B5 (the objective cell) equal to zero. Using this
procedure, find the roots of the above system starting with initial guesses of x =1
and y = 3.5.
This exercise is based on Example 7.5 in Chapra and Canale [11]. The correct pair
of roots are x=2 and y=3.
6. The volume V of liquid in a spherical tank of radius r is related to the depth h of

the liquid by:
V = πh 2 (3r −h)/3
Using VBA, develop a user-defined function that determines h at any given values
of r [m] and V [m3 ]. Check your function with r=1 m and V = 0.5 m3 . Answer: h =
0.431 m.
7. Extend the UDF developed in Section 3.2 for determining the specific heat for air
so that it can be used for other ideal gases as well. Note that in this case, the
function will have two input parameter: the name of the gas and the temperature.
8. Using suitable formulae for the thermodynamic properties of superheated steam,
develop user-defined functions with VBA for determining the specific enthalpy
and entropy of superheated steam from its pressure and temperature.
9. Using suitable formulae for the thermodynamic properties of superheated
refrigerant R134a, develop user-defined functions with VBA for determining
properties, e.g., enthalpy and entropy, of superheated R134a from its temperature
and pressure.
Optimisation analyses
The cost of energy is usually the major running cost for thermofluid systems. Therefore,
the main concern in the design of these systems is to maximise their energy efficiencies.
However, there must be a compromise between the lifetime saving in energy cost
achieved by improving the efficiency of the system and the additional cost due to this
improvement. Optimisation analyses of thermofluid systems aim to determine the most
suitable design or operating conditions for these systems. This chapter highlights the
advantages of the Excel-aided optimisation method over the traditional analytical
method and shows how the Excel-aided method can be applied in optimisation analyses
involving a single design variable or multiple design variables. The chapter also
evaluates the use of the Evolutionary method of Solver in thermofluid optimisation
analyses instead of the GRG Nonlinear method.
4.1. Analytical vs. Excel-aided optimisation of thermofluid systems

Any optimisation analysis involves two main steps: (i) developing an objective function
that quantifies the optimisation criterion in terms of the system‟s design and operating
parameters and (ii) using an optimisation method to find the values of these parameters
that optimise (maximise or minimise) the objective function. In the case of thermofluid
systems, the first step involves the development of a mathematical model for the system
based on thermofluid concepts. Compared to traditional optimisation methods that use
calculus techniques as the optimisation method, the Excel-aided method described in
this chapter allows the system‟s model to be developed with minimum idealisations and
approximations. Using Solver, method can also take into consideration any relevant
constraints on the optimisation step such as physical constraints (spatial or temporal),
economical constraints, or safety constraints. In what follows, these advantages will be
illustrated by considering the two stage air-compression system shown in Figure 4.1 for
which we want to determine the intermediate pressure (P2 ) that minimises the total
compression work.
Air
C1 C2
1
2 Intercooler 3
Cooling
water
Figure 4.1. Schematic diagrams of a two-stage air compressor with inter-stage

intercooling
4.1.1. The approximate analytical optimisation method

The development of the objective function for the traditional analytical method usually
requires simplifying assumptions to be made regarding properties of the materials and
fluids involved in the system so that the solution can be obtained analytically later. In
the present case, the system‟s model is based on the following assumptions:
1. Ideal isentropic compression processes in both stages

2. Constant specific heat for air
3. Perfect intercooling, i.e. T3 = T1
4. Zero pressure losses in the intercooler, i.e. P3 = P2
Figure 4.2 shows the resulting T-s diagram for the compression process.
T
P4
P2
4 2
P1
T 3 1
Intercooling
s
Figure 4.2. T-s diagrams of the ideal two-stage air compressor
With the above assumptions, Chapter 1 showed that the total compression work (wtotal )
of the two-stage compressor is given by:
 k 1 k 1

  P 
 c p T1 2   i    4 
P k k
wtotal  (4.1)

  1   Pi 
P

Where k is the ratio of the specific heat for air at constant pressure (cp ) to that at
constant volume (cv) and Pi is the intermediate pressure: Pi = P2 = P3 . Equation (4.1) is
the objective function for optimisation which can be used to determine the value of the
intermediate pressure that minimises the total compression work. Designating the
pressure ratio in the first and second compressor stages by rp1 and rp2 , respectively, and
the ratio (k-1)/k by α, Equation (4.1) can be put in the following simpler form:

wtotal  c p T1 2  rp1  rp2  (4.2)
The pressure ratio in the second stage rp2 can be replaced by the total pressure ratio (rp )
and the pressure ratio in the first stage (rp1 ) as follows:
P4  P4   P1  r p
rp 2      (4.3)
Pi  P1   Pi  r p1
Accordingly, Equations (4.3) can be written as:

  rp  
wtotal  c p T1 2  r p1  
   (4.4)
  r p1  
   
The analytical solution to the optimisation problem is obtained by determining the

intermediate pressure such that:
dwtotal
0 (4.5)
drp1
For convenience, let us replace rp1 in Equation (4.4) by x. Then, Equation (4.5) leads to:

d   rp  
c p T1 2  x      0

(4.6)
dx   x  
After differentiation, we get:
0  x  1

 rp ( ) x  1  0 (4.7)
Dividing by α and rearranging:
x  1  rp x  1 (4.8)
Multiplying both sides of the equation by x(α+1), we finally get:
x 2  rp , or x  rp (4.9)
Thus, the optimum intermediate pressure ratio is such that:
rp1  rp 2  rp (4.10)
This analytical solution indicates that in order to minimise the total compression work
the overall pressure ratio should be divided equally between the two compression
stages. According to Equation (4.10), the optimum pressure ratios rp1 and rp2 for a two-
stage air compression system designed to receive air at 100 kPa, 300K and discharge it
at 900 kPa are such that:
rp1  rp 2  9  3
The corresponding total compression work (w) can be determined from rom Equation
(4.1). Taking cp for air as 1.005 kJ/kg.K and k as 1.4, gives:
 1.4 1

wtotal  1.005  3002  23 1.4  = -222.35 kJ
 
4.1.2. The exact Excel-aided optimisation method

The accuracy of the analytical optimisation method described above is undermined by
the four simplifications and idealisations it makes. Figure 4.3 shows the T-s diagram for
the actual two-stage compression process.
T
P4
Px
4 2
2s
4s P1
T3
T1 3 1
Intercooling
s
Figure 4.3. T-s diagram for the two-stage air compressor
Unlike the ideal T-s diagram shown in Figure 4.2, the actual compression process
shown in Figure 4.3 allows for friction losses in the process by the increased entropy.
The diagram also allows for the imperfection of the intercooler by considering its
pressure losses and heat-transfer effectiveness. Therefore, the Excel-aided method that
adopts the actual compression model has these advantages over the analytical method.
Moreover, by using the fluid property functions provided by Thermax, the Excel-aided
method does not have to use a constant specific-heat for air. The following example
shows how the method can be applied to the two-stage compressor.
Example 4-1. Excel-aided optimisation of a two-stage air compressor

An air compressor with two stages of compression takes atmospheric air at 300K and
100 kPa and delivers it at a final pressure of 900 kPa. The adiabatic efficiency of both
stages of the compressor (ηc) is 85%. The intercooler that cools the air after the first
stage causes a pressure drop of 5 kPa and has effectiveness (ε) of 85%. Determine the
exit pressure of the first stage (Px) that minimises the total compression work.
The exact analytical model

This analytical model applies the exact variable-specific-heat method by using the
functions for ideal gases provided by Thermax. Given the inlet air temperature T1 , the
relative pressure at state 1 (Pr1 ) is determined by using the relevant function, which is
GasPr_TK. Assuming a value for the intermediate pressure (Px), the relative pressure
at state 2s (Pr2s ) can then be calculated from the following relationship [1]:
Pr 2 s  Pr1  Px / P1 (4.11)
Having determined the value of Pr2s , the ideal isentropic temperature after the first
compression stage T2s can be determined by using Thermax function GasTK_Pr.
Enthalpy at the actual temperature after the first-stage compression (h 2 ) can now be
determined as follows:
h2  h1  h2 s  h1  /  c
(4.12)
Where ηc is the adiabatic efficiency of the compressor and h 1 and h 2s are enthalpy
values at states 1 and 2s, respectively, which can be calculated by using the function
Gash_TK. Therefore, the compression work in the first stage (wc1 ) is given by:
wc1  h2  h1
(4.13)
The temperature and pressure after the intercooler (before the second compressor stage)
are given by:
T3  T2   T2  Twi  (4.14)
P3  P2  PIC (4.15)
Where ε is the effectiveness of the intercooler and ∆PIC is its pressure drop. Similarly,
the ideal exit temperature T4s and actual exit enthalpy h 4 can be calculated from:
P4 s
Pr 4 s  Pr 3  (4.16)
P3
h4  h3  h4 s  h3  /  c (4.17)
Finally, the compression work in the second stage (wc2 ) and the total compression work
(wtotal ), which is the summation of the work in the two stages, can be calculated from:
wc 2  h4  h3
(4.18)
wtotal  wc1  wc 2 . (4.19)
The Excel-aided solution

Figure 4.4 shows the sheet prepared for solving this example. The sheet reveals the
formulae entered in each cell in the calculations part, while the formula bar shows the
formula that calculates the total compression work (W_c =w_c1+w_c2) in cell L2. Cell
labelling has been used to clarify the formulae. Three Thermax functions are needed as
shown in Table 4.1. Figure 4.4 shows the calculated values at a “guessed” intermediate
pressure of 200 kPa that gives a total compression work (W_c) of 277.85 kJ.
Table 4.1.Usage of Thermax functions in the Excel model

# Function Usage
1 GasPr_TK Pr1 , Pr3
2 GasTK_Pr T2 , T4
3 Gash_TK h1 , h2 , h3 , h4
Figure 4.5 shows the total compression work obtained at various assumed values of the
intermediate pressure (P_i). The figure indicates that the total work has a minimum
value of about 310 kPa. The exact value of the intermediate pressure that minimises the
total work can be determined by using Solver. Figure 4.6 shows Solver Parameters set-
up that requires it to minimise the total compression work calculated in cell L2 by
adjusting the value of the intermediate pressure (P_i). Figure 4.7 shows the solution
obtained by the GRG Nonlinear method. As the figure shows, the optimum
intermediate pressure is 316.88 kPa at which the total compression work is 270.75 kJ.
Both values are higher than their corresponding values obtained in Example 4-1.
310
Total compression work (kJ)

300
290
280
270
260
0 200 400 600 800 1000
Intermediate pressure (kPa)
Figure 4.5.Variation of the total compression work with the intermediate pressure
Figure 4.6. Solver Parameters set-up for Example 4-1

This example demonstrates three advantages for the Excel-aided optimisation method
over the analytical method. Firstly, the Excel-aided method is easier to use because it is
applied directly to the objective function without differentiation. Secondly, it leads to
more realistic results because it can take into consideration the pressure and heat-
transfer losses in the compressors and the intercooler. Thirdly, adopting the exact
variable specific heat method makes the Excel-aided method more accurate than the
analytical method that uses a constant specific heat. Optimisation analyses of
thermofluid systems may also involve one or more of the following complications that
make the analytical method practically impossible to apply:
1. The model involves discrete-valued variables, such as the cost of pipes,

insulating material, and other equipment.
2. The optimisation analysis involves multiple parameters in the objective
function that have various constraints.
3. The model involves too many details that require the lengthy manipulations that
make analytical optimisation both tedious and inaccurate.
As the following sections illustrate, these complications make computer-aided methods,

such as the present Excel-aided method, the only viable option for optimisation
analyses of thermofluid systems.
4.2. Optimisation with a single design variable

Although many optimisation analyses of thermofluid systems involve a single design
parameter, such as the thickness of insulation or the diameter of a pipeline, the
nonlinearity of the equations involved and/or the variation of fluid properties with the
temperature and pressure make the analytical optimisation method practically
impossible to apply. This section considers four cases that illustrate how the Excel-
based platform can be used to perform the optimisation analyses. All the analyses in
this section are performed by using the GRG Nonlinear method of Solver.
4.2.1. Optimum thickness of thermal insulation

Increasing the thickness of insulation helps to increase the resistance to heat-transfer
and reduces the heat loss, but the cost of insulation also increases with the thickness.
Therefore, there is an optimum thickness beyond which adding more insulation
becomes economically unprofitable. The following example illustrates the methodology
of determining the optimum insulation thickness by using the Excel-aided optimisation
method. The example is based on Example 4-1 in Janna [2].
Example 4-2. Determining the optimum thickness of insulation around a duct

Figure 4.8 shows a circular duct of external diameter D1 that transports hot air-
conditioning air. The difference in temperature between the air inside the duct and the
surroundings causes heat loss to the surroundings that is to be minimised by covering
the duct by a layer of insulation. Due to space limitation, the outside diameter of the
insulation D2 cannot exceed 12 cm.
D2
D1
Figure 4.8. Schematic of the insulated duct
It is required to determine the most economical thickness of insulation based on the

following costs:
(a) The cost of pumping the air through the duct (Cp ) is given by:
C p  3  106 D15 [$/year] (4.20)
(b) The cost of heating the air (Ch ) is given by:
Ch  9 /  [$/year] (4.21)
Where δ is the insulation thickness in metres.
The analytical model

The total cost (CT) given by the summation of the pumping and heating costs is:
3  10 6 9 6
CT   = 3  10  9
(4.22)
D15
 D15 D2  D1 
By imposing the requirement that the maximum diameter D2 should not exceed 0.12 m,
the total cost becomes:
3 10 6 9
CT   (4.23)
D15 0.12  D1 
Differentiating Equation (4.23) with respect to D1 and equating the result to zero, Janna
[2] obtained the following solution for D1 :

D1  1.67  10 6 0.12  D1  
2 1/ 6
(4.24)
Equation (4.24), which requires an iterative solution, can be solved by using the Goal
Seek command. However, the Excel-aided method enables minimising the total cost
given by Equation (4.23) without differentiation by using Solver.
Excel solution
Figure 4.9 shows the Excel sheet developed for this example. Note that the sheet
reveals the formulae used in the calculations part in which cell labelling has been used.
The data part includes only the value of D2 (D_2 = 0.12 m). The sheet gives the total
cost for a guessed inner diameter D1 (D_1= 0.1 m). At this guessed diameter, the
insulation thickness (t) is 2 cm and the total cost (C_Total) is 450.3$. The diameter that
minimises the total cost can be found by using Solver and Figure 4.10 shows its
Parameters dialog box for this task.
Figure 4.9. Excel sheet for optimisation of the insulated duct in Example 4-2
Figure 4.10. Solver set-up for the optimisation Example 4-2
The set-up shown in Figure 4.10 requires Solver to find the values of D1 (in the
adjustable cell) that minimises the total cost (in the target cell). Initially, the GRG
Nonlinear method was used with the default options shown in Figures 3.15, i.e.,
without automatic-scaling, but Solver failed to converge to a solution. Figure 4.11
shows the optimum solution obtained with the GRG Nonlinear method and automatic-
scaling. The optimum value thus obtained, which is D1 = 0.045763 m, agrees well with
the value obtained by Janna [2].
Figure 4.11. Solver‟s solution of Example 4-2
4.2.2. Optimisation of a rectangular fin

Fins, like the one shown in Figure 4.12, are usually used to increase the rate of heat
transfer from hot surfaces by convection as in car radiators. Since the addition of fins
increases the weight of the radiator, the improvement in heat-transfer achieved by
adding the fins must not be undermined by their additional weight that increases the
fuel consumption of the car. In the following example, which is adapted from Bejan et
al [3], it is required to optimise the geometry of a rectangular fin for the maximum
possible heat-transfer rate.
T∞
Tb
Figure 4.12. A straight rectangular fin

Example 4-3. Optimisation of a rectangular fin

An electronic package includes several parallel straight rectangular fins such as the one
shown in Figure 4.12. The width of each fin is W = 2.2 cm and the fins are swept by
forced air with U∞ = 1.75 m/s and T∞ =20o C. The fin material is aluminium. Weight
limitations on the overall package permit the use of only 1 g aluminium for each fin. It
is required to determine the fin thickness (t) and length (L) that maximise the heat
transfer rate extracted per fin and the corresponding heat transfer rate from the fin per
degree of the temperature difference at the base (Tb -T∞ ).
Analytical model
Using the definition of fin efficiency (ηfin ) introduced in Appendix B, the rate of heat
transfer from a fin ( Q fin ) can be determined from:
Q fin   fin Q max (4.25)
Where Q max is the maximum possible rate of heat-transfer from the fin as given by:
Q max  hAfin Tb  T  (4.26)
Where h is the heat-transfer coefficient and Afin is the heat-transfer area of the fin
which, for a rectangular fin with convection boundary condition at the tip, is given by:
A fin  2wLc (4.27)
Where Lc is defined as:
Lc  L  t / 2 (4.28)
As shown in Appendix B, the fin efficiency for a rectangular fin with convection
boundary condition at the tip is given by:
tanh( mLc )
 fin  (4.29)
mLc
Where m is defined as:
m 2h / kt (4.30)
Where, k is the thermal conductivity of the fin‟s material. For a laminar flow, the
coefficient of heat transfer h is calculated from [4]:
k air
h  0.664 Re1 / 2 Pr1 / 3 (4.31)
w
Where, k air and Pr are the thermal conductivity and Prandtl number of air, respectively.
Excel implementation
Figure 4.13 shows the Excel sheet prepared for the solution of this example. The given
data are entered on the left-side column of the sheet. The mass of the fin is specified as
1 g of aluminium. The thickness of the fin (thickness) is initially assumed and given the
value of 2 mm. For this initial guess, the sheet calculates the Reynolds number (Re),
heat-transfer coefficient (h) and volume of the fin (Vol). The figure shows the formulae
used in the calculations.
Figure 4.13. Excel sheet for example 4-3
The length of the fin is calculated in cell E7 from:
V m fin
L  (4.32)
wt  al wt
Where, mfin is the mass of the fin (1 gm) and ρal is the mass density of aluminium. The
fin efficiency (ηfin) and maximum rate of heat-transfer from the fin (Qmax) are
determined from Equations (4.29) and (4.26), respectively. As shown in Figure 4.13,
the resulting rate of heat transfer is 0.037 W. Solver can now be used to find the
combination of t and L that gives the maximum possible rate of heat transfer from the
fin. Figure 4.14 shows the required set-up for Solver to search for the value of t that
maximise the fin‟s heat transfer rate. No constraints have been inserted for this case.
Figure 4.15 shows that the required thickness is t=0.00036 m or 0.36 mm, which agrees
with the value obtained by Bejan et al [3]. The fin‟s length is 4.6 cm and its heat
transfer rate is 0.044 W/K. The corresponding values obtained by Bejan et al [3] are
4.66 cm and 0.0436 W/K.
Figure 4.14. Solver parameters dialog box for Example 4-3
Figure 4.15. Solver solution to Example 4-3
4.2.3. Optimum outlet temperature from a heat-exchanger

Steam-turbine power generation plants consume large amounts of fuels and have to
reject significant fractions of their fuels‟ energy to the atmosphere as required by the
second law of thermodynamics. The waste heat is usually released to a near-by river or
lake or rejected directly into the atmospheric air by means of a cooling tower. Since the
cost of the cooling system constitutes a major fraction of the total installation cost, the
optimisation of its components is important for the economic feasibility of the whole
plant. The present analysis focuses on the heat-exchanger that transfers the rejected heat
from the power plant condenser to the cooling tower.

The total annual cost of the heat-exchangers (CT) consists of two parts: (i) the annual
cost of cooling water (CAW ) and (ii) the annual cost of operating the exchanger (CAO ):
CT  C AW  C AO (4.33)
The annual cost of the cooling water CAW is given by:
m
C AW  3600Vcw = 3600 c w (4.34)

Where V is the volume flow rate of the cooling water (m3 /s), m  is the mass flow rate
of the cooling water (kg/s), ρ is the density of water (kg/m3 ), cw is the cost of delivering
the cooling, water ($/m3 ), and τ is the annual number of operation hours.
The annual cost of operating the exchanger is given by:
C AO  Ac o (4.35)
Where A is the heat-exchanger area, in m2 , and co is the annual unit cost of operating
the heat exchanger in $/(m2 .year).
Equations (4.34) and (4.35) indicate that CAW and CAO depend on the cooling water
mass flow rate and exit temperature (Tco ). Given a value for the water‟s exit temperature
(Tco ), the mass flow rate of the cold water ( m c ) can be obtained from the following
equation:
m  Q / C pc Tco  Tci  (4.36)
Where Q is the rate of heat-transfer in the heat-exchanger, cpc and Tci , are the specific
heat and inlet temperatures of the cold water, respectively.
As shown in Appendix B, the surface area (A) of the heat-exchanger can be calculated
from the following heat-exchanger equation:
A  Q / UFTlm (4.37)
Where U and F are the overall heat exchanger coefficient and correction factor in the
heat-exchanger equation, respectively, and ∆Tlm is the log-mean temperature difference
(LMTD) defined in Appendix B by Figure B.2 and Equation (B.3).
The above analytical model indicates that the total annual cost of operating the
exchanger depends on the water flow rate and the area of the heat-exchanger.
Increasing the water flow rate will increase the cost of cooling water, but reduce the
required heat-exchanger area and vice versa. The optimum water outlet temperature is
that at which the total annual operation cost is minimal. The following example shows
how this exit temperature can be determined by using Excel and Solver. The example is
based on Example 9.6 in Janna [2].
Example 4-4. Optimum outlet temperature of the cooling water

A shell and tube heat exchanger uses water as a cooling medium to dispose of 1000 kW
of heat from a stream of hot water. The hot water enters the exchanger at 93.3o C and
leaves at 37.8o C. The cooling water enters the exchanger at 26.7o C. Use the data given
below to calculate the optimum cooling water outlet temperature:
U = 613 W/m2 .K
F = 0.817
cw = $1.32/m3
co = $215/(m2 .year)
τ = 7800 hour/year
Take the density of cold water (ρ) and its specific heat (Cpc) as 994 kg/m3 and 4179
J/kg.K, respectively.
Development of the Excel sheet

Figure 4.16 shows the Excel sheet developed for this example. The left side of the sheet
stores the given value for the rate of heat-transfer (Q), the heat exchanger data, the
properties of the hot and cold water, and the water and operation costs involved in the
analysis. The calculations part starts with an assumed exit temperature for the cold
water (T_c2) which is given a value of 30o C.
Figure 4.16. The Excel sheet developed for Example 4-4
Based on the assumed exit temperature for water, the sheet calculates the log-mean
temperature difference (LMTD), the area of the heat-exchanger (A_HX), the mass flow
rate of cold water (m_c), the volume flow rate of cold water (V_c), the annual water
and exchanger costs (C_AW and C_HX), and the total annual cost (C_T). Figure 4.16
also reveals the formulae involved in the calculations by the side of their respective
cells. Figure 4.16 shows that the annual cost of water is $61,453.36, the annual cost of
operating the exchanger is $14,317.57, and the resulting total annual cost is $75,770.93.
The sheet was used to calculate the heat-exchanger costs at different values of the
cooling-water outlet temperature and the results are shown in Figure 4.17. This figure
shows that, as the water temperature increases, the annual cost of water decreases but
the annual cost of operating the exchanger increases. The total annual cost initially
decreases as the water outlet temperature is increased but then increases.
80000
70000 C_AW
C_HX
Annualised costs ($/yr)
60000
C_T
50000
40000
30000
20000
10000
0
20 40 60 80 100
Cold water exit temperrature (oC)
Figure 4.17. The variation of the total cost with the cooling water outlet temperature
From Figure 4.17, it can be seen that the optimum water outlet temperature is in the
range 45 to 60o C. A more accurate estimation of the optimum cooling-water outlet
temperature can be determined by using Solver. Figure 4.18 shows the required set-up
in which two constraints have been imposed on the optimisation process to ensure a
physically meaningful result. Figure 4.19 shows the sheet with the solution found by
Solver. As the figure shows, the optimum cooling-water outlet temperature is 53.27o C
at which the total annual cost is $26,666.4.
4.2.4. Optimisation of a hot-water generation system with two steam heaters

Li and Priddy [5] presented a hot water generation system in which the water is heated
by two streams of saturated steam as shown in Figure 4.20. The incoming water that
enters at Twi is first heated in the low-pressure steam heater to an intermediate
temperature Tx and then heated in the high-pressure steam heater to the desired
temperature Two . The high-pressure steam costs more than the low-pressure steam and
the high-pressure steam heater costs more to purchase than the low-pressure steam
heater. Therefore, the total cost of the heating system, which is the summation of the
initial costs and steam costs, depends on how the heat load is divided between the two
heaters. The following example shows how the system can be optimised.
Figure 4.18. Solver set-up for Example 4-4
High-pressure steam Low-pressure steam
Tshp Ts1p
Two Tx Twi
Water
HP heater LP heater
Tshp Ts1p
Saturated liquid water
Figure 4.20. Schematic of the hot-water generation system
(adapted from Li and Priddy [5])
Example 4-5. Optimisation of a hot water generation system

The low and the high-pressure heaters in the hot-water generation system shown in
Figure 4.20 operate at 350 and 550 kPa, respectively. It is required to determine the
optimal sizes of the two heaters that minimise the total annual relative cost of the
system that operates for 2000 hours per year based on the following data:
Water flow rate ( m ) 180,000 kg/hr

Water inlet temperature (Twi ) 65o C
Water outlet temperature (Two ) 150o C
Overall heat transfer coefficient (U) 10220.667 kJ/hr-m2 -o C
Annual fixed charge rate (β) 20%
Low-pressure heater High-pressure heater

Steam cost [$/106 kJ] 5.7 5.72
Heater surface cost [$/m2 ] 108.7 109.8

The cost of each heater is the sum of its annualised initial cost (Ci ) and annual steam
cost (Cs ). For the system that consists of two heaters, the total cost (C) is given by:
C = (Ci + Cs )hp + (Ci + Cs )lp (4.38)
Where Ci is the annualised initial cost ($/year) and Cs is the annual steam cost ($/year)
and the subscripts hp and lp refer to the high-pressure heater and the low-pressure
heater, respectively. For a given heat exchanger, the steam cost is determined from the
heat transfer rate in the heater, which is also the amount of heat transferred to the water.
Using the notation shown in Figure 4.20, the rates of heat transfer in the two heaters can
be expressed as:
 m
Qlp
 C p Tx  Twi  (4.39)
 C p Two  Tx 
Q hp  m (4.40)
Where m is the mass flow rate of water. The annual steam costs for both heaters are:
C s ,lp  alp Q lp . (4.41)
C s ,hp  a hp Q hp . (4.42)
Where a lp and a hp are the steam costs in the low-pressure heater and the high-pressure
heater, respectively, in $/kJ, and τ is the operating hours per year.
The relative cost of each heater is simply the product of its heat transfer surface area (A)
and its unit cost per square metre (b). If we apply an annual fixed charge rate β, the
annual relative cost of the low-pressure heater becomes:
Ci ,lp  blp Alp . (4.43)
The transfer surface areas of the two heaters are determined by using the log-mean
temperature difference method described in Appendix B. Accordingly, the surface area
of the low-pressure heater (Alp ) is given by:
Alp  Q lp / U lp Tlm,lp (4.44)
Where ΔTlm is the log-mean temperature difference which for the low-pressure heater is
defined as:
Tlp,1  Tlp, 2
Tlm,lp 
ln Tlp,1 / Tlp, 2 
(4.45)
Where,

Tlp,1  Tslp  Tx  (4.46)
Tlp, 2  Tslp  Twi  (4.47)
Combining Equations (4.43), (4.44), and (4.45) gives:
C i ,lp 

blp . .Q lp ln Tlp,1 / Tlp, 2  (4.48)
U lp Tlp,1  Tlp , 2
Similarly, the annual relative cost for the high-pressure heater is given by:
C i ,hp 

bhp . .Q hp ln Thp,1 / Thp, 2  (4.49)
U hp Thp,1  Thp, 2
Where,
Thp,1  Tshp  Two  (4.50)
Thp, 2  Tshp  Tx  (4.51)

Excel implementation
Figure 4.21 shows the Excel sheet developed for this example. The top-left part of the
sheet shows the data provided by Li and Priddy [5]. The bottom-right part of the sheet
shows calculations of the steam and relative initial cost for the two heaters and their
total for each heater. The figure shows the calculations and total relative cost of the
system at the top right-hand side of the sheet for an intermediate temperature Tx = 65o C.
The corresponding saturation temperatures for the low-pressure and high-pressure
steams are 138.88o C and 155.48o C, respectively, as shown in cells F6 and I6.
Figure 4.21. Excel sheet for the hot-water generation system
Table 4.2 shows the heat load distributions and surface areas of the two heaters for Tx in
the range 65-135o C. It can be seen from the table that the heating load supported by the
low-pressure heater increases as the intermediate temperature Tx increases. At Tx =
65o C, when the heating load is solely supported by the high-pressure heater, the
combined heaters surface area is 207.41 m2 . However, at Tx = 135o C, when the heating
load is supported mostly by the low-pressure heater, the total surface area reaches
315.46 m2 .
Table 4.2. Heat load distributions and heater sizes

Heat load Q [kW] Heater surface area [m2 ]
Tx [o C] Total area [m2 ]
LP HP LP HP
65 0 17850 0 207.41 207.41
75 2100 15750 10.76 198.74 209.50
85 4200 13650 23.35 188.93 212.28
95 6300 11550 38.54 177.61 216.15
105 8400 9450 57.67 164.24 221.91
115 10500 7350 83.54 147.91 231.45
125 12600 5250 123.67 126.93 250.60
135 14700 3150 217.95 97.51 315.46
Figure 4.22 shows the variation with Tx of the combined initial cost (C_initial), the
combined steam cost (C_steam), and the total combined cost (C_total). The figure
shows that the minimum combined initial cost is obtained at Tx = 65o C, while the
minimum combined steam cost is obtained at Tx =135o C. If only the initial cost is
considered, then a single high-pressure heater would be preferred for the hot water
production. On the other hand, if only steam cost is considered, then only a single low-
pressure heater is to be selected. However, the variation of the total relative cost with Tx
indicates that the minimum total cost occurs at an intermediate temperature which is
around 115o C. At this temperature, the figures in Table 4.2 indicate that the high-
pressure heater takes one-third of the total heat load while the low-pressure heater
supplies the remaining two-thirds.
741000 8000
7000
Initial and total cots ($)
739000
6000
Steam cost ($)

737000 5000
4000
735000 3000
C_total 2000
733000
C_steam
1000
C_initial
731000 0
50 70 90 110 130 150
Intermediate temperature (Tx) oC
Figure 4.22. Variation of the systems costs with the intermediate temperature Tx
To precisely determine the optimum value of Tx, we can use Solver. Figure 4.23 shows
the prepared Solver Parameters dialog box for the system‟s optimisation. Two
constraints have been imposed on Solver‟s solution which are:
Tx  Ts ,l
Tx  Tw,in
Pressing the Solve button at the bottom of the dialog box will trigger Solver to iterate
and search for the value of Tx that makes the total cost assume its minimum value.
Figure 4.24 shows the solution found by Solver, which is Tx=116.5o C. At the optimum
temperature, the surface area of the LP heater is 88.428 m2 while that of the HP heater
is 145.07 m2 . The relative cost of the LP heater is 446,004.8$ and that of the HP heater
is 292,679.7$, giving a total relative cost of 738,684.4$. This result shows that steam
cost is an important variable in the optimisation of the hot-water production system. In

the present case, steam cost for the low-pressure heater is substantially lower than that
for the high-pressure heater. Therefore, the optimised solution favours the use of the
low-pressure heater even though the total area is not is not the minimum value.
Figure 4.23. Solver parameters dialog box for determining the optimum intermediate
temperature for two heat-exchangers in series
Figure 4.24. Solver solution for the hot-water generation system
4.3. Multi-variable optimisation

All the four cases considered in the previous section involved a single optimisation
variable, which is the intermediate pressure (Px) in Example 4-1, the thickness of
insulation in Example 4-2, the fin thickness in Example 4-3, and the optimum water
temperature in Examples 4-4 and 4-5. However, Solver is also capable of dealing with
optimisation analyses with multiple variables, which is the ultimate application of

computer-aided methods in optimisation analyses. The case considered in this section is
that of a cooling tower for which our design objective is to minimise its total lifetime
cost. The relevant objective function in this case involves two design variables, which
are the mass flow rate of the cooling water and the surface area of the heat-exchanger.
The example is based on Problem 11.7 given by Stoecker [6].
Example 4-6. Optimisation of the cooling tower for a power plant

A cooling-tower is to be used for removing 14 MW of heat rejected by the condenser of
a steam-turbine power plant as shown in Figure 4.25.
Cooling
Tower QR
Turbine
Tc
Condenser
Figure 4.25. Schematic for the cooling system in Example 4-6
The total lifetime cost of the system (CT) consists of three parts: (a) the first cost of
cooling tower (C1 ), (b) the lifetime cost of pumping power (C2 ) and (c) the lifetime
penalty in power production due to elevation of temperature of cooling water (C3 ). The
three costs are estimated by the following formulae:
C1  800A 0.6 (4.52)
C 2  0.0005m 3 (4.53)
C 3  270Tc (4.54)
Where A is the cooling-tower area in m2 , m is the flow rate of cooling water in kg/s,
and Tc is the temperature of water entering the condenser in o C.
It is required to determine the values of the three design variables A, m , and Tc, that
minimise the total cost of the cooling system given that the rate of heat rejection (QR ) is
known to be related to the three factors by the following empirical formula:
QR  3.7m
 1.2 ATc (4.55)
Solution
Although three variables are involved in the optimisation process, which are A, m , and
Tc, Equation (4.55) can be used to reduce the number to two variables only, viz., A and
m . Figure 4.26 shows the Excel sheet developed for this example. The only data in this
problem is the rate of heat rejection (Qcooling) which is assigned the given value of 14
MW.
The intermediate calculations start with initially assumed values for the water flow rate
(m_w=500 kg/s) and the area of the cooling tower (A=150 m2 ). Based on these assumed
values, Equation (4.55) is used to calculate the cooling-water temperature (T_cw) and
then the three costs given by Equations (4.52) - (4.54). Figure 4.26 reveals the formulae
used in these calculations.
As Figure 4.26 shows, the calculated cooling-water temperature is about 14.56o C and
the resulting total lifetime cost (C_T) is $82,601.68. The values of (m_w) and (A) that
minimise the total cost (C_T) can be found by using Solver and Figure 4.27 shows the
required set-up for the Parameters dialog box. This set-up requires Solver to minimise
the total cost (C_T) by adjusting the values of the two parameters, m_w and A.
Although no constraints have been specified, a better practice would be to apply
reasonable upper and lower limits on the two optimisation variables. The solution found
by Solver is shown in Figure 4.28. As the figure shows, the optimum mass flow rate for
the cooling water is 202.52 kg/s and the optimum area of the cooling tower is 167.95
m2 . The value obtained for the water mass flow rate agrees well with that given by
Stoecker [6], which is 202.6 kg/s. Stoecker [6] did not give the optimum value for the
heat-exchanger area.
Figure 4.27. Set-up of Solver Parameters dialog box for Example 4-6
Figure 4.28. Solution of Example 4-6 obtained with Solver
4.4. Optimisation analyses with the Evolutionary method

Chapter 3 showed that the Evolutionary method has an important advantage over the
GRG Nonlinear method which is the ability to find the global optimal point. Although
the computer time of the method was higher than that of the GRG Nonlinear method, it
could be reduced by reducing the default value for the population size. To evaluate the
advantage of the Evolutionary method with respect to optimisation analyses of
thermofluid systems, it was used to solve Examples 4.2 and 4.3 that involved a single
optimisation variable and Example 4-6 that involved two design variables. The same
sheets previously developed for these examples were used by changing the solution
method in Solver‟s Parameters form from the GRG Nonlinear method to the
Evolutionary method.
The initial trial to solve Example 4-2 with the Evolutionary method used the same
Solver set-up shown in Figure 4.10 that did not impose any limits on D1 , but the
solution failed even though imposing upper and lower bounds on the problem variables
is optional like the GRG Nonlinear method. Therefore, both upper and lower limits
were imposed as shown in Figure 4.29. With this set-up Solver obtained the solution
shown in Figure 4.30 which is identical to that of the GRG Nonlinear method shown in
Figure 4.11.
Figure 4.29. Solver set-up for the optimisation with the Evolutionary method
Figure 4.30. Optimised solution for Example 4-2 with the Evolutionary method
Figure 4.14 shows that Solver‟s set-up for Example 4-3 already imposed both lower and
upper limits on the changing variable and, therefore, no change was needed for this
example. Figure 4.31 shows the solution obtained with the Evolutionary method which
is also identical to that obtained with GRG Nonlinear method shown in Figure 4.15.
With respect to Example 4-6, it was solved with the Evolutionary method using the
same Solver‟s set-up shown in Figure 4.27, i.e., without imposing upper or lower limits
on the changing variables. Solver‟s solution for this example shown in Figure 4.32 is
also similar to that obtained with the GRG Nonlinear method shown in Figure 4.28.
With the default set-up for the Evolutionary method shown in Figure 3.16.b, in which
the number of the population size is 100, the method required more than two minutes of
computer time to solve Example 4-2. By reducing the population size from 100 to 10,
the method reached the same solution in less than a second. Similarly, the Evolutionary
method took more than 100 seconds to solve Example 4-3 with a population size of
100. By reducing the population size to 10, the method took about one second to solve
the example which is comparable to that of the GRG Nonlinear method. Example 4-6
took 65 seconds with the Evolutionary method, but this could be reduced to only 5
seconds when the population size was reduced to 10.
4.5. Closure
This chapter demonstrated the application of the Excel-based modelling platform for
optimisation analyses of thermofluid systems. The chapter initially highlighted the
limitations of the traditional optimisation method that uses calculus techniques
compared to the computer-aided optimisation method. The chapter then showed how an
Excel-aided method can be utilised for optimisation analyses that involve a single
design variable and multiple design variables. Finally, the chapter addressed the issue
of using the Evolutionary method instead of the GRG Nonlinear method in
thermofluid optimisation analyses. The optimisation analyses with a single variable and
with two variables showed that the Evolutionary did not improve those obtained earlier
with the GRG Nonlinear method.
Appendix D presents an optimisation analysis that involves an air-conditioning duct

that has two sections with different diameters and the objective of the analysis is to
optimise both diameters of the duct sections, D1 and D2 , as well as the required
insulation thicknesses, δ1 and δ2 .
References
[1] Y.A. Cengel and M.A. Boles, Thermodynamics an Engineering Approach,
[2] W.S. Janna, Design of Fluid Thermal Systems, 3rd Edition, CENGAGE Learning,
2011.
[3] A. Bejan, G. Tsatsaronis, M. Moran, Thermal Design & Optimisation, John Wiley
& Sons, 1996.
[4] Y.A. Cengel, and A.J. Ghajar, Heat and Mass Transfer: Fundamentals and
Applications. McGraw-Hill, 2011.
[5] K.W. Li and A.P. Priddy, Power Plant System Design, John Wiley and Sons, 1985.
[6] W. Stoecker, Design of Thermal Systems, McGraw-Hill, 1989.
Exercises
1. Using the analytical optimisation procedure, show that the compression ratios that
minimise the total work input for in a multi-stage compression systems with n
compression stages and a maximum compression ratio rp,max are given by:
r p1  r p 2  ....  r p , n 1  r p ,n  n r p ,max
2. An air compressor with three stages of compression takes atmospheric air at 300K
and 100 kPa and delivers it at a final pressure of 900 kPa. The adiabatic efficiency
of the three compressors (ηc) is 85%. Each one of the intercoolers that cool the air
after the first and second stages causes a pressure drop of 5 kPa and has
effectiveness (ε) of 85%. By suitably extending the Excel sheet developed for
Example 4-1, determine the exit pressures of the first and second stages, Px and Py ,
respectively, that minimise the total compression work.
3. Air, initially at a pressure of 175 kPa and a temperature of 50o C, is to be
compressed to a final pressure of 17,500 kPa by means of a two-stage compression
system that consists of a centrifugal compressor in series with a reciprocating
compressor as shown in Figure 4.P3. The advantage of this arrangement is that
centrifugal compressors can handle high-volume flow rates but develop only low
pressure ratios while reciprocating compressors are suited to low-volume flow
rates but can develop high pressure ratios.
P2 = 17,500
Intercooler
T1 = 50o C
Reciprocating
Compressor
Centrifugal
Compressor
Po = 175 kPa
To = 50o C
Qo = 15 m3 /s
Figure 4.P3. The air compression system with centrifugal and reciprocating
compressors in series
The initial flow rate of air is 15 m3 /s and its temperature after the intercooler can
be taken as the same as the inlet temperature of 50o C.
The first costs of the centrifugal and reciprocating compressors can be represented
by the following formulae:
P1
C c  70Qo  1600 ($) (A)
P0
P2
C r  200Q1  800 ($) (B)
P1
Where Cc and Cr are the first costs of the centrifugal and reciprocating
compressors, respectively, and Q0 and Q1 are the corresponding volume flow rates.
(a) Treating air as an ideal gas, develop the relevant analytical model and
objective function for optimisation in terms of the pressures and flow rates
appearing in Equations (A) and (B).
(b) Use Solver to determine the optimum intermediate pressures and the
corresponding total initial cost of the system.
This exercise is based on Problem 8.5 in Stoecker [6], p 178, Answer: minimum
cost = $24,100
4. A gas turbine operates on a regenerative Brayton cycle as shown in Figure 4.P4.

The air entering the compressor at 100 kPa, 300 K and the hot exhaust gas is used
to preheat the compressed air before the combustion chamber.
Regenerator
T6 , P6
CC
T3 , P3 T4 , P4 T5 , P5
T2 , P2
T1 , P1
Compressor Gas turbine
Figure 4.P4. The regenerative gas-turbine system
The combustion gases leave the combustion chamber (CC) at 1400 K. Take the
regenerator effectiveness as 80% and the isentropic efficiency of both compressor
and turbine as 75%. The working fluid (air) can be treated as an ideal gas and
pressure losses can be neglected.
(a) Using the functions provided Thermax for ideal gases, develop a model that
determines the thermal efficiency of the regenerative gas-turbine cycle with
the exact method of analysis.
(b) By varying the pressure ratio from 2 to 16, study the effect of pressure ratio on
the thermal efficiency of the plant.
(c) Use Solver to determine the pressure ratio that maximises thermal efficiency.
5. Two identical pumps operate in parallel as shown in Figure 4.P5 to deliver a total
of 0.01 m3 /s of water at 15o C. The head losses h L1 and h L2 in the two lines depend
upon their volumetric flow rates according to the following formulae:
hL1  2.1  107 Q12 kPa (A)
hL 2  3.6  107 Q22 kPa (B)
Where Q1 and Q2 are the respective flow rates in cubic metres per second.
0.01 m3 /s
Q Q
Figure 4.P5. Water delivery system with two pumps in parallel
Develop an Excel-aided model to determine the optimum values for the two pump-
flows that minimise the total pumping power:
W P  g Q1 hL1  Q2 hL 2  / 1000 [kW] (C)
This exercise is based on Problem 8.3 in Stoecker [6], p. 177. Answer: Q1 =0.00567
m3 /s, Q2 =0.00433 m3 /s.
6. For the circular air duct system shown in Figure 4.P6, it is required to determine
the values of the three diameters D1 , D2 and D3 that minimise the drop in static
pressure between points A and B.
16 m 12 m 20 m
Q1 Q4
D1 D2
D3 B
A Q2 Q3
Figure 4.P6. The air duct system
Use the following data:
Q1 = 2.4 m3 /s,
Q2 = 0.6 m3 /s,
Q3 = 1.0 m3 /s,
Q4 = 0.8 m3 /s,
Quantity of sheet metal available for the system, 60 m2 .
The friction factor f = 0.02
Air density ρ =1.2 kg/m3
(a) Set up the objective function and constraints in terms of D1 , D2 and D3 .

(b) Using Excel and Solver, determine the optimal diameters and the minimum
total head loss.
This exercise is based on Problem 8.9 in Stoecker [6], p 182. Answer: D1 ,= 0.468
m D2 = 0.426m and D3 .= 0.325 m. h f = 17.203 m.
Thermal insulation
Design analyses of thermal insulation demonstrate three common considerations in the

design of fluid-thermal systems which are energy conservation, safety considerations,
and economic feasibility. The analytical equations for the analyses of thermal insulation
around a pipe involve a number of parameters that include the temperature of the fluid
inside the pipe, the temperature of the surroundings, and the thickness and thermal
conductivity of the insulation. Because of the nonlinearity of these equations, the
analyses require iterative solutions. This chapter shows how Goal Seek and Solver can
be used for the analyses. While Goal Seek can be used to perform the simple analyses
that involve a single design variable, Solver can be used to deal with the analyses that
involve multiple design variables and constrained optimisation.
5.1. The analytical model for thermal insulation around a pipe

Figure 5.1.a shows a pipe that transports a fluid at a temperature (T∞i ) which is higher
than that of the surroundings (T∞o ). The temperature difference causes the hot fluid
being transported to lose part of its energy to the cooler ambient air by heat-transfer.
The rate of heat-transfer from the surface of the pipe can be reduced by covering it with
an insulation material as shown in Figure 5.1.b. Apart from saving the energy, the
insulation makes the workplace safer by reducing the surface temperature of the pipe.
T∞o , h o T∞o , ho
r1 r2 r1 r2
T∞i , h i
T∞i, hi
r3
Ts T3
(a) (b)
Figure 5.1. Fluid-transporting-pipe (a) without insulation and (b) with insulation
Without insulation, the rate of heat-transfer,  , is

Q given by:
1
  Ti  To
Q (5.1)
1
Rth
Where Rth is the thermal resistance to heat-transfer through the pipe as given by:
1 lnr2 / r1  1
Rth    (5.2)
hi A1 2Lk pipe ho A2
Where h i and A1 are inside heat-transfer coefficient and the corresponding surface area,
respectively, h o and A2 are outside heat-transfer coefficient and the corresponding
surface area, respectively, and L and k pipe are the length and thermal conductivity of the
pipe, respectively. After insulation, the reduced rate of heat-transfer, Q , is given by:
2
Q 2  Ti  To  / Rth2 (5.3)
Where Rth2 is the increased thermal resistance after the insulation as given by:
1 lnr2 / r1  lnr3 / r2  1
Rth2     (5.4)
hi A1 2Lk pipe 2Lkins ho A3
Where k ins is the thermal conductivity of the insulation and A3 is its outside surface area
of insulation.
At a steady state, the rate of heat transfer Q 2 can also be determined by applying
Newton‟s law of cooling:
Q 2  ho A3 T3  To  (5.5)
Where T3 is the temperature of the outside insulation surface. Equation (5.5) can be
used to determine the value of T3 once Q2 has been calculated by rearranging the
equation as follows:
T3  To  Q 2 / ho A3 (5.6)
By taking into consideration the cost of energy and the cost of insulation, the above
analytical model can be used to determine the value of the resulting energy saving and
the required insulation thickness from economic and safety considerations. The
following sections show how the analyses involved lead to problems that need iterative
solutions and how Goal Seek and Solver can be used to solve them.
5.2. Determining the insulation thickness from safety considerations

Safety considerations in the work place require that the surface temperatures of hot
pipes not to exceed a certain limit, e.g., 40o C. In this case, pipe insulation is used for
safety reasons. For a specified thickness of insulation, Equation (5.6) can be used
directly to determine the value of T3 , but if the thickness of insulation is to be
determined for a specified value of T3 , then it needs an iterative solution. The following
Example that shows how Excel‟s Goal Seek command can be used to solve the problem
is based on Example 7-8 given by Cengel [1].
Example 5-1. Determining the thickness of insulation from safety considerations
Hot water at T∞i = 120°C flows in a stainless steel pipe (k pipe = 15 W/m.°C) that has an
inner diameter of 1.6 cm and thickness of 0.2 cm. The pipe is to be insulated with
fiberglass (k ins = 0.038 W/m.°C) so that the temperature of the outer surface T3 does not
exceed 40°C when the ambient temperature T∞o is 25°C. Taking the heat transfer
coefficients inside and outside the pipe to be h i = 70 W/m2 .°C and h o = 20 W/m2 .°C,
respectively, determine the required thickness of insulation.
Solution
Substituting the given values of the three temperatures, T∞i , T∞o , and T3 , the two pipe
radii r1 and r2 , the two thermal conductivities, k pipe and k ins , and the two heat-transfer
coefficients h i and h o in Equations (5.3) and (5.5) leads to:
120  25 40  25 (5.7)

0.284  0.0024  4.188 lnr3 / 0.01  1 / 125.6r3  125.6r3
Although Equation (5.7) has a single unknown, which is r3 , it is nonlinear and requires
an iterative solution. Figure 5.2 shows the Excel sheet prepared for solving the equation
by using the Goal Seek command, while Figure 5.3 shows the formulae used in the
different cells.
Figure 5.2. The Excel sheet developed for Example 5-1

Figure 5.3. The formulae used in the Excel sheet of Example 5-1
For an assumed outside radius of insulation (r3 = 0.01 m), the sheet calculates the rate
of heat transfer from Equation (5.3) as Q2a in cell C6 and from Equation (5.5) as Q2b
in cell C7. The difference, Diff, between Q2a and Q2b is then calculated in cell C8.
Figure 5.2 shows that at this initial value of r3, Q2a exceeds Q2 by 68.913 W. The
value of r3 that makes Diff equal to zero can be found by using the Goal Seek
command. Figure 5.4 shows its dialog box that requires the user to specify the Target
cell, the required value, and the cell to be adjusted so as to reach this goal.
Figure 5.4. The Goal Seek command
In the present case, the target cell is C8, the required value for the target cell is zero,
and the adjustable cell is C1. By completing the form as shown and pressing the “OK”
button, the tool iterates to find the solution shown in Figure 5.5. The value determined
by Goal Seek for r3 , which is 0.17 m, agrees with the solution given by Cengel [1].
Figure 5.5. The solution found by Goal Seek for Example 5-1
There can be other reasons for imposing a lower or upper limit for the surface
temperature of insulation besides the safety consideration. For example, Helmer and
Walker [2] considered a case in which the surface temperature of the insulation material
was not allowed to drop below a certain value in order to prevent condensation of the
air-borne water vapour over the insulation surface. In this case, Goal Seek can be used
to determine the required thickness of insulation such that the surface temperature
doesn‟t fall below the wet-bulb temperature at the local condition.
5.3. Determining the insulation thickness from economic considerations
The thicker the insulation, the lower will be the loss of thermal energy to the
surroundings. However, increasing the thickness of insulation also increases its cost.
Beyond a certain thickness, the cost of insulation can outweigh the benefit due to
energy saving and make the insulation uneconomical. Determining the economically
acceptable thickness of insulation is another type of thermal insulation analyses that
require iterative solutions. The following example, which is based on Problem 7-85 in
Cengel [1], demonstrates the use of Goal Seek for this type of analyses also.
Example 5-2. Determining the insulation thickness from economic considerations

Consider a furnace whose outer surface temperature is measured to be 90°C when the
average surrounding air temperature is 27°C. The management is considering to cover
the furnace, which is 6 m long and 3 m in diameter, with fiberglass insulation (k ins =
0.038 W/m.°C) whose cost is $10/m2 per cm of thickness for materials, plus $30/m2 for
labour regardless of thickness. The furnace uses natural gas whose unit cost is
$0.50/therm input (1 therm = 105,500 kJ), and the efficiency of the furnace is 78 per
cent. The combined heat transfer coefficient on the outer surface is estimated to be h o =
30 W/m2 .°C. The management is willing to authorise the installation of the thickest
insulation (in whole cm) that will pay for itself (materials and labour) in one year.
Determine the thickness of insulation to be used and the money saved per year if the
plant operates 80 hours per week for 52 weeks per year.

Since the surface temperature of the bare furnace (Ts ) is known, the rate of heat transfer
before insulation (Q1 ) can be determined from Equation (5.1). For any insulation
thickness δ, the rate of heat-transfer after insulation (Q2 ) is given by Equation (5.3).
Taking into consideration that the diameter of the furnace is large compared to the
thickness of insulation, the vertical side of the furnace can be treated as a plane wall.
Therefore, the total thermal resistance Rth2 in Equation (5.4) can be approximated by:
1   1
Rth2     (5.8)
As  kins ho 
The annual savings (S) due to the reduction in heat-transfer losses can be determined
from:
S
Q 1 2 
    3600  c
Q f
[$/year] (5.9)
  1000 105500
Where η is the combustion efficiency of the burner, τ is number of operation hours per
year, and cf is the cost of fuel in $ per therm. The total cost of insulation Cins can be
calculated from its two parts, the material cost and the labour cost, as follows:
Cins  As   c1  c2  [$] (5.10)
Where c1 is cost of insulation material in $ per square metre per 1.0 cm thickness, c2 is
the cost of labour per square metre irrespective of thickness, and δ is the thickness of
insulation in cm. In this case, it is required that the total cost of insulation, Cins ,, will not
exceed the drop in the fuel cost of the furnace in one year, which is S.
Figure 5.6 shows the changes in the cost of insulation and the savings in fuel cost as the
thickness of insulation is varied from 1 cm to 15 cm for the furnace considered in
Example 5-2. The figure shows that the cost of insulation increases steady as the
thickness increases, but the savings in fuel cost approaches a constant value of about
$12,000 beyond a thickness of 10 cm. The figue also shows that insulation thickness at
which the cost of insulation equals the annual saving in fuel cost is about 14 cm.
14000
Insulation cost and annual savings

12000
10000
8000
($) 6000
4000
Energy savings
2000
Insulation cost
0
0 0.05 0.1 0.15 0.2
Thickness of insulation (m)
Figure 5.6. Insulation cost and annual saving in fuel cost versus insulation thickness
Goal Seek solution

Figure 5.7 shows the Excel sheet prepared for this example. The given data are shown
on the left side of the sheet with labels used to identify each item. The cells in column E
calculate the surface area of the furnace (As), the rate of heat transfer from the furnace
before insulation (Qbefore) in W, the annual energy loss due to heat transfer (Q1) in kJ,
the annual fuel energy loss due to heat transfer (Q1fuel) in Therm, and the annual cost
of energy loss (AnualCost1) in $. For an assumed insulation thickness (tins) of 0.1 m
(10 cm), the cells in column H calculate the total thermal resistance (Rtotal), and the
rate of heat transfer after insulation (Qins) in W. The cells that follow calculate the
resulting annual energy loss due to heat transfer (Q2) in kJ, the annual fuel energy loss
due to heat transfer (Q2fuel) in Therm, and the annual cost of energy loss (AnualCost2)
in $. The last column to the right determines the annual savings due to insulation
(Savings), the cost of insulation (Costins), and the net annual saving (Diff = Saving-
CostIns) in $.
Figure 5.7. Excel sheet for Example 5-2 with the inserted Goal Seek dialog box
Figure 5.7 shows that the assumed insulation thickness gives an annual saving of
$2815.407. The required insulation thickness that makes the saving equal to zero can be
determined by using Goal Seek. Figure 5.7 shows the completed Goal Seek dialog box
for this task and Figure 5.8 shows the solution found by Goal Seek. The required
insulation thickness, which is 14.04 cm, agrees with that given by Cengel [1].
Figure 5.8. Goal Seek solution to Example 5-2
5.4. Determining the optimum thickness of insulation

The management in Example 5-2 knows that the 14-cm insulation thickness pays for its
cost in one year but lasts for more than five years. Therefore, they consider this a good
investment. However, the 14-cm thickness may not be the optimum thickness which is
the one that maximises the benefit from insulation over its entire lifetime. This
optimum thickness is determined by maximising the annual revenue (R) or minimising
the total annualised cost of insulation (Ctotal ) which are given by:
R  S  Cins   [$] (5.11)
Ctotal  C fuel 2  Cins   [$] (5.12)
Where S and Cins are defined by Equations (5.9) and (5.10), respectively, Cfuel2 is the
cost of fuel after insulation, and a is the rate of amortisation. Both the maximum value
of the annual revenue and the minimum value of the total cost can be determined by
plotting the two parameters over a suitable range of insulation thickness. Figure 5.9
shows the variation of the total cost and revenue with the thickness of insulation in
Example 5-2 as it is increased from 1 cm to 5 cm with 20% amortisation rate.
10900 2100
10800 1900
10700
Total cost ($) 1700
Revenue ($)
10600
1500
10500
1300
10400
10300 Revenue 1100
10200 Total cost 900
10100 700
0 0.02 0.04 0.06
Thickness of insulation (m)
Figure 5.9. Variation of the total annual cost and revenue with the insulation thickness
Figure 5.9 shows that there a certain thickness of insulation, which is around 3 cm, that
minimises the total cost and maximises the annual revenue. Note that this thickness is
considerably less than the value determined in Example 5-2, which is 14 cm. The
optimum thickness of insulation can be determined more precisely by using Solver as
shown in the following example.
Example 5-3. The optimum thickness of insulation for a furnace

For the same thermal conditions and fuel and insulation costs in Example 5-2, find the
optimum thickness of insulation for the furnace and its payback period if the expected
write-off lifetime of the insulation is 5 years and the amortisation rate is 20%.
Solution
Figure 5.10 shows the Excel sheet developed for this example by making some changes
to the sheet developed for Example 5-2. The data part now includes the lifetime of the
insulation, which is 5 years. The last two columns of the sheet have also been modified
to include the calculations of the total annual cost (CostTotal), the annual revenue
(Revenue), and the pay-back period (PayBack) which is given by Cins /S.
For an assumed insulation thickness of 0.1 m (10 cm), Figure 5.10 shows that the total
annual cost (= Cins x α) amounts to $1989.890, while the annual revenue amounts to
$10166.73. To determine the exact optimum thickness of insulation by using Solver,
Figure 5.11 shows Solver Parameters dialog box in which the target cell has been set as
that of the total annualised cost given by Equation (5.12) as shown in the formula bar of
Figure 5.10.
Solver set-up shown in Figure 5.11 requires it to minimise the annualised cost by
adjusting the thickness of insulation (tins). As the figure shows, two constraints have
been added that require (i) the surface temperature to be lower than 40 o C and (ii) the
insulation thickness to be at least 1 cm. Figure 5.12 shows the solution found by using
the GRG Nonlinear method with its default settings.
Figure 5.12. Solver solution of Example 5-3
The optimum diameter determined by Solver is 0.032 m, or 3.2 cm, which close to the
value obtained graphically above. At the optimum thickness, the annual saving is
$11690.05, while the initial cost of insulation is 4363.9. The resulting payback period
for insulation is 0.373 years, or 4.5 months, which is considerably lower than the one-
year period required by the management in Example 5-2. Note that the surface
temperature, which is 29.4o C, is much lower than the 40o C required by safety
considerations. Since the thickness determined by Solver is not a whole number, it has
to be increased to 4 cm. The increase in thickness will allow for future increase in fuel
cost. The effect of possible changes in fuel cost on the optimum insulation thickness
can be investigated by determining the thickness at higher and lower fuel costs.
Example 5-4. The optimum thickness of insulation for a pipe

A steel pipe carries wet steam from a gas-fired boiler through a small workshop to a
process plant. To minimise the rate of heat transfer to the surrounding air, it is proposed
to insulate the pipe using glass fibre insulation with an aluminium alloy casing. The
most economic thickness of insulation is to be determined with the following data:
Pipe outside diameter = 60.3 mm

Heat transfer coefficient for outside surface of insulation = 10 W/m2 .K
Thermal conductivity of insulation = 0.07 W/m.K
Steam temperature = 200o C
Temperature of air in workshop = 15o C
Boiler efficiency = 80%
Price of gas = 0.3 p/MJ
Operation time of plant = 3000 h per annum
Lifetime of insulation = 5 years
Thickness of insulation [mm] 19 25 32 38 50 60

Cost per metre length [p] 476 531 632 763 1007 1280
The case is based on that described by Eastop and Croft [3].

The heat loss through the insulation (Q2 ) is given by Equation (5.3). Neglecting the
thermal resistance to heat transfer by conduction through the pipe and by radiation from
the outer insulation surface, the total thermal resistance becomes that due to heat
transfer by conduction through the insulation and by convection from the surface of
insulation casing to the surrounding air, i.e.:
lnD3 / D2  1
Rth 2   (5.13)
2k hoD3
The total annualised cost (CT), which is the summation of the annual heating cost and
the annualised cost of insulation, is given by:
CT    C i  C h (5.14)
Where Ci is the cost of insulation and α is the amortisation rate, which can be
approximated by the following relationship:
  1/ N (5.15)
Where N is the lifetime of insulation or period after which capital cost is written off.
Equation (5.15) determines α as 20%. The annual cost of heat lost to the surrounding
(Ch ) is given by:
 
Ch  Q2 / 106  c g    3600 / (5.16)
Where cg is the price of gas, η is the boiler efficiency, and τ is the annual operation time
of plant.
The Excel sheet

Figure 5.13 shows the Excel sheet developed for this example in which the problem
data are shown on the left side of the sheet (column B). The calculations start with a
guessed insulation thickness (δ) of to 25 mm. Based on this assumed thickness, the
sheet determines the thermal resistances due to conduction (R_ins) and due to
convection (R_con). From the total resistance (R_total = R_ins + R_con), the sheet
calculates the rate of heat transfer (Q) according to Equation (5.3). From Q and δ, the
sheet determines the annual cost of heat-loss (C_heat) and annualised cost of insulation
(C_ins), respectively. Finally, the sheet determines the total annualised cost (C_total)
for the given insulation thickness according to Equation (5.14).
Figure 5.13. Excel sheet for the optimisation of the insulated pipe in Example 5-4
The rate of heat loss (Q) is calculated as 111.3 W per metre length of the pipe, which
results in an annual cost (C_total) of 450.9 p. Adding the annualised cost of insulation
of 107.9 p, the total cost becomes 558.8 p.
To allow the sheet to determine the cost of insulation automatically for other values of
δ, the following polynomial was obtained from the given price data by using Excel
trendline feature:
Ci  359.42  1.3713  0.2329 2 (5.17)
As shown in Figure 5.13, the above formula is entered in cell H3 that calculates the
initial cost of insulation (ci_ins).
Figure 5.14 shows the variation of the cost of the annualised insulation cost (C_ins),
annual heat-loss (C_heat), and the total annualised cost (C_total) with the thickness of
insulation (δ) as obtained by changing the insulation thickness in the Excel sheet from
0.0 to 60.0 mm. It can be seen from the figure that the total cost has a minimum value at
an insulation thickness which is slightly greater than 40 mm. The optimum insulation
thickness can be determined more accurately by using Solver. Figure 5.15 shows the
set-up for Solver Parameters dialog box for finding the optimum value of δ (in the
adjustable cell) that minimises the total annualised cost C_total (in the target cell).
Based on the given data, two constraints have been added to limit the thickness of
insulation within the range 19 ≤ δ ≤ 60 mm. The solution determined by Solver is
shown in Figure 5.16.
1600
Insulation cost
1400
Fuel cost
1200
Annual cost (p) Total cost
1000
800
600
400
200
0
0 0.02 0.04 0.06 0.08
Insulation thickness (m)
Figure 5.14. Insulation cost against the thickness of insulation
Figure 5.15. Solver set-up for the pipe-insulation optimisation
Figure 5.16 shows that Solver determined the optimum thickness of insulation as 42.8
mm. Since this thickness is not available in the given data, the nearest larger thickness
should be selected; which is 50-mm. The thicker insulation is selected so as to allow for
any possible increase in fuel cost in the future. From a graph similar to that shown in
Figure 5.14, Eastop and Croft [3] also selected 50-mm as the optimum thickness of
insulation.
Figure 5.16. Optimised solution for the insulation thickness in Example 5-4
5.5. Multivariable optimisation of thermal insulation

It was possible in Example 5-3 and Example 5-4 to find the optimum thickness of
insulation by plotting the variation of the total cost with the thickness on a single chart
and picking the thickness that minimises the total cost. However, this is not applicable
when multiple design parameters are involved. The following example demonstrates the
usefulness of Solver in this case. Apart from dealing with multi-variable analyses,
Solver allows various constraints to be imposed on the optimised solution.
Example 5-5. Multivariable optimisation of thermal insulation with Solver

A 2000-m3 cylindrical steal tank is to be designed for storing chilled water at 5o C in an
environment where the average ambient temperature is 30o C. In order to minimise the
rate of heat gain from the surroundings, the vertical surface of the tank is to be insulated
by using fiberglass insulation (k = 0.038 W/m·°C) whose cost is $10/m2 per cm of
thickness for materials, plus $30/m2 for labour regardless of thickness. The top circular
side of the tank is to be insulated by a special material (k = 0.045 W/m·°C) that costs
$5/m2 per cm of thickness plus $20/m2 for labour regardless of thickness. Heat transfer
from the bottom surface is small and therefore, the bottom surface may not be insulated.
The combined heat transfer coefficient on the outer surface is estimated to be h o = 30
W/m2 ·°C. Based on the average cooling load, the refrigeration plant is estimated to
operate for 80 hour per week for 52 weeks per year. If chilling the water costs $0.087
per ton-hour (1 ton-hour = 3.5 kWh), determine the optimum tank diameter and
thicknesses of the two insulation materials taking into consideration that space
constraints limit the maximum tank diameter to 20 m.

In this case, the surface area is not known a priori but has to be determined by the
optimisation analysis itself. Given the tank volume (V), the diameter (D) and height (H)
of the tank are related by:
H  4V / D 2 (5.18)
The areas covered by the two insulation materials on the vertical surface of the tank
(A1 ) and on the horizontal top surface (A2 ) are given by:
A1  DH (5.19)
A2  D 2 / 4 (5.20)
The rate of heat transfer before insulation (Q1 ) is given by Equation (5.1) in which As is
the total surface area of the tank as given by:
As  A1  A2 (5.21)
The rate of heat-transfer after insulation is Q2 as given by Equation (5.3), but the total
resistance Rth2 now becomes:
1
Rth 2  Rins  (5.22)
ho As
Where Rins is the combined thermal resistance due to two parallel resistances related to
the two sections of insulation, which is given by:
R1 R2
Rins  (5.23)
R1  R2
Considering the large diameter of the tank compared to the thickness of insulation, the
vertical side of the tank can be treated as a plane wall. Therefore, the resistances R1 and
R2 are given by:
1
R1  (5.24)
A1kins1
2
R2  (5.25)
A2 kins2
Where δ1 and δ2 represent the thicknesses of the vertical side and top insulation layers,
respectively, and k ins1 and k ins2 are their corresponding thermal conductivities.
The total cost of insulation consists of that of vertical side (Cins1 ) and that of the top
surface (Cins2 ). Each of these two costs consists of material cost (c1 ) and labour cost (c2 )
as determined by Equation (5.10). Therefore, the total annual cost to be minimised is
given by:
Q2
Ctotal   c w    Cins1  Cins2  / N (5.26)
3.5
Where τ is the number of operation hours per year and N is the life time of insulation.
Excel solution
Figure 5.17 shows the Excel sheet prepared for this example. The data part on the left
side of the sheet shows the given information about the inside and outside temperatures
(T∞1 and T∞2 ), the volume of the tank (V), the various thermal and cost constants, the
annual operation time (Time), and the expected lifetime for insulation materials (Life).
Figure 5.17. Excel sheet for Example 5-5
For an assumed diameter of the tank (D) in cell E2, the sheet calculates the tank height
(H), the vertical area (Area1), the top area (Area2), and the total surface area of the tank
(As). It then determines the rate of heat gain before insulation (Qefore) in W, the annual
energy gain (Q1) in kW-h, and the annual added cost of water production (Cwater1) in
$. Similarly, for assumed values of the two insulation thicknesses (tins1 and tins2) in
cells H2 and H3, respectively, the cells in column H determine the rate of heat transfer
(Qins), annual energy gain (Q2), and annual added cost of water production (Cwater2)
after insulation. Finally, the sheet determines the total annual cost (CostTotal), total
annual revenue (Revenue), and payback period (PayBack).
The optimum tank diameter and the thicknesses of the two insulation layers can now be
determined by using Solver. Figure 5.18 shows Solver Parameters dialog box with the
total cost (CostTotal) as the target cell and three changeable cells which store the tank
diameter (D) and the two insulation thicknesses (tins1 and tins2). Three constraints
have been imposed on the solution to keep the tank diameter below a maximum value
of 20 m and the two insulation thicknesses above a minimum of 1 cm. With this setup
and using the GRG Nonlinear method, Solver iterates to find the solution shown in
Figure 5.19. According to this solution, the optimum values for tank diameter and
height are 19.01 m and 6.98 m, respectively. The optimum insulation thicknesses are
tins1 = 2.1 cm and tins2 = 3.2 cm. The payback period is 0.61 years.
Figure 5.19. Solver setup for Example 5-5
Note that the optimised insulation thicknesses determined by Solver are not whole
numbers. The nearest whole numbers for tins1 and tins2 are 3 cm and 4 cm,
respectively. By replacing the two guessed insulation thicknesses with these modified
thicknesses, Solver can be used to find the relevant optimum tank diameter and height.
5.6. Closure
The analyses of thermal insulation using the concept of thermal resistance require
iterative solutions when the objective is to determine the required thickness of
insulation for safety or economic considerations. The selection of the insulation‟s
material and thickness may also involve factors other than the safety and the economic
factors. This chapter showed how Excel and its two iterative tools, Goal Seek and
Solver, can be used in the analyses and optimisation of insulation systems that involve
single and multi-variable design parameters. Appendix D also presents a multi-variable
optimisation analysis that deals with an air-conditioning duct with two sections to be
insulated.
References
[1] Y. A. Cengel, HEAT TRANSFER, A Practical Approach, Second Edition,
McGraw Hill, 2007.
[2] W. Helmer and R. Walker, Engineering ethics cases with numerical problems,
Available at; http://ethics.tamu.edu/wp-content/uploads/ sites/7/2017/04/ heat-
transfer-case.pdf
[3] T.D. Eastop and D.R. Croft, Energy Efficiency for Engineers and Technologists,
Longman Scientific & Technical, 1990, Example 5-1.
Determination of the economic pipe

diameter
The pump-pipe systems used in oil and gas pipelines require various types of auxiliary
equipment for supporting and protecting the pipe system and for flow measurement and
control. The large pumps needed in these systems require a continuous supply of
electrical energy in order to overcome friction in the pipes, bends, and valves. Since
these systems are not only costly to install but also to operate, their optimisation
analyses involve both hydraulic and economic considerations. Before computers and
computer software became widely available, the engineers used an analytical method
for selecting the optimum diameter that involves arduous calculus manipulations. For
turbulent flows, the analytical method requires the use of dimensionless charts in order
to avoid iterative solutions. This chapter initially introduces the analytical method for
pipe diameter optimisation and highlights its limitations before presenting an Excel-
aided method by using Solver and demonstrating its generality and simplicity.
6.1. The objective function for pipe -diameter optimisation

The first step in any optimisation process is the formulation of an objective function for
the optimisation. The objective of optimisation for a pipeline or any industrial pump-
pipe system is to minimise the total cost of the system over its entire lifetime. For these
systems, the total cost includes three components: (i) the installation cost (ii) the
operation or running cost, and (iii) the maintenance cost. Following the method
described by Janna [1], the total annualised cost (CT) of a pump-pipe system is given by
the following equation:
CT  Lc PF  cm   COP (6.1)
Where, L is the pipe length, cPF the annualised installation cost of the pipes, fittings, and
pumps per metre-length of the pipe, cm the annual maintenance cost fittings and pumps
per metre-length of the pipe, and COP the annual cost of operation. How the three costs
are evaluated is described below.
The installation cost of a pump-pipe system

Standard pipe sizes are designated by two numbers that refer to the pipe‟s nominal
diameter and schedule, e.g., a 6-nominal and schedule 40 pipe. The nominal size is
associated with the outside diameter of the pipe, but not necessarily equal to it, while
the pipe‟s schedule refers to its thickness. Table 6.1 lists some standard pipes with their
outside and inside diameters. Schedule 40 is the standard for sizes up to nominal 12,
while for larger sizes the standard schedule is 30 or 20. Pipe costs depend on their
material and manufacturing process, nominal diameter, and schedule. Table 6.2 shows
the costs per linear metre (cP ) in $/m for commonly used pipes made of PVC [2], two
grades of stainless-steel pipes [3], and two grades of copper tubes [4].
The cost of fittings and pumps (cF ) is taken as a fraction, F, of the pipe cost:
cF  F  cP (6.2)
Table 6.1. Selected standard-size pipes (adapted from Janna [1])

Nominal Do (in) Do (cm) Di (cm) Nominal Do (in) Do (cm) Di (cm)
diameter diameter
½ 0.840 2.134 1.580 12 12.750 32.39 30.33
¾ 1.050 2.667 2.093 14 14.000 35.57 33.65
1 1.315 3.340 2.664 16 16.000 40.64 37.73
1¼ 1.660 4.216 3.504 18 18.000 45.72 43.81
1½ 1.900 4.826 4.090 20 20.000 50.80 48.89
2 2.375 6.034 5.252 22 22.000 55.88 53.97
2½ 2.875 7.303 6.271 24 24.000 60.96 59.05
3 3.500 8.890 7.792 26 26.000 66.04 64.13
4 4.500 11.43 10.23 30 30.000 76.20 74.29
6 6.625 16.83 15.41 32 32.000 81.28 79.34
8 8.625 21.91 20.27 34 34.000 86.36 84.45
10 10.750 27.31 25.46 36 36.000 91.44 89.53
Table 6.2. Costs of PVC and stainless-steel pipes and copper tubes per linear metre ($)
Nominal PVC [2] Stainless Stainless Copper Copper
diameter steel 304 [3] steel 316 [3] type L [4] type M [4]
¼ 4.59
⅜ 6.69 5.22
½ 1.94 5.64 8.33 8.10 5.94
¾ 2.36 6.95 9.28 13.15 9.64
1 3.25 9.77 11.32 18.76 14.66
1¼ 4.00 13.58 16.24 25.85 21.32
1½ 4.66 15.48 20.37 33.26 29.39
2 6.26 21.88 24.99 53.46 46.31
2½ 10.00 33.10 40.84 76.00 67.17
3 12.66 43.43 51.53 101.38 89.31
4 17.71 65.80 81.25 180.66 167.25
6 32.41 106.40 133.99 419.87
8 43.39 184.30 218.78 791.27
10 60.98 267.55 287.75
12 80.23 321.05 340.86
14 123.82
16 156.06
18 235.77
20 275.95
24 396.65
Therefore, the total annualised cost of the pipe, fittings and pumps per metre-length of
the pipe (cPF ) is then given by [1]:
 
cPF   c p  cF   1  F cP (6.3)
Where a is the amortisation rate which depends on the annual rate of interest (i) and the
number of payment years (N) according to the following relationship:
  1 N 
a  i  1     (6.4)
  1  i  
For sufficiently large values of N, the following approximate relationship can be used:
α = 1/N (6.5)
The maintenance cost

The annual maintenance cost (cm) per metre-length of the installed pipe system is taken
as a fraction, b, of the cost of the pipe, fittings and pumps [1], i.e.
cm  bcP  cF   b1  F c p (6.6)
The operation cost

The operation cost is mainly the cost of electrical energy needed by the pumps to
overcome friction losses in the pipes and fittings as well as the elevation differences in
the system. The power needed to overcome friction losses is given by:
W f  gQh f /  [W] (6.7)
Where, ρ is the fluid density, g the gravitational acceleration, Q the volume flow rate
through the pipe, h f the head loss caused by friction in the pipe including minor losses,
and η the combined pump-motor efficiency. In terms of mass flow rate ( m   Q ), the
equation becomes:
W f  m gh f /  [W] (6.8)
The power required for overcoming the difference in elevation between the inlet and
exit of the pipe is given by:
W z  m g Z 2  Z1  /  [W] (6.9)
Where Z1 and Z2 are the elevations at the inlet and outlet points, respectively. The total
annual operation cost is then given by:
 
COP  W f  W z c2 / 1000 (6.10)
Where c2 is the cost of electricity per kilowatt-hour ($/kW-h) and τ is the total number
of operation hour per year. Substituting for W f and W z from Equation (6.8) and
Equation (6.9), respectively, the total annual operation cost becomes:
COP 
h  Z
f 2 
 Z1  m gc2
(6.11)
1000
The total annualised cost of the pipeline

Substitution of the installation, maintenance, and operation costs in Equation (6.1)
yields the following equation for the total annualised cost of a pipeline:
CT    b 1  F LcP 
h  Z
f 2 
 Z1  m
 gc2
(6.12)
1000
Equation (6.12) is the required objective function for least-cost pipeline optimisation.
The equation can be used to determine the diameter at which the total pipeline cost is
minimal by using either analytical or computer-aided methods. The analytical method
described in the following section is based on that due to Janna [1].
6.2. The analytical method for pipe -diameter optimisation

The optimum pipe diameter can be obtained analytically by differentiating Equation
(6.12) with respect to the diameter (D) and equating the derivative to zero, i.e.:
CT c p  gc2 h f
   b 1  F L
m
 0 (6.13)
D D 1000 D
It should be noted that the difference in elevation (Z2 – Z1 ) does not appear as a
parameter in Equation (6.13) and, therefore, does not affect the value of the optimum
diameter (Dopt ). The solution of Equation (6.13) yields the optimum diameter, but
before the differentials of cP and h f , can be determined the analytical relationships for
these two parameters in terms of the diameter D have to be provided.
An easily differentiable relationship between the pipe cost cP and its diameter D is
given by the following power function:
c P  c1 D n (6.14)
Where c1 and n are constants the values of which have to be determined by curve-fitting
the pipe cost data. Figure 6.1 shows the power functions for the pipe materials given in
Table 6.2 obtained by plotting the data on an Excel chart and using Excel‟s trendline
feature to fit the data. Table 6.3 shows the corresponding values of c1 and n for the five
pipe types shown in Figure 6.1.
900.0
PVC Steel 304
800.0
y= 8077.8x 1.656 Steel 316 Copper "L"
700.0 Copper "M"
600.0
Pipe cost ($/m)
500.0
400.0 y = 1682.5x 1.4234 y = 637.85x 1.4951

300.0 y = 1721.6x 1.3589
y = 6763.8x1.6824
200.0
100.0
0.0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Inside diameter (m)
Figure 6.1. A plot of the pipe-cost data given in Table 6.2
Table 6.3. Values of the two constants c1 and n in Equation (6.14) for different pipe
materials
Pipe material c1 n
PVC 637.9 1.495
Stainless steel 304 1682.5 1.423
Stainless steel 316 1721.6 1.358
Copper (type L) 8077.8 1.656
Copper (type M) 6763.8 1.682
Substituting for cp from Equation (6.14) and differentiating with respect to the pipe
diameter, Equation (6.13) becomes:
 gc2 h f
n  b 1  F c1 D n1 L 
m
0 (6.15)
1000 D
Further progress with the analytical optimisation method depends on the method used
to evaluate the pipe friction h f . The following sections consider two options: (i) the
Hazen-Williams equation and (ii) the Darcy-Weisbach equation.

In the design of water pipe-systems, the pipe head loss is usually determined by using
the following Hazen-Williams equation [5]:
10.67 LQ1.852
hf  (6.16)
C 1.852 D 4.8704
Where C is a roughness coefficient that depends on the pipe‟s material and surface
condition. Typical C values for selected pipe materials are shown in Table 6.4.
Table 6.4. C-values in the Hazen-Williams equation for selected pipe materials [5]
Pipe C-value
Cast-iron (new) 130
Cast-iron (10 years) 107 – 113
Galvanised iron 120
Steel 90 – 110
Copper 130 – 140
Concrete 100-140
PVC 140
Equation (6.16) provides the necessary function for the friction loss in terms of the pipe
diameter. Substituting for h f from Equation (6.16) in Equation (6.15) yields:
10.67m gLQ1.852 c2

n  b 1  F c1 Dopt
n 1
L  4.8704 0 (6.17)
1000 C 1.852 Dopt
5.8704

Rearranging:
n  4.8704 51.9672m gQ1.852 c2 / 1000

Dopt  (6.18)
n  b 1  F c1C 1.852
The absence of the pipe length, L, in Equation (6.18) means that it also does not affect
the optimum diameter. By taking the density of water as 1000 kg/m3 , the equation can
be simplified as follows:
1
 509.7982Q 2.852 c2  n4.8704
Dopt  1.852 
(6.19)
 n  b 1  F c1C  
The following example illustrates the use of Equation (6.19) for determining the
optimum diameter with the Hazen-Williams equation.
Example 6-1. Optimum pipe diameter using the Hazen-Williams equation

A schedule 40 commercial steel pipe is used to convey water to a heat exchanger. The
flow rate of water is 1 kg/s. Calculate the optimum economic diameter and the
corresponding water velocity based on the following data:
c1 = $750/m2.2 , n = 1.2, b = 0.01, F = 7.0, N = 7 years, c2 = $0.04/kW.hr, τ = 7880 hr/yr

η = 75%
Solution
For commercial steel C = 90
m
The volume flow rate Q   0.001 m3 /s.

The amortisation rate α = 1/7 = 0.143
Substituting in Equation (6.19) gives:
1
 509.7982  0.0012.852  0.04  7880  1.2 4.8704
Dopt  
1.20.143  0.011  7   750  90  0.75 
1.852
= 0.022 m or 2.2 cm
The water velocity at the optimum diameter is calculated as follows:
Q 4Q 4  0.001
Vopt     2.6 m/s
Aopt Dopt   0.0222
2
Note that the water velocity is within the recommended range which is between 1.4 and
2.8 m/s [3]. The velocity can be used to determine the type of flow in the pipe. Taking
the absolute viscosity of water as 8.90x10-04 Pa.s, the corresponding Reynolds number
is:
0.022  2.6
Re  = 64,589.91
8.9  10 4
The value of the Reynolds number indicates that the flow is turbulent. Apart from
avoiding an iterative solution, an important advantage of Equation (6.19) is that it can
be used to determine the optimum diameter for both laminar and turbulent water flows.
Referring to the data shown in Table 6.1, we realise that the obtained optimum diameter
falls between two standard pipe diameters; which are ¾ inch and 1 inch. As a rule-of-
thump, it is recommended to select the pipe with the larger nominal diameter, which is
the 1-inch pipe.
6.2.2. Optimisation with the Darcy-Weisbach equation (laminar flow)

The Darcy-Weisbach equation is advantageous over the Hazen-Williams equation
because it can be used for various fluids including water. However, the friction factor
depends on the Reynolds number, which in turn depends on the pipe diameter. For a
laminar flow, the linear relationship between the friction factor and the Reynolds
number enables a closed-form solution to be obtained as shown below.
Using the Darcy-Weisbach equation, the pipe friction h f in Equation (6.15) can be
obtained from:
L V2 8Lm 2
hf  f  f (6.20)
D 2g g 2 2 D 5
For a laminar flow, the friction factor is given by the following linear relationship:
64
f  (6.21)
Re
Where Re is the Reynolds number ( Re  4m / D ). Substituting in Equation (6.20):
 64D  8Lm  128Lm

2
hf    2 2 5  = (6.22)
 4m  g  D  g 2D 4
Differentiating h f with respect to the diameter, D:
h f   128Lm  4  128Lm
  2 4 =  (6.23)
D 
D  g D   g 2D 5
Substituting from Equation (6.23) in Equation (6.15) we get:
m gc2  4 128Lm 2 
n  b 1  F c1 Dopt
n 1
L  0 (6.24)
1000  g 2D 5 
Rearranging:
n4


0.512  m 2 c2 
Dopt

n  b 1  F c1  2  (6.25)
Or:
 
1
 0.512  m 2 c 2  n4

 
Dopt

 n  b 1  F c1 
2
 

(6.26)
Equation (6.26) can be used to determine the optimum diameter of the pipe if all the
required data are available. It should be noted that the optimum diameter determined by
Equation (6.26) depends on the density as well as the viscosity of the fluid. Therefore,
this equation can take into consideration the effect of temperature on the fluid‟s
viscosity. The following example illustrates the procedure of using the equation.
Example 6-2. Optimum diameter for a laminar flow

Glycerine (ρ = 1252 kg/m3 , µ= 0.27 Pa.s) is to be pumped from a storage tank to a
bottling machine at a rate of 0.34 l/s via a 10-m long schedule-40 pipe made of stainless
steel 304. The pipe‟s connection to the bottling machine is 1 m higher than its inlet
from the storage tank. Determine the optimum diameter for the pipe if the other rated
data is as follows:
c2 = $0.12/kW.hr, F = 6.75, N = 10 years, η = 75%, τ = 7000 hr/yr, b = 0.01
Solution
From Table 6.3, for stainless steel 304, c1 = $1682/m2.423 and n = 1.423.
a = 1/N = 0.1
Q = 0.34/1000 = 0.00034 m3 /s
m  Q  1252 0.00034  0.42568 kg/s
VD 4m 4  0.42568 2.00738

Re    
 D   0.27  D D
Since the numerator in the above equation (≈ 2.0) is small, we assume that the flow is
laminar and use Equation (6.26) to calculate the optimum diameter:
1
 0.512  0.27  0.42568  0.12  7000  1.423  4
2
Dopt  2
1.4230.1  0.011  6.75  1682 0.75    1252 
= 0.02646 m or 2.65 cm.
We can now check the validity of our assumption that the flow is laminar by calculating
the Reynolds number using the value obtained for the diameter:
2.00738
Re   75.86
0.02646
The small value of Re confirms that the flow is laminar.
6.2.3. Optimisation with the Darcy-Weisbach equation (turbulent flow)

For turbulent flows, the relationship between the friction factor and the Reynolds
number is generally nonlinear. This makes it difficult to use the procedure for the
laminar flow described above in order to obtain an explicit formula for Dopt like
Equation (6.26). However, analytical solutions could be obtained for two limiting cases:
(i) for smooth pipes and (ii) for pipes with complete turbulence. For smooth pipes the
friction factor depends on the Reynolds number only and not on the pipe‟s roughness,
while for pipes with complete turbulence the friction factor depends on the pipe
roughness only and not on the Reynolds number. The earliest equation developed for
pipe diameter optimisation with turbulent flows in smooth pips was the Genereaux
equation [6], while Genic et al [7] described an equation that can be used for pipes with
complete turbulence.
Treating f as constant in Equation (6.20) and differentiating h f with respect to D gives:
h f  40Lm 2
 f (6.27)
D g 2 2 D 6
Substituting in Equation (6.15) and rearranging gives:
n 5 40 f m 3 c  / 1000
Dopt   2 2 2 (6.28)
n  b 1  F  c1 
Or,
1
 0.04 f m 3 c 2  n 5
Dopt   2 2
(6.29)
 n  b 1  F  c1  
Equation (6.29) can be used to determine Dopt by using a trial-and-error procedure. An

initial guess for Dopt is made based on which the Reynolds number and friction factor
are evaluated. When substituted in the right-hand side of Equation (6.29), the result
should be equal to Dopt itself. If this is not the case, a new value for Dopt is tried until a
reasonably close value is found. To avoid the trial-and-error process, Janna [1]
describes dimensionless graphs that directly determine the friction factor, for a
specified value of n, from two dimensionless numbers π 1 and π 2 defined as:
1/ 6
 128 m 2  4m  n  n  b 1  F c  2 

 1  f . Re 
n 5 1 / 6
  3 5 
  
 

1 
 (6.30)
 5    c 2 / 1000 
 /D   D  
2     (6.31)
Re   4m
D  4m 
The second dimensionless number, π 2 , is called the roughness number (Ro). Figure 6.2
shows the graph which is applicable for the case n = 1.2. Janna [1] provides two more
graphs for n = 1.0 and n=1.4. To use these graphs, the two dimensionless numbers π 1
and π 2 have to be calculated from the given problem data and used to read the
corresponding friction factor off the respective graph. The value of f thus determined
can then be substituted in Equation (6.29) to determine Dopt .
Figure 6.2. Friction factor graph for n = 1.2 (adopted from Janna [1])
The dimensionless graphs like the one shown in Figure 6.2 enable Dopt to be determined
without a trial-and-error process, but each graph is only useful for a specific value of n.
Since n differs with the pipe materials as shown in Table 6.3, a large number of such
graphs would be required to cater for all possible values of n. Th approach adopted here
is to generate a single graph that enables Figure 6.2 to be used for other values of n. The
new graph, shown in Figure 6.3, has been developed with Excel by following a similar
procedure to that described by Alciatore and Janna [8] for preparing Figure 6.2. For a
better accuracy, the following Cheng formula was used for calculating the friction
factor [1]:
2
 
  5.0452  1   
1.1098
5.8506  

f   2.0 log  log   (6.32)
  2.8257  D  Re 0.8981 
 
3.7065D Re 
 
12
11 n=1.0 n=1.4
10 n=1.6 n=1.8
9 n=2.0
8
Correction factor S
7
6
5
4
3
2
1
0
1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08
π1
Figure 6.3.Adjustment factor S for reading the friction factor from the graph
in Figure 6.2
With this graph, the friction factor can be read from Figure 6.2 at the same values of π 2 ,
but with an adjusted value, π / 1 , which is calculated from the following relationship:
1
 1/  (6.33)
S
Where S is a correction factor that is read from Figure 6.3 given the value of n. With the
friction factor thus determined, Equation (6.29) can be used to calculate Dopt following
the same procedure described earlier. The accuracy of this procedure can be checked by
recalculating the friction factor based on the obtained value of Dopt and comparing its
value with that obtained from Figure 6.3 as shown in the following example.
Example 6-3. Optimum diameter for a turbulent flow

A type-M copper tube is to be sized for an air conditioner that uses Freon-22 (ρ= 1197
kg/m3 , µ=3.51x10-4 N.s/m2 ). The flow rate is 1.0 l/s in the portion of the system that is
to be selected. Determine the optimum economic diameter for the tubing where liquid
refrigerant is being conveyed, given the following information:
c2 = $0.12/kW.hr, F = 7.0, Life time = 8 years, η = 62%, τ = 4500 hr/yr, b = 0.01
Solution
From Table 6.1, for copper tube ε = 0.0015 cm, or 0.0000015 m. From Table 6.3, for
type M copper tube c1 = $6763/m2.682 , n = 1.682
α = 1/8 = 0.125
Q  1.0 / 1000  0.001m3 /s
m  Q  1197  0.001  1.197 kg/s
4m 4 1.197 4.342103

Re   
D   3.5110 4  D D
Note the large value of the numerator in the Reynolds number indicates that the flow is
turbulent. Substitution of the above data in Equation (6.29) gives:
1
 0.04  f  1.1973 0.12  4500  1.682 5
Dopt   2
= 0.0383 f 0.15
1.6820.125  0.011  7  6763 0.62    1197 

2
The two dimensionless quantities π 1 and Ro are now calculated as follows:
 4  1.197   1.6821 / 8  0.011  7   6763 0.62  1000

1/ 6
128 1.197 2
1.2 2

1  3   

5 3.51  10  4  5
   3.51  10  
4
0.12 / 1000  4500 

= 4.25047x105
  0.0000015 3.51 10 4
Ro = = 3.455x10-10
4  1.197
The value of the friction factor can be determined by using the graph in Figure 6.2 at
the same value of Ro but an adjusted value of π 1 given by Equation (6.33). The value of
S from Figure 6.4 at n = 1.67 and π1 = 4.25x105 is approximately 3.0. Therefore, the
adjusted value π / 1 = 4.25047x105 /3.0, or 1.41683 x105 .
12
11 n=1.0 n=1.4
10 n=1.6 n=1.8
9 n=2.0
8
Correction factor S
7
6
5
4
3
2
1
0
1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 1.E+08
π1
Figure 6.4. Determining the correction factor S for Example 6-3
At π / 1 = 1.41683x105 and Ro = 3.455x10-10 Figure 6.2 gives f ≈ 0.016. Therefore, the

optimum diameter is:
Dopt  0.0383 f 0.15 = 0.0383x(0.016)0.15 = 0.02065 m or 2.065 cm
We can now determine the Reynolds number using the calculated value of the optimum
diameter of Dopt and check the friction factor obtained from Figure 6.2.
Re  4.34  103 / 0.02065 = 210,248.77
The value of Re confirms that the flow is turbulent. Using Equation (6.32), the friction
factor is:
= 0.0161
This value of f is reasonably close to the value found earlier from Figure 6.2.
6.3. An Excel-Solver method for pipe-diameter optimisation

The analytical method for determining the economic pipe diameter suffers from two
drawbacks the first of which is that it does not provide a general methodology for
dealing with both laminar and turbulent flows. Although this can be avoided by using
the Goal Seek command or Solver to perform the iterative solution required by the
turbulent flow, the analytical method still suffers from the second drawback which is
the need to use the pipe cost-diameter power relationship given by Equation (6.14)
which may not accurately represent the pipe-cost data. Figure 6.5 compares the pipe
unit price (cpo ) given in Table 6.1 for copper tubes (type M) with the estimations given
by Equation (6.14). The figure shows that the power-equation deviates from the actual
data particulalry at large diameters.
180
160
Interpl1
140
function
Pipe cost ($/m)
120
100
80
Equation (11.14)
60
y = 6763.8x 1.6824
40
20
0
0 0.05 0.1 0.15
Internal diameter (m)
Figure 6.5. Representations of the cost data for type-M copper tube
In general, determining the optimum pipe diameter by minimising the total cost does
not require the use of the power relationship which is only needed for convenience of
applying the analytical method. By using Excel, the total cost can easily be minimised
with Solver. The Excel-Solver method described in this section enables the pipe-cost
relationship to be represented by a more accurate function than the power function
either by using Excel‟s “trendline” feature to obtain a suitable polynomial or by using
the interpolation function “Interpl1” provided by the Thermax add-in. Unlike Equation
(6.14), Figure 6.5 shows that the interpolation function returns the exact values given in
the table. The use of the interpolation function is explained in Appendix C.
To illustrate the Excel-Solver method, let us go back to the pump-pipe system in

Example 6-3 and determine its optimum diameter by using Equation (6.12) instead of
Equation (6.29). The Excel sheet developed for this purpose is shown in Figure 6.6.
The given data are stored on the left part of the sheet. The Excel formula that calculates
the friction factor enables the sheet to be used for both laminar and turbulent flows. For
turbulent flows, the sheet uses the Swamee-Jain equation for the friction factor, but a
more accurate equation, such as the Cheng equation, can also be used. In order to
determine the pipe cost by the linear interpolation function, the pipe-cost data is stored
in Sheet2 of the same workbook as shown in Figure 6.7.
Figure 6.7. Cost data for the type-M copper tube in Example 6-3
The formula bar in Figure 6.6 shows how the interpolation function “Interpl1” is used
to determine the unit cost from the piper diameter. The formula is entered as follows:
cpo =Interpl1(D;Sheet2!D3:D12;Sheet2!E3:E12;9)
Based on the guessed pipe diameter of 0.04 m, the sheet calculates the Reynolds
number (Re), friction factor (f), friction losses (hf), and the power needed to overcome
friction losses (W). It then determines the pipe unit cost (cpo), annualised cost of the
pipe, pump and fittings, and annual maintenance cost (C_PT), and cost of operation
(C_op) which are stored in cells L4 – L6. The total annualised cost (C_T) in cell L8 is
the summation of C_PT and C_op. Figure 6.6 shows that the total cost for the guessed
pipe diameter of 0.04 m is about $30.55, which is not the minimum cost. Solver‟s set-
up for determining the diameter that minimises the total cost is shown in Figure 6.8.
Two constraints have been imposed for the lower and upper values of the diameter.
With this set-up, the GRG Nonlinear method of Solver failed when automatic-scaling
was used. Figure 6.9 shows the solution obtained without automatic-scaling, which is
2.093 cm. Although this solution is more accurate than the 2.065-cm diameter obtained
in Example 6-3 by using the analytical power function, the difference in not significant
since both methods led to a nominal diameter of ¾ inch.
Figure 6.9. Solver solution of Example 6-3
The developed Excel sheet can also be utilised to assess the sensitivity of the optimum
diameter to possible changes in the costs involved. By adjusting the pipe diameter in the
sheet shown in Figure 6.6, the total cost was calculated at various diameters and plotted
in Figure 6.10. This figure reveals that a pipe internal diameter of up to 2.7 cm can be
used without substantially increasing the total cost of the system. Therefore, it is wise to
select a 1-in nominal-diameter pipe so as to cater for future increases in the operating
costs.
140
C_PF
120
C_op
100 C_T
Cost ($/m)
80
60
40
20
0
0 0.01 0.02 0.03 0.04 0.05
Internal diameter (m)
Figure 6.10. Variation of pipe costs with diameter
6.4. Advantages of the Excel-Solver method over standard programming

As an educational platform for optimisation analyses, the Excel-based platform is more
effective compared to the models developed for the same purpose using standard
programming languages. This is because it gives the ability to develop the model with
less programming effort, to use Solver with its alternative solution options, and to use
user-defined functions within the model for fluid-properties and nonlinear equations.
For illustration, comparison will be made in this section with the model developed by
Akintola and Giwa [9] who used Microsoft Visual C++ to write a computer program for
determining the economic pipeline diameter. Although their computer code included
more than 80 lines, they still had to determine the optimum pipe diameter by visual
inspection of the variation of the total annualised cost with the diameter. They also had
to determine the fanning friction factor (f n ) from the following equation:
0.04
fn  (6.34)
Re 0.16
Since the equation does not involve the pipe roughness (ε), it is only suitable for smooth
pipes or for a particular pipe material. The following example, which is based on their
study-case, illustrates the convenience and effectiveness of the Excel-Solver method.
Some of the original data have been modified for generality. While the pipe cost unit
has been converted from Nigerian Naira (N) per 1 mm length to US dollar ($) per 1 m
length, the electricity cost per kW-hr has been changed from Nigerian Naira to a
slightly higher value in US dollar.
Example 6-4. Selection of optimum pipe size for a turbulent flow

A pipeline is to be made of steel pipes (ε = 0.046 mm) for transporting crude oil (ρ =
880 kg/m3 , μ = 1.1 x 10-3 Pa.s). It is required to determine the optimum pipe diameter of
the pipeline based on the following data:
c1 = $500 per metre length, c2 = $0.05 per kW.hr, n = 1.5, F = 2.25, α =10%, b = 5%,
τ = 8000 hr/yr, η = 60%.
Solution
Figure 6.11 shows the Excel sheet developed for this example by making some
modification to the sheet developed for Example 6-3. The calculations part start with an
assumed value for the pipe diameter (D) of 0.03 m. Based on this diameter, the sheet
calculates the velocity and Reynolds number in the pipe and the friction factor (f) using
Equation (1.24) and then determines the head loss (hf) and required power (Wf).
The annualised initial cost (C_PT), the operation cost (C_op), and the total annualised
cost (C_T) are calculated according to corresponding terms in Equation (6.12). At the
specified diameter, the values of these costs are $1.267, $644.391, and $645.657,
respectively. The pipe diameter that yields the minimum total cost can now be
determined by using Solver with the same set-up shown in Figure 6.8. Figure 6.12 that
shows the solution found with the GRG Nonlinear method shows that the optimum
diameter is 0.092 m at which the annualised initial cost (C_PT), the operation cost
(C_op), and the total annualised cost (C_T) per metre length of the pipeline are $6.844,
$2.054, and $8.898, respectively. The annual energy cost obtained by Akintola and
Giwa [9] was significantly lower than the present value. This could be due to the lower
friction losses estimated by their formula for the friction factor. At the same time their
annualised initial cost was higher than the present estimation; which explains why their
optimum diameter, which was about 0.034 m, was much lower than the present value.
6.5. Closure
This chapter dealt with the economic optimisation of pipeline diameters by using both
analytical and computer-aided optimisation methods. With the analytical method that
obtains the optimum pipe diameter by differentiating the total annualised-cost an
equation for the optimum pipe diameter could be obtained for water-transporting pipes
by using the Hazen-Williams equation to calculate friction losses for both laminar and
turbulent flows. However, by using the more general Darcy-Weisbach equation an
explicit equation for the optimum pipeline diameter could only be obtained for a
laminar flow. For a turbulent flow, the analytical method determines the optimum
diameter by using dimensional graphs in order to avoid the iterative solution involved.
The chapter demonstrated the generality of the computer-aided method that uses Excel
and Solver to solve the basic objective function for optimisation with the Darcy-
Weisbach equation. The analytical needs an analytic function for the price of standard
pipe sizes in terms of the diameter. By replacing the power function used in the
analytical method with a direct interpolation of the diameter-cost data, the Excel-Solver
method provides more accurate estimations than the analytical method. Finally, the
chapter demonstrated the advantages of the Excel-Solver method compared to standard-
programming models for optimisation analyses.
References
[1] W.S. Janna, Design of Fluid Thermal Systems, third edition, CENGAGE Learning,
2011
[2] USP United States Plastic Corporation, http://www.usplastic.com/ Last accessed
January 31, 2018.
[3] British Metals LLC. Brismet, http://brismet.com/ Last accessed January 31, 2018.
[4] Copper tube:www.camlee.com/wp-content/uploads/2013/06/CPRICE.pdf, Last
accessed January 31, 2018.
[5] Wikipedia contributors, https://en.wikipedia.org/wiki/Hazen%E2%80%93
Williamsequation, Last accessed January 31, 2018.
[6] R.P. Genereaux, Fluid-flow design methods, Industrial and Engineering

Chemistry, 29 (1937) 385-388.
[7] S.B. Genic, B.M. Jacimovic, and V.B. Benic, Economic optimization of pipe
diameter for complete turbulence. Energy and Buildings, 45 (2012), 335-338.
[8] D. Alciator and W.S. Janna, Modified pipe friction diagrams that eliminate trial-
and-error solutions, Proc. Of the National Fluid Dynamics Congress, Part 2, July,
1988, available at: https://www.engr.colostate.edu/~dga/dga/papers/pipefriction.
pdf. Last accessed January 31, 2018.
[9] T.A. Akintola and S.O. Giwa, Optimum pipe size selection for turbulent flow,
Leonardo Journal of Sciences, Issue 14, January-June 2009, 112- 123.
Excel-based design procedures for

pump-pipe systems and double-pipe
heat-exchangers
The design of fluid-thermal systems involves a number of interdependent design steps

such as selecting the suitable pipe diameter and the suitable pump size. Therefore,
design analyses of fluid-thermal systems clearly refect the iterative nature of the design
process. Design analyses of fluid-thermal systems also demonstrate the need for
computers and computer-based methods in the design process for both improving its
accuracy and minimizing the effort. The aim of this chapter is to show how the Excel-
based platform can be used for design analyses of two common types of fluid-thermal
systems which are pump-pipe systems and a double-pipe heat exchanger. Since design
calculations involve a number of steps that cannot be accommodated in a single Excel
sheet, these require a workbook.
7.1. Pump-pipe systems

The following procedure for designing a pump-pipe system follows that described by
Janna [1] but takes advantage of the Excel-based platform for its application. In general,
given the required fluid flow rate, the layout of the pipe system, and the various costs
involved, the design procedure goes as follows:
1. Determine the optimum pipe diameter and select a suitable standard pipe size.
2. Select the suitable pump (its type, impeller size, and motor-speed).
3. Determine the operating point for the pump-pipe system and check whether the
discharge provided by the pump meets the required value.
4. Check for cavitation and vibration (noise) limits.
The case considered for applying the design procedure for pump-pipe systems is based
on a design case given by Janna [1].

The pump and piping system shown in Figure 7.1 is to move ethylene glycol from the
lower tank to the upper tank. The pipe is all wrought iron, with threaded fittings,
schedule 40. Pipe lengths are shown in Table 7.1, Z1 = 1.22 m, and Z2 = 4.0 m. Select
a suitable pipe size and a pump for a flow rate of 50 gpm of ethylene glycol based on
the costs provided below:
Cost of the wrought iron pipe (Cpipe) can be taken as $40.0 per cm diameter per 1 m
length.
Cost of centrifugal pumps in dollar can be approximated by the following formula:
C pump  200W pump  150 (7.1)
Where W pump is the power needed to run the pump in horsepower.
Cost of electricity is $0.08 per KW-h.

Figure 7.1.
A: Globe valves (2 of them)

B: Union fittings (2 of them)
C: Plate welded to tank to suppress air entrainment in pump inlet
D: Elbows: there are five 90° elbows and two 45° elbows
Table 7.1. Pipe lengths

Pipe Length (m)
1 0.2
2 0.5
3 0.3
4 0.3
5 1.2
6 2
7 4.5
8 7
9 2
10 2
11 2
12 5
Use amortization rate of 0.2.
The system runs all the year around (8760 hours per year).

The optimum pipe diameter is determined on the basis of the fluid flow rate and the
costs of the pipes, the pump, and the cost of electricity as explained in the previous
chapter. Accordingly, an initial diameter is assumed based on which the fluid velocity is
calculated, followed by the Reynolds number, and the major and minor head losses.
These are then used to calculate the pump power and the cost of electricity. Solver can
then be used to determine the optimp pipe diameter that gives the minimum total
annualised cost taking into consideration the initial cost (cost of the pipe and the pump).
The type of the pump is determined by the specific speed (ωs ), which is a dimensionless
quantity defined as [1]:
Q1/ 2
s  (7.2)
gH 3 / 4
Where ω is the rotational speed in rad/s, Q is the required flow rate, and ∆H is the
anticipated pump head. This step determines whether the available centrifugal pump is
suitable for the job. Figure 7.2 shows the range of specific speed for the three types of
pumps; centrifugal, mixed-flow, and axial-flow.
Figure 7.2. Vane design of dynamic pump families as a function of specific speed
Once the pump-type is determined, the pump speed and impeller size can be selected
from the available data. Based on the characteristic curve of the selected pump, the
operating point of the systems can be determined. Finally, we should make sure that the
specified limits for vibration (noise) and cavitation are not exceeded by making sure
that P -------------. There is also a limit on the pressure drop due to head loss in the pipe
system which is specified per 1000 m of the pipe as shown in Table 7.2 [2]:
Table 7.2. Vibration limits for pipe pressure drop

Fluid Pressure drop limit [kPa]
Liquids 600 - 650
Gases 230 - 350
Low pressure steam 900 - 930

Considering the lengthy calculations involved in the design procedure, a workbook will
be developed in which each sheet is dedicated to one of the design steps described
earlier. The purpose of these sheets will be described.
A. Sheet 1: System description and general information

Figure 7.3 shows the fron sheet of the book that shows the general description of the
pump-pipe systems and the various costs involved. The sheet also stores the physical
properties of the fluid (density and viscosity). The total length of the pipe system is
stored as Lpipe and the difference in eleveation between pint 1 and 2 as ∆Z. Values of
K for fittings. Any changes in the data in this sheet will be automatically adopted by the
following sheets.
Figure 7.3. The system
B. Sheet 2: Determining the economic pipe diamater

The second sheet (Sheet 2), that is dedicated to determining the economic pipe
diameter, is shown in Figure 7.4. The flow rate (Q) is made equal to that given in the
front sheet so that any changes in the data in this sheet 1 will be automatically adopted.
A suitable pump efficiency is assumed. Startying with an anitial pipe diameter of D =
0.01 m, Solver … Figure 7.5 shows Solver set-up for this task. The target is to find the
pipe diameter of the pipe (D) that minimses the total annualised cost of the system
(CTotal Equation ???). The two constraints imposed on Solver‟s solution are the
mimium and maximum pipe diameers that are available. Solver‟s solution is shown in
Figure 7.6, which is 4.62 cm.
Figure 7.4. Determining the optimium pipe diameter.
Figure 7.5. Solver Parameters set-up
Figure 7.6. Solver‟s solution for the optimum pipe diameter.

Figure 7.7. Selecting the nearest optimum standard pipe diameter.
As shown in Figure 7.7, the nearest standard size for the pipe of nominal diameter 2”
which has an internal diameter of 5.252 cm. Therefore, this will be the pipe diameter to
be used in the subsequent steps. Having selected the pipe diameter, the next step is to
select the pump since the required discharge is known and the pump head can be
determined.
C. Sheet 3: Selecting the pump

Figure 7.8 shows the sheet that performs the pump-selection step. The required input
data, which is the fluid‟s flow rate is shown in cell B2. Based on this flow rate, the
sheet calculates the required pump head. The sheet calculates the fluid velocity and
friction head loss by using the standard pipe diameter and the speficfied flow rate.
Figure 7.8. Pump selection
The calculations show that the required pump head is 8.73 m. We start looking for a
supplier. Suppose that the local pump supplier deals with a single manufacturer who
provides a range of centrifugal pumps. However, the supplier currently only has the 1½
x3-6 pump which is available at two motor speeds, 1750 rpm and 3500 rpm, and four
impeller sizes. Using these two speeds and the required discharge and head loss, the
specific speed is determined from Equation (7.2) for both motor speeds. Figure 7.8
shows that the specific speeds for both motor speeds are within the range of centrifugal
pumps. We select the lower speed of 1750 rpm to save energy.
D. Sheet 4: Determimin g the opewrating point for the system

Before determining the operating point we have to select a suitable impeller size for the
selected pump. Figure 7.9 shows the pump characteristic curves for the pump at 1750
rpm and various impeller sizes. We select the suitable impleller size that gives the
highest possible efficiency. At the given flow rate and head loss, the impeller size that
is suitable from Figure 7.9 appear to be the 5.5” impller.
Figure 7.9. Performance for a 1½ x 3 – 6 centrifugal pump at 1750 rpm (adopted from
…)
We can now move to determine the actual operating point of the systems with the
selected impeller size. Figure 10 shows sheet 4 …To determine the operating point, we
have to develop a formula for the pump curve based on the curve shown in Figure 7.9.
This is done on the left side of the sheet that … By using trendline.. the following
formula:
h p = 1E+07Q3 – 96844Q2 - 142.36Q + 10.763 (7.3)
Note that at the specified flow rate, the system head loss is 3.41 m while the pump head
is 10.53 m. The operating flow rate at which the two head losses are equal can be
determined by using Solver and Figure 7.11 shows Solver set-up for determining the
operating point.
Figure 7.10. Pump selection before Solver.
Figure 7.11. Solver set-up for determining the operating point
About the set-up. Figure 7.12 shows that the solution found by Solver is 0.00338 m3 /s
or 53.631 gpm, which is more that the required flow rate by 7.31%. The requied pump
power is 379 W. The pump efficiency is 65%, which less than the assumed value of
70%.
E. Sheet 5: Checking for cavitation and exessive noise

The final step in the design procedure is to check for cavitation and exessive noise
(vibration) levels. To avoid cavitation, the pump suction pressure should be well above
the poiling pressure of the fluid at the local temperature. Figure 7.13 shows the curves
for the selected pump. The figure shows that for the selected impeller size of 5.5”, the
pressure loss at the pump suction should not exceed 0.5 m. As long as the liquid level in
the intake tank is well above the pump suction (suction head), then the possibility of
cavitation is remote.
Figure 7.12. Solver‟s solution
Figure 7.13. Performance for a 1½ x 3 – 6 centrifugal pump at 1750 rpm (adopted from
…)
The sheet number 4 (or five) that does the necessary calculations for checking the
vibration limit is shown in Figure 7.14. The figure shows that the pressure drop per
1000 m exeeds the allowable limit. If noise is an issure, then we should consider
freducing the pressure losses which could be achieved by increasing the pipe diameter.
However, increasing the pipe diameter with the same pump is bpound to increase the
flow rate further and makes it necessary to close the vales partially which will result in
energy loss over the lile time of the system. Therefore, we might consider using the
smaller pump impeller, i.e, the 5”-diameter.
Figure 7.14. Step E
7.2. Double-pipe heat exchangers

The simplicity and low opertaion cost of double-pipe heat-exchangers makes them ideal
for low-temperature heat-recovery appplications. The following case, which is based on
Example ?-? given by Cengel and Ghajar [3], will be considered for illustrating the
Excel-based design procedure.

The hot water needed by a dairy plant for milk pasteurization is supplied by a natural
gas furnace. The hot water cannot be returned to the furnace and recirculated because it
is contaminated during the process. Therefore, it is discharged to an open floor drain at
80o C. The chief engineer suggests the installation of a counter-flow double-pipe heat
exchanger so as to utilize the energy of the drained hot water for preheating the
incoming cold water as shown in Figure 7.15. The hot water flows at a rate of 2 kg/s
while the cold water flows at the rate of 2.5 kg/s. You are asked to design the heat
exchanger that maximises the saving in natural-gas cost.
2.5 kg/s Cold water

o
15 C
Hot water
2 kg/s
80o C
Figure 7.15. Schematic of a counter-flow double-pipe HE

- The plant operates 24 h a day and 365 days a year (Operation hours (τ) = 8760
h/year).
- The average temperature of the cold water entering the furnace throughout the
year is 15o C.
- Two pumps are needed for circulating the hot and the cold water in the tubes.
Pump efficiency = 75%.
The inner tube that passes the hot water is to be made of copper while the outer tube is
to be made of stainless steel. The diameters of the inner tube and the outer pipe that
maximise the saving in natural-gas cost (S) are to be determined based on the following
data:
• Cost of pump = 200W + 150 (W in HP)

• Cost of electricity = 0.12 $/kWh
• Cost of natural gas = 1.7 $/Therm
• Amortisation rate (α) = 0.2
• Furnace efficiency 80%
Based on available cost data [4], the costs per metre length can of the steel pipes and
copper tubes can be approximated by the following formulae:
Csteel  3.4249  0.0641d l arg e  0.0084d l2arg e  3.0 105 d l3arg e (Steel) (7.4)
d small  0.015
Ccopper  4  (Copper) (7.5)
0.1  0.015
The total length of the heat exchanger can be a multiple of 5 meters.

Allow for component losses by using K = 10 for every 5 meters of heat-exchanger
length in both the inner and outer tubes.

Since the exit temperature of both the cold water and the hot water are not known in
advance, we need to use the effectiveness-NTU method (refer to Appendix B) so as to
determine the rate of heat transfer in the het-exchanger and the two exit temperatures.
According to this method the rate of heat-transfer (Q) is obtained from:
Q  Q max (7.6)
Where ε is the heat-exchanger effectiveness and Q max is the maximum possible rate of
heat-transfer as given by:
Q max  C min Th,i  Tc,i  (7.7)

Cmin is the minimum of the capacities of the hot and cold water streams.
For a counter-flow double-pipe heat exchanger, the effectiveness is given by:
1  exp NTU 1  c 
 (7.8)
1  c exp NTU 1  c 
Where c and NTU stand for and , respectively. The NTU is defines as:
NTU  UAs / C min (7.9)
The above model requires the surface area and overall heat-transfer coefficient to be
determined. Both of these depend on the pipe diameters. Neglecting the thermal
resistance through the thin copper pipe, the overall heat-transfer coefficient U is given
by:
1 1 1
  (7.10)
U hi ho
The heat-transfer coefficients hi and ho can be determined from the respective Nusselt
numbers according to:
kN u
h (7.11)
Dh
Where Dh is the hydraulic diameter which for the inner tube is equal to its diameter, Di ,
and for the out tube is equal to Do -Di . The Nusselt number itself can be obtained from
the Dittus-Boetler equation:
N u  0.023R Pr (7.12)
Where n is equal to 0.3 for the hot water (cooling) and equal to 0.4 to the cold water
(heating).
Once Q is determined from Equation (7.6), the exit temperatures of the two streams can
be calculated from:
Q
Tc ,o  Tc ,i  (7.13)
Cc
Q
Th,o  Th ,i  (7.14)
Ch
The objective of the optimization is to maximize the saving in fuel cost as given by:

S  C NG  Celect   C pumps  C Steel pipe  CCopper pipe  (7.15)
Where CNG and Celect are the annual costs of natural gas and electricity, respectively, α is
the rate of amortization, Cpumps , is the cost of the two pumps, and Csteel-pipe and Ccopper-pipe
are the costs of the steel pipe and the copper tube, respectively.

The workbook developed for this analysis consists of two sheets. Figure 7.16 shows the
front sheet that the system‟s data and the values of the design parameters. The data
consists of the flow rates and temperatures of the cold and hot water stream as well as
their thermo-physical properties. Properties of water at an average temperature of
(15+80)/2, or 45o C:
Table 7.3. Properties of water at 45o C

Density (ρ) [kg/m3 ] 990.1
Viscosity (μ) 0.596x10-3
Prandtl number (Pr) 3.91
Thermal conductivity (k) 0.637
Specific heat (Cp ) 4.180
The cover sheet also shows the properties of the two pipes and the different costs
involved in the economic analysis. The two pumps are assumed to be 75% efficient.
Figure 7.17 shows the back sheet of the workbook that does the analysis.
Figure 7.16. Front sheet

Figure 7.17. Back sheet
As Figure 7.17 shows, the calculations start with assumed values for the HE length and
pipe diameters. The sheet first calculates the overall heat-transfer coefficient of the HE
(U) by applying Equation () and then determines the rate of heat-transfer (Q) and the
exit temperatures of the hot and cold streams, Th_out and Tc_out, respectively. The
sheet then calculates the different cost involved and the net saving. Figure 7.17 shows
that at the initially assumed values of the two pipe diameters, the net saving is equal to
$00000. Optimsation by Solver. Figure 7.18. Solver‟s solution is shown in Figure 7.19.
Figure 7.18. Solver setting

Figure 7.19. Solution determined by Solver
To determine a sutiable length for the HE the Excel model was used to study the effect
of HE lengh on the optmimum pipe diameters at various lengths. Figure 7.20 shows the
variation of the optimised solutions for the pipe diameters and exit temperatures of the
hot and cold water streams. While Figure 7.20.a shows that both diameters increase
with HE length, Figure 7.20.b shows that the cold water temperature reaches about
57o C at L = 50 m.
0.035 70
0.03 60
0.025 50
0.02 40
0.015 30
0.01 Series1 20 Tc_out
0.005 Series2 10 Th_out
0 0
0 20 40 60 0 20 40 60
Figure 7.20. (a) pipe diameters (b) Exit temperatures
Figure 7.21.a that shows the required power for the two pumps indicates that the power
reach increase with the length and reach their maximum values at L = 15m before
starting to drop with the length. Figure 7.21.b shows that the saving increases with
length and approaches $300,000 at L = 50 m. At this length the corresponding pipe
diameters are 2.07 cm and 3.29 cm, respectively. The nearest standard schedule 40 sizes
are Nom ¾ and Nom 1¼, respectively.
10 350000
8 300000
250000
6
200000
4 150000
Series1
Series1 100000 Series2
2 Series2 Series3
50000
0 0
0 20 40 60 0 20 40 60
Figure 7.21: (a) pump power and (b) Costs and saving
0.8
0.6
Se…
0.4
0.2
0
0 20 40 60 80
HE effectiveness
7.3. Concluding remarks

The advantages include:
a. The ability to minimise the effect of human error

b. The accuracy of the results by allowing a more sophisticated analytical model
to be used.
c. The ability to use up-to-date estimates of the economical factors relevant to the
analysis
d. The ability to adjust the design parameters easily and perform parameter
analyses
Reference:
1. Janna, W.S.(1993), Design of Fluid Thermal Systems, 3rd Edition, CRC Press,
2010.
2. Pressure limits for vibration (internet)
3. Yunis A. Cengel, Afshin J. Ghajar, Heat and Mass Transfer: Fundamentals and
Applications. 5th Edition, McGraw Hill, 2015
4. Cost of steel and copper pipes (internet).
Pipe-network optimisation
Considering the high initial and running costs of large pipe-networks, their design
considerations may involve, in addition to the economic factor, other aspects such as
the network‟s reliability and environmental impact. Moreover, their design optimisation
process may not simply seek to minimise the total owning cost of the network, but to
maximise its economic and social benefits. Therefore, even the network‟s layout may
be one of the optimisation considerations. This chapter focuses on the “least-cost”
optimisation of pipe-networks with pre-determined layouts and aims to demonstrate the
usefulness of Excel and Solver for the optimisation of both gravity-fed and pump-fed
pipe-networks. The methodology extends that applied for hydraulic analyses of pipe-
networks by adding an objective function for optimisation. Since the chance of finding
the global minimum with a deterministic method like the GRG Nonlinear method of
Solver is slim, the chapter studies effect of two factors on the optimised solution. The
first factor is the friction formula used for calculating the major pipe friction losses. In
this respect, three options are investigated; which are the Darcy-Weisbach equation
with the Swamee-Jain formula and with the Colebrook-White formula and the Hazen-
Williams equation. The second factor is the effect of Solver‟s numerical solution
options such as automatic scaling and central differencing.
8.1. The pipe-network optimisation problem

The least-cost optimisation analysis of a pipe-network seeks to minimise the total
owning cost of the network. However, the pipe-network optimisation problem is hard to
solve because of three reasons: (i) the non-linearity of the equations involved (ii) the
availability of pipes in discreet sizes and costs, and (iii) the requirement to satisfying
certain discharge and pressure criteria. For illustration, consider the two-loop gravity-
fed pipe network shown in Figure 8.1 which was first presented by Alperovits and
Shamir [1] as a test case for their network optimisation method. The network carries
water from an elevated tank to 7 consumption points via 8 pipes all of which are 1000-
m long and have the roughness of cast-iron (ε = 0.26 mm, C = 130). Table 8.1 shows
the above-ground level and the demand at each node. The minimum pressure level at
any consumption point is required to be 30 metres column of water (m.c.w). Alperovits
and Shamir [1] sought to minimise the network‟s total installation cost by selecting
from 14 available pipe sizes with the costs shown in Table 8.2.
Though much simpler than most pipe networks met in practice, the total number of
possible combinations for the 8 pipes forming the network and the 14 available pipe
sizes is 148 . Therefore, trying to determine the pipe diameters that minimise the total
cost by simply using trial-and-error would be futile. Engineers usually apply certain
rule-of-thumps, like specifying a range for the economic velocity, so as to minimise the
number of possible pipe arrangements, but the resulting pipe diameters determined this
way are unlikely to ensure optimality. Therefore, numerous attempts have been made to
develop computer models for pipe-network optimisation over the last four decades.
Unlike hydraulic analyses of pipe-networks, pipe-network optimisation is still an active
area of research. For detailed treatment of this topic, the reader may refer to the more
specialised sources listed as references [2-4].
QC C B QB
Loop 1
QE E D QD
Loop 2
QG G F QF
Figure 8.1. A two-loop gravity-fed pipe network
Table 8.1. Node data for the two-loop network

Node Node demand (m3 /h) Ground level (m)
1 (Reservoir) -1120 210
2 100 150
3 100 160
4 120 155
5 270 150
6 330 165
7 200 160
Table 8.2. Cost data for the two-loop network

Diameter Diameter Cost Diameter Diameter Cost
(in) (m) (units) (in) (m) (units)
1 0.0254 2 12 0.3048 50
2 0.0508 5 14 0.3556 60
3 0.0762 8 16 0.4064 90
4 0.1016 11 18 0.4572 130
6 0.1524 16 20 0.5080 170
8 0.2032 23 22 0.5588 300
10 0.2540 32 24 0.6096 550
8.2. The analytical model for pipe -network optimisation

Pipe-network optimisation methods have three main aspects: (i) the hydraulic-analysis
model, the objective function for optimisation, and (iii) the computer-based solution
method. The three aspects of the Excel-Solver method are described below by referring
to the two-loop pipe-network shown in Figure 8.1.
The hydraulic model

The hydraulic model is based on the two conservation principles; the conservation of
mass and the conservation of energy. Assuming the network to be under a steady-state
condition and the flow to be incompressible, the continuity equation applied to the part
of the pipe-network shown in Figure 8.1 that includes the two loops gives:
QAB = QB + QC + QD + QE + QF + QG (8.1)
Where QAB refers to the flow-rate in the pipe AB, while QB , QC, etc., to the discharge at
the nodes B, C, etc. Assuming that the flow rates in pipes BC and DE are known (or
guessed) and that the pipe flow directions are as shown in Figure 8.1, the flow rates in
the remaining pipes can be calculated from the following equations:
QBD  Q AB  QBC  QB (8.2)
QCE  QBC  QC (8.3)
QDF  QBD  QDE  QD (8.4)
QEG  QCE  QDE  QE (8.5)
QFG  QEG  QG (8.6)
The energy equation leads to the following constraints on the magnitudes and directions
of the friction losses in the two loops:
h f , BD  h f , DE  h f , BC  h f ,CE  0 Loop 1 (8.7)
h f , DF  h f , FG  h f , DE  h f , EG  0 Loop 2 (8.8)
The energy equation applied between any two nodes, I and J, also leads to:
pI pJ
 ZI   Z J  h f , IJ (8.9)
 
Where p is the pressure, γ the specific weight of water, Z the elevation, and h f,IJ the
friction head-loss in the pipe stretch joining the two nodes I and J. In terms of
piezometric heads (the sum of pressure head and elevation), the above equation can also
be written as:
HI = HJ + h f,IJ (8.10)
Where HI and HJ are the piezometric heads at node I and node J, respectively.
The objective function for least-cost optimisation

The objective of pipe-network optimisation is to determine the set of pipe diameters
that provides the required nodal loads at the minimum total cost (CT) which is given by:
NP
CT  C
1
i
(8.11)
Where Ci is the cost of pipe i and NP is the total number of pipes in the network. The
optimisation variables are the pipe diameters and the constraints are those imposed by
the energy equation, i.e. Equation (8.7) and Equation (8.8). Additional constraints are
imposed by the pipe diameters available commercially and the requirement that the
minimum pressure head at any consumption node is required to be above a specified
minimum value, i.e.
Dmin  Dopt  Dmax (8.12)
 pi   p 
     (8.13)
      min
The computer-based optimisation method

With the present Excel-based platform, the execution of the optimisation process is
divided between Excel (user-interface and built-in functions) and Solver. Starting with
guessed pipe diameters and flow rates, Excel is used to calculate the pipe flow rates that
satisfy the continuity equations given in the hydraulic model and determine the total
network cost according to Equation (8.11). Solver is used to minimise the total cost by
adjusting the pipe diameters and selected flow rates while satisfying the energy
constraints and the constraints on the pipe diameters and pressure levels. Either the
power function of Equation (6.14) or the interpolation function"Interpl1" provided by
Thermax can be used to determine the pipe costs at any set of pipe diameters. The two
optimisation methods offered by Solver, viz. the GRG Nonlinear method and the
Evolutionary method, represent two classes of pipe-network optimisation methods
currently being used [5]. Chapter 3 demonstrated the ability of the Evolutionary method
to find the global solution for a simple function that involved a single optimisation
parameter, but the more relevant cases considered in Chapter 4 and Appendix D
showed that the method does not give better results than the GRG Nonlinear method.
Moreover, pipe-network optimisation analyses are time-consuming. Therefore, only the
GRG Nonlinear method will be used in the following analyses.
Being a deterministic non-linear programming method, the GRG Nonlinear method is

likely to find a set of pipe diameters that give a locally optimised solution rather than a
globally optimised one. The optimised solution is also expected to be affected by two
factors: (i) the method used to evaluate the friction loss term (h f ) in the energy
equations, and (ii) the solution options used with the GRG Nonlinear method. In the
following analyses the effect of the friction formula is studied by using three models for
evaluating the friction loss term which are:
(i) Darcy-Weisbach equation with the Swamee-Jain formula, Equation (1.24)

(ii) Darcy-Weisbach equation with the Colebrook-White formula, Equation (1.25)
(iii) Hazen-Williams equation, Equation (1.28)
8.3. Optimisation of the two-loop gravity-fed pipe network

Three Excel sheets were developed for optimisation analyses of the two-loop pipe-
network shown in Figure 8.1 by using the three alternative methods for evaluating
friction losses and the GRG Nonlinear method of Solver. The three sheets determine the
pipe costs were by using Thermax linear interpolation function"Interpl1".
8.3.1. Darcy-Weisbach equation with the Swamee -Jain formula

Figure 8.4 shows the sheet that used the Darcy-Weisbach equation with the Swamee-
Jain formula. The data part on the left-side of the sheet stores the pipe lengths and the
known properties of water (ρ and ν). The pipe roughness is equivalent to that of cast-
iron pipes (ε = 0.26 mm).
Figure 8.4. Excel sheet developed for the D-W equation and the Swamee-Jain formula
All the pipe diameters in column D are assumed. Column E stores the pipe flow rates.
The flow rate in pipe AB is fixed at 1120 m3 /h during the optimisation process. The
flow rates in pipes BC and DE are assumed and the flow rates in the remaining 5 pipes
are calculated according to Equations (8.2) to (8.6). Based on the assumed pipe
diameters, the sheet calculates for each pipe its area (A), water velocity (V), Reynolds
number (Re), friction factor (f), and friction head-loss. Note that formula bar reveals the
formula in cell I3 that calculates the friction factor (f) using the Swamee-Jain formula.
The pipe costs are calculated in column K and summed in cell K12. At the initially
guessed diameters, the total network cost is 501,504 units. The part of the sheet on the
far right shows the nodal elevations (Znode) and calculates the nodal piezometric head
(PZnode) and pressure head (Pnode) according to Equation (8.10). Note that for node
A, PZnode = Znode since the gauge pressure is zero there.
The minimum nodal pressure is determined by using Excel's "MIN" function and stored
in cell P11. With the assumed flow rates and pipe diameters, the minimum pressure
head, which occurs at node G is 18.613 m.c.w, which is less than the specified
minimum value of 30 m.c.w. Cells I13 and I14 calculate the errors in satisfying the
energy constraints as expressed by Equations (8.7) and (8.8). Note that, with the
assumed flow rates, the errors in satisfying these constraints are relatively large. Solver
can now be used to determine the pipe diameters that minimise the total cost while
satisfying the energy constraints and pressure limit. Figure 8.5 shows Solver's set-up for
this multi-variable constrained optimisation task.
Figure 8.5. Solver set-up for optimisation of the two-loop gravity-driven pipe-network
The set-up requires Solver to minimise the total cost (C_total) by changing the
diameters of the 8 network pipes (in cells D3 to D10) as well as the flow rates in pipes
BC and DE (in cells E4 and E6). Note that the flow rates in the other 6 pipes which are
not modified by Solver automatically adjust themselves to satisfy the continuity
equations. Two of the applied constraints require Solver to satisfy Equations (8.7) and
(8.8) within a tolerance of 0.1. In addition to the requirements to satisfy the energy
constraints, Figure 8.5 shows two other constraints on the iterative process. These
include upper and lower limits on the pipe diameters, which are 0.025 m and 0.61 m,
respectively, and a lower limit on the minimum nodal pressure of 30 m.c.w. Figure 8.6
shows the solution found by Solver.
Figure 8.6. Solver solution for the D-W equation and the Swamee-Jain formula
As Figure 8.6 shows, the two loop errors have now been reduced to 0.1. All the pipe
flows are in the assumed direction except pipe FG in which the liquid flows in the
opposite direction. Also note that the minimum nodal pressure head of 30 m.c.w occurs
at two nodes F and G. The minimum total cost of the pipe network is calculated as
444,530 units, which is less than the total cost of the initial network by more than 11%.
8.3.2. Darcy-Weisbach equation with the Colebrook formula

Figure 8.7 shows the sheet using the Colebrook-White formula. The only difference
between this sheet and the previous sheet that used the Swamee-Jain formula is the
formula used for determining the friction factor. The formula bar in Figure 8.7 reveals
the formula in cell J3 that uses the Darcy-Weisbach equation with Colebrook-White
formula. Since the Colebrook-White formula is nonlinear, it is solved by using the
Newton-Raphson solver “NRM” provided by Thermax as explained in Chapter 3. The
UDF written for the Colebrook-White equation, called “Colebrook”, is listed below:
Function colebrook(x, e, Re)

„ Colebrook equation for the friction factor
colebrook = 1/Sqr(x) + (2/log(10))*Log(e /3.7 + 2.51/Re/Sqr(x))
End Function
Note that in VBA syntax the term “log” is used for the natural logarithm “ln” which is
different from Excel. Figure 8.7 shows that the values of the friction factor (f) and
friction losses (hf) calculated by this model are very close to those shown in Figure 8.4.
The value of the minimum pressure appears to be modified slightly and so are the errors
in satisfying the energy equations stored in cells J13 and J14.
Figure 8.7. Excel sheet developed for the D-W equation with the Colebrook formula
Solver was used to determine the flow rates that satisfy the energy equation with the
same set-up shown in Figure 8.5. The results shown in Figure 8.8 indicate that the
minimum nodal pressure also occurs at nodes F and G. However, the minimum total
cost found with this method is 425,599 unites, which is lower than that found by the
Swamee-Jain formula and saves more than 15% of the original network design cost.
Figure 8.8. Solver solution using the D-W equation with the Colebrook formula

Figure 8.9 shows the sheet that calculates the head-losses from the Hazen-Williams
equation. As the figure shows, this sheet is slightly different from the two previous
sheets that used the Darcy-Weisbach equation. Note the formula bar that reveals the
formula in cell H3 that calculates the friction loss in pipe according to Hazen-Williams
equation. Column H stores the values of the pipe friction losses (hf), which appear to be
reasonably close to those obtained from the D-W equation. The errors in satisfying the
energy equation, stored in cell G13 and G14, are also significant.
Figure 8.9. Excel sheet developed for the Hazen-Williams equation
Solver was used to determine the optimum pipe diameters with the same set-up shown
in Figure 8.5 and the solution is shown in Figure 8.10. The figure shows that minimum
nodal pressure occurs at nodes F and G and that the minimum cost obtained by the H-W
equation is 451,167 units, which is higher than the two previous solutions.
Figure 8.10. Solver solution using the Hazen-Williams equation

All the results presented in the previous sections were obtained by using the GRG
Nonlinear method with the default setting shown in Figures 3.15 and 3.16.a, but with
automatic-scaling. To investigate the effect of Solver options, three other solutions for
each friction equation were obtained by deactivating the automatic-scaling option and
by changing the default forward-difference (FD) scheme to the central-difference (CD)
scheme. Table 8.3 summarises the findings of the analyses and compares the four
solutions obtained with the three models with the different Solver options.
Table 8.3. Optimal Solver solutions with different Solver options

Alternative friction formulae
Alternative Solver options Average
S-J C-W H-W
FD with automatic-scaling 444.532 425.564 451.167 438.366
FD without automatic-scaling 465.146 454.855 457.891 459.297
CD with automatic-scaling 447.859 451.310 Failed 449.585
CD without automatic-scaling Failed Failed Failed
Average 456.503 443.910 454.529
All the three friction models gave solutions with the FD approximation of derivatives
with or without automatic-scaling. The CD approximation required automatic-scaling in
order to obtain results with the Darcy-Weisbach equation for both the S-J and C-W
formulae, but failed to give solutions with the H-W equation with or without automatic-
scaling. Among the three friction models, the C-W model gave the lowest average total
cost of 443.910 units and, among the four Solver combinations, the FD with automatic-
scaling gave the lowest average total cost of 438.366 units. The optimised total cost
obtained by this combination and the C-White formula, which is 425.564 units, is the
lowest total cost. Compared to the minimum cost obtained by Alperovits and Shamir
[1] for the same network, which was 497,525 units, all the three present solutions gave
considerably lower total costs.
8.4. Optimisation of a two-loop pump-driven pipe-network

Optimisation analyses of pump-driven pipe-networks differ from those of gravity-
driven networks in two respects. The first difference is that the cost of electricity
needed for operating the pump has to be included in the total cost that is to be
minimised by the solution. The second difference is that both upper and lower limits
have to be imposed on the nodal pressure. To illustrate the optimisation process for
pump-driven pipe-networks, consider the two-loop network given by Rivas et al [6] and
shown in Figure 8.11. The lengths of the 8 pipes that form the network are shown in
Table 8.4 and the elevations of its 7 demand nodes are shown in Table 8.5. Rivas et al
[6] sought to minimise the network‟s total annualised cost for the following data:
- The network transports water (ρ = 1000 kg/m3 , ν = 1.0x10-6 N.s/m2 ).

- The minimum and maximum nodal pressures are 30 and 45 m.c.w.,
respectively.
- The pipe's unit cost in € can be represented by the following power function:
cP = 1800 D1.5
Loop 1 Loop 2
Figure 8.11. Pump-powered pipe network (adopted from Rivas et al [6])
Table 8.4. Lengths of pump-driven network pipes

Pipe Length (m)
AB 2000
BC 1500
CD 2000
BG 1200
CF 2500
DE 3000
FG 1500
EF 1500
Table 8.5. Nodal elevations and demands

Node Elevation (m) Demand (l/s)
A 0 -190 (source)
B 100 40
C 85 30
D 90 20
G 100 50
F 92 20
E 78 30
- All the network pipes have roughness (ε) of 0.26 mm.

- The cost of electricity is €9.18 per kW-h, and the amortisation rate is 0.1.
- The pump has an efficiency of 75%.
- The network is used for 6000 hours per year.

The directions of the pipe flows are assumed to be as shown in Figure 8.11. From
continuity, the flow rate in pipe AB is equal to the total nodal consumption, i.e.
QAB=QB + QC + QD + QE + QF + QG (8.14)
The flows in the two pipes BC and CD (QBC and QCD ) are assumed and the flows in the
other pipes are calculated from the following continuity equations:
QBG  Q AB  QB  QBC (8.15)
QCF  QBC  QC  QCD (8.16)
QDE  QCD  QD (8.17)
QFG  QBG  QG (8.18)
QEF  QE  QDE (8.19)
The energy equation leads to the following constraints:
h f , BC  h f ,CF  h f , FG  h f , BG  0 Loop 1 (8.20)
h f ,CD  h f , DE  h f , EF  h f ,CF  0 Loop 2 (8.21)
The highest pressure head is that at the pump discharge point (p A ) after which the head
decreases as the fluid passes down the network because of friction losses. The pressure
head at node B (p B ) is related to that at the pump-discharge as follows:
p B  Z B  p A  Z A  h f , AB (8.22)
In terms of piezometric head, the above equation reads:
H B  H A  h f , AB (8.23)
In general, the energy equation applied between any two nodes I and J leads to:
H J  H I  h f ,IJ (8.24)
Where HI and HJ are the piezometric heads at node I and node J, respectively, and h f,IJ is
the summation of the friction head-losses in the pipes between the two nodes.
The objective function for optimisation

Taking into consideration the cost of electricity in the case of the pump-driven pipe
network, the total annualised cost is given by:
NP
CT   C
1
i  CE (8.25)
Where α is the amortisation rate nd CE the annual cost of electricity which is calculated
from:
Q AB p A c2
CE  (8.26)
1000 
Where γ is the specific weight of the fluid (water), QAB the flow rate in pipe AB, p A the
pressure head at the pump‟s discharge, c2 the cost of electrical energy per kW-h, τ the
annual operation time in hours, and η the efficiency of the pump.
8.4.2. The optimisation method

In addition to the lower limit imposed on the nodal pressure, for the pump-driven
network there is another constraint for Solver which is that the maximum pressure head
at any consumption node i should not exceed a specified upper value, i.e.:
 pi   p 
     (8.27)
      min
 pi   p 
     (8.28)
      max
Unlike the gravity-fed case, the pump-discharge pressure head, p A , is not known in
advance because it depends on the pump‟s power, which in turn depends on friction
losses in the pipes as well as the specified minimum and maximum pressure levels.
Therefore, p A becomes another optimisation parameter in addition to the diameters of
the 8 pipes and the flow rates in pipes BC and CD. The optimisation procedure starts by
assuming a value for p A based on which the pressure heads at other nodes can be
calculated from Equation (8.23) and Equation (8.24).
8.4.3. Optimisation analyses with alternative friction formulae

As for the gravity-fed pipe network, the present analyses of the pump-driven network
investigated the effect of the friction formulae on the optimisation results by using the
three alternative friction models. Figure 8.12 shows the Excel sheet developed for the
analysis using the Darcy-Weisbach equation with the Swamee-Jain formula. The data
part stores the lengths and assumed diameters of the network pipes together with the
other data given in the problem such as the physical properties of water, electricity cost,
etc. The flow rate in pipe AB is fixed to the value of 190 l/s, which is the sum of all the
nodal demands as given by Equation (8.14). The flow rates in two pipes, BC and CD,
are assumed and the remaining flow rates are calculated from the continuity equations,
i.e. Equations (8.15) to (8.19). Based on these initial flow rates, the sheet calculates the
velocities, Reynolds numbers, friction factors, and friction head losses using the Darcy-
Weisbach in column I.
Figure 8.12. Excel sheet developed for optimisation of the pump-driven network by
using the Swamee-Jain formula
The sheet calculates the costs of the 8 pipes in the network and the individual pipe costs
are summed into the total initial cost of the network (C_pipe) which is stored in cell
J11. The annual cost of electricity is calculated from Equation (8.26) and stored in cell
J12 based on the assumed pump discharge pressure of 200 m.c.w. The total cost
(C_total), which is the summation of the annualised initial cost of the pipes (C_pipe)
and the annual electricity cost (C_elec), is stored in cell J13. With the assumed pipe
diameters and water flow rates, the total annualised cost amounts to €986,374.
We also need to determine the nodal pressures so that the minimum and maximum
pressure levels can be identified. These are calculated and stored in the right part of the
sheet, where column M shows the nodal elevations, column N shows the nodal
piezometric heads, and column O shows the nodal pressure heads, which are the
difference between nodal piezometric heads and elevations. Cell N2 stores the assumed
piezometric pressure head at node A; which is 200 m.c.w. This cell will be one of the
cells to be changed by Solver in the optimisation process. Pressure heads at the other
nodes are calculated according to Equation (8.24). The minimum and maximum nodal
pressure heads are determined by using Excel‟s built-in functions "MIN" and "MAX",
respectively. Based on the assumed pipe diameters and flow rates, the minimum
pressure is 49.17 m.c.w. and occurs at node E. The maximum pressure level is 62.74
m.c.w. and occurs at node G. Both pressure levels are higher than the corresponding
specified minimum and maximum values of 30 and 45 m.c.w., respectively.
Figure 8.12 shows that the assumed flow rates do not satisfy the energy equation since
the summation of the friction losses through the two loops, stored in cells G12 and
G13, are relatively large. Figure 8.13 shows Solver‟s set-up for the optimisation
process. This set-up requires Solver to minimise the value of the target cell J13 by
adjusting the values of 11 cells that contain the diameters of the 8 pipes, the flow rates
in pipes BC and CD, as well as the pump discharge pressure stored in cell N2.
Figure 8.13. Solver set-up for optimisation of the pump-driven network by using the
Swamee-Jain formula
Solver parameters set-up imposes six constraints on the optimisation process. Satisfying
the energy equation, as expressed by Equations (8.20) and (8.21), leads to the first two
constraints and the following two constraints are those dictated by the specified minim
and maximum nodal pressure levels by Inequalities (8.27) and (8.28). The last two
constraints specify the minimum and maximum pipe diameters based on the standard
inner diameters shown in Table 8.1. The solution obtained by Solver with the forward-
difference approximation and automatic-scaling is shown in Figure 8.14.
Figure 8.14. Solution determined by Solver for optimisation of the pump-driven

network by with the Swamee-Jain formula
Figure 8.14 shows that both energy constraints are satisfied within the specified
tolerance of 0.1 m.c.w. The specified minimum and maximum pressure levels are also
met. With the optimised pipe diameters and flow rates, the minimum pressure occurs at
nodes D, E, and F, which is about 30.0 m.c.w, and the maximum pressure occurs at
node C, which is 39.08 m.c.w. All the pipe flows are in the assumed directions.
Two more Excel sheets were developed for this case by using the Colebrook-White
formula and the Hazen-Williams equation. The optimised solutions obtained with these
sheets are shown in Figure 8.15 for the Colebrook-White formula and Figure 8.16 for
the Hazen-Williams equation.

network by using the Colebrook-White formula

network by using the Hazen-Williams equation
Figure 8.17 compares the optimum diameters for the 8 pipes in the network as found by
Solver with the three friction equations. The figure shows slight differences between the
estimations of the three friction models with the Swamee-Jain formula giving larger
pipe diameters compared to the other two models for 6 pipes.
0.45
0.4 S-J formula

C-W formula
0.35
H-W equation
Pipe diameter (m)
0.3
0.25
0.2
0.15
0.1
0.05
0
AB BC CD BG CF DE FG EF
Figure 8.17. Optimum pipe diameters for the pump-driven network determined by using
the Swamee-Jain formula (S-J), the Colebrook-White formula (C-W), and the Hazen-
Williams equation (H-W)

As for the case of the gravity-fed pump, The Excel sheets with the three friction models
were used to determine the optimum pipe diameters with alternative Solver options. In
this case, Solver failed to find solutions with all three friction models without
automatic-scaling. Therefore, Table 8.6 shows only the results obtained by using
automatic-scaling and FD or CD approximations.
Table 8.6. Optimal network costs determined by Solver with different solution options
Alternative friction formulae
Solver options Average
S-J C-W H-W
FD with automatic scaling 647250 643883 636006 642380
CD with automatic scaling 650414 680643 660400 663819
Average 648832 662263 648203
In this case also, the FD approximation of derivatives gave a lower average total cost
than the CD approximation. However, the lowest average total annualised cost of
648203 units was obtained by the Hazen-Williams equation. The lowest annualised cost
of €636006 that saves more than 35% of the original network design cost was also
obtained by the H-W model. The highest cost of €680643 was obtained by the CD
approximation of derivatives and the Colebrook-White formula.
Solver typically needed 68 iterations in less than 2 seconds with the Swamee-Jain
formula, but with the Colebrook-White formula it needed 80 iterations that took 10
seconds. The Hazen-Williams equation typically needed 78 iterations and less than 3
seconds.
8.5. Closure
This chapter presented a method for least-cost optimisation of pipe-networks with pre-
determined layouts by using Excel and Solver. The analyses of both gravity-fed and
pump-driven pipe-networks investigated the dependence of the optimised solution
found by Solver on the method used to determine the pipe friction as well as the
particular combination of Solver options. Three methods for calculating pipe friction
losses were used in the analyses which are: the Darcy-Weisbach equation with the
Swamee-Jain formula, the Darcy-Weisbach equation with the Colebrook-White
formula, and the Hazen-Williams equation. With respect to Solver options, the analyses
used the GRG Nonlinear method with both forward-difference and central-difference
approximations of derivatives with and without automatic-scaling.
The solutions obtained with the alternative pipe-friction equations do not strongly
favour a particular equation since the largest difference between their solutions for both
gravity-fed and pump-driven pipe- networks is less than 5%. However, the analyses
with alternative Solver options showed that all the lowest total costs were obtained by
using the forward-difference approximation of derivatives with automatic-scaling. The
analyses without automatic-scaling highlighted the importance of using this option in
pipe-network optimisation without which Solver may fail to converge to the solution.
Appendix F illustrates the application of the Excel-Solver optimisation method to pipe
networks with practical complexity by considering the gravity-fed pipe network
described by Brkic [7] that has 8 loops and two storage tanks.
References
[1] E. Alperovits and U. Shamir, Design of optimal water distribution systems. Water
Res. Research, 13(6), (1977), pp 885-900.
[2] G.A. Walters, A Review of Pipe Network Optimization Techniques. In: Coulbeck
B., Evans E.P. (eds) Pipeline Systems. Fluid Mechanics and Its Applications, vol
7. Springer, Dordrecht, (1992), pp 3-13
[3] M. Shrivastava, V. Prasad, and R. Khare, Optimization techniques for water
supply network: a critical review, International Journal of Mechanical
Engineering and Technology (IJMET), Volume 5, Issue 9, September (2014), pp.
417-426
[4] G. C. Dandy, A. R. Simpson, and L.J. Murphy, A review of pipe network
optimisation techniques, Presented at Watercomp '93, The 2nd Australasian
Conference on Technical Computing in the Water Industry, 30 March - 1 April

1993, Melbourne, Australia
[5] X. Zhou, D.Y. Gao, and A.R. Simpson, Optimal Design of Water Distribution
Networks by Discrete State Transition Algorithm, Engineering Optimization,
48(4):1-26, April 2015
[6] A. Rivas, T. Gomez-Acebo, J.C. Ramos, The application of spreadsheets to the
analysis and optimization of systems and processes in the teaching of hydraulic
and thermal engineering, Computer Applications in Engineering Education 14(4),
2006. Available at https://doi.org/10.1002/cae.20085
Thermodynamic and economic

optimisation of the air-bottoming
cycle
Gas-turbines‟ exhausts carry away with them significant amounts of energy that can be
recovered and used for improving their thermal efficiency in a number of ways. One
such option is the air-bottoming (AB) cycle in which the exhaust-gas energy is used as
the high-temperature source for a low-pressure gas turbine. Although the quantity of the
exhaust-gas energy is significant, its availability is relatively low. Therefore, the
economic feasibility of the air-bottoming cycle depends on its effectiveness in utilising
the exhaust-gas energy and on the additional cost of its equipment. Therefore, the
thermodynamic performance of the system has to be carefully balanced with its cost.
This chapter deals with thermodynamic and economic optimisation analyses of the air-
bottoming cycle. For the thermodynamic optimisation, the objective is to maximise the
thermal efficiency while for the economic optimisation, it is to minimise the total
annualised cost of the system. The analyses that apply the exact variable specific-heat
method by using Thermax property functions compare two methods for modelling the
heat-exchanger which is a key component in the air-bottoming system.
9.1. The analytical model for the air-bottoming gas-turbine system

Figure 9.1 shows a schematic diagram of a power generation plant that works on the
air-bottoming cycle in which the two gas turbines work in series to produce work from
the fuel energy supplied only to the high-pressure turbine. According to Korobitsyn [1],
the bottom cycle adds 20–30% to the power output of the plant and boosts the plant‟s
efficiency by up to 45%. Compared to the combined Brayton-Rankine (BR) cycle, the
AB cycle doesn‟t require abundant supply of water or expensive water treatment
facilities. Considering its other advantages such as the low initial cost, the shorter
installation period, and the ability to run unmanned, the AB system is an attractive
option for small power plants of up to 50 MWe [2].
Fuel
3
C.C.
2
Air
CT TT G
1
Heat exchanger
5 4
2b 4b
3b
Air
CB TB G
1b
Figure 9.1. Schematic diagram of the air-bottoming gas-turbine plant

Given the intake-air temperature (T1 ) and the pressure ratio of the top cycle (rpT), the
compressor‟s work in the top-cycle (wcT) per unit mass flow rate of air is given by:
wcT  h1  h2 s  /  cT (9.1)
Where ƞcT is the isentropic efficiency of the top-cycle‟s compressor and h 2s is the
enthalpy after compression assuming the compression process to be isentropic. In the
following analyses, air is treated as an ideal gas. Using the variable specific-heat
method, the temperatures T2s can be determined from the corresponding relative
pressure Pr2s which can be obtained from [3]:
Pr 2 s  Pr1  rpT (9.2)
Where Pr1 is the relative pressures at T1 . Given the temperature at state 3 (T3 ), the heat
input in the combustion chamber per unit mass flow rate in the top cycle (q in ) is
obtainable from:
qin  h3  h2  (9.3)
The flow rate in the turbine of the top cycle includes an additional flow due to the mass
of the fuel. Accordingly, turbine work (wtT) in the top cycle per unit mass flow rate of
air is given by:
 
wtT  1  m f h3  h4 s tT (9.4)
Where mf is the mass of the fuel burnt per unit flow rate of air and ƞtT is the isentropic
efficiency of the top-cycle‟s turbine. The mass of the fuel mf is given by:
q in
mf  (9.5)
LHV   cc
Where ƞcc is the combustion efficiency. Treating the combustion products as pure air
and assuming the expansion process in the top-cycle‟s turbine to be isentropic, the
temperatures T4s is determined from the relative pressures Pr4s which is given by:
Pr 4 s  Pr 3 P4 / P3  (9.6)
Where Pr3 is the relative pressures at T3 .
The model for the bottom cycle closely follows that for the top cycle. Assuming that
mFR kilograms of air go through the bottom cycle for each kilogram of air in the top
cycle, the compressor work (wcB ) and turbine work (wtB ) in the bottom cycle per unit
mass flow can be calculated from:
wcB  mFR h1B  h2 B  /  cB (9.7)
wtB  mFR h3B  h4 B tB (9.8)
The relative pressures Pr2sB and Pr4sB can be calculated from:
Pr 2 sB  Pr1B  rpB (9.9)
Pr 4 sB  Pr 3B P3 B / P4 B   Pr 3B P3B / P1B  (9.10)
Where Pr1B and Pr3B are the relative pressures at T1B and T3B , respectively, and rpB is the
pressure ratio in the boom cycle.
The overall net work per unit mass flow rate of air in the top cycle (wnet ) is given by:
wnet  wtT  wcT   wtB  wcB  (9.11)
The overall thermal efficiency and net output power ( W ) of the combined AB cycle
(η) are then obtained from:
  wnet / qin (9.12)
W  maT wnet / 1000 [MW] (9.13)
Where maT is the mass flow rate of air in the top cycle. Equation (9.13) can be
rearranged to give the mass flow rate of air in the top cycle (maT) as:
maT  W  1000 / wnet [kg/s] (9.14)
Thermodynamic analyses of gas-turbine systems are usually simplified by neglecting

the frictional losses in the combustion chamber and heat-exchanger. However, when
considering the economics of the system these practical factors must be accounted for.
Therefore, in the present analyses the pressures at points 3 and 4 are given by:
P3  P2  Pcc (9.15)
P4  P1 /(1  PHX , g ) (9.16)

Where ∆Pcc, and ∆PHX,g are the pressure losses in the combustion chamber and the gas-
side of the heat-exchanger, respectively. Similarly, the pressure at point 3b is given by:
P3B  P2 B  PHX , a (9.17)
Where ∆PHX,a is the pressure loss in the air-side of the heat-exchanger.
9.2. Modelling the gas-to-air heat-exchanger

Unlike T3 , which has a specified value, T3b has to be calculated based on the
temperature of the exhaust gas in the top cycle that enters the heat-exchanger and that
of the compressed air in the bottom cycle by adopting a suitable model for the heat-
exchanger [4]. Since the outcomes of both the thermodynamic and economic
optimisation analyses of the AB system depend on this temperature, proper modelling
of the heat exchanger is required. Figure 9.1 shows the heat exchanger as a counter-
flow device at one side of which the hot exhaust gas from the top cycle enters at T4
while the compressed air of the bottom cycle enters from the other side at T2b . The exit
temperatures of the hot stream T5 and the cold stream T3b depend on the size and
effectiveness of the heat-exchanger as well as the flow rates and specific heats of the
two streams. Two models for determining the exit temperatures have been tested in the
present analyses that apply (a) the heat-exchanger effectiveness method and (b) the
pinch-point method [2,5].
9.2.1. The effectiveness model

Thermodynamic analyses usually determine the exit temperatures of the heat-exchanger
by using the following equation [3]:
Tco  Tci   Thi  Tci  (9.18)
Where Tci and Tco are the inlet and exit temperature of the cold stream, respectively, Thi
and Tho are their corresponding values for the hot stream, and ε is the heat-exchanger‟s
effectiveness. For most heat exchangers used in practice, ε is below 85% [6]. The exit
temperature of the hot stream can then be determined from the balance of energy over
the heat-exchanger:
m c
hho  hhi  hco  hci  (9.19)
m h
Where m  h and m c are the mass flow rates of the hot and the cold streams, respectively.
Values of the enthalpies h hi , h co , and h ci are determined using property tables or property
relationships. In the present analyses, this can be done by using the property function
“GasTK_1h” provided by Thermax with air as the fluid. Similarly, the temperature Tho
can be determined once h ho is known. Figure 9.2 shows the variation of the net specific
work and thermal efficiency with the mass flow rate ratio ( m FR  m c / m h ) as obtained
by using Equations (9.18) and (9.19). The figure indicates that both the thermal
efficiency and net work continuously decrease with increasing the mass flow rate in the
bottom cycle, m c ; which suggests that m c should be zero. As the following analyses
show, the optimum value for mFR is actually close to unity.
500 0.50
Net specific work (kJ/kg)
400 0.40
Thermal efficiency
300 0.30
200 0.20
w_net
100 0.10
η
0 0.00
0 0.5 1 1.5 2
Ratio of mass flow rates (m FR =m B/m T)
Figure 9.2. Variation of the net specific work and thermal efficiency of the AB system
with the ratio of the mass flow rates according to the simple effectiveness model
For constant specific heats, Equation (9.19) can be rewritten as:
m c c pc
Tho  Thi  Tco  Tci  (9.20)
m h c ph
The multiplication of the fluid‟s mass flow rate ( m ) with its specific heat (cp ) in
Equation (9.20) is called the heat capacity of the fluid (C), i.e. Ch  m h c ph and
Cc  m c c pc . In optimisation analyses of the AB cycle, the heat capacity of the hot gas
can be less or more than that of the cold air by significant margins. To take this into
consideration, Czaja et al [7] used an improved effectiveness model by using Equation
(9.18) only when Ch > Cc. When Ch ≤ Cc, the following relationship is used instead:
Tho  Thi   Thi  Tci  (9.21)
In this case, the exit temperature of the cold gas (Tco ) is obtained from its enthalpy (h co ),
which is determined by the following energy-conservation equation:
m h
hco  hci  hhi  hho  (9.22)
m c
9.2.2. The Pinch analysis model

Figure 9.3 shows the variation of the temperatures in a counter flow heat-exchanger in
which the hot and cold streams do not undergo phase-change. The pinch point is the
point at which the temperature difference between the two streams is at its minimum
value. For the heat-exchanger shown in Figure 9.3 the pinch point is located either at
the hot-stream‟s entry or the cold-stream‟s entry. If the heat capacitance of the hot
stream (Ch ) is greater than the heat capacitance of the cold stream (Cc), the pinch point
would be located at the hot stream entry and the cold stream exit.
∆Tp if Ch >Cc
Thi
Temperature
Tco ∆Tp if Cc>Ch
Tho
Tci
Figure 9.3. The pinch point in a counter-flow heat-exchanger without phase change
Pinch analyses provide an alternative method for determining the temperatures of the
hot and cold streams exiting the heat-exchanger. According to this method, the exit
temperature of the cold stream (Tco ) is determined from:
Tco  Thi  T p (9.23)
Where ∆Tp is the designed pinch temperature difference as shown in Figure 9.3. The
enthalpy of the hot stream (h ho ) is then determined by using Equation (9.19). However,
if the heat capacitance of the cold stream is greater, then the exit temperature of the hot
steam (Tho ) is determined first by:
Tho  Tci  T p (9.24)
Energy balance over the heat-exchanger then enables h co to be determined according to

Equation (9.22). This model which is simpler than the modified ε-model has been
adopted by Saghafifar and Poullikkas [2] in the analysis of the air-bottoming cycle and
by Bejan et al [5] in the analysis of a cogeneration system.
9.3. Thermodynamic optimisation of the AB cycle

For economic feasibility of the AB system a compromise must be made between the
thermodynamic improvement of the system and the additional cost needed. Both of
these factors depend on a number of operating parameters that include the pressure
ratios in the upper and lower cycle, the efficiencies of the upper and lower turbines and
compressors, the size and effectiveness of the heat exchanger. These parameters can be
optimised with the objective of either maximising the system‟s thermal efficiency
(thermodynamic optimisation) or minimising its total cost (economic optimisation). To
illustrate the concepts underlying both optimisation analyses, an AB system will be
considered that produces 30 MWe of net power at the basic design data given in Table
9.1. This section deals with the thermodynamic optimisation analysis.
Table 9.1. Basic design data and initial conditions for the AB system
Parameter Assigned value
Lower heating value (LHV ) of fuel (natural gas) 50,010 kJ/kg
Inlet-air temperature 300K
Inlet-air pressure 101.3 kPa
Pressure ratio in the top cycle 10
Inlet temperature of the top cycle turbine 1520K
Pressure ratio in the bottom cycle 4
Inlet temperature of the bottom cycle turbine 850K
Compressors isentropic efficiency 86%
Turbines isentropic efficiency 86%
Combustion chamber efficiency 98%
Combustion chamber pressure drop 5%
Heat exchanger pressure drop (air side) 5%
Heat exchanger pressure drop (gas side) 3%
Two Excel sheets were developed for the AB system using the two heat-exchanger
models discussed in the previous section. Figure 9.4 shows the sheet that uses the
modified effectiveness model. The upper part of the sheet includes the basic data and
calculation for the top cycle while the lower part stores the data of the bottom cycle and
its calculations. Thermax functions are used to determine fluid properties at the various
states. The results shown in Figure 9.4 are those for a pressure ratio in the top cycle
(P_ratio) of 4 and a mass-flow ratio in the bottom cycle (MFR) of 1. The value of the
heat-exchanger effectiveness (ε) is set to a value that makes T3b = 850K. The
appropriate value of ε is determined by using Goal Seek as shown in Figure 9.5. The
appropriate value determined by Goal Seek is ε = 0.756.
Figure 9.4. Excel sheet for the analysis of the AB system with the ε-model
Figure 9.5. Goal Seek set-up for determining the appropriate heat-exchanger
effectiveness
The formula tab in Figure 9.4 reveals the formula used to determine the heat-exchanger
exit temperature (T_3b) according to Equation (9.18) or Equation (9.21). Figure 9.4
shows that for each 1 kg/s of air flow, the top cycle requires 1036.599 kJ of heat input
(Q_in) to produce 333.4589 kJ of net work (w_nett). Therefore, the thermal efficiency
of the top cycle standing alone (η_top) is 0.321685. The exhaust gas leaves the top
cycle (T_4) at 973.2916K and passes part of its energy to the compressed air of the
bottom cycle that enters the heat exchanger (T_2b) at 467.2093K and leaves (T_3b) at
850K. The heated air then expands in the bottom-cycle turbine to produce a net work
(w_netb) of 66.0615 kJ in the bottom cycle per kg of air flow in the top cycle. The
overall thermal efficiency (η_overall) at the specified pressure ratio is 0.385361, which
is higher than that of the top cycle alone by about 19.79%.
Figure 9.6 shows the sheet developed for the pinch method using the same pressure
ratio and mass flow rate ratio as those of the ε-model. The formula tab in the sheet
reveals the formula used to determine the heat-exchanger exit temperature (T_3b)
according to Equation (9.23) or Equation (9.22). The temperature difference at the
pinch-point (∆T_pinch) is determined by using Goal-Seek such that the value of T3b is
the same as that obtained by the ε-method, i.e., 850K. The value thus obtained is
∆T_pinch = 123.2916K. It should be pointed out that this pinch-point temperature
difference is rather high compared to the value used by Saghafifar and Poullikkas [2],
which is only 9K.
Figure 9.6. Excel sheet for the analysis of the AB cycle with the pinch method
Figure 9.6 shows that the heat input (Q_in), net work in the top cycle (w_nett), thermal
efficiency of the top cycle standing alone (η_top), net work in the bottom cycle
(w_netb), and overall thermal efficiency (η_overall) are basically the same as those
obtained by the ε-model. Similarly, the overall efficiency of the air-bottoming cycle is
higher than that of the top cycle alone by about 19.79%.
Among the many deaign parameters that affect the thermodynamic and economic
merits of the AB systems the most important factors are the pressure ratio and the mass
flow-rate in the bottom cycle. Keeping the mass flow ratio of air in the bottom cycle
(mFR) equal to 1.0, the two Excel sheets were used to determine the variation of the
combined net specific work (w_net) and thermal efficiency (η_overall) as the pressure
ratio of the bottom cycle (rpb ) was varied from 2 to 8. As shown in Figure 9.7, the
results indicate that the ε-model estimates that the maximum net work and the
maximum thermal efficiency occur approximately at a pressure ratio of 4, but the
pinch-analysis model predicts a lower value for the optimum pressure ratio in the
bottom cycle. With the pressure ratio in the bottom cycle fixed at 4, the two models
were then used to determine the variation of the combined net specific work and overall
thermal efficiency with the mass flow ratio in the cycle (mFR ). Figure 9.8 shows the
results obtained for 0.2 ≤ mFR ≤ 2.0. The figure indicates that the optimum value for mFR
at which the AB cycle yields its maximum net specific work and thermal efficiency is
about 1.0. Solver allows the AB system to be optimised while including other design
parameters in the optimisation process.
420 0.4
400 0.38
Thermal efficiency
380
0.36
360
0.34
340 ε-model
ε-model
320 0.32
Pinch-model Pinch-model
300 0.3
0 2 4 6 8 10 0 2 4 6 8 10
Pressure ratio Pressure ratio
(a) (b)
Figure 9.7. Effect of the bottom-cycle rp on (a) the combined net specific work and (b)
the thermal efficiency
420 0.4
400
Thermal efficiency
0.38
380
0.36
360
0.34
340 ε-method
ε-method
320 0.32
Pinch method Pinch method
300 0.3
0 1 2 3 0 1 2 3
Mass-flow ratio Mass-flow ratio
(a) (b)
Figure 9.8. Effect of mass flow ratio on (a) the net specific work and (b) the thermal
efficiency
For thermodynamic optimisation of the AB system either Equation (9.11) or Equation

(9.12) can be used as the objective function. Nine design variables are considered in the
following optimisation analyses of the system, which are the two pressure ratios in the
top and bottom cycles (rp and rpb ), the top turbine inlet temperature (T3 ), the isentropic
efficiencies of the two compressors (ƞcT and ƞcB ), the isentropic efficiencies of the two
turbines (ƞtT and ƞtB ), the mass flow ratio (mFR), and the heat-exchanger parameter
relevant to the model used, i.e. ε or ∆Tp . Solver can take into consideration practical
limitations on these design variables by applying relevant constraints on the solution
process. Figure 9.9 shows Solver‟s set-up used with the Excel sheet that applies the ε-
model in which the overall thermal efficiency (η_overall) is treated as the objective for
optimisation with Equation (9.12) as the objective function.
Figure 9.9 Solver‟s set-up to maximise the overall thermal efficiency with the ε-model
Figure 9.9 also shows that ten constraints are imposed on the optimisation process. The
first constraint ($O$14=30) limits the system‟s output to 30 MW while the second and
third constraints limit the maximum compressor pressure ratios in the top and bottom
cycles to 16. The fourth constraint limits the maximum turbine‟s inlet temperature to
1550K and the fifth constraint limits the minimum exhaust gas temperature (T_5) to
400K (127o C) which is required in order to avoid water condensation [5]. The size and
effectiveness of the heat exchanger is also limited by practical constraints that must be
respected by the relevant heat-exchanger model. For the ε-model, the sixth constraint
limits the heat-exchanger effectiveness (ε) to 0.85. The last four constraints limit the
isentropic efficiencies of the compressors sand turbines in the top and bottom cycles to
89.0% and 90.0%, respectively. Although not followed in Figure 9.9, it is a good
practice to impose both lower and upper practical limits on all the design variables
involved in the optimisation process.
Both the GRG Nonlinear and the Evolutionary methods of Solver suit the optimisation
analysis. As shown in Chapter 3, the Evolutionary method stands a better chance to find
the global optimum solution than the GRG Nonlinear method. Considering the
advantage of the Evolutionary method and that the present optimisation analyses
involve only nine variables, at initial attempt was made to use the Evolutionary method
instead of the GRG Nonlinear method. However, the trial showed that method was
unstable and gave different solutions for the same data apart from taking long computer
time. Therefore, both the thermodynamic and economic optimisation analyses were
performed by using the GRG Nonlinear method. Figures 9.10 and 9.11 show the
optimum solutions found with the two heat-exchanger models and Table 9.2 compares
the different system parameters with their initial values in the basic design given in
Table 9.1.
Figure 9.10. Solver solution with the effectiveness model
Figure 9.11. Solver solution with the pinch-analysis model
As the figures in Table 9.2 show, both the ε-model and the pinch models led to the
maximum specified limits for the gas-turbine‟s inlet temperature (T3 ) and the top-cycle
pressure ratio (rpT). Both models obtained a pressure ratio in the bottom cycle which is
lower than 5 and the optimum mass flow rate in the bottom cycle obtained by both
models is the same as that in the top cycle, i.e. MFR = 1. When the isentropic
efficiencies of the compressors and turbines were treated as optimisation variables, they
reached their corresponding maximum values. The ε-model also gave the maximum
value for the heat-exchanger effectiveness (ε).
Table 9.2 shows that the optimisation analyses by both models led to a significant
increase in the thermal efficiency which reached 47.7% according to the effectiveness
model and 49.3% according to the pinch model. Note that the thermal efficiency of the
basic design of the AB cycle is only 38.5% while that of the top gas-turbine cycle alone
is only 32.169%. Also note that both models predicted the exhaust-gas temperature to
be higher than the imposed limit of 400K.
Table 9.2. Optimum design variables for the AB system determined by the two models
Basic design ε–model Pinch model
T3 1520 1550‡ 1550‡
rpT 10 16‡ 16‡
rbB 4.0 4.625 4.613
mFR 1.0 1.0 1.0
ƞcT 0.86 0.89‡ 0.89‡
ƞtT 0.86 0.9‡ 0.9‡
ƞcB 0.86 0.89‡ 0.89‡
ƞtB 0.86 0.9‡ 0.9‡
ε [-] 0.756 0.85‡
T5 607.384 551.624 500.402
ƞ 0.385 0.477 0.493
ffi
Specified upper limit
9.4. Economic optimisation of the AB cycle

The thermodynamic optimisation analysis maximised the thermal efficiency of the AB
system, but led to idealised solutions since almost all the design variables shown in
Table 9.2 reached their maximum values. Although the resulting increase in thermal
efficiency would lead to remarkable savings in fuel cost, the extra initial investment
required by the additional equipment may not be economically feasible. When the
objective of the optimisation process becomes minimising the total owning cost of the
system instead of maximising the system‟s efficiency, the process will only allow for
the feasible increase in thermal efficiency that pays for itself. The total owning cost
consists of the initial cost of the system and its running cost over its lifetime. The initial
cost of the system is the purchase cost of the equipment which is paid at the initial stage
of the project while the running cost is mainly the cost of fuel in the case of a power-
generation system. Accordingly, the objective function for the optimisation is [2]:
CTotal  CequipCRF  C fuel (9.25)
Where CTotal stands for the total owning cost, Cequip for the cost of equipment, CRF for
the capital recovery factor, φ for the maintenance cost factor, and Cfuel for the annual
fuel cost. By accounting for the costs of the individual components that constitute the
AB system and the unit cost of the fuel, Equation (9.25) can be written as:
n
Z i  CRF  
CTotal    m f c f LHV (9.26)
i 1 3600 N
Where Zi ,i = 1,2,..., n, is the capital investment cost of each component constituting the
ABC system, N is the annual number of operation hours, m  f is the rate of fuel
consumption per second, and cf is the cost of fuel per energy unit.
Following Saghafifar and Poullikkas [2], in the following analyses the capital
recovery factor (CRF) and the maintenance factor φ are taken as 19.1% and 1.03,
respectively. The annual number of hours of operation of the plant N is assumed to be
7446 h/year [5]. Knopf [8] provided the following formulae that express the purchase
costs of the equipment as functions of their thermodynamic parameters, in US$:
ac rcT m aT
Z cT  ln rcT  (9.27)
0.9   cT
a r m
Z cB  c cB aB ln rcB  (9.28)
0.9   cB
Z tT 
at m g
0.92   tT

lnrtT 1  e 0.036 T3  54.4   (9.29)
Z tB 
at m aB
0.92   tB

ln rtB  1  e 0.036 T3 B 54.4   (9.30)
Z cc 
acc m a
0.995  P3 / P2

1  e 0.018 T3 26.4   (9.31)
 m g h4  h5  
0.6
Z HX  a HX   (9.32)
 0.018Tlm 
Where ZcT, ZcB , ZtT, ZtB , Zcc, and ZHG stand for the purchase costs of the top-cycle
compressor, bottom-cycle compressor, top-cycle turbine, bottom-cycle turbine,
combustion chamber, and heat-exchanger, respectively, rcT and rcB stand for the
compressor pressure ratios in the top and bottom cycles, respectively, and rtT and rtB
stand for the turbine pressure ratios in the top and bottom cycles, respectively. The log-
mean temperature difference (∆Tlm ) in Equation (9.32) is calculated from:
Tlm 
T5  T2b 
(9.33)
T4  T3b 
Equation (9.32) is based on a heat-exchanger with an overall heat-transfer coefficient of
18 W/m2 .K [5]. The constants a c, a t , a cc, and a HG appearing in the cost formulae (9.27) –
(9.32) vary with the year of purchase. The following table shows values of the constants
as given in Bejan et al [5].
ac at a cc a HX
71.1 479.34 46.08 4122
The economic optimisation analysis of the AB system required certain extensions to the
two Excel models developed for the thermodynamic optimisation. Figure 9.12 shows
the sheet that applies the ε-model. Two cells have been added to the sheet shown in
Figure 9.4, which are O17 and O18, to show the surface area of the heat exchanger
(A_HX) and the total cost (C_total). The data and calculations of the economic model
have been placed in a separate sheet of the workbook as shown in Figure 9.13.
Figure 9.12. The front sheet of the workbook developed for the ε-model
Figure 9.13. The second sheet of the workbook developed for the ε-model
As with thermodynamic optimisation, the inlet temperature of the turbine in the bottom
cycle (T_3b) for both models is required to be 850K at the initial design parameters
shown in Table 9.1. Figure 9.12 shows the sheet of the ε-model with the specified
pressure ratio in the bottom cycle (P_ratiob) of 4 and mass-flow ratio (MFR) of 1.0.
The heat-exchanger area and the total cost determined by this model at MFR = 1 are
12965.95 m2 and 0.307 $/s, respectively. The extended sheet was used to determine the
variation of the different costs of the system with the mass-flow ratio MFR and the
results obtained are shown in Figures 9.14 and 9.15.
4.0E+06 4.0E+06
Z_cB
Cost ($Million)
3.0E+06 3.0E+06 Z_tB
Cost ($Million)
Z_HX
2.0E+06 Z_cT 2.0E+06
Z_tT
1.0E+06 Z_cc 1.0E+06
0.0E+00 0.0E+00
0 1 2 0 1 2
Mass-flow rate ratio (MFR) Mass-flow rate ratio (MFR)
(a) (b)
Figure 9.14. Variation of the equipment purchase costs as determined by the ε-model:
(a) the top cycle, (b) the bottom cycle
0.50
C_initial
0.40 C_fuel
C_total
Cost ($/s)
0.30
0.20
0.10
0.00
0 1 2
Mass-flow rate ratio (MFR)
Figure 9.15. Variation of the equipment, fuel, and total costs as determined by the ε-
model
Figure 9.14 indicates the costs of the turbines and compressors are the two main
components in the total system cost. According to this model, the turbine and
compressor costs in the bottom cycle increase continuously as the mass flow rate in the
bottom cycle is increased but the costs of the turbine and compressor in the top cycle
have a minimum value at MFR ≈ 1.0. Figure 9.14 also shows that the purchase cost of
the heat-exchanger reaches its maximum at MFR ≈ 1.0. Figure 9.15 that shows the
variation of the equipment purchase cost, the fuel cost, and the total cost with the mass
flow rate in the bottom cycle shows that the fuel cost is the main item in the total cost
and that the minimum values of the fuel cost and the total cost occurs when the mass-
flow rate in the bottom cycle is around 1.0.
The Excel sheet using the pinch-analysis model was extended in a similar manner to
that of the ε-model. Figure 9.16 and Figure 9.17 show the variations of the different
costs with the ratio of the mass flow rates in the bottom cycle as obtained with this
model. The figure shows almost identical trends to those produced by the ε-model. The
purchase cost of the heat-exchanger also has a maximum value at MFR ≈ 1.0.
4.0E+06 4.0E+06
Z_cB
Cost ($Million)
Cost ($Million)
3.0E+06 3.0E+06 Z_tB

Z_HX
2.0E+06 Z_cT 2.0E+06
Z_tT
1.0E+06 Z_cc 1.0E+06
0.0E+00 0.0E+00
0 1 2 0 1 2
Mass-flow rate ratio (MFR) Mass-flow rate ratio (MFR)
(a) (b)
Figure 9.16. Variation of the equipment purchase costs as determined by the pinch
model: (a) the top cycle, (b) the bottom cycle
0.50
C_initial
0.40 C_fuel
C_total
Cost ($/s)
0.30
0.20
0.10
0.00
0 0.5 1 1.5 2
Mass-flow rate ratio (MFR)
Figure 9.17. Variation of the equipment, fuel, and total cost determined by the pinch-
model
Figure 9.18 shows Solver set-up used to minimise the total cost of the AB system. This
setup involved the same design variables and constraints applied for the thermodynamic
optimisation shown in Figure 9.9, but with the objective of the optimisation being
changed to minimising the total cost instead of maximising the thermal efficiency.
Figures 9.19 and 9.20 show the solution found by Solver with the effectiveness model
and the pinch model, respectively.
Figure 9. 18. Solver Parametrs setup for the economic optimisation of the AB system
Figure 9.19. Solver solution for the economic optimisation with the ε-model
Figure 9.20. Solver solution for the economic optimisation with the pinch model
For comparison, the total annualised costs of the basic design and the
thermodynamically optimised system were obtained by using the sheets of the two
models developed for economic optimisation and the results are shown in Figures 9.21
and 9.22. Figure 9.21 shows that the thermal efficiency of the thermodynamically
optimised (TO) system obtained by both models is significantly higher than that of the
basic-design system, but the efficiency of the economically optimised (EO) system is
only slightly better.
0.60
Basic
0.50
TO
Thermal efficiency
0.40
EO
0.30
0.20
0.10
0.00
ε-model Pinch model
Figure 9.21. The thermal efficiencies obtained by for the thermodynamically optimised
(TO) and economically optimised (EO) AB system compared to that of the basic design
0.60
Basic
0.50
Annualised cost ($/s)
TO
0.40
EO
0.30
0.20
0.10
0.00
Figure 9.22. The total annualised costs obtained for the thermodynamically optimised
(TO) and economically optimised (EO) AB system compared to that of the basic design
Figure 9.22 that compares the total annualised costs of the thermodynamically
optimised and the economically optimised systems with those of the basic design shows
that economic optimisation leads to lower equipment costs and, therefore, lower
annualised total cost.
Table 9.3 summarises the results of the economic optimisation with the two heat-
exchanger models compared to the basic design. Unlike the thermodynamic
optimisation results, the pressure ratios, turbine inlet temperatures, and heat-exchanger
effectiveness parameters all fall below the respective maximum limits. Consequently,
the economically optimisation leads to a thermal efficiency which is lower than that
obtained by thermodynamic optimisation, but higher than that of the basic design.
Compared to the basic design, the optimised solution with the ε-model reduced the total
annualised cost of the system by 11.2% while the pinch model reduced it by 10.3%.
Table 9.3. Economic optimisation solutions obtained by the two models for the AB
system
Basic design ε-model Pinch model
TITT [K] 1520 1485.605 1496.244
rpT 10.0 10.758 9.936
rpB 4.0 3.939 3.941
ƞcT 0.86 0.827 0.832
ƞtT 0.86 0.90* 0.90*
ƞcB 0.86 0.851 0.860
ƞtB 0.86 0.830 0.839
mFR 1.0 1.0 1.0
ε/∆Tp [K] 0.756/123.292K 0.795 121.688
T5 [K] 607.384 570.561 601.525
ƞ 0.385 0.396 0.391
ƩZ equipment [$] 9,537,006 6,100,440 6,201,882
Cfuel [$/s] 0.242 0.232 0.235
C_inital [$/s] 0.070 0.045 0.046
C_total [$/s] 0.312 0.277 0.280
*
Specified upper limit
9.5. Solver time

The optimisation analyses using the two heat-exchanger models were performed on a
PC with an Intel Celeron CPU N2840 @ 2.16 GHz processor with a 64-bit operating
system. The GRG Non-linear option of Solver was used with forward-difference
approximation of derivatives and automatic-scaling. Table 9.4 shows typical number of
iterations and computation times required by each method for the thermodynamic and
the economic optimisation analyses. The figures of Table 9.4 show that the numbers of
iterations required by both heat-transfer models were well below the maximum number
of iterations in the default setting, which are 100 iterations. The figures also show that
the pinch model has lower average number of iterations and computer time than the ε-
model for both the thermodynamic and the economic optimisation analyses. Also note
that the economic optimisation requires less number of iterations and CPU time than the
thermodynamic optimisation by both models.
Table 9.4. Typical number of iterations and CPU times in seconds for Solver
optimisation with the two heat-exchanger models
Iterations Time (s) Iterations Time (s)
Thermodynamic analysis 17 4 13 3
Economic analysis 11 4 9 3
9.6. Closure
This chapter demonstrates the capability of the Excel-based platform to perform multi-
variable thermodynamic and economic optimisation of fluid-thermal systems. By
determining the thermodynamic fluid properties with Thermax functions, the Excel-
based platform enables exact variable specific-heat method to be used. The optimisation
analyses of the air-bottoming system highlight the difference between the objectives
and outcomes of thermodynamic optimisation and economic optimisation analyses.
While the thermodynamic optimisation yield an idealised system with a high thermal
efficiency, the economic optimisation yield a system with the lowest total cost for the
desired capacity but with less thermal efficiency. For the present case, both the
thermodynamically optimised and the economically optimised systems could reach a
higher thermal efficiency than the basic design.
The thermodynamic and economic optimisation analyses of the air-bottoming system

compared the performance of two heat-transfer methods for modelling the gas-to-air
heat-exchanger. One of these models applied an extended heat-exchanger effectiveness
method while the other applied the pinch-point method. Although the analyses showed
that the optimised solutions obtained by the two heat-exchanger models are almost
identical, the pinch-point model is advantageous over the effectiveness model in terms
of implementation simplicity and computer time.
References
[1] M.A Korobitsyn,. New and advanced energy conversion technologies. Analysis
of cogeneration, combined and integrated cycles, ISBN 90 365 1107 0, 1998.
[2] M. Saghafifar, A. Poullikkas, Thermo–economic optimization of air bottoming
cycles, Journal of Power Technologies, 95(3) (2015) 211–220
[4] L. Pierobon & F. Haglind, Design and optimization of air bottoming cycles for
waste heat recovery in off-shore platforms. Applied Energy, Vol. 118, 2014, 156–
165.
Wiley & Sons, 1996.
[6] Y.A. Cengel and A.J. Ghajar. Heat and Mass Transfer: Fundamentals and
Applications. McGraw Hill, 2011
[7] D. Czaja, T. Chmielniak, S. Lepszy, Selection of Gas Turbine Air Bottoming

Cycle for Polish compressor Stations, Journal of Power Technologies , Vol. 93,
2013, No. 2, 67–77
[8] F.C. Knopf, Modeling, analysis and optimization of process and energy systems,
John Wiley & Sons, 2012.
10
Analyses of the gas-turbine cycle with

co-generation and inlet-air cooling
Gas turbines are preferable over steam-turbines and combined-cycle systems for
meeting the peak-load demands on electricity utilities and for providing small and
medium-size industries with power. Gas turbines installed for these purposes are
usually of the open-cycle type in which the exhaust-gas energy is wasted. With the
increasing cost of energy and the environment call for energy efficiency, the utilisation
of this waste energy becomes more attractive. This chapter considers two options for
utilising the waste energy which are cogeneration and inlet air cooling. The chapter
presents least-cost optimisation analyses for the two options. Both options involve a
heat-recovery steam generator (HRSG) to be installed. As in the analyses of the AB
system considered in the previous chapter, the HRSG is modelled by using two heat-
exchanger models which are the effectiveness model and the pinch-analysis model. The
exact analysis method is also adopted here by using the Thermax add-in to determine
the air and gas properties.
10.1. Cogeneration
Cogeneration is the generation of power and heat from the same source. Cogeneration
systems are divided into topping-cycle systems and bottoming-cycle systems. An
example of the topping-cycle cogeneration systems is the one shown in Figure 10.1 that
recovers the waste energy from the gas-turbine for steam generation. Bottoming-cycle
systems are those in which the waste heat from an industrial application, such an
industrial furnace, is used to generate electricity.
Saturated
Feedwater steam
8 9
Air preheater
7 6 5
HRSG
Combustion
products
Natural gas
2
3 10
4
Net
Combustion power
chamber 30 MW
Air
1 compressor Gas turbine
Air
Figure 10.1. Cogeneration system (adapted from Bejan et al [1])
To maximise the utilisation of the exhaust-gas energy, an air-preheater is also used in

the system shown in Figure 10.1. By increasing the temperature of the compressed air
before going to the combustion chamber, the air preheater reduces the fuel consumption
of the plant. Therefore, the additional costs due to the HRSG and air preheater are
compensated for by the production of steam and the reduced fuel consumption.
10.1.1. The thermodynamic model

The analytical model for the gas-turbine itself is basically the same as that described in
Sections 9.1 and 9.2 for the top-cycle of the air-bottoming system, but some extensions
are required in order to allow for the air preheater and the steam production process in
the HRSG. The state of the compressed air at 2 is determined from the given inlet-air
temperature and pressure ratio of the compressor. Using the ε-model, the exit
temperature of the compressed air from the air-preheater (T3 ) is calculated from:
T3  T2   T5  T2  (10.1)
Using the pinch model, T3 is calculated from:
T3  T5  T p (10.2)
The rate of heat-transfer in the HRSG is given by:
QHRSG  ms h9  h8  (10.3)
Where ms is the rate steam-production in the HRSG per unit mass flow rate of air in
the compressor and h 9 and h 8 are the enthalpies of saturated steam at the given pressure
and water at the given temperature, respectively. The enthalpy at point 7 is then
obtained from QHRSG as follows:
h7  h6  QHRSG / m g (10.4)
Where m g is the mass flow rate of the products of combustion per unit mass flow rate
of air in the compressor.
Following Bejan et al [1], the HRSG is modelled as shown in Figure 10.2 with two
parts: an economiser (Ec) and an evaporator (Ev). To avoid vaporisation in the
preheater, the temperature of the liquid water exiting the economiser (Txw) must be kept
below the saturation temperature by a reasonable margin which is set to 15K, i.e,:
Txw  T9  15 (10.5)
Feedwater Saturated steam

T8 T9
Txg T6
T7
Combustion Ec Ev
products
Txw
Figure 10.2. Temperature change in the economiser and evaporator of the HRSG
The sensible energy absorbed by the water in the economiser (Qec) is obtained from:
Qec  ms hxw  h8  (10.6)
Since this amount of energy must be provided by the exhaust gas, the enthalpy of the
gas leaving the evaporator and entering the economiser (h xg ) is obtained from:

hxg  h7  ms / m g Qxw (10.7)
For economic considerations, the temperature of the gas leaving the evaporator (Txg )
should be kept higher than the saturation temperature by a reasonable margin, which is
also taken as 15K, i.e.
Txg  T9  15 (10.8)
The energy absorbed in the evaporator is given by:
Qev  QHRSG  Qec (10.9)
10.1.2. The objective function for optimisation

The objective function for optimisation is still that of the total cost given by Equation
(9.26), but the cost now includes those of the compressor, the air preheater, the
combustion chamber, the turbine, and the heat-recovery steam generator. The purchase
costs of the first four components are given by Equations (9.27) to (9.32). The
following formula for the purchase cost of the HRSG was given by Bejan et al [1]:
 Q 
0.8
 Q ev 
0.8

Z hrsg  a1   ec     1.2 
 a2 m s  a3 m g (10.10)
 Tlm, ec   T


lm, ev  
 
Where Q ec and Q ec are the rates of heat-transfer in the economiser and evaporator,
respectively, in kW and ∆Tlm,ec and are ∆Tlm,ev the log-mean temperature differences in
the economiser and evaporator, respectively, which are given by:
T7  T8   Txg  Txw 

Tlm, ec 
lnT7  T8  / Txg  Txw 
(10.11)
Txg  Txw  T6  T9 

Tlm, ev 
lnTxg  Txw  / T6  T9 
(10.12)
The values of the three constants in Equation (10.10), a 1 , a 2 , and a 3 are 6570 $/(kg/s)0.8 ,
21276 $/(kg/s), 1184.4 $/(kg/s)1.2 , respectively [1].
10.2. Optimisation analysis of the cogeneration system

The cogeneration system considered for analysis is similar to that given by Bejan et al
[1] which produces 30 MW of electric power and 14 kg/s of saturated steam at 20 MPa.
Table 10.1 shows the basic design data and initial conditions for the system.
Table 10.1. Basic design data and initial conditions for the cogeneration system [1]
Parameter Assigned value
Inlet-air temperature 300K
Inlet-air pressure 101.3 kPa
Pressure ratio in the turbine 10
Inlet temperature of the turbine 1520K
Compressor isentropic efficiency 86%
Turbine isentropic efficiency 86%
Combustion chamber efficiency 98%
Combustion chamber pressure drop 5%
Heat exchanger pressure drop (air side) 5%
Heat exchanger pressure drop (gas side) 3%
Lower heating value of natural gas (LHV ) 50,010 kJ/kg
The overall heat transfer coefficient for the air preheater based on air-side surface area
is 18 W/m2 .K and those for the economiser and evaporator of the HRSG are 50 and 250
W/m2 .K, respectively. The water is supplied at 25o C. The annual number of hours of
operation of the plant N is 7446 h/year and the capital recovery factor (CRF) and the
maintenance factor φ are 19.1% and 1.03, respectively.
Two Excel sheets were developed for the cogeneration system using the alternative
heat-exchanger models described in Section 9.2. The values used for ε and ∆Tp are such
that the temperature T3 is 850K. Figure 10.3 shows the Excel sheet that applies the ε-
model. The data part on the left column of the sheet now adds the mass flow rate and
pressure of the generated steam production (msteam and P_steam), the inlet
temperature of water (T_8), and the minimum difference between the water temperature
leaving the economiser and the saturation temperature at the specified pressure
(∆T_xw). The formula bar reveals the formula that determined the temperature of the
compressed air leaving the preheater (T_3) according to Equation (10.1). The sheet then
determines the state of the fluids at each point and calculates the net work, heat input,
etc. Figure 10.4 shows the back sheet of the model that applies the economic model to
determine the various costs of the system. This sheet uses the log-mean temperature
difference method to calculate the areas of the three heat-exchangers: the air preheater
(A_HX), the economiser (A_EC), and the evaporator (A_EV).
Figure 10.3. The Excel sheet developed for the cogeneration system using the ε-model
Figure 10.4. The back sheet developed for the cogeneration system using the ε-model
Based on the initial values of the design variables shown in Table 10.1, the sheet
determines the annualised fuel cost, initial cost, and total cost as 0.2339, 0.07849, and
0.31241$/s, respectively. The total purchase cost of the equipment is 10,595,722$.
Figure 10.5 shows the front Excel sheet developed for the cogeneration system using
the pinch-analysis model with the formula bar showing the relevant formula that uses
Equation (10.2) to determine the temperature T3 .
Figure 10.5. The Excel sheet developed for the cogeneration system using the pinch-
analysis model
Figure 10.6 shows Solver set-up for minimising the total annualised cost with five
design variables; the compressor ratio (P_ratio), the turbine‟s inlet temperature (T_4),
the isentropic efficiency of the compressor (ƞ_c), the isentropic efficiency of the turbine
(ƞ_t), and the effectiveness of the air preheater (ε). The figure shows the constraints
imposed on the five design variables and the exhaust gas exit temperature (T_7). A
constraint is also imposed on the water temperature leaving the economiser (T_xg)
which has to be greater the water saturation temperature (T_9) by 15K in accordance
with Equation (10.5). Figure 10.7 shows the solution found by Solver with this setup.
Figure 10.6. Solver set-up for optimising the cogeneration system using the ε-model
Figure 10.7. Solver solution with ε model
The sheet using the pinch-analysis model was developed following the same procedure
and Solver was used to minimise the total annual cost. Figure 10.8 shows the optimised
values of the turbine inlet temperature (TIT) and compressor pressure ratio determined
by both heat-exchanger models. The figure shows that both models gave values that are
lower than the basic design values, but the ε-model led to lower values of the two
parameters than those obtained by the pinch model.
1530 10.5
ε-model ε-model
1520
Pinch model 10.0 Pinch model
Pressure ratio
1510 Basic Basic

TIT (K)
1500 9.5
1490
9.0
1480
1470 8.5
(a) (b)
Figure 10.8. Optimised values of the turbine inlet temperature (TIT) and pressure ratio
0.9 0.88
ε-model ε-model
Compressor efficiency
0.88
Turbine efficiency
0.9 Pinch model Pinch model

Basic 0.87 Basic
0.8
0.87
0.8
0.86
0.8 0.86
0.8 0.85
(a) (b)
Figure 10.9. Optimised isentropic efficiencies of the turbine and compressor
Figure 10.9 compares the optimised isentropic efficiencies of the compressor and
turbine to the basic design values. The figure shows that both models reduced the
turbine efficiency, but increased the compressor efficiency. Here also, the ε-model led
to lower values of the two parameters than those obtained by the pinch model.
Figure 10.10 shows the optimised the annualised costs and equipment purchase costs as
determined by the two models. Figure 10.10.a shows the optimised annualised initial
cost, fuel cost, and total annualised cost determined by the two models compared to
those of the basic design. The figure shows that both models predicted minor changes in
fuels costs but remarkably lower initial costs. The net effect is that both models gave
lower total costs compared to the basic design. While the ε-model reduced the total
annualised cost of the optimised design by 7%, the pinch model reduced it by 4%.
0.40 5.0E+06
ε-model ε-model
Annualised costs ($/s)
4.0E+06
0.30
Equpment costs ($)
Pinch model Pinch model
Basic 3.0E+06
0.20
2.0E+06
0.10 1.0E+06
0.00 0.0E+00
Z_c Z_t Z_cc Z_AHX Z_HRSG
C_fuel C_initial C_total
(a) (b)
Figure 10.10. Optimised equipment purchase costs and annualised system costs:
Z_c=compressor, Z_t=turbine, Z_c= combustion chamber, Z_AHX =air preheater
Figure 10.10.b shows that the two main items in the equipment cost are those of the
compressor and the turbine. For both items, the ε-model gave lower costs than those
obtained by the pinch-analysis model. However, for the HRSG, the air-preheater, and
the combustion chamber, which cost less than the compressor and turbine, the ε-model
resulted in higher costs than those determined by the pinch-analysis model.
Table 10.2 shows the optimised values obtained by the present model for selected
system parameters and their corresponding values obtained by Bejan et al [1]. Two
salient differences between the present analytical model and that used by Bejan et al [2]
force us to expect some differences between their results. The first reason is that Bejan
et al [1] treated the product of combustion as a mixture of ideal gases, but the present
model treated it as pure air. This explains why the mass flow rates of the air and the fuel
obtained by the present model for the same basic design setup are different from those
obtained by Bejan et al [1]. The second reason is that the economic optimisation of
Bejan etl al [1] is based on exergoeconomic analyses, while the present optimisation is
a least-cost optimisation. Figure 10.11 shows a qualitative comparison between the
present values and their corresponding values obtained by Bejan et al [1].
Table 10.2. Basic and optimised values of selected parameters in the cogeneration
system compared to the corresponding values obtained by Bejan et al [1]
Parameter Design Value Bejan et al [1] Present
TO CO analysis
Compressor pressure ratio 10.0 16.0 6.0 9.98
Compressor isentropic efficiency 86.0 88.0 81.1 82.3
Turbine isentropic efficiency 86.0 90.0 84.7 86.4
Air preheater outlet temperature 850 792.4 903 860.8
Combustion products temperature 1520 1550.0 1462.9 1482.9
Air mass flow rate 91.28(95.5)* 74.23 122.20 106.0
Fuel mass flow rate 1.64(1.52)* 1.51 1.83 1.57
* Bracketed design values are those calculated by the present model
Figure 10.11. Deviations of the optimised values obtained by the present model from
the corresponding basic design values compared to those reported by Bejan el al [1]
It is remarkable to note that the present model modified the design parameters in a
manner which is similar to that of the cost optimisation (CO) solution obtained by
Bejan et al [2]. The only exception in this respect is the turbine isentropic efficiency.
Note that the thermodynamic optimisation (TO) analysis increased the compression
ratio, the combustion-products temperature, and isentropic efficiencies of both the
compressor and turbine, but reduced the air and fuel flow rates in the basic design,
while the CO analysis did exactly the opposite.
10.3. The gas-turbine with inlet-air cooling

Cooling the air before the compressor of a gas-turbine improves the turbine‟s thermal
efficiency by reducing the lower temperature in the system. It also increases the
turbine‟s output by increasing the density and mass flow rate of the air. Figure 10.12
shows a gas-turbine power generation plant equipped a vapour- absorption refrigeration

(VAR) system for cooling the inlet air. The waste energy in the exhaust gas is extracted
in a HRSG to produce steam which is used to supply the heat needed for operating the
VAR system. The system utilises part of the energy carried away by the exhaust-gas to
raise the temperature of the compressed air before going to the combustion chamber.
Air preheater
7 6 5
HRSG
Combustion
products Natural gas
2
3 10
8
9 4
Steam loop
Combustion Net
chamber
power
Air Absorption
1 Air Gas turbine
0 refrigeration
compressor
Air cooler
Figure 10.12. Gas turbine with inlet-air cooling system (adapted from Burmeister [2])
Burmeister [2] analysed an air-cooled gas-turbine system such as that shown in Figure
10.12 with the objective of finding the sizes for the HRSG and the air preheater that
minimise the total cost (CT) of the system which is given by:
CT = fuel cost + air-cooler cost + boiler cost + air-preheater cost - sale of power
In terms of the relevant system parameters the objective functions can be written as:
CT = Cf Q + Cr Qr + CB AB + Cx Ax – Cw W (10.13)
Where Cf , Cr , CB , Cx, and Cw are unit costs for the fuel, the refrigeration system, the
boiler, the air preaheater, and the output power, respectively, Q and Qr stand for the
energy of the fuel burnt in the combustion chamber the amount of heat transferred to
the VAR system, respectively, and W is the power output of the plant.
10.3.1. The approximate analytical model

For an ideal Brayton cycle and applying the constant-specific heat method, Burmeister
[2] obtained the following equations for the three quantities W, Q, and Qr :
 
W / md C p T4  T5   1  T1 / T5 (10.14)
 
Q / md C p T4  T5   T4 1  T1 / T5  / T4  T5   x1  /1  x1  (10.15)
Qr /md T4  T5   COP / T4  T5 1  exp x 2  

T5  x1T1T4 / T5  / 1  x1   TB  (10.16)
The cooled-air temperature T1 in the above equations can be calculated from:
T1  T0 1  x1   COP1  exp x 2 T5  TB 1  x1  /

1  x1  COP1  exp x2 x1T4 / T5  (10.17)
Where x1 and x2 are defined as follows:
U x Ax
x1  (10.18)
md C p
U B AB
x2  (10.19)
md C p
Where Ux and UB are the overall heat-transfer coefficients of the air preheater and the
boiler, respectively, and the two areas Ax and AB are the surface areas of the air-
preheater and the boiler, respectively. For specified values of ma, Ux, UB, and Cp , the
values of x1 and x2 determine the heat-transfer areas of the regenerator and the boiler Ax
and AB , respectively.
The above analytical model enables the required values of x1 and x2 to be determined
easily by using Solver to minimise the total cost given by Equation (10.13) with x1 and
x2 as the changing variables. Alternatively, we can use Excel to analyse the ideal
Brayton cycle with the approximate constant specific-heat method and thus determine
the quantities W and Q. The quantities Qr and T1 can also be determined by modelling
the VAR system as a heat-exchanger. This approach, which would be more general
because the optimisation problem can be solved for any set of data, is left as an exercise
to the reader. However, the approximate method is inaccurate because of the
idealisations and approximations involved. As shown below, by using Thermax
property functions the Excel-based model uses the variable specific-heat method while
taking into consideration the deficiencies of the various system components. Moreover,
the model can be used to determine not only the optimum sizes of the air-preheater and
HRSG, but also the optimum steam flow rate in the VCR system.
10.3.2. The exact analytical model

This model uses the variable specific-heat method and follows the same steps described
in Section 9.1 for the top-cycle of the air-bottoming system. Accordingly, the net power
from the turbine is given by:
W  ma h4  h5   h2  h1  (10.20)
Where, ma is the mass flow rate of the air in the compressor. The exit temperature of the
compressed air from the regenerator (T_3) is determined by energy balance:
h3  h2  h5  h6  (10.21)
The amount of heat added in the combustion chamber is given by:
Q  ma h4  h3  (10.22)
The amount of heat-removal in the regenerator is given by:
Q x  ma h5  h6  (10.23)
The amount of heat-removal in the boiler (HRSG) is given by:
QB  m s h9  h8  = ms h fg (10.24)
Where ms is the mass flow rate of the steam and h fg is the latent heat for vaporisation of
water at the specified boiler temperature or pressure. The amount of heat-removal in the
VAR cooling system is given by:
Qr  COP  Q B (10.25)
Qr  ma h1  h0  (10.26)
The enthalpy of the chilled air can be obtained from:
h1  h0  Qr / ma (10.27)
The chilled-air temperature T1 can be determined from h 1 by using the relevant

Thermax property function. The enthalpy of the exhaust gas from the HRSG can be
determined from:
h7  h6  QB / m g (10.28)
The outlet temperature of the exhaust gas (T7 ) can be obtained from h 7 by using the
relevant Thermax property function. The regenerator effectiveness (ε) can be calculated
from:
T3  T2
 (10.29)
T5  T2
Note that this analytical model does not require a heat-transfer submodel for the
regenerator or HRSG like the cogeneration system. The following section shows how
the Excel-based model can be used with Thermax property functions and Solver for
optimisation analyses of the air-cooled gas-turbine system.
10.4. Optimisation analyse s of the gas-turbine with inlet-air cooling

The gas-turbine considered for this analysis is similar to that described by Burmeister
[2] which has a mass flow rate of air of 45 kg/s, pressure ratio of 3.8, and turbine-inlet
temperature (TIT) of 1100K. The ambient air temperature To is 300 K and the steam
temperature in the HRSG (TB ) is 380K (107o C). The following data has also been
adopted from Burmeister [2]:
COP = 1, UB = 0.25 kW/m2 K, Ux = 0.082 kW/m2 K, Cw = $350/kW, Cf = $50/kW, Cr =

$100/kW, CB = $1000/m2 , Cx = $100/m2
While Burmeister [2] used a constant value or T5 of 750K, in the present analyses T5 is
calculated by the analytical model. For the purpose of comparison with the results of
Burmeistyer [2], the analysis will initially be conducted using the ideal Brayton cycle.
10.4.1. Optimisation analysis with the ideal cycle

Figures 10.13 and 10.14 shows the Excel sheet developed for analysing the air-cooled
system using the analytical model described above. The data part of the main sheet
stores the gas-turbine particulars as well as the pressure and temperature of the air at the
atmospheric condition, the temperature of the steam in the boiler, and the COP of the
VAR system. Note that isentropic efficiencies of the compressor and turbine are made
100% because the cycle is ideal. Also note that the data part that includes the pressure
losses in the combustion chamber, the air side of the regenerator, the gas side of the
regenerator, and the HRSG gives them zero values. The back sheet shown in Figure
10.14 calculates the areas of the regenerator (A_x) and the boiler (A_B) and then
determines the total cost (Ctotal) according to Equation (10.13) as shown in the formula
bar of this sheet.
The calculations part in the main sheet starts with assumed values for the air
temperature after the regenerator (T_6) and the mass flow rate of steam in the VCR
system (m_steam). The pressures and temperatures of the air and products of
combustion are then determined at each state of the cycle. The formula bar in Figure
10.13 reveals the formula used to calculate the temperature T5 using Thermax property
function regenerator effectiveness (ε) from Equation (10.29).
Figure 10.13. The main Excel sheet for the air-cooled system with the ideal cycle
Figure 10.14. The back Excel sheet for the air-cooled system
Solver was used to determine the values of T6 and ms that yield the lowest total cost
with the Parameters setup shown in Figure 10.15. A number of constraints have been
imposed on the optimisation process in order to exclude physically or economically
unacceptable solutions. The first constraint shown in the figure limits the lowest
chilled-air temperature to 265K (-8o C). The following four constraints limit the range
for T6 from 400K to 900K and the range for ms from 0.0 kg/s to 10 kg/s. The last two
constraints limit the regenerator‟s effectiveness, ε, to be in the range of 50% to 85%.
Figure 10.16 shows the optimum solution found by Solver and Table 10.3 compares the
solution obtained by Solver (variable-k) for selected key parameters with their
corresponding values obtained by using the approximate method (constant-k) described
in Section 10.3.1. Although the figures in the table show some differences, they validate
the present model since the difference between the optimum total costs is less than 2%.
Figure 10.15. Solver set-up for optimising the inlet-air cooling system
Figure 10.16. Solver solution for the inlet-air cooling system with the ideal cycle
Table 10.3. Comparison of the results obtained by the constant specific heat model
(constant-k) and the present variable specific heat model (variable-k)
Constant k Variable k
Regenerator exit temperature T6 (K) 614.0 466.0
Chilled-air temperature T1 (K) 280.84 277.48
Specific netw work, W, (kJ/kg) 218.9 228.2
Specific heat-addition, Q, (kJ/kg) 446.2 419.7
Power output (MW) 9.85 10.27
Thermal efficiency 0.49 0.54
2
Area of regenerator (m ) 1379.1 3127.4
Area of boiler (m2 ) 29.1 11.7
Total cost ($) -2,190,999 -2,224,579
10.4.2. Optimisation analysis with actual cycle

Figure 10.17 shows the sheet developed for the actual cycle in which the isentropic
efficiencies are not allowed to reach 100%, but assigned initial values of 86% for each.
Note that the data part also includes the pressure losses in the combustion chamber, the
air side of the regenerator, the gas side of the regenerator, and the HRSG. These figures
are borrowed from the cogeneration case considered earlier as given by Bejan et al [1].
Comparison with Figure 10.13 for the ideal cycle shows that these losses cause
significant drop in the power of the turbine and its thermal efficiency which result in a
significant increase in the total annual cost. Solver was used to determine the values of
T6 and ms that yield the lowest total cost with the same Parameters setup shown in
Figure 10.15. Figure 10.18 shows the optimum solution found by Solver and Table 10.4
shows values of the key parameters as determined by Solver compared to their
corresponding values obtained by the ideal cycle as shown in Figure 10.16.
Figure 10.17. The main Excel sheet for the air-cooled system with the actual cycle
Figure 10.18. Solver solution for the inlet-air cooling system with the actual cycle
Table 10.4. Key parameters in the optimised inlet-air cooling system with ideal cycle
and the actual cycle
Parameter Ideal cycle Actual cycle
Regenerator exit temperature T6 (K) 466.06 520.06
Mass flow rate of steam (kg/s) 0.453755 0.658
Chilled-air temperature T1 (K) 277.48 267.3
Compressed air temperature T3 (K) 724.7555 757.97
Exhaust-gas exit temperatur T7 (K) 444.14 488.53
Power output (MW) 10.27 5.68
Thermal efficiency (%) 54.4 32.9
Effectiveness of regenerator 0.85 0.77
Area of regenerator (m2 ) 3127.352 1896.115
Area of boiler (m2 ) 11.6716 14.83757
Total cost ($) -2,224,579 -771,466
The figures in Table 10.4 show that actual-cycle model increased the optimum
intermediate temperature from 466,06K to 520.06K. As a result, the area of the
regenerator is reduced while that of the boiler is increased and the required mass flow
rate of steam increased from 0.454 kg/s to 0.658 kg/s. The actual cycle decreased the
power output of the gas turbine from 10.27 to 5.68 MW and the thermal efficiency from
54.4% to 32.9%. Therefore, the optimum total cost increased from – 2.22M to – 0.77M.
Also note that the required chilled-air temperature, T1 , dropped from 277.48K to
267.3K, but the temperature of the existing gas, T7 , increased from 444.14K to
488.53K.
10.5. Closure
This chapter dealt with optimisation analyses of two options for recovering and utilising
the waste heat in the gas-turbines‟ exhaust gas by installing: (a) a cogeneration system
and (b) a vapour-absorption refrigeration system for cooling the inlet air. While the
Thermax add-in provided the property functions needed for exact analyses of the
thermodynamic cycles relevant to the two systems, Solver provided the required tool
for the multi-variable constrained optimisation process. The analysis of the
cogeneration system compared the optimised solutions obtained by an extended
effectiveness model and a pinch-analysis model for the regenerator. Compared to the
basic design, the ε-model reduced the total annualised cost of the optimised design of
the cogeneration system by 7% while the pinch model reduced it by 4%.
The analysis of the air-cooled system showed that the Excel-based model, which takes
into consideration the deficiencies of the various system components, can be used to
determine the optimum sizes of the air-preheater and HRSG as well as the optimum
steam flow rate in the VCR system. Comparison of the optimised solutions obtained
with the ideal cycle and the actual cycle showed that the actual cycle gave more
realistic estimations by decreasing the power output thermal efficiency of the gas
turbine, which results in a reduced total annual revenue from the plant.
References
Wiley & Sons, 1996.
[2] L.C. Burmeister, Elements of Thermal-Fluid System Design, Prentice Hall, 1989.
[3] J. E. Kraft, Turbine Inlet Cooling System Comparisons, Energy-Tech, 36-37,
August 2006
Appendices
Appendix A: Properties of air and water at atmospheric pressure
Table A.1. Properties of air at at various temperatures †

T ρ cp μ ν k α
o Pr
C kg/m3 kJ/kg.K kg/m.s m2 /s W/m· o C m2 /s
-150 2.866 0.983 8.636x10-6 3.013x10-6 0.01171 4.158x10-6 0.7246
-100 2.038 0.966 1.189x10-5 5.837x10-6 0.01582 8.036x10-6 0.7263
-50 1.582 0.999 1.474x10-5 9.319x10-6 0.01979 1.252x10-5 0.7440
-40 1.514 1.002 1.527x10-5 1.008x10-5 0.02057 1.356x10-5 0.7436
-30 1.451 1.004 1.579x10-5 1.087x10-5 0.02134 1.465x10-5 0.7425
-20 1.394 1.005 1.630x10-5 1.169x10-5 0.02211 1.578x10-5 0.7408
-10 1.341 1.006 1.680x10-5 1.252x10-5 0.02288 1.696x10-5 0.7387
0 1.292 1.006 1.729x10-5 1.338x10-5 0.02364 1.818x10-5 0.7362
5 1.269 1.006 1.754x10-5 1.382x10-5 0.02401 1.880x10-5 0.7350
10 1.246 1.006 1.778x10-5 1.426x10-5 0.02439 1.944x10-5 0.7336
15 1.225 1.007 1.802x10-5 1.470x10-5 0.02476 2.009x10-5 0.7323
20 1.204 1.007 1.825x10-5 1.516x10-5 0.02514 2.074x10-5 0.7309
25 1.184 1.007 1.849x10-5 1.562x10-5 0.02551 2.141x10-5 0.7296
30 1.164 1.007 1.872x10-5 1.608x10-5 0.02588 2.208x10-5 0.7282
35 1.145 1.007 1.895x10-5 1.655x10-5 0.02625 2.277x10-5 0.7268
40 1.127 1.007 1.918x10-5 1.702x10-5 0.02662 2.346x10-5 0.7255
45 1.109 1.007 1.941x10-5 1.750x10-5 0.02699 2.416x10-5 0.7241
50 1.092 1.007 1.963x10-5 1.798x10-5 0.02735 2.487x10-5 0.7228
60 1.059 1.007 2.008x10-5 1.896x10-5 0.02808 2.632x10-5 0.7202
70 1.028 1.007 2.052x10-5 1.995x10-5 0.02881 2.780x10-5 0.7177
80 0.9994 1.008 2.096x10-5 2.097x10-5 0.02953 2.931x10-5 0.7154
90 0.9718 1.008 2.139x10-5 2.201x10-5 0.03024 3.086x10-5 0.7132
100 0.9458 1.009 2.181x10-5 2.306x10-5 0.03095 3.243x10-5 0.7111
120 0.8977 1.011 2.264x10-5 2.522x10-5 0.03235 3.565x10-5 0.7073
140 0.8542 1.013 2.345x10-5 2.745x10-5 0.03374 3.898x10-5 0.7041
160 0.8148 1.016 2.420x10-5 2.975x10-5 0.03511 4.241x10-5 0.7014
180 0.7788 1.019 2.504x10-5 3.212x10-5 0.03646 4.593x10-5 0.6992
200 0.7459 1.023 2.577x10-5 3.455x10-5 0.03779 4.954x10-5 0.6974
250 0.6746 1.033 2.760x10-5 4.091x10-5 0.04104 5.890x10-5 0.6946
300 0.6158 1.044 2.934x10-5 4.765x10-5 0.04418 6.871x10-5 0.6935
350 0.5664 1.056 3.101x10-5 5.475x10-5 0.04721 7.892x10-5 0.6937
400 0.5243 1.069 3.261x10-5 6.219x10-5 0.05015 8.951x10-5 0.6948
450 0.4880 1.081 3.415x10-5 6.997x10-5 0.05298 1.004x10-4 0.6965
500 0.4565 1.093 3.563x10-5 7.806x10-5 0.05572 1.117x10-4 0.6986
600 0.4042 1.115 3.846x10-5 9.515x10-5 0.06093 1.352x10-4 0.7037
700 0.3627 1.135 4.111x10-5 1.133x10-4 0.06581 1.598x10-4 0.7092
800 0.3289 1.153 4.362x10-5 1.326x10-4 0.07037 1.855x10-4 0.7149
900 0.3008 1.169 4.600x10-5 1.529x10-4 0.07465 2.122x10-4 0.7206
1000 0.2772 1.184 4.826x10-5 1.741x10-4 0.07868 2.398x10-4 0.7260
1500 0.1990 1.234 5.817x10-5 2.922x10-4 0.09599 3.908x10-4 0.7478
2000 0.1553 1.264 6.630x10-5 4.270x10-4 0.11113 5.664x10-4 0.7539
†
Adopted from Y. A. Cengel and A. J. Ghajar, Heat and Mass Transfer:
Fundamentals and Applications. 5th edition, McGraw-Hill, 2015.
Table A.2: Properties of water (saturated liquid) at various temperatures
o cp ρ μ ×104 k
C o 3 Pr
kJ/kg. C kg/m kg/m.s W/m· o C
0 4.225 999.8 17.90 0.566 13.25
5 4.207 999.7 15.26 0.576 11.15
10 4.195 999.2 13.10 0.585 9.40
15 4.187 998.7 11.39 0.594 8.03
20 4.180 997.6 10.08 0.602 7.00
25 4.179 996.3 8.96 0.611 6.13
30 4.176 995.3 8.03 0.619 5.41
35 4.174 994.0 7.24 0.627 4.83
40 4.174 992.0 6.56 0.633 4.33
45 4.174 990.1 6.00 0.639 3.92
50 4.175 988.2 5.52 0.645 3.57
55 4.179 985.5 5.09 0.650 3.27
60 4.179 983.3 4.71 0.654 3.01
65 4.183 980.6 4.34 0.659 2.76
70 4.185 977.9 4.07 0.664 2.57
75 4.189 974.8 3.81 0.667 2.39
80 4.193 971.6 3.57 0.671 2.23
90 4.201 965.3 3.19 0.676 1.98
100 4.211 958.3 2.83 0.682 1.76
110 4.223 950.9 2.56 0.685 1.59
120 4.237 942.9 2.34 0.685 1.45
140 4.276 926.1 1.96 0.685 1.23
160 4.326 907.0 1.74 0.681 1.11
180 4.382 886.7 1.54 0.676 1.02
200 4.452 864.3 1.39 0.667 1.00
230 4.576 828.4 1.21 0.648 0.86
260 4.731 785.2 1.07 0.616 0.83
Appendix B: Methods of heat-transfer analyses for heat-exchangers and fins
Heat exchangers and extended surfaces (fins) are two important equipment for energy
systems that demonstrate the application of heat-transfer principles. This appendix
provides the basic equations needed for their design and application analyses. With
respect to heat-exchangers, there are two types of analyses: sizing analyses and rating
analyses. The appendix explains the difference between the purposes and
methodologies of these two types of analyses.
B.1. Sizing analysis of heat-exchangers and the LMTD method

Consider the double-pipe heat-exchanger shown in Figure B.1 in which the cold and the
hot fluids pass in opposite directions. Figure B.2 shows the variation of the fluid
temperatures in this counter-flow heat-exchanger.
Hot in
, Thi
Cold in
,Tci
Cold out Hot out

Tco Tho
Figure B.1. Schematic of a counter-flow double-pipe heat-exchanger
∆T1
Temperature
Thi
Tco
∆T2
Tho
Tci
Figure B.2. Temperature variation in a counter-flow double-pipe heat-exchanger
Neglecting any exchange of heat with the surrounding environment, the conservation
energy principle implies that in a steady operation the rate of heat-transfer lost by the
hot stream ( Q ) is equal to that gained by the cold stream:
Q  m h c ph Thi  Tho   m c c pc Tco  Tci  (B.1)
Where m  , cp , Ti, and To are the mass flow rate, the specific heat, the inlet temperature,
and the exit temperature, respectively, and the suffices h and c refer to the hot stream
and the cold stream, respectively.
The purpose of a sizing analysis is to determine the required surface area for a heat
exchanger. The effective surface area of the heat-exchanger (A) is related to the rate of
heat transfer between the two streams ( Q ) by the following equation:
Q  UATlm (B.2)
Where U is the overall heat-transfer coefficient of the heat-exchanger and ∆Tlm is the
log-mean temperature difference given by [1]:
T1  T2
Tlm  (B.3)
lnT1 / T2 
Where ∆T1 and ∆T2 are the temperature differences between the two fluids at both ends
of the heat exchanger as shown in Figure B.2.
Combining Equation (B.1) and Equation (B.2) gives:
 h C ph Thi  Tho   m
UAFTlm  m  c C pc Tco  Tci  (B.4)
Equations (B.1) to (B.4) form the log-mean temperature difference (LMTD) method
which is useful for sizing analyses of a heat exchanger with a known overall heat
transfer coefficient and fluid streams with known mass flow rates, specific heats, and
inlet and outlet temperatures. Alternatively, the LMTD method can be used to
determine the overall heat-transfer coefficient for a heat-exchanger with a known area.
For multi-path heat exchangers and compact heat exchangers, Equation (B.2) is
modified as follows [1]:
Q  UAFTlm (B.5)
Where F is a correction factor the value of which depends on the heat-exchanger type
as well as the temperatures of the two fluids. Values of F can to be obtained from
relevant charts or calculated from relevant formulae. However, F = 1 for simple types
of heat exchangers and for condensers and boilers of all types. The basic requirement
for applying the LMTD method is that the inlet and outlet temperatures of both the hot
and cold fluid steams are given.
B.2. Rating analysis of heat-exchangers and the effectiveness -NTU method

The objective of a rating analysis is to determine the exit temperatures of the two fluid
streams for a given heat-exchanger. Since the exit temperatures are not known in
advance, the LMTD method described cannot be used. Instead, the analyses use the
effectiveness-NTU method that determines the rate of heat-transfer from [1]:
Q  Q max (B.6)
Where ε is the heat-exchanger‟s effectiveness, Q is the actual heat-transfer rate in the

heat exchanger, and Q max is the maximum possible heat-transfer rate which is given by:
Q max  C min Th,i  Tc,i  (B.7)
Where Cmin is the minimum heat capacities Ch and Cc among the two fluid streams
defined as:
C h  m h c ph (B.8)
C c  m c c pc (B.9)
If the value of the heat-exchanger effectiveness ε is given, then the actual rate of heat-
transfer in the exchanger can be calculated from Equations (B.6) and (B.7).
The determination of ε requires using appropriate formulae or dimensionless charts. For

condensers and boilers of all types, ε can be obtained from [1]:
  1  exp NTU  (B.10)
Where NTU is the number of transfer units given by:
NTU  UA / C min (B.11)
Once the actual heat-transfer rate ( Q ) is known, the outlet temperatures of the hot and
cold fluids can be obtained from:
Tho  Thi  Q / c ph (B.12)

Tco  Tci  Q / c pc (B.13)
B.3. Fins, fin efficiency, and fin effectiveness

The rate of heat transfer by convection between a solid surface at a temperature Ts and a
flowing fluid at a temperature T∞ is governed by Newton‟s law of cooling:
Q conv  hAs Ts  T  (B.14)
Where, h is the heat-transfer coefficient and As is the surface area. The above equation
indicates that the rate of heat transfer can be increased by increasing the heat-transfer
coefficient (h) and/or the surface area (As ). When increasing the heat-transfer
coefficient by increasing the speed of the fluid or by using a different fluid is not
possible, increasing the area becomes the only viable way. The surface area can be
increased by attaching fins to the surface as shown in Figure B.3.
T∞ , h qconv Insulation
Tb qcond
Figure B.3 Schematic of a straight rectangular fin with an insulated boundary

condition at the tip
The fin efficiency

Figure B.3 shows a fin of length L attached to a vertical surface that has a temperature
Tb . At steady state, the heat transfer entering the fin from its base must be balanced by
the heat transfer leaving the surface of the fin by convection plus any heat transfer
leaving the fin at x=L.
The highest temperature in the fin is at its base, while the lowest temperature is at its
tip. Therefore, both the temperature and the rate of heat-transfer from the fin drop
gradually along the length of the fin. The fin efficiency (ηfin ) is defined as the ratio
between the actual heat-transfer rate from the fin (Qfin ) to the maximum possible heat
transfer rate (Qfin,max) that would be obtained if the entire fin were at the base
temperature, i.e.,
Q fin
 fin   (B.15)
Q fin ,max
The maximum heat transfer from the fin (Qfin,max) is given by:
Q max  hAfin Tb  T  (B.16)
Where Afin is the fin‟s surface area (Afin =pL).
The actual rate heat transfer from the fin (Qfin ) can be obtained by applying Fourier law
at the base of the fin [1, 2]:
dT
Q   kAc (B.17)
dx x 0
For the straight rectangular fin with an insulated tip as shown in Figure B.3, Equation
(B.17) gives [1, 2]:
Q  hPkA fin  0 tanh mL (B.18)
Combining Equations (B.15), (B.16), and (B.18) gives the following equation for the
fin efficiency of a straight rectangular fin with an insulated tip:
hpkAc  0 tanh mL tanh mL

 fin  = (B.19)
hpL o mL
Table B.1 shows the fin efficiency for the straight rectangular fin with three possible
boundary conditions at the tip which are: (i) insulated tip, (ii) specified temperatures,
and (iii) convection boundary condition [1, 2].
Table B.1.The fin efficiency for a straight rectangular fin with three boundary
conditions at the tip
Case Boundary condition at x = L  fin
dT tanh mL
(i)
dx x 0
mL
1
(ii) TL = T∞
mL
tanh mLc
;
 TL  T 
dT h
(iii) mLc
dx xL k
Lc = L +t/2
Fin efficiencies for other fin shapes can be obtained from relevant equations or non-
dimensional charts [1,2]. The usefulness of the fin-efficiency concept is that, if the fin
efficiency is known, the actual heat transfer from the fin can be calculated from:
Q fin   fin  Q fin ,max (B.20)
The fin effectiveness

Another criterion that is used to measure the performance of fins is the fin effectiveness
which is defined as the ratio between the actual rate of heat transfer with the fin to the
rate of heat transfer from the base without the fin, i.e.:
Q fin Q fin
 fin   (B.21)
Q no fin hAb Tb  T 
While the value of the fin efficiency is always less than 100%, the value of the fin
effectiveness is always more than 100%. A fin effectiveness of more than 100%
indicates that the fin has enhanced the rate of heat-transfer, while an effectiveness of
100 indicates that the addition of fin to the surface has no effect on the heat transfer. An
effectiveness of less than 100 is not possible.
References
[1] Y. A. Cengel and A. J. Ghajar, Heat and Mass Transfer: Fundamentals and
Applications. 4th edition, McGraw-Hill, 2011
[2] J. P. Holman, Heat Transfer, 10th edition, McGraw-Hill. 2010
Appendix C: The numerical tools provided by Thermax
The Thermax add-in provides, in addition to its functions for fluid properties, a
Newton-Raphson solver for non-linear equations and two interpolation functions for
tabulated data. This appendix shows why these numerical tools are needed in
thermofluid analyses and how they are used with the help of relevant examples.
C.1. The Newton-Raphson solver “NRM”

Excel‟s cell formulae become too restrictive and inconvenient to use when dealing with
iterative solutions and optimisation analyses that involve implicit equations. This
occurs, for example, when dealing with type-2 or type-3 pipe-flow problems, which
themselves require iterative solutions, by using the following Colebrook-White (C-W)
equation that determines the Darcy friction factor (f) in a turbulent pipe flow:
1  / D 2.51 
 2.0 log10    (1.25)
f  3.7 Re f 
 
In this case, Excel has to deal with two nested iterations; an inner iteration to determine
f and an outer iteration to determine the pipe‟s diameter or the flow rate.
Another nonlinear equation in thermofluid analyses is the following Benedict-Webb-

Rubin (BWR) equation of state [1]:
P = Ru T   B0 Ru T  A0  C 0  1  bRu T  a  a  c 1   e  / v~ 2

(C.1)
v~  T 2  v~ 2 v~ 3 v~ 6 v~ 3T 2  v~ 2 
Where Ru is the universal gas constant, P is the absolute pressure, T is the absolute
~
temperature, and v is the molar specific volume. The BWR equation is one of the most
~
accurate equations of state, but it is implicit in v and requires an iterative solution to
determine the molar specific volume. Therefore, its use in thermodynamic optimisation
analyses faces a similar problem to that of the C-W equation in type-2 or type-3 flow
problems.
The Newton-Raphson solver provided by Thermax is a custom function developed with

VBA for solving nonlinear equations such as the C-W equation and the BWR equation.
To illustrate the use of this “NRM” solver, let us use it to determine the error of the
ideal-gas law compared to the BWR equation of state by calculating the specific
volume of carbon dioxide (CO 2 ) at 0.2 MPa and temperatures in the range 273 – 373K.
Figure C.1 shows the Excel sheet developed for this purpose and Table C.1 reveals the
formulae used in the sheet.
Figure C.1. Excel sheet for determining the specific volume by the ideal-gas law
compared to the BWR equation of state
Table C.1. The formulae used in the Excel sheet using the NRM solver
T v_ideal v_BWR error_v_ideal
273 =Ru*D3/P =NRM("BWR",E3,P,D3) =ABS(F3-E3)/F3*100
The formula bar in Figure C.1 reveals the following Excel formula in cell F3 that uses
~
the NRM solver to calculate v at 273K:
=NRM("BWR",E3,P,D3)
Note that the function NRM requires four input arguments: "BWR", E3, P, and D3. The
first argument, “BWR”, is the name of another user-defined VBA function for the
particular nonlinear equation to be solved; which in this case is the BWR equation. The
second argument, E3, is an initial guess for the dependent variable in the nonlinear
equation; i.e., v~ in Equation (C.1). The third and fourth arguments, P and D3, are
values of two independent variables in the nonlinear equation; which are P and T in
Equation (C.1). For the C-W equation, they are the values of ε/D and Re. The function
“BWR” that applies the BWR equation is listed below:
Function BWR(x, P, T)
Dim Ru, a, A0, b, B0, c, C0, alfa, gama
'Carbon dioxide, CO2
Ru = 8314
a = 13.86
A0 = 277.3
b = 0.00721
B0 = 0.04991
c = 1511000.0
C0 = 14040000.0
alfa = 0.0000847
gama = 0.00539
BWR = P - Ru * T / x + (B0 * Ru * T - A0 - C0 / T ^ 2) / (x ^ 2) + (b * Ru * T
- a) / x ^ 3 + a * alfa / x ^ 6 + c / (x ^ 3 * T ^ 2) * (1 + gama / x ^ 2) * Exp(-
gama / x ^ 2)
End Function
The different constants given in the BWR function are those of carbon dioxide (CO 2 ).
Note that the value determined by the ideal-gas law in cell E3 is used as an initial guess
for the BWR equation. Figure C.2 shows the deviation of the ideal-gas law from the
BWR equation in estimating the specific volume.
0.50
Error of the ideal gas law(%)
0.45
0.40
0.35
0.30
250 300 350 400
Temperature (K)
Figure C.2. Errors in the specific volume estimations by the ideal-gas law compared to
the BWR equation of state for carbon-dioxide
C.2. The interpolation functions “Interpol1” and “Interpol2”

The two functions provided by Thermax for the interpolation of tabulated data, called
Interpol1 and Interpol2, use linear and quadratic equations, respectively. These
functions enable tabulated fluid properties to be used in parametric studies and
optimisation analyses. For example, Figure C.3 shows an Excel sheet in which three
properties of engine oil are given at various temperatures: the density (ρ), specific heat
(cp ), and kinematic viscosity (ν).
Figure C.3. Using the interpolation functions for tabulated data
The interpolation functions can be used to calculate the three oil properties at any
temperature within the range of tabulated data. Both Interpol1 and Interpol2 require
four input arguments referred to as: X, XX, YY, and Ndata. Their meanings are as
follows:
- X passes the value of the independent property (i.e. temperature) at which the
value of the dependent property (i.e. viscosity) is to be determined.
- XX and YY are vectors that store the tabulated data (in this case, the
temperature, B5:B13, and viscosity values, E5:E13.
- Ndata passes the number of entries in the tabulated data. For the oil properties
shown in Figure C.3, Ndata = 9.
The formula bar in Figure C.3 reveals the formula in which the linear interpolation
function Interpol1 is used to determine the viscosity at 90o C. The following example
shows how the interpolation functions can be used in a typical analysis.
Example C-1. Effect of oil temperature on the drag force over a flat plate
Engine oil at 20°C flows over the upper surface of a 5-m-long flat plate as shown in
Figure C.4. If the oil velocity is 2 m/s, plot the variation of the total drag force (FD ) per
unit width of the entire plate FD with oil temperatures in the range 40 - 150o C.
T∞ = 40 – 150o C
V = 2 m/s
Oil FD
A Ts = 20o C
L=5m
Figure C.4. Flow of oil over a flat plate (adapted from Cengel and Ghajar [2])

The drag force (FD ) over a flat plate is due to friction, which is given by [2]:
FD  C f AV 2 / 2 (C.2)
Where Cf is the friction coefficient (not to be confused with the friction factor f), A is
the surface area of the plate, and ρ and V are the density and velocity of oil,
respectively.
The value of the friction coefficient depends on whether is flow is laminar or turbulent
as indicated by the Reynolds number (ReL). For the flow over a flat plate, the critical
Reynolds number above which the flow becomes turbulent is about 5x105 . For laminar
flows, the friction coefficient is given by:
C f  1.328 Re L0.5 Re L  5  10 5 (C.3)
For turbulent flows, it is given by:
C f  0.074 Re L0.2 5  10 5  Re L  10 7 (C.4)
Note that fluid properties are evaluated at the average temperature (Ts +T∞ )/2, e.g., if the
oil temperature is 60o C, the oil viscosity and density are evaluated at 40o C.
Solution with Excel

Figure C.5 shows the Excel sheet developed for this example. The data part shows the
given information concerning the oil temperature (T_oil), plate temperature (T_plate),
oil velocity (Velocity), and plate length (Length). Based on the specified oil
temperature, the sheet calculates the average temperature (T_average) and the
corresponding oil density (ρ_oil) and kinematic viscosity (ν_oil). The sheet then
calculates the Reynolds number (Re) and friction coefficient (Cf). The single final
result, which is the drag force (F_D), is shown on the right-side of the sheet. The figure
reveals the formulae entered in each cell in the calculations part and the formula bar
reveals that used for determining the drag force in cell I2.
Figure C.5. The Excel sheet developed for Example C-1
The friction coefficient is calculated depending on the value of the Reynolds number by
using the following nested-If formula in cell F7:
Cf=IF(Re<500000,1.328*Re^-0.5,0.074*Re^-0.2)
The above formula determines whether to use Equation (C.3) or Equation (C.4) to
calculate the friction coefficient depending on the value of the Reynolds number.
Figure C.5 shows the calculations at an oil temperature of 40o C, which is the initial oil
temperature in the required range. At the average temperature, which is 30 o C, the sheet
uses the linear interpolation function Iinterpol1 to determine the values of the oil
density and viscosity. At this temperature the calculated drag force, F_D, is 88.445 N.
By inserting a different value for the oil temperature, say 50o C, the sheet automatically
updates its calculations. Figure C.6 shows a plot of the drag force determined at
different oil temperatures in the range 40o C to 150o C.
The linear and quadratic interpolation functions are useful in thermofluid analyses not
only for determining fluid properties, but also for the interpolation of other types of
tabulated data that is required by such analyses. Table C.2 shows the cost per metre of
galvanised-steel air-conditioning ducts for different diameters of the duct. By
permitting automatic determination of the duct diameter from the tabulated data, the
interpolation functions are useful for optimisation analyses that aim to determine the
optimum duct diameter.
100
80
Drag force (N)
60
40
20
0
0 50 100 150 200
Oil temperature (oC)
Figure C.6. Variation of the drag force with oil temperature
Table C.2. Cost of galvanised-steel air-conditioning ducts

Duct diameter Cost per linear
(m) metre ($)
0.10 9.0
0.15 11.5
0.20 14.5
0.25 17.0
0.30 22.5
0.35 29.0
0.40 34.0
0.45 40.0
0.50 50.0
References
[2] Y.A. Cengel and A.J. Ghajar. Heat and Mass Transfer: Fundamentals and
Applications. McGraw Hill, 2011
Exercises
10. Using the data for properties of air at atmospheric pressure given in Appendix A,
develop an Excel sheet that can be used to determine the kinematic viscosity of air
at any given temperature in the range 200 – 1000K by using:
a. The trendline feature of Excel

b. The linear-interpolation function (Interpl) provided by Thermax.
11. Develop a VBA function to determine the friction factor from the Colebrook-
White equation and use it with the NRM solver to determine the frictional losses
(h f ) in a circular pipe that carries air at 20o C with the following data:
D = 25 cm, L = 150 m, V = 7 m/s, k s = 0.045 mm.

Appendix D: Optimisation analysis of an air-conditioning duct
This appendix supplements Chapter 4 that dealt with optimisation analyses of

thermofluid systems in general and Chapter 5 that dealt with thermal insulation in
particular. The appendix preents a multi-variable optimisation analysis of an air-
conditioning duct with two sections to be insulated. For the purpose of comparison, the
case is solved by both the GRG Nonlinear method and the Evolutionary method.
Case description
Conditioned air enters the 30-m air-handling duct shown in Figure D.1 at P1 = 100 kPa
and T1 =80o C. The flow rate at the entrance is Q1 = 0.7 m3 /s, part of which (Q3 =0.3
m3 /s) is discharged at a point 16 m downstream of the duct entrance. The remaining
part (Q2 = 0.4 m3 /s) is discharged at the end. Ambient temperature (T∞2 ) is 15o C and the
outside heat-transfer coefficient h 2 is 30 W/m2 .o C. The air duct is to be assembled from
1-m-long prefabricated units made of 3-mm galvanised sheet metal.
L1 L2
Q1 Q2
D1 D2
Q3
Figure D.1. The uninsulated air-conditioning duct with two sections
To minimise energy losses to the surroundings, it is decided to insulate the duct with
fibre-glass. It is required to determine the diameters of the two duct sections (D1 and
D2 ) and the thicknesses of insulation (δ1 and δ2 ) that minimise the total owning cost
based on the data provided below.
Duct data:
Thermal conductivity of the duct material (k d ) = 18 W/m.o C
Duct roughness (ε) = 0.045mm
Cost of 1-m unit (cd ) is as shown in Table D.1.
Table D.1. Unit cost of the duct body [1]

Diameter (Di ) in m 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Cost per 1 m length ($) 9 11.5 14.5 17 22.5 29 34 40 50
Insulation (Fibre-glas) data:

Thermal conductivity (k s ) = 0.04 W/m.o C
Insulation cost (cs1 ): 30 $/m2 per cm of insulation
Labour cost (cs2 ): 10 $/m2 (irrespective of thickness)
Energy costs:
Cost of electricity (cE ): 0.12 $/kWh
Cost of fuel (cF): 0.5 $/therm (1 therm = 105500 kJ)
Capital recovery factor (i) = 0.15
Operation data:
365 days per year 24 hours per day
Janna [2] considered the case of an un-insulated duct that involved only two design
parameters and could be solved analytically. With the addition of insulation to the duct
the optimisation process in the present case involves four parameters which make the
case more difficult to solve analytically than the case considered by Janna [2].

A larger duct diameter reduces the friction loss in the duct and, therefore, the cost of
electricity needed by the circulation fan. However, a smaller duct diameter gives a
smaller surface area, which reduces the heat loss from the duct and the cost of
insulation. Since the duct has two sections with different diameters, the optimisation
process involves four parameters; D1 , δ1 , D2 and δ2 . The total annual cost of the
insulated air-duct (CTotal ) consists of three components as expressed by:
CTotal  C I  i  C E  C F ($) (D.1)
Where:
CI = initial cost of the duct itself plus the cost of insulation

i = capital recovery factor
CE = cost of electricity consumed by the fan in order to overcome friction in the duct
CF = cost of fuel needed to make-up for the heat loss to the surrounding air
How the three components of the total cost are evaluated is explained below.
a) Initial cost
The initial cost (CI) has two parts: (1) the cost of the duct itself (Cduct ) and (2) the cost
of insulation (Cins ). The two parts are given by:
Cduct = D1  c d ,1   L1  D2  c d , 2   L2 (D.2)
Cins = D1 L1    1  D2 L2    2  cs1  D1 L1   D2 L2  c s 2 (D.3)
Where:
D1 = Diamter of the first section of the duct (m)

D2 = Diamter of second section of the duct (m)
L1 = Length of the first section of the duct (m)
L2 = Length of second section of the duct (m)
δ1 = thickness of insulation in the first section of the duct (m)
δ2 = thickness of insulation in the second section of the duct (m)
c d = cost of 1-m duct unit which depends on the diameter ($/m)
2 2
c s ,1 = cost of insulation per m that varies with insulation thickness ($/m .cm)
2 2
c s , 2 = cost of labour per m that depends on insulated surface area only ($/m )
b) Annual cost of electricity for the pump:
C E  W fan (kW )  time(hr)  c E ($ / kW .hr) ($) (D.4)
Where W fan is the power consumed by the air-circulation fan and cE is the electricity
tariff in $/kW.hr. The total power consumed by the fan ( W fan ) is determined from:
 air g h f ,1Q1  h f , 2 Q2 
W fan  (kW) (D.5)
1000
The friction head losses in the two duct sections (h f,1 and h f,2 ) are obtained from the
Darcy-Weisbach equation, Equation (1.21). The friction factor (f) in each section of the
duct, which depends on the Reynolds number in the section, is obtained from the
Swamee-Jain equation (Equation 1.24).
c) Annual cost of fuel for the heater:
Q Total [W ]   [hr ]
CF   cF ($) (D.6)
1000  105500
Where cF is the cost of fuel in $/therm and τ is the total number of operation hours in a
year. The total heat loss ( Q Total ) is the sum of the heat loss in both sections, i.e., Q1 and
Q 2 , where the heat loss ( Q ) in each section is calculated according to the formula:
Q  T1  T 2  / Rth (W) (D.7)
Where T∞1 and T∞2 are the inside and outside air temperatures and Rth is the total
thermal resistance of the series resistances shown in Figure D.2.
r2 T∞2
r1
r3
T∞1
T1
T2
T3
T1 T2 T3
T∞1 T∞2
Ri R1 R2 Ro
Figure D.2. Thermal resistances of the insulated duct with two sections
The total thermal resistance (Rth ) is given by:
1 lnr2 / r1  lnr3 / r2  1
Rth     (m.K/W) (D.8)
2r1 h1 L 2k d L 2k s L 2r3 h2 L
Where r1 , r2 and r3 are the radii shown in Figure D.2.
Note that the value of the outside heat-transfer coefficient (h 2 ) is constant, but the inside
heat-transfer coefficient (h 1 ), which depends on the air velocity, changes with the inside
diameter of the duct and therefore, has to be determined for each diameter from:
h1  k air Nu / D1 (W/m2 .o C) (D.9)
For fully developed turbulent flow in tubes, the Nusselt number (Nu) is calculated from
the Dittus–Boelter equation, Equation (1.31), with n = 0.3 [3].
Optimisation with the GRG Nonlinear method

Figure D.3 shows the Excel sheet developed for this case. The problem data is shown in
the two columns at the left side of the sheet. The calculations part has two columns for
the two sections of the duct. The calculations start by guessed values for the two
diameters (D_1 and D_2) and two insulation thicknesses (δ_1 and δ_2) in both sections
of the duct. To allow Excel to automatically calculate the cost of duct unit (cd ) when the
two duct diameters are changed, the following formula for cd was obtained by using
Excel‟s trendline feature from the data shown in Table D.1:
cd  7.6881  1.7814 D  169.91D 2 (D.10)

Figure D.3. Excel sheet developed for the air-conditioning duct
The results part determines the total friction loss (hf_total), the fan power (Power), and
the different costs involved before determining the total annual cost (C_total) for the
guessed values of D_1=0.3m, D_2=0.2m, δ_1=0.1m and δ_2=0.1m.
Solver can now be used to find the optimum values for D_1, D_2, δ_1 and δ_2. As
shown in Figure D.4, the set-up box for Solver requires it to minimise the total cost
(C_total), which is the target cell, by changing the values of the two diameters (D_1 and
D_2) and the two insulation thicknesses (δ_1 and δ_2), which are the adjustable cells.
Figure D.4. Solver set-up for the air-conditioning duct

Figure D.5 shows the Excel sheet with the solution found by the GRG Nonlinear
method of Solver. The optimised dimensions found by Solver are shown in Table D.2
which also shows the different cost involved. The nearest dimeters are D1 =0.4 m and
D2 =0.3 m. Both insulation thicknesses are ≈ 0.3 m. The total annual cost sums up to
479.7 $.
Figure D.5. Solver solution for the air-conditioning duct
Table D.2. Solver solution for the insulated duct with four and three design variables
Optimised values Optimised values
determined with four determined with three
design variables design variables
(D1 , D2 , δ1 and δ2 ) (D1 , δ1 and δ2 )
D1 (m) 0.4142 0.4525
D2 (m) 0.2951 0.2586
 1 (m) 0.2970 0.2953
 2 (m) 0.3023 0.3056
Cduct ($) 128.5113 132.1191
Cins ($) 95.7092 95.2828
CE ($) 182.7827 186.4329
CF ($) 72.6776 72.3330
CTotal ($) 479.6808 486.1678
As a rule-of-thump, air-conditioning engineers frequently determine the duct areas from

the ratio of flow rates. Therefore, the duct diameters D1 and D2 are related as follows:
D2  Q3 / Q1  D1 (D.11)
Equation (D.11) allows us to solve the optimisation problem with only three variables,
which are D1 , δ1 and δ2 . The solution determined by Solver with three design variables
by using Equation (D.11) is also shown in Table D.2. Comparison with the solution
obtained without Equation (D.11) indicates that the rule-of-thump leads to a larger D1
and a smaller D2 . Although the insulation thicknesses on the two section are only
marginally affected, the figures in the table show that the total cost (486.2$) has
increased due to the increase in the initial cost of the duct (132.1$) as well as the
annual cost of electricity (186.4$). By suitably adjusting the given data, the Excel sheet
can be used to study the effects of electricity cost, fuel cost, or capital recovery factor
on the opimised solution.
Optimisation with the Evolutionary method

Chapter 5 showed that the optimised solutions obtained by the Evolutionary method of
Solver are not significantly different from those obtained by the GRG Nonlinear
method. To support this finding, this case was solved by using the same Excel sheet but
changing the solution method from the GRG Nonlinear method to the Evolutionary
method. Figure D.7 shows the solution obtained by the Evolutionary method. As the
figure shows, the solution is identical to that obtained with the GRG Nonlinear
method shown in Figure D.5. With the default options of the Evolutionary method
shown in Figure 3.15 and Figure 3.16.b, the Evolutionary method required more than
110 seconds, but by reducing the population size from 100 to 10, the method took about
24 seconds to solve the case.
Figure D.7. Solver‟s solution with the Evolutionary method

Concluding remarks
This appendix demonstrated the capability of the Excel-based modelling platform to
deal with practical optimisation analyses of fluid-thermal systems. The presented
analysis involved lengthy calculations, discrete-valued costs, and multi-variable
objective functions. In spite of these complications, the Excel-aided optimisation
method did not require the simplifications usually needed for applying the traditional
analytical method that depends on calculus methods.
Although not needed for the case considered in the present analysis, the Thermax add-
in enables the Excel-based platform to perform the optimisation analyses with
significant changes in fluid properties such as those of power generation and
refrigeration systems.
References
[1] The duct shop, https://www.theductshop.com/shop/catalog-galvanized-sheet-
metal-duct-c-1_3.html, last accessed 24/12/2015.
[2] W.S. Janna, Design of Fluid Thermal Systems, 3rd Edition, CENGAGE Learning,
2011.
[3] Y.A. Cengel, and A.J. Ghajar, Heat and Mass Transfer: Fundamentals and
Applications. McGraw-Hill, 2011.
Appendix E. Optimisation analysis of a pump-pipe system
This appendix serves as a bridge between Chapter 6, that shows how Excel and Solver
can be used for determining the optimum pipe diamater, and Chapter 7 that deals with
design analyses of pump-pipe systems. This appendix shows how the Excel-Solver
method can be applied for two steps in the design process that determine the
optimumum pipe diameter and the opertaing point for the pump-pipe system. The case
considered in the analysis is based on one of the cases given by Schumack [1].
E.1. The pump-pipe system

An air conditioning system is designed to provide cooling water at 3°C to the condenser
of a power plant as shown in Figure E.1. The water flow-rate through the pump is 12.5
gallon per minute (gpm). The commercial steel pipe from point 1 to point 2 is 20 m
long and has 15 elbows and 6 valves. The elevation of point 2 is 10 m higher than that
of point 1. The water goes through a 20 m of ½ inch pipe through twenty 180° bends
within the chiller. The commercial steel pipe from point 3 to point 1 is 30 m long.
Chiller
2
Heat exchanger
KL = 30
Warm water
1
Pump
Cold water
Figure E.1.
The initial cost of the system is that of the pump and pipes and the operating cost is the
cost of electric power required to run the pump. Costs of the heat exchanger, the chiller
– including the cost of pipe 2-3, and the fittings (i.e., the valves and elbows) need not be
considered in the analysis. The cost of the commercial steel pipe (Cpipe) can be taken as
$3.4 per cm diameter per 1 m length. The cost of centrifugal pumps in dollar can be
approximated by [1]:
C pump  200W pump  150

(E.1)
Where W pump is the power in horsepower needed to run the pump.

The cost of electricity is $0.10 per KW-h.
(a) Develop a model for the system in Excel and use Solver to determine the optimum
diameter for pipes 1-2 and 3-1 that minimises the annualised total cost of the
pump-pipe system based on the data given above and assuming that the system
runs for 8760 hours per year. Take the amortization rate (α) as 0.2 and the pump
efficiency (η) as 65%. Use Kvalve = 0.15, Kelbow= 0.6, and Kbend = 0.6.
(b) Select the nearest standard pipe diameter and, using this diameter to determine the
required pump power at its opertaing point given that it has the following
charactersitic curve:
h pump = 49.657 – 12909.0Q - 8.0x106 Q2 - 4.0x109 Q3 (E.2)
Where h pump is the head delivered by the pump and Q is the pump‟s discharge.
E.2. Determination of the optimum pipe diameter

Figure E.2 shows the front sheet in the workbook (Sheet 1) developed for the analysis.
The sheet stores the general information about the pump-pipe system such as the water
flow rate in gallons per minute (Q_gpm), the lengths of pipes 1-2, 2-3, and 3-1. It also
stores the densoty and viscosity of water at the average temperature. The roughness
hights for the steaal pipe (ks_pipe) and for the chiller pipe (ks_chillr) are stored
together with friction coefficients for the valve (K_valve), the elbow (K_elbow), the
bend (K_bend), and the heat-exchanger (K_HE). The cost of electricity per kilowatt-
hour (c2) and the amortization rate (α) are also stored in this sheet.
Figure E.2.
Figure E.3 shows the second sheet in the workbook (Sheet 2) that determines the
optimum pipe diameter. For the sake of the analysis, the sheet converts the water flow
rate from gallpns per minute into cubic metre per second as Q. Starting with an initial
pipe diamter, D_pipe, of 0.01 m, the sheet calculates the water velocities in the three
pipes, the corresponding Reynolds number, and friction factors. Based on these values,
the sheet determines the major and minor friction losses in the pipes, the total friction
head loss (hf_total), the required pump power (Power), the initial costs of the two pipes
and the pump, the annual cost of electricty (C_elec), and the total annualised cost
(C_total) given by:
Ctotal = α(Cpipe + Cpump ) + Celec (E.3)
Figure E.3.
Figure E.3 shows that at the initial pipe diameter, the total annulaised cost is equal to
11574.74 dollar. Following the procedure described in Chjapter 6, Solver can be used to
determine the optimum diameter by minmising the total annulaised cost and Figure E.4
shows the requires set-up.
Figure E.4. Solver‟s set-up

Figure E.5 shows the solution determined by Solver with this set-up and the GRG
Nonlinear method. According to this solution, the optimum diameter for the two pipes
is 0. 035722 m, or 3.572 cm. The nearest standard schedule 40 pipe is the 1½ inch
nominal pipe with internal diameter Di of 4.090 cm.
Figure E.5. Solver solution
E.3. Determination of the operating point for the pump

The actual water flow rate and head delivered by the pump depend on its charactertic
curve. Using the pipe diameter found in part E.2, the operating point for the system can
be determined by polotting the pump‟s characteristic curve and the system curves for
various values of the water flow rate. Figure E.6 shows the third sheet in the workbook
(Sheet 3) developed for determining the operating point.
Figure E.6
The sheet was used to calculate the friction head loss in the systems and the head
delivered by the pump by changing the pump‟s discharge from 0.0001 to 0.001 m3 /s
and the result is shown in Figure E.7.
60
50
40
Head (m) 30
y = -4E+09x 3 - 8E+06x 2 - 12909x + 49.657
20
Series1
10 Series2
Poly. (Series2)
0
0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012
Discharge (m 3 /s)
Figure E.7.
Figure E.7 shows that the operating point, which is the intersection between the system
curve with the pump charaterstic curve, is around Q = 0.00079 m3 /s and h p = 33 m.
The operating point can be determined precvisely by using Solver. Figure E.8 shows the
set-up that requires Solver to iterate to determine the value of Q (the changing varaible)
that satisfies the following constraint:
Hf = hp
Figure E.9 shows the solution sound by Solver. Which is Q = 0.000791 m3 /s and h p =
32.45 m. The flow rate given by the pump is 12.561 m3 /s, which is slightly more than
the required flow rate of 12.5 m3 /s. The required pump power is 0.50698 kW.
Figure E.8.
Figure E.9.
E.4. Conclusions
References:
[1] M. Schumack, Solution of complex pipe flow problems using spreadsheets in an
introductory fluid mechanics course, Proceedings of the 2004 American Society
for Engineering Education Annual Conference & Exposition, Session 3666, Pages
9.1108.1 - 9.1108.13. Available at https://peer.asae.org/solution-of-complex-pipe-
flow-problems-using-spreadsheets-in-an-introductory-fluid-mechanics-course
[2] https://www.plumbingsupply.com/ed-frictionlosses.html
Appendix F: Optimisation analysis of a pipe network with eight loops
This appendix applyies the Excel-Solver method presented in Chapter 8 for

optimisation analyses of pipe-networks to a network of practical complexity. The case
considered is the gravity-driven network presented by Brkic [1] shown in Figure F.1.
The network distributes water from two reservoirs to four demand points via 19 cast
iron pipes (ε = 0.00026 m) that are laid on a flat area with no variation in elevation.
Table F.1 shows the lengths of the 19 network pipes.
Figure F.1. A pipe-network with eight loops (adopted from Brkic [1])
This network will initially be analysed hydraulically by using the same initial pipe
diameters used by Brkic [1] so that comparison can be made with his hydraulic solution
with the initial pipe diameters and flow rates shown in Table F.1. Solver will then be
used to determine the pipe diameters and flow rates that minimise the total cost
following the method described in Chapter 8. Both analyses witll study the effect of the
method used to evaluate friction losses on the results obtained by Solver. The
optimisation analysis also studies the effect of Solver‟s solution options on the results.
F.1. The hydraulic model

Following the procedure described in Chapter 8 for the analysis of pipe networks using
Excel and Solver, the flow rates in 8 selected pipes were specified and those in the
remaining 11 pipes were determined by applying the continuity equation at each
junction. Solver is then used to adjust the initial flow rates in order to satisfy the energy
equation. The assumed flow-directions for all the pipes are shown in Figure F.1. The
initially assumed flow rates are those for pipes AB, BC, CD, DE, EF, HI, JK, and KL.
The flow rates in the remaining 11 pipes are calculated from continuity as follows:
QFG  QGK  QGH (F.1)

Table F.1. Pipe data and initial flow rates in the 8-loop network [1]
Brkic [1] Length (m) Initial Initial flow
Pipe
notation diameter (m) rate (m3 /h)
/1/ AB 457.2 0.305 173.32
/2/ BC 304.8 0.203 150
/3/ CD 365.8 0.203 130
/4/ DE 609.6 0.203 6.6
/5/ EF 853.4 0.203 100
/6/ FG 335.3 0.203 0.28
/7/ GH 304.8 0.203 16.88
/8/ HI 762.0 0.203 13.56
/9/ AI 243.8 0.203 200
/10/ IJ 396.2 0.152 50
/11/ JK 304.8 0.152 70
/12/ KL 335.3 0.254 51.96
/13/ EL 304.8 0.254 32.96
/14/ HJ 548.6 0.152 3.32
/15/ BJ 335.3 0.152 23.32
/16/ GK 548.6 0.152 17.16
/17/ CK 365.9 0.254 20
/18/ FL 548.6 0.152 9
/19/ DL 396.2 0.152 10
QGH  QHI  QHJ (F.2)

QIA = 373.32 - QAB (F.3)
QIJ  Q AI  136.44  QHI (F.4)
QLE  126.36  QDE  QEF (F.5)
QJH  QIJ  QBJ  QJK (F.6)
QBJ  Q AB  QBC (F.7)
QGK  159.12  QJK  QKL  QCK (F.8)
QCK  QBC  QCD (F.9)
QFL  QEF  90.72  QFG (F.10)
QDL  QCD  113.4  QDE (F.11)
The energy equation imposes 8 constraints on the iterative solution, one for each loop.
Taking the clockwise flow direction as positive, the constraints are:
h f , AB  h f , BJ  h f , AI  h f , IJ  0 Loop 1 (F.12)
h f , IJ  h f , HJ  h f , HI  0 Loop 2 (F.13)
h f , BC  h f , CK  h f , BJ  h f , JK  0 Loop 3 (F.14)
h f , JK  h f , GK  h f , GH  h f , HJ  0 Loop 4 (F.15)
h f ,CD  h f , DL  h f , KL  h f ,CK  0 Loop 5 (F.16)
h f , FG  h f , GK  h f , KL  h f , FL  0 Loop 6 (F.17)
h f , DE  h f , EL  h f , DL  0 Loop 7 (F.18)
h f , EF  h f , FL  h f , EL  0 Loop 8 (F.19)
To investigate the effect of the formula for the friction-factor on the analysis results,
two Excel sheets were developed for the hydraulic analyses of this network one of
which applies Darcy-Weisbach equation with the Swamee-Jain (S-J) formula, while the
other applies the Colebrook-White (C-W) formulae. Figure F.2 shows the sheet that
uses the S-J formula. The data part stores the water density (ρ) and viscosity (visc), the
gravity constant (g), and the pipe roughness (ε). The length and diameter for the 19
pipes are stored in columns E and F, respectively. The sheet calculates the cross-
sectional area for each pipe in column G. Column H stores the initial flow rates in the
pipes. The 8 coloured (shaded) cells in this column store the assumed flow rates while
those determined according to Equation (F.1) – (F.11) are not coloured. Note that the
calculated pipe flow rates are the same as those given by Brkic [1] and Table F.1.
Figure F.2. The Excel sheet developed for analysing the pipe-network with eight loops
by using the Swamee-Jain formula
The calculated pipe friction losses are stored in column L. The formula bar in Figure
F.2 reveals the formula for determining the error in satisfying the energy equation in
loop 1 according to Equation (F.12). Since the initially assumed flow rates in the pipes
satisfy only the continuity equation but not the energy equation, most of the errors
calculated by Equations (F.12) – (F.19) in column O are not close to zero. The flow
rates that minimise the errors in the energy equations to sufficiently small values can be
found by Solver and the set-up for the task is shown in Figure F.3.
Figure F.3. Solver set-up for the analysis of the pipe-network with eight loops
Note that the “Set Objective ” slot is left blank and the variables to be changed are the
flow rates in the 8 selected pipes. There are 8 constraints which require the absolute
values of the errors in satisfying Equations (F.12) – (F.19) to be reduced to a maximum
tolerance (Maxerror) of 0.001. Attempts were made to use the Evolutionary method of
Solver for the present analyses, but the method failed to converge to a solution even
with the different numerical options offered by Solver. The GRG Nonlinear method
also failed to converge to a solution with automatic-scaling. Figure F.4 shows the
solution found with GRG Nonlinear method but without automatic-scaling. Note that
all the 8 constraints are now satisfied. Figure F.5 compares the flow rates with Solver
with those obtained by Brkic [1]. The figure shows that the values calculated by the S-J
formula agree well with those obtained by Brkic [1] who used the node-loop method
and solved the resulting linear system with Excel matrix functions.
The second sheet that uses the Colebrook-White formula was developed following a
similar procedure and Solver was used to determine the flow rates that satisfy the
energy equations. Figure F.6 shows the solution obtained with the GRG Nonlinear
method of Solver with this formula. Figure F.5 that also shows the flow rates calculated
by this formula shows that their values are close to those obtained with the Swamee-
Jain formulae and agree well with those obtained by Brkic [1].
Figure F.4. Solver solution for the pipe-network using the Swamee-Jain formula
200
Brkic[1]
150 S-J
Flow ratee (m 3 /h)
C-W
100
50
0
EL
AI
CK
DL
BJ
GH
EF
AB
CD
JK
-50
Figure F.5. Comparison of the flow rates obtained by the present analyses for the eight-
loop pipe-network with those given by Brkic [1]
Figure F.6. Hydraulic solution found with Solver using the Colebrook-White formula
F.2. The economic optimisation analysis

The objective function for the economic optimisation of the pipe network is to minimise
the total owning cost as given by:
NP
CT  C1
i
(F.20)
Where Ci is the cost of pipe i and NP is the total number of pipes in the network. The
pipe costs used in the present analysis are those given in Table 12.2. The pressures at
the four demand nodes, which have to be not less than 30 m.c.w., are given by:
PD  PA  h f , AB  h f , BC  h f ,CD (F.21)
PF  PE  h f , EF (F.22)
PI  PA  h f , AI (F.23)
PK  PI  h f , IJ  h f , JK
(F.24)
The present optimisation analysis investigated the effect of the friction formula and the
numerical options given by Solver on the optimised solution. The two Excel sheets
developed for the hydraulic analysis were extended for the economic optimisation by
calculating the total cost of the network. A third Excel sheet was also developed by
using the Hazen-Williams equation. Figure F.7 shows the sheet that applies the
Swamee-Jain formula. Column N has been added to the analysis sheet to calculate the
pipe costs from the data stored in Sheet 2. The sheet uses Thermax linear interpolation
function "Interpl1" to determine the pipe costs and the formula bar reveals the formula
in cell Q16 that calculates the cost of pipe AB.
Figure F.7. Excel sheet for economic optimisation of the 8-loop pipe network
Column S stores the initial pressures in nodes A, D, E, F, I, and K. Initially, the
pressures at the two nodes A and E are given values of 50 m.c.w. and the pressures at
the four demand nodes are calculated according to Equations (F.21) – (F.24). The
minimum and maximum diameters are identified and stored in cell Q13 and Q14,
respectively. Cell S9 determines the minimum nodal pressure that has to be greater than
the specified limit of 30 m.c.w. The total network cost is stored in cell Q16. With the
initial pipe diameters shown in Table F.1, the total cost of the network is 441.409 units.
Solver can be used to simultaneously minimise the total cost and satisfy the energy
equations. The required dialog box is shown in Figure F.8.
Figure F.8. Solver set-up for the optimisation of the eight-loop pipe-network
Note that the 19 pipe diameters have been added to the “Changing Variable Cells”. Two
constraints have also been added to specify the minimum and maximum pipe diameter
to 0.025 and 0.61 m, respectively. The last constraint expresses the lower limit for the
pressure at the demand nodes, which is 30 m.c.w. The solution found by Solver using
forward-difference approximation and automatic-scaling is shown in Figure F.9.As the
figure shows, the minimum total cost for the pipe-network obtained by the S-J formula
is 221,874 units. The minimum pressure occurs at node K, which is 30 m.c.w. Also note
that the initial pressures at the supply nodes A and E have not been changed.
Following the same steps, Solver was used with the other two Excel sheets that used the
Colebrook-White formula and with the Hazen-Williams equation and the corresponding
solutions are shown in Figures F.10 and F.11, respectively. Figure F.10 shows that the
minimum total cost obtained by the Colebrook-White formula is 411.851 units, which
is considerably higher than that obtained with the Swamee-Jain formula. As shown in
Figure F.11, he minimum total cost obtained by the Hazen-Williams equation, which is
242.234 units, is also higher than that obtained with Swamee-Jain formula.
Figure F.9. The optimised network obtained with the Swamee-Jain formula
Figure F.10. The optimised network obtained with the Colebrook-White formula
Figure F.11. The optimised network obtained with the Hazen-Williams equation
Figure F.12 compares the pipe diameters obtained by the three friction equations.
Unlike the hydraulic analysis, the figure shows great variations between the results of
the three friction models. The figure shows that the Swamee-Jain formula gave the
smallest diameters for almost all the pipes and, therefore, the least total cost. Table F.2
compares the costs of the initial pipe diameters given in Table F.1 using the cost data
given in Table 12.2 with the present costs obtained by the three friction equations.
0.35
0.3 S-J
C-W
0.25
Pipe diameter (m)
H-W
0.2
0.15
0.1
0.05
0
G…
EL
AI
CK
FL
FG
DL
BJ
HJ
IJ
GK
EF
AB
KL
CD
DE
BC
HI
JK
Figure F.12. Optimised pipe diameters
Table F.2. The optimised total costs compared to the initial cost of the network
Pipe Initial cost S-J C-W H-W Brkic [1]
AB 50.02 26.96 30.82 27.33 30.98
BC 22.97 19.06 23.12 22.75 22.90
CD 22.97 19.93 30.90 19.18 14.94
DE 22.97 7.93 22.75 12.15 12.07
EF 22.97 13.27 24.23 19.29 11.26
FG 22.97 6.92 19.11 9.53 13.16
GH 22.97 6.83 18.01 8.77 18.28
HI 22.97 8.48 19.68 9.95 12.54
AI 22.97 18.28 16.44 15.50 35.91
IJ 15.96 8.93 15.64 7.49 2.99
JK 15.96 15.53 21.67 10.36 11.76
KL 31.99 15.99 23.72 12.94 9.77
EL 31.99 15.40 52.84 12.33 15.37
HJ 15.96 4.17 14.74 7.38 11.50
BJ 15.96 13.86 16.28 7.52 15.42
GK 15.96 7.89 15.02 8.57 9.54
CK 31.99 2.07 15.26 13.24 16.09
FL 15.96 5.02 15.82 11.14 7.81
DL 15.96 5.36 15.81 6.84 8.74
Total cost 441.41 221.87 411.85 242.23 281.04

The figures in Table F.2 show that the lowest total cost for the network was that
obtained by the Darcy-Weisbach equation and the Swamee-Jain formula. By reducing
the total cost from 441.41 units to 281.04 units, this friction formula saves almost 50%
of the initial network cost. The cost obtained by the Hazen-Williams equation is
comparable to that of the Swamee-Jain formula, but the Colebrook-White formula
failed to reduce the initial cost. Table F.2 also shows the costs of the optimised
diameters obtained by Brkic [1] based on the cost data given in Table 12.2. His
optimum solution reduced the total cost to 281.04 units, which is slightly higher than
the cost obtained by the Swamee-Jain formula and the Hazen-Williams equation.
F.3. Effect of Solver options on the optimisation results

The three Excel sheets developed for the optimisation analysis were also used to
investigate the effect on the optimised solution by the set of Solver‟s solution options.
For this investigation, the GRG Nonlinear method was used with forward-difference
(FD) and central-difference (CD) approximations of derivatives and with or without
automatic-scaling. Tables F.3 to F.5 show the minimum network total cost, number of
Solver iterations, and computer time for the three friction models. Without automatic-
scaling, the CD approximation failed to converge to a solution with all three friction
equations. Therefore, this combination is execluded from the comparison.
Table F.3. Optimal Solver solutions with the Swamee-Jain friction formula
Option Total cost No of CPU time
iterations (second)
FD with automatic scaling 221.874 122 17
FD without automatic scaling 244.796 115 17
CD with automatic scaling 221.947 126 21
Table F.4. Optimal Solver solutions with the Colebrook-White friction formula
iterations (seconds)
CD with automatic scaling*
* Terminated after 1100 iterations without convergence
Table F.5. Optimal Solver solutions with the Hazen-Williams friction equation
iterations (seconds)
CD with automatic scaling 452.507 63 14
Although the figures in the three tables show considerable variations in the estimations
by the three friction models, they show that the FD approximation yields better results
than the CD approximation. This agrees with the results presented in Chapter 8 for a
smaller gravity-fed pipe network. However, the tables show that the Swamee-Jain
formula obtained the lowest total cost, which is 221,874 units. In average, the Swamee-
Jain formula also gave the best result for this case. Note that for the gravity-fed pipe
network considered in Chapter 8, the lowest total cost was obtained by the Colebrook-
White formula. For the present case, the Colebrook-White formula gave higher total
costs compared to the other two models.
The automatic-scaling option increased the minimum total cost for the S-J formulae, but
decreased it for the other two friction models. Note that with automatic-scaling the two
Darcy-Weisbach options gave comparable solutions of about 245 units. Even with
automatic-scaling, the CD approximation failed to converge to a solution with the
Colebrook-White formula that took longer computer time compared to the other two
friction models.
F.4. Concluding remarks

This appendix demonstrated the ability of Excel and Solver to deal with hydraulic and
optimisation analyses of medium-size pipe-network of practical complexity. The
analyses with alternative friction formula show that the results of the hydraulic analysis
are not sensitive to the method used, but the optimisation analyses show considerable
variations in the results obtained by the three friction formulae. Although the
optimisation analyses with different friction formulae do not favour a particular friction
equation, those with different Solver options indicate that the default forward-difference
approximation is preferred to central-difference approximation. Regarding the use of
automatic-scaling, the analyses show that it gives mixed results with the three friction
models. Finally, it should be mentioned that attempts were also made to solve the
optimisation problem with the Evolutionary method, but the method failed to converge
to a solution even with different solver options available.
References
Appendix G: Comprehensive exercises and projects
This appendix provides comprehensive exercises for optimisation analyses of fluid and
thermal systems. Most of these exercises are based on relevant cases given by Schumak
[1] and Burmeister [2]. Standard textbooks on the design of fluid-thermal system, such
as Janna [3] and Stoecker [4], can be consulted for more information about the basic
principles involved in the analytical models and for validating the analyses results.
G.1. Fluid systems

1. An air conditioning system is designed to provide cold water at 3°C to a series of
chillers as shown in the Figure G.1. The water‟s flow rate through the pump is 50
gallon per minute (gpm). The pipe from point 1 to point 2 is 18 m long and has 15
elbows and 6 valves. The details of the pipe runs between point 2 and the four
chillers are as follows:
Run A: L = 100 m, 15 elbows, 6 valves

Run B: L = 80 m, 10 elbows, 5 valves
Eun C: L = 60 m, 8 elbows, 3 valves
Run D: L = 50 m, 6 elbows, 3 valves
Chillers
A (4)
B 3
2
C
Heat exchanger D
KL = 30
Figure G.1. Schematic of an air conditioning system with four chillers (adopted
from Schumak [1])
The cold water goes through a 20 m of ½ inch pipe through twenty 180° bends
within each chiller. The pipe length from point 3 to point 1 is 25 m.
For the same costs, economic factors, and pump-efficiency given in Appendix E
develop a model for the system in Excel and use Solver to determine the diameters
for the pipes 1-2, 3-1, and runs A, B, C and D with the goal of providing equal
flowrates through the chillers at the lowest auualised total cost. This case is based
on Case 4 given by Schumak [1].
2. Extend the Excel sheet developed for the hydraulic analysis of the pipe-network in
Problem 8.2 for least-cost optimisation of the network by using the pipe-cost data
given in Table 8.2. Use the GRG Nonlinear method to study the effect Solver‟s
solution options on the optimised solution.
3. Extend the Excel sheet developed for the hydraulic analysis of the pipe-network in
Problem 8.3 for least-cost optimisation of the network by using the pipe-cost data
given in Table 8.2. Use the GRG Nonlinear method of Solver to study the effect on
the optimised solution due to the method used to determine friction losses, i.e., the
Darcy-Weisbach equation or the Hazen-Williams equation.
G.2. Thermal systems

4. The system shown in Figure G.2 is to be designed for transferring heat from
condensing steam at 80o C to a low-boiling fluid that boils at 20o C by means of a
circulating water loop and a heat exchanger on each side. The required rate of heat
transfer, q, is 65 kW and the overall heat transfer coefficient, U, is 0.03 kW/(m2 .
K) for both heat exchangers. It is required to optimise the total cost of the system
over its lif time (CT), which is given by:
Vapour T2 Vapour
80o C o
20 C
A1 A2
Heat exchanger 1 Heat exchanger 2
T1
Liquid Liquid
o o
80 C 20 C
Figure G.2. Run-around water loop for heat-recovery (adopted from Stoecker [4])
CT = Cpurchase + Cpumping = Ch (A1 + A2 ) + Co m 3 (A)
Where A1 and A2 are the areas of the two heat exchangers, Ch is the unit cost of the
heat exchanger over their lifetime in $/m2 , Co is the unit cost of the pumping power
in $/(kg/s)3 , and m is the mass flow rate of circulated water in kg/s.
Over the lifetime of the system, the cost of the heat exchanger Ch = $80/m2 and the
cost of the pumping power Co = $28,474 s3 /kg3 .
Develop an Excel sheet for determining the total cost of the system and use Solver
to determine the values of A1 , A2 , and m that minimise the total cost.
Also determine the optimum values for T1 , T2 , the heat-exchangers‟ effectiveness

(ε), and the total lifetime costs Cpurchase and Cpumping .
This case is based on that given by Stoecker [4], but using the cost data given by
Burmeister [2], p 370. Answer: A1 = A2 = 86 m2 , m = 0.406 kg/s, ε=0.78, T1 =31o C,
T2 =69o C, Cpurchase = $13,660, and Cpumping =$1,900.
Solution procedure:
 Assume reasonable values for T1 and T2

 Determine the mass flow rate of water ( m ) from:
 
  q / c pw T2  T1 
m (B)
Where q is the specified heat-transfer rate, 65 kW.

 Determine the heat exchanger areas by using the long-mean temperature
difference method presented in Appendix B:
A  q / U lm  (C)
Where U is the specified overall heat transfer coefficient and ∆Tlm is the
log-mean temperature difference for the respective heat exchanger.
 Use Solver to minimse the total cost by adjusting the two water
temperatures T1 and T2 within the following limits:
20 ≤ T1 ≤ 80 o C, 20 ≤ T2 ≤ 80 o C.
 Use the LMTD method to determine the optimum values of T1 , T2 , and ε.
5. The facility shown in Figure G.3 is contemplated for cooling 1.2 kg/s of air from
95 to l0o C using two stages of air compression. The air is first cooled in a
precooler and then in a refrigeration unit. Water passes through the condenser of
the refrigeration unit, then into the precooler, and finally to a cooling tower, where
heat is rejected to the atmosphere. The flow rate of the cooling water is 2.3 kg/s
and it leaves the cooling tower at a temperature 24°C.
Water 2.3 kg/s q3
Cooling
Tower
Tw2 Compressor power P, kW
o
24 C
Tw1
Air q2 q1
95o C Ta3
10o C
1.2 kg/s
Precooler Refrigeration unit
Figure G.3. Compressed-air cooling system (adapted from Burmeister [2])
The total first cost of the system (CT,0 ) comprises the cost of the refrigeration unit,
precooler, and cooling tower, designated x1 , x2 , and x3 , respectively. The
expressions for these costs, in dollars, are given below:
x1  48q1 (A)
50q 2
x2  (B)
Ta 3  Tw1
x3  25q 3 (C)
Where q 1 , q 2 , and q 3 are the rates of heat transfer shown on Figure G.3 in kW.
The power required by the compressor of the refrigeration unit (P) is given by:
P = 0.25q 1 (kW) (D)
Both q 1 and P must be absorbed by the condenser cooling water passing through
the refrigeration unit.
Develop an Excel sheet for calculating the total cost of the system and use Solver
to determine the minimum first cost for the system.
This case is based on an example in Burmeister [2], p 310. Solution: x1 = 1452.76,

x2 = 491.31, x3 = 2739.63, C T,0 = 4683.7.
The following is a suggested solution procedure:
 Guess the value of q 2

 Calculate the corresponding air temperature Ta3 from energy balance:

Ta3  95  q 2 / c pa m a  (E)
 Calculate the value of q 1 from:
q1  / m a c pa Ta3  24 (F)
 Calculate the water temperatures Tw1 and Tw2 from the values of q 1 and q 2
and the specific heat for water.
 Calculate the three costs x1 , x2 , and x3 and then the total cost C T,0 .
 Use Solver to determine the minimum total cost by changing the value of
q 2 while applying the following constraint:
T3a = T1w + ∆
Where ∆ is a suitable temperature difference, e.g., 5o C.
6. Figure G.4 shows a wood-frame for a house wall in which glass mineral fibre is
used for insulation. The construction engineer is concerned that at winter time the
temperature of the insulation in the wall may fall below the dew point temperature
of the air in the room. If this happens, the water vapour will diffuse through the
wall and condense out in the glass fibre insulation, thus, causing its damage.
Insulation Plywood
Sheet rock
Sliding
Ti, hi T o , ho
Figure G.4. A wood-frame house wall construction

The wood siding on the outside surface of the wall is followed by plywood
sheathing. The inside surface is constructed of sheet rock. Material properties and
thicknesses are given in Table G.1.
Table G.1. Wall material properties

Material Thickness (cm) k (W/m-K)
Wood 0.95 0.14
Plywood 1.27 0.12
Glass Fibre 9.0 0.038
Sheet Rock 0.95 0.17
The glass mineral fibre insulation forms the main thermal resistance in the wall
between the wood framing members (not shown). The equivalent thermal
resistance circuit for the system is shown in Figure G.5.
Ti T5 T4 T3 T2 T1 To
Ri Rsr Rins Rply Rs Ro
Figure G.5. Thermal-resistance circuit for the wall
(a) Develop an Excel sheet to determine the inside and outside temperatures of
the isolation, T4 and T3 , for winter conditions of Ti = 23 °C and To = 15 °C
when the values for the outside and inside heat-transfer coefficients are h o
= 34 W/m2 K, and h i = 9.3 W/m2 K.
(b) Use the Excel sheet developed for part (a) to determine whether the two
temperatures T4 and T3 can fall below the corresponding dew-point
temperatures under winter weather conditions of Tinside= 23°C, Toutside =
15°C, indoor relative humidity of 50%, and various outdoor humidity
levels in the range 50 - 90%.
This case is based on that provided by Harris et al [5].
7. Figure G.6 shows a schematic diagram of a gas-turbine with an inlet-air precooler

before the compressor and an air preheater before the combustion chamber. The
system improves the thermal efficiency of the gas-turbine by utilising the waste
energy in the exhaust gas for: (i) cooling the air before the compressor with a
vapour-absorption refrigeration (VAR) system and (ii) preheating the compressed
air before the combustion chamber (regeneration). The improved efficiency
reduces the fuel consumption and increases the power output.
Air preheater
HRSG
(AB ), (TB ) (Ax)
Combustion
products
Fuel
T3 (Q)
T5
Steam loop T2
Combustion Net
T4 power
chamber
(Qr ) (W)
VAR
Air system Gas turbine
T0 T1
Air precooler
Figure G.6. Gas turbine with an air preheater and a VAR inlet-air cooling system
(adapted from Burmeister [2])
The total cost (CT) of the system over its lifetime is given by:
CT = Cf Q + Cr Qr + CB AB + Cx Ax – Cw W (A)
Where Cf , Cr , CB , Cx, and Cw are unit costs for the fuel, refrigeration, HRSG, air
preaheater, and power, respectively. Q, Qr , and W stand for the fuel‟s energy, the
amount of heat transferred to the VAR system, and the power output of the plant,
respectively.
The total cost of the system can be treated as a function of two dimensionless
quantities x1 and x2 which are given by:
U x Ax
x1  (B)
md C p
U B AB
x2  (C)
md C p
Where Ux and UB are the overall heat-transfer coefficients of the air preheater and
the boiler (HRSG), respectively, the two areas Ax and AB are the surface areas of
the air-preheater and the boiler (HRSG), respectively, and md and Cp are is the
mass flow rate and the specific heat of the air, respectively.
It is desired to find the sizes (areas) for the heat-recovery steam generator (HRSG)
and the air preheater that minimise that minimise the total cost of the system based
on the following data [2]:
md = 45 kg/s, COP = 1, UB = 0.25 kW/m2 K, Ux = 0.082 kW/m2 K, Cp = 1 kJ/kg K,

To = 300 K, T4 = 1100 K, T5 = 750 K, TB = 380 K, Cw = $350/kW, Cf = $50/kW, Cr
= $100/kW, CB = $1000/m2 , Cx = $100/m2 .
The analytical model:

Applying the constant-specific heat method for an ideal Brayton cycle, Burmeister
[2] obtained the following equations for the three quantities W, Q, and Qr :
 
W / md C p T4  T5   1  T1 / T5 (D)
 
Q / md C p T4  T5   T4 1  T1 / T5  / T4  T5   x1  /1  x1  (E)
Qr /md T4  T5   COP / T4  T5 1  exp x 2  

T5  x1T1T4 / T5  / 1  x1   TB  (F)
The cooled-air temperature T1 in the above equations can be calculated from:
T1  T0 1  x1   COP1  exp x 2 T5  TB 1  x1  /

1  x1  COP1  exp x2 x1T4 / T5  (G)
Where COP is the coefficient of performance of the VAR system.
Finally, the actual total cost F in $ is given by:
F= [md Cp (T4 –T5 )]CT (H)
This case is based on Problem 4.10 in Burmesier [2] who gave the following
answer: x1 = 2.46, x2 = 0.174, and F= -2,191 x106 $.
8. Since the ideal Brayton cycle does not account for the losses in the compression
and expansion processes, the analytical method applied in problem G.8 can be
made more realistic by the following modification:
(a) The ideal temperatures after the compression and expansion processes, T2s and
T5s , are determined by using the constant specific-heat method from:
k 1
T2 s  T1  rp k
(A)
 
k 1
T5 s  T4  1 / r p k
(B)
Where rp is the pressure ratio of the gas-turbine (use rp =3.82) and k is the ratio of
specific heats for air which can be taken as k = 1.4. The actual temperatures, T2 and
T5 , are then calculated from:
T2  T1  (T2 s  T1 ) /  c (C)
T5  T4  (h4  T5 s )  t (D)
Where, ηc and ηt are the isentropic efficiencies of the compressor and the turbine,
respectively. Use suitable values for ηc and ηt .
(b) The net work and the two rates of heat transfer are given by:
W= Cp [(T4 – T5 ) –(T2 – T1 )] (E)
Q = Cp (T4 – T3 ) (F)
Qr = COPxQ (G)
(c) The chilled-air temperature, T1 , can be determined either by using Equation (G)
of Problem G.8 or by modelling the VAR system as a heat-exchanger and using
the LMTD method described in Appendix B.
Develop and Excel sheet for the gas-turbine system based on the above analytical
model and use Solver to minimise the total cost. Compare your results with those
obtained by Burmeister [2] and comment on any differences.
References
[3] M. Schumack, Solution of complex pipe flow problems using spreadsheets in an
introductory fluid mechanics course, Proceedings of the 2004 American Society
for Engineering Education Annual Conference & Exposition, Session 3666, Pages
9.1108.1 - 9.1108.13. Available at https://peer.asae.org/solution-of-complex-pipe-
flow-problems-using-spreadsheets-in-an-introductory-fluid-mechanics-course
[4] L.C. Burmeister, Elements of Thermal-Fluid System Design, Prentice Hall, 1998.
[5] W.S. Janna, Design of Fluid Thermal Systems, third edition, CENGAGE Learning,
2011
[6] W. Stoecker, Design of Thermal Systems, McGraw-Hill, 1989.
[7] C. E. Harris, Jr. M. S. Pritchard, M. J. Rabins. Engineering Ethics.1995.

Wadsworth Publishing Co. Belmont CA
Subject Index
Activation of Solver, 53, 56 Energy equation, 8, 91, 136, 140, 151,

Air-conditioning, 7, 45, 113, 314, 326, 154, 155, 158, 161, 169, 171, 174,
328 175, 182, 183, 184, 185, 187, 188,
Air-conditioning duct, 326 189, 280, 281, 285, 286, 289, 290,
Amortisation rate, 247, 248, 258, 262, 292, 293, 336, 356, 357, 359
289, 290, 317, 370 Evolutionary, 55, 56, 58, 59, 60, 63, 64,
Automatic scaling, 278, 295, 366, 367 65, 75, 105, 126, 127, 128, 129, 334
Backward difference, 200 Evolutionary method, 55, 56, 58, 59,
Boundary conditions, 196, 197, 198, 60, 63, 64, 65, 105, 126, 127, 128,
199, 201, 202, 205, 213, 216, 219, 333, 334
220, 227, 229, 235, 237, 238, 306, Excel built-in functions, 29
344 Excel formula, 24, 31, 41, 69, 70, 98,
Brayton, 131 308, 309
Brayton cycle, 131 Excel-aided optimisation, 105, 108,
Built-in function, 24, 25, 69 109, 113
Cavitation, 90, 91, 93 Feed-water, 5, 6
Centrifugal pump, 38, 142, 143, 144, Finite-difference method
146, 147, 150, 370 Explicit scheme, 17, 20, 36, 44, 49,
Chezy, 9 213, 350
Chezy-Manning equation, 9 Fins
Circular calculations, 18, 23, 38, 41, Fin efficiency, 11, 12, 13, 115, 116,
44, 48, 196, 226 237, 301, 304, 306
Colebrook, 15, 80, 97, 98, 99, 308 Formula bar, 23, 24, 25, 28, 29, 40, 69,
Colebrook equation, 9, 15, 80, 97, 98, 89, 92, 110, 309, 312, 313
99, 308 Forward-difference, 77, 226, 287, 293,
Constrained iterative solutions, 90 296, 363, 366, 367
Continuity equation, 2, 154, 161, 169, Fourier number, 219, 226, 231
174, 184, 185, 187, 280, 335, 336, Friction coefficient, 313, 314
337, 338, 341, 356, 359 Friction factor, 8, 9, 15, 16, 80, 81, 82,
Convection heat-transfer, 10, 103 91, 97, 98, 133, 164, 308, 313, 329
Convection heat-transfer analyses, 86 Friction formula, 278, 282, 362, 366,
Cooling tower, 119, 123, 124, 125, 126, 367
373, 374 Functions for matrix operations, 32
Cooling water, 101 Global minimum, 58, 59, 278
Correction factor, 303 Goal Seek
Cross-flow, 13, 14 Iterative solution, 18, 19, 23, 38, 39,
Darcy-Weisbach equation, 8, 9, 31, 80, 40, 41, 44, 48, 51, 53, 80, 81, 82,
91, 329 83, 85, 86, 89, 94, 99, 101, 102,
Dittus, 11, 87, 330 103
Dittus-Boelter equation, 11, 87 GRG Nonlinear, 55, 56, 57, 58, 59, 60,
Double-pipe, 301, 302 63, 64, 65, 77, 111, 112, 115
Economic diameter, 140, 262, 268 GRG Nonlinear method, 55, 56, 58, 59,
Economic pipe diameter, 10, 255, 270 60, 63, 64, 65, 77, 111, 112, 115,
122, 123, 126, 127, 333
Hardy-Cross method, 17, 168, 177, 335 NTU method, 303

Hazen-Williams equation, 9 Nusselt number, 11, 87, 330
Heat conduction equation Objective function, 333
steady, 36, 214, 231, 237 Optimisation analyses
Heat generation, 196, 201, 202, 204, Solver, 5, 14, 16, 17, 19, 53, 56, 75,
205, 207, 208, 213, 226, 344 112, 127, 129, 311, 333
Heat-exchanger, 13, 119, 120, 121, Optimisation analyses, 16, 112
122, 126, 302, 303 Optimum thickness, 113, 320
Heat-exchangers Optimum thickness of insulation, 113,
Optimisation, 13, 119, 325 316
Ideal gas Pipe network, 17, 150, 151, 168, 169,
law, 15, 21, 70, 94, 132 170, 173, 175, 186, 190, 191, 278,
Ideal-gas mixtures, 83 279, 282, 285, 288, 290, 291, 296,
Insulation, 11, 12, 16, 56, 112, 113, 297, 335, 338, 340, 361, 362, 367
114, 123, 316, 317, 318, 319, 320, Power function, 259, 270, 272, 275,
326, 327, 328, 329, 330, 331, 333 282, 289
Insulation, 113, 319, 327 Prandtl number, 11, 87, 330
Interpolation functions, 18, 75, 301, Pump curve, 10, 144, 148, 161
311, 312, 314 Pumps, 5, 7, 10, 132, 135, 142, 154,
Isentropic efficiency, 378 155, 156, 157, 158, 159, 160, 161,
Iterative solution 165, 166, 169, 183, 188, 256, 257,
Constrained, 81, 83, 89, 92, 94, 99, 258, 369
101, 103, 114 R134a, 15, 16, 25, 26, 78
Iterative solutions Rankine cycle, 16
Goal Seek, 16, 97 Raphson, 15, 18, 75, 80, 99, 308
Labelling, 23, 27, 48, 82, 110, 114 Rating analyses, 16, 301
Linear systems, 23, 33, 36 Reynolds number, 8, 11, 16, 31, 81, 82,
LMTD, 13, 101, 120, 122, 301, 302, 87, 91, 98, 313, 314, 329
303, 322, 373, 379 Roughness number, 266
Logical functions, 31 Shell-and-tube, 13, 14, 101, 102
Log-mean temperature difference, 13, Simplex LP method, 60, 62
120, 122, 302, 322, 373 Sizing analyses, 301, 302
Macros, 18 Soave, 15, 75
Major friction, 8 Soave-Redlich-Kwong equation, 75
Matrix functions, 23, 60 System curve, 10, 143, 144
Matrix inverse method, 206 System of linear equations, 36, 49, 50,
Matrix inversion, 18, 33, 36, 209, 223 61, 196, 206, 210, 214, 218, 223,
Minor friction, 8, 10 224, 232, 347
Moody diagram, 8 TDMA, 218, 223, 224, 350, 353, 354
Multi-variable, 95, 173, 250, 283, 316, The LMTD method, 302
333 The ε-NTU method, 301, 303
Newton-Raphson solver, 18, 75, 80, 99, Thermal resistance
308 circuit, 12, 87, 88, 240, 241, 245,
Nonlinear equation, 38, 75, 80, 97, 98, 246, 251, 254, 317, 329, 330, 375
187, 308, 310
NTU, 13, 101, 303, 304
Thermax1, 18, 19, 71, 72, 73, 74, 75, Type-1 problems, 80, 81
80, 84, 85, 86, 97, 99, 101, 102, 109, Type-2 problems, 80
110, 132 Type-3 problems, 81
Tower, 124, 373 VBA, 18, 65, 66, 67, 68, 70, 75, 76, 78,
Trendline feature, 45 98
Tri-diagonal matrix algorithm, 218, ε-NTU- method, 13
223, 353

Book2c With Cover Page v2

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Book2c With Cover Page v2

Uploaded by

Copyright:

Available Formats

Accelerat ing t he world's research.

Optimisation Analyses of Fluid-

Related papers Download a PDF Pack of t he best relat ed papers 

Comput er-Aided T hermal Design Opt imisat ion

Design Analyses of Fluid-Thermal Systems

Design Analyses of Fluid-Thermal Systems

Ghada This book is dedicated to my beloved family:

The Thermax add-in can be downloaded from:

Microsoft Excel is a general-purpose application that is taught to engineering students

3. Solver , VBA, and Thermax 47

5. Thermal insulation 105

6. Determination of the economic pipe diameter 124

7. Excel-based design procedures for pump-pipe systems and double-pipe heat-

7.1.2. The analytical model

9. Thermodynamic and economic optimisation of the air-botomming cycle

10.4. Optimisation analyses of the gas-turbine with inlet-air cooling

w Work-done per kg of the working fluid, usually kJ

Dimensionless Numbers and Groups

1.1. A review of thermofluid principles

q  w  hout  hin  (1.1)

w1  h2  h1   c p T2  T1  (1.2)

w2  h4  h3   c p T4  T3  (1.3)

The total compression work (wtotal ) is then given by:

wtotal  w1  w2  c p T2  T1   T4  T3  (1.4)

Assuming perfect intercooling, i.e., T3 = T1 , Equation (1.4) can be rearranged as:

wout  wHPT  wLPT (1.9)

win  wP1  wP 2 (1.10)

wHPT  h1  h2  (1.11)

wP1  1  y h5  h4  (1.13)

yh2  1  y h5  1  h6 (1.15)

wnet  wout  win (1.16)

qin  h1  h7  (1.17)

  wnet / qin (1.18)

Fraction of extracted steam (y)

1.1.2. Fluid dynamics

Figure 1.5. Schematic diagram of a simple pump-pipe system

f  64 / Re Re < 2300 (1.22)

Where Re is the Reynolds number defined as:

The Colebrook equation is an example of the implicit equations met in thermofluid

f  0.790 ln Re 1.64

1.1.3. Heat transfer

Nu  0.023 Re 0.8 Pr n (1.31)

Figure 1.6. Schematic for an insulated metal pipe

Q  Ti  T  / Rth [W] (1.32)

Where Rth is the combined thermal resistance to heat-transfer by conduction,

Figure 1.9. Circular fins attached to a metal pipe

Another important application of heat-transfer principles is that related to the design

method and the effectiveness-number of transfer units (ε-NTU) method. Complex

Figure 1.8. A parallel-flow shell-and-tube exchanger

Figure 1.9. A cross-flow exchanger with both streams unmixed

1.2. Advantages of computer-aided thermofluid analyses

A. Avoiding excessive simplification of the model

combustion (IC) engines. Traditional air-standard models of IC engines, such as the

By comparison, computer-aided models of IC engines closely mimic the behaviour of

B. Accurate representation of fluid properties and processes

Another important implicit equation for thermofluid analyses is the Colebrook-White

Error in specific volume (%)

C. Dealing with iterative solutions and optimisation analyses

D. Analyses involving complex models

Although the solution is mainly based on the two well-known principles of

Mathematically complex thermofluid models that need computer-aided numerical

1.3. The Excel-based modelling platform for thermofluid analyses

functions not provided by Excel or the Thermax add-in