Thermodynamic Optimization of

Complex Energy Systems

NATO Science Series
A Series presenting the results of activities sponsored by the NATO Science
Committee. The Series is published by 10S Press and Kluwer Academic
Publishers, in conjunction with the NATO Scientific Affairs Division.

A. Life Sciences 10S Press

B. Physics Kluwer Academic Publishers
C. Mathematical and Physical Sciences Kluwer Academic Publishers
D. Behavioural and Social Sciences Kluwer Academic Publishers
E. Applied Sciences Kluwer Academic Publishers
F. Computer and Systems Sciences 10S Press

1. Disarmament Technologies Kluwer Academic Publishers

2. Environmental Security Kluwer Academic Publishers
3. High Technology Kluwer Academic Publishers
4. Science and Technology Policy 10S Press
5. Computer Networking 10S Press

NATO-PCO-DATA BASE

The NATO Science Series continues the series of books published formerly in the NATO ASI
Series. An electronic index to the NATO ASI Series provides full bibliographical references (with
keywords and/or abstracts) to more than 50000 contributions from internatonal scientists publish-
ed in aII sections of the NATO ASI Series.
Access to the NATO-PCO-DATA BASE is possible via CD-ROM "NATO-PCO-DATA BASE" with
user-friendly retrieval software in English, French and German (WTV GmbH and DATAWARE
Technologies Inc. 1989).

The CD-ROM of the NATO ASI Series can be ordered from: PCO, Overijse, Belgium

Series 3. High Technology - VoI. 69

Thermodynamic Optimization of
Complex Energy Systems
edited by

Adrian Bejan
Duke University,
Durham, North Carolina, U.S.A.
and

Eden Mamut
Ovidius University,
Constanta, Romania

Springer-Science+Business Media, B.v.

Proceedings of the NATO Advanced Study Institute on
Thermodynamics and the Optimization of Complex Energy Systems
Neptun, Romania
13-24 July 1998

A C.I. P. Catalogue record for this book is available from the Library of Congress.

ISBN 978-0-7923-5726-1 ISBN 978-94-011-4685-2 (eBook)

DOI 10.1007/978-94-011-4685-2

Printed on acid-free paper

AII Rights Reserved

© 1999 Springer Science+Business Media Dordrecht
Originally published by Kluwer Academic Publishers in 1999
Softcover reprint of the hardcover 1st edition 1999
No part of the material protected by this copyright notice may be reproduced or utilized in
any torm or by any means, electronic or mechanical,including photocopying, recording or
by any intormation storage and retrieval system, without written permission from the
copyright owner.
Table of Contents

Preface ix

Acknowlewdgements Xlll

E.P. GYFTOPOULOS
Presentation of the foundations of thermodynamics in about twelve
one-hour lectures 1

E.P. GYFfOPOULOS
Pictorial visualization of the entropy of thermodynamics 19

A.BEJAN
Thermodynamic optimization of inanimate and animate flow systems 45

A.BEJAN
Constructal flow geometry optimization 61

M.J. MORAN
Fundamentals of exergy analysis and exergy-aided thermal systems
design 73

G. TSATSARONIS
Strengths and limitations of exergy analysis 93

G. TSATSARONIS
Design optimization using exergoeconomics 101

A. VALERO, L. CORREAS and L. SERRA

On-line thermoeconomic diagnosis of thermal power plants 117

J. SZARGUT
Exergy in thermal systems analysis 137

E. SCIUBBA
Allocation of finite energetic resources via an exergetic costing
method 151

E. SCIUBBA
Optimisation of turbomachinary components by constrained
minimisation of the local entropy production rate 163
vi

A. OZTURK
Available energy versus entropy 187

R.L. CORNELISSEN and G.G. HIRS

Exergy analysis in the process industry 195
R.L. CORNELISSEN and G.G. HIRS
Exergetic life cycle analysis of components in a system 209
I. DINCER and M.A. ROSEN
The intimate connection between exergy and the environment 221
Y.A. GadUS and O.E. ATAER
Effect of variation of environmental conditions on exergy and on power
conversion 231
Y.A. GadUS
Bond graphs and influence coefficients applications 241
P.F. MATHIEU
Repowering options for existing power plants 251
P.F. MATHIEU
Cogeneration based on gas turbines, gas engines and fuel cells 261
A.1. LEONTIEV
Gas dynamics cycles of thermal and refrigerating machines 271

Z. FONYO and E. REV

Energy saving techniques in distillation thermodynamic efficiency
and energy conservation 279
E.MAMUT
Second law based optimization of systems with thermomechanial
dissipative processes 297
V.BADESCU
Solar energy conversion into work: simple upper bound efficiencies 313

A. SEGAL
Thermodynamic approach to the optimization of central solar energy
systems 323
Y. IKEGAMI and H. UEHARA
Thermodynamic optimization in ocean thermal energy conversion 335
A. DE VOS
How to unify solar energy converters and Carnot engines 345
 vii

S. SIENIUTYCZ
Optimal control for multistage endoreversible engines with heat and
mass transfer 363
M.L. FElDT
Thermodynamics and optimization of reverse cycle machines 385
M. COSTEA, M. FElDT and S. PETRESCU
Synthesis on Stirling engine optimization 403

B.ANDRESEN
Minimizing losses-tools of finite-time thermodynamics 411

P. SALAMON
Physics versus engineering of finite-time thermodynamic models and
optimizations 421
J.V.C. VARGAS
Optimization and simulation of time dependent heat driven refrigerators
with continuous temperature control 425
c. WU
Intelligent computer-aided design, analysis, optimization and
improvement of thermodynamic systems 437

G. STANESCU
A study of the large oscillations of the thermodynamic pendullum by
the method of cubication 445
Author Index 453
Subject Index 459
Preface

This book provides a comprehensive assessment of the methodologies of

thermodynamic optimization, exergy analysis and thermoeconomics, and their
application to the design of efficient and environmentally-sound energy systems.
The contributors cover a wide spectrum of educational backgrounds and research
orientations, from engineering to physics, and from fundamentals to actual
industrial applications. Each chapter was presented to a select and highly-animated
audience of 100, at the NATO Advanced Study Institute held on July 13-24 in
Neptun, Romania.
The chapters are organized in a sequence that begins with pure thermo-
dynamics and progresses toward the blending of thermodynamics with other
disciplines (e.g., heat transfer and cost accounting) in the pursuit of more realistic
models and more significant optimization results. This is also how this very active
field has developed in time. In this sequence, three methods of analysis and
optimization stand out, and are documented through specific examples and
applications.
First, entropy generation minimization combines from the start the laws of
classical thermodynamics with principles of transport phenomena (e.g., heat transfer
and fluid mechanics) in the construction of simple models. Entropy generation is
expressed as a function of the physical parameters (geometry, size, materials) of
the device. Their minimization is then carried out subject to global constraints
such as overall size and time of operation. The specific tutorial examples collected
in this book show how the method became established in the decades of the '50s,
'60s and '70s in engineering. The significant growth that has occurred in the '80s
and '90s, and the new titles that have been attached recently to the method
(endoreversible, exoirreversible, finite-time, finite-size, finite-resources) are also
documented (see e.g., Am. J. Phys., 1994, Vol. 62, pp. 11-12 and 1. Non-Equilibrium
Thermodynamics, 1996, Vol. 21, pp. 239-242). It is shown that system "structure"
springs out of the optimization subject to global constraints.
Second, exergy (or availability) analysis shows how the destruction of "useful"
energy can be traced at the component and sub-component level, by accounting for
irreversible flows, interactions, and regions of nonequi1ibrium. The contributors
stress the importance of describing the exergy-reference environment (the "dead"
state) unambiguously and completely. They also document the merits of the
method with respect to pinpointing system features that could be improved
thermodynamically. Presented are also new techniques (computer-aided, modular,
artificial intelligence), and several exergy-based efficiencies for evaluating
thermodynamic performance and the improvements that result from optimization.
ix
x

Third, thennoeconomics (or exergoeconomics) combines thennodynamics with

principles of cost accounting, and the ultimate objective is cost minimization. The
cost of each component and stream is related to its exergy content. Capital costs of
the components and the environmental impact are taken into account systemati-
cally. The authors present several computer-based modular techniques for imple-
menting the method. They also document the expansion of the method into life-
cycle analysis, and its application to specific industrial examples of complex
energy systems.
In summary, this book reviews the current directions in a field that is both
extremely important in practice and alive intellectually. Additionally, new
directions for research on thermodynamics and optimization are unveiled. For
example, the occurrence of shape and structure in natural flow systems is now
reasoned on the basis of geometry optimization subject to global constraints (con-
structal theory).

* * *

We close with a few thoughts directed to student readers of this book. The
Neptun institute was truly a historic event in a field that is experiencing tremendous
growth. The invited contributors to this volume are arguably the best known
authors in the field. No less than six of the invited speakers have won the premier
thermodynamics award bestowed by the American Society of Mechanical
Engineers International-the James Harry Potter Gold Medal. They have written
books and chronicled our discipline. They have come together to pass on to the
next generation the method, with all its strengths, limitations and areas for real
growth.

1800 1900 2000

Heat transfer Entropy generation

through
heat and flu id flow

Engineering thermodynamics

Figure 1. The predicted merger of classical thermodynamics with heat transfer and fluid mechanics
(A. 8ejan, Entropy Generati(m through Heat and Fluid Flow, Wiley, New York, 1982).
 xi

Creative individuals have a tendency to overstate the importance of their

contributions, especially in the early stages when they must pursuade others that
the contributions are both novel and valuable. Students are advised to read this
book critically and to question authority. Conflicting claims and disagreements
between the authors of the following chapters can be clarified by reading the
original sources. By knowing the history of your discipline, your work can be more
original and valuable to others.
Thermodynamic optimization has made spectacular progress. The prediction
made graphically in 1982 in the first textbook dedicated the method (Fig. 1) has
come true: classical thermodynamics has merged with heat transfer and fluid
mechanics into a method of modeling and optimization subject to global
constraints. This method unites education with research, engineering with physics,
and fundamentals with applications. There is even a rush to teach thermodynamics
with heat transfer in a single course, as was predicted in Fig. 1. Today, this broad
field is being enlarged still more by additional considerations such as cost
accounting, life-cycle analysis, computer-aided design, and natural self-
organization and self-optimization (constructal theory). The path is open to new
ways and new ideas.

July 1998 Adrian Bejan

Duke University
Durham, NC, USA

Eden Mamut
Ovidius University
Constanta, Romania
Acknowledgement

We wish to thank the following sponsors of the NATO Advanced Study Institute on
Thermodynamics and the Optimization of Complex Energy Systems, Neptun, Romania,
July 13-24, 1998:

NATO Scientific and Environmental Affairs Division

Ovidius University, Constanta
Duke University
Neptun - Olimp S.A.
Romanian Electric Power Authority (RENEL)
Tomis S.A.
Santierul Naval Constanta
Petromidia S.A.
Danubius Travel Agency
Adrian & Mary Bejan

We also thank the following colleagues for their personal contributions before and
during the institute:

Adrian Bavaru, President, Ovidius University, Constanta

Aureliu Leca, President, Romanian Electric Power Authority
Constantin Bratianu, General Director, Ministry of National Education, Romania
Dumitru Filip, General Director, s.c. Neptun-Olimp s.a.
Florentin Filote, Technical Director, s.c. Neptun-Olimp s.a.
Ya1cin Gogiis, International Center for Applied Thermodynamics
Adina Gogan, Assistant Professor, Ovidius University, Constanta
Teodor Popa, Assistant Professor, Ovidius University, Constanta
Ion Aldea, Assistant Professor, Navy Scientific Researcher Center, Constanta

The Scientific Commitee of the institute was:

Adrian Bejan, Duke University, USA

Elias P. Gyftopoulos, Massachusetts Institute of Technology, USA
Sadik Kakac, University of Miami, USA
Eden Mamut, Ovidius University, Romania
Michael J. Moran, The Ohio State University, USA
George Tsatsaronis, Technical University of Berlin, Germany

This book and the preliminary version used in Neptun were prepared by

Deborah Alford, Duke University

Madalina Ursu, Ovidius University
Rita Avram, Ovidius University

A.B. &E. M.
xiii
PRESENTATION OF THE FOUNDATIONS OF THERMO-
DYNAMICS IN ABOUT TWELVE ONE-HOUR LECTURES

E.P. GYFTOPOULOS
Massachusetts Institute of Technology
Department of Nuclear Engineering, Room 24-111
Cambridge, MA 02139, USA

1. Introduction
In many expositions of thermodynamics, heat is introduced at the outset of the
logical development as an intuitive and self-evident concept, independent of the
laws of the subject [1, 2].
For example, Feynman [3] describes heat as one of several different forms of
energy related to the jiggling motion of particles stuck together and tagging along
with each other [3, pp. 1-3 and 4-2], a form of energy which really is just kinetic
energy - internal motion [3, p. 4-6], and is measured by the random motions of
the atoms [3, p. 10-8]. Tisza [4] argues that such slogans as "heat is motion," in
spite of their fuzzy meaning, convey intuitive images of pedagogical and heuristic
value. Brush [5] entitles his two-volume history of the kinetic theory of gases "The
kind of motion we call heat."
There are at least two problems with these illustrations. First, work and heat
are not stored in a system. Second, and perhaps more important, concepts of
mechanics are used to justify and make plausible the notion of heat which is beyond
the realm of mechanics. In spite of these logical drawbacks, the trick works because
at first the student finds the idea of heat harmless, even natural. But the situation
changes drastically as soon as heat is used to define a host of new concepts, less
natural and less harmless. Heat is contrasted to work and used as an essential
ingredient in the first law. The student begins to worry because heat is less definite
than and not as operational as work.
An attempt to address the first problem is made in some expositions. Landau
and Lifshitz [6] define heat as the part of an energy change of a body that is not
due to work. Guggenheim [7] defines heat as an exchange of energy that differs
from work and is determined by a temperature difference. Keenan [8] defines heat
as that which transfers from one system to a second system at lower temperature,
by virtue of the temperature difference, if the two are brought into communication.
Similar definitions are adopted in other textbooks.
These definitions, however, are ambiguous, and none addresses the basic prob-
lem, that is, the existence of exchanges of energy that differ from work, and cannot
be justified by mechanics. Such exchanges are one of the striking results of ther-
modynamics, and are due to entropy being a property of matter. Hatsopoulos
and Keenan [9] have pointed out explicitly that without the second law heat and

A. Bejan and E. Manila (eds.), Thermodyntlmic OptimiZlltion o/Complex Energy Systems, 1-18.
© 1999 Kluwer Academic Publishers.
2

work would be indistinguishable and, therefore, a satisfactory definition of heat is

unlikely without a prior statement of the second law.
In our experience, whenever heat is introduced before the first law, and then
used in the statement of the second law and in the definition of entropy, the student
cannot avoid but sense ambiguity and lack of logical consistency. This results in
the wrong but unfortunately widely spread conviction that thermodynamics is a
confusing, ambiguous, and hand-waving subject.
During the past three decades of teaching thermodynamics to students from
all over the globe, we have sensed a need to address these concerns. In response,
we have composed an exposition [10] in which we strive to develop the basic con-
cepts without ambiguities and logical inconsistencies, building upon the student's
sophomore background in introductory physics.
The basic concepts and principles are introduced in a novel sequence that elim-
inates the problem of incomplete definitions, and that is valid for both macroscopic
and microscopic systems, and for both stable or thermodynamic equilibrium states
and for states that are not stable equilibrium. The laws of thermodynamics are
presented as partial complements to the incomplete law of mechanical dynamics.
Heat plays no role in the first law, the definition of energy, the second law, the
definition of entropy, and the general concepts of energy and entropy exchanges
between interacting systems. It emerges as a consequence of these concepts and
laws. The discussion that follows is a summary of the key aspects of the novel
logical sequence that we propose in our exposition. Though it makes no reference
to the ideas of quantum thermodynamics, the exposition is construed so as to be
entirely compatible with quantum theory. Such compatibility is absolutely essen-
tial to the resolution of the dilemma about the relation between mechanics and
thermodynamics.
Experience at the University of Brescia, Italy, and MIT, Tufts University,
Northeastern University, Washington University in St. Louis, and the University
of Florida in the USA indicates that the fundamental ideas can be presented in
about twelve one-hour lectures, with sufficient time for explanations and illustra-
tions short of elaborate proofs. In what follows, we provide a syllabus of the twelve
lectures without the necessary numerical and graphical illustrative examples given
in [10]. However, the use of numerical and graphical illustrations is absolutely
necessary and strongly recommended.
A variant of the sequence of the topics is to begin with the energy versus
entropy graphs of Lecture 11 without defining any term. This approach gives the
students a visual conception of the ideas that are presented in Lectures 1 to 10,
and stimulates interest in the details of the course.

2. Lecture 1
2.1. GENERAL THERMODYNAMICS
We define general thermodynamics or simply thermodynamics as the study of mo-
tions of physical constituents (particles and radiations) resulting from externally
applied forces, and from internal forces (the actions and reactions between con-
stituents). This definition is identical to that given by Timoshenko and Young
about mechanical dynamics [11]. However, because of the second law, the defini-
tion encompasses a much broader spectrum of phenomena than mechanical dy-
namics. We will see that thermodynamics accounts for phenomena with both zero
 3

and positive values of entropy, whereas classical mechanics and ordinary quantum
mechanics account only for phenomena with zero values of entropy.

2.2. KINEMATICS: CONDITIONS AT AN INSTANT IN TIME

In kinematics we give verbal definitions of the terms system, environment, prop-

erty, and state so that each definition is valid without change in any physical
theory, and involves no statistics attributable to lack of information. The defini-
tions include innovations.

2.2.1. System
A system is defined as a collection of constituents. For our purposes, we consider
only one constituent which is determined by the following specifications: (a) The
type of the constituent and the range of values of its amount; (b) The type and the
range of values of the parameters that fully characterize the external forces exerted
on the constituent by bodies other than the constituent itself, such as the volume of
a container. The external forces do not depend on coordinates of bodies other than
those of the constituent of the system. For our purposes, we consider only volume
as a parameter; and (c) The internal forces between particles of the constituent,
such as intermolecular forces. The internal forces depend on the coordinates of all
the interacting particles.
Everything that is not included in the system is the environment. We denote
the amount of the constituent by n, and the value of the volume by V. At any
instant in time, the amount of the constituent and the volume have specific values.
We denote these values by n and V with or without additional subscripts.

2.2.2. Property
By themselves, the values of the amount of the constituent and of the volume at
an instant in time do not suffice to characterize completely the condition of the
system at that time. We also need the values of properties at the same instant
in time. A property is an attribute that can be evaluated at any given instant in
time by means of a set of measurements and operations that are performed on the
system and result in a numerical value - the value of the property. This value
is independent of the measuring devices, other systems in the environment, and
other instants in time.
Two properties are independent if the value of one can be varied without
affecting the value of the other.

2.2.3. State
For a given system, the values of the amount of the constituent, the value of the
volume, and the values of a complete set of independent properties encompass all
that can be said about the system at an instant in time and about the results of
any measurements that may be performed on the system at that same instant. We
call this complete characterization of the system at an instant in time the state
of the system. Without change, this definition of state applies to any branch of
physics.
4

3. Lecture 2

3.1. DYNAMICS: CHANGES OF STATE IN TIME

3.1.1. Equation of Motion

The state of a system may change in time either spontaneously as the constituent
tries to conform to the external and internal forces or as a result of interactions with
other systems, or both. A system that experiences only spontaneous changes of
state is called isolated. A system that is not isolated interacts with the environment
in a number of different ways, some of which may result in net flows of properties
between the system and the environment. For example, collision between a system
and the environment results in the flow or transfer of momentum to or from the
system.
The relation that describes the evolution of the state of an isolated system -
spontaneous changes of state - as a function of time is the equation of motion. Cer-
tain time evolutions obey Newton's equation which relates the total force F on each
system particle to its inertial mass m and acceleration a so that F = mao Other
evolutions obey the time-dependent Schroedinger equation, that is, the quantum-
mechanical equivalent of Newton's equation. Other experimentally observed time
evolutions, however, do not obey either of these equations. So the equations of
motion that we have are incomplete. The discovery of the complete equation of
motion that describes all physical phenomena remains a subject of research at
the frontier of science - one of the most intriguing and challenging problems in
thermodynamics.
Many features of the equation of motion have already been discovered. These
features provide not only guidance for the discovery of the complete equation but
also a powerful alternative procedure for analyses of many time-dependent, prac-
tical problems. Two of the most general and well-established features are captured
by the consequences of the first and second laws of thermodynamics presented in
subsequent lectures.

3.1.2. Interactions and Processes

Rather than through the explicit time dependence which requires the complete
equation of motion, a change of state can be described in terms of: (a) the in-
teractions that are active during the change of state; (b) the end states of the
system, that is, the initial and final states; and (c) conditions on the values of
properties of the end states that are consequences of the laws of thermodynamics,
that is, conditions that express not all, but the most general and well-established
features of the complete equation of motion. Each interaction is characterized by
means of well-specified net flows of properties across the boundary of the system.
For example, after defining the properties energy and entropy, we will see that
some interactions involve the flow of energy across the boundary of the system
without any flow of entropy, whereas other interactions involve the flows of both
energy and entropy. Among the conditions on the values of properties, we will see
that the energy change of a system must equal the energy exchanged between the
system and its environment, whereas the entropy change must not be less than
the entropy exchanged between the system and its environment.
 5

4. Lecture 3
4.1. ENERGY AND ENERGY BALANCE

Energy is a concept that underlies our understanding of all physical phenomena,

yet its meaning is subtle and difficult to grasp. It emerges from a fundamental prin-
ciple known as the first law of thermodynamics but is not a part of the statement
of that law.

4.2. FIRST LAW OF THERMODYNAMICS

The first law asserts that any two states of a system may always be the initial
and final states of a change (weight process) that involves no net effects external
to the system except the change in elevation between ZI and Z2 of a weight, that
is, a mechanical effect. Moreover, for a given weight, the value of the expression
M g(ZI - Z2) is fixed only by the end states of the system, where M is the inertial
mass of the weight, and g the gravitational acceleration.

4.2.1. Definition of Energy

One consequence (theorem) of the first law is that every system A in any state Al
has a property called energy, with a value denoted by the symbol E I . The energy
EI can be evaluated by a weight process that connects Al and a reference state
Ao to which is assigned an arbitrary reference value Eo so that

(1)

Energy is an additive property, that is, the energy of a composite system is the
sum of the energies of its subsystems. Moreover, energy either has the same value
at the final time as at the initial time if the process is a zero-net-effect weight
process, or remains invariant in time if the process is spontaneous. In either of
these two processes, Z2 = ZI and E(t2) = E(tr) for time t2 greater than tI, that
is, energy is conser·ved. Energy conservation is a time-dependent result. Here it
is obtained without use of the general equation of motion. It is noteworthy that
energy is defined for any system (both macroscopic and microscopic) and for any
state, and is not statistical.

4.2.2. Energy Balance

In the course of interactions, energy can be exchanged between a system and its
environment. Denoting by EA+- the amount of energy exchanged between the
environment and system A in a process that changes the state of A from Al to
A 2 , we can derive the energy balance. This balance is based on the additivity of
energy and energy conservation, and reads

(2)

where EA+- is positive if energy flows into A. In words, the energy change of system
A must be accounted for by the net energy crossing the boundary of the system.
6

4.3. MASS BALANCE

The values of energy disclosed by the first law are relative because the choice of
the reference value Eo is arbitrary. We can assign absolute values by using the
theory of special relativity and the concept of (inertial) mass.
In general, and in contrast to energy, mass is neither additive nor conserved.
However, in the absence of nuclear reactions, and creation and annihilation re-
actions, the mass changes caused within a system by energy exchanges between
either constituents and the electromagnetic field or by chemical reactions, or both,
are negligible with respect to the mass of the system. As a result, we establish a
very useful tool, the mass balance, i.e.,

(3)

where mAt- is the net amount of mass exchanged with the environment of system
A. The mass mAt- is positive if mass flows into A.

4.4. UNITS OF BOTH ENERGY AND MASS

Provide a table of the different units and the conversion factor from one unit to
another for both energy and mass.

5. Lecture 4
5.1. TYPES OF STATES
Because the number of independent properties of a system is infinite even for a
system consisting of a single particle with a single translational degree of free-
dom - a single variable that fixes the configuration of the system in space -
and because most properties can vary over a range of values, the number of pos-
sible states of a system is infinite. To facilitate the discussion of these states, we
classify them into different categories according to their time evolutions. This clas-
sification brings forth many important aspects of thermodynamics, and provides
a readily understandable motivation for the the introduction of the second law of
thermodynamics.
An unsteady state is one that changes as a function of time because of interac-
tions of the system with its environment. A steady state is one that does not change
as a function of time despite interactions of the system with its environment. A
non equilibrium state is one that changes spontaneously as a function of time, that
is, a state that evolves in time without any effects on or interactions with the
environment. An equilibrium state is one that does not change as a function of
time while the system is isolated - a state that does not change spontaneously.
An unstable equilibrium state is an equilibrium state that may be caused to pro-
ceed spontaneously to a sequence of entirely different states by means of a minute
and short-lived interaction that has only an infinitesimal temporary effect on the
state of the environment. A stable equilibrium state is an equilibrium state that
can be altered to a different state only by interactions that leave net effects in the
environment of the system. These definitions are identical to the corresponding
definitions in mechanics but include a much broader spectrum of states than those
encountered in mechanics.
 7

Starting either from a nonequilibrium state or from an equilibrium state that

is not stable, energy and only energy can be transferred out of a system and
affect a mechanical effect without leaving any other net changes in the state of
the environment. For example, a charged electricity storage battery can supply
energy to an electrical appliance. In contrast, experience shows that, starting from
a stable equilibrium state having any value of energy, the mechanical effect just
cited is impossible. For example, an internally discharged battery can have the
same amount of energy as a charged battery and yet the former cannot energize
an electrical appliance. This impossibility is one of the most striking consequences
of the first and second laws of thermodynamics. It is consistent with innumerable
experiences. The second law will be introduced in the next section.

6. Lecture 5
6.1. STABILITY AND THE SECOND LAW

To a scientist or engineer familiar with mechanics, the existence of stable equilib-

rium states at any value of energy is not self-evident. It was first recognized by
Hatsopoulos and Keenan [12] as the essence of all correct statements of the second
law of thermodynamics. We concur with this recognition.

6.2. SECOND LAW OF THERMODYNAMICS

The second law asserts that, among all the states of a system with a given value of
energy E, and given values of the amount n of the constituent and of the volume
V, there exists one and only one stable equilibrium state.
The existence of a stable equilibrium state for each set of values E, n, V and,
therefore, the second law of thermodynamics cannot be derived from the laws of
mechanics. In mechanics among all the states of a system with a fixed value of
the amount of the constituent and a fixed value of the volume, the only stable
equilibrium state is that of lowest energy, that is, the state of least potential
energy and zero kinetic energy. In contrast, the second law avers that a stable
equilibrium state exists for each value of the energy. It follows that the class
of states contemplated by thermodynamics is broader than the class of states
contemplated by mechanics.

6.2.1. Impossibility of Perpetual Motion Machine of the Second Kind

The existence of stable equilibrium states for various conditions of a system has
many theoretical and practical consequences. One consequence is that, starting
from a stable equilibrium state of any system, no energy can be extracted to
affect solely a mechanical effect while the values of the amount of the constituent
and of the volume of the system experience no net changes. If the extraction
were possible, the energy could be returned to the system, and accelerate it to
a nonzero speed. Thus the stable equilibrium state would have been changed to
another state without any effect on the environment. Such a change violates the
definition of stable equilibrium. Because the concept of stable equilibrium is part
of the second law, such a violation is also a violation of that law. This consequence
is often referred to as the impossibility of the perpetual motion machine of the
second kind (PMM2). In some expositions of thermodynamics it is taken as the
8

statement of the second law. Here, it is only one aspect of both the first and the
second laws.

7. Lecture 6

7.1. ADIABATIC AVAILABILITY

For given values of n and V, another consequence of the two laws is that not all
states can be changed to the state of minimum energy of the system by means
of interactions which result solely in the rise of a weight - interactions which
have solely a mechanical effect on the environment. This is a generalization of
the impossibility of a PMM2 because the initial state is not a stable equilibrium
state. In essence we prove the existence of another property that we call adiabatic
availability and denote by IV.
The adiabatic availability IV 1 of a system in state Al represents the largest
amount of energy that can be transferred from the system to a weight in a reversible
process, as A changes from initial state Al to a stable equilibrium state Ao. The
value of IV is nonnegative.

7.1.1. Reversible and Irreversible Processes

Among other distinctions, a process may be either reversible or irreversible. A
process is reversible if it can be performed in at least one way such that both the
system and the environment can be restored to their respective initial states. A
process is irreversible if it is impossible to perform it in such a way that both the
system and its environment can be restored to their respective initial states.

7.2. GENERALIZED ADIABATIC AVAILABILITY

The concept of adiabatic availability can be generalized to a reversible process

of system A in the course of which the only effect on the environment is the
rise or fall of a weight as the system changes from state Al with values nl and
Vl to a stable equilibrium state Ao with values no and Vo. We define the energy
exchanged between the system and the weight in the course of the process just cited
as generalized adiabatic availability. The energy exchanged between the system
and the weight is optimum. It is the largest if transferred from the system to the
weight, and the smallest if transferred from the weight to the system. For simplicity,
we denote generalized adiabatic availability by the same symbol IV as adiabatic
availability. If energy is transferred out of the system, the adiabatic availability is
denoted by a positive number, and if transferred into the system, by a negative
number.
Like energy, \)! is a well-defined property of any system in any state. Unlike
energy, neither adiabatic availability nor generalized adiabatic availability is ad-
ditive. If nl = no and Vi = Vo, it is noteworthy that the generalized adiabatic
availability reduces to the adiabatic availability.
 9

8. Lectures 7 and 8
8.1. ADIABATIC AVAILABILITIES OF A SYSTEM AND A RESERVOIR

Additive properties are very useful because they facilitate the discussion of prop-
erties of composite systems - systems that consist of two or more subsystems. In
striving to define an additive availability, we investigate the adiabatic availabilities
of a composite consisting of a system and of a special reference system called a
reservoir. These availabilities are generalizations of the concept of motive power of
fire first introduced by Carnot [13]. They are generalizations because he restricted
the composite to consist of two reservoirs with fixed values of both the respective
constituents and the respective volumes, and we do not use his restrictions.

8.2. MUTUAL STABLE EQUILIBRIUM

If a composite of two or more subsystems is in a stable equilibrium state, the

subsystems are said to be in mutual stable equilibrium. For example, if system C
is in a stable equilibrium state and is a composite of two systems A and B, we say
that A is in mutual stable equilibrium with B.
In contrast, if each of two systems is in a stable equilibrium state, their compos-
ite is not necessarily in a stable equilibrium state and, therefore, the two systems
are not necessarily in mutual stable equilibrium.

8.3. RESERVOIR

A reservoir is an idealized kind of system that provides useful reference states

both in theory and in applications, and that behaves in a manner approaching
the following limiting conditions: (a) it passes through stable equilibrium states
only; (b) in the course of finite changes of state, it remains in mutual stable
equilibrium with a duplicate of itself that experiences no such changes; and (c)
at constant values of both the constituent and the volume of each of two reservoirs
initially in mutual stable equilibrium, energy can be transferred reversibly from
one reservoir to the other with no net effect whatsoever on the environments of
the two reservoirs.

8.4. AVAILABLE ENERGY

We consider a composite of system A in state AI, and reservoir R in state R I .

Without net changes of the values of both the amounts of constituents and the
volumes of A and R, we raise the question: What is the largest energy that can
be exchanged between the composite and its environment so that the only effect
on the latter is the rise of a weight?
The answer is the adiabatic availability of the composite and, therefore, con-
forms to the following assertions: (a) it is limited and equal to the energy exchanged
during a reversible weight process in which system A starts from state AI, interacts
with both the weight and the reservoir, and ends being in mutual stable equilib-
rium with the reservoir; (b) it is the same for all reversible weight processes such
that system A begins from a given state Al and ends in a state in which A and R
are in mutual stable equilibrium; (c) it is independent of the initial state of R; and
10

(d) it is the same for all reservoirs that are initially in mutual stable equilibrium
with R.
We denote this largest value hi{ of, and conclude that OR is a property of the
composite of A and R. We call 0 available energy.

8.5. GENERALIZED AVAILABLE ENERGY

We consider weight processes of a composite of system A and reservoir R in which

the values of both the amount of the constituent and of the volume of A experience
net changes. For example, system A is in state Al with values nI and VI and, at
the end of a weight process of the composite that affects only the elevation of
a weight in the environment of the composite, the values are no and Vo. Again,
system A is in state A2 with values n2 and V2 and at the end of the weight process
of the composite the values are the same no and Vo as in the first example. With
these specifications, we raise the question: Under the conditions just cited, what
is the optimum amount of energy that can be exchanged between the composite
and the weight?
The answer is the generalized adiabatic availability of the composite. It is
the largest if the energy transfer is from the composite to the weight, and the
smallest if the energy transfer is from the weight to the composite. For simplicity,
we denote the optimum by the same symbol OR as for available energy, and call
it the generalized available energy of the composite of A and R. For state Al with
nI and VI, and final values no and Vo , the value of OR is denoted as Of. For state
A2 with n2 and V2 , and final values again no and Vo , the value of OR is denoted
Or. For nI = n2 = no and VI = V2 = Vo , and a given reservoir R, the generalized
available energy is identical with the available energy.
It is noteworthy that generalized available energy is defined for any state of
any system regardless of whether the state is steady, unsteady, nonequilibrium,
equilibrium, or stable equilibrium, and regardless of whether the system has many
degrees of freedom or one degree of freedom, or whether the size of the system
is large or small. In addition, like energy, OR is an additive and nonstatistical
property of the composite system of A and R.

8.5.1. Energy and Generalized Available Energy Relations

In principle, given any two states Al and A2 of system A and a reservoir R, we
can tabulate the values of the energies EI and E 2 , and the generalized available
energies Of and Or. For a process of A alone from state Al to state A2 that affects
only the elevation of a weight in the environment of A, that is, a weight process of
A and not of the composite of A and R, we can prove the following. If the process
of A alone is reversible, then

(4)

and if the process of A alone is irreversible, then

(5)
 II

9. Lecture 9
9.l. ENTROPY AND ENTROPY BALANCE

Entropy is the concept that distinguishes thermodynamics from all other branches
of science and engineering. It is an additional dimension of the space defined by
the properties of a system, and has important, strange and unusual characteristics
that have attracted the attention of thousands of scientists and engineers, and that
have stimulated animated discussions.

9.1.1. Definition of Entropy

A system A in any state Al has the two properties: energy E 1 , and generalized
available energy nf with respect to a given auxiliary reservoir R. These two prop-
erties determine a third one we call entropy and denote by the symbol 8 1 . It is
a property in the same sense that energy is a property, or momentum is a prop-
erty. It can be evaluated by means of the auxiliary reservoir R, a reference state
Ao, with energy Eo and generalized available energy n~, to which is assigned a
reference value 80, and the expression

(6)

where C"R is a well-defined positive constant that depends on the auxiliary reservoir
R only. Entropy 8 is shown to be independent of the reservoir, that is, indeed the
reservoir is auxiliary and is used only because it facilitates the definition of 8. It
is also shown that 8 can be assigned absolute values that are nonnegative.
The concept of entropy introduced here differs from and is more general than
that in practically all textbooks. It does not involve the concepts of heat and
temperature which have not yet been defined; it is not restricted to large systems;
it applies to both macroscopic and microscopic systems, including a system with
one particle with only one (translational) degree of freedom, that is, even one
particle has entropy; it is not restricted to stable or thermodynamic equilibrium
states; it is defined for both stable equilibrium and not stable equilibrium states
because both energy and generalized available energy are defined for all states; and
most certainly, it is not statistical because both energy and generalized available
energy are not statistical.
The dimensions of 8 depend on the dimensions of both energy and CR. It
turns out that the dimensions of CR are independent of mechanical dimensions,
and are the same as those of temperature (defined later). The unit of CR chosen
in the International System of units is the kelvin, denoted by K. So entropy can
be expressed in J / K or other equivalent units.
Because both energy and generalized available energy are additive, entropy is
also an additive property. In contrast, whereas energy remains constant in time if
the system experiences either a spontaneous process or a zero-net-effect interaction
only with a weight, using relations 4 t06 we prove that entropy remains constant
if either of these processes is reversible, and increases if either of these processes
is irreversible. These two features are known as the principle of non decrease of
entropy.
The spontaneous entropy creation or increase during an irreversible process is
called entropy generated by irreversibility. It is positive. Like energy conservation,
12

entropy non decrease is a time-dependent result which here is obtained without use
of the unknown general equation of motion.

9.1.2. Entropy Balance

Like energy, entropy can be exchanged between systems by means of interactions.
Denoting by SA<.- the amount of entropy exchanged between system A and the
environment as a result of all interactions involved in a process in which the state
of A changes from Al to A 2 , we derive a very important analytical tool, the entropy
balance, that is,
(7)
where Sirr is positive or at least zero and represents the entropy generated spon-
taneously within system A during the time interval from tl to t2 required to affect
the change from state Al to state A 2 . Spontaneous entropy generation within a
system occurs if the system is in a state that is not stable equilibrium as the
constituent tries to conform to both the external and internal forces, and tends
towards stable equilibrium.

10. Lecture 10
10.1. STABLE EQUILIBRIUM STATES AND THE STATE PRINCIPLE
An isolated system admits an infinite number of states that share the same values
of energy E, amount of constituent n, and volume V. Some of these states are
nonequilibrium, some are equilibrium and, according to the second law, one and
only one is a stable equilibrium state. It follows that the value of any property of
the system in a stable equilibrium state is uniquely determined by the values E,
n, V, that is, it can be written as a function of E, n, V only. This result is known
as the stable equilibrium state principle or simply the state principle.

10.2. THE FUNDAMENTAL RELATION

A system in general has an infinite number of independent properties. If we focus
on the special family of states that are stable equilibrium, however, the state
principle asserts that the value of each of these properties is uniquely determined
by the values E, n, V or the values of an equivalent set of three other independent
properties. In contrast, for states that are not stable equilibrium, the values of E,
n, V are not sufficient to specify the values of all independent properties.
In general, one property of great interest is the entropy of stable equilibrium
states which, according to the state principle, must satisfy the relation of the form
S = S(E, n, V) (8)
that is, S must be solely a function of E, n, V. The function S(E, n, V) is called the
fundamental relation, and is concave - has a negative second partial derivative
- with respect to each of the independent variables E, n, V. The entropy of each
unique stable equilibrium state is larger than that of any other state with the same
values E, n, V. This result is known as the highest entropy principle.
Upon solving the fundamental relation for E, we find the energy relation, that
is, we express the energy as a function of S, n, V
E = E(S, n, V) . (9)
 13

Traditionally, if the energy of a system in a stable equilibrium state depends only on

S, n, V, it is called internal energy, and is denoted by U. However, this terminology
must not be misconstrued as implying that some systems have both internal and
external energy.

10.3. TEMPERATURE, CHEMICAL POTENTIAL, AND PRESSURE

Upon writing the differentials of the fundamental relation and the energy relation,
we find

dS (~~) n,V dE+ (~~) E,V dn+ (;~) E,n dV (10)

dE = (~~) n,V dS + (~~) S,V dn + (~~) S,n dV (11)

Thus, we define

T = (8E) = temperature; p = _ (8E) = pressure;

8S n,v 8V S,n

/-L = (~!) s,v = chemical potential .

Moreover, upon solving equation (10) for dE and comparing the result with equa-
tion (11), we find

T= 1
(8S18E)n,v
/-L = -T(8S18n)E,v ; p = T(8S18V)E,n .

The concepts of temperature, chemical potential, and pressure are valid only for
stable equilibrium states, and are very useful because, if two systems are in mutual
stable equilibrium, then the temperature, chemical potential, and pressure of one
system is equal to the temperature, chemical potential, and pressure of the other,
respectively. In other words, we have temperature equality, chemical potential
equality, and pressure equality.
It is noteworthy that either of the differentials
dS = dEIT - (/-LIT) dn + (pIT) dV (12)

dE T dS + /-L dn - p dV (13)
is a mathematical relation between properties of neighboring stable equilibrium
states, and not the result of any interactions. Temperature is measured on different
scales. The same is true for chemical potential and pressure.

11. Lecture 11
11.1. ENERGY VERSUS ENTROPY GRAPHS
Because they are defined in terms of the values of the amount of the constituent,
the volume, and a complete set of independent properties, states can in principle
14

Line of the

ij~)1I, v=T
Slope =
E,I------

o Entropy S

Figure 1. Typical energy versus entropy graph. All concepts of the introductory lectures,
including work and heat interactions, can be illustrated on this graph.

be represented by points in a multidimensional geometrical space with one axis

for the amount, one for the volume, and one for each independent property. Such
a representation, however, would not be enlightening because the number of inde-
pendent properties is infinite. Nevertheless, useful information can be summarized
by first cutting the multidimensional space with a surface corresponding to given
values of both the amount of the constituent and the volume, and then projecting
the result onto a two-dimensional plane - a plane of two property axes. Here
we discuss the energy versus entropy plane which illustrates many of the basic
concepts of thermodynamics.
For given values of the amount of the constituent and the volume, we project
the multi-dimensional state space of the system onto the E versus 8 plane. This
projection includes both stable equilibrium states and other states that are not
stable equilibrium. The laws of thermodynamics imply that the projection must
have the shape of the cross-hatched area shown in Figure 1, that is, all the states
that share the given characteristics have property values that project on the area
between the vertical line denoted as the line of zero-entropy states, and the curve
of the stable equilibrium states.
Each point either inside the cross-hatched area or on the vertical line 8 g =
o represents a large number of states. Each such state has the same values of
amount of the constituent, volume, energy and entropy, but differing values of
other properties, and is not a stable equilibrium state. It can be any state except
a stable equilibrium state.
A point on the convex curve of stable equilibrium states represents one and
only one state. For each of these states, the value of any property is uniquely
determined by the values of the amount of the constituent, the volume, and either
E or 8 of the point on the curve.
 15

The zero-entropy states correspond to states contemplated in mechanics. All

other states belong to the broad class of states implied and required by the second
law or, more specifically, the existence of entropy as a property. So mechanics can
be regarded as a special case of thermodynamics - zero-entropy physics.
The stable-equilibrium-state curve represents relations S = S(E, n, V) and
E = E(S, n, V). Each stable equilibrium state has either the lowest energy for
a given entropy, or the highest entropy for a given energy. So, stable equilibrium
states can be regarded as either lowest energy physics for a given entropy, or
highest entropy physics for a given energy.
Because each stable equilibrium state is unique, the temperature T =
(8Ej8S)n,v at each point on the curve. Temperature is not defined for states
that are not stable equilibrium because then E, S, n, V are independent and,
therefore, the partial derivative of E with respect to S is meaningless.
Starting from a stable equilibrium state Ao, the system cannot transfer energy
to affect a solely mechanical effect on the environment without net changes in n
and V because no state of lower energy exists that has an entropy equal to or
greater than the entropy of Ao. This is a graphical illustration of the impossibility
of PMM2. It is sometimes expressed as the nonexistence of a Maxwellian demon.
The vertical distance of state Al from the stable equilibrium state curve is the
adiabatic availability IT! I. The vertical distance of Al from a tangent to the curve
of slope TR is the available energy
reservoir R at temperature TR.
nr of system A in state Al with respect to

Generation of entropy occurs as an isolated system changes its state sponta-

neously from Al to Ao. All other aspects of thermodynamics can be illustrated on
suitable E versus S, or E versus n, or E versus V graphs.

11.2. THE THIRD LAW OF THERMODYNAMICS

For the given values n, V, the energy Eg is the lowest energy - ground-state
energy - for which the system can exist. It corresponds to zero entropy and
zero temperature. If we were using quantum-theoretical concepts, we would be
able to show that all ground-energy states have the same temperature Tg , and
that the value of this temperature is equal to zero. Instead, we note that this
important conclusion of the quantum-theoretical treatment cannot be drawn as a
logical conclusion of the statements of the first and second laws, but here must
be presented as an additional fundamental postulate. It is known as the third law
of thermodynamics or the Nernst principle and can be stated as follows. For each
given set of values of the amount of the constituent and the volume of a system,
there exists one stable equilibrium state with zero temperature.

12. Lecture 12
12.1. INTERACTIONS
Interactions between a system and its environment can have a great variety of
effects. Some result in exchange only of energy between the system and its en-
vironment at constant system volume. Other interactions result in exchanges of
energy and entropy, but neither the amount of the constituent nor the value of
the volume are affected. Still others may result in exchanges of energy, entropy,
and the constituent, as well as in changes in values of the volume. Such exchanges
16

result in changing the states of the system and its environment. In addition, if as
a result of interactions a system is brought to a state that is not stable equilib-
rium, this state may evolve spontaneously toward a stable equilibrium state, thus
causing further changes in the values of the properties of the system.
Knowing how much of the change of a property is due to exchanges with
other systems and how much to spontaneous generation or destruction within
the system is very important to understanding the performance of the system.
For example, if the entropy of a system increases solely because of spontaneous
generation - irreversibility -- such an increase implies imperfections and is subject
to improvement by redesign.
An ingenious method for identifying the amount of entropy generation by ir-
reversibility is by specifying each interaction in terms of the net exchanges that it
causes at the boundary between the system and its environment, and comparing
these exchanges to the changes of properties of the system. For this reason we clas-
sify interactions into different categories, such as exchanges involving energy and
the constituent but no entropy, and energy and entropy but not the constituent.
Two types of interactions are work and heat. An interaction that results in a net
exchange of energy between a system and its environment but no exchanges of
entropy and the constituent is called a work interaction. The amount of energy
exchanged is called work, and denoted by W-+. Work W-+ is positive if energy
flows out of the system. An interaction that results in net exchanges of energy and
entropy but no constituent between the system and an environment at fixed tem-
perature TQ, that is, a reservoir, and such that the ratio of the amount of energy
exchanged and the amount of entropy is equal to T Q , is called a heat interaction.
The energy exchanged with the system is called heat, and denoted by Q+-. Heat
Q+- is positive if energy flows into the system.

12.2. PRACTICAL USE OF BALANCES

Interactions can be combined with the various balances to provide relations be-
tween changes of properties of a system and property flows at the boundary of the
system. For example, if as a result of heat and work interactions system A changes
its state from A 1 to A 2 , the energy, entropy and mass balances are
(E2 - E 1 ) system A E A+- = QA+- _ W A -+ (14)

(15)

(16)
It is noteworthy that here the energy balance is just a very special consequence
of the two laws of thermodynamics and, hence, cannot be called "the first law."
Similarly, the entropy balance is just a very special consequence of the two laws
of thermodynamics and, hence, cannot be called "the second law."
Work and heat are ingenious concepts. For given end states of a system, they
allow the quantitative distinction between entropy generated by irreversibility and
entropy exchanged via interactions with other systems. As such, these two con-
cepts provide practical means for identifying opportunities to reduce the entropy
generation by irreversibility and, hence, improve the performance of the system.
 17

The identification of these opportunities would be missed if heat were defined as

just any interaction that is not work, i.e., any nonwork interaction. The usefulness
of balances 12 to 14 is increased if the initial and final states - end-states -
of the system are assumed to be stable equilibrium states. Then, the differences
(E2 - El )system and (82 - 8 dsystem can be expressed in terms of temperatures and
pressures, and many practical results can be computed.

13. Concluding ReIllarks

In our experience with undergraduate teaching in an engineering school, we man-
age to develop the preceding sequence in the first 12 hours of the course. After
these lectures, we proceed with applications of the state principle and of energy,
entropy, and mass balances to heat engines, simple systems, Euler and Maxwell
relations, the phase rule, two-phase states, bulk-flow interactions, energy conver-
sion devices, and other standard topics of thermodynamics. We find that exposing
the undergraduate students to all the precise definitions and also to some of the
key proofs, instead of just saying "it can be proved" on all occasions, has two
advantages. First, the students gain a precise idea of the nature of the logic in-
volved in thermodynamics and the explicit role of its laws. Second, the students
build up much more confidence in all the results if they have seen the precise and
rigorous logic that is behind at least for some of them. This is what is missing
from traditional expositions.

References
1. Keenan, J.H. and Shapiro, A.H. (1947) History and exposition of the laws of ther-
modynamics, Mech. Eng., 69:915-921.
2. Bejan, A. (1988) Advanced Engineering Thermodynamics, John Wiley & Sons, New
York, pp. 89-96.
3. Feynman, R. (1963) Lectures on Physics, Addison-Wesley, Vol. 1.
4. Tisza, L. (1966) Generalized Thermodynamics, MIT Press, p. 16.
5. Brush, S.G. (1976) The Kind of Motion We Call Heat, North Holland, Vols. 1 & 2.
6. Landau, L.D. and Lifshitz, E.M. (1980) Statistical Physics, Pergamon Press, p. 45.
7. Guggenheim, E.A. (1967) Thermodynamics, 7th Ed., North Holland, p. 10.
8. Keenan, J.H. (1941) Thermodynamics, John Wiley & Sons, p. 6.
9. Hatsopoulos, G.N. and Keenan, J.H. (1965) Principles of General Thermodynamics,
John Wiley & Sons, p. xxiii.
10. Gyftopoulos, E.P. and Beretta, J.P. (1991) Thermodynamics: Foundations and Ap-
plications, Macmillan.
11. Timoshenko, S. and Young, D.H. (1948) Advanced Dynamics, McGraw-Hill, p. 1
and p. 106.
12. Hatsopoulos, G.N. and Keenan, J.H., op. cit., p. 367.
13. Carnot, S. (1960) Reflections on the Motive Power of Fire, in E. Mendoza (ed.),
Dover Publications.
PICTORIAL VISUALIZATION OF THE ENTROPY OF
THERMODYNAMICS

E.P. GYFTOPOULOS
Massachusetts Institute of Technology
Department of Nuclear Engineering, Room 24-111
Cambridge, MA 02139, USA

1. Introductory Remarks

The term entropy was coined by Clausius [I]. He said: "I have felt it more suitable
to take the names of important scientific quantities from the ancient languages
in order that they may appear unchanged in all contemporary languages. Hence
I propose that we call S the entropy of the body after the Greek word rpo7r7} ,
meaning transformation. I have intentionally formed the word entropy to be as
similar as possible to the word energy since the two quantities that are given these
names are so closely related in their physical significance that a certain likeness in
their names has seemed appropriate."
In the scientific literature, Clausius' perceptive and important observation that
the energy and entropy of thermodynamics are closely related has not always been
heeded. Since its inception, the concept of energy has been accorded a universal
meaning that underlies our understanding of all physical phenomena. In contrast,
the term entropy has been associated with many statistical concepts and mathe-
matical expressions. Because these concepts and expressions differ from each other,
the common name is a source of confusion [2,3] and has confounded understand-
ing of many issues not only about the relations among various entropies but, more
importantly, about both the physical meaning of the entropy of thermodynamics
and the idea of transformation.
In addition to the issues raised by the association of the term entropy with
different concepts, many scientists and engineers have expressed concerns about
the completeness and clarity of the usual expositions of thermodynamics. For ex-
ample, Obert [4] writes: "Most teachers will agree that the subject of engineering
thermodynamics is confusing to the student despite the simplicity of the usual
undergraduate presentation."
Intrigued and challenged by the issues and concerns just cited, over the past
few decades we have made an attempt to clarify both the theoretical foundations
of thermodynamics and the experimental evidence in support of these foundations,
and have reached conclusions that differ from the interpretations presented in the
scientific literature. We have discussed the new viewpoint first as a unified quantum
theory of mechanics and thermodynamics in Refs. [5-10], and then as a nonquantal
exposition of the foundations and applications of thermodynamics in Ref. [11].
19
A. Bejan and E. Mumut (eds.J, Thermodynamic Optimization o/Complex Energy Systems, 19-44.
© 1999 Kluwer Academic Publishers.
 20

In this paper, we recapitulate our findings about both the thermodynamic

definition of entropy, and its quantum-theoretic expression. Then, we provide a
pictorial visualization of entropy in terms of the quantum-theoretic probability
density function in ordinary space, that is, in terms of the shape of the constituents
of a system. Moreover, we show that the generation of entropy (irreversibility) is
due to spontaneous changes or transformations of the shape of constituents as they
try to conform to the external and internal forces of the system. The spontaneous
changes can occur until the entropy assumes its largest value. At that condition,
the spatial shape of the constituents yields zero values in any direction of both
the translational momentum and the internal structure angular momentum of each
atom or molecule. In other words, in thermodynamic equilibrium, all motions cease
to exist - everything is at rest.
Many scientists and engineers interested in applications may feel that a gen-
eral, rigorous, noncircular, and complete description of thermodynamics, and a
quantum-theoretic explanation of the foundations of the subject are not impor-
tant in solving practical problems. This view is only partially correct, and if in-
sisted upon may deprive us of opportunities to make new discoveries and major
practical contributions. Some of the reasons for the preceding comment are as
follows: (i) The entropy of thermodynamics is a cornerstone of the profession of
mechanical engineers as well as of all other engineering and scientific disciplines.
For example, if it were not for the possible spontaneous increase of entropy, hu-
manity would have no problems of exhaustion of natural resources of both energy
and raw materials, no problems of pollution, and no need for so many profession-
als preoccupying themselves with such problems; (ii) The popular meaning of the
word energy does not represent the concept of energy of thermodynamics but the
concept of available energy or exergy, and available energy is decisively controlled
by entropy rather than the energy defined in thermodynamics. So, understand-
ing entropy is of paramount importance to all concerned about energy systems
and processes; (iii) In our works we have expressed the laws of thermodynam-
ics in terms of the concepts of space, time, and force or inertial mass only, and
have proven many theorems of these laws without inconsistencies, incompleteness,
circular arguments, and ambiguities. As a result, we have discovered that thermo-
dynamics and quantum theory are facets of a new unified paradigm of physics, and
that conventional quantum mechanics (the quantum mechanics taught to under-
graduate students) is a special case of the unified theory (zero entropy physics);
(iv) The rigorous exposition of thermodynamics helped us to discover that en-
tropy is a nonstatistical property of the constituents of any system in any state.
Moreover, the unified quantum theory of mechanics and thermodynamics helped
us to determine the quantum- theoretic analytical expression for this important
property; (v) As a result of the quantum-theoretic considerations, we have ob-
tained both a pictorial visualization of the mathematical expression of entropy as
a measure of the shape of the constituents of a system, and a correlation between
spontaneous shape changes and entropy generation (irreversibility). Hopefully, the
visualization and the correlation will stimulate interest in better understanding
of the foundations of thermodynamics both without and with quantum theory;
and (vi) Educators, practitioners, and students have a scientific and professional
responsibility to pay more attention to the rigorous, noncircular, and non confusing
understanding and practice of thermodynamics of all systems and all states. Such
understanding and practice are lacking now. A wider and more frequent apprecia-
tion of both the beauty, generality, and power of thermodynamics, and the correct
 21

use of its fundamental principles is not a matter of presenting a new theory. It is an

imperative that will help us address and solve many problems of great importance
to humanity and its planet Earth.

2. The Entropy of Thermodynamics

Among the many novel results of the nonquantal exposition [11] is the recognition
that the entropy of thermodynamics is a nonstatistical property of a system, in
the same sense that inertial mass, or energy, or momentum is a nonstatistical
property of a system, and that this entropy is well-defined for all systems, both
macroscopic and microscopic, and all states, both thermodynamic equilibrium and
not thermodynamic equilibrium.
Because the nonquantal exposition is not widely known yet, a brief summary of
the laws of thermodynamics and some of their theorems is presented in Appendix
A, including the definition of entropy by the relation

(1)

where Si and Ei are the entropy and energy of state A of system A, respectively,
nf' is the generalized available energy of state Ai with respect to reservoir R, and
CR. a well-defined positive constant.
Based on the characteristic features of Ei and nf' listed in Appendix A, we
find that the entropy of thermodynamics must conform to the following criteria:
(i) be well-defined for every system (both macroscopic and microscopic), and every
state (both stable equilibrium and not stable equilibrium); (ii) be independent of
the reservoir R; (iii) be invariant in all reversible adiabatic processes, and increas-
ing in all irreversible adiabatic processes; (iv) be additive; (v) be nonnegative,
and vanish for all the states encountered in mechanics; (vi) for given values of
energy, amounts of constituents, and (external) parameters, admit one and only
one state that corresponds to the largest value of entropy; (vii) for given values
of amounts of constituents, and parameters, the graph of entropy versus energy
of stable equilibrium states be concave and smooth; (viii) for a composite 0 of
two subsystems A and B, must be such that the constrained maximization proce-
dure for 0 (criterion vi) yields identical thermodynamic potentials (for example,
temperature, total potentials, and pressure) for all three systems A, B, and 0;
(ix) for stable equilibrium states, must reduce to relations that have been estab-
lished experimentally and that express the entropy in terms of energy, amounts of
constituents, and parameters, such as the relations for ideal gases.
In due course, we prove that the preceding reasonable and relatively simple cri-
teria reject all but one of the analytical expressions for entropy that have appeared
in the scientific literature. In fact, even the one and only acceptable expression re-
quires a profound and radical reinterpretation of its independent variables.

3. Unified Quantum Theory of Mechanics and Thermodynamics

In many textbooks on quantum mechanics, probabilities associated with measure-
ment results are derived from a normalized wave function V(x) or, equivalently,
from either a Dirac ket ''I'} or a projector 'V}(V', where x are the spatial coordi-
nates of the constituents of a system, and ('II' is the bra of ket ''I') [12]. For example,
22

for a system with one degree of spatial freedom x, the probability density function
of measurement results of x is given by /V(X)/2 = (V/x)(x/V) = /(X/V)/2.
In statistical quantum mechanics, probabilities associated with measurement
results are derived from a density operator P = Ei ex;Pi which represents a mixture
of quantal probabilities derived from projectors Pi = /Vi)(Vi/ for i = 1,2, ... , and
statistical (informational) probabilities ex; which reflect the lack of information
about some or all aspects of the state of a system.
In general, the foundations and theorems of the unified quantum theory of
mechanics and thermodynamics differ from those of the ordinary expositions of
quantum mechanics and statistical quantum mechanics. Among the many novel
concepts and results, two deserve special emphasis at this point: (i) In contrast
to statistical quantum mechanics, a novel concept of the unified quantum theory
is that its postulates - quantum-theoretic and thermodynamic - do not apply
to density operators that represent mixtures of quantal probabilities derived from
projectors Pi = /vi)(vd, and statistical (informational) probabilities ex; that reflect
the lack of information about some or all aspects of the state of a system. Instead,
the unified theory avers that the laws of physics apply only to density operators
each of which is construed as the seat of exclusively quantal probabilities, that
is, only to a P that can be represented by a homogeneous ensemble of identical
systems, identically prepared. Homogeneous is an ensemble in which the proba-
bilities of results of measurements on any member are represented by the same
density operator P as those on any other member. Accordingly, experimentally
(in contrast to algebraically) the ensemble cannot be decomposed into statistical
mixtures of projectors or other non-projector density operators. The concept of
homogeneous ensemble was introduced by von Neumann [13]. But he assumed
that it applies only to projectors (Pi = pf), whereas here the concept is extended
to all self-adjoint, nonnegative definite, linear, unit trace density operators. Each
such operator satisfies the relation P 2: p2 j and (ii) As already indicated, in con-
trast to the plethora of expressions for entropy that have been proposed in the
scientific literature over the past fourteen decades, we find that only one satisfies
the criteria listed in the preceding section. Because the unified quantum theory
of mechanics and thermodynamics is not widely known yet, a brief summary of
both the quantum-theoretic postulates and some of their theorems is presented in
Appendix B, including a discussion of the fact that, of the known expressions for
entropy S, the only acceptable is

S = -kTr[plnp] (2)

where k is Boltzmann's constant, Tr stands for the trace of the operator that fol-
lows, and p is a density operator which can be represented solely by a homogeneous
ensemble. It is noteworthy that, if p cannot be represented by a homogeneous en-
semble, then -k Tr [p In p] does not represent the entropy of thermodynamics.
For given values of energy, parameters, and amounts of constituents, if p is a
projector (wave function) then S = 0, if P is not a projector but corresponds to
a state which is not stable equilibrium (not thermodynamic equilibrium) then S
has a positive value smaller than the largest possible for the given specifications,
and if p corresponds to the unique stable equilibrium state, then S has the largest
value of all the entropies of the system which share the given values of energy,
parameters, and amounts of constituents.
 23

4. Pictorial Visualization of Entropy

In many textbooks, the probability density function associated with measurement
results of the spatial coordinates of the constituents of a system is interpreted as
the spatial shape of the constituents of the system [14,15], and the shape is used
to calculate atomic, molecular, and ionic radii [16], and to explain the formation
of various compounds. The probability density function of the spatial coordinates
enters also in the evaluation of the entropy S (equation (2)) and, therefore, we can
think of entropy as a measure of the spatial shape of the constituents of a system,
and thus achieve a pictorial visualization of a concept that has been puzzling
scientists and engineers over more than a century.
A few graphs of surfaces of constant probability density are shown in Figure 1.
They are derived from probability density functions which represent energy eigen-
functions (energy eigenprojectors) of the electron in a hydrogen atom. They have
been calculated by Brandt and Dahmen [15]. In the notation Pnl m of each shape,
P denotes the probability density function as a function of the Cartesian coordi-
nates x, y, z of the electron, subscript n denotes the energy quantum number and
appears in the energy eigenvalue relation

En = -13.61/n 2 (eV) , (3)

subscript £ denotes the angular momentum quantum number (0:::; £ :::; n - 1) and
appears in the square of the total angular momentum eigenvalue relation

(4)

L2 is the square of the total angular momentum, and 11, Planck's constant divided by
21r, and subscript m denotes the z-component of the angular momentum quantum
number (-£ :::; m :::; £) and appears in the eigenvalue relation

(5)

where Lz is the z-component of the angular momentum. The probability density

function equals the square of the absolute magnitude of the electron energy eigen-
function as a function of x, y, z.
If the normalized wave function '1'( x, y, z) of the electron is not an energy
eigenfunction, and has energy Eelectron = ('I'I Helectron 1'1') , then the probability
distribution function in x, y, z-space is again given by 1'1'1 2 , and a graph of the
electron shape can be made in a manner analogous to that described for energy
eigenfunctions.
Regardless of whether a wave function is an eigenfunction of a Hamiltonian
operator H, or another operator that represents an observable, or not an eigen-
function of any operator, the probability density function can be visualized as a
shape in Cartesian space, and the value of the entropy S of any such shape equals
zero.
For a nonzero value of S, we must consider a density operator which does not
correspond to a projector 1'1') ('1'& or wave function '1'. The choices are infinite. For
our purposes, we have selected Pelectron which corresponds to the stable equilibrium
state of the electron of a hydrogen atom, at electron energy Eelectron [17]. Because
the energy eigenstates of the electron in a hydrogen atom are degenerate, the
 24

~311 ~

Figure 1. Surfaces of constant probability density P3l m = 0.0002 in full x, y, z-space of

the electron in a hydrogen atom [15].

probability density function P~lectron of the electron in a stable equilibrium state

of energy Eelectron is given by the relation (see Appendix B)

o
Pelectron = '~
" 0 Pn
Pnn (6)
n

where

(7)
 25

Pn (8)

gn = n 2 and is the degeneracy of the energy eigenvalue en, ~ is determined by the

energy by means of the relation

Eelectron = L
n
P~n en (9)

and Pnlm is the probability density of the energy eigenstate with quantum numbers
n, £, m (0 :::; £:::; n - 1 and -£:::; m :::; i). As shown in Appendix B, ~ = l/kT, where
T is the temperature of the electron. Because ~ is determined by Eelectron, it follows
that T is also determined solely by Eelectron.
Using graphs such as in Figure 1, we find

(10)

1
P4 = 16 (P400 + P410 + P420 + P430)
2
+ 16 (P411 + P421 + P422 + P431 + P432 + P433) (11)

For the purposes of this paper, Dahmen and Stroh [18] graphed P3' P4 and
Pn for any value of n. As any quantum-thermodynamicist might have expected,
Pn is spherically symmetric for n = 1,2, ... (Figure 2) and, therefore, the prob-
ability density function P~lectron or the corresponding electron shape around the
proton is spherically symmetric. The spherical shape has at least two important
implications. First, the angular momentum of the electron in any direction equals
zero. Zero angular momentum means that the electron is not moving around the
proton nucleus of the hydrogen atom, as the term stable equilibrium suggests and
electromagnetic theory requires for equilibrium to be achieved. It is noteworthy
that there exists an infinite number of spherical shapes of the electron around the
proton which correspond to the same value Eelectron of the energy but an upper
limit ofn in equation (6) smaller than infinity. Each such shape is an unstable equi-
librium probability distribution function, and upon a minute perturbation tends
spontaneously to the shape P~lectron.
The second important implication is that the value of the entropy that corre-
sponds to the shape P~lectron is larger than that determined by any other density
operator P -::j:. P~lectron and such that 'If P Helectron = 'If P~lectron Helectron (see,
however, Ref. 17).
If the change of shape from P to P~lectron occurs, the change from a low value
S(p) to the largest value S(P~lectron) is spontaneous, and the process is irreversible.
There are, however, spontaneous shape changes such as either from one wave
function to another, or from PI to P2 -::j:. PI but at constant values of energy
(Tr PI Helectron = 'If P2 Helectron), amounts of constituents, and parameters, for
which S(P2) = S(PI) and, then, the process is reversible.
Completion of the discussion of a hydrogen atom in a finite-size box requires
also the consideration of the shapes arising from the translational degrees of free-
dom of the proton nucleus. The energy eigenvalues, eigenfunctions, and shapes
 26

Figure 2. Surface of constant probability density Pn in full x, y, z-space of the electron

in a hydrogen atom, arbitrary units for any value of n [18].

of a structureless proton in a box, and the procedure for combining the trans-
lational shapes with the electron shapes are discussed in many texts, including
Hatsopoulos and Gyftopoulos [19]. For the sake of brevity, we will not repeat this
discussion here except for only one comment. If the probability density function
of the proton is derived from an energy eigenfunction, the value of the speed of
the proton in any direction is zero (see Appendix B). So, in a stable equilibrium
state, both the electron velocities and the proton velocities are all zero - nothing
moves, that is, the concept of stable equilibrium in the unified theory is the same
as the concept of stable equilibrium in classical mechanics. However, there exists
an important difference between the two theories. Whereas in classical mechanics
the stable equilibrium state of a system corresponds to the lowest energy of the
system, in the unified theory there exists one stable equilibrium state for each set
of values of energy, amounts of constituents, and parameters (see Appendix A).

5. Concluding Relllarks

In closing, we wish to reiterate that the entropy of thermodynamics is: (i) a nonsta-
tistical property of the constituents of a system, in the same sense that inertial mass
is such a property; (ii) valid for any system, both macroscopic and microscopic,
 27

including a system with only one particle; (iii) valid for any state, thermodynamic
equilibrium or not; and (iv) a measure of the geometric shape of the constituents of
the system, even if the system consists of only one particle. Moreover, irreversibil-
ity is due to the spontaneous change of the shape of the constituents as they try
to conform to the external and internal forces of the system.

Acknowledgements
I am very thankful to Professors Doctors S. Brandt and H.D. Dahmen, and to
Mr. T. Stroh for Figures 1 and 2.

Appendix A. Thermodynamics
A.I. GENERAL REMARKS
Gyftopoulos and Beretta [11) have composed a novel exposition in which all basic
concepts of thermodynamics are defined completely and without circular argu-
ments in terms of the mechanical concepts of space, time, and force or inertial
mass. Many of these definitions are new. The order of introduction of concepts
and postulates is: system (types and amounts of constituents, forces between con-
stituents, and external forces or parameters); properties; states; the first law; en-
ergy (without work and heat); energy balance; classification of states in terms
of time evolutions; stable equilibrium states; second law (without temperature,
heat, and entropy); generalized available energy; entropy of any state, stable equi-
librium or not, in terms of energy and generalized available energy and not in
terms of temperature and heat; entropy balance; fundamental relation for stable
equilibrium states only; temperature, total potentials, and pressure in terms of
energy, entropy, amounts of constituents and parameters for stable equilibrium
states only; the third law; work in terms of energy; and heat in terms of energy,
entropy, and temperature. All concepts and postulates are valid for all systems
(both macroscopic and microscopic), all states (both stable equilibrium and not
stable equilibrium), and require no statistical probabilities.

A.I.I. Definition
We define general thermodynamics or simply thermodynamics as the study of
motions of physical constituents (particles and radiations) resulting from exter-
nally applied forces, and from internal forces (the actions and reactions between
constituents). This definition is identical to that given by Timoshenko and Young
about mechanical dynamics [20). However, because of the second law, the definition
encompasses a much broader spectrum of phenomena than mechanical dynamics.

A.I.2. Kinematics: Conditions at an Instant in Time

In kinematics we give verbal definitions of the terms system, property, and state so
that each definition is valid without change in any physical theory, and involves no
statistics attributable to lack of information. The definitions include innovations.
To the best of our knowledge, they contradict neither any theoretical principle nor
any experimental result.
A system is defined as a collection of constituents confined by a nest of internal
intermolecular forces and external forces or parameters. Without modification, this
28

definition applies to all paradigms of physics. The term paradigm is used in the
sense of Kuhn [21].
Everything that is not included in the system is the environment.
For a system with r constituents, we denote their amounts by the vector n =
{n1 ,n2, ... ,nr}. For a system with external forces described by s parameters we
denote the parameters by the vector f3 = {j31, (32, ... , (38}. One parameter may be
volume, (31 = V.
At any instant in time, the amount of each constituent, and the parameters of
each external force have specific values. We denote these values by nand f3 with
or without additional subscripts.
By themselves, the values of the amounts of constituents and of the parameters
at an instant in time do not suffice to characterize completely the condition of the
system at that time. We also need the values of a set of independent properties
at the same instant in time. The value of an independent property can be varied
without affecting the values of other properties. Each property is an attribute
that can be evaluated at any given instant in time (not as an average over time)
by means of a set of measurements and operations that are performed on the
system and result in a numerical value - the value of the property. This value
is independent of the measuring devices, other systems in the environment, and
other instants in time.
For a given system, the values of the amounts of the constituents, the values
of the parameters, and the values of a complete set of independent properties
encompass all that can be said about both the system at an instant in time, and
the results of any measurements that may be performed on the system at that
same instant in time. We call this complete characterization of the system at an
instant in time the state of the system. This definition of state is novel and, without
change, applies to any branch and any paradigm of physics.

A.2. DYNAMICS: CHANGES OF STATE IN TIME

The state of a system may change in time either spontaneously due to the internal
and external forces, or as a result of interactions with other systems, or both. The
relation that describes the evolution of the state of an isolated system - sponta-
neous changes of state - as a function of time is the equation of motion. Certain
time evolutions obey Newton's equation which relates the total force F on each
system particle to its inertial mass m and acceleration a so that F = mao Other
evolutions obey the time-dependent Schroedinger equation, that is, the quantum-
mechanical equivalent of Newton's equation. Other experimentally observed time
evolutions, however, do not obey either of these two equations. So the equations
of motion that we have are incomplete. The discovery of the complete equation
of motion that describes all physical phenomena remains a subject of research at
the frontier of science - one of the most intriguing and challenging problems in
physics [2, 5, 8-10,22-24]. Many features of the equation of motion have already
been discovered. These features provide not only guidance for the discovery of the
complete equation but also a powerful alternative procedure for analyses of many
time-dependent, practical problems. Two of the most general and well-established
features are captured by the consequences of the first and second laws of thermo-
dynamics discussed later.
 29

A.3. ENERGY AND ENERGY BALANCE

Energy is a concept that underlies our understanding of all physical phenomena,
yet its meaning is subtle and difficult to grasp. It emerges from a fundamental
principle known as the first law of thermodynamics. The first law asserts that any
two states of a system may always be the initial and final states of a change (weight
process) that involves no net effects external to the system except the change in
elevation between Zl and Z2 of a weight, that is, a mechanical effect. Moreover, for
a given weight, the value of the expression M g(Zl - Z2) is fixed only by the end
states of the system, where M is the mass of the weight, and 9 the gravitational
acceleration.
The main consequence of this law is that every system A in any state Al has a
property called energy, with a value denoted by the symbol EI (Ref. 11, Sec. 3.4,
pp. 32-33). The energy EI can be evaluated by means of a weight process that
connects Al and a reference state Ao to which is assigned an arbitrary reference
value Eo so that
El = Eo - Mg(Zl - zo) (A-I)
Energy is shown to be an additive property (Ref. 11, Sec. 3.6, pp. 34-35), that
is, the energy of a composite system is the sum of the energies of its subsystems.
Moreover, it is also shown that energy has the same value at the final time as at the
initial time if the system experiences a zero-net-effect weight process, or remains
invariant in time if the process is spontaneous (Ref. 11, Sec. 3.7, pp. 35-37). In
either ofthe last two processes, Z2 = Zl and E(t2) = E(tl) for time t2 greater than
tl, that is, energy is conserved. Energy conservation is a time-dependent result. In
Ref. 11, this result is obtained without use of the general equation of motion.
Energy can be exchanged between systems by means of interactions. Denoting
by EA+- the amount of energy exchanged between the environment and system
A in a process that changes the state of A from Al to A 2 , we can derive the
energy balance. This derivation is based on the additivity of energy and energy
conservation (Ref. 11, Sec. 3.8, pp. 37-38), and reads

(A-2)

In words, the energy change of a system must be accounted for by the energy
transferred across the boundary of the system. The energy EA+- crossing the
boundary of A is positive if energy flows from the environment to system A.

A.4. TYPES OF STATES

Because the number of independent properties of a system is infinite even for a sys-
tem consisting of a single particle with a single translational degree of freedom -
a single variable that fixes the configuration of the system in space - and because
most properties can vary over a range of values, the number of possible states of a
system is infinite. The discussion of these states is facilitated if they are classified
into different categories according to time evolutions. This classification brings
forth many important aspects of physics, and provides a readily understandable
motivation for the introduction of the second law of thermodynamics.
The classification consists of: unsteady states; steady states; nonequilibrium
states; and equilibrium states (Ref. 11, Sec. 4.1, pp. 53-58). An equilibrium state
is one that does not change as a function of time while the system is isolated -
30

a state that does not change spontaneously. An unstable equilibrium state is an

equilibrium state that may be caused to proceed spontaneously to a sequence of
entirely different states by means of a minute and short-lived interaction that has
only an infinitesimal effect on the state of the environment. A stable equilibrium
state is an equilibrium state that can be altered to a different state only by interac-
tions that leave net effects in the environment of the system. These definitions are
identical to the corresponding definitions in mechanics but include a much broader
spectrum of states than those encountered in mechanics. The broader spectrum is
due to the second law discussed later.
Starting either from a non equilibrium state or from an equilibrium state that
is not stable, energy can be extracted from a system and affect a mechanical effect
without leaving any other net changes in the state of the environment. In contrast,
experience shows that, starting from a stable equilibrium state, no energy can be
extracted from the system that would result in the mechanical effect just cited.
This impossibility is one of the most striking consequences of the first and second
laws of thermodynamics. It is consistent with innumerable experiences.

A.5. GENERALIZED AVAILABLE ENERGY

The existence of stable equilibrium states is not self-evident. It was first recognized
by Hatsopoulos and Keenan [25] as the essence of all correct statements of the
second law. Gyftopoulos and Beretta (Ref. 11, Ch. 4, pp. 53-66) concur with
this recognition, and state the second law as follows (simplified version): Among
all the states of a system with a given value of energy, and given values of the
amounts of constituents and the parameters, there exists one and only one stable
equilibrium state. The second law cannot be derived from the laws of mechanics.
Within mechanics, the stability analysis yields that among all the allowed states
of a system with fixed values of amounts of constituents and parameters, the only
stable equilibrium state is that of lowest energy. In contrast the second law avers
the existence of a stable equilibrium state for each value of the energy.
The existence of stable equilibrium states for various conditions of a system
has many theoretical and practical consequences. One consequence is that, starting
from a stable equilibrium state of any system, no energy is available to affect a
mechanical effect while the values of both the amounts of constituents and the
parameters of the system experience no net changes (Ref. 11, Sec. 4.5, pp. 64-65).
This consequence is often referred to as the impossibility of a perpetual motion
machine of the second kind (PMM2). In some expositions of thermodynamics, it
is taken as the statement of the second law. In the new exposition, it is only one
aspect of both the first and the second laws.
At fixed values of both the amounts of constituents and the parameters, an-
other consequence is that not all states of a system can be changed to the state of
minimum energy by means of solely a mechanical effect. This is a generalization
of the impossibility of a PMM2, and is represented by a property called adiabatic
availability, and denoted by tJI (Ref. 11, Secs. 5.2 to 5.4, pp. 73-76). The concept
of adiabatic availability can be generalized to the concept of generalized adiabatic
availability (Ref. 11, Secs. 5.5 and 5.6, pp. 77-80). The latter represents the opti-
mum amount of energy that can be exchanged between a system and a weight in
a weight process in which the respective initial and final values of the amounts of
constituents and/or the parameters differ. Generalized adiabatic availability dif-
 31

fers from energy. Like energy, this property is well defined for any system in any
state. Unlike energy, it is not additive (Ref. 11, Sec. 5.3.8, p. 75).
In striving to define an additive property that captures the important features
of generalized adiabatic availability, Gyftopoulos and Beretta introduce a special
reference system, called a reservoir, and discuss the possible weight processes that
the composite of a system and the reservoir may experience. Thus, they disclose
a third consequence of the first and second laws, that is, a limit on the optimum
amount of energy that can be exchanged between a weight and a composite of a
system and a reservoir R - the optimum mechanical effect. They call the opti-
mum value generalized available energy (Ref. 11, Sec. 6.8, pp. 95-97), denote it by
n R , and show that it is additive (Ref. 11, Sec. 6.9.6, pp. 98-99). It is a generaliza-
tion of the concept of motive power of fire first introduced by Carnot because he
assumed that both systems of the composite acted as reservoirs with fixed values
of their respective amounts of constituents and parameters, whereas Gyftopoulos
and Beretta do not use this assumption. The definition of a reservoir is given in
the article on the foundations of thermodynamics by E. P. Gyftopoulos in this
volume.
If the net exchanges between a system and its environment involve only energy,
the process experienced by the system is called adiabatic. For an adiabatic process
of system A only, it is shown that the energy change El - E2 of A and the
generalized available energy change nf - nf of the composite of A and reservoir
R satisfy the relations (Ref. 11, Sec. 6.9, pp. 97-99):
If the adiabatic process of A is reversible: EI - E2 = nf - nf (A-3)
If the adiabatic process of A is irreversible: El - E2 < nf - nf (A-4)
A process is reversible if both the system and its environment can be restored to
their respective initial states. A process is irreversible if the restoration just cited
is impossible.
It is noteworthy that energy and generalized available energy are defined for
any state of any system, regardless of whether the state is steady, unsteady, equilib-
rium, nonequilibrium, or stable equilibrium, and regardless of whether the system
has many degrees of freedom or one degree of freedom, or whether the size of the
system is large or small.

A.6. ENTROPY AND ENTROPY BALANCE

A system A in any state Al has many properties. Two of these properties are:
energy E 1 , and generalized available energy nf with respect to a given auxiliary
reservoir R. These two properties determine a third property called entropy, and
denoted by the symbol 8. It is a property in the same sense that inertial mass is a
property, or energy is a property, or momentum is a property. For a state AI, 8 1
can be evaluated by means of an auxiliary reservoir R, a reference state Ao, with
energy Eo and generalized available energy nfj , to which is assigned a reference
value 80, and the expression

(A-5)

where CR. is a well-defined positive constant that depends on the auxiliary reservoir
R only. Entropy 8 is shown to be independent of the reservoir (Ref. 11, Sec. 7.4,
32

pp. 108-112), that is, 8 is a property of system A and the reservoir is auxiliary
and is used only because it facilitates the definition of 8. It is also shown that 8
can be assigned absolute values that are nonnegative, and that vanish for all the
states encountered in mechanics (Ref. 11, Sec. 9.8, pp. 137-138).
Because energy and generalized available energy satisfy relations (A-3) and
(A-4), the entropy defined by equation (A-5) remains invariant in any reversible
adiabatic process of A, and increases in any irreversible adiabatic process of A.
These conclusions are valid also for spontaneous processes and for zero-net-effect
interactions. The latter features are known as the principle of nondecrease of en-
tropy. Both a spontaneous process and a zero-net-effect interaction are special
cases of an adiabatic process of system A.
The entropy created during an irreversible process as a state changes in time is
called entropy generated by irreversibility. It is positive. The entropy nondecrease
is a time-dependent result. In the exposition of thermodynamics in Ref. 11, this
result is obtained without use of the general equation of motion. Because both
energy and generalized available energy are additive, equation (A-5) implies that
entropy is also additive (Ref. 11, Sec. 7.2.2, pp. 103-104).
Like energy, entropy can be exchanged between systems by means of interac-
tions. Denoting by 8 M - the amount of entropy exchanged between the environ-
ment and system A in the course of a process that changes the state of A from Al
to A 2 , we derive a very important analytical tool, the entropy balance (Ref. 11,
Sec. 7.3, pp. 106-108), that is,

(A-6)

where 8irr is positive or at least zero and represents the entropy generated sponta-
neously within system A during the time interval from tl to t2 required to affect
the change from state Al to state A 2 . Spontaneous entropy generation within a
system occurs if the system is in a state that is not stable and the system forces
precipitate the natural tendency towards stable equilibrium. The entropy 8 A t-
crossing the boundary of A is positive if entropy flows from the environment to
system A. The dimensions of 8 depend on the dimensions of both energy and CR.
It turns out that the dimensions of CR are independent of mechanical dimensions,
and are the same as those of temperature. Temperature is defined later.

A.7. STABLE EQUILIBRIUM STATES

It is shown that among the many states of a system that have given values of the
energy E, the amounts of constituents n, and the parameters {3, the entropy of the
unique stable equilibrium state that corresponds to these values is larger than that
of any other state with the same values E,n,{3 (Ref. 11, Sec. 8.2, pp. 119-120).
Moreover, the entropy of a stable equilibrium state must be a function solely of
E, n, (3, that is,
8 = 8(E, n, (3) (A-7)

Equation (A-7) is called the fundamental relation (Ref. 11, Sec. 8.3, pp. 120-
124). For states that are not stable equilibrium, 8 depends on more variables than
E, n, {3, that is, equation (A-7) is not valid.
 33

The fundamental relation is concave with respect to energy (Ref. 11, Sec. 9.4,
pp. 131-132), that is,
8S)
(8E2 2
:S 0 (A-B)
n,/3
and analytic in each of its variables E, n, f3 (Ref. 11, Sec. B.3, pp. 120-124).
Moreover, the fundamental relation is used to define other properties of stable
equilibrium states, such as temperature T (Ref. 11, Ch. 9),

~ = (:!) n,/3
(A-9)

total potentials fl-i for i = 1,2, ... , r (Ref. 11, Ch. 10, pp. 147-151)

fl-i=-T ( 8S)
-
8ni E,n,/3
for i = 1,2, ... , r (A-lO)

and pressure p (Ref. 11, Ch. 11, pp. 157-162)

p -
-
T(8S)
8V E,n,/3
for /31 = V= volume. (A-11)

Detailed discussion of properties of stable equilibrium states are given in Ref. 11,
Chapters B-11.

A.B. COMMENT
The concept of entropy introduced here differs from and is more general than that
in all textbooks except Ref. 11. It does not involve the concepts of temperature
and heat; it is not restricted to large systems; it applies to both macroscopic
and microscopic systems, including a system with one spin, or a system with one
particle with only one (translational) degree of freedom; it is not restricted to stable
(thermodynamic) equilibrium states; it is defined for both stable equilibrium and
not stable equilibrium states because energy and generalized available energy are
defined for all states; and most certainly, it is not statistical - it is a property of
matter and not a measure of the lack of information of an observer.

A.9. INTERACTIONS
Work, heat, bulk flow, and diffusion interactions, and their use in the balance
equations are discussed in Refs. [11] and [26] and will not be repeated in this brief
review.

Appendix B. Unified Quantum Theory of Mechanics and Thermody-

namics
B.1. STATISTICAL DESCRIPTIONS
Ever since the enunciation of the first and second laws of classical thermodynamics
by Clausius more than 130 years ago, the question of the relation between classi-
cal thermodynamics and mechanics has been the subject of intense investigations
34

and controversy. Invariably, Maxwell's seminal ideas prevail [27], that is, " ... the
molecules in a vessel full of air at uniform temperature are moving with velocities
by no means uniform ... " , and "In dealing with masses of matter, while we do not
perceive the individual molecules, we are compelled to adopt what I have described
as the statistical method of calculation, ... ". Though deeply rooted in classical
mechanics, these ideas playa major role in conventional quantal explanations as
well.
Specifically, in statistical quantum mechanics, the dominant view currently
held about the physical significance of classical thermodynamics is based on the
interpretation of a "thermodynamic equilibrium state" as a composite that best
describes the knowledge of an observer possessing only partial information about
the "actual state" of a macroscopic system. The "ith actual state" at any instant
in time is defined as the set of quantal probabilities derivable from an energy
eigenprojector Pi = Ifi)(fd, or from an energy eigenfunction fi, where Ifi) is an
energy eigenket of conventional quantum mechanics. Because observers are uncer-
tain about the actual state, they consider all possible "actual states" and assign
to each of them a statistical probability (Xi. The assignment is achieved by using a
hypothesis in addition to the laws of conventional quantum mechanics. The com-
bination of the two types of probabilities is a mixture characterized by an overall
density operator P = I:i CX;Pjl where I:i (Xi = 1, and P > p2. The theories that
have evolved pursuant to the view just cited are called informational, though the
general idea is the foundation of each statistical interpretation of thermodynamics
proposed to date.
If at an instant in time, Pi for i = 1,2, ... , and P are represented by ensembles
of identical systems, von Neumann has shown that the ensemble for each Pi is and
must be homogeneous [13], that is, each member of the ensemble is assigned the
same Pi as any other member, whereas the ensemble for P is heterogeneous, that
is, only a fraction (Xi of the members of the P ensemble is assigned the projector Pi
(Figure B-1). In addition to representing the fractions of the Pi'S in p, the statistical
probabilities (Xi enter in the evaluation of an informational measure of uncertainty,
a subjective entropy, such as S = - k I:i (Xi In (Xi , where k is Boltzmann's constant.

B.l.l. Comments
Statistical theories of thermodynamics yield many correct and practical results.
For example, they yield the canonical, grand-canonical, Boltzmann, Bose-Einstein,
and Fermi- Dirac distributions, and predict the equality of temperatures of sys-
tems in mutual stable equilibrium, the Maxwell relations, and the Gibbs equation
[28,29]. Mutual stable equilibrium is defined in Ref. 11, Sec. 6.2, pp. 86-87. (See
also article on the foundations of thermodynamics by E.P. Gyftopoulos in this
volume.) Despite these successes, the premise that entropy is a subjective charac-
teristic of the knowledge of a partially informed observer rather than a property
of a system leaves much to be desired in the light of many accurate, reproducible
and nonstatistical experiences, such as the mixing of hot and cold substances, the
characteristics of an internally discharging electricity storage battery, the Peltier
effect, and chemical reactions. In all these experiences, entropy plays a dominant
and decisive role which is entirely independent of whether an observer is informed
or misinformed. As pointed out by Schroedinger and others [30-32], the concep-
tual foundations of statistical interpretations of thermodynamics are not on solid
ground. For example, they seem to require abandonment of the concept of state
 35

HETEROGENEOUS ENSEMBLE

[EJ lliJ [EJ §] §]

P # PI
[EJ lliJ [EJ §] §]
P # P2
[EJ lliJ @!J §] §]

FRACTION al FRACTION a2
OVERALL DENSITY P = alPI + a2P2
Figure B-1. Representation of a heterogeneous ensemble.

of a system, a cornerstone of traditional physical thought. Again, they foreclose

opportunities for the development of a sound theory of nonequilibrium. The reason
for the foreclosure is that each statistical theory considers either Newton's equa-
tion or Schroedinger's equation as the relation that specifies the evolution of the
"actual state" in time, but faces insurmountable conceptual difficulties to propose
deterministic equations for the evolutions of the ai's in time. In fact, any argument
that considers the <X.;. 's as indices of ignorance but proposes a deterministic evolu-
tion of this ignorance in time is an oxymoron. Though the successes of mechanics,
equilibrium thermodynamics, and the mathematical formalism of statistical me-
chanics leave no doubt about the validity of the numerical results, the need for
a coherent physical theory capable of encompassing these same results within a
sound unified conceptual framework continues to be an interesting challenge.

B.2. THE UNIFIED THEORY

Intrigued by the experiences, ideas, and concerns just cited, Hatsopoulos and
Gyftopoulos [5-8] have proposed a resolution of the dilemmas and paradoxes that
have preoccupied generations of physicists over more than a century in their at-
tempts to rationalize the relation between mechanics and thermodynamics. The
resolution differs from all statistical interpretations of thermodynamics, in gen-
eral, and from Maxwell's explanation, in particular. This resolution eliminates the
need for the statistical estimates ai, and is the quantum-theoretic underpinning
of the exposition of thermodynamics summarized in Appendix A, that is, the
exposition which asserts that thermodynamics is a general, nonstatistical or non-
informational theory of all physical phenomena. The basis of the resolution is a
unified quantum theory of mechanics and thermodynamics which without modifi-
cation encompasses all systems (both microscopic and macroscopic), and all states
(both thermodynamic equilibrium and not thermodynamic equilibrium). The key
 36

HOMOGENEOUS ENSEMBLE

[IJ[IJ[IJ
[IJ[IJ[IJ
[IJ00

OVERALL DENSITY = P

Figure B-2. Representation of a homogeneous ensemble.

for the elimination of the statistical probabilities a.; is the recognition that the only
density operators p > p2 that are subject to the laws of physics (quantum the-
oretic and thermodynamic) are those that can be represented by a homogeneous
ensemble. In such an ensemble, every member is assigned the same p as any other
member (Figure B-2) and experimentally (in contrast to algebraically) p cannot be
decomposed into a statistical mixture of either projectors or density operators dif-
ferent from p. The impossibility of decomposition is analogous to von Neumann's
conclusion that a projector cannot be decomposed into a statistical mixture of
states of classical mechanics. Moreover, and perhaps more importantly, the exten-
sion of the concept of homogeneity to density operators p > p2 is accomplished
without radical modifications of the quantum-theoretic postulates and theorems
about observables, measurement results, values of observables, and densities or
probabilities of measurement results. Key concepts - definitions, postulates, and
theorems - of the unified theory are discussed briefly below. They are included
here because of the emphasis that must be given to the concept of homogeneous
or unambiguous ensemble.

B.3. KINEMATICS
The term kinematics refers to the definitions of the terms system, property, and
state, all at an instant in time.

B.3.l. System
The meaning of the term system is discussed in Appendix A. The quantum-
theoretic representation of a system is as follows.

B.3.2. System Postulate

To every system there corresponds a complex, separable, complete, inner prod-
uct space, a Hilbert space 1{. The Hilbert space of a composite system of two
distinguishable subsystems 1 and 2, with associated Hilbert spaces 1£1 and 1£2,
respectively, is the direct product space 1£1 18> 1£2 [331.
 37

B.3.3. Homogeneous or Unambiguous Ensemble

At an instant in time, an ensemble of identical systems is called homogeneous
or unambiguous [8] only if upon subdivision into sub ensembles in any conceivable
way short of measurements, each sub ensemble yields in every respect measurement
results - spectra of values and frequency of occurrence of each value within a
spectrum - identical to the corresponding results obtained from the ensemble.
For example, the spectrum of energy measurement results and the frequency of
occurrence of each energy measurement result obtained from any sub ensemble
are identical to the spectrum of energy measurement results and the frequency
of occurrence of each energy measurement result obtained from an independent
ensemble that includes all the subensembles.

B.3.4. Preparation
A preparation is a reproducible scheme used to generate one or more homogeneous
ensembles for study.

B.3.S. Property
The meaning of the term property is discussed in Appendix A. It refers to any
attribute of a system that can be quantitatively evaluated at an instant in time
by means of measurements and specified procedures. All measurement results and
procedures are assumed to be precise.
Without any modifications the meanings of the concepts of homogeneous en-
semble, preparation, and property are valid in all paradigms of physics.

B.3.6. Observable
From the definition just cited, it follows that each property can be observed, that
is, evaluated. Traditionally, however, in quantum theory, a property is called an
observable only if it conforms to the following mathematical representation.

B.3.7. Correspondence postulate

Some linear Hermitian operators A, B, ... on Hilbert space H, which have complete
orthonormal sets of eigenvectors, correspond to observables of a system [33].
As explained by Park and Margenau, the content of this postulate is slightly
different from that of its analogues in typical axiomatics inspired by the work of
von Neumann. In its original form, the correspondence postulate included both
of the following statements: (i) every Hermitian operator corresponds to a physi-
cal observable; and (ii) every observable has a Hermitian operator representative.
Superselection rules introduced by Wick et al. [34] exclude certain Hermitian op-
erators from being observable. By replacing the word every in statement (i) by
the word some, superselection rules are satisfied. Compatibility of simultaneous
measurements introduced by Park and Margenau [33] excludes certain observables
from corresponding to Hermitian operators. In addition, in a unified theory of
mechanics and thermodynamics other properties are observable, such as temper-
ature, but correspond to no Hermitian operators. By replacing the word every in
statement (ii) by the word some the asymmetry between observables and oper-
ators is embraced. It is clear that the correspondence postulate as stated earlier
accommodates both the asymmetry between operators and observables and the
asymmetry between observables and operators.
38

B.3.8. Measurement Act

A measurement act is a reproducible scheme of measurements and operations on
a member of an ensemble. The result of such an act is a precise - error free -
number associated with an observable.
If a measurement act is applied to each and every member of a homogeneous
ensemble, the results conform to the following mathematical representation.

B.3.9. Mean-value Postulate

If a measurement act of an observable represented by Hermitian operator A is
applied to each and every member of a homogeneous ensemble, there exists a
linear functional m(A) of A such that the value of m(A) equals the arithmetic
mean of the ensemble of A measurements, that is,

m(A) = (A) = Eia;/N for N --+ 00 (B-1)

where a; is the error-free measurement result of the measurement act applied to

the ith member of the ensemble, and (A) another notation for m(A) [33].

B.3.1O. Mean-value Theorem

For each of the mean-value functionals m(A) of a system at an instant in time,
there exists the same Hermitian operator p such that [33]

m(A) = (A) = Tr [pAl (B-2)

The operator p is known as the density operator or the density of measurement

results of observables. The concept of the density operator was introduced by von
Neumann [35] as a statistical average of projectors. In contrast, here p is restricted
to homogeneous ensembles and, therefore, it is exclusively quantum-theoretic. The
operator p is proven to be Hermitian, nonnegative-definite, unit trace and, in
general, not a projector [8,22,36], that is,

p > 0; Trp =1 ; (B-3)

B.3.11. Probability Theorem

If a measurement act of an observable represented by operator A is applied to each
and every member of a homogeneous ensemble characterized by p, the probability
or frequency W(an) that the results will yield eigenvalue a., is given by the relation

(B-4)

where An is the projection onto the subspace belonging to a."

Alan} = a.,Ia..,} for n = 1,2, .... (B-5)

and Ian} the nth eigenket of operator A.

B.3.12. Measurement Result Theorem

The only possible result of a measurement act of the observable represented by A
is one of the eigenvalues of A (equation B-5).
 39

Though the statements of the mean-value postulate, and the probability and
measurement result theorems are practically the same as those given by Park and
Margenau [33], here the contents of the statements differ from those of Park and
Margenau because of the restriction of p to a homogeneous ensemble. The impor-
tance and necessity of this restriction in the unified theory cannot be overempha-
sized.

B.4. COMMENT ON THE PICTORIAL REPRESENTATION OF A

HOMOGENEOUS ENSEMBLE
Because no conceivable decomposition of a homogeneous ensemble short of mea-
surements can yield component sub ensembles with different measurement results
- spectra of values and frequency of occurrence of each value - the density
operator p can be assigned to each member of the homogeneous ensemble as
shown in Figure B-2. This assignment introduces no ambiguities, that is, a ho-
mogeneous ensemble can be safely thought of as consisting of identical systems
each of which has the same operators A, B, ... that represent observables, the
same values (A), (B), ... of observables, and the same density operator p. In fact,
this unambiguous characterization of a member at an instant in time is all that
can be said at that time.

B.4.l. State
The meaning of the term state is discussed in Appendix A. In essence, it is all
that can be said about a system at an instant in time. In view of the discussion
of the meaning of Figure B-2, the mathematical representation of state in the
unified theory consists of a set of Hermitian operators A, B, ... that correspond to a
complete set of independent observables - the value of an independent observable
can be varied without affecting the values of other observables - and the relations
(A) Tr[pA] 2.:i ai/ N
(B) Tr [pB] 2.:i bi/N (B-6)

In these relations, either the density operator p is specified a priori and the values
of the observables are calculated, or the values of all the independent observables
2.:i ai/ N, 2.:i bi/ N, ... are measured, and a unique density operator is calculated.
The density operator is unique because equations (B-6) are linear in p.
Because the idea is sometimes overlooked, it is very important to emphasize
that the quantum-theoretic value of an observable is always determined by the
corresponding equation in set (B-6) and not by any particular measurement result.
A measurement act yields an eigenvalue. But no correct quantum postulate or
theorem asserts that this eigenvalue was necessarily the value of the observable
prior to measurement. An ensemble of measurement results of an observable is
needed in order to ascertain its value.

B.4.2. Dynamics
Under the heading of dynamics there is only one concept, the equation of mo-
tion. As discussed in Refs. [5,11]' the equation of motion of conventional quantum
40

mechanics in the Schroedinger picture is incomplete because it prescribes only a

unitary evolution in time of a projector Pi = p? for i = 1,2, ... , and such an evo-
lution corresponds to a reversible adiabatic process only. In a unified theory, the
equation of motion must account for unitary changes in time of density operators
that are not projectors (p > p2), for reversible adiabatic processes that are not
unitary and, of course, for irreversible processes. Until a complete equation of mo-
tion is universally accepted by the scientific community, three postulates provide
a partial substitute for the purposes of the unified theory - equation (B-7) and
the first and second laws of thermodynamics. The substitute is partial because it
covers only some of the requirements of the unified theory.

B.4.3. Limited Dynamical Postulate

Hatsopoulos and Gyftopoulos [5) postulate that unitary transformations of p in
time obey the relation
-dp
dt
i
= --[Hp
h
- pH) (B-7)
where H is the Hamiltonian operator of the system. The unitary transformation
of p satisfies the equation
pet) = U(t, to) p(to) U+(t, to) (B-8)
where U+ is the Hermitian conjugate of U and, if H is independent of t,
U(t, to) = exp[-(i/h)(t - to) H) (B-9)
and, if H is explicitly dependent on t,

dU~~ to) = -(i/h) H(t) U(t, to) (B-lO)

Though equation (B-7) is well known in the literature as the von Neumann
equation, here it must be postulated for the following reason. In statistical quantum
mechanics [37), the equation is derived as a statistical average of Schroedinger
equations, each of which describes the evolution in time of a projector Pi in the
statistical mixture represented by p, and each of which is multiplied by a time
independent statistical probability (Xi. In the unified theory, p is not a mixture of
projectors and, therefore, cannot be derived as a statistical average of projectors.
It is noteworthy that the dynamical postulate is limited or incomplete because all
unitary evolutions of p in time correspond to reversible adiabatic processes. But
not all reversible adiabatic processes correspond to unitary evolutions of p in time
[7), and not all processes are reversible.

B.4.4. The First and Second Laws of Thermodynamics

A partial relief to the limitations of the incomplete dynamical postulate just cited
is provided by adding to equation (B-7) two more statements, the first law and the
second law of thermodynamics. These statements are given in Appendix A. The
quantum-theoretic postulates and theorems, and the two laws of thermodynamics
provide the conceptual framework for the exposition of the unified quantum theory
of mechanics and thermodynamics, a theory that applies to all systems and all
states. Moreover, the quantum-theoretic concepts lurk behind every aspect of the
exposition of thermodynamics summarized in Appendix A.
 41

It is noteworthy that the third law of thermodynamics is not needed because

it is inherent in the quantum theoretic foundations.

B.4.5. Entropy
On the basis of the new exposition of thermodynamics presented in Ref. [11] and
summarized in Appendix A, and the unified theory presented in Refs. [5-8], and
summarized in this appendix, in Ref. [38] we prove that of all the expressions for
entropy S that have been proposed in the literature the only one that satisfies all
the necessary criteria is given by the relation
S=-kTh[plnp] (B-11)
provided that p is exclusively represented by a homogeneous ensemble. If p is
represented by a heterogeneous ensemble, then equation (B-11) does not represent
the entropy of thermodynamics.
For given values of energy, amounts of constituents, and parameters, if p is a
projector then S = 0, and if p corresponds to the unique stable equilibrium state
required by the second law then S has the largest value of all the entropies of states
that share the given values of energy, amounts of constituents, and parameters. If,
as it is usually done for projectors, we interpret a density operator p as the shape
of constituents of a system, then the entropy of the system is a special measure of
the shape with values ranging from zero to a maximum for each set of values of
energy, amounts of constituents, and parameters.
If we adopt the measure of shape interpretation for entropy, an interesting con-
comitant ensues. Let us assume that the paradigm of the unified quantum theory
was conceived prior to that of classical mechanics, and that a physicist wished to
approximate quantum theoretic results by classical concepts. We can safely pre-
dict that he would have done an excellent job because for macroscopic systems
with highly degenerate eigenkets, densities of measurement results of practically
all observables can be approximated by the Dirac delta function 8(q -qo)8(p-po)
of space coordinates q and momenta p. Though highly accurate, such an approx-
imation would be inadequate because it does not include the concept of shape of
the constituents of the system and, therefore, provides neither the mathematical
representation for the concept of entropy as a property of the constituents, nor the
possibility of change of this mathematical representation over a range of values.
This is another aspect of the inadequacy of classical mechanics to accommodate
the concepts of thermodynamics.

B.4.6. Density Operator of a Stable Equilibrium State

We can find the density operator pO of a thermodynamic or stable equilibrium
state Ao of system A by maximizing the entropy S subject to the constraints
Th p = 1 and (H) = Th [pH] = given value E . (B-12)
For simplicity, we assume that the system has only volume as a parameter, and only
one constituent with an amount n equal to an eigenvalue of the number operator
of the constituent. Moreover, we use a different subscript i even for orthonormal
projectors that correspond to the same eigenvalue.
The constrained maximization solution is proven to be [39]

(B-13)
42

where

and ~ is determined by the value of the energy E because

We can show that

(B-14)

where the subscript "c" stands for fixed values of all the energy eigenvalues £1,
£2,···,and the subscript "0" for state Ao, that is, the partial derivative is taken
along the stable equilibrium state locus for fixed values of parameters (fixed c) and
fixed amount of the constituent n at state Ao. But for stable equilibrium states,
the partial derivative [(8S/8E)e,nL is defined as the inverse temperature of Ao.
Accordingly
~ = l/ kTo (B-15)

B.4. 7. Translational Velocity of a Molecule

We consider a system A in a stable equilibrium state Ao with energy E, number of
molecules n ~ 1, and volume V. For such a state, the value (Pk) of the momentum
of a single molecule in the spatial direction Xk is given by the relation

(B-16)

where the third of these equations results from the fact that the energy eigenkets
are orthonormal and, therefore,

(B-17)

and the last equation from the relation

for all k and m . (B-18)

The proof of equation (B-18) is straightforward. First, we observe that the Hamil-
tonian operator H of the system and the momentum operator Pk of a particular
molecule satisfy the commutation relation

(B-19)

where M is the mass of the molecule. Next, upon defining for all k and m

(B-20)

(B-21)
 43

we can readily prove that [40]

(~Xk)m (~H)m ~ nl(€mIPkl€m}l/2M (B-22)
But for a system with a finite extension Lx along the coordinate axis of Xk, and
an energy eigenket I€m), we have
o < (~Xk)m < Lk and (~H)m = 0 (B-23)
and so equality (B-18) is proved.

References
1. Clausius, R. (1976) In The Second Law of Thermodynamics, J. Kestin (ed.), Dow-
den, Hutchinson and Ross, Inc., pp. 186-187.
2. Wehrl, A. (1978) General properties of entropy, Rev. Mod. Phys., 50, 2, 221-260.
3. Kline, S.J. (1997) The semantics and meaning of the entropies, Report CB-1, Dept.
of Mech. Eng., Stanford University.
4. Obert, E.F. (1960) Concepts of Thermodynamics, McGraw-Hill, New York.
5. Hatsopoulos, G.N. and Gyftopoulos, E.P. (1976) A unified quantum theory of me-
chanics and thermodynamics. Part I. Postulates, Found. Phys., 6, 15-31.
6. Hatsopoulos, G.N. and Gyftopoulos, E.P. (1976) A unified quantum theory of me-
chanics and thermodynamics. Part I1a. Available energy, Found. Phys., 6, 127-141.
7. Hatsopoulos, G.N. and Gyftopoulos, E.P. (1976) A unified quantum theory of me-
chanics and thermodynamics. Part lIb. Stable equilibrium states, Found. Phys., 6,
439-455.
8. Hatsopoulos, G.N. and Gyftopoulos, E.P. (1976) A unified quantum theory of
mechanics and thermodynamics. Part III. Irreducible quantal dispersions, Found.
Phys., 6, 561-570.
9. Beretta, G.P., Gyftopoulos, E.P., Park, J.L., and Hatsopoulos, G.N. (1984) Quan-
tum thermodynamics. A new equation of motion for a single constituent of matter,
Nuovo Cimento, 82B, 2, 169-191.
10. Beretta, G.P., Gyftopoulos, E.P., and Park, J.L. (1985) Quantum thermodynamics.
A new equation of motion for a general quantum system, Nuovo Cimento, 87B, 1,
77-97.
11. Gyftopoulos, E.P. and Beretta, G.P. (1991) Thermodynamics: Foundations and Ap-
plications, Macmillan, New York.
12. Shankar, R. (1994) Principles of Quantum Mechanics, 2nd Ed., Plenum Press, New
York.
13. von Neumann, J. (1955) Mathematical Foundations of Quantum Mechanics, Prince-
ton University Press, New Jersey.
14. Leighton, R.B. (1959) Principles of Modern Physics, McGraw-Hill, New York.
15. Brandt, S. and Dahmen, H.D. (1995) The Picture Book of Quantum Mechanics,
2nd Ed., Springer-Verlag, New York.
16. Slater, J. (1963) Quantum Theory of Molecules and Solids, McGraw-Hill, New York.
17. There is an inconsistency between the calculation of the eigenstates of the proton
and those of the electron of a hydrogen atom. The former are calculated for a proton
confined in a finite size container, whereas the latter are calculated for a Coulomb
potential extending over all space. As a result the operator P~lectrDn is not well
defined because Tr exp( -~HelectrDn) is unbounded, and the entropy of the electron
is infinite. For details see: Conway, J.B. (1985) A Course in Jilunctional Analysis,
Springer-Verlag, New York.
18. Dahmen, H.D. and Stroh, T. (1998) Private communication.
19. Hatsopoulos, G.N. and Gyftopoulos, E.P. (1979) Thermionic Energy Conversion,
Vol. II, MIT Press, Cambridge, Massachusetts.
44

20. Timoshenko, S. and Young, D.H. (1948) Advanced Dynamics, McGraw-Hill, New
York.
21. Kuhn, T. (1970) The Structure of Scientific Revolutions, 2nd Ed., Chicago Univer-
sity Press, Chicago.
22. Park, J.L. and Simmons, R.F. (1983) In Old and New Questions of Physics, Cos-
mology, and Theoretical Biology: Essays in Honor of Wolfgang Yourgrau, A. van
der Merwe (ed.), Plenum Press, New York, pp. 289-308.
23. Korsch, H.J. and Steffen, H. (1987) Dissipative quantum dynamics, entropy produc-
tion and irreversible evolution towards equilibrium, J. of Phys., A20, 3787-3803.
24. Lindblad, G. (1976) On the generators of quantum dynamical semigroups, Commun.
Math. Phys., 48, 119-130.
25. Hatsopoulos, G.N. and Keenan, J.H. (1965) Principles of General Thermodynamics,
Wiley, New York.
26. Gyftopoulos, E.P. (1997) Fundamentals of analyses of processes, Energy Conv.
Mgmt, 38,15-17, 1525-1533.
27. Maxwell, .J.C. (1871) Theory of Heat, Longmans, Green and Co., London.
28. Mcquarrie, D.A. (1973) Statistical Thermodynamics, Harper and Row, New York.
29. Callen, H.B. (1985) Thermodynamics and an Introduction to Thermostatics, 2nd
Ed., Wiley, New York.
30. Schroedinger, E. (1935) Discussion of probability relations between separated sys-
tems, Proc. Cambridge Philos. Soc., 31, 555-563, and (1936) Probability relations
between separated systems, Proc. Cambridge Philos. Soc., 32, 446-452.
31. Park, J.L. (1968) Nature of quantum states, Amer. J. Phys., 36,211-226, and (1988)
Thermodynamic aspects of Schrodinger's probability relations, Found. Phys., 18, 2,
225-244.
32. .Jancel, R. (1963) Foundations of Classical and Quantum Statistical Mechanics,
Pergamon Press, Oxford, England, 1963.
33. Park, .J.L. and Margenau, H. (1968) Simultaneous measureability in quantum the-
ory, Int. J. Theor. Phys., 1, 3, 211-283.
34. Wick, G.C., Wightman, A.S., and Wigner, E.P. (1952) The intrinsic parity of ele-
mentary particles, Phys. Rev., 88, 101-105.
35. von Neumann, .J., op. cit., pp. 313-316.
36. .Jauch, J.M. (1973) Foundations of Quantum Mechanics, Addison Wesley, Reading,
MA.
37. Tolman, R.C. (1938) The Principles of Statistical Mechanics, Oxford University
Press, London, Great Britain.
38. Gyftopoulos, E.P. and Qubuk~u, E. (1997) Entropy: Thermodynamic definition and
quantum expression, Phys. Rev. E, 55, 4, 3851-3858.
39. Katz, A. (1967) Principles of Statistical Mechanics, W.H. Freeman, San Francisco,
CA.
40. Jauch, .J.M. op. cit., pp. 160-163.
THERMODYNAMIC OPTIMIZATION OF INANIMATE AND
ANIMATE FLOW SYSTEMS

A.BEJAN
Duke University
Department of Mechanical Engineering
and Materials Science
Durham, NC 27708-0300, USA

1. An Old and Prevalent Natural Phenomenon

Thermodynamic optimization is, literally, the search for the best thermodynamic
performance subject to present-day constraints. This philosophy is very old. It has
been with us throughout the history of the heat engine. Its reach however, is much
broader and more permanent. We can be sure that the performance of power plants will
continue to improve in time, in the same way that, in time, a sudden rain fall will
generate an efficient flow structure (dendritic channels), not Darcy flow alone.
In the broader sense, optimization of performance has been the driving force behind
the development of all machines, starting with the wheel and the pulley. Physiologists
and geophysicists too are coming to the conclusion that optimization of performance
has been the generating mechanism for geometric form and size in all plants, animals
and natural flow structures.
If thermodynamic optimization is such an old and prevalent natural phenomenon,
then what is new? New is the streamlining of the study of thermodynamic optimization
into a simple and coherent discipline that is now recognized as an integral part of
contemporary thermodynamics. The constant streamlining, simplification and
generalization of individual observations and applications has always been the flow we
call science, or knowledge. Exergy analysis, thermodynamic optimization * and
thermoeconomics are the main methods of this new discipline. Thermodynamicists
have developed these methods for the past one hundred years or more. Significant
growth has occurred in the wake of the energy crisis, as is shown by the chronological
sequence of books written in this field beginning with 1982 [1 - 10].
In this chapter I review the most basic aspects of thermodynamic optimization,
with particular emphasis on why this method is different than pure thermodynamics. I
then rely on some of the simplest models in order to stress that the method also requires
modeling, which is a construction that precedes optimization. I conclude with
illustrations from my most recent applications of the method, in a sequence that
continues in my accompanying chapter on Constructal Flow Geometry Optimization.

* Thermodynamic optimization is also practiced under the equivalent titles of entropy generation
minimization (EGM), finite-time (finite-size, finite-resources), and endoreversible (exoirreversible)
thermodynamics. There is absolutely no fundamental difference between these lines of work: see sections
3 and 8.

4S
A. Bejan and E. Mamut (eels.), Thermodynmnic Optimization o/Complex Energy Systems, 45-60.
© 1999 KlIlwer Academic Publishers.
46

2. Thermodynamics Combined with Heat Transfer and Fluid Mechanics

It is instructive to begin with a brief look at why in thermodynamic optimization we

must rely on heat transfer and fluid mechanics, not just thermodynamics. Consider the
most general system-environment configuration, namely a system that operates in the
unsteady state. Its instantaneous inventories of mass, energy, and entropy are M, E,
and S. The system experiences the net work transfer rate W, heat transfer rates
(Qo, QI,···Qn) with n + I temperature reservoirs (To, T), ... ,Tn ), and mass flow rates
(min' m out ) through any number of inlet and outlet ports. Noteworthy in this array of
interactions is the heat transfer rate between the system and the environmental
(atmospheric) temperature reservoir, 00' on which we focus shortly.
The thermodynamics of the system consists of accounting for the first law and the
second law [4],
dE n. . . .
- = I, Qi - W + I, mh - I, mh (1)
dt i=O in out

S·gen = -dS - ~
k
Qi ~.
-
~.
- k ms + k ms;::: ,
0 (2)
dt i=O Ti in out

where h is shorthand for the sum of specific enthalpy, kinetic energy, and potential
energy of a particular stream at the boundary. In Eq. (2) the total entropy generation rate
Sgen is simply a definition (notation) for the entire quantity on the.left-hand side of the
inequality sign. We shall see that it is advantageous to decrease Sgen' and this can be
accomplished only by changing at least one of the quantities (properties, interactions)
specified along the system boundary.
We select the environmental heat transfer 00 as the interaction that is allowed to float
as Sgen varies. Historically, this choice was inspired (and justified) by applications to
power plants and refrigeration plants, because the rejection of heat to the atmosphere
was of little consequence in the overall cost analysis of the design. Eliminating 00
between Eqs. (1) and (2) we obtain

. d n[ TO]' ()
W=--(E - ToS)+ I, 1 - - Qi + I, m(h - Tos) - I, m h - Tos - TOSgen (3) .
dt i=1 Ti in out

The power output in the limit of reversible operation (Sgen = 0) is

Wrev = - i.(E - ToS) +

dt i=1
±(l-TO)Qi +I,m(h - Tos) - I,m(h - Tos)
Ti in out
(4)

In engineering thermodynamics each of the terms on the right-hand side of Eq. (4) is
recognized as an exergy of one type or another [2, 8], and the calculation of Wrev is
known as exergy analysis. Subtracting Eq. (3) from Eq. (4) we arrive at the Gouy-
Stodola theorem,
Wrev - W= ToSgen (5)

In Eq. (5) Wrev is jued because all the heat and mass flows (other than (0) are fixed.
Pure thermodynamics (e.g., exergy analysis) ends, and EGM begins, with Eq. (5).
 47

The lost power (W rev - W) is always positive, regardless of whether the system is a
power producer (e.g., power plant) or a power user (e.g., refrigeration plant). To
minimize lost power when Wrev is fixed is the same as maximizing power output in a
power plant, and minimizing power input in a refrigeration plant. This operation is
also equivalent to minimizing the total rate of entropy generation. If Wrev is not held
fixed while the system design is being changed, then Eq. (5) and the calculation of Sgen
lose their usefulness.
The critically new aspect of the EGM method-the aspect that makes the use of
thermodynamics insufficient, and distinguishes EGM from exergy analysis-is the
minimization of the calculated entropy generation rate. To minimize the irreversibility
of a proposed design, the analyst must use the relations between temperature differences
and heat transfer rates, and between pressure differences and mass flow rates. He or she
must express the thermodynamic nonideality of the design Sgen as a function of the
physical characteristics of the system, namely to finite dimensions, shapes, materials,
finite speeds, and finite-time intervals of operation. For this the analyst must rely on
heat transfer and fluid mechanics principles, in addition to thermodynamics. Only by
varying one or more of the physical characteristics of the system, can the analyst bring
the design closer to the operation characterized by minimum entropy generation subject
to size and time constraints.

3. The Method and Its History

To appreciate the origins and history of the method of thermodynamic optimization, it

is important to note that the field of low temperature refrigeration was the first where
irreversibility minimization became an established method of optimization and design
[1]. It is easy to prove that the power required to keep a cold space cold is equal to the
total rate of entropy generation times the ambient temperature, with the observation that
the entropy generation rate includes the contribution made by the leakage of heat from
To into the cold space [1]. The structure of a cryogenic system is in fact dominated by
components with heat transfer irreversibility, that leak heat, e.g., mechanical supports,
radiation shields, electric cables, and counterflow heat exchangers. The minimization of
entropy generation along a heat leak path consists of optimizing the path in harmony
with the rest of the refrigerator of liquifier.
Figure 1 shows a mechanical support of length L that connects the cold end of the
machine (TL) to room temperature (TH). The rate of entropy generation inside the
support shown as a vertical column is [1,9],

S. = fT -dT
QH (6)
gen T T2
L

where it is important to note that the heat leak 0 is allowed to vary with the local
temperature T. The local heat leak decrement dO is removed by the rest (the internal
part) of the installation, which is modeled as reversible. The heat leak is also related to
the local temperature gradient and conduction cross-section A,

Q. =kA dT (7)
dx'

where the thermal conductivity k(T) decreases toward low temperatures. Rearranged and
48

integrated, Eq. (7) places a size constraint on the unknown function Q(T),

(8)

According to variational calculus, the heat leak function that minimizes the S
=
integral (6) subject to the finite-size constraint (8) is QoPt (A.k)I12T. The LagraJg~
multiplier A. is determined by substituting QoPt into the constraint (8):

.
Q = (~f.TH kl/2 dT)k 1l2T (9)
opt L TL T '

. A( kl/2
. =L- fTL -T
S gen,mlO dT
TH
)2
(10)

Equation (6) was provided by thermodynamics and Eq. (7) by heat transfer: together
they prescribe the optimal design (9, 10), which is characterized by a certain distribution
of intermediate cooling effect (dQ I dT)opt. Any other design, Q(T), will generate more
entropy and will require more power in order to maintain the cold end of the support at
T L . Quantitative and older examples from cryogenic engineering are given in [1, 3, 4].
Together, Eqs. (6) and (8) illustrate the backbone of the method of thermodynamic
optimization subject to a physical constraint, as it was practiced in engineering before
1974 when the preceding example was published [11]. Key advances had been made as
early as the 1950s in heat exchanger optimization [12], cryogenics [13, 14], solar
thermal energy conversion [IS], and power maximization in power plants [16 - 24].
These advances outlined the field before the energy crisis, long before the growth that
occurred in the 1980s and 1990s. For example, Curzon and Ahlborn's first physics
paper on power maximization [25] is a rediscovery of the idea published 18 years earlier
by Chambadal [16] and Novikov [19]. Physics papers followed Curzon and Ahlborn in
the 1980s: there is absolutely no difference [26, 27] between this line of work and the
constrained thermodynamic optimization developed by the pioneers and illustrated in this
chapter (see also section 8).

L, TH----~----~i------

I
I I
I I
ill I I

0_
Q+dQ
Reversible

.+d., T:,d:==~=?

~~
II {J. II
I
Q
I
____~IL-__~IL-__________
n TL

Figure 1. Mechanical support with variable heat leak and intennediate cooling [1].
 49

4. Maximum Power from Flow between Two Reservoirs

Let us consider the peer-refereed and published criticism [28, 29] of Chambadal,
Novikov, and Curzon and Ahlborn's power plant model, and their maximization of
power output. We consider this model in a sequence of three analogs: mechanical [9],
electrical [9], and thermodynamic, Fig. 2.
Imagine that we have access to two infinitely large reservoirs of a fluid at different
pressures (PH, Pd, and that we install a turbine between the reservoirs such that the
stream mproduces the power W [9]. The stream flows with pressure drops (PH - PHC
and PLC - Pd along the two ducts that connect the turbine to the reservoirs. The actual
pressure difference across the turbine (PHC - PLC ) is smaller than the overall pressure
difference (PH - Pd. The simplest flow resistance model is represented by the linear
relations

(11)

where the flow resistances (RH' Rd are related to the physical characteristics of the ducts
and fluid. For simplicity, we assume that the fluid is incompressible (density p) and
that the turbine is reversible. The power output is simply W = m
(P HC - PLC> / p:
this quantity can be maximized by varying m, that is by imagining-as needed-
trickles or gushes of fluid, depending on what yields a higher W. We can do this
because we imagined that the reservoirs are infinite, i.e., the size of m has no absolute
effect on the reservoirs and our ability to access them. It is easy to show that maximum
power output occurs when the pressure difference across the turbine is exactly half of the
overall pressure difference,

(12)

The electrical analog of this example is a reversible electrical motor connected to

two constant-voltage lines (VH' V L). The voltage drop across the motor is V HC - V LC .

RH' rit RH I

reversible reversible
turbine ~ w electrical ~
motor
PLC V LC

RL rit RL I

(a) (b)

Figure 2. Mechanical, electrical and thermodynamic systems producing power while bridging the gap
between two reservoirs of pressure, voltage and temperature.
50

If the resistances of the conductors between motor and reservoirs are RH and R L, then we
may use the linear model V H - V HC = RHI and V LC - V L = RLI, where I is the current.
The power delivered by the motor, I(VHC - VLd, can be maximized by selecting the
steady-state current that we need. We can do this because we have free access to V H and
V L, and because V H and V L are insensitive to I. Maximum power occurs when the
current is such that the equivalent ofEq. (12) holds, (V HC - VLdopt = (VH - Vdl2.
The thermodynamic analog of this dream is the reversible heat engine sandwiched
b~tween two infinitely large entities at different temperatures (TH, T d. The heat input
(QH) and the heat rejection require finite temperature differences,

(13)

where the thermal resistances (RH' R L) are determined by the dimensions and materials
of the systems that effect the two thermal contacts, T wengine and engine-T L. Internal
engine reversibility means that Q H / T H = Q L / T L . The power output W = QH
(1 - T LC / T HC) can. be maximized by varying Q H freely. It turns out that the
optimum heat input QH.opt is such that maximum power is reached when

( THC) (TH)"2
T LC opt = TL (14)
The three-way analogy is complete. The questioning of the thermodynamic
optimum, Eq. (14), concerns its realism. In the optimization of a power plant, is the
heat input-t~e fuel burnt-a degree of freedom? Do we have access to an infinite
reservoir of QH in the same way that we can attach toy turbines and toy electrical
motors to the 'Yater and power lines in the house? The answer is no in power plant
designs where QH is an expensive commodity (e.g., fossil fuel). We may contemplate
situations in which the power plant can be placed in the vicinity of two "environments"
at different temperatures (e.g., solar power, geothermal, hydroelectric applications), but
even then the access to these temperature reservoirs is limited by economic and
geophysical constraints. An example of these limitations is given in the next section.
The thermodynamic optimum of Eq. (15) has also been reported as an efficiency at
maximum power 11 = Wmax / QH,Opt> which is the conclusion of Chambadal [16],
Novikov [19] and Curzon and Ahlborn [25],
"2
11=1-
(
~~ ) (15)

This result is questionable not only for the free- Q H assumption on which it is based,
but also conceptually: to use an efficiency (a figure of merit) ratio W / QH makes no
sense when the denominator varies freely. It has been shown that the free- Q H modeling
error is also responsible for the observation that in Curzon and Ahlborn's model the
operating point of maximum power is not the same as the point of minimum entropy
generation [29], when the entropy generation is summed only over the space between TH
and TL in Fig. 2c. The reason is clear if we re-examine Eq. (5): when Q H is free to
vary, the term Wrev = Q H (l - T L / T H) is also free to vary.
Gyftopoulos [28] pointed out that the approximate agreement between Eq. (15) and
the efficiencies of old power stations is simply a coincidence. It turns out that
 51

agreement with the reported efficiencies can also be reached when we hold QH fixed, and
use the same power plant model as in Fig. 2c [8,9]. For better description, we use the
thermal conductances UHA H = Rill and ULA L = ReI, where U H and U L are the
overall heat transfer coefficients based on the two heat transfer contact areas, AH and
A L , respectively. If the overall size of the heat exchange equipment is constrained [4]

(constant) (16)

then the power output of the model can be maximized with respect to how UA is split
into UHA H and ULA L. The optimal way is (UHAH)opt = (ULAdopl' and the
maximum power <:>utput is represented by the maximum efficiency (i.e., maximum
power per unit of QH):

11.
max
=I - (.:!l)/(1
TH
-~)
TH UA
(17)

Wp,H

- I==::::J-
Wp,L Wmax

- t-----'-

reversible
compartment

Figure 3. Power plant model with two finite-size capacity flow rates [30].
52

This efficiency is expectedly lower than the Carnot effic!ency associated with the same
temperature extremes. When the dimensionless group Q H I (T H UA) is a constant of
order 0.1, the T1max values agree well with the reported efficiencies of ten power plants
[10]. The constancy of the QH / (T H UA) group makes sense because in the real world
both QH and UA must scale with the overall size of the power plant.
Is there a class.of power plants in which the heat input Q H is as free to vary as the
heat rejection rate QL? The trouble with finding an answer to this question in Fig. 2c
is that in that model the physical origin of QH is not shown: Fig. 2c is incomplete as
a model of a power plant. A possible alternative is the power plant model shown in the
lower-right detail of Fig. 3 [30]. Here the heat input QJ-I is drawn from a hot single-
phase stream of original temperature T H' namely Q H = C H (T H - T HC)' where
C H = (mcp)H is the capacity rate of the stream. It is assumed that the heat exchanger
between the hot stream and the reversible compartment is sufficiently large such that the
outlet temperature of the hot stream is practically equal to the hot-end temperature of the
reversible compartment, THe. .
The cold end of the power plant is modeled similarly. The heat transfer rate Q L is
rejected to a cold stream of original temperature T L and capacity rate C L = (mc p k.
When the cold-end heat exchanger is sufficiently large, the outlet temperature of the cold
stream is nearly equal to the c~ld-end temperature of the reversible compartment, such
that the heat rejection rate is Q L = C L (T LC - T d. Models similar to Fig. 3 have
been recognized in the literature. For example, a model with one stream at the hot end
was proposed in [4], p. 382.
It can be shown that the models of Fig. 2c and 3 are analogous analytically: the
thermal conductances U HAH and U LAL are replaced by the finite capacity rates C H and
C L. The optimization with respect to variable Q H leads to Eq. (14) and

(18)

which shows that the maximum power Wm can be increased further by increasing C H
and C L. Equation (18) also shows that the maximum power output is limited by the
smaller of the two capacity flow rates: to see this, imagine that C H « C L such that
Eq. (19) approaches the limit Wm == C H(T ~12 - T L/2)2. The search for a larger power
output sends us in the direction of larger streams (CH,Cd, i.e., in the direction of large
fluid reservoirs of temperatures T Hand T L, and, especially, large flow rates. On the
practical aspects of this limit we focus next.
Consider now the complete power plant model shown in the upper part of Fig. 3,
where the two-stream model has been connected to the fluid reservoirs T Hand T L' It is
useful to think of Fig. 3 as the simplest possible model of a geothermal or ocean
thermal energy conversion (OTEC) power plant. Even when the fluid reservoirs are
infinite, the flow rates cannot be increased indefinitely: the pumping power
(W p,H + Wp,L) must be subtracted from the power output of the power plant (W m)'
Let us assume that the two fluids are incompressible liquids, such that the pumping
power requirement of each fluid can be expressed as Wp = mLl}> / (PT1n)' In this
expression T1n is the second law efficiency of the pump (assumed constant; also known
 53

as pump isentropic efficiency, II p)' m is the mass flow rate, p is the density of the
respective incompressible liquid, and AI> is the pressure drop along the duct,
AI> = r mn. In this expression r is the effective resistance to fluid flow, and n covers
the range from laminar flow (n = I) to turbulent flow in the fully rough regime
(n == 2). We treat rand n as two constants that are known from the design of the duct.
The net power output of the installation is Wnet = Wm - Wp,H - Wp,L.' where Wm
is the maximized power output shown in Eq. (18). It can be shown that W net becomes

·
W = LlTCHC L _ R Cn+1 R Cn+1 (19)
net C +C H H - L L
H L

where LlT = (T //2 - T [12)2, and RH,L = (r / Plln c~ + 1)H,L' The net power output
has maxima with respect to both C H and C L : the expressions for the optimal capacity
rates [30] show that the flow path with the lower R value should be assigned to the
stream with the larger C value. The net power output for C H.opt and C L.opt is

Wnet .m = n[LlT/(n + l)t+l)/n[R~/(n+2) + R~/(n+2)tn+2)/n (20)

This power output was maximized three times: with respect to QH (or THe) in the
internal design of the model of Fig. 3, and with respect to C H and C L . Equation (20)
shows that Wnet,m is larger when RH and RL are smaller, and that the maximum net
power is limited by the larger of the two flow resistances.

5. Maximum Power from a Stream of Hot Exhaust

Another way to provide a freely varying heat input to a power plant is by placing the
hot surface Ts(x) of the power cycle in contact with a hot stream that is being dumped
into the ambient, Fig. 4 [31]. The length traveled by the hot stream (L) is proportional
to the heat transfer area, A = P L, where p is the heat transfer area per unit of flow path

o x L

Figure 4. Power plant model with unmixed hot stream in contact with a
nonisothermai heat transfer surface [31].
54

length. The power producing compartment is a succession of many infinitesimal

reversible compartments of the kind shown in the center of the figure. The infinitesimal
power output is .dW = [l - To ITs(x)] dQH' where the temperature is plotted on the
ve~ical, and dQH = mCpdT. The heat transfer through the heat exchanger area is
dQH = [T(x) - Ts (x)] Vpdx. Combining these equations and integrating from x = 0
to x = L = Alp while treating V as a constant we arrive at the finite-area constraint and
the total power output:
dT VA
- - - = - - = Nt (21)
T - Ts mc p u

(22)

An alternate route to calculating the power output W is to apply the Gouy-Stodola

theorem to the larger system (extended with dashed line) in Fig. 4:
W = W rev - To Sgen. The reversible-limit power output W rev corresponds to the
reversible cool~ng of the stream from THall the way down to To. The entropy
generation rate Sgen is the total amount associated with the larger system, and is due to
two sources: the temperature difference T - Ts, and the finite temperature difference
required by the external cooling rate Qe = mcp (T out - To). These two contributions
are represented by the two terms in the expression

S·gen = f (1
T
TH - - -I ).mc pdT + (.mc p ITo
n-- + -Qe ) (23)
uu' Ts T Tout To
To maximize W is equivalent to minimizing Sgen, because W rev is fixed. The
entropy generation expression (23) is subject to the size constraint (21). There are two
degrees of freedom in the minimization of Sgen: the shape of the surface temperature
function Ts(x), and the place of this function on the temperature scale (i.e., closer to T H
or To). The second degree of freedom is alternately represented by the value of the
exhaust temperature Tout.
It was shown in [31] that the optimal hot-stream temperature distribution T(x) is
exponential in x, and so is the temperature Ts(x) along the hot end of the system that
converts the heating into mechanical power. At any x, the temperature difference across
the heat exchanger is proportional to the local absolute temperature. The optimal
solution can be implemented in practice by using a single-phase stream in place of the
Ts(x) surface: this stream runs in counterflow relative to the hot stream m. The
counterflow imbalance (the ratio between the capacity flow rates of the two streams) is
the result of thermodynamic optimization.

6. Optimum Breathing and Heart Beating

An illustration of natural thermodynamic optimization subject to global constraints is

provided by the characteristic, finely tuned frequencies of physiological processes such as
breathing and heart beating [10,32]. Consider a simple model of how the lung works.
During the inhaling time t( the chest cavity experiences the volume increase V, while
 55

drawing in atmospheric air (To, Po) through the nasal Pllssages with the time averaged
mean velocity U I. This flow is made possible by the lower-than-atmospheric pressure
(po - API) maintained inside the lungs during the inhaling process. The pressure
difference varies monotonically with the mean inhaling velocity, API = rUt, where r
is the fluid resistance of the air passages, and n varies from n == 1 in laminar flow to
n == 2 in turbulent flow. Mass conservation during the tl-Iong process requires that
PoUIA f tl = PIV, where Af is the flow cross-sectional area of the nasal passages, po
is the air density at atmospheric conditions, and PI is the density of the inhaled air
(evaluated at Po - API and body temperature).
The exhaling process is characterized by the excess pressure AP2 inside the chest
cavity, AP2 = r U;, where r is the same nasal flow resistance, and U 2 is the mean
exhaling velocity. The duration of the exhaling process is t2, therefore mass
conservation requires that PI U 2Aft2 = PI V. The total work done by the thorax muscles
during the inhaling and exhaling cycle (t I + t2) is W = (API + AP2)V, The cycle-
averaged power consumed by the thorax muscles is W = W / (tl + t2)' By combining
these equations with the assumption that Po "" PI , the average power reduces to
. V n + I t l- n +tin
W = r --. (24)
At tl +t2
Equation (24) shows that the power requirement decreases monotonically as either tl
or t2 increases. Effortless breathing corresponds to the longest inhaling and exhaling
strokes possible. It turns out, however, that both tl and t2 must be finite, as required by
the chief function of the breathing process: to facilitate the transfer of oxygen between
the freshly inhaled air and the vascularized tissue beneath the surface of the pulmonary
passage. The transfer of O2 is by mass diffusion on both sides of the surface. The scale
analysis of mass diffusion begins with writing that durin~ the inhaling interval tl the
O 2 mass flux is of order j - DAC /8, where 8 - (Dtl) /2 is the mass penetration
distance associated with t l . We find that the oxygen mass transferred during the tl
interval is m = j A t l . or that m - A tl DAC / (Dtl )1/2, where A is the total contact
(mass transfer) area of the air passages. The important quantity is the mass transfer rate
averaged over one breathing cycle, m= m / (tl + t2), or m- A DII2 AC tl'2 /
(tl + t 2 ). This m result places a constraint on the inhaling and exhaling time
intervals, forcing both to be finite,
ll2
t -- _____
-1 m =K (constant) (25)
tl + t2 ADI/2AC
Finally, consider the behavior of the average power requirement, Eq. (24), subject to
the mass transfer rate constraint (25). Eliminating t2 and solving aw / atl = 0 we
obtain the optimal inhaling time tl,opt for minimum breathing power consumption:

(1 - Ktt;Pt)[(-h- _1)n + 1] _ _n (1 < n < 2) (26)

Ktl,opt 2n + 1

The chief implication of this result is that the scale of tl,opt is K-2, In view of Eq. (25),
we also conclude that tl,opt and t2,opt are of the same order of magnitude, which is in
56

agreement with the large volume of observations catalogued and correlated in the
biology literature [10]. The breathing time intervals are proportional to

(27)

Equation (27) also explains why the breathing interval increases with the size of the
animal. Specifically, Eq. (27) predicts that the breathing time intervals must vary as
(A I m)2. We need two relations: (i) the relation between contact area (A) and body
size (M), and (ii) the relation between metabolic rate (or m) and body size. Relation (ii)
may be derived by modeling the animal as an open system in the steady state [10],
which leads to m = (constant) XM2I3. The numerous empirical data are correlated even
better by m =(constant) XMO.76.
For the relation (i) between mass transfer area and body size, we proceed
theoretically by arguing that the thickness of the tissue penetrated by mass diffusion
during the time scale tl,opt (or t2,opt) is 0 - (Dtl,opt)ll2. The total volume of the tissue
penetrated by mass diffusion during this time is V - A O. Finally, since the volume V
must be proportional to the body size M, we conclude that A 0 is proportional to M,
i.e., that A = (constant) x MI ti,~!" Substituting this relation and m - MO. 76 into Eq.
(27) we predict that the breathing time increases with the body size, and that this
increase is relatively slow, (tlo t2)opt = (constant) x MO.24. This relation agrees
amazingly well with the empirical correlation (allometric law) known for t} as a
function of M in physiology [10].

7. Optimal Solar Collector Temperature

There are many directions in which the method of thermodynamic optimization has been
extended and perfected, notably in the fields of cryogenics, heat transfer, storage systems,
power generation, refrigeration, and solar energy conversion. These and other directions
are illustrated in this book. Solar energy conversion has been studied under

sun T. - - - - - -

Q.

collector Tc

user
-.-.. ¢ : reversible
powerplanl

ambient

Figure 5. Solar power plant model with collector-ambient heat loss

and collector-power cycle heat exchanger [33].
 57

the banners of two fundamental problems. One is concerned with establishing the
theoretical limits of converting thermal radiation into work, or calculating the
exergy content of radiation. The other problem deals with the delivery of maximum
power from a solar collector of fixed size [33]. This problem has also been solved
in many subsequent applications [I, 9, 10, 34], which are united by an important
design feature: the collector operating temperature can be optimized.
This optimization opportunity is illustrated with the help of Fig. 5. A power
plant is driven by a solar collector with convective heat leak to the ambient. The
heat leak is assumed to be proportional to the collector-ambient temperature
difference, Q o == (UA)c(Tc - To)· The internal heat exchanger between the
collector and the hot end of the power cycle (the user) is modeled similarly,
Qo == (UAMTc - Tu)' There is an optimal coupling between the collector and the
power cycle such that the power output is maximum. This design is presented by
the optimal collector temperature [I, 33]
Tc,opt 9~;' + R9 max
- - == --"-=----""''''- (28)
To 1+ R
where R == (UA)c I (UA)j and 9 max == Tc,max ITo is the maximum (stagnation)
temperature of the collector. This optimum has its origin in the trade-off between
the Carnot efficiency of the reversible part of the power plant (1- TofT u) and the
heat loss to the ambient, Q o . The power output is the product Q(1 - To IT u)
When Tc < Tc,opt the Carnot factor is too small. When Tc > Tc,opt the heat inp~t Q
drawn from the collector is too small, because the heat loss to the ambient Q o is
large.
Corresponding optimal couplings have been identified in solar-driven power
plants of many power-cycle designs, extraterrestrial power plants, and refrigeration
systems [9].

8. Finite Time and Endoreversible: New Names for Old Ideas

The current state of thermodynamic optimization subject to size and time

constraints is reviewed at length in [9] and [34 - 36]. Additional examples are given
in this book in the chapter on Constructal Flow Geometry Optimization. The earlier
work was reviewed in [1 - 4] and [26, 27].
The published record is very clear. It shows that by the end of the 1970s the
method had been used by many independent schools in Europe and North America,
over the course of three decades, and in several fields: cryogenics, heat transfer,
power, refrigeration, storage systems, solar energy, and education. Steady-state and
time-dependent systems had been optimized. Size and time constraints had been
used. Optimization criteria had called for entropy generation minimization and
power output maximization in power plants, as well as power input minimization in
refrigeration plants. Endoreversible models had been used at the macroscopic
level (e.g., [16, 19]) and at the infinitesimal level (e.g., Fig. 1). The method had
become a self-standing graduate course with its own textbook [1].
On this rich background came in 1983 Andresen's surprising claim that "the
field of finite-time thermodynamics was started in 1975 by Professor R. Stephen
Berry, Peter Salamon and myself." This claim is false, not because Andresen et
58

et al.'s first paper was actually in 1977, not in 1975, and not because it was merely
a follow up to Curzon and Ahlborn 1975 paper [25], which it referenced. It is false
because of Andresen's own definition of the supposedly new field: "Finite-time
thermodynamics is the extension of traditional thermodynamics to deal with
processes which have explicit time or rate dependencies. These constraints, of
course, imply a certain amount of loss, or entropy production [37]." Andresen is
clearly referring to an already established method: the combined thermodynamics
& heat transfer optimization method illustrated here in section 3. My own way of
recognizing the existence of this field (this in 1982, before Andresen's claim), was
through the triangle diagram of Fig. 6.
Some of the work done by followers of Curzon and Ahlborn has been criticized
for specific technical reasons in the peer-refereed literature, in both physics [29, 34,
39] and engineering [28, 38]. One aspect of this type formed the subject of section
4 in this chapter. Endoreversible-the assumption that the generation of entropy is
located at the internal interfaces between the assumed building blocks of the
system-is similar to the old concept of local thermodynamic equilibrium [40]: the
foundation of all of our heat transfer and fluid mechanics. The local
thermodynamic equilibrium model is very clear in Fig. 1 (the shaded element),
which is why this model was and continues to be used routinely in the
thermodynamic optimization of heat transfer systems [1], either macroscopic (e.g.,
Fig. 5) or distributed infinitesimally (e.g., Fig. 1). The place for future discussions
of such aspects is the peer-refereed literature. These discussions, however, tend to
obscure the much more important issue, which is the purported novelty of the
method behind the work of Curzon and Ahlborn's followers.
It is also worth noting that none of this work represents "new physics." Each
result such as Eq. (9) and Eq. (14) is case specific. It corresponds specifically not
only to the type of system that is being modeled (e.g., insulation vs. power plant)

Entropy
generation
through
heat and flu id flow

Thermodynamics

Figure 6. The interdisciplinary field covered by thermodynamic optimization in constrained systems with
heat and fluid flow irreversibilities [1].
 59

but also the personal talent and taste for simplicity and beauty exhibited by the modeler.
All the physics that is needed continues to be covered by old thermodynamics, heat
transfer and fluid mechanics, combined. The models and the associated optimization
results occupy more and more the field of applications. This is the work that along
with Andresen et al.'s more recent contribution continues to fill the triangular area
depicted in Fig. 6.
New physics are the new principles that are waiting to be formulated in order to
cover in a predictive sense the natural phenomena that so far have evaded determinism.
One new principle is the constructallaw [10] of geometric shape and structure in natural
systems far from internal equilibrium. This direction is outlined in this book in the
chapter on Constructal Flow Geometry Optimization. This new extension of
thermodynamics covers the physics behind concepts such as optimization, purpose,
necessity (the fittest), time direction and fractals, and unites physics with engineering
and biology.

Acknowledgement. This work was supported by the National Science Foundation

and the Air Force Office of Scientific Research.

References
1. Bejan, A. (1982) Entropy Generation through Heat and Fluid Flow, Wiley, New York.
2. Moran, M. J. (1982) Availability Analysis: A Guide to Efficient Energy Use, Prentice-Hall,
Englewood Cliffs.
3. Feidt, M. (1987) Thermodynamique et Optimisation Energetique des Systemes et Procedes,
Technique et Documentation, Lavoisier, Paris.
4. Bejan, A. (1988) Advanced Engineering Thermodynamics, Wiley, New York.
5. Sieniutycz, S. and Salamon, P., eds. (1990) Finite-Time Thermodynamic.f and Thermoeconomics,
Taylor and Francis, New York.
6. De Vos, A. (1992) Endoreversible Thermodynamics o/Solar Energy Conversion, Oxford University
Press, Oxford.
7. Radcenco, V. (1994) Generalized Thermodynamics, Editura Tehnica, Bucharest.
8. Bejan, A., Tsatsaronis, G. and Moran, M. (1996) Thermal Design and Optimization, Wiley, New
York.
9. Bejan, A. (1996) Entropy Generation Minimization, CRC Press, Boca Raton.
10. Bejan, A. (1997) Advanced Engineering Thermodynamics, second ed., Wiley, New York.
11. Bejan, A. and Smith, J. L., Jr. (1974) Thermodynamic optimization of mechanical supports for
cryogenic apparatus, Cryogenics 14, 158-163.
12. McClintock, F. A. (1951) The design of heat exchangers for minimum irreversibility, Paper no. 51-
A-108, Winter Annual Meeting, ASME.
13. Grassmann, P. and Kopp. J. (1957) Zur gunstigsten Wahl der Temperaturdifferenz und der
Wiinneiibergangszahl in Wiinneaustauschem Kiiltetechnik 9, 306-308.
14. Martynovskii, V. S., Cheilyakh, V. T. and Shnaid, T. N. (1971) Thermodynamic effectiveness of a
cooled shield in vacuum low-temperature insulation, Energ. Transp. 2; also in Ahem, J. E. (1980)
The Exergy Method of Energy Systems Analysis, Wiley, New York.
15. Miiser, H. (1957) Thermodynamische Behandlung von Electronenprozessen in Halbleiter-
Randschichten, Z Phys. 148, 380-390.
16. Chambadal, P. (1957) Les Centrales Nucieaires, Armand Colin, Paris.
60

17. Chambada1, P. (1958) Le choix du cycle thermique dans une usine generatrice nucleaire, Revue
Generale de I'Electricite 67, 332 - 345.
18. Chambadal, P. (1963) Evolution et Applications du Concept d'Entropie, section 30, Dunod, Paris.
19. Novikov, I. I. (1957) The efficiency of atomic power stations, Atomnaya Energiya 3,409; English
translation in (1958) J. Nuclear Energy 117, 125 - 128.
20. Vukalovich, M. P. and Novikov, I. I. (1972) Thermodynamics, Mashinostroenie, Moscow.
21. Novikov, I. I. and Voskresenskii, K. D. (1977) Thermodynamics and Heat Tran.~fer, Atomizdat,
Moscow.
22. Novikov,l. I. (1984) Thermodynamics, Mashinostroenie, Moscow.
23. EI-Wakil, M. M. (1962) Nuclear Power Engineering, McGraw-Hili, New York.
24. EI-Wakil, M. M. (1971) Nuclear Energy Conversion, International Textbook Company, Scranton,
PA.
25. Curzon, F. L. and Ahlborn, B. (1975) Efficiency of a Carnot engine at maximum power output,
American Journal (!fPhysics 43,22-24.
26. Bejan, A. (1994) Engineering advances on finite time thermodynamics, American Journal of
Physics 62, January, 11-12.
27. Bejan, A. (1996) Notes on the history of the method of entropy generation minimization (finite time
thermodynamics), Journal (!fNon-Equilibrium Thermodynamics 21,239-242.
28. Gyftopoulos, E. P. (1997) Fundamentals of analyses of processes, Energy Conversion and
Management38,1525-1533.
29. Bejan, A. (1996) Models of power plants that generate minimum entropy while operating at
maximum power, American Journal (!fPhysics 64, 1054-1059.
30. Ikegami, Y. and Bejan, A. (1998) On the thermodynamic optimization of power plants with heat
transfer and fluid flow irreversibilities, Journal of Solar Energy Engineering 120,139-144.
31. Bejan, A. and Errera, M. R. (1998) Maximum power from a hot stream, International Journal of
Heat and Mass Tran.ifer 41, 2025-2036.
32. Bejan, A. (1997) Theory of organization in nature: pulsating physiological processes, International
Journal (if Heat and Mass Tran.~fer 40,2097-2104.
33. Bejan, A., Kearney, D. W. and Kreith, F. (1981) Second law analysis and synthesis of solar
collector systems, Journal (!f Solar Energy Engineering 103, 23-30.
34. Bejan, A. (1996) Entropy generation minimization: the new thermodynamics of finite-size devices
and finite-time processes, Journal (!fApplied Physics 79, 1191-1218.
35. Sieniutycz, S. and Shiner, J. S. (1994) Thermodynamics of irreversible processes and its relation to
chemical engineering: second law analyses and finite time thermodynamics. Journal of Non-
Equilibrium Thermodynamics 19, 303-348.
36. Bejan, A. (1996) Method of entropy generation minimization, or modeling and optimization based
on combined heat transfer and thermodynamics, Revue Generale de Thermique 35, 637-646.
37. Andresen, B. (1983) Finite time thermodynamics, Physics Laboratory II, University of Copenhagen,
p.3.
38. Moran, M. J. (1998) On second-law analysis and the failed promise 0 finite-time thermodynamics,
Energy 23,517-519.
39. Sekulic, D. P. (1998) A fallacious argument in the finite time thermodynamics concept of
endoreversibility, Journal (!fApplied Phyiscs 83,4561-4565.
40. Berg, C. A. (1992) Conceptual issues in energy efficiency, in Valero, A. and Tsatsaronis, G., eds.,
Proceedings of the International Symposium on Efficiency, Costs, Optimization and Simulation of
Energy Systems (ECOS'92), Zaragoza, ASME, New York, pp. 7-16.
CONSTRUCT AL FLOW GEOMETRY OPTIMIZATION

A.BEJAN
Duke University
Department of Mechanical Engineering
and Materials Science
Durham, NC 27708-0300, USA

1. Optimal Size

In the field of heat transfer, the method of thermodynamic optimization or entropy

generation minimization (EGM) brings out the inherent competition between heat-
transfer and fluid-flow irreversibilities in the optimization of devices subjected to overall
constraints. Consider the most elementary heat exchanger passage [1] which is
represented by a duct of arbitrary cross-section (A) and arbitrary wetted perimeter (p).
The engineering function of the passage is specified in terms of the heat transfer rate per
unit length (q') that is to be transmitted to the stream (m); that is, both q' and m are
fixed. In the steady state, the heat transfer q' crosses the temperature gap LlT formed
between the wall temperature (T + LlT) and the bulk temperature of the stream (T). The
stream flows with friction in the x direction; hence, the pressure gradient (-dP/dx) > O.
Taking as thermodynamic system a passage of length dx, the first law and the second
law state that

mdh = q'dx and S· -' ds - q' >0 (1)

gen - m dx T + LlT - ,

where S'gen is the entropy generation rate per unit length. Combining these statements
with the assumption that the fluid is single phase (dh = T ds + v dP), we obtain

S' = q'LlT + ~(_ dP) = q'LlT + ~(_ dP) > 0 (2)

gen P(l + LlT IT) pT dx - T2 pT dx-

The denominator of the first term on the right-hand side has been simplified by
assuming that the local temperature difference LlT is negligible when compared with the
local absolute temperature T. The heat exchanger passage is a site for both flow with
friction and heat transfer across a finite LlT; the S~en exp~ession th~s has two terms,
each accounting for one irreve~sibility mechanism, S'gen = S'gen"H + S'gen,M"
A remarkable feature of the S'gen expression (2) and of many like it for other simple
devices is that a proposed design change (for example, making the passage narrower)
induces changes of opposite signs in the two terms of the expression. An optimal trade-
61
A. Bejan and E. Mamut (eds.), Thermodynamic Optimization o/Compiex Energy Systems, 61-72.
@ 1999 Kluwer Academic Publishers.
62

off then exists between the two irreversibility contributions, an optimal design for
which the overall measure of exergy destruction (S~en) is minimal while the system
continues to serve its specified function (q', rit). This trade-off becomes visible
analytically when heat-transfer and fluid-friction information [2] is added to Eq. (2), Le.,
when ~T and (-dP/dx) are replaced by the Nusselt number and friction factor
expressions that characterize the duct. For example, it was shown [1] that when the
cross-section is round and the flow is fully developed and turbulent, Eq. (2) can be
arranged as

-08 4.8
S~en
Ns = ....,.-,-==--- 0.856 ( ~ ) + 0.144 ~
( )
(3)
Sgen,min Re opt Re opt

where the ratio on the left-hand side is known as the entropy generation number Ns [1],
and Re = 4m/(1tJ.ID). Equation (3) shows that S~en has a sharp minimum with
respect with the size of the duct, D.
The literature contains a growing number of internal flow configurations that have
been subjected to the same type of thermodynamic optimization: reviews can be found
in [3,4]. External flow configurations are governed by the same competition between
heat-transfer and fluid-flow irreversibilities. The overall entropy generation rate
associated with external flow is

(4)

where QB,TB,Too,FD,andU oo are, respectively, the instantaneous heat transfer rate

between the body and the fluid reservoir, the body surface temperature, the fluid reservoir
temperature, the drag force, and the relative speed between body and reservoir. This
remarkably simple result proves again that poor thermal contact (the first term) and fluid
friction (the second term) contribute hand in hand to degrading the thermodynamic
performance of the external convection arrangement. One area in which Eq. (4) has
found application is the problem of selecting the size and number (density) of fins for
the design of extended surfaces [1].
Most of the literature on the thermodynamic optimization of heat transfer devices is
devoted to more complex systems that incorporate the elemental features (internal,
external) mentioned in this section. Examples include heat exchangers of several
configurations (counterflow, parallel flow, cross-flow) with single-phase or two-phase
flow [3,4].

2. Optimal Internal Spacings

Even simpler is the sizing of a system in which a single transport mechanism causes
the irreversibility, for example, heat transfer. When the heat current is imposed,
minimizing the entropy generation means minimizing the resistance to heat flow.
For example, in the cooling of electronic packages, both the volume and the heat
generation rate that is distributed uniformly over that volume are fixed. The heat current
 63

is removed by a single-phase stream, with natural or forced convection. The geometric

arrangement of components can then be optimized such that the hot-spot temperature is
minimal. The plate-to-plate spacing (or number of columns) is free to vary. If the
spacing is too large, however, there is not enough heat-transfer area and the hot-spot
temperature becomes high; when the spacing is too small, the coolant flow rate
decreases and the hot-spot temperature is again high. Between the two lies an optimal
spacing-an optimal package architecture-that minimizes the thermal resistance
between the system and the environment. This method of geometric optimization of
internal architecture is known as the intersection of the asymptotes method [5, 6].
In this section we illustrate this principle for natural convection cooling, Fig. 1.
We assume that the flow is laminar, the board surfaces are sufficiently smooth to justify
the use of heat transfer results for natural convection over vertical smooth walls, and the
maximum (hot spot) temperature Tmax is the scale that represents the temperature at
every point on the board surface. The method consists of two steps. In the first, the
two extremes are identified: the total heat transfer rate (Q) expressions in the small-D
limit and the large-D limit. In the second step the two asymptotes are intersected for the
purpose of locating the D value that maximizes Q.
In the D ~ 0 limit we can use with confidence Q1 = m)c p (T max - T ~) for the
heat transfer rate extracted by the coolant from one of the channels of spacing D. Note
that in each channel T~ is the inlet temperature, Tmax is the outlet temperature, and m)
is the mass flow rate through a single channel. To calculate m) we rely on the classical
solution for Hagen-Poiseuille chimney flow between two vertical plates [5]

(5)

(6)

where W is the third dimension of the volume, and U is the vertical velocity averaged
across the D-wide channel. The fluid is modeled as Boussinesq incompressible with the

I- -I

r
L

D
g~
H ......

1 t t t t t t t t t
T
-
Figure 1. Fixed volume with a variable number of vertical plates cooled by natural convection [5, 6].
64

density p, kinematic viscosity v, and coefficient of volumetric thermal expansion ~.

The remaining parameters are defined in Fig. 1. The total rate of heat transfer removed
from the package is 0 = n 01' where n =LID, hence

(7)

Consider next the opposite limit in which D is large enough so that it exceeds the
thickness of the thermal boundary layer that forms on each vertical surface [5], namely
D > H[g~H3(Tmax - Too)/(av)]-1/4 where a is the fluid thermal diffusivity, and the
Prandtl number is of order 1 or greater. In this limit the center region of the board-to-
board spacing is occupied by fluid of temperature Too. The number of distinct boundary
layers is 2n = 2L1D. The heat transfer rate through one boundary layer is
hHW(T max - Too) for which the average h~t transfer coefficient h is furnished by the
correlation for laminar flow [2, 5], namely hH 1 k = O. 517Ra~4, where RaH = g~(Tmax
- Too )H 3/(av). The total rate of heat transfer extracted from the entire package is 2n
times larger than hHW(T max - Too),

0= 2(L/D)HW(T max - Too)(k/H)0.517Ra~4 (8)

Figure 2 shows that the two asymptotes (7, 8) intersect above what would be the
peak of the actual 0 (D) curve. It is not necessary to know the actual O(D) relation.
The optimal spacing Dopt for maximum 0 is approximately

-1/4
D opt IH == 2.3 [g~(Tmax - Too)H3/(av) ] (9)

This estimate reproduced within 20 percent the optimal spacing deduced based on more
exact and lengthier methods, such as the maximization of the O(D) relation, or the

\ i"
natural convection
. .
the small-D limit: Q- D2

\J
i" "\
,.,.
. / " ", ' ' / the Iarge-D limit:
.....
Q _ D- 1
i

--'-0_..
o ~ ____J -______________________ ~

o
D

Figure 2. The maximization of the total heat transfer rate removed by natural convection
from a fixed volume [5, 6].
 65

finite-difference simulations of the complete flow and temperature fields in the package.
An order of magnitude estimate for the maximum heat transfer rate can be obtained by
substituting D opt in Eqs. (7) and (8): Qrnax ~ 0.45k(Trpax - Tca)(LW IH)RaU2.
The unequal sign is a reminder that the peak of the actual Q(D) curve falls under the
intersection of the two asymptotes (Fig. 2). The intersection of asymptotes method
has been used to determine the optimal spacing of elemental features of several other
shapes [5]. One example is the spacing between horizontal cylinders in a fixed volume
(for example, tubes in a compact heat exchanger).
A completely analogous geometric optimization rules the internal spacings of
volumes bathed by forced convection. This problem was solved more recently [7], and
for the sake of conciseness we review only its main result. The analysis can also be
found in [5, 6]. We assume that the coolant is forced to flow by the pressure difference
I1p, and that the plate length in the direction of flow is L. The intersection of the small-
D and large-D asymptotes yields the optimal spacing formula

(to)

This result agrees within 20 percent with the value obtained by locating the maximum
of the actual Q(D) curve, and is adequate when the board surface can be modeled as
isothermal. For uniform-flux surfaces, the 2.7 factor is replaced by 3.2 on the right side
of formula (to).
Compare (10) with (9) and notice the symmetry between the non dimensional
optimal spacings for forced convection and natural convection. Note further that the role
that in the natural convection formula is played by RaH, in forced convection is played
by the new dimensionless pressure drop number [5] defined as IlL = I1p. V I (Ila.).
This and a related dimensionless group were also identified in external forced convection,
thermodynamic optimization, and electronic cooling, and named Bejan number, Be [8].
If the stack is exposed to (surrounded by) a flow approaching with the velocity U ca
parallel to L, we may approximate I1p with (1I2)pU2 in (10) to estimate the optimal
spacing D opt I L= 3.2Pr-1/4 Re LI12 , where ReL = UooLlv. This estimate is supported
by numerical studies of stacks with several types of distributions of heat generation on
the board surfaces (uniform, flush mounted sources, protruding sources, etc.) [9]. The
same method was applied recently to determine the optimal spacing between cylinders in
cross-flow in a fixed volume [5], and to predict the formation of regularly placed cracks
in solids that shrink upon cooling or drying [to].

3. Volume-to-Point Flow, and Constructal Tree Networks

The research frontier in heat exchangers and the cooling of electronics is being pushed in
the direction of smaller and smaller package dimensions. There comes a point where
miniaturization makes convection cooling impractical, because the ducts through which
the coolant must flow take up too much space. The only way to channel the generated
heat out of the electronic package is by conduction. This conduction path will have to
be very effective (of high thermal conductivity, kp), so that the temperature difference
between the hot spot (the heart of the package) and the heat sink (on the side of the
66

package) will not exceed a certain value. Conduction paths also take up space. Designs
with fewer and smaller paths are better suited for the miniaturization evolution. The
fundamental problem that was proposed in [II] is: "Consider a finite-size volume in
which heat is being generated at every point and which is cooled through a small patch
(heat sink) located on its boundary. A finite amount of high conductivity (kp) material
is available. Determine the optimal distribution of kp material through the given
volume such that the highest temperature is minimized."
The discovery made in [II] is purely geometric: every portion of the given volume
can have its shape optimized, such that its resistance to heat flow is minimal. This
design principle applies at any volume scale, and to other forms of transport (fluid,
electric charge, mass species). The volume-to-point path was determined in a sequence
of steps consisting of shape optimization and subsequent construction (assembly,
grouping). Very important is the time arrow of this construction, which points from
small to large. It starts from the smallest building block (elemental system), and
proceeds toward larger building blocks (assemblies, constructs). It was shown that
determinism vanishes if the direction is reversed, from large to small. To emphasize the
link between determinism and the direction from small to large, and as a reminder that
theory runs against fractal thinking, the geometric optimization principle proposed in
[11] and summarized in [12] was named "constructal theory."
Consider a volume (V) that generates heat volumetrically at the uniform rate q''',
which is collected over the volume and led to a heat sink on its boundary. For
simplicity, assume that the heat-flow geometry is two-dimensional over the area A, .
hence V = AW where W is the third dimension of the volume. Most of A is occupied
by the material of low conductivity ko, which generates q'''. A small fraction of A
(namely Ap) is to be covered by high-conductivity material.
We regard A as a collection of subsystems of various sizes (Ao, AI, A 2, ... ), and
begin with the smallest such system (Ao), which is shown in Fig. 3. The defining
feature of this "elemental system" is that its heat current (qo = q"'AoW) is collected by a
single kp blade. The other important feature is that the size Ao is known and fixed
(dictated by manufacturing constraints) while the shape HolLo may vary. The objective
of the following analysis is to determine the optimal shape such that the overall
resistance between Ao and the heat sink [the Do patch 10cated at (0,0)] i~ minimal.
An analytical solution is possible in the limit k » 1, where k = kp/ko is a
known constant of the conducting composite. Symmetry and the requirement to
minimize resistance are why the Do-thin blade is p~aced on the longer of the two axes of
symmetry of Ao (note Ho > La in Fig. 3). When k » 1, heat flows vertically through
the ko material and horizontally through the kp material. This allows us to develop the
expression for the hot-spot temperature T(Lo, ± Hol2) as a sum of two terms: the
temperature drop through ko from (Lo, Hol2) to (Lo,O), plus the temperature drop
through kp from (Lo,O) to (0,0). The result is [11, 12]:

(11)

Next, we note that the ratio DoIHo is another constant of the conductive composite,
because it accounts for the proportion in which kp material is allocated to Ao, namely
 67

ApolAo = DoLo/(HoLo). We call this constant <1>0 = DoIHo «1. With this, Eq. (11)
shows clearly that the overall volume-to-point resistance can be minimized with respect
to the shape of the elemental system (HolLo):

_ )-1/2
(Ho I LO)OPI = 2 (k<l>o
1
<1 ( ATO ko) _
= ----;-;'")1/2 (12, 13)
2 (k<l>o
q'" A o mID
.

An important feature of (12, 13) is that the resistance through the ko material matches
the resistance through the kp material: the two terms have the same value on the right-
hand side of Eq. (13). This principle of equipartition of resistance applies at every
volume scale.
As we proceed toward larger scales, we can only assemble elemental volumes of
size Ao, which have to be connected so that their heat currents can be collected into a
single stream, Fig. 4. This first construct (AI) contains nl elemental volumes, AI =
nlAo. The number nl (or the shape HIIL I) is a degree of freedom. The elemental heat
currents are collected by a new blade of kp material and thickness D I.
When nl is large, the first construct conducts heat in almost the same way as the
elemental volume. The only difference is that the ko regions of Fig. 4 are now replaced
by a composite material. The effective thermal conductivity of this composite in the
vertical direction is kl = kpDo/Ho. The hot-spot temperature difference ATI occurs
between the far corner (LI' H1/2) and the heat sink (0,0). Nothing has changed, except
the notation in Eq. (11) which now reads:

(14)

The geometric optimization too is a repeat of what went on at the elemental level.
First, it is convenient to rewrite Eq. (14) in a form that shows the effect of n),

ko' q'" Do
~u

i
0 x
Lo
Lk
p ko' q'"

Figure 3. Elemental volume with volumetric heat generation and one high-conductivity path along its
axis of symmetry [11, 12].
68

(15)

where DI = DI / A~/2 . This expression shows how the overall volume-to-point

resistance of the first construct depends on nl: Resistance can be minimized through
optimal growth,

__ )1/2
( kDI
nl,opt = -"--(-_"":""'-)"-;1/";;'"8 » ( ATlkO) (16, 17)
2114 k~o
I - qA'"
I mm
= (--
2 3/4 kDl
)112(_k~o )3/8

Beginning with the first construct, the overall resistance can be minimized for a
second time..: When the total kp volume is constrained, ApI = DILl + nlA po , the
parameters DI and ~o cannot be increased independently. Writing ~1 = ApI/AI (=
constant), and noting that Apo =_ ~oAo and HI =2L I,opt, we fi~d that the _ApI constraint
provides a relation between DI and ~o, namely ~I = DI / [2 1/2 (k~O)1/4] + ~o·
~inimizing _(17) subject to the ~ 0 we find, in order, ~o,opt = ~I /2,
D I,opt = 2-3/4kI/4",5/4
'1'1 ' and

_ )112 1
nl,opt = ( k~1 /2 » I ( ATlkO) = (18, 19)
q'" A I mm k~1

where the subscript ( )mm indicates the result of double minimization. Combining
DI,oPt with Do =~oHo,opt, and using the Ho,opt result listed in the preceding section, we
fi!ld the optimal ratio between the thicknesses of the kp blades: (DI/Do)opt =
(k ~ 112) 112. The unexpected geometric feature that follows from this double
optimization is that the external aspect ratio is exactly equal to the integer 2, namely

q'" DI Do"
I
1+ t L
~J
f !,oPt~
-..x
0
Lk
p q'" i Ll

-Ho/2

Figure 4. The first construct: a large number of elemental volumes (Fig. 3) connected to a central high-
conductivity path [II, 12].
 69

(HI 1 LI)O t = 2; in other words, it is independe'!,t of the design parameters of the

conductiv6 composite built into the first assembly (k, «I> I)'
To fill an even larger volume, we connect a large number (n2 = A 2/A I ) of first
constructs and arrive at the second construct shown in Fig. 5. The question, again, is
how many building blocks to assemble into a second construct, so that we continue on
the path that we started-a path where resistance is minimized geometrically at every
step. The analysis begins with the analog of Eq. (15), which in the case of the overall
thermal resistance of the second construct is an expression that can be rearranged as
follows,

(20)

where D2 = D21 AII2. The temperature difference AT2 is measured between the hot-
spot (the upper right corner in Fig. 5) and the heat sink (the left end of the D2 blade).
The minimization of (20) with respect to n2 yields

_ )112
n2,opt = 23/4 (D2 1 «1>2 (21,22)

The second optimization consists of balancing «1>1 against D2 while taking into
account the kp-material constraint of the second c~mstruct, Ap2 =D2L2 + n2ApJ, which,
after we define «1>2 = A p2/A 2 , becomes «1>2 =} -112 D2 + «1>1' The results of minimizing
(22) for a second time are «I>l,opt = «1>2 12, D 2,opt = 2 -1/2 «1>2' and

2
n2,opt = 2 ( AT2 ko) (23,24)
q'" A 2 mm

as well as (H21 L 2)opt = I and (D21 DI)opt = 2. The conclusion that n2,opt = 2 is
important because it is the first time that dichotomy (bifurcation, pairing) has been
predicted. The best second construct is the one that contains the smallest number of

q'" DI
..

q'"

Figure 5. Second construct containing a large number of first constructs [II, 12].
70

first constructs, and that number is 2. Contrary to the present theory, dichotomy is
assumed at the start of tree network analyses, for example, in physiology [13]. In
constructal theory the integer 2 is a result, not an assumption.
This single principle of geometric optimization and subsequent construction
(growth) is repeated toward stepwise larger volume scales. The end result is shape and
structure-the optimized architecture-of the composite (kolkp) that connects the sink
point to the finite-size volume. The infinity of points of the given volume is
"connected" to the sink because at the smallest volume scale the transport is volumetric,
by thermal diffusion through the low-conductivity material. At larger scales the
transport is via channels (streams) of high conductivity. Diffusion comes first and
streams later.
Astonishing is not that the high-conductivity channels form a tree (a network
without loops) but that each feature of the tree is deterministic, the result of a single
principle of optimization. This conclusion runs against the currently accepted doctrine
that natural structures are nondeterministic, the result of chance and necessity. The
discovery then is not the tree but the principle that generates this structure.

4. Shape and Structure in Natural Flow Systems

The hand-in-glove geometry formed by slow (disorganized) and fast (organized) flow
regimes unites all the optimized volume-to-point flows. The slow regime is the
"glove," and the fast regime is the "hand." In the flow of oxygen through a mammal,
the slow, shapeless flow is volumetric mass diffusion through the tissues, while the
faster regime is stream flow through blood vessels and bronchial passages. The length
scale of the smallest volume element from which constructs begin is known. It is the
distance to which oxygen diffuses into the tissue during one breathing time interval or
heart beat.
The same hand-in-glove geometry characterizes all turbulent flows. Viscous
diffusion in the smallest volume elements is accompanied by the macroscopic structure
of faster streams known empirically as "eddies." The length scale of the smallest
volume element (or the smallest eddy) is known from the distance of viscous diffusion
during the first roll [5]. Pure diffusion is replaced by the first eddy, in other words, by
diffusion combined with the first stream, when the stream motor can advance the mixing
process farther and faster than diffusion alone. The evolution of constructs of stepwise
larger sizes can be observed in time by looking at any shear layer, jet or plume. The
oldest structure (shape) in the field of vision is the smallest eddy, which is swept
intermittently from the laminar tip section of the flow. The youngest structure is the
largest eddy, which is a construct of smaller constructs of all the preceding sizes.
Most natural tree networks accommodate the flow of fluid or electricity. In fluid
trees the small-scale volumetric flow is by slow diffusion (for example, Darcy flow in
the wet banks of the smallest rivulet), while the larger-scale flow is organized into faster
conduits [12]. In the electric tree visible during lightning, new features are three-
dimensionality and the unsteady (one shot) nature of the flow. Additional electric tree
networks are found in living systems, for example, the neural dendrites of the nervous
system. Street patterns and urban growth have been predicted based on a kinematics
analog of the resistance minimization shown in this chapter. That analog is the
 71

minimization of the travel time between one point and a finite-size area (an infinity of
points) [12].
The resistance minimization principle is raised to the rank of law by the structure of
all the tree networks anticipated by constructal theory. This law can be summarized as
follows [11, 12]: "For a finite-size system to persist in time (to live), it must evolve in
such a way that it provides easier access to the imposed currents that flow through it."
This statement has two parts. First, it recognizes the natural tendency of imposed
currents to construct shapes, i.e. paths of optimal access through constrained open
systems. The second part accounts for the improvements of these paths, which occurs
in an identifiable direction that can be aligned with time itself.
This formulation of the principle refers to a system with imposed steady flow, as in
the heat flow examples discussed in this chapter. If the system discharges itself to one
point in unsteady fashion, then the constructal minimization of volume-to-point
resistance is equivalent to the minimization of the time of discharge, or the
maximization of the speed of approach to internal eqUilibrium (uniformity, zero flow).
If the volume is unbounded, the constructs compound themselves and continue to spread
indefinitely. Complexity continues to increase in time because "growth" is the path to
resistance minimization. Examples are the jet in a fluid reservoir, and the dendritic
crystal that grows into a subcooled liquid [12].
All the structured phenomena mentioned at various stages in this chapter, including
the round cross-section of the blood vessel and the proportionality between river width
and depth, can be predicted based on constructal theory [12]. The theoretical heat-tree
networks [11] were subjected to numerical tests and refinements in [14]. Constructal
trees for fluid flow through a heterogeneous porous medium were developed in [15, 16].
Analogous trees of streets that provide minimum-time travel between an area and one
point are described in [17-19]. Constructal theory also accounts for the onset, structure
and heat transfer rate of turbulent flow [12, 20] and Benard convection [21], and for the
structure in cracked solids that are cooled volumetrically [10]. Three-dimensional
constructal networks are presented in [16, 18, 22]. The time-dependent discharge
between a volume and one point-the lightning tree problem-was added to constructal
theory in [23]. The deterministic development in time of a river drainage basin is
presented in [24]. Reviews of the applications of constructal theory are provided in [12,
17, 25].

Acknowledgment. This work was supported by the National Science Foundation.

References

I. Bejan, A. (1982) Entropy Generation through Heat and Fluid Flow, Wiley, New York.
2. Bejan, A. (1993) Heat Transfer, Wiley, New York.
3. Bejan, A. (1996) Entropy Generation Minimization, CRC Press, Boca Raton, FL.
4. Bejan, A. (1996) Entropy generation minimization: the new thermodynamics of finite-size devices
and finite-time processes, Journal (~"Applied Physics 79, 1191-1218.
5. Bejan, A. (1995) Convection Heat Transfer, second edition, Wiley, New York.
6. Bejan, A. , Tsatsaronis, G. and Moran, M. (1996) Thermal Design and Optimization, Wiley, New
York.
72

7. Bejan, A. and Sciubba, E. (1992) The optimal spacing of parallel plates cooled by forced
convection, International Journal of Heat and Mass Transfer 35, 3259-3264.
8. Petrescu, S. (1994) Comments 0 the optimal spacing of parallel plates cooled by forced convection,
International Journal (!f Heat and Mass Transfer 37, 1283.
9. Bejan, A. (1996) Geometric optimization of cooling techniques, chapter I in Kim, S. J. and Lee, S.
W., eds., Air Cooling Technology.tllr Electronic Equipment, CRC Press, Boca Raton, FL.
10. Bejan, A. Ikegami, Y. and Ledezma, G. A. (1997) Constructal theory of crack pattern formation
for fastest cooling, International Journal (It Heat and Mass Transfer 41, 1945-1954.
II. Bejan, A. (1997) Constructal-theory network of conducting paths for cooling a heat generating
volume, International Journal (if Heat and Mass Tran~fer 40,799-816.
12. Bejan, A. (1997) Advanced Engineering Thermodynamics, second edition, Wiley, New York.
13. Thompson, D'A. W. (1942) On Growth and Form, Cambridge University Press, Cambridge, UK.
14. Ledezma, G. A., Bejan, A. and Errera, M. R. (1997) Constructal tree networks for heat transfer,
Journal of Applied Physics 82, 89-100.
15. Bejan, A. and Errera, M. R. (1997) Deterministic tree networks for fluid flow: geometry for
minimum flow resistance between a volume and one point, Fractals 5, 685-695.
16. Bejan, A. (1997) Constructal tree network for fluid flow between a finite-size volume and one
source or sink, Revue Generale de Thermique 36, 592-604.
17. Bejan, A. (1996) Street network theory of organization in Nature, Journal of Advanced
Transportation 30, 85-107.
18. Bejan, A. and Ledezma, G. A. (1998) Streets tree networks and urban growth: optimal geometry
for quickest access between a finite-size volume and one point, Physica A 255, 211-217.
19. Bejan, A. and Tondeur, D. (1998) Equipartition, optimal allocation, and the constructal approach to
predicting organization in nature, Revue Generale de Thermique 37, 165-180.
20. Bejan, A. (1998) Questions in fluid mechanics: natural tree-shaped flows, Journal of Fluids
Engineering 120, 429-430.
21 Nelson, R. A., Jr. and Bejan, A. (1998) Constructal optimization of internal flow geometry in
convection, Journal (if Heat Tran.fter 120, 357-364.
22. Ledezma, G. A. and Bejan, A. (1998) ConstClJctal three-dimensional trees for conduction between
a volume and one point, Journal (dReat Transfer 120, 977-984.
23. Dan, N. and Bejan, A. (1998) Constructal tree networks for the time-dependent discharge of a
finite-size volume to one point, Journal (!f Applied Physics, 84, 3042-3050.
24. Errera, M. R. and Bejan, A. (1998) Deterministic tree networks for river drainage basins, Fractals
6,245-261.
25. Bejan, A. (1998) Constructal optimization of paths for heat transfer, in Tupholme, G. E. and Wood,
A. S., eds., Mathematics (!f Heat Tran.~ter, Clarendon Press, Oxford, UK.
FUNDAMENTALS OF EXERGY ANALYSIS AND EXERGY-AIDED
THERMAL SYSTEMS DESIGN

M.J. MORAN
The OhiD State University
Department of Mechanical Engineering
Columbus, OH 43210, USA

1• Introduction

A close reading of the early literature of the second law of thermodynamics originating
from the time of Carnot shows that second law reasoning in today's engineering sense - -
optimal conversion of inputs to end use - - was present from the beginning. Interest in
this theme occasionally flickered in the interim, but it was kept alive and now forms the
basis for contemporary applications of engineering thermodynamics. The bonds that have
been forged in recent years between thermodynamics and economics have made the theme
of optimal use more explicit and have allowed second law methods to be even more
powerful and effective than ever before for the design and analysis of thermal systems.
The development of second law analysis owes considerably to the efforts of a great
many towering figures in the history of science whose contributions have advanced our
present understanding immeasurably. Among these are Carnot, Clausius, Gibbs, Kelvin,
and Planck. Significant contributions to engineering thermodynamics have been provided
by the likes of Stodola, Bosnjakovic, Keenan, and Obert. Still, the bulk of the literature
of second law analysis in engineering has appeared in the last quarter-century, and most of
those who have contributed remain active somewhere over the globe, including Szargut,
Gyftopoulos, Gaggioli, EI-Sayed, Valero, Tsatsaronis, Bejan, and many more.
Today, second law analysis has several offshoots, including exergy analysis,
thermoeconomics, entropy-generation minimization, and finite-time thermodynamics
(or endoreversible thermodynamics). The first three are well-embedded in contemporary
thermodynamics. Many experts attest to their pedagogical value. These methods are
increasingly being used in engineering practice and even finding their way into handbooks
[1]. As they are now familiar, the mere use of such methods is no longer sufficiently
novel to stir interest; rather, it is the results elicited from them that have become the
focus. In short, these methods have achieved a level of maturity, and the spotlight is now
on what can be achieved with them [2]. Some additional evolution surely will occur, but
as we begin a new millennium we can regard these methods as established, and expect
them to be used increasingly for the design and analysis of thermal systems. By
comparison, finite-time (endoreversible) thermodynamics faces an uncertain future,
stemming from fundamental flaws, over-reliance on highly simplified models and lack of
engagement with real-world considerations [3].
The present discussion provides an introduction to thermal system design from a
second law perspective featuring exergy. 'Since exergy measures the true thermodynamic
values of the work, heat, and other interactions between a system and its surroundings as
well as the effect of irreversibilities within the system, exergy is also a rational basis for
assigning costs. This is called exergy costing. Although the importance of considering
73
A. Bejan and E. Mamllt (eds.), ThermodynDmic Optimization o/Compla Energy Systems, 73-92.
© 1999 K1IIWer Academic Publishers.
 74

costs is emphasized throughout the current presentation, an indepth discussion of exergy

costing is not provided, and readers are referred to [2].
Essential background material concerning engineering thermodynamics is reviewed
in Sec. 2. Various aspects of exergy analysis are then surveyed in Sec. 3. Guidelines for
improving thermodynamics effectiveness are listed in Sec. 4. The role of exergy analysis
in thermal system design is considered in Sec. 5. Closing comments are provided in Sec.
6. References are listed in Sec. 7.

2. Engineering Thermodynamics Background

Engineering applications involving exergy are generally analyzed on a control volume

basis. Accordingly, the control volume formulations of the mass, energy, and entropy
balances presented in this section play important roles. These are provided here in the
form of overall balances assuming one-dimensional flow. Equations of change for mass,
energy, and entropy in the form of differential equations are also available in the literature
[4].
For applications in which inward and outward flows occur, each through one or
more ports, the extensive property balance expressing the conservation of mass principle
takes the form

(la)

where dm/dt represents the time rate of change of mass contained within the control
volume, mi denotes the mass flow rate at an inlet port, and me denotes the mass flow
rate at an exi t port.
Energy is a fundamental concept of thermodynamics and one of the most
significant aspects of engineering analysis. Energy can be stored within systems in
various macroscopic forms: kinetic energy, gravitational potential energy, and internal
energy. Energy can also be transformed from one form to another and transferred between
systems. Energy can be transferred by work, by heat transfer, and by flowing matter.
The total amount of energy is conserved in all transformations and transfers. The
extensive property balance expressing the conservation of energy principle (first law)
takes the form

(2a)

where U, KE and PE denote, respectively, the internal energy, kinetic energy, and
gravitational potential energy of the overall control volume, respectively. The right side
of Eq. (2a) accounts for transfers of energy across the boundary of the control volume.
Energy can enter and exit control volumes by work. Because work is done on or
by a control volume when matter flows across the boundary, it is convenient to separate
 75

the work rate (or power) into two contributions. One contribution is the work rate
associated with the force of the fluid pressure as mass is introduced at the inlet am
removed at the exit. Commonly referred to as flow work. this contribution is accounted
for by mi(ptVt) and me(Peve). respectively. where p denotes pressure and v denotes
specific volume. The other contribution. denoted as Wcu' includes all other work effects.
such as those associated with rotating shafts. displacement of the boundary. and electrical
effects. Wcv is considered positive for energy transfer from the control volume.
Energy can also enter and exit control volumes with flowing streams of matter. On
a one-dimensional flow basis. the rate at which energy enters with matter at inlet i is
ml (ui + Vi2 /2 + gZi). where the three terms in parentheses account. respectively. for
the specific internal energy. specific kinetic energy. and specific gravitational potential
energy of the substance flowing through port i. In writing Eq. (2a). the sum of the
specific internal energy and specific flow work at each inlet and exit is expressed in terms
of the specific enthalpy h( = u + pv). Finally. Q accounts for the rate of energy transfer
by heat. and is considered positive for energy transfer to the control volume.
The second law of thermodynamics can be expressed in terms of entropy or.
alternatively. in terms of exergy (Sec. 3). Like mass and energy. entropy can be stored
within systems and transferred across system boundaries. However. unlike mass am
energy. entropy is not conserved. but generated (or produced) by irreversibilities within
systems. The extensive property balance for entropy takes the form

ciS
dt
(3a)
rate of rates of rate of
entropy entropy entropy
change transfer generation

where dS / dt represents the time rate of change of entropy within the control volume.
The terms mis i and mese account. respectively. for rates of entropy transfer into am
out of the control volume accompanying mass flow. Qj represents the time rate of heat
transfer at the location on the boundary where the instantaneous temperature is T). am
QiT) accounts for the accompanying rate of entropy transfer. Sgen denotes the time
rate of entropy generation due to irreversibilities within the control volume.
When applying the entropy balance in any of its forms. the objective is often to
evaluate the entropy genciation term. However, the value of the entropy generation for a
given process of a system usually does not have much significance by itself. The
significance is normally determined through comparison: The entropy generation within
a given component might be compared with the entropy generation values of the other
components included in an overall system formed by these components.
To evaluate the entropy transfer term of the entropy balance requires information
about both the heat transfer and the temperature on the boundary where the heat transfer
occurs. The entropy transfer term is not always subject to direct evaluation. however.
because the required information is either unknown or not defined. In such applications. it
may be convenient. therefore. to enlarge the system to include enough of the immediate
 76

surroundings that the temperature on the boundary of the enlarged system corresponds to
the ambient temperature Tamb • The entropy transfer rate is then simply Q/Tamb •
However, as the irreversibilities present would not be just those for the system of interest
but those for the enlarged system, the entropy generation term would account for the
effects of internal irreversibilities within the system and external irreversibilities within
that portion of the surroundings included within the enlarged system.
For control volumes at steady state, Eqs. (1a) - (3a) reduce, respectively, to read

(1b)
e

~QJ ~. ~ . .
o = L.J - + L.J misl - L.J mese + Sgen (3b)
j TJ i e

The mass, energy and entropy balances generally must be supplemented by

appropriate thermodynamic property data [1, 5]. For some applications a momentum
equation expressing Newton's second law of notion for control volumes is also required.
Data for transport properties, heat transfer coefficients, friction factors, and so on also
may be needed for a comprehensive engineering analysis.

3. Exergy Analysis

In this section essential aspects of exergy analysis are surveyed, For further details, see
[2,6-9].

3.1. DEFINING EXERGY

An opportunity for doing work exists whenever two systems at different states are placed
in communication because. in principle, work can be developed as the two are allowed to
come into equilibrium. When one of the two systems is a suitably idealized system called
an exergy reference environment or simply, an environment, and the other is some
system of interest, exergy is the maximum theoretical useful work (shaft work or
electrical work) obtainable as the systems interact to equilibrium, heat transfer occurring
with the environment only. (Alternatively, exergy is the minimum theoretical useful
work required to form a quantity of matter from substances present in the environment and
to bring the matter to a specified state.) Exergy is a measure of the departure of the state
of the system from that of the environment, and is therefore an attribute of the system and
environment together. Once the environment is specified, however, a value can be
assigned to exergy in terms of property values for the system only, so exergy can be
regarded as an extensive property of the system.
Exergy can be destroyed and generally is not conserved. A limiting case is when
exergy would be completely destroyed, as would occur if a system were to come into
 77

equilibrium with the environment spontaneously with no provision to obtain work. The
capability to develop work that existed initially would be completely wasted in the
spontaneous process. Moreover, since no work needs to be done to effect such a
spontaneous change, the value of exergy can never be negative.
Although the term environment is selected to suggest kinship with the natural
environment, the exergy reference environment used in any particular application is a
thermodynamic model. Models with various levels of specificity are used to evaluate
exergy. The exergy reference environment is typically regarded as composed of common
substances existing in abundance within the Earth's atmosphere, oceans, and crust. The
substances are in their stable forms as they exist naturally, and there is no possibility of
developing work from interactions - physical or chemical - between parts of the
environment. Although the intensive properties of the environment are assumed to be
unchanging, the extensive properties can change as a result of interactions with other
systems. Kinetic and potential energies are evaluated relative to coordinates in the
environment, all parts of which are considered to be at rest with respect to one another.
For computational ease, the temperature To and pressure Po of the environment are
often taken as standard-state values, such as 1 atm and 25°C(77°P). However, these
properties may be specified differently depending on the application. To and Po might be
taken as the average ambient temperature and pressure, respectively, for the location at
which the system under consideration operates. Or, if the system uses atmospheric air, 10
might be specified as the average air temperature. If both air and water from the natural
surroundings are used, 10 would be specified as the lower of the average temperatures for
air and water. Further discussion of the exergy reference environment is provided in Sec.
3.3.
When a system is in equilibrium with the environment, the state of the system is
called the dead state. At the dead state, the conditions of mechanical, thermal, am
chemical equilibrium between the system and the environment are satisfied: the pressure,
temperature, and chemical potentials of the system equal those of the environment,
respectively. In addition, the system has no motion or elevation relative to coordinates in
the environment. Under these conditions, there is no possibility of a spontaneous change
within the system or the environment, nor can there be an interaction between them. At
the dead state, the value of exergy is zero. Another type of equilibrium between the
system and environment can be identified. This is a restricted form of equilibrium where
only the conditions of mechanical and thermal equilibrium must be satisfied. This state of
the system is called the restricted dead state. At the restricted dead state, the fixed
quantity of matter under consideration is imagined to be sealed in an envelope impervious
to mass flow, at zero velocity and elevation relative to coordinates in the environment,
and at the temperature 10 and pressure Po'

3.2. CONTROL VOLUME EXERGY RATE BALANCE

Exergy balances can be written in various forms, depending on whether a closed system or
control volume is under consideration and whether steady-state or transient operation is of
interest. Owing to its importance for a wide range of applications, an exergy rate balance
for control volumes at steady state is presented here.
78

Exergy can be transferred across the boundary of a control volume by three means:
exergy transfer associated with work, exergy transfer associated with heat transfer, am
exergy transfer associated with the matter entering and exiting the control volume. All
such exergy transfers are evaluated relative to the environment used to define exergy.
Exergy is also destroyed by irreversibilities within the control volume. At steady state

(4a)
j e

rates of rate of
exergy exergy
transfer destruction

or

(4b)

where Wcv is the work rate excluding the flow work. Qj is the time rate of heat transfer
at the location on the boundary of the control volume where the instantaneous
temperature is T j . The associated rate of exergy transfer is

(5)

In Eqs. (4), the subscripts i and e denote inlets and outlets, respectively. The exergy
transfer rates at control volume inlets and outlets are denoted, respectively, as Ei = miei
and Ee = meee. Finally, ED accounts for the time rate of exergy destruction due to
irreversibilities within the control volume. The exergy destruction rate is related to the
entropy generation rate by

The specific exergy transfer terms e i and ee are expressible in terms of four
. I exergy, e PH
components: PhYSlca k · · exergy e KN , potentIa
metlc . 1
exergy e PT , --..I
'1UJ

. I
Chemlca exergy e
CH
:

(7)
As discussed in [2], the first three components are evaluated as follows:

(8a)
 79

KN V2
e (8b)
2

PT
e = gz (8c)

where h and S denote, respectively, the specific enthalpy and specific entropy, and V arx:I
z denote velocity and elevation relative to coordinates in the environment, respectively.
The subscript 0 denotes the restricted dead state.

3.3. CHEMICAL EXERGY

To evaluate the chemical exergy, the exergy component associated with the departure of
the chemical composition of a system from that of the environment, the substances
comprising the system are referred to the properties of a suitably selected set of
environmental substances. For discussion of alternative sets of substances tailored to
particular applications, see [6, 8]. Exergy analysis is facilitated, however, by employing a
standard environment and a corresponding table of standard chemical exergies. Standard
chemical exergies are based on standard values of the environmental temperature 10 arx:I
pressure Po - for example, 298.15 K(25°C) and 1 atm, respectively. A standard
environment also consists of a set of reference substances with standard concentrations
closely reflecting the chemical makeup of the natural environment. The reference
substances generally fall into three groups: gaseous components of the atmosphere, solid
substances from the lithosphere, and ionic and nonionic substances from the oceans. The
chemical exergy data of Table 1 correspond to two alternative standard exergy reference
environments, called here model I and model II, that have gained acceptance for
engineering evaluations.
Although the use of standard chemical exergies greatly facilitates the application of
exergy principles, the term standard is somewhat misleading since there is no one
specification of the environment that suffices for all applications. Still, chemical exergies
calculated relative to alternative specifications of the environment are generally in good
agreement. For a broad range of engineering applications the simplicity and ease of use of
standard chemical exergies generally outweigh any slight lack of accuracy that might
result. In particular, the effect of slight variations in the values of 10 and Po about the
values used to determine the standard chemical exergies reported in Table 1 can be
neglected.
80

TABLE I Standard chemical exergy, e CH (kJ / kInol)

of various substances at 298 K and PO.
Substance Formula Modella Model lib
Nitrogen N 2 (g)
640 720
Oxygen 3,950 3,970
°2(g)
Carbon dioxide CO 2 (g)
14,175 19,870
Water H 2 O(g)
8,635 9,500
45 900
H 2°(l)
Carbon (graphite) 404,590 410,260
CIs)
Hydrogen H 2 (g) 235,250 236,100
Methane 824,350 831,650
CH.(g)
Ethane 1,482,035 1,495,840
C 2 H 6 (g)
Methanol 715,070 722,300
CH 3 OH(g)
710,745 718,000
CH 3 OH(l)
Ethyl alcohol 1,348,330 1,363,900
C 2 H s OH(g)
1,342,085 1,357,700
C 2H s OH(l)

a. See [1,10) Po = 1.019 atm.

b. See [1, 9) Po = 1.0 atm.

°
A common feature of standard exergy reference environments is a gas phase.
intended to represent air. that includes N2 • 2 , CO 2 , H 2 0(g), and other gases. Each
e e
gas i present in the gas phase is at temperature 10 and partial pressure PI = xlPo'
e
where the superscript denotes the environment and x7 is the mole fraction of gas i in
the gas phase. Referring to the device at steady state shown in Fig. 1. the standard
chemical exergy for a gas included in the environmental gas phase can be evaluated as
follows: Gas i enters at temperature 10 and pressure Po. expands isothermally with heat
transfer only with the environment. and exits to the environment at temperature 10 am
partial pressure x7 po. The maximum theoretical work per mole of gas i would be
developed when the expansion occurs without internal irreversibilities. Accordingly. with
energy and entropy balances together with the ideal gas equation of state. the chemical
exergy per mole of i is
e
-CH - 1 Xi Po
el = -RTo n - -
Po
-
= -RTo In XI
e (9)

The details of the derivation are left as are exercise. As an application of Eq. (9)
consider CO 2 with x~o
2
= 3.3 x 10-4 and To = 298.15 K. as in [9]. Inserting
 81

values gives egg 2

= 19.870 kJ / kmol, which is the value listed for CO 2 in Table
1 (Model II).

.
Wcv

Gas kat
To· Po

'- ------7-------
To

Figure 1. Device for evaluating the chemical exergy of a gas.

The chemical exergy of a mixture of n gases each of which is present in the

environmental gas phase can be obtained similarly. We may think of a set of n devices
such as shown in Fig. I, one for each gas in the mixture. Gas i, whose mole fraction in
the gas mixture at 10, Po is xi' enters at 10 and the partial pressure xi po. As before,
the gas exits to the environment at 10 and the partial pressure xr po. Paralleling the
previous development, the work per mole i is -RTo In( x7 IX i ). Summing over all
components~ the chemical exergy per mole of mixture is
e
_CH - ~ Xi
e mlxture = - RTo L.J Xi I n -
Xi

This expression can be written alternatively with Eq. (9) as

(10)

Equation (10) remains valid for mixtures containing gases other than those present in the

e:
reference environment and can be extended to mixtures (and solutions) that do not adhere
to the ideal gas model [8]. In such applications, the terms H are selected from a table

of standard chemical exergies.

As a further illustration of the evaluation of chemical exergy, consider the case of
hydrocarbon fuels. In principle, the standard chemical exergy of a substance can be.
evaluated by considering an idealized reaction of the substance with other substances for
which the chemical exergies are known. For the case of a pure hydrocarbon fuel CaHb
at To, PO' refer to the system shown in Fig. 2, where all substances are assumed to enter
82

and exit at To. po. For operation at steady state. heat transfer only with the environment.
and no internal irreversibilities. a systematic application of extensive property balances
results in the following alternative expressions [2]:

(lla)

-CH
+ ae co
(b)_CH (a + -b)_CH
+ - eH 0 - eo (llb)
2 2 2 4 2

Where the subscript F denotes the fuel. HHV denotes the higher heating value (H 2 0 is
a liquid). s denotes absolute entropy. and 9 denotes the Gibbs function.
With Eqs. (11). the chemical exergy of CaHb can be calculated using the standard
°
chemical exergies of 2 , CO 2 , and H2 0(l). together with selected property data: the
fuel higher heating value and absolute entropies. or Gibbs functions of formation. As an
application, consider the case of carbon. With a = 1, b = O. Gibbs function data from
[7] and standard chemical exergies for CO 2 and 02 from Table I (Model 11), Eq. (llb)
gives e~H = 410.280 kJ / krnol, which agrees closely with the value listed in Table I
(Model II). Similarly, for the case of hydrogen (a = O. b = 2) we have
e~: = 236. 095 kJ / kmol. When tabulated chemical exergy data are lacking, as for
example in the cases of coal, char, and fuel oil, the approach of Eq. (lla) can be invoked
using a measured heating value and an estimated value of the fuel absolute entropy

.
determined with procedures discussed in the literature. For an illustration, see [2] .
Wcv

.
Qcv
Figure 2. Device for evaluating the chemical exergy of C. H b
 83

3.4. EXERGETIC EFFICIENCY

The exergetic efficiency (second law efficiency, effectiveness, or rational efficiency)

provides a true measure of the performance of a system from the thermodynamic
viewpoint. To define the exergetic efficiency both a product and a fuel for the system
being analyzed are identified. The product represents the desired result of the system
(power, steam, some combination of power and steam, etc.). Accordingly, the definition
of the product must be consistent with the purpose of purchasing and using the system.
The fuel represents the resources expended to generate the product and is not necessarily
restricted to being an actual fuel such as a natural gas, oil, or coal. Both the product and
the fuel are expressed in terms of exergy.
For a control volume at steady state whose exergy rate balance reads

(12)

the exergetic efficiency is

E = -Ep = 1 - --"'-----'=-
ED + EL (13)
EF EF
where the rates at which the fuel is supplied and the product is generated are EF and Ep
respectively. ED' and EL denote the rates of exergy destruction and exergy loss,
respectively. Exergy is destroyed by irreversibilities ~ithin the control volume, and
exergy is lost from the control volume via stray heat transfer, material streams vented to
the surroundings, and so on.
The exergetic efficiency shows the percentage of the fuel exergy provided to a control
volume that is found in the product exergy. Moreover, the difference between 100% and
the value of the exergetic efficiency, expressed as a percent, is the percentage of the fuel
exergy wasted in this control volume as exergy destruction and exergy loss. To apply Eq.
(13), decisions are required concerning what are considered as the fuel and the product.
Table 2 provides illustrations for four common components. Similar considerations are
used to write exergetic efficiencies for systems consisting of several components, as for
example a power plant.
Exergetic efficiencies can be used to assess the thermodynamic performance of a
component, plant, or industry relative to the performance of similar components, plants,
or industries. By this means the performance of a gas turbine. for instance, can be gauged
relative to the typical present-day performance level of gas turbines. A comparison of
exergetic efficiencies for dissimilar devices - gas turbines and heat exchangers, for example
- is generally not significant, however. The exergetic efficiency is generally more
meaningful. objective, and useful than other efficiencies based on the first or second law
of thermodynamics, including the thermal efficiency of a power plant, the isentropic
efficiency of a compressor or turbine, and the effectiveness of a heat exchanger. The
thermal efficiency of a cogeneration system is misleading because it treats both work and
heat transfer as having equal thermodynamic value. The isentropic turbine efficiency does
not consider that the working fluid at the outlet of the turbine has a higher temperature
(and consequently a higher exergy that may be used in the next component) in the actual
process than in the isentropic process. The heat exchanger effectiveness fails to identify
the exergy destruction associated with the pressure drops of the heat exchanger working
fluids.
84

TABLE 2. The Exergetic Efficiency for Selected Components at Steady State a

Thrbine or Extraction Compnssor, Heat

Expander Thrbine Pump, or Fan Exchangerb

Hot stream
2

Cold
stream
Componenl

E, W W E, -E, E, -E,
E, E, -E, E,-E, -E) W E) -Eo
W W E, -E, E, -E,
£
E, -E, E, -E, -E, W E) -Eo

a) For discussion, see [2].

b) This definition assumes that the purpose of the heat exchanger is to heat the
cold stream (T1 ~ To)'

3.5. CASE STUDY PROBLEM

As an illustration of exergy analysis, consider the gas turbine cogeneration system shown
in Fig. 3, which produces process steam and a net power output. Table 3 provides exergy
and exergy destruction data for the cogeneration system determined using concepts
introduced in Secs. 3.2, 3.3; see [2] for details.
Using the approach of Sec. 3.4, the overall exergetic efficiency is calculated as the
percentage of the fuel exergy supplied to the system that is recovered in the product of the
system. Identifying the product of the cogeneration system as the sum of the net power
developed and the net increase of the exergy of the feedwater.

E=
l"net + (£9 - £8)
ElO

30 MW + 12.75 MW = 0.503(50.3%) (14)

85MW
 85

\V,...lT_--,
Feedwater
f9\ Saturated Vapor.
® 20 bar . 14 kgls
r--L _ _ _ Air Preheater

Combustion
Products
Heat - Recovery
Steam Generator

~_.~ Net Power

30MW

CD
Air

Figure 3. Gas turbine cogeneration system

where the exergy values are obtained from Table 3. The exergy carried out of the system
at state 7, amounting to 3.27 percent of the fuel exergy, is regarded as a loss.
The exergy destruction and loss data summarized in Table 3 clearly identify the
combustion chamber as the major site of thermodynamic inefficiency. The next most
prominent site is the recovery steam generator (HRSG). Roughly equal contributions to
inefficiency are made by the gas turbine, air compressor, and the loss associated with
stream 7. The gas compressor is an only slightly small contributor.
The exergy destructions in these components stem from one or more of three
principal internal irreversibilities associated, respectively, with chemical reaction, heat
transfer, and friction. All three irreversibilities are present in the combustion chamber,
where chemical reaction and heat transfer are both significant and friction is of secondary
importance. For the HRSG and pre heater the most significant irreversibility is related to
the stream-to-stream heat transfer, with friction playing a secondary role. The exergy
destruction in the adiabatic gas turbine and air compressor is caused mainly by friction.
Although combustion is intrinsically a significant source of irreversibility, a
dramatic reduction in its effect on exergy destruction cannot be expected using ordinary
means. Still, the inefficiency of combustion can be ameliorated by preheating the
combustion air and reducing the air-fuel ratio. For the HRSG and preheater, the effect of
the irreversibility associated with heat transfer tends to lessen as the minimum
temperature difference between the streams, (~T)min' is reduced. The exergy destruction
within the gas turbine and the air compressor decreases as the isentropic turbine arxf
compressor efficiencies increase, respectively. The significance of the exergy loss
associated with stream 7 reduces as the gas temperature reduces.
Although these considerations provide a basis for implementing practical
engineering measures to improve the thermodynamic performance of the cogeneration
system, such measures have to be applied judiciously. Measures that improve the
thermodynamic performance of one component might adversely affect another, leading to
86

TABLE 3. Exergy and exergy destruction data for the gas

turbine cogeneration system of Fig. 3.

Exergy Data: I
Mass Flow Rate T p
State Substance (kgls) (K) (bar) E
(MW)
I Air 91.28 298.15 1.013 0.00
7 Combustion 92.92 426.90 1.013 2.77
products
8 Water 14.00 298.15 20.000 0.06
9 Water 14.00 485.57 20.000 12.81
10 Methane 1.64 298.15 12.000 85.00

Exergy Destruction Data:

Component MW (%)2
Combustion chamber 3 25.48 29.98
Heat recovery steam generator 6.23 7.33
Gas turbine 3.01 3.54
Air preheater 2.63 3.09
Gas compressor 2.12 2.49
Overall plant 39.47 46.43 4

I. Determined relative to Table I (Model I).

2. Exergy destruction within the component as a percentage of the exergy entering the plant with
the fuel.
3. Includes the exergy loss accompanying heat transfer from the combustion chamber, which is
estimated as two percent of the lower heating value of the fuel.
4. An additional 3.27 percent of the fuel exergy is carried out of the plant as state 7 and is charged
as a loss.

no net overall improvement. Moreover, measures to improve thermodynamic

performance invariably have economic consequences. The objective in thermal system
design normally is to identify the cost-optimal configuration, requiring consideration of
both thermodynamic and economic imperatives, as discussed in Secs. 4.1 and 5.2.

4. Guidelines for Improving Thermodynamic Effectiveness

Various methods can be used to improve thermodynamic effectiveness. All such methods
should achieve their objectives cost-effectively, of course. To improve thermodynamic
effectiveness it is necessary to deal directly with inefficiencies related to exergy destruction
and exergy loss. The following guidelines, called here Sarna' s principles [11,12], provide
valuable advice for pursuing such an objective:

• There is nothing wrong with expending exergy if something useful is obtained

in return.

• Some exergy destructions and exergy losses can be avoided, others cannot.
Efforts should be centered on those that can be avoided.

4.1. OTHER GUIDELINES

The' primary contributors to exergy destruction are chemical reaction, heat transfer,
mixing, and friction, including unrestrained expansions of gases and liquids. To deal with
them effectively, the principal sources of inefficiency not only should be understood
qualitatively, but also determined quantitatively, at least approximately. Design changes
 87

to improve effectiveness must be done judiciously, however, for the cost associated with
different sources of inefficiency can be different. For example, the unit cost of the
electrical or mechanical power required to provide for the exergy destroyed owing to a
pressure drop is generally higher than the unit cost of the fuel required for the exergy
destruction caused by combustion or heat transfer.
Since chemical reaction is a significant source of thermodynamic inefficiency, it is
generally good practice to minimize the use of combustion. In many applications the use
of combustion equipment such as boilers is unavoidable, however. In these cases a
significant reduction in the combustion irreversibility by conventional means simply
cannot be expected, for the major part of the exergy destruction introduced by combustion
is an inevitable consequence of incorporating such equipment. Still, the exergy
destruction in practical combustion systems can be reduced by minimizing the use of
excess air and by preheating the reactants as in Fig. 3. In most cases only a small part of
the exergy destruction in a combustion chamber can be avoided by these means.
Consequently, after considering such options for reducing the exergy destruction related to
combustion, efforts to improve thermodynamic performance should focus on components
of the overall system that are more amenable to betterment by cost-effective conventional
measures. In other words, apply Sama's second principle.
Nonidealities associated with heat transfer also typically contribute heavily to
inefficiency. Accordingly, unnecessary or cost-ineffective heat transfer must be avoided.
Additional guidelines follow:

• The higher the temperature T at which a heat transfer occurs in cases where
T > To' the more valuable the heat transfer and, consequently, the greater the need
to avoid heat transfer to the ambient, to cooling water, or to a refrigerated stream.
• The lower the temperature T at which a heat transfer occurs in cases where
T < To ' the more valuable the heat transfer and, consequently, the greater the
need to avoid direct heat transfer with the ambient or a heated stream.
• Ext:rgy destruction associated with heat transfer between streams varies inversely
with the temperature level. Accordingly, the lower the temperature level, the
greater the need to minimize the stream-to-stream temperature difference.

Although irreversibilities related to friction, unrestrained expansion, and mixing are

often secondary in importance to those of combustion and heat transfer, they should not
be overlooked, and the following guidelines apply:

• Minimize the use of throttling; check whether power recovery expanders are a
cost-effective alternative for pressure reduction.
• A void processes using excessi vel y large thermodynamic dri ving forces (differences
in temperature, pressure, and chemical composition). In particular, minimize the
mixing of streams differing significantly in temperature, pressure, or chemical
composition.
• The greater the mass rate of flow, the greater the need to use the exergy of the
stream effectively.
• The lower the temperature level, the greater the need to minimize friction.
88

4.2. RELATION TO ENVIRONMENTAL IMPACT

For industries where energy resources are a major contributor to operating costs, an
opportunity exists for cost savings by improving thermodynamic effectiveness using
means such as discussed in Sec. 4.1. This is a well-known and largely accepted approach
today. A related but less publicized aspect concerns effluent streams. The waste from a
plant is often not an unavoidable result of plant operation but a measure of its
inefficiency: The less efficient a plant, the more unusable by-products it produces, ani
conversely. Effluents not produced owing to greater efficiency require no costly cleanup
and do not impose a burden on the environment.
Cleanup efforts havt: customarily featured an end-oJ-the-pipe approach that addresses
the pollutants emitted from stacks, ash from incinerators, thermal pollution, and so on.
Increasing attention is being given today to what goes into the pipe. however. This is
embodied in the concept of design Jor the environment (DFE) , in which the
environmentally preferred aspects of a system are treated as design objectives rather than
as constraints. The aim in DFE is to anticipate negative environmental impacts ani
engineer them out. In particular, efforts are directed to reducing the creation of waste ani
to managing materials better, using methods such as changing the process technology
and/or plant operation, replacing input materials known to be sources of toxic waste with
more benign materials, and doing more in-plant recycling. Exergy analysis, with its
rational approach to costing and well-defined measures of true efficiency, is especially
suited for use in DFE. Still, this is an aspect of exergy analysis that has lagged
somewhat and deserves more development by the thermodynamics community.
A related area of application involving exergy that merits further development is in
assessing environmental impact. An underlying idea is that the exergy of an effluent
stream may serve as index of the stream's influence on nature. That is, the exergy of
such a stream might be an indicator of its potential for driving damaging processes in the
natural environment, Alternatively, the exergy of an effluent stream might be correlated to
observed types of environmental damage. Though occasionally mentioned in the
literature over the past two decades, exergy-aided assessment of environmental impact
largely remains in its infancy. Recent efforts have suggested intriguing possibilities [13-
16J, but much remains to be done.

5. Thermal Systems Design

In this section, the life-cycle design process is considered together with the role of exergy
analysis in design.

5.1. LIFE-CYCLE DESIGN PROCESS

Engineering design requires highly-structured critical thinking and active communication

among the group of individuals, the design team, whose responsibility is the development
of the design of a thermal system. The typical design project begins with an idea that
something might be worth doing: a primitive problem proposed in recognition of a need
or an economic opportunity_ The life-cycle design process then involves several stages:

1. Understanding the problem: "What" not "How."

 89

2. Concept development: "How" not "What."

3. Detailed design: Sizing, costing, optimization.
4. Project engineering: Equipment purchase, fabrication.
5. Service: Startup, operation, eventual retirement.

The influence of exergy analysis and exergy costing is felt the greatest at stages 2 and 3.
The primitive problem that engineers receive may be only a general statement of a
need or opportunity. For example

Primitive problem: To provide for a plant expansion, additional power and steam
are required. Determine how the power and steam are best supplied.

A fundamental aspect of the design process makes it mandatory that knowledge

about a project be generated as early as possible: At the outset, knowledge about the
project and its potential solutions is at a minimum, but designers have the greatest
freedom of choice because few decisions have been made. As the design process evolves,
knowledge increases but design freedom decreases because earlier decisions tend to
constrain what is allowable later. The early design effort should focus, therefore, on
establishing an understanding of, and the need for, the project.
Accordingly, the first stage of the design process is to develop understanding. At
this stage the object is to determine what qualities a system should possess and not how
to achieve them. The question of how enters later. Specifically, the primitive problem
must be defined, and the requirements to which the system must adhere must be
determined. Not all requirements are equally important. The list of requirements
normally can be divided into hard and soft constraints: musts and wants. The musts are
essential and have to be met by the final design. The wants vary in importance and may
be placed in rank order by the design team. Although the wants are not mandatory, some
of the top-ranked ones may be highly desirable and useful in screening alternatives later.
Each requirement should be expressed in measurable terms, for the lack of a measure often
means that the requirement is not well understood. Moreover, only with numerical values
can the design team evaluate the design as it evolves.
For the primitive problem above, let us assume for illustration purposes that the
first stage of the design process results in the following requirements, though many
others might apply: the power developed must be at least thirty megawatts. The steam
must be saturated or slightly superheated at 20 bar and have a mass flow rate of at least
fourteen kilograms per second. The system must adhere to all applicable environmental
regulations. Subject to these, the total cost is to be minimized.
In the concept development stage, the design team shifts from the what-to-do
phase to the how-to-do-it phase, and considers how the primitive problems can be solved.
Experience plays an important role at this juncture, for successful solutions to similar
problems can provide the basis for thinking about the current case. Artificial intelligence-
based procedures such as recently reported [17] also may prove to be helpful.
The concept development stage can lead to a number, perhaps a great number, of
plausible alternative solutions differing from one another in basic concept and detail.
Because the methods of engineering analysis and optimization can only polish specific
solutions, alternatives must be generated having sufficient quality to merit polishing.
 90

Major design errors more often occur owing to a flawed alternative being selected than to
subsequent shortcomings in analysis and optimization.
Alternative solution concepts for the current problem would include, but not
necessarily be limited to the following: Produce all of the steam required, and

• generate the full electricity requirement, or

• generate a portion of the electricity requirement and purchase the balance of

the electricity needed from a utility, or

• generate more than the electricity requirement, and sell the excess electricity
to the utility.

Each cogeneration alternative might be configured using a steam turbine system, a gas
turbine system, or a combined steam and gas turbine system, and the type of fuel (coal,
natural gas, oil) introduced still more flexibility. Alternative concepts such as these must
be screened to eliminate all but the most worthwhile.

5.2. EXERGY-AIDED DESIGN

After preliminary screening of the above alternative solution concepts, let us suppose for
illustration purposes that the following have been retained for further screening an:!
evaluation:

• Employ a coal-fired steam turbine cogeneration system.

• Employ a gas turbine cogeneration system using a natural gas as the fuel.

• Employ a combined steam and gas turbine cogeneration system using

natural gas as the fuel.

These three alternatives, which are shown schematically in Fig. 1.6 of [2], would be
subject to further screening and evaluation until a preferred design emerges. Exergy
methods can be especially effective at this point of the design process. Exergy methods
may enter the design process on a qualitative basis through design guidelines, as provided
in Sec. 4, and on a quantitative basis through application of thermoecollomics.
Thermoeconomics combines thermodynamic and economic principles for
engineering decision making. For instance, thermoeconomics can be applied to optimize
an entire system or specific variables in a single component. Costing in
thermoeconomics is based on exergy, which in most cases is the only rational basis for
assigning costs to the exergy streams of a thermal system. Through thermoeconomic
analysis, a system's real cost sources can be identified. Such information is useful in
developing design changes that improve overall cost effectiveness.
When thermoeconomic principles are applied, the gas turbine cogeneration system
emerges as the preferred alternative. In keeping with the design guidelines of Sec. 4, a
thermodynamically improved version of this cogeneration configuration can be achieved
by including an air preheater. The result is the case-study gas turbine cogeneration
 91

system of Fig. 3 and Table 3. This configuration provides the starting point for a
thermoeconomic optimization.
A thermodynamic optimization aims at maximizing the exergetic efficiency:
minimizing the exergy destruction and exergy loss. The objective of a thermoeconomic
optimization, however, is to minimize costs, including costs related to exergy destruction
and loss. A thermodynamically-optimal design is generally much different than a cost-
optimal design. In particular, the total cost associated with the thermodynamically-
optimal design is always higher than for the cost-optimal design. It is also noteworthy
that the optimization of a thermal system seldom leads to a unique solution corresponding
to a global mathematical optimum. Rather, acceptable alternative solutions may be
feasible. Different solutions developed by different design teams may be equally
acceptable and nearly equally cost effective. As a rule, the more complex the system
being optimized, the larger the number of acceptable solutions.
Although the details are beyond the scope of the current presentation, the
discussion of the cogeneration system case study initiated with the above primitive
problem is brought to closure in [2] with the presentation of a cost-optimal design (see
Sec. 9.5 of [2]). Referring again to life-cycle design, such a result provides a solid basis
for completing the detailed design stage. Though important from an engineering
perspective, the remaining aspects of the life-cycle design process are not detailed here for
brevity, however.

6. Closure

Over the past two centuries, and especially in the final quarter of this century, the use of
the second law for the analysis and design of thermal systems has evolved to a state of
relative maturity. Today, engineers have powerful second-law methods at their disposal.
Although such methods can be significant aids, it is worth observing that the successful
completion of a thermal design project is a complex undertaking, requiring principles
from many disciplines: engineering thermodynamics, fluid mechanics, heat and mass
transfer, mechanical design, automatic controls, mathematics, and economics, to name
just a few. Experience and intuition also play roles, and often a bit of luck is helpful.

7. References
1. Moran, M.l (1998) Engineering Thermodynamics (Sec. 2) in F. Kreith (ed), Mechanical Engineering
Handbook. CRC Press, Boca Raton.
2. Bejan, A, Tsatsaronis, G. and Moran, M. (1996), Thermal Design & Optimization, Joim Wiley & Sons, New
YorK.
3. Moran, M.J. (1998) On Second-Law Analysis and the Failed Promise of Finite-time Thermodynamics,
Energy Int. J.,23, 517-519.
4. Bird, R.B., Stewart, W.E. and Lightfoot, E.N. (1960) Transport Phenomena, Jolm Wiley & Sons, New
York.
S. Moran, M.J. and Shapiro, H.N. (1996) IT: interactive ThermodynamiCS, Software to accompany
Fundamentals a/Engineering ThermodynamiCS, 3rd ed, Jolm Wiley & Sons, New Yark.
6. Moran, M.J. (1989) Availability Analysis - A Guide to EffiCient Energy Use, ASME Press, New York.
7. Moran, M.J. and Shapiro, H.N. (1996) Fundamentals o/Engineering Thermodynamics, 3rd ed., Jolm Wiley
& Sons, New York.
8. Kotas, T.J. (1995) The Exergy Method ofThermal Plant AnalYSis Krieger, Melbourne, FL.
9. Szargut, J., Morris, D.R., and Steward, F.R. (1988) Exergy Analysis of Thermal, Chemical and
Metallurgical Processes, Hemisphere, New Yark.
10. Ahrendts, J. (1980) Reference states, Energy Int. 1., 5,667-677.
92

II. Sarna, DA (1998) Department of Chemical Engineering, University of Massachusetts Lowell, Lowell, MA
01854 USA Personal Communication.
12. Sarna, D.A (1995) The Use of Second Law ofThennodynamics in Process Design, ASME Journal o/Energy
Resources Technology. 117, 179-185.
13. Rosen, MA, and Dincer, I. (1997) On Exergy and Environmental Impact, Int. J. Energy Research. 21, 643-
654.
14. Creyts, IC. and Carey, YP. (1997) Use of Extended Exergy Analysis as a Tool for Assessment of the
Environmental Impact of Industrial Processes, in Proceedings 0/ the ASME Advanced Energy Systems
Division, AES-VoL37. ASME, New York, 129-137.
IS. Ayres., R.u., Ayers, L.W., and Martinas, K. (1998) Exergy, Waste Accounting, and Life-Cycle Analysis,
Energy Int. J., 23, 355-363.
16. Markarytchev, S.V. (1998) Environmental Impact Analysis of ACFB-Based Gas and Power Cogeneration,
Energy Int. J. (to appear)
17. Sciubba, E. (1998) Toward Automatic Process Simulators, ASME Journal o/Engineering/or Gas Turbines
and Power, 120, 1-16.
STRENGTHS AND LIMITATIONS OF EXERGY ANALYSIS

G. TSATSARONIS
Institute for Energy Engineering
Technical University of Berlin
Marchstr. 18. 10587 Berlin. Germany

1. Introduction

An exergy analysis identifies the location, the magnitude and the sources of thermodynamic
inefficiencies in a thermal system. This information, which cannot be provided by other
means (e.g., an energy analysis), is very useful for improving the overall efficiency and cost
effectiveness of a system or for comparing the performance of two systems. This paper deals
with the use of exergetic variables, i.e., those variables calculated in an exergy analysis to
characterize each component and some streams of the system. Their critical review will
identify the strengths and the limitations of an exergy analysis. The following discussion is
limited to systems at steady-state operation. The results can easily be extended to other
systems.
The following variables are calculated in an exergy analysis for the kth component of a
system:

Exergy destruction rate

(1)

Exergy loss rate EL.k

Exergy destruction ratio

YD.k (2)

• Exergy loss ratio

h,k (3)

Percentage of exergy destruction

(4)

exergetic efficiency

(5)

93
A. Bejan and E. Mamut (eds.), Thermodynamic Optimization o/Complex Energy Systems, 93-100.
© 1999 Kluwer Academic Publishers.
94

Many exergy analyses reported in the literature calculate the values of some of the above
variables and reach some usually reasonable conclusions regarding system improvement but
without checking whether the proposed changes would lead indeed to an improvement. In
addition, little attention has been given in many exergy analyses to the structure of the system
and the mutual interdependencies among its components. The reader of such articles could
ask the well justified question "And now that I know how much exergy is destroyed in each
component of a system what do I do?" The question remains often unanswered, and some
critics, not always in good faith, use this situation to discredit the concept of exergy analysis.
The purpose of this article is to shed some light on some aspects associated with this analysis.
In the following, the strengths and limitations of exergy analysis applied to evaluate, optimize
or develop a thermal system are briefly discussed.

2. Evaluation

The following three questions need to be answered when reviewing the exergetic variables
used to evaluate the thermodynamic performance of system components:

1. Which variable best characterizes the performance of a component from the

thermodynamic viewpoint?
2. Which variable should be used to compare the performance of similar components in the
same system or in different systems?
3. Which variable should be used to compare the performance of dissimilar components?

To facilitate the following discussion, let us consider the system shown in Fig. I, which
consists of the cO.mponents A and B. The f~el of component A (EF,A) is equa! to the fuel of
the total system (EF,tot)' The product of A (Ep,A) is the fuel of co~ponent B (EF,B)' whereas
the product ofB (Ep,B) is also the product of the overall system (Ep,tot) and is kept constant.
To further simplify the presentation, we also assume that there are no exergy losses:

(6)

COMPONENT A
(ED A' YA' y;. £Al

COMPONENT B
(Eo,B' YB' YB'£B>

Ep,B= Ep,tot = constant

Figure J. System in which the product of one component is the fuel of the next component.
 95

Thus. all thennodynamic inefficiencies are caused by the exergy destruction within
components A and B. Then. by applying the exergy balance

(7)
and Eqs (2). (4) and (5) to components A and B we obtain the following relations

ED.A = EEp •,o, (J... - I). E.

E D.B
=EP.IOI . (1- - 1 )
EB
(8.9)
B A

YD.A 1 - EA' YD.B = EA (1 - EB ) (10. 11)

.
YD.A
1 - EA
. .
YD.B
(1 - EB)EA
(12.13)
1 - EAEB - EAEB

Equation (8) demonstrates that the rate of exergy destruction in component A depends not
only on the efficiency of the same component (EA ). but also on the exergetic efficiency of
component B (E B ). Thus. the rates of exergy destruction should be used very cautiously to
characterize the perfonnance of system components because. in general. a part of the exergy
destruction occuring in a component is caused by the inefficiencies of the remaining system
components (exogenous exergy destruction). The total exergy destruction within a component
is the sum of the exogenous exergy destruction and the endogenous exergy destruction. i.e.
the exergy destruction due exclusively to the component being considered assuming that all
remaining components operate with exergetic efficiencies of 100%. In complex thennal
systems it is very difficult and costly to separate these two parts of the exergy destruction
within a system component. Only in the component where Ep •,01 is generated is the
exogenous exergy destruction zero (see Eq. (9) for component B of the system shown in
Fig.I).
Similarly Eq. (11) shows that the exergy destruction ratio of component B depends on the
exergetic efficiencies of both components A and B. Here only that component where EF.lol is
supplied to the entire system has a YD value which is independent of the perfonnance of the
remaining components (see Eq. (10) for the component A). The cautiousness to be associated
with the use of YD is not reduced if the exergy destruction in the kth component is related to
the product (instead of the fuel) of the overall system as the following equations demonstrate
for the exergy destruction ratios for the components of the system shown in Fig. 1

(14)

I I
YD.A YD,B - 1 (15. 16)

The variab~e Y~.k provides a clear characterization only of the perfonnance of the component
in which Ep •101 is generated (see Eq. (16) for the component B). Neither does variable Y;,k
(Eq. (4» provide an unambiguous characterization of the perfonnance of the kth component
96

as Eqs. (12) and (13) indicate for the system of Fig. 1.

The problems associated with the use of the variables ED,k' YD,k' y~,k and Y;,k during a
thermodynamic evaluation of the kth component stem from the fact that these variables are
affected by the structure of the system. In an attempt to consider these structural effects, Beyer
[2-3] suggested the use of structural coefficients. The calculation of these coefficients
requires the use of simulation software, and it is time consuming. These coefficients have not
yet found appropriate attention in practical applications. Some structural effects are
considered in the application of exergoeconomic techniques to the diagnosis of the operation
of thermal systems [1,12,13].
To answer the questions formulated in the beginning of this section we can conclude that
the only variable that unambiguously characterizes the performance of a component from the
thermodynamic viewpoint is an appropriately defined exergetic efficiency. Details about the
appropriate definition of exergetic efficiencies can be found in [4-6]. The exergetic efficiency
should also be used to compare the performance of similar components operating under
similar conditions in the same system or in different systems. For the comparison of dissimilar
components the only variable that may be used (with the previously mentioned caveat in
mind) is the exergy destruction ratio YD k (or Y~,k if EF.tol remains constant).
In evaluating the thermodynamic pertormance of thermal systems it is always useful to
know what part of the exergy destruction in each component could be avoided in a real
process. This distinction between avoidable (E;~) and unavoidable (E::':) exergy
destruction [7], although associated with many subjective decisions, has the potential to
further facilitate an exergy-based evaluation and optimization procedure. By using a modified
exergetic efficiency e; defined through
• AV •
ED•k + EL•k
1 - (17)
• . UN
EF,k - ED,k

with . . UN . AV
ED,k = ED,k + ED,k (18)

we are able to make a more realistic assessment of the potential for improving the kth
component from the thermodynamic viewpoint. In addition, this modified exergetic efficiency
enables the comparison of dissimilar components with respect to this potential.
However, the major contribution of an exergy analysis to the evaluation of a system comes
through an exergoeconomic evaluation [I, 8] that considers not only the inefficiencies but
also the costs associated with these inefficiencies and the investment expenditures required to
reduce the inefficiencies.

3. Optimization

Before we discuss the use of exergetic variables for optimization purposes we should clearly
state that a procedure with the final objective to maximize the thermodynamic efficiency in
the design of a new system has no practical value and should be considered only in
conjunction with other objectives such as, for example, the minimization of costs and
pollutant emissions.
There is no structured procedure available for the iterative improvement ("optimization")
of the efficiency of a thermal system, Only several guidelines and heuristic rules (e.g., [1,9])
may be cautiously applied. However, they do not always lead to an improvement. In addition,
 97

a minimum experience and intuition are usually required to avoid misinterpretations or

wrong applications of the rules and guidelines.
Usually in studies reporting the results of exergy analyses it is suggested to start the
iterative process for improving the overall efficiency by improving the component with the
highest destruction rate. It is apparent that this suggestion does not always lead to a
thermodynamic improvement. For example, in the cogeneration system considered
throughout Reference [1], the combustion chamber is the component with the highest exergy
destruction rate with a YD value of more than 30%. In the first workable design (the "base
design") the exergy destruction rate is 25.75 MW and the exergetic efficiency is 79.8%,
whereas in the thermodynamically optimal design these values are 23.81 MW and 78.8%,
respectively (see Table 9.2 in [1]). Thus, if we started with the base design and just improved
the efficiency of the combustion chamber, we would probably not improve the overall
efficiency. An iterative approach in which the components are ranked according to their
avoidable exergy destruction might provide an improvement of this iterative optimization
procedure.
Here again the full potential of conducting an exergy analysis for optimization purposes
is used when the exergy analysis is applied as the first step in an iterative exergoeconomic
optimization procedure [I, 8] whose objective is to minimize the costs of generating the
products of a system.
At this point a few words need to be said about two other related concepts that are also
applied for optimization: the pinch analysis and the finite-time thermodynamics. A pinch
analysis by providing an easy to use formal approach can be effectively applied for the
evaluation and thermodynamic optimization of a heat-exchanger network and its integration
with other components. For these applications a pinch analysis could nicely complement an
exergy analysis. The elements of exergy analysis that have been incorporated into an extended
so-called "pinch analysis" applied to any thermal system simply provide an additional proof
of the superiority of exergy analysis in comparison with a "pinch analysis" when dealing with
systems other than heat exchanger networks and/or with cost minimization. As far as the
finite-time thermodynamics concepts are concerned, it is sufficient to say that their value for
the optimization of real-world systems is practically zero.

4. Process Development

This area of exergy analysis applications has not yet received the attention it deserves. The
decision-making process during the preliminary design phase may be significantly simplified
if the exergetic efficiencies of components, processes, and/or subsystems are known. In this
way the overall exergetic efficiency of various concepts can be roughly estimated, thus
enabling a very quick indentifIcation of the concepts that are expected to have efficiencies
higher than a given minimum. This feature is particularly useful in the development stage of
complex thermal systems.
In the following I would like to illustrate how exergy-based estimates assisted or perhaps
guided me in developing a novel concept. At the end of 1981, I was evaluating various
concepts for reducing the exergy destruction in combustion processes by avoiding the direct
contact of oxygen with the fuel [10). In this case, an oxygen carrier (e.g., euO) reacts with the
fuel, thus, providing the oxygen required for the combustion (e.g., euo is reduced to eu).
The intermediate chemical component obtained in the previous reaction (e.g., eu) is
subsequently oxidized with air to provide the oxygen carrier (e.g., euO).
One of the processes I was studying at that time was the following one in which for
simplicity carbon (instead of coal) is used as fuel:
 98

C + 2H2S04 - CO2 + 2S02 + 2H20 (RI)

2S02 + 02 + 3.76N2 - 2S03 + 3.76N2 (R2)

2S03 + 2Hp - 2H2S04 (R3)

The net result obtained through the combination of reaction steps R I-R3 is the stoichiometric
combustion of carbon with air:

(R4)

The exergetic evaluation of the concept defined by reaction steps R 1 through R3 showed that
the exergy destruction in the process using reaction RI is much smaller than the exergy
destruction in the processes associated with reactions R2 and R3.
One day I attended a seminar presentation by a doctoral student who was studying the
following hydrogen generation process, the so-called "Westinghouse" process:

2H20 + S02 - H2 + H2S04 (RS)

1
H2S04 -H20 + S02 + -02 (R6)
2
Reaction RS is conducted in an electrolytic cell. The net result of reactions R5 and R6 is the
splitting of water through the supply of high-temperature (1 ISO K) thermal energy and small
amounts of electric power:

(R7)

In this seminar presentation it was mentioned that the exergy destruction in the electrolytic
step R5 is considerably lower than in step R6. Immediately after hearing that, I came up with
the idea of combining the reaction steps from each of the above processes with the lowest
exergy destruction, that is steps RI and RS, into a new process:

C + 2H2S04 - CO 2 + 2S02 + 2Hp (RI)

4H20 + 2S02 - 2H2 + 2H2S04 (RS)

net result: (R8)

Before the presentation was over, by combining elements from the two original flow
diagrams, I had already developed a rough flow diagram of what later became the HECAP
process (Hydrogen and Electric Power generation using CArbonaceous materials and high-
temperature Process heat), Fig. 2, [11]. The only thing I could be sure of at that moment was
that the new concept, if it could be realized, should have a high overall thermodynamic
efficiency. Subsequent evaluations confirmed the high exergetic efficiency (over 70%) and
some significant environmental advantages of the new concept: no NO x formation, flue gas
consisting almost exclusively of CO2, and no sulfur emissions to the environment regardless
of the amount of sulfur contained in the carbonaceous fuel (e.g., coal). Later also the main
difficulties associated with the realization of this concept became apparent: the lack of (a) gas
 99

turbines in which the gas products of reaction R 1 could be expanded, (b) gas cleaning devices
for the same gas mixture before expansion, and (c ) an appropriate source for the high-
temperature thermal energy that according to the plans at that time should be provided by a
high-temperature nuclear reactor. The above case is a good example of how even qualitative
information from an exergy analysis can assist an engineer in inventing a new process or
concept. An energy analysis cannot provide a similar enhancement of creativity. Without the
information about the relative exergy destruction there would be no incentive for combining
steps Rl and R5 to develop a more complex process which uses two fuels (e.g., coal and
high-temperature thermal energy) and generates two products (electricity and hydrogen).

He

26

5 6
HIGH·
TEMPERATURE
15 He NUCLEAR
REACTOR
HPGT LPGT 7

16

Figure 2. Row diagram of the HECAPprocess.

5. Conclusions

Exergy analysis is without a doubt a powerful tool for developing, evaluating and improving
thermal systems, particularly when this analysis is part of an exergoeconomic evaluation.
However, the lack of a formal procedure in using the results from an exergy analysis is one of
the major reasons for it not being very popular among practitioners. This represents
simultaneously a weakness and a strength of the method. A strength because an exergy
analysis can be conducted for every thermal system and because the engineers control the
evaluation and optimization process and, thus, they could still use their intuition and creativity
as opposed to a formal optimization procedure in which they do not know how and why
specific optimization results were obtained. A specific method (e.g., pinch analysis)
100

developed in a fonnal way for a specific system (e.g., heat-exchanger networks) always has
the potential of gaining a wider acceptance among practitioners than the more general method
of exergy analysis.
Future studies using exergetic and exergoeconomic variables need to focus more on the
mutual interdependencies among the components of a thennal system, and on the calculation
of the unavoidable exergy destruction for each component of a system. Today it is not
sufficient any more to just present the results of an exergy analysis without using them to
improve the system being considered.
Exergy analysis also has a high pedagogical value because it shows to the students what
the real thennodynamic inefficiencies are and what causes them. The results from the analysis
assist the students in developing ideas with respect to what can be done to reduce these
inefficiencies. Teaching the concept of exergy analysis should focus on that strength and on
the appropriate calculation, interpretation and use of the exergetic variables.

References
1. Bejan, A., Tsatsaronis, G., and Moran, M. (1996) Thermal Design and Optimization, J. Wiley, New York.
2. Beyer, J. (1970) Strukturuntersuchungen - notwendiger Bestandteil der Effektivitiitsanalyse von Wiirme-
verbrauchersystemen Energieanwendung 19, 358-361.
3. Beyer, 1. (1974) Struktur wiirmetechnischer Systeme und tikonomische Optimierung der Systemparameter
Energieanwendung 23, 274-279.
4. Tsatsaronis, G. (1995) On the efficiency of energy systems in Y.A. Gtigii~, A. Oztiirk and G. Tsatsaronis (eds.)
Proceedings of the International Conference on Efficiency, Costs, Optimization, Simulation and
Environmental Impact of Energy Systems, ECOS '95, Istanbul, Turkey, July 11-14,53-60.
5. Lazzaretto, A. and Tsatsaronis, G. (1996) A general process-based methology for exergy costing in A.B.
Duncan, J. Fiszdon, D. O'Neal and K Den Braven (eds.) Proceedings '!f" the ASME Advanced Energy Systems
Devision, 36, ASME, New York, 413-428.
6. Lazzaretto, A. and Tsatsaronis, G. (1997) On the quest for objective equations in exergy costing in ML
Ramalingam, J.G. Lage, V.c. Mei, and J.N. Chapman (eds.), Proceedings 'if the ASME Advanced Energy
Sy..rems Devision, 37, ASME, New York, 197-210.
7. Feng, X, Zhu, Xx. and Zheng, J.P. (1996) A practical exergy method for systems analysis, Proceedings of the
19961ntersociety Energy Converson Engineering Conference, IEEE, New York, 2068-2071.
8. Tsatsaronis, G. (1998) Design optimization using exergoeconomics, Presentation at the NATO Advanced Study
Institute Thermodynamics and the Optimization of Complex Energy System." July 13-24, 1998, Neptun,
Constanza, Romania.
9. Szargut, J. and Sama, D.A. (1995) Practical rules of the reduction of energy losses caused by the
thermodynamic imperfection of thermal processes, Proceedings of !TEC '95, the Second International
Thermal Energy Congress, Agadir, Marocco, June 1995,782-785.
10. Knoche, KF. and Richter, H. (1968) Verbesserung der Reversibilitiit von Verbrennungsprozessen, Brenn.,toff-
Wiirme-Kraft 20, 205-210.
11. Tsatsaronis, G. and Spindler, K (1984) Thermodynamic analysis of a new process for hydrogen and electric
power production by using carbonaceous fuels and high-temperature process heat, Int. 1. Hydrogen Energy 9,
595-601.
12. Lozano, M.A. and Valero, A (1993) Theory of the exergetic cost, Energy - The International lournal18, 939-
960.
13. Stoppato, A and Lazzaretto, A. (1996) The exergetic analysis for energy systems diagnosis, Proceedings 'if the
Engineering Systems Design and Analysis (ESDA) Conference PD-Vol. 73, I, ASME, New York, 191-198.
DESIGN OPTIMIZATION USING EXERGOECONOMICS

G. TSATSARONIS
Institute for Energy Engineering
Technical University of Berlin
Marchstr. 18, 10587 Berlin, Germany

1. Introduction

Exergoeconomics, this common branch of mechanical and chemical engineering, represents

a unique combination of an exergy analysis and a cost analysis, to provide the designer or the
operator of an energy-conversion plant with information not available through conventional
energy, exergy or cost analyses but crucial to the design and operation of a cost-effective
plant. Exergoeconomics may be defined as an exergy-aided cost-reduction method. In
addition, exergoeconomics is a very powerful tool for understanding the interconnections
between thermodynamics and economics, and, thus, the behavior of an energy conversion
plant from the cost viewpoint [1].
Design optimization of a thermal system means the modification of the structure and the
design parameters of a system to minimize the total leveIized cost of the system products
under boundary conditions associated with available materials, financial resources, protection
of the environment, and government regulation, together with the safety, reliability,
operability, availability and maintainability of the system. A truly optimized system is one for
which the magnitude of every significant thermodynamic inefficiency (exergy destruction and
exergy loss) is justified by considerations related to costs or is imposed by at least one of the
above boundary conditions. A thermodynamic optimization, which aims at minimizing the
thermodynamic inefficiencies, may be considered as a subcase of design optimization.
Appropriate formulation of the optimization problem is usually the most important and
sometimes the most difficult step of a successful optimization study. In optimization problems
we distinguish among independent variables, whose values are amenable to change; these are
the decision variables and the parameters whose values are practically fixed by the particular
application. In optimization studies, only the decision variables may be varied. The
parameters are independent variables that are each given one specific and unchanging value
in any particular model statement. The variables whose values are calculated from the
independent variables using a mathematical model are the dependent variables.
We will initially consider an almost ideal case: A complete thermodynamic model and a
complete economic model are available for the optimization of an energy system. In addition,
the structure of the system might be considered as optimal, that is, either there are no
alternative structures or each alternative structure is considered to be inferior to the present
structure. Finally, we assume that an analytical or a numerical optimization technique (e.g.,
[2-4]) may be used to directly optimize the decision variables. Only in this ideal case, the
calculation of exergy-based variables is unnecessary since the optimization can be conducted
directly using the available models and techniques. In practical applications, however,
thermal systems cannot usually be optimized as in this ideal case. The reasons include the
following:

Some of the input data and functions required for the thermodynamic and, particularly, the
economic model might not be available or might not be in the required form. For
101
A. Bejan and E. Mamut (eds.), Thermodynamic Optimization of Complex Energy Systems, 101-115.
© 1999 Kluwer Academic Publishers.
102

example, it is not always possible to express the purchased-equipment costs as a function

of the appropriate thermodynamic decision variables.
The analytical and numerical optimization techniques are applied to a specified structure
of the thermal system. However, a significant decrease in the product costs may be
achieved through changes in the structure of the system. It is not always practical to
develop a mathematical optimization model for every promising design configuration of
a system.
Even if all the required information is available, the complexity of the system might not
allow a satisfactory mathematical model to be formulated and solved in a reasonable time.

In these cases, that is, when any of the assumptions made for the ideal case is not fulfilled, the
application of exergoeconomic techniques may provide significant benefits for the
optimization process. The more complex the thermal system, the larger the expected benefits,
particularly when chemical reactions are involved.
Exergoeconomics interacts with several other areas during the optimization procedure.
Fig. I shows schematically some of these interactions: Exergoeconomics uses results from the
synthesis, cost analysis and simulation of thermal systems and provides useful information for
the evaluation and optimization of these systems as well as for the application of expert
systems to improve the design and operation of such systems.

2. A General Exergoeconomic Methodology

1his section presents the main features of a general methodology that can be used to evaluate
and iteratively optimize the design of a thermal system. Sections 2.1 through 2.3 discuss the
procedure for calCUlating the costs associated with all material and energy streams in the
thermal system being considered. This information is used to calculate the characteristic
variables of each system component that are discussed in section 2.4.

2.1. EXERGY COSTING

For a system operating at steady state there may be a number of entering and exiting streams
as well as both heat and work interactions with the surroundings. Associated with these
transfers of matter and energy are exergy transfers into and out of the system and exergy
destructions caused by the irreversibilities within the system. Since exergy measures the true
thermodynamic value of such effects, and costs should only be assigned to commodities of
value, it is meaningful to use exergy as a basis for assigning costs in thermal systems. Indeed,
exergoeconomics rests on the notion that exergy is the only rational basis for assigning costs
to the interactions a thermal system experiences with its surroundings and to their sources of
inefficiencies within it. We refer to this approach as exergy costing. In exergy costing a cost
rate is associated with each exer~y transfer. Thus, f«;lr entering and exiting streams of matter
with associated exergy transfers E j and Ee , power W, and the exergy tranfer associated with
heat transfer E we write, respectively
q

(1,2)

(3,4)

Here c j ' ce ' c w ' and cq denote average costs per unit of exergy, for example in $ per
gigajoule ($101).
 103

process thermodynamic economic

simulation analysis

mathematical expert systems

optimization and
methods fuzzy systems

optimization of energy-c:onversion plants

Figure 1. Interactions of exergoeconomics with other areas of engineering and optimization procedure.

Exergy costing does not necessarily imply that costs associated with streams of matter are
related only to the exergy rate of each respective stream. Non-exergy related costs can also
affect the total cost rate associated with material streams. Examples include the cost rates
associated with a treated water stream at the outlet of a water treatment unit, an oxygen or
nitrogen stream at the outlet of an air separation unit, a limestone stream supplied to a gasifier
or fluidized-bed reactor, iron feedstock supplied to a metallurgical process, and an inorganic
chemical fed into a chemical reactor.
Accordingly, when significant non-exergy-related costs occur, the total cost rate
associated with the material streamj, denoted by <rar, is given by

CJTOT = C.J + C~
J
(5)

where C. is the cost rate directly related to the exergy of stream j (e.g., MS. (1) and (2»
J • NE ';NE
whereas C. is the cost rate due to nonexergetic effects. The term C. represents a
. J fi h '
convement way or c argmg non-exergy-related costs from one component J to 0 ther
components that should bear such costs. More details about C are given in [1] and [5].
NE

2.2. COST BALANCE

Exergy costing involves cost balances usually formulated for each component separately. A
cost balance applied to the kth system component shows that the sum of cost rates associated
with all exiting exergy transfers equals the sum of cost rates of all entering exergy transfers
plus the appr~89ate charges due to capital investment CZ;I) and oper!lting and maintenance
expenses (Zk ). The sum of the last two terms is denoted by Zk' Accordingly, for a
component receiving a heat transfer and generating power, we write

E Ce, k + Cw,k =C q,k + E C",k + it (6)

104

This equation simply states that the total cost of the exiting exergy transfers equals the
total expenditure to obtain them: the cost of the entering exergy streams plus the capital and
other costs. Note that when a component receives power (as in a compressor or a pump) the
term Cw•k would move with its positive sign to the right side of this expression. The term C .k
would appear with its positive sign on the left side if there is a heat transfer from the
component. Cost balances are generally written so that all terms are positive.
Introducing the cost rate expressions of Eqs. (1)-(4), Eq. (6) becomes

~
L.- IeE) +
~. • k
ew.k Wk = eq,k Eq.k + ~
L.- Ie
~ j
E)k + t k
j
(7)
• I

The exergy rates (Ee' W, Eq , and E) exiting and entering the kth. component are
calculated in an exergy analysis [1] conducted at a previous stage. The term Zk is obtained by
first calculating the capital investment and O&M costs associated with the kth component and
then computing the levelized values of these costs per unit of time (year, hour, or second) of
system operation [1, 5-8].
The variables in Eq. (7) are the levelized costs [1] per unit of exergy for the exergy
transfers associated with the kth component (ee,k' cw,k' Cq,k' and c). In analyzing a
component, we may assume that the costs per exergy unit are known for ali entering streams.
These costs are known from the components they exit or, if a stream enters the overall system
consisting of all components under consideration, from the purchase cost of this stream.
Consequently, the unknown variables that need to be calculated with the aid of the cost
balance for the kth component are the costs per exergy unit of the exiting streams (e.,k) and, if
power or useful heat are generated in that component, the cost per unit of exergy associated
with the transfer of power (eW,k) or heat (cq,k)' Usually, some auxiliary equations are necessary
to calculate these costs.

2.3. AUXILIARY COSTING EQUATIONS

Different approaches for formulating the auxiliary equations are suggested in the literature,
each one having a certain degree of subjectivity (see, for example the references given in [9]).
The method recommended for obtaining the least subjective auxiliary equations is detailed in
[9] and [10] and summarized in the following:
The fuel and product used in the definition of the exergetic efficiencies of the components
are formulated through the following procedure:
(I) The exergy carriers, consisting of all material and energy streams are identified.
(2) Depending on the purpose of the study, the total exergy or any of its components (e.g.,
thermal, mechanical and chemical exergies) are considered.
(3) The exergy streams are considered to be separated into two categories:
- continuous exergy streams, and
- interrupted exergy streams.
The purpose of the component being considered is used as a criterium to decide whether
an exergy stream is continuous or interrupted.
(4) At this point the fuel and the product for each component are defined using the
following criteria:
For continuous exergy streams, every decrease in an exergy stream between the inlet
and the outlet of the component is included in the fuel whereas every exergy increase
between the inlet and the outlet of the component is included in the product.
For interrupted exergy streams, the exergy at the inlet of the component belongs to the
fuel, and the exergy at the outlet of the component is part of the product.
(5) The auxiliary costing equations are formulated as follows:
 105

On the fuel side

for continuous exergy streams the F rule is applied, whereas
for interrupted exergy streams no cost equation is necessary.
On the product side
for continuous exergy streams the P rule is used, and
for interrupted exergy streams the P postulate is applied.

The F and P rules and the P postulate are presented in the following.

Frule
The F rule is based on the simple concept that the cost removal from a stream by a component
must be equal to the cost at which the removed exergy was previously supplied to that stream.
For continuous exergy streams, the F rule is expressed as

c/kE (Ei: k.y -

y- I
Eo~t'kJ = ECi;'k,y(Ei: k.y -
Y- I
E,;'t,k,Y) (8)

where X represents the exergy form, k the component, and y the exergy stream. NS is the total
number of exergy streams of the Xth exergy form entering the kth component. The average
cost c;
k is defined by
, NS
~
LJ
(Cin.k,y
x _ CX )
out,k.y
y-I
(9)

Eo~t,k,y and C:"t,k,y are the exiting exergy and cost rates, respectively, associated with the Xth
exergy form of the Xth material or energy stream, which enters the kth component with the
·X ·X
respective values Ein,k,y and Cin,k,y'

Prule
The P rule states that each exergy unit is supplied to all the streams associated with the
product of the kth component at the same average cost. Thus, for continuous exergy streams
y and x we write
M T en
cP,k,y =cP,k,y =cP,k,y =cP,k (10)

and
en en
cP,k,x =cP,k,y =cP,k (11)

Here c P,k is the cost per exergy unit of the product for the kth component.

P postulate
The P postulate derives directly from the P rule and is applied when at least one interrupted
exergy stream participates in the definition of the product: If both the Xth and 2th exergy
forms of the same material stream (y) at the outlet are interrupted, the auxiliary equation to be
used for this stream is
106

x z
Cout,k,y = Cout,k,y (12)

For the same exergy form (X) of two different interrupted outlet streams (y and x) we write
X X
C"ut,k,y = C"ut,k,x (13)

If the Xth exergy form of the yth material stream is continuous and is part of the product and
the 2th exergy form (of the xth material stream) being also part of the product is interrupted,
then we use
X Z
CP,k,y -C
- (lut,k,x (14)

The approach just outlined for obtaining the auxiliary costing equations is general, objective,
and illustrates the interdependency between the definition of exergetic efficiency (or the
purpose) of a component and the formulation of the auxiliary equations.
By combining the cost balances for all system components with the auxiliary costing
equations and by solving the obtained linear system of equations, we can calculate the cost
associated with each material and energy stream in the system.

2.4. EXERGOECONOMIC EVALUATION

The exergoeconomic evaluation is conducted at the system component level using the
variables presented in the following for the kth component. From the exergy analysis we
know the

• rate of exergy destruction, ED k

(15)

• exergetic efficiency Ek

(16)

• and exergy destruction ratio Yk

YD,k = ED,k / EF,tot (17)

In addition, we calculate the following variables from the exergoeconomic analysis:

• cost per unit of fuel exergy

(18)

• cost per unit of product exergy

Cp,k = Cp,k / Ep,k (19)

• cost rate associated with exergy destruction CD k

(20)
 107

• cost rate associated with exergy 10ssesCL •k

CL,k = CF.kEL,k (21)

• cost rate associated with capital investment i kC1

• OM
• cost rate associated with operating and maintenance expenses Zk

• the sum i k of the last two cost rates

(22)

• relative cost difference 'k

(23)

• exergoeconomic factor f

(24)

The values of CF, k depend on the relative position of the kth component in the system and on
the interconnections between the kth component and the remaining components. As a general
rule, the closer the kth component to the product (fuel) stream of the overall system, the larger
(smaller) usually the value of CF k'
Pr
When a~opriate cost functions can be developed [1, 8, 1 the cost optimal exergetic
efficiency £k and the cost optimal relative cost difference 'kO for the kth component can
be approximated by

(25)

where

(26)

and

(27)

Here ~ is the capital recovery factor, Yk is a coefficient that accounts for the part of the fixed
operating and maintenance costs depending on the total capital investment associated with the
kthcompone~t(TClk) while B k , n k and m k are constants used to express TCl k as a function
of £k and Ep •k ; 't is the average annual time of plant operation at the nominal capacity.
Approaches using Eqs. (25) - (27) are still in the development stage.
An exergoeconomic analysis and evaluation identifies and compares the real cost sources
in a system [given by Eqs. (20) - (22)], illustrates the cost formation process within a system,
calculates separately the cost at which each product stream is generated, and, most
108

importantly, assists in the effective cost optimization of a thermal system, particularly its
design. Examples of the design optimization are given in [1, 6-8, and 11-16]. This point is
further discussed in the next section.

3. An Iterative Exergoeconomic Procedure for Optimizing the Design of a Thermal

System

The conventional approach in optimizIng complex systems is to iteratively optimize

sybsystems and/or ignore the influence of some structural changes and decision variables.
Only energy-based principles, or the results from an exergy analysis at most, are used in the
optimization procedure. An alternative to this approach is an iterative exergoeconomic
optimization technique that consits of the following seven steps:

1. In the fist step a workable design is developed. The guidelines presented in [1,8, 11 and
16-18] may assist in developing a workable design that is relatively close to the optimal
design. The use of these guidelines can reduce, therefore, the total number of iterations
required.
2. A detailed exergoeconomic analysis and evaluation, and a pinch analysis are conducted
for the design configuration developed in the previous step. The results are used to
determine design changes that are expected to improve the design being considered. In
this step, and in steps 3-7, we consider, however, only changes in the decision variables
that affect both the exergetic efficiency and the investment costs. The remaining decision
variables are optimized in step 8.
3. If the system has one or two components for which the sum of the cost rates (ik + CD,k)
is significantly higher than the same sum for the remaining components, the designs of
these components are modified to approach the corresponding cost-optimal exergetic
efficiency, Eqs. (25) and (26). To apply these equations, we assume that the costs per
exergy unit remain constant for all inlet streams. This step is meaningful only for
components in which each of the terms i k and CD,k has a significant contribution to the
costs associated with the respective component. If not, step 3 should either be omitted
or, preferably, replaced by an efficiency maximization procedure when Cn,k is the
dominating cost rate, or an investment cost minimization procedure when Zk is the
dominating cost rate.
An alternative to using Eqs. (25) - (27) is to optimize one or two major decision
variables for each of the most important components by graphically minimizing the
relative cost difference (Eq. (23)) calculated for the components at various values of the
corresponding decision variables. In this case it is unnecessary to assume that the costs
per exergy unit remain constant for each inlet stream. This approach was used in [12]
and [13] to optimize the gasification temperature in an integrated gasification-combined-
cycle power plant.
4a. For the remaining components, particularly the ones having a relatively high value of the
sum (ik + CD,k)' the relative deviations of the actual values from the cost-optimal
values for the exergetic efficiency and relative cost difference are calculated:
OPT
A Ek - Ek 100 (28)
L.lEk = - " - - - -
OPT
Ek

(29)
 109

OPT OPT •
Here ek and rk are the actual values and ek and rk are the cost-optImal values for
the exergetic efficiency and the relative cost difference, respectively. The cost-optimal
values are calculated using Eqs. (25) to (27) for the design being currently analyzed.
The design of these remaining thermoeconomically important components is modified
ae
to reduce the values of k and ark' If conflicting design changes are suggested from
the evaluation of different components, the design changes for the components with the
higher (t + C ) values prevail.
k D,k
4b. If Eqs. (25) - (27) are not available, the exergoeconomic factor tk should be used to
identify the major cost source (capital investment or cost of exergy destruction).
a. If the tk value is high, investigate whether it is cost effective to reduce the capital
investment of the kth component at the expense of the component efficiency.
b. If the tk value is low, try to improve the component efficiency by increasing the
capital investment.
5. Eliminate any subprocesses that increase the exergy destruction or exergy loss without
contributing to the reduction of capital investment or of fuel costs for other components.
6. Consider improving the exergetic efficiency of a component if it has a relatively low
exergetic efficiency or relatively large values for the rate of exergy destruction, the
exergy destruction ratio, or the exergy loss ratio.
7. Based on the results from steps 2-6, a new design is developed and the value of the
objective function for this design is calculated. If in comparison with the previous design
this value has been improved, we may decide to proceed with another iteration that
involves steps 2 through 7. If, however, the value of the objective function is not better
in the new design than in the previous one, we may either revise some design changes
and repeat steps 2 through 7 or proceed with step 8.
8. In this step, we use an appropriate mathematical optimization technique to optimize the
decision variables that affect the costs but not the exergetic efficiency. At the end of this
step the cost-optimal design is obtained.
9. Finally, a parametric study may be conducted to investigate the effect on the
optimization results of some parameters used and/or assumptions made in the
optimization procedure.

When applying this methodology, it is important to recognize that the values of all
thermoeconomic variables depend on the component types: heat exchanger, compressor,
turbine, pump, chemical reactor, and so forth. Accordingly, whether a particular value is
judged to be high or low can be determined only with reference to a particular class of
components. It is also important to consider the effects of contemplated design changes in one
component on the performance of the remaining components. These effects may be
determined either by inspection of the system flowsheets or by using a simulation program.
The application of steps I, 2, 4a, and 7 to a cogeneration system is presented in [II],
which is one of four papers [11, 19-21] dealing with exergoeconomic optimization
techniques applied to the same cogeneration system. This system is also used in the case study
discussed in the following section to demonstrate the application of steps I, 2, 4b, and 7.
Examples of exergoeconomic evaluations and improvements of complex thermal systems are
discussed in [12-15].

4. Case Study

Fig. 2 presents the base-case design of a cogeneration system that develops a net power
output of 30 MW and provides 14 kgls of saturated water vapor at 20 bar. The first five
110

columns of Table 1 show relevant thennodynamic and economic data. The second column of
Table 2 shows the purchased-equipment costs (PEC) for each component in the base-case
design. These costs are obtained from the cost equations given in Appendix B of [I]. The
remaining direct costs, as well as the indirect costs, are estimated using average factors. The
total capital investment of the cogeneration system in the base case is estimated at
approximately 46 million mid-I 994 dollars. Table 7.9 in [I] summarizes the parameters and
assumptions used in the economic analysis, which is based on the revenue-requirement
method [22].
The year-by-year economic analysis results in the 1evelized annual costs for fuel
($10.4 x 106 ), operating and maintenance ($5.9 x 106 ), and carrying charges ($10.5 x 106 )
for a levelization time period of 10 years. The values given in the previous sentence are
the corresponding levelized current-dollar costs obtained for the base case. The levelized
costs are used as input data for the therrnoeconomic analysis and optimization. The cost flow
rates in the system are obtained by dividing the levelized annual costs by the number of hours
of system operation per year.
The methodology introduced above will now be applied to the case study cogeneration
system. The objectives are to identify the effects of the design variables on the costs and
suggest values of the design variables that would make the system more cost effective. The
key design variables - the decision variables - for the cogeneration system are the compressor
pressure ratio P21PI' the isentropic compressor efficiency 11.lc ' the isentropic turbine
efficiency 11.1,/, the temperature of the air entering the combustion chamber T3 , and the
temperature of the combustion products entering the gas turbine T4 .

4.1. FIRST ITERATION

The following nominal values of the decision variables correspond to the first workable
design (base-case design) developed for the cogeneration system of Figure 2 and Tables I, 2:

pip I = 10, 11.,c = l1s' = 0.86, T3 = 850K, T4 = 1520K.

The last two columns of Table 1 and the last 10 columns of Table 2 summarize the values of
the thennoeconomic variables calculated for each component of the cogeneration system for
the base-case design. In accord with the 'presen.ted methodology, the components are listed in
order of descending value of the sum CD + Z.
The combustion chamber, the gas turbine, and the air compressor have the highest
values of the sum t + CD and are, therefore, the most important components from the
thermoeconomic viewpoint. The low value of the variable f for the combustion chamber
shows that the costs associated with the combustion chamber are almost exclusively due to
exergy destruction. A part of the exergy destruction in a combustion chamber can be avoided
by preheating the reactants and by reducing the heat loss and the excess air, but this usually
leads only to a small reduction in the exergy destruction. For simplicity, we assume here that
the heat loss cannot be further reduced. The excess air is determined by the desired
temperature T4 at the inlet to the gas turbine. The temperature T4 is a key design variable for
it affects both the perfonnance of the entire system (exergy destruction in the combustion
chamber, gas turbine, air preheater, and HRSG, and exergy loss associated with stream 7) and
the investment costs of the components.
An increase in the heat transfer rate in the air preheater achieved through an increase in
temperature T3 also results in a decrease of the exergy destruction in the combustion chamber.
Thus, the temperature T3 is also a key design variable because, in addition to the combustion
chamber, it affects the exergy loss associated with stream 7 as well as the perfonnance and
 III

Saturated Vapor
Feedwater 20 bar, 14 kg/s
8
Air Preheater
7

Natural Gas
Heat-Recovery
Steam Generator

2 3 5

Net Power
11 Power to Air Com ressor 12

t
30MW

1 Air Compressor Gas Turbine

Air

Figure 2. Base-case design of the cogeneration system.

investment costs of the air preheater and the heat recovery steam generator. Holding all other
decision variables constant, the higher the temperature T j the smaller the average temperature
difference in the air preheater and the heat-recovery steam generator. A decrease in the
average temperature difference in these heat exchangers results in an increase in both the
exergetic efficiency and the capital investment for each heat exchanger.
To summarize: By considering measures for reducing the high cost rate associated with
the exergy destruction in the combustion chamber of the cogeneration system, two key design
variables have been identified, temperatures T3 and 7;. An increase in these temperatures
reduces the CD value for the combustion chamber and other components but increases their
capital investment costs.
Tuming next to the gas turbine, which has the second highest value of the sum + CD' t
the relatively large value of factor f suggests that the capital investment and O&M costs
dominate. The capital investment costs of the gas turbine depend on temperature T4 , pressure
t
ratiop21Pl' and isentropic efficiency 11 ..1 • To reduce the high value associated with the gas
turbine, we should consider a reduction in the value of at least one of these variables.
The air compressor has the highestfvalue and the second highest relative cost difference
r among all cOlIlJ?Onents. Thus, we would expect the cost effectiveness of the entire system to
improve if the Z value for the air compressor is reduced. This may be achieved by reducing
the pressure ratio P2 Ipl and/or the isentropic compressor efficiency 11 ..c .
The heat recovery steam generator has the lowest exergetic efficency and the highest r
value among all the components. As the f value indicates, almost 45% of the relative cost
t
difference is caused by the value in this component, with the remaining 55% caused by
exergy destruction. Thus, we might conclude that a decrease of the exergy destruction in the
HRSG could be cost effective for the entire system even if this would increase the investment
costs associated with this component. The exergy destruction in the HRSG can be reduced by
decreasing the values of T6 and T7• A decrease in the value of T7 also results in a decrease in
the exergy loss from the total system. In terms of the decision variables, temperatures T6 and
T7 may be reduced by increasing Ts and/or decreasing T4 at fixed values of the remaining
decision variables.
112

TABLE I.Mass flow rate, temperature, pressure, exergy rate, and cost data
for the streams of the cogeneration system

State Stream MassAow Temperature Pressure ExergyAow CostAow Cost per Exergy
Rate Rate Rate Unit
rit T p E t c
(kg/s) (K) (bar) (MW) (Sib) ($/OJ)

Air 91.28 298.1 1.01 0.000 0 0

2 Air 91.28 603.7 10.13 27.538 2756 27.80

3 Air 91.28 850.0 9.62 41.938 3835 25.40

4 Combustion 92.92 1520.0 9.14 101.454 5301 14.51

products

5 Combustion 92.92 1006.2 1.10 38.782 2026 14.51

products

6 Combustion 92.92 779.8 1.07 21.752 1137 14.51

products

7 Combustion 92.92 426.9 1.01 2.773 145 14.51

products

8 Water 14.00 298.1 20.00 0.062 0 0

9 Water 14.00 485.6 20.00 12.810 1256 27.23

10 Methane 1.64 298.1 12.00 84.994 1398 4.57

II Power to air 29.662 2003 18.76

compressor

12 Net power 30.000 2026 18.76

The relatively high value off in the air preheater suggests a reduction in the investment
costs of this component. This can be achieved by decreasing T J'
. It should be noted, however, that changes suggested by the evaluation of this component
should only be considered if they do not contradict changes suggested by components with a
larger value of CD + i.
Summarizing the foregoing conclusions, the following changes in the design variables are
expected to improve the cost effectiveness of the system:
Increase the value of T, as suggested by the evaluation of the combustion chamber and
HRSG.
Decrease the pressure ratio P2 IpI (and thus P4 Ips) and the isentropic efficiencies 11.,e'
and 11.,,' as suggested by the evaluation of the air compressor and gas turbine.
Maintain T 4 fIxed, since we get contradictory indications from the evaluation of
the combustion chamber on one side and the gas turbine and HRSG on the other side.

4.2. SECOND rIERATION

Contemplating the effects of changes made in accord with the above list in the values of the
design variables used in the fIrst iteration, the following new values are selected for the
second iteration:
 113

TABlE 2. Values oflhe purchased-i!quipment costs (PEC) and the thermoec~nomic variables for the base design case
(T3 = 850K; T. = 1520K; PIP. = 10; 1'J,c = 1'J" = 0.86)*

Component PEC E E YD c. cp Co Z CD +Z r f
(10· $) [%] [MW] [%] [$/01] [$/01] [$/hI [$/hI [$Ib] [%] [%]

Combustion 0.34 80.37 25.48 29.98 11.45 14.51 1050 68 1118 26.7 6.1
Chamber

Gas Turbine 3.74 95.20 3.01 3.54 14.51 18.76 157 753 910 29.2 82.7

Air Compressor 3.73 92.84 2.12 2.50 18.76 27.80 143 753 896 48.2 84.0

HRSG 1.31 67.17 6.23 7.33 14.51 27.36 326 264 590 88.5 44.8

Air Preheater 0.94 84.55 2.63 3.09 14.51 20.81 137 189 326 43.4 57.9

* For the overall plant we have CP••• = $3617Ib and CI..,., = t, = $ 1451b.

TABLE 3. Comp6nent data for the second iteration case: (T3 =870 K; T4 = 1520 K; .P2IP. =9; 1'J", =85 %; 1'J" =85 % )*.
The values given in parentheses refer to the third itemtion case considered here
(T] =91OK; T. = 1480K; pip. =7; 1'J", =0.83) **

Component E ED Yo c. cp Co Z CD +2: r f
[%] [MW] [%] [$/GJ] [$/GJ] [$/hI [$/hI [$Ib] [%] [%]

Combustion 80.3 25.93 29.77 10.50 13.26 980 72 1052 26.3 6.8
Chamber (81.3) (27.47) (29.92) (9.42) (11.71) (931) (55) (986) (24.4) (5.5)

Gas Turbine 94.9 3.18 3.66 13.26 16.97 152 647 799 28.0 81.0
(94.3) (3.69) (4.01) (11.71) (13.75) (155) (296) (451) (17.5) (65.6)

Air 92.1 2.34 2.69 16.97 23.96 143 546 689 41.2 79.2
Compressor (90.5) (2.99) (3.25) (13.75) (18.38) (148) (324) (472) (33.6) (68.7)

HRSG 66.6 6.40 7.35 13.26 25.60 305 261 566 93.1 46.1
(67.6) (6.10) (6.65) (11.71) (23.51) (257) (284) (541) (100.7) (52.5)

Air 84.7 3.15 3.62 13.26 18.94 150 206 356 42.9 57.8
Preheater (85.6) (4.97) (4.90) (11.71) (16.53) (190) (275) (464) (41.2) (59.2)

Overall 49.1 41.01 47.09 4.57 21.80 675 1922 2597 377.0 74.0
Plant (46.6) (44.79) (48.79) (4.57) (19.06) (736) (1424) (2160) (317.2) (65.9)

* For the overall plant in the new design case we have CP••, =$33551b and CI..,.. =C7 =$ I 571b.
** For the overall plant in this design case we have CP•••, =$29341b and CI..,., =C7 =$167lb.

T3 = 870K, T4 = 1520K (unchanged), P/PI = 9, 11.« = 85%, 11.1, = 85%.

The new values of the thermoeconomic variables for each component are summarized in
Table 3. In the new design. the combusti~n ch~ber, the gas turbine and the air compressor
also have the highest values of the sum CD + Z and are, therefore. still the most important
components from the thermoeconomic viewpoint. The high cost rate associated with the
combustion chamber can be reduced by increasing the values of Tj and T4• In the evaluation
of the cogeneration system we should. however, consider that the value of the combustion
 114

t
chamber will always be the highest among the CD + values for the components of the
cogeneration system.
The gas turbine now has the highestJvalue. The reduction in this value from 82.7% in the
base design to 81.0% in the new design is relatively small compared with a target value of
below 75%. This observation suggests (1) a significant decrease in the values of 11.1"/ and/or
pz /p I - that is, a decrease that is greater than the decrease in these variables in the previous
step: from 86% to 85% and from 10 to 9, respectively, and (2) a reduction in the values of
T 4 • Note that the decrease in 1; value contradicts the corresponding suggestion from the
combustion chamber.
The high values of the exergoeconomic factor J and the relative cost difference r for the air
compressor suggest a decrease in the values of the decision variables PZ/Pl and 11.I.c '
The anticipated increase in the exergetic efficiency of the HRSG (see first iteration) was
not realized because of the interdependence of the components: The reduction in the values
of pz /Pl , 11.1"c' and 11,,/ for the compressor and the turbine leads to an increase in the
temperature differences (and therefore a decrease in the exergetic efficiency) of the HRSG.
Thus, the HRSG thermoeconomic evaluation suggests that the T3 value increases and the T4
value decreases.
The relatively high value ofJin the air preheater suggests a reduction in the T3 value. As
noted in the first iteration, however, changes suggested by the evaluation of this component
should only be considered if they do not contradict changes suggested by components with
a higher value of the sum CD + t.
Summarizing the foregoing suggestions from the thermoeconomic evaluation of each
component, the following changes in the decision variables are expected to improve the cost
effectiveness of the cogeneration system:
• Increase the value of T3 as suggested by the evaluation of the combustion chamber and
HRSG.
• Decrease the pressure ratio PZ/Pl and the isentropic efficiencies 'Il.,< and'll", as suggested
by the evaluations of the air compressor and gas turbine.
• Decrease the temperature T4 as suggested from the evaluations of the gas turbine and the
HRSG.

4.3. TIIIRD ITERATION

To illustrate the effect of the suggested changes in the decision variables on the overall costs,
we use the following new set of values for the design variables:

T3 = 91OK, T4 = 1480K, P2/Pl = 7, 'Il.,C = 0.83 and 'Il,,/ = 0.83.

The results from the thermoeconomic analysis for the last set of values are summarized in
Table 3 by the values given in parentheses. A comparison of the corresponding values shown
in Table 3 demonstrates the improvement in the cost effectiveness of the last design case. As
a result of these ch~nges, the value of the objective function CP,I(J/ is reduced from $3355/h
to $2934/h. The C7 value has increased from $157/h in the new case to $167~h in the. last
case. This increase is, however, outweighed by the decreases in the values of CD,k + Zk'

4.4. ADDITIONAL ITERATIONS

Some additional iterations conducted in a similar way are necessary to further decrease the
value of the objective function and/or establish a nearly-optimal design case. The cost-
optimal values of the decision variables [1] are
 115

T3 = 9 10K, T4 = 1463.0K, pz'Pt = 5.77, l1l'c = 0.811 and 11." = 0.845.

With these values we obtain for the objective function CP•,O, = $28701h. For the cost-optimal
case, the exergetic efficiency of the overall system is 45.0%, the cost rate associated with the
exergy loss is $2051h, and the pinch temperature difference in the heat-recovery steam
generator is 49.7 K.

References
I. Bejan, A., Tsatsaronis, G., Moran, M. (1996) Thermal Design and Optimization, J. Wiley, New York.
2. Reklaitis, G.V., Ravindran, A., Ragsdell, K.M. (1983) Engineering Optimization, J. Wiley, New York.
3. Edgar, T.F., Himmelblau, D.M. (1988) Optimization of Chemical Processes, McGraw-HiI, New York.
4. Papalambros, P.Y., Wilde, OJ. (1988) Principle.• of Optimal Design-Modeling and Computation, Cambridge
University Press, Cambridge.
5. Tsatsaronis, G., Winhold, M., Stojanoff, e.G. (1986) Thermoeoconomic analy.•is of a gasification-combined-
cycle power plant, EPRI AP-4734, RP 2029-8, Final Report, Electric Power Research Institute, Palo Alto, CA.
6. Tsatsaronis, G., Winhold, M. (1984) Thermoeoconomic analysis of power plants, EPRI AP-365I, RP 2029-
8, Final Report, Electric Power Research Institute, Palo Alto, CA.
7. Tsatsaronis, G., Winhold, M. (1985) Exergoeconomic analysis and evaluation of energy conversion plants,
Energy-Int. 1.10, 69-94.
8. Tsatsaronis, G. (1993) Thennoeconomic analysis and optimization of energy systems, Prog. Energy Combust.
Sci. 19, 227-257.
9. Lazzaretto, A. and Tsatsaronis, G. (1996) A general process-based methology for exergy costing, in A.B.
Duncan, J. Fiszdon, D. O'Neal, and K. Den Braven (eds.), Proceedings of the ASME Advanced Energy
Systems Devision, AES-Vol. 36, AS ME, New York, 413-428.
10. Lazzaretto, A. and Tsatsaronis, G. (1997) On the quest for objective equations in exergy costing, in M.L.
Ramalingam, J.G. Lage, V.e. Mei, and J.N. Chapman (eds.), Proceedings of the ASME Advanced Energy
Systems Devision, AES-Vol. 37, ASME, New York, 413-428.
II. Tsatsaronis, G. and Pisa, 1. (1994) Exergoeconomic evaluation and optimization of energy systems, Energy-
Int. 1. 19,287-321.
12. Tsatsaronis, G., Lin, L., Pisa, J., Tawfik, T. (1991) Thermoeconomic design optimization ofa KRW-based
IGCC power plant, Final Report, prepared for the US Department of Energy, Morgantown Energy Technology
Center, DE-FC2 1-98MC2601 9
13. Tsatsaronis, G., Lin, L., Pisa, J., Tawfik, T. (1992) Optimization of an IGCC power plant - Parts I and II, in
R.F. Boehm et al. (eds.), Thermodynamics and the Design, Analysis and Improvement of Energy System..,
AES-Vol. 27, ASME, New York, 37-67.
14. Tsatsaronis, G., Lin, L., and Tawfik, TJ. (1994) Economic evalutation ofa KRW-based IGCC power plant,
Eng. Gas Turbines Power 116, pp. 300-306.
15. Tsatsaronis, G., Krause, A., Lin,and L., Tawfik, T. (1992) Thermoeconomic evaluation of the design of a
pressurized fuidized-bed hydroretorting plant, Final Report prepared for the Institute of Gas Technology and
the Department of Energy, DE-AC21-87MCII089.
16. Lin, L., Tsatsaronis, G. (1993) Cost optimization of an advanced IGCC power plant concept design, in H.1.
Richter (00.) Thermodynamic .• and the Design, Analy.• is and Improvement of Energy Systems, AES-Vol. 10,
ASME, New York, 156-166.
17. Linnhoff, B. et al. (1982) A User Guide on Process Integration for the Efficient Use of Energy, The
Institution of Chemical Engineers, Rugby, Warks.
18. Sama, D.A. (1993) The use of the second law of thennodynamics, in J. Szargut, Z. Kolenda, G. Tsatsaronis,
and A. Ziebik (eds.), Proceedings of the International Conference ENSEC '93, ASME' New York, 53-76.
19. Frangopoulos, c.A. (1994) Application of the thennoeconomic functional approach to the CGAM problem,
Energy-Int. 1. 19, 323-342.
20. Valero, A., Lozano, M.A., Serra, L., and Torres, C. (1994) Application of the exergetic cost theory to the
CGAM problem, Energy-Int. 1.19,365-381.
21. von Spakovksy, M.R. (1994) Application of engineering functional analysis to the analysis and optimization
of the CGAM problem, Energy-Int. 1. 19,343-364.
22. EPRI Technical Assessment Guide (TAG lM) (1991) Electric Power Research Institute, TR-100281, 3,
Revision 6.
ON-LINE THERMOECONOMIC DIAGNOSIS OF THERMAL
POWER PLANTS

A. VALERO, L. CORREAS and L. SERRA

CIRCE - Center of Research for Power Plant Efficiency
Maria de Luna 3,50015 Zaragoza, SPAIN

1. Introduction
Diagnosis is the identification of what is wrong with something that is not
working properly. The main question that any diagnosis methodology must
answer to, when applied to the design or operation of an energy system can be
formulated as follows:
Where, how and which part of the consumed resources can be saved by
keeping the quantity and specifications of the final product constant.
Thermoeconomic diagnosis is the only Second Law based technique devoted
to operation analysis. The exergy balance of an installation allows us to allocate
and calculate irreversibilities in the production process and to identify the
equipment which affect overall efficiency and the reasons thereof. This
information, although useful, has proved not to be enough. In practice, when
attempting to achieve energy savings in an installation, we must consider that not
all irreversibility can be avoided. Thus, the technical possibilities for exergy
savings are always lower than the theoretical limit of thermodynamic exergy
losses. Moreover, the local exergy savings which can be achieved in the different
units or processes of an installation are not equivalent. The same decrease in the
local irreversibility of two different plant components leads, in general, to
different variations of the total plant energy consumption.
At this point it is necessary to introduce a concept that will help us to relate
causes of irreversibility with their effects at a local and a global level. This is
attained by means of a representation of the plant, not like its physical structure,
composed of equipment that exchange material and energetic flows, but as a
function of the objective that each component accomplishes in the utility. This
representation, commonly called productive structure, contains the information
about component interaction - or groups of components - by means of
exchanging resources in order to obtain a certain product.
117
A. 8ejan and E. Mam"t (eds.), Thermodynamic OptimiVUion o/Complex Energy Systems, 117-136.
@ 1999 Kluwer Academic PublishLrs.
 118

From this point we will be ready to know how any malfunction in the plant
affects the additional consumption of resources, which is the final objective of the
thermoeconomic diagnosis.
This paper first introduces the theory behind thermoeconomics and then
explains an actual diagnosis system that has been developed according to the
principles outlined.

2. Theoretical Background

2.1. EXERGETIC COST ANALYSIS

For a given physical system of components and flows, thermodynamic
properties such as mass flow rate, pressure, temperature, composition, enthalpy,
entropy can be determined. Using these properties and choosing a reference or
dead state, the energy and exergy flows for the system can be found. If these
flows are considered to be fuels, products or losses of the system's components
(where the difference between fuels and products represent losses), the efficiency
of each system component can be defined as the ratio of its products to its fuels.
These are the basic concepts used to analyze an energy system with energy
and exergy balances. An energy balance indicates how the fuels' energy is
distributed throughout the system as well as the location of losses which make
the energy of the component products lower than that of their fuels. These energy
losses, however, only represent losses from a component or the system and give
no indication of losses due to the degradation of energy (irreversibilities) within
each system component. In order to account for these losses, exergy balances are
needed. These balances account for the quality of energy present and allow one to
determine where and how much energy could be theoretically saved within the
system itself.
A further enhancement of the more conventional analysis described above is
the introduction of exergetic costs. These costs take into account the fact that the
technical possibilities for exergy (and energy) savings are always lower than the
theoretical thermodynamic limits and that the same decrease in local
irreversibilities for two different system components leads, in general, to different
variations in total system fuel consumption. Since exergy is a departure from
equilibrium with the environment or reference state [I], it represents the
minimum quantity of work needed to produce a product from the reference state.
Thus, the exergy of a physical flow is a potential which does not depend on the
process used to obtain this flow. Because real processes in energy systems are
only possible with irreversibilities, the exergy of the resources of a component is
always greater than that of its products. In other words, the exergy needed to
produce a physical flow, i.e. the exergetic cost [1, 2], is always greater than the
 119

exergy which the flow contains. Minimizing the exergetic cost per unit of exergy,
i.e. the unit exergetic cost of a flow, is a tool which can be used in the
thermodynamic analysis and/or optimization of a system's operation. This unit
cost for the product flow of a subsystem (or component) is set by both the
subsystem's (or component's) internal irreversibilities and by the irreversibilities
of the subsystems upstream. Thus, exergetic cost theory permits one to analyze
how each subsystem is exergetically interconnected with other subsystems and
explicitly determines, i.e. quantifies, which irreversibilities have the greater
impact on the system's exergy efficiency, something not explicitly apparent from
an exergy analysis alone.
A set of rules developed by Valero et al. [I, 2] allows one to rationally and
uniquely calculate all the product exergetic unit costs as well as exergetic costs
for all the subsystems or components which make up a system. They are as
follows:
1. the exergetic cost is a conservative property;
2. all costs generated by the productive process must be included in the cost of
the final products so that in the absence of external assessment, system losses
to the environment have a zero exergetic cost;
3. in the absence of an external assessment, the exergetic cost of each flow
entering the system is equal to its exergy;
4. if the output flow of a unit is part of the fuel of this unit, then it is understood
that its unit exergetic cost is the same as that of the input;
5. if a component has a product composed of several flows, then the same unit
exergetic cost is assigned to all of them;
6. if a component produces a residue, then its unit exergetic cost is negative and
equal to the amount of external resources needed to dispose of it per unit of its
exergy content.
Rule 1 can be justified by considering a subsystem with i fuels, j products and
k losses to the environment. Applying this rule to the subsystem gives the
following exergy and exergetic cost balances:
LF; = LlJ + Llk
j k
(1)

Lk;) . F; = Lk;j . P
. j
j (2)

where the k* are unit exergetic costs values; F, P, I the exergy values of fuel,
product, and irreversibilities respectively. The product amounts are set by
downstream subsystems and losses to the environment while the exergetic unit
costs of the product flows exiting the subsystem are set by both the subsystem's
internal transformation process of fuel flow exergies and the fuels' exergetic unit
 120

costs. The latter are set by the irreversibilities of upstream subsystems. The
transformation process depends on the internal irreversibilities of the subsystem
in question. Since both input as well as transformation costs are directly
transferred to the product flows, there is no internal cost generation. Exergetic
cost is, thus, a conservative property.
To perform an exergetic cost analysis, it is useful to define a productive or
causal structure, the counterpart to the physical structure used to calculate the
energy and the exergy flows of a system. A productive structure is a schematic
representation of the plant based on the fuel-product concept. Flows that connect
components to one another are component fuels and products but not necessarily
physical flows. When defining a productive structure, component fuels and
products must be defined in terms of exergy.
Productive and dissipative units, junctions and bifurcations (branches») are
the elements which make up a productive structure. Productive units have the
capacity for energy transformations which consume resources and produce one
or more energy products required by a system's environment or by another unit.
The goal of dissipative units is preparing a product, disposing of wastes or
interacting with productive units in order, for example, to improve system
efficiency, as is the case for the condenser2.
The thermoeconomic model (mathematical representation of the productive
structure) consists of a group of characteristic equations which express for all the
components appearing in the productive structure each inlet flow as a function of
the outlet flows and a set of internal parameters modeling the component
behavior, i.e.:
Unit j: F. = Ki] • Pj (3)
Junction j: F. = rij • Pj (4)
Branching point j: F j = L P; (5)

where K has the meaning of exergy consumption of the unit, and r is the
equivalent parameter in the case of junctions, called the exergy ratio.
Considering the representation of the Exergetic Cost Theory, explained in [2,
3,5], the characteristic equations can be expressed for unitj as

) Junctions and bifurcations are fictive components unless they happen to correspond to real components such as
flow mixers and splitters. However, mixers should be considered as units rather than junctions as there is a
degradation of exergy in the mixing process, resulting in an increase of entropy between mixers' input and output
flows.
2 The condenser permits the system (steam turbine cycle) to effectively comply with the Second Law of
thermodynamics by transferring heat (with entropy as the carrier) to the environment in an efficient manner (i.e. at
reduced pressure). This function of transferring heat to the environment is equivalent to the selling of negentropy
by the condenser to the other components in the steam cycle turbine [4].
 121

(6)
n
Pj = OJ j + L q ji . F; (7)
i=l

where kj is the unit exergetic consumption of component j, cq is a product that

exits the plant, and qji, called production coefficients, are combinations of the rji
coefficients. Eq. (6) infonns about
• the productive function of each component, i.e. what is its production,
• what does the component need (fuel) to develop its productive purpose, and
• the thennodynamic efficiency of the process taking place in the component.
Equation (7) is the structural equation, that contains how the resources
consumed by the plant are distributed through the plant components, i.e. it
infonns about how components are connected since a productive point of view.
In vectorial fonn, Eq. (6) and (7) transfonn into:
F = KD·P (8)
P = Ps + < PF > F (9)
The diagonal matrix KD (n x n) contains the unit exergetic consumption of the
productive and dissipative units, <PF> is a matrix (n x n) containing the
coefficients qji from Eq. (7), and P s are the products leaving the plant, expressed
as vector.
Given that each unit and each junction has a primary product flow, the total
number of these flows equals the total number of internal products whose unit
exergetic cost k; must be calculated.
Knowing the costs of the external resources k: and applying the chain rule of
derivatives to the set of characteristic equations the following system of
equations is obtained:

Unitj:k;.j = L);.i ·ki (10)

Junction j: k p*
.J
= ""
4...J
k p* .'. • rI).. (11)

Branching point j: k ;.J =k ;,j (12)

from which the cost of all flows in the productive structure can then easily be
calculated [7, 8].
 122

2.2. MALFUNCTION AND IMPACT ON FUEL ANALYSIS

The cost calculations outlined in the previous section can be used to analyze the
given state of a plant. However, when comparing two different states, such as
those for design and off-design conditions, a better comparison can be made by
performing a malfunction and impact on fuel analysis. This type of analysis, also
known as the theory of perturbations [7, 9], determines how the costs of the
resources vary when the internal parameters of an installation, the unit costs of
resources consumed or the external demand vary.
For example, let us consider a productive structure with E external resources
consumed, N productive and dissipative units and J junctions. Each primary flow
is assigned a branch, from which the flow is distributed. Thus, the number of
branches is equal to M = E + N + J
Furthermore, the exergetic cost of the fuel used by a plant can be expressed as
E
Co=Lk;.Fej (13)
j=1

Applying Eq. (8) and Eq. (9) to Eq. (13), and deriving by the internal
parameters of the units, x, we get:
oC O _ tk • oKn tk. o<PF>
ox - F ox P+ P ox F (14)
Also a equivalent formula in terms of product can be written:

oC o = (tc
ox e
oKn +tk*
ox P
i« PF>
ox
Kn)].p (15)

And in terms of the unit exergy consumption:

ceo
-=kp PJ
• (16)
OK. ,1
I)

It means that the increase of resources cost due to a variation of the efficiency
or the technical production coefficients is proportional to the cost of fuel used to
obtain the product, and to the production of component,
So, we can express the additional cost incurred due to an infinitesimal
perturbation in the state of the plant as:
e n s
dCo = L~(Ce,idki +kidce,i)+ L k;,iPjdKij + Lk;.n-s+idmi (17)
i=1 i.j=1 i=1

Here we like to focus the problems related with thermoeconomic diagnosis of

the operation of thermal systems. That is, we wish to compare the real cost of the
energy resources consumed with that which would have applied if the plant has
 123

the same external demand in design conditions. Therefore let d~ = O. We will

also assume that de e,l. = O. So previous expression is simplified to:

dC o= ~(e
L..
j=1
o.++ ~ kp*dK)P
e,1.dK ,J L..,I
i,j=1
IJ J (I8)

This equation allows to separate the additional fuel consumption dC0' into the
contributions of each unit and even each production coefficient. The performance
test to be carried out on the installation should provide the actual values of the
production coefficients, so we can calculate their deviation with regard to the
design conditions, In a unit, the positive deviation of the productive coefficients
dk..P , called malfunction, implies an additional fuel consumption to produce a
IJ J
given quantity of product.
The preceding Eq. (18) quantifies the system perturbations which affect fuel
consumption. Together with the exergetic cost balances, Eq. (8), they form the
basis for a diagnostic thermoeconomic analysis of an existing system operating
under off-design conditions.

2.3. THERMOECONOMIC OPERATION DIAGNOSIS

Cost accounting is the technique that records, measures and reports about how
much things cost. It has already been a long time since cost accounting became a
recognized profession and today many techniques are applied to calculate costs
of products and services, as well as to estimate the deviations from the pursued
targets. Decision making at every level is always supported by an evaluation of
the benefits for the organization.
In a parallel manner to monetary cost accounting, the energy cost accounting
has been developed for systems in which energy plays an important role.
Thermoeconomics, the branch of Thermodynamics that has adapted concepts of
Economics, consists of a set of techniques devoted in general to energy saving,
among which we remark operation diagnosis.
The thermoeconomic diagnosis of the operation of an energy system,
explained and applied to actual utilities in [7] and [9], consists of identifYing the
degradation in the performance of the system and the assessment of its effects in
terms of additional fuel consumption.
The application of the thermoeconomic diagnosis to complex plants becomes
a very useful tool for the management, plant engineering, operation and
maintenance departments. Moreover, if this tool has been integrated with the
plant information management system, continuously processing plant data and
spreading the generated information to the different users, the enhancement
potential and derived benefits amplifies in several orders of magnitude.
124

Generally speaking, the possible applications of the thennoeconomic

operation diagnosis system are outlined as follows:
• On-line operation costs, and economical assessment of the departure from the
best attainable operation.
• Optimization of operation strategies as a function of the market's demand of
electricity.
• Economical evaluation of corrective actions (opening turbines for
maintenance, compressor cleaning, soot blowing)
• Objective assessment of the cost (not the price) of every product and
subproducts of the plant.
• Assessment of operators efficacy.
• Continuous instrument monitoring.
Depending on the end user, and hence, on the decision level to which the
diagnose is referred to, the system has to be customized. That is, an operators
manager needs certain infonnation and with a different frequency than a plant
manager or a maintenance technician. It is important to remark that the system
remains the same, only the way to present the results changes.

3. Description of the System

Valero and coworkers have already been installing and developing two real-time
diagnosis systems, whose aim is to demonstrate the practical application of the
Thennoeconomic Theories.
The first one was installed in Endesa's Teruel conventional pulverized-coal
power plant, and focused its man-machine interface to operators, as explained in
[10]. The scope of the second system covers not only operation but also will
adapt to other potential users such as plant management or maintenance
department. The application is currently being tested at the 330 MW-rated
Combined Cycle located at Puertollano and will be expanded to cover the whole
IGCC facility in the next phase. Puertollano IGCC is owned and operated by
Elcogas, a consortium of the principal european producers. The Project has been
supported by the EU Thermie Program for Clean Coal Technologies.
The design of the thennoeconomic diagnosis system has been accomplished
in a modular fashion. The system requires in a general case four modules, which
are depicted in Fig. 1:
1. Data Processing Utility: in connection with the plant data acquisition system,
in charge of real-time data retrieval, checking against validation rules and
historical data storage.
 125

2. Perfonnance Tests: a procedure, nonnally based on perfonnance test codes,

that determines the actual state of the plant with the higher attainable accuracy,
with regard to the available instrumentation.
3. State of Reference: a validated model of the plant which represents the state of
reference for any operation mode, ambient conditions or feedstock compositions.
4. Thennoeconomic Diagnosis Model: that allocates and assesses the increased
consumption of resources compared to the one predicted by the state of reference
and explains the underlying causes.

THERMOECONOMIC
DIAGNOSIS

STATE of
PERFORMANCE
REFERENCE
TEST
MODEL

Figure 1. Modular design of the thennoeconomic diagnosis system.

Given a steady-state period, characterized by a certain operation mode, a set of

raw instrument data are checked for consistency, validated and processed. These
infonn about the actual state of the plant, and allow to quantitative analyses
through the following successive steps:
• Comparison with a reference case (target) obtained with the same operating
conditions.
126

• Identification of inefficiencies or malfunctions (performance degradation) of

sub-systems or components of the plant.
• Evaluation of the causes for the component malfunctions.
• Assessment of the extra-operating cost due to malfunctions, and the cost
impact of appropriate maintenance actions.

3.1. DATA PROCESSING UTILITY

The first module performs the tasks concerning connection to the data sources,
and retrieval and filtering of data sets that will be processed by the Performance
Test module. Basically, it can be decomposed in:
1. Data source access, continuous performance monitoring and data sets
retrieval.
2. Preprocessing and validation.
3. Data storage.
I I
SJ
r ___
1~
Plant
Management

1
Quenes

Auto~8ted
Reports
"t
ThermoecOflolT1lc DIagnoSIs

(~~
Filtered Valu&S Cont5J'rMion
.1 Dilll:lbese

~~dotion~ ,".,"m~' OJ O1aracl'l;lnzabon

-f .,-
Historic
RDBMS -
-------,--- Mean Values (5')

L-c;~~=:~eon--::L.
'-"
Slandard---.i'continuOLl~ 'IE-- 30l --
ExWnal
Interface ~anitorinv ODBC- Datab.ses

T
CommunlC:atlons
Protocol

Figure 2. General architecture of data processing utility.

Main data sources usually are plant acquisition systems, or even the
distributed control systems themselves, that perform a propietary
communications protocol. Interfacing directly with them implies developing a
tailor-made layer to interact. However, there are some commercial applications,
 127

database engines actually, that have already implemented various specific

propietary communications protocols and allow standard queries through SQL or
ODBC support, thus avoiding the development of a particular solution. Other
existing databases (feedstock analysis, meteorology, emissions) should also be
accessed.
The data retrieval tool must continuously monitor the plant performance,
because it is necessary to discriminate steady states, where it makes sense to
apply the thermoeconomic diagnosis. Surveillance of instantaneous values of the
main variables determine stability, as a function of their high-frequency
variability, as shown in Fig. 3. For a Combined Cycle, main variables are power
production of both steam and gas turbine.

~i
,I,. ~.)-:;:~....:.
;,;;;,_::;Ii:;;';;;J~'
.l.,j

i~~1ifiI'tl· ~!
Steam Turbine PtY.¥er !Minimum . \

-.--~j\j\

steady state Period

.!

Figure 3. Steady-state period detection.

It is important to remark that raw data from plant instrumentation are

measured values, and hence, not actual values. A certain accuracy and precision
must be taken into account, depending on the instrument and its calibration,
which may degrade as time goes by.
Mean values for raw data are generated over a period (every 5 to 10 minutes
typically), and then undergo a series of tests that determine departures on actual
instrument readings from the expected ones. First, every measurement is
compared against the historically most probable value, as a function of operating
mode.
Moreover, relationships between measurements allow to build validation
rules. These involve some comparison between instrument readings, like e.g. a
difference in pressure measurements upstream and downstream a certain
128

equipment that usually is 1,3 bar, or the power consumption of a certain pump
which is proportional to plant load. Validation rules must be particularly
formulated for every utility, depending on available instrumentation. The most
important feature is their sensibility to detect biases in single instruments.
As a result of these tests, the measurements that have failed any of the
validation rules are marked as suspicious and recorded. Depending on the
severity ofthe departure from the expected value, like a null value, the data could
be replaced by their most probable value for the actual operating mode, so as not
to introduce a clearly inconsistent input to the calculations thereafter. At this
point, the preprocessed data set is ready to be sent to the Performance Test
module.

Figure 4. Instrumentation monitoring.

The Diagnosis system features a historical database complementary to the

real-time database from which is eventually fed, that is not reliable for long term
storage. One must remark that typically control systems handle rates of data of
over some thousands of measurements refreshed at a frequency of some seconds.
Actually, not all this information is relevant in the long term, and thus can be
somewhat volatile. But it is necessary to record all the processed information, in
order to conduct analysis of tendencies, very useful to maintenance or operation,
or for energy cost accounting with an adequate periodicity for a financial
department. In this sense, a complete record for steady-state as well as for
 129

transient periods, containing all the necessary information characterizing it, IS

stored in a relational database.

Figure 5. Relationships between measurements.

3.2. PERFORMANCE TESTS

The Performance Test module assures the calculation of a complete set of
equipment parameters that reflect the actual state of operation. There are mainly
two possible strategies:
• Conventional: based on the election of a minimum set of reliable instruments,
generally normalized by recognized International Codes, such as ASME, ISO
or DIN. It is not a matter of this paper to deepen in the widely known
Performance Test Codes for gas or steam turbines, steam generators, or
HRSGs, [11-15].
• Data reconciliation: applies error optimization techniques [16, 17], in order to
take advantage of all the available plant information. There are usually some
instrumentation redundancies, normally very little, but anyway useful so that
combining all measurements, weighted with an estimation of their accuracy,
the global uncertainty decreases. This method requires an exact knowledge
and monitoring of all available instrumentation: instrument type, measuring
conditions, signal processing.... Combining these factors yields an estimation
 130

of the individual uncertainties. A further fine-tuning of the estimations, based

on the operating experience, is however necessary.

Figure 6. Perfonnance test results.

Global plant performance monitoring calculated by means of a direct method

does not require a big amount of input measurements, that on the other hand are
usually reliable enough (e.g. net power and natural gas mass flow). Thus, a data
reconciliation procedure is not justified in this case.
However, taking into account that the thermoeconomic diagnosis tries to find
out a causality, equipment by equipment, from a doubtful instrumentation, is
very convenient to minimize error propagation through a more robust calculus
procedure, that does not rely on just the exact amount of information to solve the
energy and mass balances. In the case of a steam generator, gas mass flow is not
measured and thus should be obtained from an indirect method. The more
information, the more accurate result would be obtained.
Another advantage of data reconciliation techniques is the open equation
procedure implicitly implemented to solve the constraints and energy and mass
balances, that permits the developer to focus in the set of constraints itself and
not in the converging criteria.

3.3. STATE OF REFERENCE MODEL

For a given operating conditions, the Diagnosis system requires two set of values
that characterize the thermodynamic behavior of the plant components: a so
 131

called state vector, from the validated measurements, and another one with the
state of reference of the plant. The first one is obtained from the Performance
Test module, and the second from a model that provides the State of Reference,
that is, the better possible operation of the plant for a given conditions.
The State of Reference model consists of a model of the plant, that emulates
the physical behavior of the components for every usual operating mode and
react against changes in boundary conditions, such as ambient conditions or
feedstock composition.
The model has been elaborated by means of an open-equation procedure with
non-linear constraints, which is solved as a whole with a quasi-Newton
algorithm. This method assures a perfect coherence of energy and mass balances
and composition, and a fast convergence, because the initial guesses, taken from
the results of the Performance Test module, are sufficiently close to the solution.
As a first approach, the model is based on design data, but has to be validated
against actual data of the considered optimum plant performance. Generally,
Acceptance Tests set a very adequate reference to compare with.

3.4. EXERGETIC COST AND IMPACT ON FUEL ANALYSIS

As it was already mentioned, the analysis of the interactions between the various
components of a utility develops into the so called productive structure. This
contains the production objective for each component (e.g. heating or evaporating
water, increasing gas pressure, ... ) related to all resources consumed (e.g.
temperature decrease of another fluid, electric power, pressure losses to
overcome resistance), and to other secondary effects, the ones that are somehow
beneficial (subproducts, like e.g. temperature rise when compressing a gas), or
those that need even more resources to be disposed or eliminated (residuals, like
e.g. S02 from combustion or entropy increase in a steam cycle).
The productive structure of a combined cycle is showed in Fig. 7, as an
example of the great complexity that arises in any real utility.
Exergetic Cost Theory allows us to calculate exergetic costs not only for
every product going out of the plant, but for every stream or every process that
takes place, as it can be seen in Fig. 8.
Monetary costs, even taking into· account pay-off and O&M costs, are
calculated easily, as a direct application of the Thermoeconomic Theory.
 IN
-
TOHRSG~ N

DeNOx
Steam ...

Auxiliary
Systems
To Steam
Cycle
Steam Turbine
Net Power RHTR
+ CooIingAir~r
\
---l'LP SHTR

Imported
r - - - - Steam

To COnd Pump
Flash Tank &
Seal Steam
Condenser

Flash Tank

Figure 7. Productive structure of the combined cycle.

 133

Figure 8. Cost calculation results.

The assessment of the higher specific consumption compared to the reference

case is obtained calculating impact on fuel at the component level, from which
interpretation the adequate corrective actions should be adopted. Fig. 9 shows
impact on fuel dis aggregated by components for two actual cases, where a deep
evaluation of two operation strategies is conducted.
This working case compares two slight different operation strategies, based
on a higher ratio of fuel gas to steam for DeNOx abatement, and the set point at
a heater inlet (changed from 100° C to 60° C). Moreover, ambient conditions
varied from one state to the other, what affected a different condenser pressure.
The disaggregated results revealed the following effects:
• Set point change, that evidently enhanced heat generator efficiency, implies
saving 0,4 MW.
• Energetic cost, and hence economic, of requiring a higher steam consumption
for DeNOx abatement has been assessed as 2,2 MW. This allows an
economical optimization that meets adequate emission levels and maximal
efficiency.
• The biasing effect of comparing two states in two slight different ambient
conditions is isolated and quantified. A worse condenser vacuum affects the
low pressure steam turbine, assessed as 1,9 MW.
 134

-1000 o 1000 2000 3000 4000 5000

Impact on Fuel (MW)

Figure 9. Disaggregated impacts for a comparison of two operation strategies.

Generally speaking, the system offers a wide variety of possible applications,

as mentioned above. Is the final user who determines the scope of the tool, that
basically must satisfy his needs of organizing the adequate information at the
corresponding level. Evidently, already cost increase allocation and
quantification at a component level is a step ahead the systems considered state-
of-the-art.

4. Results and Conclusions

This paper has presented the motivation of applying a on-line thermoeconomic
diagnosis system to enhance energetic resources utilization and thermal power
plant management, as well as the main features of the system. Two relevant
conclusions must be outlined.
First, quality of mput data must be absolutely assured, by means of a
measurements monitoring system and historical data analysis supported by an
adequate data base. Moreover, it has been demonstrated that data reconciliation
techniques are the most suitable for calculations based on process signals, that
have not necessarily a good quality. This characteristic is common to all tools
that are fed with measurement data, and hence the application needs certain
techniques, which have been developed in recent years.
Second, the methodology that supports thermoeconomic diagnosis is unique,
because of having a different approximation than other applications.
 135

Thermoeconomic approximation is based on a priori analysis, proactive, that is,

it previews the phenomena to be diagnosed. The fundamental feature consists of
the detailed analysis of the equipment interactions. The systematic cause-effect
relationships are materialized into the productive structure. Once given confident
input data, malfunction causes and quantification are obtained.
The practical implementation at the Combined Cycle has demonstrated that
thermoeconomic diagnosis is becoming a promising industrial tool that integrates
the technician's point of view with the interest on enhancing utility's benefits.
For the research team at CIRCE, the thermoeconomic diagnosis system
implies successfully reaching the summit of a research field that began more than
ten years ago, with the development of the Exergetic Cost Theory· and the
Structural Theory, theories that could become industrial practice.

Acknowledgments. The authors would like to acknowledge the management of

Elcogas and of Endesa's Teruel power plant for their faithful collaboration at
every moment, specially to Dr. Manuel Trevino, chief executive officer of
Elcogas, and Dr. Javier Pisa, head of the Technology Group at Elcogas, for his
constant and active support to this idea. Last but not least, we acknowledge all
the coworkers and researchers within the development of the thermoeconomic
diagnosis, log. Angel Martinez, log. Pedro Casero, log. Luis M. Blasco, Dr.
Scarpellini and the collaboration of Dr. Miguel A. Lozano, Dr. Javier Collado
and log. Juan L. Bartolome.

5. References
I. Lozano, M. A. (1987). Metodologia para el analisis exergetico de calderas de vapor en
centrales tennicas, Doctoral Thesis. Universidad de Zaragoza, Zaragoza, Spain.
2. Lozano, M. A. and Valero, A. (1993). Theory of the exergetic cost, Energy Vol. 18, No.
9, pp. 939-960.
3. Valero, A., Munoz, M. and Lozano, M. A. (1986). A general theory of exergy saving,
ASME. AES, Vol. 2-3 Computer-Aided Engineering and Energy Systems Vol. 3:
Second Law Analysis and Modeling, pp. 1-21 Editors: RA. Gaggioli Book No. H034 I C
4. Lozano, M. A. and Valero, A (1993). Thennoeconomic analysis of gas turbine
cogeneration systems, ASME. Winter Annual Meeting Symposium on thermodynamics
and the design. analysis and improvement of energy systems. November 28-December 3,
Session IT General Thennodynamics & Energy Systems
5. Valero, A., Lozano, M. A, Serra, L. and Torres C. (1994). Application of the exergetic
cost theory to the CGAM problem, Energy Vol. 19, No.3, pp 365-381.
6. Torres, C. (1991). Exergoeconomia simb61ica. Metodologia para el analisis
tennoecon6mico de los sistemas energeticos, Doctoral Thesis, Universidad de Zaragoza,
Zaragoza, Spain.
7. Lozano, M. A., Bartolome, J.L., Valero and A., Reini, M. (1994). Thermoeconomic
diagnosis of energy systems, Flowers 94, Florence World Energy Research Symposium.
July 6-8, Florence, Italy.
136

8. Valero, A., Serra, L. and Lozano, M. A. (1993). Structural Theory of Thennoeconomics,

ASME. Winter Annual Meeting Symposium on thermodynamics and the design, analysis
and improvement of energy systems. November 28-December 3, Session n General
Thennodynamics & Energy Systems.
9. Schwarcz, P., Lozano, M. A., Von Spakovsky, M.R. and Valero, A. (1997). Diagnostic
Analysis Of A PFBC Power Plant Using A Thermoeconomic Methodology. TAIES'97:
Thermodynamic Analysis and Improvement of Energy Systems Proceedings of IntI.
Conference JWle 10-13, Beijing, China Eds. Cai, Ruixian, M.J. Moran, Zhang, S., Xiao,
Y. World Pubs. Corp. Beijing, China, ISBN 7-5062-3264-Znx.6 pp. 240-249.
10. Valero, A., Lozano, M. A. and Bartolome, J.L. (1995). On Line Monitoring Of Power
Plants Performance, Using Exergetic Cost Techniques. Exergoeconomical Analysis and
Optimization in Chemical Engineering, EUROTHERM Seminar No. 41, July 17-18,
Aachen, Germany.
II. ASME (1964). Performance Test Codes: Steam Generating Units, PTC 4.1. The
American Society of Mechanical Engineers.
12. ASME (1970). Performance Test Codes Report: Simplified Procedures for Routine
Performance Tests ofSteam Turbines. The American Society of Mechanical Engineers.
13. ASME (1976). Performance Test Codes: Steam Turbines, PTC 6. The American Society
of Mechanical Engineers.
14. ASME (1981). Performance Test Codes: Gas Turbine Heat Recovery Steam Generators.
PTC 4.4. American Society of Mechanical Engineers.
15. British Standard (1989). British Standard Specification for Gas Turbine acceptance
tests. ISO 2314.
16. Stephenson, G.R. and Shewchuk, C.F. Reconciliation of Process Data with Process
Simulation (1986). AIChE Journal, Vol. 32, No.2, pp. 247-254.
17. Heyen, G., Marechal, E. and Kalitventzeff, B. (1996). Sensitivity Calculations and
Variance Analysis in Plant Measurement Reconciliation. Computers Chemical
Engineering Vol. 20 Sup. S539-S544.
EXERGY IN THE THERMAL SYSTEMS ANALYSIS

J. SZARGUT
Technical University of Silesia
Institute of Thermal Technology
Konarskiego 22, 44100 Gliwice, Poland

1. Introduction

Exergy analysis has been developed as a result of application of the second law of
thermodynamics to the investigation of the thermodynamic imperfection of industrial
thermal processes. Precursors of exergy analysis were Gouy [1] and Stodola [7]. They
formulated independently the law determining the loss of ability to perform work due
to the thermodynamic irrevesibility. The term exergy has been introduced by Rant [4]
whose works initiated a very intensive development of this field of technical
thermodynamics. In sixties the exergy analysis has been developed mainly in Europe,
in eighties the interest in this analysis increased distinctly in America. The bibliography
of exergy analysis contains thousands of papers, among them many monographic books.

2. Definition of Exergy and Exergy Losses

Various kinds of energy differ in their quality characterized by their ability to perform
work. Exergy is a measure of quality of various kinds of energy. It has been defined as
a maximum ability to perform work in the existing natural environment. Very clear
definition of exergy has been proposed by Riekert [5], with small supplementation of
the author of present paper:
Exergy expresses the minimal consumption of work necessary to produce the
investigated material with required parameters from commonly appearing components
of the surroundings nature, in reversible processes, using the environment as a single
source of heat.
Similarly as in the case of energy, it should be distinguished between the exergy of
matter crossing the control boundary of the system and exergy of matter contained
within the boundary (difference similar to that between internal energy and enthalpy)
[8].
Exergy does not fulfill the law of conservation. Every irreversible process causes an
irrecoverable loss of exergy tlB. which can be determined by means of the Gouy-
Stodola law:
(1)

137
A. Bejan and E. Momut (eds.), Thermody1lDl1lic Optimizalion o/Complex Energy Systems, 137-150.
© 1999 Kluwer Academic Publishers.
138

where To is the environmental temperature and l:: t!..S is the sum of entropy increases
of all kinds of matter taking part in the process.
As stated Sama [6] an exergy loss has to be accepted in order to ensure a limited
value of the investment expenditures. E.g. a limited area of heat transfer in a heat
exchanger can by attained, if there exists some temperature difference between the
heating and heated fluid in every point of the heat transfer surface. But every
temperature difference during the heat transfer evokes some exergy loss. Hence exergy
loss should always have some economical justification. If such a justification does not
exist, it indicates that the exergy loss results only from an error in engineering art.
On the other hand, every exergy loss causes some decrease of the useful effect of the
process or an increase of the consumption of driving means (if the required useful effect
is given). Hence, the exergy loss indicates always the possibility of thermodynamic
improvement of the process, but the profitability of such an improvement should be
checked by means of an economical analysis (because is usually connected with some
increase of investment expenditures).
Main causes of exergy losses are: friction, heat transfer with finite temperature
difference, diffusion, combustion. The lower is the temperature in the process, the
greater is the deleterious influence of irreversibility. E.g. in processes realized below
the environmental temperature exergy loss due to the friction is greater than the heat of
friction. The analysis of exergy losses is not easy, because the change of irreversibility
in some link of the system evokes changes of exergy losses also in other links of the
system.
The exergy losses can be divided into internal and external ones. Internal exergy
losses are connected with the internal,parts of the system, appear inside the control
boundary of the system and can be easily calculated by means of Eq. (1). External
exergy losses result from the discharge of waste products of the process to the
environment. Exergy of the waste products is destroyed in the environment. Hence the
external exergy loss equals to the exergy of the considered waste product. The
calculation of external exergy loss by means of Eq. (1) is difficult, it is easier to
determine it by mean of the exergy of waste product. The appearance of external exergy
losses indicates that the primary causes of exergy losses act often far from the place of
appearance of exergy loss. Often it is difficult to decide, which is the cause of exergy
loss. E.g. the external exergy loss with combustion gases flowing out from the boiler
can be understood as a result irreversible phenomena inside the boiler or as a result of
the lack of utilization after the boiler.

3. Exergy Balance and Exergy Efficiency

Exergy is exempt from the law of conservation, so the exergy balance must be closed
by the internal exergy loss:
(2)

where Bd is the exergy delivered to the system with the streams of matter, Bau,Baw
 139

denote, respectively, the exergy carried out from the system with useful products and
waste products, A.Bs is the increase of exergy of the system, A.Bq denotes the increase
of exergy of the heat source being in contact with the system, L is the work performed
by the system and 6B i denotes the internal exergy loss, due to the irreversibilities inside
the system.
Exergy carried out with waste products expresses the external exergy losses. Increase
of exergy of the external heat source can be determined by means of the Carnot-factor:

T-T.
A.B =_Q _ _o (3)
q T

where Q is the heat taken from the external heat source and T is the temperature of
heat source measured on the system boundary.

Heat Sourc.
~~T

Figure 1. Band diagram of the exergy balance.

For a steady-state process in which flow velocities, chemical compositlons and

thermal parameters are constant at all points of the system, exergy of the system
remains constant and exergy balance can be related to the time unit:

Bd=Bau +Baw +:EA.Bq +£+6B.I (4)

where the dotted terms denote quantities per time unit.

Exergy balance can be presented in the form of band diagram, Fig.l [2]. The width
of the bands is proportional to the exergy values. The band of the internal exergy loss
has a width increasing inside the system boundary from zero to the value resulting from
Eq.(l).
Ratio of the useful exergetic effect to the consumption of driving exergy expresses
the exergetic efficiency called also rational efficiency or effectiveness:
 140

(5)

where Bdf,Bdr denote, respectively, the exergy of delivered driving materials (e.g.
fuels) and of nondriving raw materials, t:.Bsu is the useful exergy increase of the
system, -t:.Bqf' t:.BqU denote, respectively, the exergy decrease of the source of driving
heat and increase of the serviced source of heat, -Lf,Lu are, respectively, the driving
work and useful performed work.
Exergetic efficiency is always smaller than one. In some processes its value is
negative. In such a case the degree of thermodynamic perfection "p can be calculated
in order to characterize the irreversibility of the process. It can be defined as a ratio of
the carried out useful exergy to the delivered exergy. When calculating "p the transit
exergy should be omitted. It comprises the exergy components not taking part in the
thermodynamic changes (e.g. the chemical exergy of the solvent when analyzing the
chemical processes in solutions).
The main task of exergy analysis is to detect and quantify the causes of
thermodynamic imperfection. The mentioned causes are characterized by the connected
exergy losses.
Szargut and Sama [16] elaborated 20 practical rules facilitating the reduction of
exergy losses. The mentioned rules take also into account the mutual influence of the
irreversibility of processes appearing in the large network.
Influence of exergy losses appearing in particular parts of the process on its useful
effect can be examined by means of the coefficients of sensibility:

(6)

where Bu'''B are, respectively, the useful exergy effect and exergy efficiency of the
process, oBk'£k are, respectively, the total and relative exergy loss in the k-th part of
the process.

4. Economic Applications of Exergy

Exergy losses appearing in particular links of the analyzed system are not of the same
importance. In course of thermodynamic processes the specific economic value of
exergy increases. E.g. in a steam power plant the specific economic value of fuel
exergy is smaller than that of live steam, and the last one is smaller than the specific
economic value of electricity produced. Hence the exergy losses are the more harmful,
the farther is located the analyzed link in the chain of partial processes.
The cited statement stimulated many authors to propose economic applications of
exergy. The new field of exergy analysis, called thermoeconomics has been developed
 141

[20],[21],[22],[23]. However exergy is a thermodynamic concept, not an economic one

[9]. It should not be assumed, that the economic value is proportional to the exergy.
E.g. specific exergy of pure gold is smaller than specific exergy of pure iron.
The economic applications of exergy usually base upon the determination of specific
cost of exergy in particular points of the considered system. However such a
determination is not possible without additional assumptions, if the system contains
cycles. Consider e.g. a simple power plant containing a steam boiler, a turbine, a
condenser and a pump. Four independent economic balances can be formulated for this
system, e.g. three balances for particular links and a balance for the entire system. The
balance of entire system enables us to calculate the specific cost of electricity produced
(the investment costs and the cost of fuel are given). There remain three independent
balances with four unknown quantities (specific cost of exergy of steam before and after
the turbine, specific cost of water exergy before and after the pump). Hence if the
specific costs of exergy in all the characteristic points of the system are calculated, some
conscious or unconscious assumption has been made.
The determination of the specific cost of exergy is usually used for the partition of
production costs between the products of cogeneration. Such a partition can be made by
means of purely economic tools i.e. using the principle of avoided costs. In every
cogeneration process a major product can be distinguished. Its demand determines the
location of the plant and the production rate. The additionally fabricated product can be
called a by-product if it substitutes a major product of another specialized process. The
production cost of the by-product can be determined according to the production costs
avoided in the substituted specialized process.
The principle of the avoided expenditures can be also used for determination of the
consumption of driving exergy burdening the fabrication of major product of a
4cogeneration (combined) process:

b1m =_1
G (B-G
I b
~) (7)
m "Bs

where blm is the specific consumption of driving exergy burdening the major product,
Gm,G b are, respectively, the amount of the major product and by-product, BI is the
overall consumption of driving exergy, bb denotes the specific exergy of the by-product
and "Bs is the exergetic efficiency of the avoided process of separate fabrication of by-
product.
Despite the mentioned objections, exergy can be used as an auxiliary tool for solving
some economical problems. E.g. the heat lost from the pipeline of live steam in a steam
power plant has a better quality, than the mean quality of heat delivered inside the
boiler. Also the economic value of the lost heat is greater than the mean value. It can
be approximately evaluated by means of exergy:
142

(8)

where km,k are, respectively, the mean specific cost of the heat delivered inside the
boiler and local specific cost of the heat lost from the pipeline and Tm , Tare,
respectively, the mean temperature during heat absorption inside the boiler and local
temperature in the pipeline.

5. Cumulative Consumption of Exergy and Partial Exergy Losses

Consumption of exergy connected with the fabrication of the considered product

appear not only in the plant fabricating this product, but also in plants delivering semi-
finished products and raw materials for the final production process. Total consumption
of exergy of natural resources connected with the fabrication of the considered product
and appearing in all the links of the network of production processes, has been called
cumulative consumption of exergy CExC [12]. Analysis of CExC can be made in a
global scope or in scope of a particular country. In the last case the influence of the
import of semi-finished product on the domestic consumption of exergy should be
investigated. The calculations can comprise all the natural resources of exergy or only
the unrestorable resources. The last case is usually more interesting. The CExC-values
can be calculated separately for particular kinds of primary exergy or can comprise the
sum of primary exergy. Szargut [17] proved, that it is not necessary to take into account
the human labor when calculating the CExC-value, because it would denote a double
introduction of the same quantity. The mentioned statement results e.g. from the
analysis of a simple system presented in Fig.2. The system contains coal mines, the
industry and the society. The society consumes immediately a part Ck of the coal
production and the entire production of industry. The part C, of coal production
consumes the industry. The consumption of human labor amounts to L; in the industry
and Lc in coal mines.

I'
,.------, ,
(Society
\ . \
' . . 1--
Lc/

Figure 2. Scheme of a simplified system.

 143

If we take into account the CExC-value burdening the human labor, the CExC-values
characterizing the unit of coal and of industrial product are:

(9)

where b c' ,bi' are, respectively, the CExC-values burdening a unit of coal and of
industrial product, bc is the specific exergy of coal, Ic,Ii denote, respectivelyu, the
specific consumption of human labor in coal mines and in industry, r is the CExC-
value burdening an unit of human labor and Ci ,! denote, respectively, the consumption
of coal in the industry and amount of industrial products.
From eqs (9) it results:
• C. C. C. (10)
b.I = --'- (b + 1 r) + I.r = --'- b + r(l
J C C I J C
--'- + 1.)
cJ I

Total consumption of the primary exergy is:

(11)

where C is the total production of coal.

On the other hand it is evident, that the total consumption of primary exergy is C bC •
So the balance of CExC is closed, if the quantity r does not appear in eqs.(8). Hence
the mentioned statement is correct.
CExC can be calculated by means of the process analysis (sequence method) or by
means of the set of balance equations. The process analysis (sequence method) begins
in the final process of the fabrication of the considered product and goes through the
processes of production of semi-finished products, fabrication of machines and
equipment until the extraction of raw materials from natural resources. The sequence
method enables us to analyze only one product, but can be applied if the connections
between the considered links of the technological network are week (the consumption
of the considered product in preceding links is small). The method of balance equations
is more general, but requires more work.
The balance equation of CExC bases upon the statement, that the CExC-value of the
product of considered link results from the CExC-values burdening the semi-finished
products and by-product and from the exergy of natural resources extracted from nature
in the considered link. If the CExC-value are calculated in a scope of a particular
country, the set of balance equations has a form:
144

(12)

where b; ,bi' ,b; are, respectively, the specific CExC-values burdening the major
product ofj-th and i-th process and the r-th imported semi-finished product, aij'a,j are,
respectively, the coefficients of consumption of i-th domestic and r-th imported semi-
finished product per unit of j-th product, fij is the coefficient of the by-production
expressed in units of the substituted i-th major product, per unit of the j-th product, and
~j denotes the exergy ofthe natural resources extracted from nature in thej-th process,
per unit of the j-th product.
An exact method of calculation of b; has been formulated by Szargut [13] and based
upon the statement, that the financial means for import are gained by export. Hence the
CExC-value burdening the monetary unit of the import equals to that of export. An
exact method is rather complicated. An approximate method can be based upon the
assumption that the structure of export is similar to that of domestic production. From
this assumption it results:

b'=D~ (13)
, 'NGP
where D, is the specific monetary cost of the imported material, B is the total
consumption of exergy of unrestorable natural resources in the considered country, per
year and NGP is the national gross product in monetary units per year.
The difference between the CExC-value and the exergy of the considered product can
be partitioned into partial exergy losses PEL appearing in particular links of the network
of production processes [14]. Such a partition is possible if the calculation of CExC
comprises all the kinds of exergy extracted from nature, i.e. taken from 2
unrestorable or restorable, domestic or foreign sources:

(14)

where b; is the specific cumulative consumption of exergy burdening thej-th product,

and 6b/cj is the partial exergy loss connected with the fabrication of the j-th product and
appearing in the k-th link of the network of production processes.
Analysis of the partial exergy losses indicates the possibilities to decrease the CExC-
value. It can be attained by improving the links with great PEL-values, by changing the
technology of production of some semi-finished products or by changing the kind of
some semi-finished products. The calculation of partial exergy losses can be made only
in a global scope or for products fabricated without distinct consumption of the imported
materials and energy carriers.
The coefficients of consumption aij' ~j appearing in Eq.(12) relate to the unit of
major j-th product together with by-products. E.g. in a heat-and-power plant the
 145

coefficient of coal consumption relates to the unit of useful heat together with the
cogenerated amount of electricity. In order to calculate the partial exergy losses the net
coefficients of consumption of semi-finished products an natural resources should be
determined:

Ajj =a jj - LJ;,jA jp (15)

p

PNj = Pj - L!;,jP NP (16)

p

where Ajj,A jp are, respectively, the coefficients of a net consumption of i-th product per
unit ofthej-th andp-th major product and PNj,P Np are, respectively, the coefficients
of the immediate net consumption of natural resources per unit of the j-th and p-th
major product.
Equations (15) and (16) can be presented in a matrix form:
(17)

where E is the unitary diagonal matrix.

The coefficients of net consumption can be used for calculation of the CExC-values
similarly as the coefficients of gross consumption:
(18)

We can check the correctness of eqs (17) introducing them into Eq.(18):
b*T[E -a(E +f)-I] =BT(E +f)-I (19)

Multiplying Eq.(19) by the matrix (E +f) we obtain a formula (12) with omission of
the imported semi-finished products.
The local net exergy loss obNj related only to the unit of major product of the j-th
process, results from the exergy balance of thej-th process. It contains the internal and
external exergy losses:

(20)

or in a matrix form:
(21)

Partial net exergy losses result from the cumulative consumption of particular products
per unit of the considered product:
 146

(22)

where SNkj is the cumulative net consumption of k-th product per unit of j-th product.
The value of SNkj results from the matrix relation:
(23)

The sum of partial exergy losses fulfills Eq.(l4), but only then, when the balance
equation (12) does not contain the CExC-values of imported products.

Example 1. In order to calculate the partial exergy losses connected with the production
of coke and liquid pig iron, a simplified system comprising a coal mine, a iron-ore
mine, a source of natural gas, a power plant, a coke plant, a ore-sinter plant and blast
furnace plant has been investigated [15]. A coke plant is equipped by a dry quenching
installation and a steam turbine driving an electric generator. The mentioned plants have
only week connections with other industrial plants. Four by-products appear in the
considered system: coke-oven gas, tar and electricity produced in the coke plant and the
blast-furnace gas. It has been assumed that the coke-oven gas and tar substitute the
natural gas in the proportion 0.514 kmol n.g.lkmol c-o.g. and 0.0499 kmol n.g.lkg tar.
The basic part of the blast-furnace gas substitutes a natural gas in the proportion 0.114
kmol n.g.lkmol b-f.g. The peak part of the produced blast-furnace gas is burnt in the
boiler house and substitutes a hard coal in the proportion 3.62 kg coallkmol b-f g. The
consumption of steel in coal mines has been substituted by means of the consumption
of pig iron.
Following values of the specific exergy and of the coefficients of consumption and
by-production have been assumed (the mentioned values of the substitution ratio have
been taken into account):
1) Coal mine b l = 24.7 MJ/kg,
all =0.0058kgJkg,a31 =0.OOOO41kmol/kg,a41 =0.175MJ/kg,a 71 =0.004kgJkg.
2)Iron-ore mine b2 =0.35 MJ/kg,
a32 =0.00023 kmoljkg,a 42 =0.0735 MJ/kg
3) Extraction of natural gas b3 = 802 MJ /kmol,
a 43 = 4.17 MJ/kmol.
4) Power plant b4 = 1 MJ/MJ, a l4 = 0.156 kg/MJ.
5) Coke plant b5 = 28.95 MJ/kg,
a l5 = 1.60 kg/kg, a 35 = 0.00458 kmollkg, 135 = 0.0112 kmollkg,
a45 = 0.0958 MJ/kg, h5 = 0.391 MJ/kg.
6) Ore sinter plant b6 = 0.73 MJ/kg,
a 26 = 0.827 kg/kg, a36 = 0.0002247 kmollkg, a 46 = 0.1073 MJ/kg,
a56 = 0.0612 kg/kg.
7) Blast furnace plant b7 = 8.75 MJ/kg liquid pig iron,
 147

a17 = 0.036_6 kg/kg, /17 = 0.0214 kg/kg, a27 = 0.206 kg/kg,

a37 = 0.00441 kmollkg, /37 = 0.00549 kmol/kg, a47 = 0.0568 MJ/kg,
a S7 = 0.571 kg/kg, a67 = 1.594 kg/kg.
Calculation results of the partial exergy losses connected with the production of coke
and pig iron, are presented in Table 1.

TABLE 1. Partial net exergy losses

connected with the production of coke and liquid pig iron (in % of CExC).

Partial
Coefficient of consumption net
Material Process exergy
gross net unit loss
ali - fu Aii %

Coke coal mine 1.600 1.539 kg/kg 1.803

iron ore mine - - - 0.007
natural gas extraction -0.00602 0.0011 kmollkg 0.002
coke plant
power plant - - - 11.128
sinter plant -0.2952 0.0677 MJ/kg 2.918
blast furnace plant - - - 0.048
- - - 0.160

cumulative degree op - - - 83.934

perfection

Pig iron coal mine 0.0152 0.0366 kg/kg 1.802

iron ore mine 0.206 0.206 kg/kg 1.577
natural gas extraction -0.00108 -0.0038 kmollkg 0.077
coke plant
power plant 0.571 0.571 kg/kg 10.251
sinter plant 0.0568 0.0327 MJ/kg 6.684
blast furnace plant 1.594 1.594 kg/kg 10.351
- - - 34.315

cumulative degree of - - - 34.943

perfection

6. Ecological Cost

The inevitable depletion of unrestorable natural resources is very dangerous for the future
existence of mankind. Exergy can be accepted as the common quality measure of all the
natural resources. Therefore the cumulative exergy consumption of unrestorable natural
resources, termed ecological cost, has been proposed as a measure of their depletion
[10],[11],[17],[18]. Set of Equations (12) can be used forthe calcu1ationofec010gical cost
but the quantity Pj should comprise only the unrestorable resources and additionally the
deleterious ecological impact of waste products should be taken into account.
The waste products of every process have usually a deleterious impact on the
 148

environment. Three kinds of this impact can be named: the corrosion of buildings, machines
and equipment, the reduction of agricultural and forest production and the damages in
human health. The corrosion is most important. All the mentioned negative effects should
be compensated by an additional consumption of useful products substituting the lost useful
products or preventing the possible damages [17]. The determination of the coefficients
expressing the additional consumption of useful products is very difficult. Also the
assumption, that the ecological losses are proportional to the exergy of waste products is not
justified. E.g. CO has e relatively great exergy, but does not cause any corrosion of useful
products. Therefore Szargut [18] proposed to evaluate the deleterious impact of waste
products by means of their monetary index of harmfulness. The additional exergy
consumption of unrestorable natural resources due to the emission of waste products (e.g.
in kJ/kg) can be expressed as:

(24)

where B - see Eq.(13), a k is the monetary index of harmfulness of the k-th waste product,
and Pk denotes the annual emission of the k-th waste product in the country.
Eq.(24) takes into account the reduction of national product due to the emission of waste
products (the last term in the denominator).
After the introduction of the coefficients ~k the set of balance equations determining the
value of specific ecological cost takes the form:

(25)

where Pj' Pi denotye, respectively, the specific ecological cost of the major product ofJ-th
and i-th process, bsj is the exergy consumption of the s-th unrestorable natural resource, per
unit of the J-th product, Pkj denotes the emission of the k-th waste product per unit of the
J-th product and Pr is the specific ecological cost of the imported r-th semi-finished product,
Eq.(13).
Set of Eqs.(25) can be separately formulated for a group of production processes having
week connections with the remaining part of the network of production processes.
The ecological cost of the imported semi-finished products (e.g. fuels) is smaller than
their exergy, if in the exported goods prevail the highly processed products or products
obtained with a great participation of restorable exergy resources (agricultural or forest
products).
 149

Example 2. The annual exergy consumption of domestic unrestorable natural resources in

Poland contains the exergy Bhe of hard coal, Bbe of brown coal, Bng of own natural gas,Bcr
of own crude oil, Bs of sulfur and Beu of copper ore. In 1995 the mentioned values were:
B=Bhe +Bbc +Bng +Bcr +Bs+Beu =
=3315.109 +579.109 + 145.109 +22-109 +47.109 + 15.109 =412HO~J/a
The national gross product was 119.8· 109 $/a. The ecological cost per unit of national
gross product was 34.4 MJ/$. The specific cost of the imported natural gas was 2.24
$/kmol. So the ecological cost of the imported gas was 77.1 MJ/kmol (about 10 times
smaller than the specific exergy). Hence the import of natural gas distinctly decreases the
depletion of the domestic unrestorable natural resources [19].

Example 3. The annual emission of the main waste products (S02, NO. and dust) was in
1995 in Poland:
P so =2337'103 ,PNO = 1120'103 'Pd = 130S·103 t/a.
1 •
Following monetary indices of harmfulness have been accepted:
a so = 1500, a NO = 1500, a p = 300 $/t.
1 •
The indices of ecological cost of waste products, resulting from Eq.(24) are:
~so
1
=49305, ~NO• =49305, ~p = 10190 MJ/t.

7. Final Remarks

The ecological cost can be used for the optimization of the design- and operational
parameters of thermal plants and installations from the view point of the minimization of
depletion of unrest orable natural resources. However some problems should be additionally
solved, e.g.:
1. The methodology of optimization should be developed. The problem of the necessity to
introduce some kind of discount calculation has to be solved [3].
2. The minimization of ecological cost would prefer simple processes with a great
consumption of human work, but a small consumption of natural fuels. The human labor
does not appear in the calculation of ecological cost. Presumably the consumption of human
work should appear as a limitation in the optimization of the ecological cost.
3. It has to be concluded, how to take into account the consumption of natural resources
recoverable with some delay (e.g. the consumption of timber).
A most effective method of the economy of unrestorable natural resources results from
the utilization of restorable resources. It should be however not forgotten, that in the
processes of utilization of restorable resources appears also some consumption of
unrestorable resources.
150

8. References

I. Gouy G. (1889) Sur l'cnergie utilisable. Journal de physique 11,8,501-518.

2. Grassmann P. (1959) Die Exergie und das Flullbild der technisch nutzbaren Leisrung, Allg. Wiirmetechn. 9,
No 4/5, 79-86.
3. lantovski E. (1997) What is exergonomics? (on a branch of engineering thermodynamics), Proc.
FLOWERS'97, Florence, 1163-1174.
4. Rant Z. (1956) Exergie, ein neues Wort fur "technische Arbeitsfiihigkeit", Forsch. Ing.- Wes. 22, No I,
36-37.
5. Riekert L. (1974) The efficiency of energy utilization in chemical processes, Chern. Eng. Sci. No 29, 48-51.
6. Sarna D. (1992) A common sense 2nd law approach to heat exchanger network design, Proc. ECOS '92 ,
Zaragoza, 329-338.
7. Stodola A. (1898) Die Kreisprozesse der Gasmaschine, Z.d. VDI 32, No 38, 1086-1091.
8. Szargut J. (1956) Potential balance of physical processes resulting from the 2nd law of thermodynamics (in
Polish), Arch. Budowy Masz;yn 3, No 3, 231-276.
9. Szargut J. (1967) Grenzen fiir die Anwendungsmoglichkeiten des Exergiebegriffs, Brennstoff-Wiirrne-Kraft
19, No 6, 309-313.
10. Szargut J. (1978) Minimization of the consumption of natural resources, Bull. Pol. Acad., Sci. Techn., No
6,611-615.
11. Szargut J. (1986) Application of exergy for the calculation of ecological cost, Bull. Pol. Acad., Sci. Techn.,
No 7-8, 475-480.
12. Szargut J. and Morris D.R. (1987) Cumulative exergy consumption and cumulative degree of perfection of
chemical processes, Energy Research, No 11, 245-261.
13. Szargut J. (1987) Influence of the imported goods on the cumulative energy indices. Bull. Pol. Acad., Sci.
Techn. 35, No 9-10, 591-595.
14. Szargut J. (1988) Exergy losses in the chains of technological processes. Bull. Polish A cad. , Sci. Techn. 36,
No 7-8, 513-521.
15. Szargut J. and Majza E. (1989) Thermodynamic imperfection and exergy losses of the production of pig iron
and steel, Arch. Hutnictwa, No 2, 197-216.
16. Szargut J. and Sarna D.A. (1995) Practical rules of the reduction of energy losses caused by the
thermodynamic imperfection of thermal processes, Proc. The Second Intern. Thermal Energy Congress,
Agadir,782-785.
17. Szargut J. (1995) Exergy and ecology, Proc. of the Workshop" Second Law Analysis of Energy Systems",
Roma, 11-19
18. Szargut J. (1997) Depletion of the unrestorable natural exergy resources, Bull. Pol. Acad., Techn. Sci. 45,
No 2, 241-250.
19. Szargut J. and Stanek W. (1998) Influence of the import of natural gas on the depletion of unrestorable
domestic natural resources (in Polish), Gospodarka Paliwami i Energiq, No 2, 2-4.
20. Tribus M. and Evans R. (1962) Thermoeconomics, UCLA Repon, No 52-63.
21. Tsatsaronis G. and Pisa J. (1992) Exergoecomomic evaluation and optimization of energy systems, application
to the CGAM problem, Proc. ECOS'92, Zaragoza.
22. Valero A., Torres C. and Serra L. (1992) A general theory of thermoeconomics, part I, structural analysis,
Proc. ECOS'92, Zaragoza 1992, 137-145.
23. Valero A. and Lozano M.A. (1992) A general theory of thermoeconomics, part II, the relative free energy
function, Proc. ECOS'92, Zaragoza, 147-154.
ALLOCATION OF FINITE ENERGETIC RESOURCES
VIA AN EXERGETIC COSTING METHOD

E. SCIUBBA
Dipartimento di Meccanica e Aeronautica,
Universita di Roma 1,
via Eudossiana 18 - 00184 Roma,ltaly

1. Introduction

This paper describes a method for the assessment of an "exergetically optimal"

resource allocation and distribution. Allocating a certain resource R means to make it
available in a given amount at a finite or infinite number of "consumption centres",
and to find the optimal siting of the production andlor of the consumption points, so
that the resource is distributed as needed, under the specified constraints, and its
distribution pattern is such that it minimises the overall exergy consumption caused by
the technological chain of processes the resource undergoes from its extraction to its
final use (including disposal of residues). Two general formulations are possible: in the
discrete problem, the number of consumption sites as well as that of the production
centres is finite, and usually rather small. In the continuum problem, while the number
of production sites remains small, the possible consumption points are distributed
uniformly over the entire region of interest, so that in effect the resource must be made
available everywhere in the domain. In both cases, the practical implementation of the
method consists in the minimisation of an objective function that contains the
extraction, preparation, transportation, distribution and environmental costs of the
resource: the important novel feature here is the fact that these costs are expressed in
exergetic units, i.e., they are represented by the amount of exergy that is expended for
that particular operation. The constraints are topological (the orography and the
geography of the region), technical and environmental, and are imposed in the form of
exergetic penalty functions. For the continuum formulation, a similarity with a
diffusion problem can be invoked to extend the mathematical analogy to the solution of
a general dual dispersion equation: the reader is referred to a companion paper [10].

2. Problem Position

Consider a generic geographical area D in which N Production Centres Pi (i=I,2, ...N)

i.
for a single resource R are located (Fig. 1 In the same area, but of course in different
sites, there will also be M Consumption Points Cj (j=1,2, ... M) for that resource. In the

I Th~ map of Romania has been downloaded via Internet from MapQuest"'. The location of Production and
Consumption centers is arbitrary and hypothetical, and serves only as an example of application of tOO method
pro~d in this lecture.

151
A. Bejan and E. Mamut (eels.), Thermodynamic Optimization o/Complex Energy Systems. 151-162.
© 1999 Kluwer Academic Publishers.
 152

entire course of this Lecture, we shall make the following assumptions:

• each one of the N Production Centres is characterised by a specific production cost
g, (x,y,mi) whose structure is known;
• the transportation cost of the resource from Production Centre to Consumption Site
is a function of the equivalent distance d iJ and of the quantity of resource made
available at Consumption Site, mij;
• the equivalent distance diJ is a computable function of the actual distance rij, of the
oro-geographic conditions and of the actual transportation line and means;
• all "costs" are to be intended as exergetic, and not monetary, costs. The specific
exergetic equivalent cost we shall refer to is the total "exergetic content" of the unit
mass of resource, whereby by exergetic content we mean the sum of the
thermodynamic exergy of the untapped resource and of all exergetic expenditures
incurred in during the processes of extraction, refinement, transportation,
packaging and final distribution (including disposal of residues);
• transportation mass losses are negligible;
• the problem is at steady-state.

The following three formulations shall be considered:

M
I. given the Cj (x,y,mj) find the optimal siting for the Pi'S for a specified m tat = :Lmj ;
j=1
2. given the Pi(x,y,m;), find the optimal siting of the Cj's for the prescribed
N
I1ltat=2: m; ;
1=1
3. given Cix,y,mj) and Pi(x,y,mi), find the optimal resource allocation (the distribution
of the m;'s among the consumption sites) such that the overall exergetic cost is
minimised and the mass flow rate constraints are abided by.

A fundamental distinction must be made depending on the demand structure for the
resource R: if the demand is concentrated, i.e., if there are few and relatively wide
apart Consumption Sites, a discrete formulation of the problem is in order. If, on the
other hand, the demand for R is distributed, i.e., almost everywhere in D there are
potential Consumption Sites for R, then a continuum approach is needed. The two
cases are discussed in Sections 3 and [10] respectively.

3. The Discrete Problem: Concentrated Consumption Sites

3.1. OPTIMAL SITING OF THE PRODUCTION CENTRES FOR A KNOWN

DISTRIBUTION OF CONSUMPTION SITES

3.1.1. One Production Centre, one Consumption Site

The specific exergetic cost k of resource R at location C is given by the sum of the
production, transportation and "ex1ernal" costs:
k = g(xp,yp,m) + f(dpc,m)+e(d,m) (1)
Since by hypothesis f, g and e are known or computable (see Appendix 2), it will be
always possible to draw iso-cost curves on the map of D. An inhibitory constraint may
 153

be active (there is a certain area Ac around the Consumption Site in which no

Production Centre may be located), and the solution is obvious: P must be localised on
a point lying on the lowest-k curve AND not contained in the forbidden area Ac. A
simple (hypothetical) example is shown in Fig. 2.

3.1. 2. Multiple Production Points, one Consumption Site

The total exergetic cost K of resource R at location C is given by:
N
K = Lm;[g(x; ,y;.m;)+ f(d;com; )+e(d;c,m;)] (2)
;=)
where m; is the mass flow rate reaching C from Pi. Since f, g and e are known or
computable, the problem can be restated:
N
K = ~Aimi (3)
;=1

where Ai is a function of (XiSi) and (Xc,Ye) only, through die, gi and eic. A mass
conservation constraint applies:
(4)
This formulation corresponds to a linear programming problem in its canonical form
(the minimum of K must be found). Again, taking into consideration possible
inhibitory constraints, the solution in this case is: the P; 's must be localised on any set
of N points which lie on the lowest K-curve and are not contained in any of the
forbidden areas Ago A simple example is shown in Fig. 3. Notice that in this case
additional constraints on the locations of the Pi's relative to each other might be
imposed ("proximity" constraints, see Appendix I).

3. /.3. Multiple Production Points, multiple Consumption Sites

The total exergetic cost K of the resource at each location Cj is given by :
N
K j = Lm,Ag(x;,y;,mj )+ f(dif,mj)+e(dif,m j )] (5)
;=)
N
where mj =Lmif . The total exergetic cost is found by summing over the indexj:
;=)
NM
K tot = LLmif[g(X;,y;,m j ) + f(dif,mj)+e(d;j,m)] (6)
;=)j=1

and the problem is again one of linear programming (the minimum of KID. must be
found). There are two types of mass-conservation constraints in this case:
M
m; =Lmif (N additional constraints) (7)
j=1
N
mj =Lmq (M additional constraints) (8)
;=1

and N congruency constraints, that require that for each Production Site
M
Lit/if ::;; m;,mar' Again, possible inhibitory and proximity constraints must also be
j=1
 154

taken into account. The solution prescribes that the Pi'S must be localised on any N of
,o,
the points which lie on the lowest-K curve AND are not contained on any of the
forbidden areas A g •

3.2. Optimal siting of the Consumption Sites for a specified distribution ofthe
Production Centres

The general discrete alIocation problem formulated here is clearly symmetric with
respect to "sources" and "sinks": if in the expressions hitherto computed for the
exergetic cost the indices p and c (or i and j) are inverted, the corresponding formulae
give the solution of the symmetric problem, i.e., the one in which the location and
strength (productivity) of the Production Centres are both known, and the solution
consists in finding the optimal siting for the Consumption Points. Therefore, no
specific discussion of this case is offered here, and the reader is referred to the previous
Section (and to [3] for a more detailed treatment).

3.3. - Optimal allocation ofa known amount of resource generated at N Production

Sites and distributed to M Consumption Centres

The problem here is to find the optimal combination of the mij, i.e. to determine how
much resource ought to be generated from each production centre Pi and transported to
each consumption site Cj . The total cost is again given by equation 6), and the same
constraints given by equations 7) and 8) apply. The N congruency constraints are the
same as welI, so that the Linear Programming problem is formally identical to that
illustrated in Section 2.1.3.

4. Conclusions

A new formulation of the resource alIocation problem has been outlined in this paper.
Several simplifYing assumptions have been made, most of them aimed at linearising
the problem. The folIowing conclusions can be drawn:
I. For resources that admit of localised Consumption Sites, discrete optimisation
methods based on linear programming techniques can be efficiently employed to
achieve at least a quasi-optimal alIocation; such resources are mineral ores, fossil
fuels, and in general alI those materials that must undergo substantial
transformations before being suitable for industrial or commercial use;
2. For resources that admit of uniformly distributed Consumption Sites, the method
can stiU be applied, and admits of a very powerful and elegant mathematical
formulation based on the dual form of the dispersion equation. The details are
developed in [10]. The resources amenable to be treated under a dispersion analogy
are typicalIy those that can be consumed as-such: food, commercial goods and the
like, water, electricity or district utilities;
3. It must be stressed that the possibility of practical application for this entire
approach depends on the availability of a coherent and reliable cost basis. The cost
functions employed must alIow for the consistent incorporation of different
expenses occurred in distinct locations and in diverse phases of the extraction-
transportation-distribution chain; moreover, they must alIow for the quantification
 155

of physically different "costs" (value of the Gibbs' function for a fuel, energy used
for transportation, extraction costs of a resource, etc.) in a homogeneous way. This
cannot be achieved using monetary costs. The exergetic cost adopted here (first
proposed by [12,13,14]) not only leads to the formulation of a proper, complete and
homogeneous objective function, but it evaluates the flux of resource for its value,
and not for its price: the difference is substantial, and carries far-reaching
implications if correctly applied to the analysis of an industrial system.

Since a hierarchy of cases can be easily constructed (a resource R from N wells is

allotted among M pre-treatment stations; then, each one of these stations becomes the
"producer" of an R-derivative which is distributed among MJ chemical industries, each
one of which becomes a "producer" of several types of further derivatives, and so on
until final consumer production/distribution is reached), the method described in this
Lecture can be repeatedly applied to the entire chain of resource extraction,
(production), transformation and distribution, to investigate the possibility of existence
of "optimal paths" in the resource use.

5. References

I. A)TeS R. (1995): Life cycle analysis: a critique, Res. conservation and recycling, n.14.
2. .-\zzarone F.,Langer P. (1996): Optimal Resource Allocation Model, TR-CIRCVS 010996, Vniv. ofRoma.
3. Azzarone F.,Sciubba E. (1995): Analysis of the energetic and exergetic sustainability of complex systems,
in AS ME AES·35.
4. Evans R.L. (1980): Thermo-economic isolatIOn and Essergy Analysis, Energy, vol.5 n.8-9.
5. Gaggioli R.A.,EI-Sayed Y. (1987: A Critical review ofsecond-law costing methods, in ASME vo1.I00236.
6. Gong M.,Wall G. (1997): On Exergetics, Economics and optimization of technical processes to meet
environmental conditions, Proc. T AIES'97, Beijing (China).
7. Marchuk G.!. (1986): Mathematical models In environmental problems, North Holland.
8. Oni AA.,Cambd A.8. (1985): Generalised thermodynamic approach to scarce energy resources
allocation through the disjointed incrementalism algorithm, in ASME-AES-1.
9. Paolocci S.,Sciubba E. (1994): A new criterion/or power plant siting based on a dual formulation of the
dispersion ofpollutants in the atmosphere, ASME-AES vo1.33.
10. Sciubba E. (1998): A novel exergetic costing method for determining the optimal allocation of scarce
resources, Proc. Contemporary Problems of Thermal Engineering, A.Ziebik ed., Gliwice (Poland).
II. Szargut J. (1997): Depletion of Unrestorable Natural Exergy Resources, Bull. Polish Ac.ofSci.
12. Szargut J,Moms D.R.,Stewart F.R. (1988): Exergy analysis of thermal, chemical and metallurgical
processes. Hemisphere, Washington. DC.
13. Valero A.,Lozano M.A.,Munoz M. (1986): A general theory of exergy saving, Part 1: on the exergetic
cost. in ASME-AES-2/3.
14. Wall G. (1986): Exergy: a useful concept within resource accounting, Ph.D.Th., Chalmers T.V., Goteborg,
Sweden.
15. Ziebik 1..Gwozdz J.,Presz K. (1994): Matrix method for calculating the unit costs of energy carriers as a
co-ordination procedure In the optimisation ofindustrial energy systems, in ASME PD-64/3.
 156

Appendix 1. The Solution Method

To begin with, let us observe that the general fonnula given in equation 6 can be
applied to the simple cases of only one Production Centre or only one Consumption
Site by proper contraction of the corresponding index. Therefore, we will take equation
6 to represent the general solution to the optimal resource allocation problem studied
here. This equation is non-linear (the cumulative yearly resource production appears in
the specific cost functions), but can be quasi-linearised by devising a proper quasi-
continuous fonnulation for the cost functions (see Appendix 2). The constraints
considered here are of three different types:
I) mass conservation (strong constraint):
la - the total mass flow of resource from each Production Centre Pi must be equal
to the sum of the (at most) M mass flow rates supplied by Pi to the various Cj
it is connected to;
Ib - conversely, the total mass flow of resource into each Consumption Site Cj must
be equal to the sum of the (at most) N mass flow rates supplied to Cj by the
various Pi it is connected to;
2) congruence (weak constraint): the maximum extractive capacity of each Production
Centre m~max shall not be exceeded;
3) minimum distance ofany Production Centre from any Consumption Site (weak con-
straint). These are the so-called inhibitory constraints, and set a lower bound to the
physical distance riJ between any Pi and any Cj;
4) minimum distance between different Production Points (weak constraint). These are
the so-called proximity constraints, and set a lower bound to the physical distance
between each pair of Pi'S.

The solution procedure can then be outlined as follows:

I. the region under examination is discretised according to two distinct criteria: first,
areas of similar oro-geographical conformation are detected, in which the penalty
factor y (a coefficient ~l by which the physical distance rij must be multiplied to
find the equivalent distance dij) can be assumed to take a constant value; second,
areas of similar environmental fragility (see Appendix 2) are identified, in which
the marginal environmental exergetic cost takes constant values. The mesh adopted
in the final calculations is then the finer of the two;
2. the cost functionals are computed, with the fonnulae reported in Appendix 2;
3. the constraints are imposed in discrete fonn. Weak constraints are always imposed
in their equivalent strong fonn (e.g.: both "less than or equal to" and "greater than
or equal to" are imposed as "equal to");
4. a standard linear programming algorithm is employed to find the minimum of the
exergetic cost expressed by equation 6.

The errors deriving from the discretisation procedure are unknown and difficult to
estimate a priori: therefore, the solution algorithm yields in practice only a quasi-
optimal allocation. These computational errors have two different sources: first, they
are caused by the functional approximations introduced by the quasi-linearisation of
the cost functions; second, they are originated by the discretisation approximations
(like for instance that of applying a constant penalty factor to a relatively large and
 157

non-homogeneous geographical area).

Appendix 2. The Cost Functions

A2.1. THE EXTRACTIONIREFININGIPRODUCTION COST

This cost is associated with each one of the N production centres, and is assumed to be
of the form:
cl "
m + C,....1. + c'" + c4 ,I
g I = _.1 ~,l
(A. I)

both g , c:. C3 and c.are expressed in kJlkg, and CI in kW .

.'12.1.1. Fixed costs, CI

The fixed costs CI include land-, factory buildings-, infrastructures and facilities costs.
In real life. they show a marked dependence on the total production rate of resource
(mR) and to the global yearly output ~. In the discretisation employed here, they have
though taken to be constant (Fig. 4) .

.'12.1.1. Cost a/the Resource, c]

An intrinsic specific cost (per unit mass) c] can be attributed to the resource R itself, to
represent its exergetic value in the conditions in which R is extracted. A distinction is
necessary between renewable and non-renewable resources: while the marginal
exergetic value of a renewable resource can be taken to be independent of the mass
consumed, a non-renewable resource has a marginal exergetic cost that increases with
the mined amount (Fig. 5).

A 2.1. 3. Process costs, c3

In real resource tapping operations, process costs C3 are a function of both total yearly
yield (~) and maximum productivity (mIJ. Here, a stepwise linearisation of C3 with mR
and ~ has been adopted (Fig. 6), which amounts to "discretising" scale effects .

.'12.1.4. Operational costs, C4

These costs represent the amount of exergy expended in the chain of mechanical,
thermal and chemical processes that bring the resource R from its "natural" to its
"final" state. For a given technology, scale effects can be assumed to be negligible: a
constant function of mR is proposed here.

A2.2. THE TRANSPORTATIONIDISTRIBUTION COST

This cost pertains to the distribution network, and has been assumed to have the
structure:
;; = t\j + tv + t3.,d;, (A.2)
i.e .. the functional dependence has been split into a fixed (capital) cost flo a "friction"
effect t: and a distance/difficulty cost f3: f, tl and t2 are expressed in kJlkg, t3 is in
kJ/(kg*m)=mls 2 Notice that dy=;rij, the penalty factor y depending on the oro-
geographical conditions of the region. The total transportation cost for N Production
Centres serving M Consumption Sites is:
158

NM
F= I Imif/;j (A3)
i=1 j=1

A2.2.1. Fixed cost. tl

This cost represents the (specific) exergy spent to set up the transportation/distribution
network. It has been assumed here to depend only on the transportation means, but in
reality it varies (at times discontinuously) with both the total annual yield and with the
mass flow rate of the resource on that specific route (Fig. 7) .

.42.2.2. 'Friction" cost. t2

This cost coefficient depends on the means of transportation employed and on the
technological level of the distribution chain. As an example, Fig. 8 represents the
functional dependence of t 2 on m ij for a hypothetical transport of the same resource via
barge and by truck. Notice that there is an implicit correlation between the penalty
factor y and t2, as shown by the fact that the transport by barge in the upstream
direction carries a higher exergetic cost than that in the downstream direction, similar
in principle to the difference between the exergetic costs of transportation by truck on
flat (cheaper) or hilly (more expensive) terrain shown in Fig. 9.

.42.2.3. Distance cost. t3

This cost coefficient depends mainly on the physical distance between the Production
and the Consumption Sites. A functional dependence on the type and state of terrain,
climate, etc. must be also considered, by introducing the corrected (dij ) instead of the
physical (rij) distance. This marginal cost coefficient is also dependent (for the same
resource) on the type of transportation means employed.

A2.3. THE ENVIRONMENTAL COST, e

Production and transportation costs are not the only expenditures that are incurred in
when allocating a resource. There are other costs, unaccounted for as yet, namely the
environmental pollution generated by the eXlJloitation of natural resources, and the
costs associated with it (which may include social- and health effects as well). An
advantage of the allocation method just described is that it allows a quantitative
estimate of environmental effects to be included in the objective function. To begin
with, it has been assumed here that the harmful effects of the release of waste products
into the environment can be expressed quantitatively in terms of the exergetic value of
the actions necessary to nUllify their effects. This is not a new assumption: it has been
proposed -with slightly different meanings- by several Authors [1,2,6,10,11,14]. A
discussion on the methods of quantification of the exergetic value of a waste stream is
outside of the limits of this Lecture: we shall assume that such a quantification can be
carried out and that it produces univocal results, and adopt a heuristic criterion, by
dividing the task into three sub-tasks:
a) determine the chemical and physical nature of the released waste stream;
b) estimate the portion of the biosphere that is directly affected by this release;
c) calculate the already existing concentration of waste products of any kind in the
geographical area directly affected by the "new" release.

The proposed cost assessment procedure is then the following steps. First, the
 159

environmental costs are considered to be directly related to the clean-up costs, i.e., to
the costs of the re-establishment of the pristine environmental situation: this "pristine"
environmental situation is taken to be the one existing in the region D prior to the
introduction of the Production Centres Pi and of the corresponding transportation lines.
Second, the environmental cost is split in two terms: one is a direct measure of the
exergetic costs of the clean-up, and the other is a quantification of the present state of
(environmental) affairs in the region. This last cost component quantifies the concept
that the cumulative effects of a "new" pollution source are more costly if the region is
already polluted. It is clear that a higher mass flow rate of "new" releases would
produce higher restoration costs, and it is also clear that release of wastes in areas
which have already been contaminated by previous releases carries an extra cost
penalty: this can be quantified by introducing a coefficient of environmental fragility,
and assuming that the exergetic cost of environmental damage be a function of site
fragility and of the mass flow rate of the new release (Fig. 10). The environmental cost
can therefore be computed with the following formula:
MDa oj
ei = . +bj (A.4)
mij

where l1AD is the area of the portion of D on which the cost is being calculated, aoJ is
the fragility coefficient (kW/m2) and bj (kJ/kg) is the marginal cost of the new release.
The total cost for N Production Centres serving M Consumption Sites is:
NM
E = LLmijeij (A.5)
i=1 j=l

and can be derived as the sum of several "route" integrals (Fig. 10).
In the method presented here, the environmental cost has been considered as a
constraint for the allocation procedure: all other parameters being the same, the siting
of the Production Centres (considered as concentrated pollution sources) is optimal
when they are located in the least fragile areas. If several areas of equal fragility are
available, then the cost function is the discriminating parameter (see [2] for an
example of application).

Figure 1. Problem position (two possible production centers shown).

160

Figure 2. One production center, one consumption site: iso-cost curves (Cj+l > Cj).

Fiiiure 3. Two production centers, one consumption site: iso-cost curves.

c, (kW)

o • fl (kgly)

Fiiiure 4. Fixed cost codlicient c, as a function ofth~ yearly mass output of resource.
 161

c: (b)
(kJ/kg)

. . .
[------------------~--------~-------------. (a)

o ill: IiJ (kg/s)

Figure 5. Resource cost coefficient c, for a renewable (a) and non-renewable (b) resource.

- - ----------~~---
...;
- - m (kg/S)

ms
Figure 6. Process cost coefficient Cj as a function of the production mass flow rate and of the yearly output (fll > fl').

tI
(kJ/kg)
b

o m (kg/s)
FiKure 7. Fi,,~d .:ost coefficient tl for transportation by pipeline (a) and by truck (b)
162

(kJ~g) (b)

---------,
f-----'

.. _ .. _ .. ,.. .. - .. - .. .,. .. _ .. _ .. ..., (a2)

o m (legis)

Figure 8. ·'Friction·' cost t2 for transportation by barge up- (a.) and downstream (a2),
and by truck (b)

I.,
[kJ/(1Il kg)]
,.. .. _.. _.. - .. _..
:---------~

;...---------:

o m(legis)
Figure 9. Distance cost coefficient h for transportation by barge (a) and by truck
on hilly (b.) and flat (b2) terrain

§.'""""'"1 Environmental Cost

~";iH1
:;UlL~
b (kJ/kg)

c.----:,
1"'1
..... _,', J

m
I111111111 cenlcr

Dimensionless
Fragility physical
Factor a (kW/m') distance

Figure 10. Example of calculation of the environmental cost for two different routes.
OPTIMISATION OF TURBOMACHINERY COMPONENTS BY
CONSTRAINED MINIMISATION OF THE LOCAL ENTROPY
PRODUCTION RATE

E. SCIUBBA
Dipartimento di Meccanica ed Aeronautica,
Universita di Roma J - "La Sapienza"
via Eudossiana J8 - 00 J84 Roma, Italy

1. Introduction

This Lecture presents some reflections on the practical measure of thermodynamic

irreversibilities affecting the interactions of a process fluid with a part of a
turbomachine: the main contention here being that the entropy generation rate is the
only thermodynamic quantity that can be used to correctly estimate viscous and
thermal losses, and to compare both the absolute and the relative performance of
different configurations. An entropic analysis (i.e., the calculation of the entropy
generation rates associated with the various types of "losses") can be performed at a
bulk, at a microscopic and at a macroscopic level. The bulk level is the one usually
employed in lumped formulations, where either the entire system or portions thereof are
considered as a control volume, and the entropy generation rate is computed via the
usual thermodynamic expressions derived by direct manipulation of the constitutive
relations. The microscopic or local level is the method advocated here: it consists of
separating the two different forms of irreversible entropy production in fluid flow, due to
thermal (St) and viscous (sv) effects respectively, and of expressing them as functions of
the local velocity- and temperature gradients and of the local temperature. The
macroscopic or global level consists in integrating the results of the microscopic
analysis over the domain of interest: it produces a value of the total entropy generation
rate quantitatively very close to that obtained by the bulk method, but different in
principle from it, because calculations at the global level take explicitly into account the
non-uniformity of the medium.
This lecture describes in detail a general procedure by which cases of practical
engineering relevance can be numerically evaluated in order to obtain useful
quantitative design information about the relative performance of different
configurations. An optimisation procedure is presented, whose objective function is the .
global entropy generation rate and which contains appropriate integral design
163
A. Bejan and E. Mamut (eds.), Thermodynamic OptimiZlltion a/Complex Energy Systems, 163-186.
© 1999 Kluwer Academic Publishers.
164

constraints, like the maximum material temperature (weak constraint, applies to both
fixed and rotating passages) and the overall useful enthalpy drop of the working fluid in
the passage (strong constraint, applies to rotating passages and to heat exchanging
elements). The discontinuous optimisation applications implemented to-date are limited
to a trial-and-error technique, but the thermodynamic basis upon which they rest can be
rigorously formalised, and this is one of the objectives of the Lecture. The proposed
optimisation technique consists of four steps:
1. consider an existing, realistic geometry and regard it as a "zero configuration";
2. compute the thermo-fluiddynamic field around it and calculate the entropy
generation rates in the passage, recording the value of the overall enthalpy drop of
the gas as well as the value and location of the maximum material temperature;
3. keeping all other parameters fixed (in particular, maintaining the same kinematic,
dynamic and thermal boundary conditions), slightly perturb the geometry of the
passage, by introducing arbitrary modifications in the blade profile, in the tube
diameter and pitch, in the number and location of cooling holes, etc.;
4. compute anew the velocity and temperature fields and, if necessary, tune the inlet
conditions so that the integral constraints are abided by; recompute the entropy
generation rates and compare them to those of the previous configurations.

By iterated application of the very strong heuristics which can be derived from the
entropy generation rate, an "optimal" configuration can be found that still satisfies the
design specifications but is affected by minimal losses. Some applications of the
procedure are discussed, for realistic gas turbine blade profiles, both in 2-D and in 3-D,
and for a finned-tube heat exchanger. Such an approach was implicitly already
contained in the first formulation of an "entropy minimisation strategy" outlined in [1]:
the innovative character of the analysis presented here consists in the fact that local
entropy generation rates are extracted from the flow field and can thus be directly used
as objective function in a (discrete) pseudo-optimisation procedure. The method results
in a very powerful tool for the design of "better" turbomachinery components, because
the ratio of the global value of the entropy generation rate (i.e., of its integral over the
entire computational domain) to the total enthalpy drop across the blading or the heat
exchanging channel is proportional to the cascade efficiency vz. to the heat exchanger
effectiveness, and because the local entropy generation values, properly split into their
viscous and thermal components, give a complete and thermodynamically correct
indication of the "problem areas" in which design changes ought to be introduced to
improve component performance.

2. The Calculation Procedure for the Local Entropy Generation Rates and the
Entropy-Based Constrained Minimisation Procedure

The governing equations of the phenomena described here are the Navier-Stokes
equations, the continuity- and the energy equation; the formulation applies to the
laminar or turbulent, steady, incompressible flow of a fluid of constant properties. The
governing equations and the numerical techniques are extensively discussed in
[3,14,15,19]. With the exception of the results presented in Section 3.3.1, the entropy
generation rates are computed from the converged flow field using the following
equations:
 165

S,. = _!T [(~~)2

ax + (av)2
ay
+ (~~)2] + ![(au. + a",)2 + (~'" + aw)2 + (au + aw)2]
az 2 ax ay az ay az ax
(1)

(2)

In this Lecture these two contributions are handled separately, and indicated by
viscous- and thermal entropy generation rate respectively. Notice that these expressions
are dimensionless: to compute the dimensional specific entropy generation rates, one
must mUltiply svand Sf respectively by the so-called entropy scales:
~U2 k~,,-,
S = ---- and S, = - - (3), (4)
v rt1T L2
The value of the entropy generation rate in a finite domain can be computed from a
global point of view, using the perfect gas approximation:

sH2 = d( SH2) = c In(-~)- R In( P2.) (5)

dt I;
I' PI

or can be computed in a local perspective, considering the thermal and VISCOUS

contributions to the overall dissipation [1]:

SI->2= f[v ~<I>T + k~a.J'i1TP}

.-..
-2--
f
V= (SvSV+S'Sf)dV (6)
T v
In the right-hand side of equation 6, the first term in the integrand represents the
irreversible entropy production rate due to local shear (<I> being the so-called dissipation
function, which includes in this case the contributions of the turbulent stresses), while
the second term represents the irreversible entropy production due to local thermal
imbalances.
The procedure proposed in this work can be now described and justified on the basis
of the above remarks. For a given heat-exchanger or cascade configuration, if a
complete quantitative description of the velocity- and temperature fields is available,
one can compute Sv and Sf locally, i.e., in every point of the flow field; their integral over
the entire flow domain (equation 6) gives then a value which is numerically very close to
the SJ-2 obtained by equation 5, and is therefore a measure of the work lost in viscous-
and thermal irreversibilities in the flow; notice however that the two contributions in
equation 6 can be computed separately, while this would not be possible if one used
equation 5; furthermore, the integral in equation 6 explicitly includes local
inhomogeneities in the u;'s and T.
Now, if the configuration is modified, for instance by adding a row of cooling holes
on the leading edge of the turbine vanes, or by increasing the fin spacing in the heat
exchanger, the same calculation can be repeated, and the new values of Sv and St, as well
as their integral over the entire flow domain, will allow a direct comparison of the
relative performance of the modified configuration with respect to the original one, and,
what is more important, will show how the map of the viscous- and thermal
irreversibility distribution in the flow domain depends explicitly on the design changes.
The information thus gathered is of significance to the designer, who can use it to:
166

a) choose the configuration which produces the lower total irreversibility (lower
entropy generation rate, equation 6);
b) determine -by inspection- where design changes might be desirable to reduce
irreversible losses under the applicable design constraints;
c) determine -also by inspection- which design changes are more apt to reduce
irreversible losses at off-design conditions.

The following Sections describe several applications of this procedure to representative

2-and 3-D geometries of turbine cascades and heat exchangers.

3. Examples of application of the Optimisation Procedure

3.1. AIR-COOLED GAS TURBINE NOZZLE: CONFIGURA nON STUDY

Consider a two-dimensional section of a gas turbine first-row stator blade (figure 1): for
the purpose of the present study, the profile can be taken as being representative of the
radial section at mid-span of a generic gas turbine vane, internally cooled by a radial
flow of air (perpendicular to the plane of the drawing) through three separate internal
slots, equipped with cooling holes from which the cooling air is discharged into the
mainstream gas. The effectiveness of film cooling on the stator vanes depends on nozzle
geometry and on the location and number of the cooling holes (unsteadiness effects have
been not addressed here): to choose the most appropriate design configuration, it is very
important to fully understand the flow characteristics in the channel, which in this
context means the combination of number, type and location of the cooling holes which
produces the lowest thermal stresses compatible with the type of materials and
operation. But, there is another face to the medal: any discrete holes film cooling
configuration will produce losses in the expander, which decrease its performance. I
The difference between the configurations analysed in this study consists in the
presence (or absence), location and geometry of one or more rows of cooling holes
(represented in this 2-D schematization by their longitudinal midsection). In the
configuration without cooling holes ("a" here below) the cooling air is assumed to be
discharged from the vane tip.

3.1.1. Cascade Data

The relevant geometric parameters of the cascade are shown in Fig. I. Details on the
material properties and on the flow characteristics are given in [14]. The dimensionless
characteristic flow parameters are shown in Table I.

I A major contribution to these losses is given by the reduced mass flow in the compressor, which requires a
portion of the expander's shaft work to be used to pressurise air which then bypasses combustion; but there
are also pressure losses in the cooling circuit, which reduce the effectiveness of the convective cooling, and
mixing losses at the ejection points (holes), which disturb the boundary layer/mainstream interface and
modify the endwaillwake interaction (thereby also reducing the cooling effectiveness). We shall consider only
losses related to the blade passage, and neglect all others.
 167

TABLE I. Dimensionless flow parameters for case study 3.1.1.

Lrer = 0.15 m Re = 86000

Drer = 103 mls Pr =0.7 (at T = 1300 0 K)
T ref = 300 OK Pe = 60200
.1.T ref = 1000 OK Br = 154

3.1.2. Boundary Conditions

The boundary conditions are the same for all four configurations, and are shown in
Table 2.

TABLE 2. Boundary conditions for case study 3.1.1.

Inlet velocity turbulent, symmetric profile with V _=V I = 103 mls

Inlet temperature plug profile, with T~=1300 K (this value was varied to
achieve constant blade enthalpy drop, see Table 3);
Inlet turbulent quantities v3
k~ = 0.03 v? £~ = 5.10-9 .. ~ ;
Lc
Blade inner wall T;"side = 1118 K;
temperature
Outflow conditions aT = ..aw ak af
__ . = . - =. - = 0 .
an an an an '
Wall values (see Section 2) u=v=O, k=O, f= lim
y~O
fwl

3.1.3. The Numerical Procedure

Since circumferential periodicity is assumed, only a portion of the 2-D cascade (a "cell"
containing one blade) is actually computed. All relevant details about the procedure and
the numerics of this example are given in [14].

3.1.4. Results and Discussion

The velocity and temperature fields in the fluid and the blade temperature contours are
presented respectively in Fig. 2,3,4. The general features of the profiles are in qualitative
agreement with the flow patterns one would expect in such a cascade (see [14] for a
complete discussion). The entropy generation rates are shown in Fig. 5 and 6, while their
integral values over the entire flow domain are reported in Table 3. The following
conclusions can be drawn:
1) both 51' and 5, are almost entirely produced in the boundary layer around the blade
(Fig. 9,10, but see also [14 D, and are caused by high local gradients at the wall and
in the wake; a "better" fluiddynamic blade design could reduce the viscous entropy
generation rate to a certain extent, but not of a substantial amount (the turbulent
boundary layer and the turbulent wake cannot be simply dispensed with!). Thermal
irreversibilities, on the other hand, can be substantially reduced by avoiding the
strong temperature gradients at the blade walls: from this point of view, and
pending numerical validation of this conclusion, transpiration cooling (i.e.,
168

continuous ejection of the cooling air through permeable vane walls) would be the
best solution. At any rate, the thermal gradients are lower (cooler blade) for
configuration dl (Fig. 3).
2) Among the four configurations examined here, the one with the highest St is the
internally cooled blade (without discrete hole film cooling), and this was expected,
because this configuration also attains the highest wall temperatures, and therefore
the highest thermal gradients. What was not expected, however, is that Sv can be
lowered by the addition of cooling holes. A possible a posteriori physical
explanation is that the velocity gradients at the wall are significantly reduced on a
large portion of both the suction and the pressure side by the thin layers of cooling
air which remain attached to the wall and act as a sort of "slip layers" between the
viscous sublayer and the mainstream flow. The numerical values shown in Table 3
demonstrate that an acceleration of the trailing edge cooling flow (case dl) brings
about an improvement in the efficiency (sv decreases), but that an excessive
acceleration (case d2) has just the opposite effect. Adopting this point of view,
transpiration cooling could slightly increase the cascade performance, by decreasing
the viscous losses as well. As remarked in [14], these conclusions have to be
carefully checked by specific numerical simulations, and must be taken here only as
"workable hypotheses".
3) Flattening the leading edge of the blade (case d) increases Sv as well, because the
upstream section of the blade is affected by higher stagnation point losses.
4) The local Bejan number, Be = sl(st + sv), which is clearly a synthetic measure of the
relative magnitude of the two irreversibilities, attains a value of about 0.5 at the wall,
and of 0.4 - 0.6 around the cooling holes, showing that in these flow regions the
thermal and viscous contributions to the irreversible losses are comparable. In the so-
called "quasi-inviscid regions", which surround the boundary layer at the wall and
the mixing layers around the wake, Be reaches a peak value of about 0.95, showing
the diminishing importance of Sv and the effects of strong convective thermal
mixing. Upstream of the vane and in the so-called "isothermal core", which spans the
central portion of the interblade passage, Be'" 0.05.
5) The average or integral Bejan number, Be, takes values in the vicinity of 0.5 for all
configurations (see Table 3).

TABLE 3. Global results of the entropy generation calculations for case study 3.1.1
(Note: Sv' St and SlOt in WloK, Ah in J/kg, Tn = 300 OK).

Sub-Case St Sv Slot r. 11 L\h; L\hreal Be

design ("a'') 76 84.6 160.6 0.318 0.871 70.8 61.7 0.47
"b" 59.3 76.3 135.7 0.26 0.849 72.7 61.7 0.44
"e" 65.5 77.3 152.8 0.288 0.888 69.5 61.7 0.46
"d 1" 64.5 73.7 148.3 0.27 0.860 71.7 61.7 0.47
 169

3.2. INFLUENCE OF THE ANGLE OF ATTACK ON THE IRREVERSIBLE

LOSSES IN A GAS TURBINE ROTOR BLADE

Consider a two-dimensional section of a highly loaded gas turbine first-row rotor blade
(Fig. 7): for this configuration, extensive experimental data are available (see [15] for
details). Again, the profile is representative of the radial mid-section of the vane, which
is internally cooled by a radial flow of air through three separate internal slots. In all
cases examined here, the cooling air is assumed to be discharged from the vane tip. The
unsteadiness due to rotation effects is not addressed by these steady-state simulations:
attention is placed to the response of the rotor to off-design operating conditions, which
in this context means to study the effects of a variation of the angle of attack u l (angle
between the relative velocity and the geometrical tangent to the profile at the leading
edge), which takes the values _8.4°, -4.4°, -0.4° (design value), +3.6° and +7.6°.

3.2.1. Cascade Data

The midplane section of the blade on which the calculations were performed is shown in
Fig. 7, and the detailed description of the physics of the flow is given in [15]. The values
assumed of the dimensionless characteristic flow parameters are the same as those
shown in Table 1: Re has now been calculated for the relative flow (U ref = W ref = WI =
103 m1s).

3.2.2. Boundary Conditions

The boundary conditions (which are the same for all four configurations, except for the
flow direction at the inlet boundary) are identical to those displayed in Table 3.

3.2.3. The Numerical Procedure

The method, the code and the general features of the numerical solution are the same as
those of example 3.1: all relevant details are offered in [15].

3.2.4. Results and Discussion

The details of the numerical solution are not shown here: the general large-scale features
of the flow fields are similar for different angles of attack, and the differences are barely
detectable in the maps of Sv and St given in [15]). What is of interest here is to examine
in detail the entropy production rates and their distribution, and to predict the
performance of a cascade on the basis of its total entropy production. The following
remarks apply:
1. a variation of u l induces a modification of the local characteristics of the boundary
layer in the vicinity of the stagnation point, and the distribution of the viscous
entropy generation rate changes accordingly: one would expect that a departure from
the design conditions brings about an increase of the irreversible losses, i.e., that the
value of the integral of Sv over the entire domain increases with ilu (where ilu =
no matter whether the sign of ilu is positive or negative. Figure 8
Udesign - Uoff-design),
shows that this is indeed the case, and that the increase is almost quadratic (the usual
loss correlations predict a quadratic dependence of the viscous losses on uJ The
thermal entropy production, on the other hand, is not substantially affected by the
170

angle of attack, and only minor variations result from the distortion of the boundary
layer in the proximity of the stagnation point.
2. Blade performance, measured by the irreversibility coefficient I; [6], remains almost
constant for small variations of CI.., about its design value, but decreases substantially
for angles of attack far from it. This means that the blade leading edge is capable of
accommodating small variations in the flowrate without severe performance
deterioration.
3. The cooling effectiveness is only slightly affected by a variation of the angle of
attack (the blade surface heat flux profiles do not vary appreciably with CI..,. see
[IS]),: but the thermal entropy production varies of a non-negligible amount, as
shown by its global value (integral of Sf over the entire domain, figure 8). The
conclusion is that the temperature distribution around the blade, though contributing
significantly to the overall irreversibility of the expansion process, is nevertheless
not a relevant parameter in this optimisation.
4. Neither the local Bejan number, Be = stl(sv + Sl), nor its integral counterpart, Be,
show substantial variations with CI.." confirming that the ratio of the two irreversible
entropy production rates does not depend on the flowrate coefficient <1>= V m/U, at
least within the range covered by the present investigation.

These conclusions -as remarked in [15]- ought to be viewed with some caution for
three reasons: first, the simulation being at steady-state, the details of the flow are
obviously valid only as "statistical averages", and the time-dependent details of the
solution are entirely neglected; second, the turbulence has been modelled by a 2-D k-e
closure, so that the behaviour of the turbulent quantities in the vicinity of the wall has
been imposed a priori by the choice of model parameters. Now, most of Sv and St are
produced in the sub- and inner layers, and therefore the results might be strongly model-
dependent. Last, a 2-D simulation undoubtedly fails to account for both the cooling
jet/mainstream interactions and the "centrifugal" displacement of the streamlines, which
are essentially three-dimensional phenomena: so, it may well be that there are features of
both the temperature and the entropy profiles in and around the blade that cannot be
captured but with 3-D simulations.

3.3. THREE-DIMENSIONAL SIMULATIONS

The objections raised against 2-D simulations have been addressed, as suggested in
[14], in two separate and independent investigations [10,20], the first one for steady and
the second for unsteady flow in a first stage of a gas turbine. The two methods employed
different turbulence modelling techniques, and neither of them took shock-related effects
into account.

3.3.1. Steady, Uncooled Gas Turbine Rotor

A synthesis of the results presented in [20] for a blade with the midplane section shown
in figures 9 is offered in Fig. 10: the aerodynamic losses in the rotor have been
calculated independently (by integrating equation 5 over the relevant areas of the
computational domain) so that their relative importance could be assessed. The rotor
 171

blade is uncooled, so that the results are in effect of interest for the choice of the profile
shape, and can be used to directly test the effects of design changes. The use of equation
5, and its integration over separate areas of the flow, is questionable, because it is
difficult to imagine that there is no mutual influence among losses of different types;
furthermore, viscous and thermal losses cannot be properly separated. Nevertheless, the
study clearly demonstrates the feasibility of the method as a design tool.

3.3.2. Unsteady, Air-Cooled Gas Turbine Stage

It is well known from textbook analyses that unsteady effects linked to relative rotation
are important for a series of theoretical and practical reasons. The optimal design of a
cooled vane, therefore, is likely to be affected by the fully 3-D flow characteristics in the
entire stage (nozzle + rotor): to understand these characteristics, a three-dimensional
calculation was conducted on a complete first stage [10]; the entropy generation
contours obtained by that simulation are shown in Fig. 11. The entropy generation rate
was computed using equation 6. The loss patterns appear clearly from the entropy
contours: losses are relatively low in the central portion of the channel, where both Sv
and Sl are low; the contribution of the thermal gradients is noticeable around the blade,
and the local "high loss" spots correspond to the cold air injection into the main stream.
At the leading edge, an additional contribution by Sv increases the irreversible losses;
downstream, wake effects (again, mainly due to the Sv contribution) are noticeable. This
study is impressive, in that it is a real fuB-scale demonstration of the applicability of the
technique to industrial cases.

3.4. OPTIMAL CONFIGURATION OF A SINGLE FINNED TUBE IN A GAS-TO-

LIQUID HEAT EXCHANGER

In the cooling of an air stream in a finned-tube heat exchanger (for which both the tube
external diameter and the fin thickness are prescribed), there are two competing effects
to be accounted for: the friction losses, which decrease with increasing fin spacing, and
the overaB heat transfer, which prescribes the ratio of the fin diameter to the fin spacing.
In this case, an "optimal" configuration can be found by entropic analysis which is
different (and more effective) than the one obtained by separately minimising the
pressure losses or maximising the heat transfer per meter of tube.

3.4.1. Description of the Analysed Configurations

The initial configuration is the "optimal" one suggested by current engineering practice,
which dictates the choice of the outer fin diameter as a function of the tube external
diameter and of the fin spacing (see [19] for details). The configuration is depicted in
Fig. 12 and Table 4: in the numerical simulation, the tube diameter dt as well as the fin
thickness t have been kept constant, while both the fin protrusion [ and the fin spacing s
have been varied. The values assumed for the air- and material properties and the
dimensionless characteristic flow parameters are not reported here, but can be found in
[3,19].
Four different fin protrusion lengths ({ = 10,15,20 and 25 mm) have been considered,
and for each one ten different values for the fin spacing have been computed: all in all,
forty different configurations have been numerically evaluated. The fluid flowing inside
172

of the tube has been assumed to have an internal heat transfer coefficient much higher
than that of the external air, so that the external tube surface could be considered
isothermal. The Reynolds number based on the tube diameter, calculated with the fluid
and material properties described in detail in [3,19], is Re = 966, and therefore it can be
safely assumed that a laminar solution will suffice. The problem is axisymmetrical, so
that it is convenient to model only a portion of the domain: we chose a control volume
that included a slice of tube of axial length equal to (s+t)/2 and of radial width equal to
twice the fin external radius (see Fig. 12).
TABLE 4. Configuration data for case study 3.4.

3.4.2. Boundary Conditions

The boundary conditions (Table 5) are the same for all configurations, the only
difference being of course the aspect ratio of the finned tube.
TABLE 5. Boundary conditions for the finned tube case (refer to Fig. 19).

inlet velocity (section "A") plug profile with V_= 0.6 rnIs
inlet temperature (section "A") plug profile with T_ = 293 K
tube outer wall temperature (section "E") Tw = 333 K
outflow conditions (section "e")
a-Ya'n = 0 ("f' being the generic variable
u,v,w,T)
symmetry conditions sections "B" and "I"

3.4.3. The Numerical Procedure

The computational domain is shown in Fig. 13: since axial periodicity is assumed, only a
portion of the tube (a "cell" containing one fin) is actually computed; in addition, the
problem being symmetrical with respect to the longitudinal midplane through the tube
axis, only half of the tube/fin arrangement has been modelled. Details on the numerical
technique are given in [19].

3.4.4. Results and Discussion

A representative view of the velocity and temperature fields is shown in Fig. 14. These
results are in excellent agreement with an unrelated numerical and experimental data on
similar configurations [9]: for what the discussion of the flow structure is concerned, the
reader is referred to [3]. The results prompt the following remarks:
1. the heat flux from a single fin is shown in Fig. 21: for a fixed fin thickness it
increases -as expected- with the fin protrusion f, and also increases with fin spacing,
tending asymptotically to the "isolated fin" limit. As a consequence, the heat flux per
meter of tube, Qmeter = Qfi.!(S+ t), exhibits a maximum (Fig. 17, top) which
 173

corresponds approximately to a spacing of 0.025 to 0.03 m, which is identical with

the "optimal" value derived analytically in [4];
2. the viscous dissipation is -due to the low Re- of orders of magnitude lower than its
thermal counterpart: considering that the entropy generation rates scale defined by
equations 3 and 4, the order of magnitude of their ratio will be:

SO.t k gas!1T
.. - ""--- = 155000 (15)
SO,v ~U2
and the ratio between the computed macroscopic values (obtained by integrating the
local values over the entire flow domain) is about the same. With increasing fin
spacing, the velocity gradients decrease, and so does the viscous entropy generation
rate per meter of tube (Fig, 16);
3, The thermal entropy generation (Fig, 17, bottom) attains its maximum in the thermal
boundary layer and in the thermal wake of the finned tube. The thermal entropy
production per meter, on the other hand, displays a maximum which is attained for
much wider fin spacing than the corresponding maximum in the heat flux. This is
due to the development of the thermal boundary layer in the interfin space, which
gives origin to two distinct temperature profiles in the channel between fins,
depending, in a way that we are not going to discuss here, on a sort of entry length
parameter [19].

The analysis of this finned tube configuration leads to the following conclusions:
I) for the low Reynolds number considered here, Sf is much higher than Sv (their ratio
actually is of the same order of magnitude as the Brinkman number): thus, it has to
be expected that in most laminar flows, as long as the temperature difference is not
very small, the dominant loss will be that due to the thermal irreversibilities, In
particular, for air between 300 and 800 OK, Ilik ",(0.71+0.65)10-3, so that U2/ilT
ought to be of order 1000 for the two entropy production rates to be of the same
order of magnitude (while in the present example U 2/ilT = 0.012!).
2) The "optimal" fin spacing based on the maximisation of the heat flux per meter of
tube, which for the present configuration would be chosen around 2.5 mm (figure
17), still gives origin to a high thermal entropy generation rate, From a 2-nd Law
viewpoint, it would be better to use a larger fin spacing, thus reducing the
irreversible losses, and to accept the corresponding decrease in overall heat flux per
meter. In [19] a possible thermo-economic criterion for choosing the fin spacing is
presented which suggests to take, for each fin protrusion r, the spacing s for which
the thermal entropy generation rate per meter is half of its peak value: in the cases
analysed in this study, this would correspond to s - 5+6 mm.

4. Conclusions

A new procedure to assess the losses in a process or transformation in which a fluid

exchanges energy with a thermo-mechanical component is described in this paper, As a
performance parameter, the total entropy generation rate is proposed. This quantity
174

offers several advantages in evaluating the effectiveness of an energy conversion

process:
1) sro,can be computed numerically from known flow fields (where local values for
both velocity and temperature are available);
2) proper integral constraints can be imposed (like for example the overall enthalpy
drop across the cascade, or the maximum blade temperature) to represent realistic
operational conditions;
3) s,tJ1 can be conveniently split into its viscous- and thermal contributions, thus
allowing better insight in the flow phenomena;
4) s,tJ1 ontains information at two physical levels: at a local level, it shows not only
where irreversibilities are present, but also to what extent they are sensitive to design
changes; at a global -or integral- level, it gives a measure of the "degree of
irreversibility" of the flow process in the cascade, which enables a designer to assess
the relative performance of different configurations;
5) s,tJ1 can be therefore readily linked to the real efficiency of the machine of which the
cascade is a component through -for example- the "entropy loss" coefficient [6], or
the Gouy-Stodola lost work.

The procedure has been applied to realistic 2-D and 3-D models of turbomachinery
components, and for some of the configurations the geometry was varied to verify the
claim made in Section 1 that it would be possible to compute the "classical"
performance parameters on the basis of the irreversible losses generated in the flow. The
results confirm and complement those described in [14,15,19], and the conclusions
which were stated in those previous studies can be repeated here:
\- the procedure (which was applied to a stator vane in [14] and to a rotor vane in
[10,15,20) is indeed applicable to realistic 3-D analyses of an entire stage;
11- there are areas which require further investigation. First of all, the turbulence model
(especially the wall functions) is likely to affect the absolute numerical value of the
computed entropy generation rates; furthermore, unsteady and three-dimensional
effects can change the picture, as end wall and wake losses, tip leakage effects, and
interactions with upstream flow disturbances affect the relative magnitude of svand
Sf;

iii- in general, the procedure is only as good as the flow field from which the velocity-
and temperature data were extracted: if -however- experimental data about the flow
fields are known, the calculation of S,. and S, is straightforward.
iv- design decisions always involve some form of economic (sub-)optimisation: it
appears indeed possible (see also [19]) to introduce "cost figures" in the entropic
analysis and derive "engineering optima".

Since compressible flow simulations results are not yet available for an entropic
analysis like the one described here, these remain open issues as well; but it is likely that
2-D compressible simulations ought to be addressed before going into complete 3-D
simulations, which are still extremely demanding in terms of computer resources. On the
basis of the experience gained in several applications of the method outlined in this
paper (and described in more detail in the literature quoted here), the author
 175

recommends that more case studies be analysed, to better focus the numerical procedure
and to gain deeper insight in the local flow phenomena.
176

et=37.5°

.e=O.15m

Figure 1. Mid-plane section of an air-cooled first stage nozzle blade: geometry definition.

REFER. VECTOR
-+O.2147E+Ol
MAX.VEC.PLOT·D
0.2147E+Ol
AT NODE 454
COLOR CODE:

i1
VELOCrTV

S111
0.107 I!81
+01
0.716 +00
0.358 +00

Figure 2. Velocity vector field for Cass Study 3.1, Sub-case "a".
 177

CASE A: OLAUE IEH1iEi'iATUIlE flUor lLE CASE 0: OLADE TEMPEnAlUflE PIlOFILE

L. L.

CASE C: BLADE TEMPERA lURE PROF lLE

, CASE 0: OLAOE fEIO'EnA1UII[ f'1IlIl·ILE
'-"\

L. =L~._____
Figure 3. Blade temperature contours for Case 3.1.
178

VELOCITY VECTOR FIELD

,
L.

Figure 4. Details of the velocity vector fields for Sub-cases "b", "d l " and "d 2"·
 179

v [seous OHPOPY USER FUNC rr ON

CONfOUR PLOT

~
R a.'738E'03
o O.II~5C:-O'
.= 0 17)7E"n4
H O.2369E>04
.J a.JOuIE-Q4
L O.3633E-Q4

- 0 .o792€ _("J
- 0 7'2"E-0',
Z- 0 ;JOSSE'OJ
I - O. S6a 7E -I) ,
J - 0 931910-0-'
5 - 0.99SIE-O'
7 - O. 1OSSE -05
9 - O.lt2lE"05
a.; \95< ,05
O. ~2·1E~ :-~':5
-'''SEE PAL',TCvf
MrNr~~~04
o .COCCOc~J!J
>ux[>!I.i"
o. 1a5J6~ .. ·.:5
SCRE"" Ll;!u
X:-.4[N 0 td7E~0t
X~.o\X ;).! ·~2=:"0 1

I, Y!'-I(N O.2·J9E-,j~
·'.~'X J237~-'):

IL, r[O"';:> :3. vl

VISCOUS ENTROPY GENERATION USER FUNCT [ON

CONTOUR PLOT
LEGEND
A - o 57i5E·O'l
B a IIlJE·O·1
C O.2888E-O'1
;1

F - 0 5]~JE .'J,'
a-a. 75\JClE":'~ I
H - O.6663E·\;J I
[ - O. ~(ll BE • cJ" I
J - O.I097E'O~·,
M[N[MUM
a.ooOOOE-OO
MAX (MUM
0.1 \550E-05

SCflE"N LCM(TS
XM [N a 723E Hla
XMAX O.90aE.OQ
YM(N 0.2S2€.I)\
, YMAX O.27SE.O\

L,
F [OAP .6: 0 1

Figure 5. Viscous entropy generation rates for Case 3.1 (details).

180

USER FUNCrrON
CONTOUR PLOT
~
4 - 0.4699E'03
8 - O. 1470E+OJ
C - 0.2449E+04
fI \} . :I.;,~· i: • :'.;
r: 0 ':'!;':;!"0 ,1
F - O. 5369E'0.
G - O. 6366E +0 4
H - O. 7346E'0.
[ - 0.8328E+0.
J - 0.9300£'04
M[N[MUM
O.OOOOOE·OO
MAX [MUM
0.9797610'04

SC~EEI'l I r,'HT~
XM[1'l 0.133E'01

.
XI~'X 0.139E'01
YIHN 0.2IJE'01
YM.J.X 0 225::;001

L. F [OAP 6.01

THERMAL ENTROPY GENERAT[ON USER FUNC T rON

CONTOUR PLOT
LEGEND
A - 0.:'O()5E>03
8 - 0.150;'E>0·1
C - O. 250 JE '0 4
U .. : .;~ . ,:1
:' " ,;';1,',; . ,.;

i' - 0.:506E'04
G - 0.6507E'0.
H - 0.7506E'04
I - O. 9509E >0'
J - 0.9510(;.04
M[NrMU.~
O.OOOOOE·OO
MAX [MUM
o . 100 l! E -05

SCREEN LIM [IS

XM[1'l0.727E·00
XMAX 0.970E+00
YM [N O. 252E'0 I
YMAX 0.276E'01
1
F[OAP 6. a1
L.
Fixure 6. Thermal entropy generation rates fOT Case 3.1 (details).
 181

v
u

Lx
Figure 7. Mid-plane section of an air-cooled first stage rotor blade:
geometry definition for Case 3.2.

2,3 . , - - - - - - r - - - - - r - - - - - r - - - - ,

2,1

1,9 ~----t-----+----_f_-~~y

1 ,7 +----=::"-c--t-----+----_f_+~~__I

1,5 ~~~----~~------~--------~--~----~

0,9 + - - - - - t - - - - - I - - - - - _ f _ - - - - t

0,7 +-----1-----1------1-----1
-8,4 -4,4 -0,4 3,6 7,6
a.

I-o-SUSt,d -X-Sv/Sv,d -o-StoUStot,d -6-Y'!;,S I

Figure 8. Case 3.2: variation of the global entropy generation rates
as a function of the angle of attack.
182

Figure 9. Case Study 3.3.1: geometry definition (midplane section).

Other Jiiiiiiiiiiiiiiiiiii~--------------------------------l
Wake Loss

Casing Vortex

Hub Vortex

Tip Leak

Press. Side b.1.

Sucl.Side b.l.

Casing b.1.

Hub b.l. e!!!!!!!!!!!!~!!!!!!!!!!-._----_----_---.J

o 5 10 15 20 25
entropy generation rates in %

Figure 10. Case Study 3.3.1: entropy generation rates attributed to the single losses (adapted from [20]).
 183

Fi/iure 11. Case Study 3.3.2: entropy generation rates on the midplane section.

Fi/iure 12. A "unit element" of a water-to-finned tube heat exchanger.

184

Figure 13. Case Study 3.4: the computational mesh.

Figure 14. Case Study 3.4: velocity and temperature fields for the original design configuration.

3.50

3.00

.'-2.51....)
9'-2(.",)
.'-1.5(...)
if ,.,(....)
1.00

0.50

2 l 4 5 e 7
Inl..,1n o""cl.,. (mm)

Figure 15. Case Study 3.4: heat flux from one single fin.
 185

120 ...
VISCOUS ENTROPY PER METER

100
><
E 80
......
~
>-
g- 60
.l:
i
ft
~ 40
~
-;:
20

0
0 2 3 5 6 7 8
int.rfin spacing (mm)

Figure 16. Case Study 3.4: viscous entropy generation rates per meter of tube.

800

700

600

ESOO

~ 400
-;
~300

200

100 .......-.. ~~ . ~. . ~-......-..t--..~-. . . . .L. .-.. -

0+---~--~--~--_+--_+--_+--_4--~
a 2 J 5 6 7 8
Inloriin spacing (mm)

140,000
THERMAL ENTROPY PER METER

120.000·+·················,...,····-·····~···············, .•................ , ...•......... .,.

><
E 100.000

~ 030.000
>-
"-
e
~ 60.000
-;
g 40.000
:;
20.000

o 2 J ~ 5 6
lntoriln opaclng (mm)

Figure 17. Case Study 3.4: thermal entropy generation rates and heat flux per meter of tube.
186

5. References

I. A.Bejan: Entropy Generation through Heat and Fluid Flow, Wiley, New York, 1982.
2. A.Bejan, E.Sciubba: The optimal spacing of parallel plates cooled by forced convection, Inti. J. (!t" H. &
M. Tran.~f, 35, n.l2, 1992.
3. P.L.Benedetti, E.Sciubba: Numerical calculation (It" the local rate (!t" entropy generation in the flow
around a heatedfinned tube, in ASME AES-3, 1993.
4. A.E.Biermann, H.H.Ellenbrock Jr.: The design of fins for air-cooled cylinders, NACA Rep. 726, 1976.
5. RJ.Boyle: Navier-Stokes analysis of turbine blade heat transfer, J.ofTurbotnach., 113, Jul. 1991.
6. J.D.Denton: Loss mechanisms in turbomachines, J.(!t"Turbomach., 115, Oct. 1993.
7. DJ.Domey, R.L.Davis: Navier-Stokes analysis of turbine blade heat transfer and performance, J.(!t"
Turbomach., 114, Oct. 1992.
8. M.K.Drost, M.D.White: Numerical predictions of local entropy generation in an impinging jet, ASME
J.H.T., 113, 1991.
9. A.Haught, M.S.Engelman: Numerical and experimental simulation of air flow heating in a tube fin heat
exchanger, in ASME HTD-I03, 1988.
10. A.Holzwarth: Private Communication, April 1997.
II. T.VJones, C.M.B.Russell: Local heat transfer coefficients on finned tubes, in ASME HTD-21, 1981.
12. S.Mereu, E.Sciubba, A.Bejan: The optimal cooling of a stack of heat generating boards with fixed
pressure drop, flowrate or pumping power, Inti. J. (!t"H. & M. Tran.ff, 36, n.15, 1992.
13. J.Moore, J.G.Moore: Entropy production rates from viscous flow calculations, Part 1- a turbulent
boundary layer, ASME paper 83-GT-70, 1983.
14. G.Natalini, E.Sciubba: Entropy generation rates in air-cooled gas turbine nozzles: a numerical
configuration study, Proc. IGTI-COGEN-TURBO, 1994.
15. G.Natalini, E.Sciubba: Entropy generation in a 2-D cascade at different angles of attack: a numerical
study. Proc. IGTI-COGEN-TURBO, 1997.
16. S.Paoletti, F.Rispoli, E.Sciubba: Calculation of exergetic losses in compact heat exchanger passages, in
ASMEAES-I0I2,1989.
17. E.Sciubba: Numerical calculation of local irreversibilities in compact heat exchangers, in Proc. NATO-
TIBTD Workshop 011 2nd Law, Erciyes Univ., Kayseri 1990.
18. E.Sciubba: Entropy generation rates as a true measure of thermal and viscous losses in thermo-
mechanical components and systems, Pro£.". Bucharest Heat Tran.~t"er and Thermodynamics Workshop,
Polytechnic Univ. of Bucharest, 1994.
19. E.Sciubba: A minimum entropy generation procedure for the discrete pseudo-optimisation of finned tube
heat exchangers, Rev.Gen.Therm., vo1.35, 1996.
20. TJ.Sullivan: Novel aerodynamic loss analysis technique based on CFD predictions of local entropy
production, SAE Paper 951430, 1995.
21. K.Takeishi, S.Aoki, T.Sato, K.Tsukagoshi: Film cooling on a gas turbine rotor blade. J.(!t"Turbomach.,
114,Oct.l992.
AVAII.ABLE ENERGY VERSUS ENTROPY

A.ozTORK
jstanbul Technical University
Mechanical Engineering Department
GUmii~suyu, jstanbul80191 - Turkey

1. Introduction

As known, the second law of thennodynamics is an entropy law. The foundation of the
second law of thennodynamics is usually attributed to Clausius and Kelvin, around
1850, after S.Carnot's principles "heat cannot be converted completely and
continuously into work" [1]. In Classical Thennodynamics, the first law of
thennodynamics is introduced and then followed with a discussion of the defmitions of
the reversible cycle (Carnot cycle) and --related to Carnot cycle-- absolute temperature.
Next, the Kelvin-Planck statement and Clausius statement for the second law of
thennodynamics are introduced. Then the Clausius inequality

f8~Q ~o (1)

is proved in different ways [2-5]. Entropy is then defined and the entropy equation is
fonnulated as the second law of thennodynamics (Fig. 1). The application of the second
law of thennodynamics to the various processes has been presented in different ways,
such as, (i) the net entropy increase (or entropy generation, or entropy production. or
entropy creation, etc.). (ii) irreversibility, (iii) lost work, and finally (iv) the loss of
available energy (or exergy).
Usually, it is considered that the available energy equation is the combination of the
first and the second laws of thennodynamics. in other words, the combination of the
energy and entropy equations [6]. Indeed, if those two equations are written for control
mass or control volume. the available energy equation can be obtained with some
mathematical and thennodynamical procedures [5,6]. Even though some authors. like
Obert [7], have concluded that "The available energy of the isolated system decreases
in all real processes (and is conserved in reversible processes)," traditionally all
scholars introduce the formulations the available energy equation after the entropy
equation.
This work introduces a new and conceptually easier approach based on a direct
proof of the known statement of inequality. In this approach definitions are provided
for concepts like the available energy of heat (maximum work produced from heat) and
the available energy of work (useful work). Then the known inequality is proven
187
A. Bejan and E. Mamut (eds.), Thermodynamic Optimization o/Complex Energy Systems, 187-194.
@ 1999 Kluwer Academic Publishers.
188

directly in terms of the available energy of heat and the available energy of work.
In the literature. this well-known inequality statement is usually induced after
Clausius inequality, in terms of entropy and with long computations based on concepts
such as reversible work, maximum reversible useful work, and so on. The statement as
developed by this author is proved directly and is thus another version of the second law
of thermodynamics. I have termed this statement "the new inequality" [8].

t
The First Law of Thennodynamics

(OE Q + oEw )= 0

..... ...
........
.........
New Inequality Clausius Inequality

--------~-------

'tI

Available Energy Equation

--·---IL-__ E_n_tr_o_p_y_E_q_ua_t_io_n_.....

-----. new way of obtaining available energy equation

known way of obtaining available energy equation

Figure J. Illustration of available energy versus entropy.

 189

2. New Inequality

Recalling the reversible cycle (Carnot cycle) and the existence of absolute temperature,
the maximum work which can be achieved from the amount of heat (EQ) at the
temperature level T with respect to To ambient temperature is called the available
energy ofheat
(2)

where, 1lco = 1- To / T is the Carnot efficiency. According to the second law of

thermodynamics, the actual work obtained from the same amount of heat is less than the
available energy of heat and if the process were completely reversible they could be
equal (8Ew ~ 8Ko).
The available energy ofthe work is
8Kw = 8Ew + Po dV (3)
which is actually the useful work.
Now, let us determine the amount of work done on a control mass during the
infinitesimal reversible process shown in Fig. 2. Let us assume the amount of heat
exchange of the control mass is 8EQ during this process. For only internal
irreversibilities, the heat exchange can be done with the ambient by means of reversible
engine. Applying the first law and the definition of the absolute temperature scale to the
reversible engine the following equations can be written, respectively

Po, To

Figure 2. The illustration of the new inequality and Clausius inequality.

(4)

-8E Q 8E QO
--+--=0 (5)
T To
where, 8EQ is the heat transfer to the control mass and therefore the minus sign (-)
comes for the heat transfer of the reversible engine. The sign convention is used in this
work, as both the heat and the work transferred to the control mass is considered
positive. From the Eqs.(2), (4) and (5) eliminating 8EQO and 8EQ • the work of the
190

reversible engine can be found as follows,

8E wc = (1- ~) 8E Q = 8K Q (6)

On the other hand, the work of the control mass, from Eq.(2), can be written as follows
(7)

Now, in a manner which is conceptually very simple, the work done on the control
mass and on the reversible engine can be written in terms of available energy of heat
and the available energy of work, from the Eqs.(6) and (7)
(8)

Recalling the Kelvin-Planck statement for the second law of thermodynamics "It is
impossible to construct a heat engine using one heat reservoir to produce the work, ..
so, work can be converted to heat and the analytical statement is reduced as follows. by
means of our sign convention
f8E w ~0 (9)

Applying the Kelvin-Planck statement to the Eq.(8), the new inequality for the cyclic
process of a control mass

(10)

is developed directly, because V is a property of the control mass and the cyclic integral
of a property is zero. Now, from this, the second law of thermodynamics can be stated
as follows: "During the cycle of a control mass, the available energy flow to and/or
from the system as heat and/or work can not be negative." In other words, available
energy is destructed because of internal irreversibilities. If the cycle is internally
reversible there is no destruction of available energy.
With the combination of the flrst law of thermodynamics

(11)

Clausius inequality can be derived from this new inequality using the Eqs. (2) and (3).

f[(l- ~) 8E Q +{8E w +PodV)]~O (12)

Thus both inequalities are equivalent.

3. Available Energy - Availability (Available Internal Energy)

The available energy of a system may be deflned as follows: the maximum possible
useful work that could be achieved by bringing the system to a dead state --a state in
equilibrium to its environment- by ideal processes. If the available energy of a system
 191

is denoted by K, the available internal energy (B) which is an extensive property of a

substance is similar to internal energy, however it is also associated with environmental
properties, such as Po and To. Then, subtracting available kinetic energy and available
potential energy of the system from the available energy, the availability is detennined
as follows:

(13)

where, m is the mass, V is the velocity, z is the elevation of the system, Zo is the
elevation of dead state which can be taken as zero, and g is the gravitational
acceleration.
The available energy of a control mass, in this work, will be directly derived from
the new inequality in like manner to the realization of energy from the first law of
thennodynamics (Eq.ll). Consider any two arbitrary reversible processes of a system,
lA2 and IB2, which are between the initial state 1 and the final state 2. Witll the
application of a third reversible process 2C 1, each of these two processes becomes
reversible cycles; tllen tlle new inequality can be applied (Fig. 3). The new inequality
will be zero as follows.
___ ~-.2

1
Figure 3. Two reversible cycles used to determine the available energy as a property.

[ ( 8K Q +8Kw)+ S:J8KQ +8K w}=O (14a)

fB(8K Q +8K w}+ lie (8K Q +8K w}=O (l4b)

Subtracting Eq.(l4b) from Eq.(l4a) it is concluded tllat

1 2
IA
(8KQ +8Kw) = r2 (8KQ +8Kw) = 0
.h
(15)

This shows that the integral of (8Ko+8K w)rev is the same for all reversible paths
between state 1 and state 2 since these two processes (A and B) have been chosen
arbitrarily. This means the value of this integral depends only on the end states, but not
the path. Thus, one can conclude that the quantity of (8KQ+8K w)rev defines the change
of a thennodynamic property that is obviously the available energy of the control mass:

dK = (8K Q +8K w) = dB + dK KE +dK PE (16)

rev
192

By substituting Eqs.(2) and (3) into Eq.(16) and neglecting both the availa.ble kinetic
and potential energies, the differential change of available internal energy,

dB=[(I- To}mQ +(8E w +PodV)] (17)

T rev

and eliminating the heat by means of the first law of thermodynamics and with the only
moving boundary work 8Ew = -P dV, differential of available internal energy per unit
mass can be determined as

(18)

and then substituting internal energy equation into the above equation,

db=(l- ;)CvdT+[(T-To)(:)v -(P-Po)]dV (19)

can be obtained without entropy as well as the other properties [9]. On the other hand,
arranging the Eq.(17) by means of the first law of the thennodynamics (8E Q+8E w = dU)
8E Q)
dB=dU+PodV-To( T (20)
rev
and using the definition of entropy the well-known relation can be derived easily
dB = dU + Po dV - To dS (21)

4. Available Energy Equation

Consider two arbitrary processes of a system between end states 1 and 2, and that one is
reversible (A) and the other irreversible (B). These two cyclic processes are then
concluded with an arbitrary reversible process (C) (Fig. 4). Because the cycle 1A2C 1 is
reversible,

r2
JIA
(8K Q +8Kw)
rev
+ rl
J2C
(8K Q +8Kw)
rev
=0 (22a)

can be written similar to Eq.(14a). Applying the new inequality to the irreversible
cycle, it gives the result

r2
JI B
(8K Q +8Kw)
iff
+ r
J2C
l
(8K Q +8K w ) rev > 0 (22b)

Subtracting Eq.(22b) from Eq.(22a) the following inequality can be written:

r
2
JIA
(8K Q + 8K w)
rev
< r (8K
2
JIB Q + 8K w)
irr
(23)
 193

Since the process IA2 is reversible, by substituting Eq.(16), without using entropy we
are able to obtain the available energy equation for a system

Figure 4. The reversible and irreversible cycles used to develop the available energy equation.

dK :::;; (BK Q +BKw) (24)

and the available energy loss for a process.

K L12 ,int = f(BK Q +BKw) +K, -K2 (25)

In the Eq. (24), equality assigns for a reversible process and the inequality for an irre-
versible process. The available energy loss, in these equations, are due to internal
irreversibilities. If the control mass contacts more than one reservoir having
temperatures of TRi , the total loss of available energy, including external
irreversibilities, is obtained as follows:

KLI2 = ~(l- ;: )E Qi ,12 + EWuse.l2 + K, -K2 (26)

by adding and subtracting the summation of L(EQirrRi) where, EQi s are the heat transfer
interactions with the heat reservoirs TRI'

5. Conclusion

This is a new approach for obtaining availability which is a property of substances and
the available energy equation directly from the new inequality without involving the
concept of entropy. This considerably simplifies the traditional procedures and these
concepts become much easier to understand, especially for students who had often
found the traditional procedure ambiguous. Previously some authors had tried not to
use entropy, however that approach had made the discussion even more obscure [10].
The new inequality is much simpler and more meaningful than Clausius inequality
because it involves the available energy of heat and work instead of "transformation",
194

QIT. The new inequality is also superior in that it resembles the first law of
thermodynamics (Eq. 11).
By means of the new inequality, the second law of thermodynamics becomes easily
understandable to everyone, not only just to students of the subject.
Of course, at this point, we can not abandon the concept of entropy, but initially we
can discuss the available energy equation, and then follow with the entropy equation.
Obviously, the equation
dK = 8K Q +8K w (27)

for a reversible process and

dK = 8Kw (28)
are achieved for a reversible adiabatic process from the Eq.(24). For an adiabatic
control mass, the second law of thermodynamics results
dK ::;; 8Kw (29)
and for an isolated system
dK ::;;0 (30)
the available energy of the system decreases and the change of available energy
becomes zero for equilibrium.
Finally, in parallel to Clausius' famous expression, we can say that "the energy of
the universe is constant, but the available energy of the universe decreases
continuously. "

References

I. Carnot, S. (1824) Reflections on the motive power of fire. and on machines fitted to develop that power,
in J. Kestin, (ed.), The Second Law of Thermodynamics Benchmark Papers on Energy. Vol. 5. Dowden.
Hutchington & Ross. Inc.1976.
2. Van Wylen G.J., and Sonntag, R.E. (1986) Fundamentals orClasslcal Thermodynamics. 3rd Ed., Wiley.
3. Lee, F.L. and Sears, F.W. (1963) Thermodynamics, Second Ed., Addison-Wesley.
4. Wark, K. (1983) Thermodynamics. 4th ed., McGraw-Hill.
5. Oztiirk, A. and Klh~, A. (1993) Thermodynamics With Solved Problems, 3rd Ed., yaglayan (in Turkish).
6. Bejan, A. (1988) Advanced Engineering Thermodynamics. Wiley.
7. Obert, E. F. (1960) Concepts of Thermodynamics. McGraw-Hill.
8. Oztiirk, A. (1996) Teaching second law without entropy, in G. Svedberg et.al. (eds.) Efficiency. Costs.
OptimizatIOn, Simulation and Environmental Aspects of Energy Systems, ECOS'96 June 22-27,
Stockholm-Sweden.
9. Oztiirk, A. (1997) Some thermodynamic relations involving availability not entropy, in G. Manfrida et.al.
(eds.) Florence World Energy Research Symposium, FLOWERS '97, July 30-August I, Florence-Italy.
10. Lucca, G. (1991) On the opportunity, from the didactic point of view, to derive entropy from exergy, in
Reistad, G.M. et.al. (eds.), Second Law AnalYSis-Industrial and EnVironmental Applications, AES-
Vol.25/HTD-Vol.J9J, Winter Annual Meeting of ASME, the Heat Transfer Division, Dec.I-6, Atlanta,
Georgia.
EXERGY ANALYSIS IN THE PROCESS INDUSTRY

R.L. CORNELISSEN and G.G. HIRS

University of Twente, Dept. of Mechanical Eng.,
Chair of Energy Technology
P. O. Box 217, 7500 AE Enschede, The Netherlands

1. Introduction

The method of exergy analysis, as it is used nowadays, has been developed by Szargut
[1]. He and his co-workers have analysed several plants systematically. In his method
the complete production chain for the product has been analysed. Another
systematically analysis of process plants is performed by Hedman [2] in the USA. His
work is limited to the analysis of plants only, without adding new elements to the
method of exergy analysis.
Nowadays, the use of exergy analysis in the process industry in the Netherlands is
increasing steadily. Most big companies in this industry have experienced, often with
help of the universities, sometimes without, the use of exergy analysis. The reactions of
the companies are positive. Shell, Akzo Nobel and DSM, three main chemical
companies have performed exergy analysis of several plants. On basis of this analysis,
several improvements are suggested. However, the financial pay back period of these
improvements is in many cases too long, because investments for energy saving are
often high in existing plants. So, it is suggested to implement the use of exergy analysis
in the design process of new plants.
Hoogovens, the Dutch steel maker, has used exergy analysis to analyse a new
production process of steel making [3]. With exergy analysis a comparison is made with
the blast furnace process. Furthermore, the locations of the losses are found in the two
processes. On basis of this knowledge, the existing process can be optimised, while for
the new process more insight is gained on which components major improvements are
possible. Uni1ever has used exergy analysis for analysing its products. The complete
life cycle of the washing process has been carried out [4]. The insights of this analysis
are used to develop better washing powders. New products designs have been tested on
basis of exergy analysis. The results have been used to improve the products and even to
stop a product development.
In this article a cryogenic air separation plant is analysed. The analysis is performed
on a plant of Hoek Loos in the Rotterdam harbour area.
195
A. Bejan and E. Mamut (eds.), Thermodynamic Optimization of Complex Energy Systems, 195-208.
© 1999 Kluwer Academic Publishers.
196

2. Method

The exergy analysis consist of the following steps:

1. Setting up of the mass balance. This balance has to be closed

2. Setting up of the energy balance. This balance has also to be closed. Quite often,
(parts of) the energy balance is used to set up the material balance.
3. Carrying out of the exergy analysis.

For performing an exergy analysis a reference state has to be chosen. In the Netherlands
the reference system of Szargut [1] is mostly used. The standard environmental
conditions are a reference temperature of 298.15 K and a pressure of 1 atm. However,
sometimes it is necessary to change these conditions. For example, in the case of a
power plant, which use cooling water of 15°C. Problems arise when two reference
conditions are required. For example, as in the just mentioned power plant the
temperature of the inlet air is 100e. A choice between the reference temperature of 10
or 15°C has to be made. In both cases one of the inputs, water or air, will have an
exergy content. No systematic approach for this situation has been developed yet.
Most exergy analyses for process plants are performed with flowsheeters. The most
common flowsheeter is ASPEN PLUS. A special exergy calculating routine,
EXERCOM, has been developed for ASPEN PLUS by STORK ENGINEERING [5].
The routine facilitates the performing of a exergy analysis greatly.
On basis of the results of the exergy analysis improvements are suggested.
Suggestions for improvements are based on two criteria, the exergy losses and the
efficiency of the components. On basis of the exergy losses the maximum amount of
exergy saving is known, while the efficiency of the components gives an indication of
the technical possibilities for improvements. No systematic method for improvements
has been developed yet, as is the case for pinch technology.

3. Cryogenic Air Separation Plant

The analysed plant uses the principle of two-column separation. This is based on a low
and a high pressure distillation column. The condenser of the high pressure lower
column (LC) is cooled by the reboiler of the low pressure upper column (UC). This
principle is widely used for cryogenic air separation plants. The flowsheet of the plant is
shown in Figure 1. The incoming air is compressed to 6.2 bar and after cooling in the
spray cooler (SC) water and carbon dioxide are removed by adsorption in the molecular
filters. Subsequently, the air is cooled in the main heat exchanger (MHE) to -172 °C and
enters the lower column (LC), which has 54 stages. Here, the air is separated into the
top product nitrogen, which is partly liquefied in the condenser and the bottom product
(B), which is a mixture of 38 vol.% oxygen, 60 vol. % nitrogen and 1.6 vol. % argon.
The gaseous part of the nitrogen (TN) is liquefied in a special section. Here, the
nitrogen is compressed to 46 bar with the aid of compressor 1,2 and 3. Compressor 3 is
driven by the hot expansion turbine (HET) and compressor 2 by the cold expansion
turbine (CET). The liquefaction of the stream takes place in heat exchanger 2 (HE 2).
 LlQUEFACTION UNIT DlSTD..LAnON UNIT ARGON PURIFYING UNIT

~~
.....
:!l
~
'"
(1) :l"
(1)

"
'0> (1)
~
..., 0
a""
o· So
:l (1)

'" l
_ :0>
0> 0>
:l -
~~
(1)
p..

a
~
g
n°
~.

' -0
--.I
-
198

Further cooling of the liquid nitrogen takes place in the throttling valve (TV). The
nitrogen stays almost completely liquefied after passing this valve and enters the lower
column again. The bottom product and the liquid nitrogen leaving the lower column are
throttled to ca. 1.5 bar in valve I(JT 1) and valve 2 (JT2) and enter the upper column
(UC), which has 96 stages. Of this column the top product is gaseous (GAN) and liquid
nitrogen (LIN) and the bottom product is gaseous (GOX) and liquid oxygen (LOX).
One of the side products is unpurified argon, consisting of argon with oxygen and
nitrogen. Most of this oxygen and nitrogen returns from the crude argon distillation
column to the upper column. The other side product is purge gas (PC), which is used to
clean the molecular filters. It consists mainly of nitrogen and the gas is heated to l70°C
before cleaning the filters. The top product of the crude argon distillation column
(CRAR) enters the argon purifying unit, where it is compressed to 3 bar and
deoxygenated by burning the oxygen with hydrogen. After cooling, separating the water
and further cooling to -180°C the argon is further purified in the pure argon distillation
column (PA). Compression of the products to the desired pressure is left out in this
analysis. The flow and product specification are displayed in Table 1. Because of the
flexibility of the plant variations of the flow are possible. The analysis has been
perfonned with the flow given in Table 1.

TABLE 1. Flow and product specification.

Components Flow(lll.kg/s)
Air 16.39
Gaseous Oxygen (GOX) 2.82 99.5
Liquid Oxygen (LOX) 0.70 99.6
Gaseous Nitrogen 6.18 99.95
(GAN)
Liquid Nitrogen (LIN) 2.86 99.999
Liquid Argon (LAR) 0.13 99

4. Rational efficiency of the different units

The general fonn of the rational efficiency is defined as the ratio of the desired exergy
output to the exergy used [6]:
E desired output
(1)
Eused

where
Eused = E desired output +j (2)

and with help offonnula (2) can (1) by rewritten in

 199

j
If =1--.- (3)
Eused

For the compressors the following formula is used

(4)

where W is the work input of the compressor. For turbines the rational efficiency is
defined as the work output divided by the exergy decrease of the incoming and outing
flows, in formula form

WOUI
Ifturbines =. . (5)
EOUI -Ein

For the sake of simplicity no distinction is made in the rational efficiency of the heat
exchanger for the thermal and mechanical component. The rational efficiency is then
defined as

Ecold,out - Ecold,in (6a)

If HE, heating = . .
Ehot,in - Ehot,out

for operating above To and

. .
Ecold,out-E cold,in
If HE, heating = . . (6b)
Ehot,in - Ehot,out

for operating below To.

No chemical reactions occur in the heat exchangers, so the difference in exergy of the
streams is taken, because the chemical component will have no influence on the rational
efficiency. Because no distinction is made between the thermal and mechanical
component in the case of the throttling valve it was not possible to define a desired
product, so no efficiency could be calculated. No desired product can be defmed in the
case of mixing and also no rational efficiency can be obtained.

For the air separation and cleaning unit the exergy balance l is

I Streams STl ,in and STl ,out refer to the incoming and outgoing streams of steam turbine I (STl). This
notation is used further on for more components displayed in Fig. I.
 200

The desired product is

Edesiredoutput = E AIRIN - E AIR (8)
which leads to the following expression for the exergy used:

The rational efficiency is then

EAIRIN - EAIR (10)

or in the rewritten fonn by (3)

If ACC=
E
AIRIN - AIR E
---'==-'-----,----
EAIRIN- EAIR +j ACC

For the lower distillation column the exergy balance is

(11)

The desired product is the increase of the chemical component

E desired output = (L Ef- E~ix) (12)

fr
=E~+E~ +E;' -EXIRC -E~QN

where the mixture is the incoming stream(s) and the fractions are the leaving streams.
E~QN and E;' are equal and can be left out for further analysis. This leads to the
following expression for the exergy used:

. _ ( . ph _ '" . ph) .Q
Eused - Emix £oJ Ei + Econd (13)
fr
_( . ph . ph . ph . ph . ph ) .Q
- EAIRC+ ELQN-EB - ED - ETN +Econd

The rational efficiency for the lower distillation column is then

 201

1fI LC = ( . ph . ph . ph . ph . ph ) .Q (14)
EAIRC+ ELQN-EB - ED - ErN +Econd

For the upper distillation column the exergy balance is

(15)

where the fractions (fr) are GANC, LIN, CRAR, PC, GC and LOX. For the upper
column, including the crude argon distillation column, the desired product is

Q
• (" • 0
Edesiredoutput = .t-Ei -Emix +Econd
• 0 ) .
(16)
Ir

where mix is JTl, in and JT2, out and this results to the following expression for the
exergy used:

. _ ( . ph " . ph) (17)

Eused - Emix-.t-Ei
Ir

The rational efficiency for the upper distillation column is then

(" . 0 . 0 ) • Q
.t- Ei - Emix + Econd
Ir
(18)
1fI uc = ( . ph _" . ph)
Em,x .t-E,
Ir

Because there is no heat exchange with the environment, the rational efficiency for the
distillation section is

(Lg-E~IRC)
Ir
1fI DIST = -(""-E·"""p,:-,---"-E-·P-:-h-) (19)
AIRC .t- i
Ir

as can be derived by setting up the exergy balance. The fractions are the flows leaving
the section. The same formula can be applied for the argon purifying unit by replacing
AIRC by CRAR in (19). The outgoing fraction is then only the LAR. The top product is
released and dissipated in the environment.
The desired output of liquefaction unit is seen as the exergy increase of the outgoing
flow compared to the incoming flow because of the cooling and liquefying of the
nitrogen. By setting up the exergy balance and defming the desired exergy output and
thus the exergy used as just described the following formula is obtained
202

(20)

The exergy balance for the whole air separation plant is given by:

Eair+ L ESteam,in +
all
Wfreon = frail
L (E~~l,i+ En + L ESteam,out+ L ii
all all
(21)

where LESteam,in and LESteam,out are all steam in- and outputs of the plant,
all all
respectively, Eair is the exergy of the incoming air and L E~~l,i means the physical
frail
exergy of the fractions leaving the distillation plant, which are GAN, LIN, GOX, LOX
and LAR as displayed in Figure 1. This physical exergy can be used in other processes.
From (9) it is clear that the exergy used is:

E used = L E Steam,in - L E Steam,out +

all all
Wfreon (22)

This leads with reference to (2) and with the aid of expressions (21) and (22) to the
following expression for the desired exergy output:

E desired output = L (E~~t.i+ En - E air (23)

frail

Using (1), the rational efficiency for the whole air separation plant is given by:

'" (.ph '0) .

.... Eoul,i+Ei - Eair
frail
If/plant (24)
L ESteam,in - L ESteam,out + Wfreon
all all

or from (3)
'" (·ph '0) .
.... EOUl,i+Ei - Eair
frail
If/plant = - - " - - : - h - - - - - - - (25)
L (E~ut.i+Ef)- Eair+Lii
frail all

It could be argued that the desired product of the plant is the chemical exergy increase
only. This would lead to the following formula for the whole plant
 203

(26)
If/ plant = '" . ph . ph "'. '" . W·
£.... E out.i - E air + £.... E Steam, in - £.... E Steam,out + freon
frail all all

5. Exergy Analysis

5.1. AIR COMPRESSOR AND CLEANING UNIT

The main exergy loss, 1708 kW, is caused by the compression of the air to 6.2 bar as
can be seen in Table 2. The exergy loss of steam turbine 1 (STl), which delivers the
required 4.3 MW of power for the air compressor (AC), is 708 kW (see Table 3). The
consumption of steam of 400°C and 40 bar is 15.5 ton a hour. The rational efficiency of
steam turbine 1 as calculated with (5) is 0.86. For cooling the air to 7°C the electricity
use for the freon unit is 110 kW freon ). Most of the exergy increase due to the cooling (W
of the air is lost in the molecular filter, because the air is here heated to just below the
environmental temperature. The molecular filters are cleaned by the hot purge gas. By
the heating of the purge gas to 170°C with steam of 210°C and 11 bar and dumping the
gas into the environment after the cleaning of the filters 225 kW of exergy is lost. The
rational efficiency of the air compressor and cleaning unit is calculated to be 0.48 with
formula 10.

TABLE 2. The in- and outgoing flows of the air compressor (AC).

Components Temp.
Press. ····························F1ov/i .................... ·······Enthajpy . . . ······Exergy
. (i!.l~C:). ..J~Il.~) (k~/sc:~l Q11 ~~2 (inJc...Wl
AIR 20.0 1.01 16.39 -2638.9 0.7
Work input 4334.9 4334.9
Heat transfer to envir. -3362.6 0
AIRP 79.4 6.20 -16.39 1666.7 -2627.8
Total 0.1 1707.8
I) The minus sign means the stream is leaving the AC

TABLE 3. The in- and outgoing flows of steam turbine I (STl) .

Press. . .......... ···F1owl)

.....................................................................•............. -.....•.. -............ -...........................................................................-
Temp. Enthalpy Exergy
(in DC)
Components
in
Steam (STITn)-- ---- - . .. 400.0 ---
(in bar)
40.00 ..
. __ (kg/sec) _..
4.31
(in kW)_ . _ . (in Kw)
-54921 5497.3
Steam out (Stout) 29.6 0.04 -4.31 59256 -454.1
Work output -4335 -4334.9
Total o 708.3
204

5.2. MAIN HEAT EXCHANGER (MHE)

The exergy losses are caused by temperature difference between the hot and cold
streams and pressure losses. The mean temperature difference is 4.2 K. The rational
efficiency is 0.86 as calculated by (6b), where AIRIN is the hot stream. In Figure 1 INA
and INB are displayed as one stream (IN). In reality Ml is split in the main stream INA
and a small stream INB in the main heat exchanger (MHE) and combined at constant
temperature after heat exchanger 1 (HE 1). The exergy calculations take into account
this split as can be seen in Table 4.

TABLE 4. The in- and outgoing flows of the main heat exchanger.

Components
............•...••................................................-•.....
Temp. Press. 'lilowi j'" ·················Enth;!ipy·· Exergy
(in 0C) (in bar) (kg/sec) (in kW) (in kW)
AIRIN 20.0 6.20 16.20 -110.1 2567.6
GOXC -178.1 1.67 2.82 -527.4 81.3
Purge cold (PC) -180.6 1.43 3.49 -743.8 665.9
MI -177.5 5.85 3.60 -795.1 I 195.1
GANC(GC) -182.0 1.40 6.20 -1349.6 1292.2
AIRC -172.6 5.85 -16.20 3373.5 -4688.0
GOX 17.0 1.47 -2.82 21.9 -426.5
INA (IN) 17.0 5.75 -3.50 34.9 -626.3
INB (IN) -84.0 5.75 -0.10 11.7 -20.7
GAN 17.0 1.30 -6.20 53.9 -293.4
PURGE 17.0 1.43 -3.49 30.1 -145.6
Total -0.3 333.2
········WThe·minusslgn·meansthestrearn·isleavingthe··MHE:
5.3. DISTILLATION UNIT

Because of its complex nature no complete flow scheme, including all stream data, has
been given in this article. In Table 5 it can be seen how the different components
operate by showing the exergy losses and rational efficiencies. The heat transfer
between the reboiler of the upper column and the condenser of the lower column cause
an exergy loss of 82 kW due to the temperature difference of 1.1 K. This results
according to (6b) in a rational efficiency of 0.98. The throttling valves (ITI and IT2)
between the columns causes an exergy loss of 125 kW. The exergy loss in the lower
column (LC) is 62 kW and the exergy loss in the upper column (DC) and crude argon
column is 487 kW. Of the exergy loss in the crude argon column 108 kW is caused in
the condenser by the temperature difference of 2.8 K between the coolant and the top of
the column. The chemical exergy increase of the products of the lower and upper
distillation column is 242.7 kJ and 428.5 kJ, respectively. Using (14) and (18) the
rational efficiency of the lower and upper column were found to be 0.80 and 0.92,
respectively. The high efficiency of the upper column is caused by the high exergy
transfer from its reboiler to the condenser of the lower column. An exergy loss of 32
kW takes place in the heat exchanger (not displayed in Figure 1), which cools the
products of the lower column, B and D, before going into the valves which lead to the
 205

upper column. Using (6b) a rational efficiency of 0.88 is obtained for this heat
exchanger. The rational efficiency of the distillation section is 0.46 as calculated by (19)
with a chemical exergy increase of67l.2 kJ.

TABLE 5. Results of the exergy analysis in the distillation unit.

Unit exergy losses (in kW) rational efficiency

heat exchange reboiler/condenser 82 0.98
Throttling valves columns 125
Lower distillation column 62 0.80
Upper distillation column 487 0.91
Heat exchanger 32 0.88
Total 788 0.46

5.4. LIQUEFACTION UNIT

Because of its complex nature the incoming and outgoing flows of the different
components are not given. In Table 6 it can be seen how the different components
operate by showing the exergy losses and rational efficiencies.

TABLE 6. Results of the exergy analysis in the liquefaction unit.

Unit exergy losses Rational efficiency

(in kW)
Five stages compressor (CI) 1965 0.67
Steam turbine 1021 0.86
Compressor 2 (C2) 205 0.70
Compressor 3 (C3) 141 0.69
Capacity mismatch 344
Hot expansion turbine (HET) 241 0.77
Cold expansion turbine (CET) 428 0.61
Heat exchanger I (HE I) 256 0.72
Heat exchanger 2 (HE 2) 173 0.90
Mixers 5
Throttling valve (TV) 74
Total 4853 0.25

For the compression of the nitrogen to 30 bar a five stage compressor with intercooling
(Cl) is used. Steam turbine 2 (ST2), which delivers the required 6.0 MW for the
compressor, uses 20 ton steam a hour of 420 DC and 40 bar. A capacity mismatch
between compressor 3 (C3) , which compresses to 46 bar, and the hot expansion turbine
(RET) results in a exergy loss of 344 kW. The relative low efficiency of the turbines is
caused by the fact that the expansion of the flows takes place below the environmental
temperature. The temperature difference between the isentropic temperature and the real
outlet temperature leads to a great exergy loss due to the low temperature. Especially, in
the case of the cold expansion turbine (CET), where the outlet temperature is -177 DC.
206

The throttling process in the valve (TV) causes a temperature decrease of 3.4 K. The
low exergy loss in the valve is achieved because the flow enters the valve in the liquid
phase and leaves the valve still almost completely liquid.

5.5. ARGON PURIFYING UNIT

The exergy loss in the heat exchanger (H.E.) and compressor is 1.2 kW and 2.9 kW,
respectively. The deoxygenation of the stream with hydrogen with an exergy of 65 kW
causes an irreversibility of 27 kW. The exergy loss taking place due to the cooling of
argon flow from 950°C to 15°C leads to an exergy loss of 43 kW. The heat is not used
somewhere else. The exergy loss in the pure argon distillation column is 10 kW. A
purge stream of 0.8 kW leaves the column at the top. The rational efficiency of this unit
is 0.017 according to (19). No rational efficiencies are calculated for the separate
components due to the small exergy losses of these components.

5.6. TOTAL

The overall exergy loss is 8810 kW as can be seen in Table 7. The total exergy of the
products is 3514 kW, resulting in an overall rational efficiency of 0.28 according to
(25). The efficiency is so high due to the great physical exergy of the products of 2765
kW, mainly the physical exergy of the liquid products, which is 2602 kW. If only the
chemical exergy of the products is seen as the desired output, the rational efficiency
becomes 0.071 by using (26).

TABLE 7. Results of the exergy analysis .

Unit . exer~xlosses(ink\V) ratioTl~l . efiiciencx .....

A;rcompressorandciean;;;!i 2751 0.48
Main heat exchanger 333 0.86
Distillation unit 788 0.52
Liquefaction unit 4853 0.25
Argon purifying unit 85 0.02
Total 8810 0.28

6. Results

More than half of the exergy loss is taking place in the liquefaction unit, while almost
one third is lost in the air compression unit. Minor exergy losses are taking place in the
distillation unit and heat exchangers. The greatest exergy losses are caused by the
compressors and to a lesser extent by the turbines. The relatively low rational efficiency
of the turbines operating in the cryogenic region is striking, while the efficiency of the
heat exchangers is very high.
 207

7. Improvements

By using the excess work in the case of the capacity mismatch 344 kW can be saved.
The increase of the polytropic efficiency of the air compressor (AC) from 0.70 to 0.85
will decrease the power use by 880 kW. This will give an associated exergy saving in
steam turbine I (STl) of 139 kW. The increase of the polytropic efficiency of the
nitrogen compressor (CI) from 0.75 to 0.85 leads to a decrease of the power use by 759
kW. It gives also an associated exergy saving of 108 kW in steam turbine 2 (ST2). The
rational efficiencies of both compressors (AC and CI) becomes 0.77. The increase of
the polytropic efficiency of 0.83 of the cold expansion turbine (CET) to the same value
as the hot expansion turbine (HET), will give an exergy saving of 75 kW and results in
a rational efficiency of 0.66 for the cold expansion turbine.
The rational efficiency of the heat exchangers is high to very high. So the
possibilities for exergy saving are very limited. Only the exergy loss in heat exchanger I
(HE I) of the liquefaction unit could be halved be improving the heat integration. This
would lead to an exergy saving of about 125 kW.

8. Conclusions for the Cryogenic Air Separation

The exergy analysis of the cryogenic air distillation plant has pinpointed and quantified
the exergy loss in the different plant sections. The biggest exergy loss is caused by the
compressors. This exergy loss can be reduced by almost one half by using better
compressors. Including other suggested improvements the exergy loss can be reduced
by 25%. However, the cryogenic air separation unit itself is well designed from an
exergetic point of view and for a further improvement of the rational efficiency of air
separation alternative processes have to be used or developed.

References
1. Szargut, J., Morris, D.R. and Steward, F.R. (1988) Exergy Analysis 0/ Thermal, Chemical, and
Metallurgical Processes, Hemispere Publishing Corporation.
2. Hedman, B. (1981), Application o/the Second Law o/Thermodynamics to Industrial Processes, Ph.D.
thesis, Drexel University,
3. Oriffioen, A. (1996), Exergy Analysis of the Blast Furnace, Corex and CCF process, Master thesis,
University of Twente.
4. Mozes, E, Cornelissen, R.L., HiTS, 0.0. and Boom, R.M. (1997) Exergy Analysis of the Conventional
Washing Machine. Proc. 0/ Flowers '97, (Manfrida, Eds), Florence, 931-937.
5. Hinderink, A.P., Kerkhof, PJ.M., Lie, A.B.K. (1996) De Swaan Arons, J. and Van der Kooi, HJ.,
Exergy Analysis with a F10wsheeting Simulator-I. Theory; Calculating Exergies of Material Streams,
Chemical Engineering Science, Vol. 51, No 20, 4693-4700.
6. Kotas, T.J. (1995) The Exergy Method of Thermal Plant Analysis, 2-nd edition, Krieger Publishing
Company, USA.
208

9; Nomenclature

E,& = exergy, molar exergy i-th item

GAN gaseous nitrogen in incoming
GOX gaseous oxygen JT Joule-Thomson valve
LAR liquid argon LC lower distillation column
LIN liquid nitrogen LIQ liquefaction unit
I irreversibility min minimum
p pressure mix mixture
R universal gas constant out outgoing
T temperature PA = pure argon distillation column
Ijf = rational efficiency PC = purge cold
PH = purge hot
Subscripts S STEAM
Scf spray cooler
AC air compressor ST steam turbine
ACC air compressing and cleaning unit TN top nitrogen
CET cold expansion turbine UC upper distillation column
cold cold stream YOU vacuum distillation unit
CRAR crude argon
cold cold stream Superscripts
cond condenser
DIST distillation section 0 = environmental state, chemical
HE = heat exchanger component
HEN heat exchanger network ph = physical component
HET hot expansion turbine Q heat
hot hot stream L1T thermal component
fr fractions rate with respect to time
hot = hot stream
EXERGETIC LIFE CYCLE ANALYSIS OF COMPONENTS IN A
SYSTEM

R.L. CORNELISSEN and G.G. HIRS

University of Twente, Dept. of Mechanical Eng.,
Chair of Energy Technology
P.O. Box 217, 7500 AE Enschede, The Netherlands

1. Introduction

Exergy analysis and life cycle analysis have been developed separately.
Exergy analysis has been described extensively in the books of Kotas [1] and
Szargut [2]. Life cycle analysis (LCA) has been described by Consoli et al.
[3] and Heijungs et al. [4]. The latter one gives a detailed methodology to use
in a LCA. The methodology in the LCA includes the effects of all the phases
of the production, use and recycling on the environment. In this paper the
methodology has been performed using only one criterion which is the
minimisation of the life cycle irreversibility associated with the delivery of
domestic hot water. The complete LCA involves other factors e.g. pollution of
air and water, noise, etc., which were not considered here. The concept of
cumulative exergy, introduced by Szargut, uses the method of accumulation of
the exergy consumption to a defined point in the life cycle analysis (Szargut et
al. [2]. The cumulative exergy consumption of a product takes into account all
the exergy destruction for the manufacture of the product. However, in this
method the exergy destruction associated with the disposal of the product and
the influence of recycling which cause changes in the exergy destruction are
not taken into account.
The ELCA method is described by Cornelissen [5]. A comparison between
the LCA and the ELCA is performed for the porcelain mug and the disposable
polystyrene cup. It is shown that the life cycle irreversibility, the exergy loss
during the life cycle, is the parameter for the depletion of natural resources.
Furthermore, the general results of the ELCA and the LCA were similar. More
research has to be performed to show in which cases this takes place. It is
expected that in energy intensive processes or components operating in these
processes the results of the ELCA and the LCA are similar [6J. However, the
ELCA is less time consuming and gives more insight in the location of the
losses. On basis of this knowledge better improvements can be suggested.
Because of its widespread use the heat exchanger has been selected as an
example. Bejan [7] studied extensively the optimisation of a heat exchanger,
excluding exergy destruction associated with use of materials and cumulative
exergy of generation of heat and power. His approach uses the concept of
209
A. Bejan and E. Mamut (eels.), Thermodynamic Optimization o/Complex Energy Systems, 209-219.
© 1999 Kluwer Academic Publishers.
210

entropy generation minimisation. An extension to his approach to include

material use has been made by Aceves-Saborio et a1. [8]. They took into
account the cumulative exergy of the material, but did not include the
irreversibility due to the pressure drops.
Tondeur and Kvaalen [9] have shown that in the case of heat exchangers or
separation devices involving a given heat transfer and achieving a specified
transfer duty, the total entropy produced is minimal when the local rate of
entropy production is uniformly distributed along space variables and time.
In this article the optimal design of a component, a heat exchanger, has
been obtained on basis of the concepts of ELCA. This means the minimisation
of the life cycle irreversibility of the heat exchanger.

Optimisation of a Heat Exchanger

1.1. THE HEAT EXCHANGER

The heat exchanger analysed is a balanced counter flow heat exchanger, which
is used in a district heating system to heat the domestic tap water. The inner
tube carries the cold stream and the surrounding outer tube carries the hot
stream. An equal mass flow of the hot and cold water has been assumed. The
inner and outer tube have been constructed from copper and steel,
respectively. The combined annular heat exchanger is helically wound as
shown in Figure 1. The influence of winding on the pressure drop has been
neglected. The thermal insulation of the heat exchanger has been assumed to
be perfect.

Figure 1. The analysed heat exchanger.

 211

1.1.1. Theory
The following formula can be derived for the irreversibility in the heat
exchanger due to the stream to stream heat transfer and pressure drops:

(1)
with

/ . LIT = T 0 [ mc
.
p
1n--+mc
Tl,out . 1 T2,out]
p n-- (la)
Tl,in T2,in
and
'M> m m
/ =-(PI,in - PI,out)+-(P2,in- P2,out) (1 b)
p p
The heat exchanger effectiveness is:
e TUn - Tl,out T2,in - T2,out
(2)
T1.in-T2,in Tl,in-T2';n

»
In a nearly ideal heat exchanger limit N tu 1 and therefore l-e ~ N;;1 , where
N tu = aA /(mc p ) . Neglecting the heat resistance of the tube wall we have
according to Bejan [7]

rm c 1
l
'AT
2 2 T2
P m2 c P2 T 2 1 . [2
T T2
+ =To·mc - - + - -
/ =To
al A a2 A J P Ntu,1 Ntu ,2
(3)

iT 2,in - TI,ini
with T
~TI,inT2,in .

By using the force balance inside the tube(s) we can write for the pressure
drops:
L
!1 PI = l'1,in -l'1,out =211 (Re) . P . ui 2 DI (4a)

(4b) in (lb) yields

 212

(5)

In the turbulent flow region in tubes and annular spaces with a limited
temperature difference of 5 K for liquids between the bulk fluid and pipe
surface temperature we have according to Chapman [10]
aDh n 08
Pu-Dh
NU=T=0.023Re· Pr =0.023· ( - P -
)08. .(TCpP )n (6)

with n= 0.3 or 0.4 for cooling or heating, respectively. Experimental data of

Kays and London [11] gives a similar relation.

The friction factor in tubes according to the friction law of Blasius is given in

( j
Rogers and Mayhew [12] as
O.25
0.0791 P
f(Re) =--=0.0791· - - (7)
Re 0.25 -D
Pu h

Substituting (6) and (7) in (5) and using A ~ 1t·L·D\ we obtain

T 14 .3 m·1.2 [C1.6
p T 2 r,,0.4 D 10.8 c1.7
p T2 ,,0.5 (D 2 + D 1 + 2d 1 )0.8 Dh,2 1
/. t'J.T
°
r
= 7r 0.2 ).,0.6 L + ).,0.7 L· D1
(8a)

(8b)

Cumulative losses due to heat and power generation. Irreversibilities in heat

exchangers are associated with external exergy destruction, for example, to
overcome the frictional losses in the heat exchanger a pressure difference is
needed. The generation of heat and power is associated with exergy
 213

destruction, which has to be taken into account in the optimisation, according

to the LCA. We assume the following situation, based on the district heating
system in Enschede, The Netherlands. In this system a cogeneration heat and
power plant, a steam and gas turbine plant (STAG), is used, which has an
exergetic efficiency of 50%, when there is no useful heat production. The
exergetic efficiency of the extraction type of plant stays nearly constant when
a part of the steam is used for the district heating system. Irreversibilities
associated with the building of the combined heat and power plant and the
transport of the fuel, natural gas, are neglected.
The exergetic efficiency of the heat transport to the houses is estimated to
be 0.5 for a widespread distribution net l (wide net) and 0.75 for a very dense
distribution net l (dense net). The exergy destruction in the wide net is greater
because of more power needed to overcome the frictional pressure drops and
more heat transfer to the environment. A great part of the exergy destruction
in the heat transport takes place because of the temperature difference in the
heat exchanger between the main transport tube and the local distribution net.
It has been assumed that the exergy destruction associated with the heat
transport is independent of the district heating water temperature. The
exergetic efficiency of the pumps is assumed to be 0.7.
Hence the exergetic cost for the analysed heat exchanger of pressure rise,
k p , and heat, kT' can be calculated to be 2.85 and 4 for the wide net and 2.85
and 2.66 for the dense net, respectively. The exergetic cost of a product is the
amount of exergy which is needed for the production of one exergy unit of the
product.
For the exergy destruction associated with the operation of the heat
exchangers we have
(9)

Irreversibilities associated with the use of the material. The life cycle flow diagram of
the heat exchanger is displayed in Figure 2. The exergy destruction in each process is
shown in Table 1.

TABLE I. Exergy destruction associated with the

production of material and manufacturing of tubes.

Process Is ( MJ/kg) Ieu ( MJ/kg)

Primary process 10.5 60
Secondary process 4.4 20
Manufacturing process 5.7 15

The data listed in Table 1 is obtained from Cornelissen [5].

I This is a hypothetical situation. In reality the exergy losses are higher because the peak in heat demand is
supplied by auxiliary heating boilers using natural gas. Their exergetic efficiency is very poor.
214

The heat exchanger is located in an insulated box made of polyurethane foam

(PUR) with a thickness of 0.10 m. The cumulative energy consumption for the
production of PUR with a density of 30 kg/m 3 is 98 MJ/kg according to
Kindler et al. [13]. The exergy content is estimated to be 27 MJ/kg on basis of
the lower heating value according to Kindler et al. [13]. So the cumulative
exergy destruction is 71 MJ/kg (C PUR ). The box containing the heat exchanger
has two outer sides of 0.70 meter. The height of the box is calculated from the
length of the heat exchanger. The heat exchanger has been helical wound in
three coaxial tubular layers. The mean diameter of the tube windings is 0.45
meter. No recycling of the PUR has been assumed.

Natural resources

Primary material

Heat exchanger

Disposed material
S"OOd'~Pro""J I

s"ood'Tm~
Figure 2. Life cycle of the heat exchanger.

The exergy destruction associated with the manufacture of the heat exchangers is due to
the production of copper tube and steel tube, welding and the production of the PUR-
foam box.

. M cuCcu + M Fe Cs + LCw,S + M PURCpUR

Iman = t

~ [(~ + O.5dl)dlPCu(Cman,Cu + XCuCsec,Cu + (1- xcu)Cpri,cu)]

 215

(10)

in which t is the operating time of the heat exchanger during its life cycle and
x is the recycling ratio which is the proportion of secondary material, i.e.
material which is recycled.

2. Results

From the above considerations we obtain an expression for the total life cycle
irreversibility, which has to be minimised.

j LC = j oper + j man (11 )

Where joper and jman are stated in (9) and (10), respectively.

The following operating parameters have been assumed for the heat
exchanger. The incoming temperature of the cold domestic tap water is 15°C.
The domestic tap water is heated to 65°C. The temperature of the incoming
district heating water is variable. The environmental temperature, To, is 25°C.
The operating time for the heat exchanger is 30 minutes a day on full load for
10 years. The mass flow of the district heating water as the domestic water is
0.1 kg/so The mean temperature of the inlet and outlet streams is used for the
heat capacity, viscosity and thermal conductivity of water. The wall thickness
of the inner and outer tube are 0.8 and 2 mm, respectively. The recycling ratio
is set to be 0.9 for the copper and steel parts of the tube.

2.1. REFERENCE CONFIGURA nON

As a reference situation a domestic water heat exchanger is taken with the

hypothetical fixed length of 30 meter. The optimised inner and outer tube
diameter are 8.56'10- 3 m and 1.29'10- 2 m for the wide net and 9_04'10- 3 m and
l.36·10- 2 m for the dense net. The life cycle irreversibility of the heat
exchanger in full operation is 1.82'10 3 W for the wide net and l.34·10 3 W for
the dense net. Resulting from a t.T, t.P 1 and t.P 2 of 5.24 K, 1.34 bar and 7.57
bar for the wide net and 5.57 K, l.03 and 5.48 bar for the dense net,
respectively. Where t. T is the temperature difference between the hot and cold
stream, which is constant, because we have a balanced counter flow heat

2The value of the viscosity is strongly temperature dependent. The viscosity at 15°C, 70 °C and 80°C is
1.14'10-3 Pa's, 0.406'10-3 Pa's and 0.357.10-3 Pa's, respectively. So the assumption of the mean temperature
will cause a deviation from the real situation.
216

exchanger. T'.out, which can be calculated by T 2. in - ~T, is set to 65°C by use

of iterative calculations.

2.2. RESULTS OF OPTIMISATION

The minimisation of j LC for the 3 variables, D" D2 and L gives the minimum
value of life cycle irreversibility per heat exchanger in the wide net of the
district heating system of 1.06'10 3 W. The optimum geometrical parameters
were found to be D, = 1.07'10. 2 m, D2 = 1.69'10. 2 m and L = 105.8 m. The ~T,
~P, and ~P2 of this optimised heat exchanger are 1.95 K, 1.62 bar and 3.65
bar, respectively.

The dense net life cycle irreversibility is 8.76,10 2 W for D, = 1.07'10.2 m, D2

= 1.69'10. 2 m and L = 87.0 m. The ~T, ~P, and ~P2 of the optimised heat
exchanger are 2.93 K, 1.34 bar and 3.00 bar, respectively.

We see that for optimal geometrical parameters the tube diameters are
independent of the efficiency of the distribution net, the dense or wide net.
However, the length of the tube is strongly dependent on the type of
distribution net. The components of the life cycle irreversibility in the heat
exchanger are displayed in Table 2.

TABLE 2. Components of life cycle irreversibility in Watts.

Component Wide net dense net Wide net dense net
(L=30m) (L=30m)
Thermal 1451 1033 514 422
Mechanical 255 186 151 124
Manufacture 114 118 395 330
Total 1820 1337 1060 876

The contribution of the use of copper, steel and PUR-foam to the

irreversibility associated with the manufacture is 39%, 42% and 19% for the
wide and dense net optimisation and 32%, 33% and 35% for the both heat
exchangers with l:l length of 30 meter, respectively.

2.3. DISCUSSION

The effect of the tube diameters and the length on life cycle irreversibility for
the optimal geometrical parameters is shown in Figure 3. The cross sections of
flow areas of the inner and outer passages are set fixed in the ratio of 1: 0.863
to get a 3 dimensional figure. This ratio has been obtained for the optimisation
of the wide net. In Figure 3 can be seen that the life cycle irreversibility rate
rises for smaller tube length and inner tube diameter. At zero length or at zero
tube diameter the life cycle irreversibility rate becomes of course infinite. The
effects of the diameters and the length on j LC is greatest at their smallest
 217

values. The optimal inner tube diameter varies from around 0.8'10- 2 m for
smaller lengths to 1.1'10- 2 for longer lengths of the tube.

L
(in meters)

0.0) 0.0)5 0.02

z
D) (in meters)

Figure 3. Life cycle irreversibility (in Watts) against length

and diameter of the heat exchanger for the wide net

The analysis carried out in this work is only applicable to the situation when
both the heat and the power are supplied from a cogeneration plant. If the heat
is provided by a source which produces only heat from combustion of fossil
fuel the exergy saved in the heat exchanger will be lost in the heater. There is
no similar trade-off between exergy saving during operation and exergy use
during construction of the heat exchanger as in the case when heat and power
are generated separately. The saving in exergy resulting from the use of a
longer heat exchanger requiring lower heating water temperature can not in
this case be utilised elsewhere e.g. for generating more electricity.
It is possible that a heat exchanger with more than one heated water tube,
arranged in parallel inside the larger outer tube carrying the heating, may lead
to a more efficient heat exchanger process. However, this matter must be left
for a further investigation.
The Reynolds numbers of the inner and outer flow are 19.8'103 and
7.15'10 3 for the optimal configuration of the heat exchanger, respectively,
when the mean temperature of the inlet and outlet streams is taken for the
viscosity of the water. For the fixed length optimisation the Reynolds numbers
are between 7.72'10 3 and 24.4'10 3 for the inner and outer flow, so the
condition for the turbulent flow region is fulfilled.
Heat transfer inside the heat exchanger box, which leads to exergy
destruction, is neglected. If this is taken into account an insulation directly
wound around the tubes would probably be more effective than the assumed
218

box. However, if this is the case the analysis shown would not changed
drastically, because the exergy use during construction would still increase
when the tubes become longer.

3. Conclusions

With the Exergetic Life Cycle Analysis, the combination of exergy analysis
and life cycle analysis, the optimal design of a heat exchanger can be
obtained. For all energy systems where there is a trade-off between exergy
saving during operation and exergy use during construction of the energy
system this method should be adopted to get the true optimum from the point
of view of conservation of exergy reservoirs of natural resources.
In the case under study the optimal design parameters of the heat
exchangers are obtained under the specified conditions. The dense net, which
is a more energy efficient heat supply system than the wide net, has the same
inner tubes diameters as the wide net whilst the length of the heat exchanger is
smaller for the former than for the latter. The dense net has lower life cycle
irreversibility due to the manufacture of the heat exchanger compared to the
wide net, because less exergy is saved by the same increase of exergy use due
to the manufacture. In the optimised situation the life cycle irreversibility is
more uniformly distributed between the component irreversibilities than in the
fixed length of 30 meters situation.
In general we can conclude that the thermodynamic optimisation of the
design parameters of a subsystem is dependent of the thermodynamic
efficiency of the whole system and that the different components of the life
cycle irreversibility of heat exchangers are more uniformly distributed when
there are less restrictions on the design parameters for the optimisation.

4. Nomenclature

A = heat transfer area T2 = outer tube temperature

Cp = heat capacity Ii = mean velocity of the fluid in the tube
C = cumulative exergy destruction x = recycling ratio of the material
D = inner diameter of the tube
d = thickness of the tube wal1 Greek letters
I = irreversibility or exergy destruction
L = length of the tubes a = heat transfer coefficient
M = mass e = effectiveness
m= mass flow p = density
Nu = NusseIt number A= coefficient of thermal conduction
N tu = number of heat transfer units J1. = dynamic viscosity
P = pressure
Pr = Prandtl number Subscripts
Re = Reynolds' number
T = temperature o= environmental
T/ = inner tube temperature 1 = inner tube of the heat exchanger
 219

2 = outer tube of the heat exchanger pri = primary

Cu = copper PUR = polyurethane foam
h = hydraulic sec = secondary
S = steel tot = total
in = inlet w = welding
LC = life cycle
man = manufacturing Superscripts
mat = material
oper = operating AP = mechanical component
out = outlet LIT = thermal component

5. References

I. Kotas, T.J., (1995) The Exergy Method of Thermal Plant Analysis, Krieger Publishing
Company, Melbourne, Florida, USA.
2. Szargut, J., Morris, D. R. and Steward, F.R., (1988) Exergy Analysis of Thermal, Chemical
and Metallurgical Processes, Hemispere Publishing Corporation.
3. Consoli, F., et aI., (1993), Guidelines for Life Cycle Assessment of Products: A 'Code of
Practice', SETAC, Brussels, Belgium.
4. Heijungs, R. et aI., (1992) Environmental Life cycle Analysis of Products, Center for
environmental studies, University of Lei den, The Netherlands.
5. Cornelissen, R.L., (1997) Thermodynamics and Sustainable development, the Use of Exergy
Analysis and Reduction of Irreversibility, Ph.D. thesis, University ofTwente.
6. Cornelissen, R.L., Bourdri, J.C. and Kalf, M.A., (1998) The value of the ELCA above the
LCA, KEMA/University of Twente/CCS, in Dutch, in press.
7. Bejan, A., (1982) Entropy Generation through Heat and Fluid Flo"H/ John Wiley & Sons.
8. Aceves-Saborio, S., Ranasinghe, J. and Reistad, G.M., (1989) ASME Journal of Heat
Transfer 111,29 - 36.
9. Tondeur, D. and Kvaalen, E., (1987) Industrial & Engineering Chemistry Research 26,
50- 56.
10. Chapman, A.J., (1974) Heat Transfer, Macmillan, New York.
11. Kays, W.M. and London, A.L., Compact Heat Exchangers, third edition, McGraw-Hili
Book Company.
12. Rogers, G.F .C. and Mayhew, Y.R., (1967) Engineering Thermodynamics, Work and Heat
Transfer, Longmans, London.
13. Kindler H. and Nikles A., (1980), Energy for thePproduction of Materials - Method of
Calculation and Energy use of Plastics (in German), Kunststoffe 70, 802-807.
THE INTIMATE CONNECTION BETWEEN EXERGY AND
THE ENVIRONMENT

I. DINCER 1 and M.A. ROSEN 2

J Department of Mechanical Engineering, KFUPM, Dhahran 31261,
Saudi Arabia.
2 Department of Mechanical Engineering, Ryerson Polytechnic
University, Toronto, Ontario M5B 2K3, Canada.

1. Introduction

It is well known that there is always an environmental cost associated with the thennal,
chemical, and/or nuclear emissions which are a necessary consequence of carrying out the
processes that give benefits to mankind. The environmental impact of emissions is
reduced by increasing the efficiency of resource utilization. Sometimes, in practice, this is
referred to as "energy conservation." However, increasing efficiency generally entails
greater use of materials, labor, and more complex devices. The additional cost may be
justified by the added security associated with a decreased dependence on energy
resources and by the social peace obtained through increased productive employment.
In 1970s, there was a primary focus on the relationship between energy and
economics. At that time, the linkage between energy and the environment did not receive
as much attention. As environmental concerns, such as acid rain, ozone depletion and
global climate change, became major issues in the 1980s, interest in the link between
energy utilization and the environment became more pronounced (especially in the late
1980s and early 1990s). Since then, there has been increasing attention on this connection.
Despite many studies concerning the close relationship between energy and the
environment, there have been limited works on the link between the exergy and
environment concepts [e.g., 1-5].
Many suggest that the impact of energy resource utilization on the environment is
best addressed by considering exergy. The exergy of a quantity of energy or a substance is
a measure of its usefulness, quality or potential to cause change. Exergy appears to be an
effective measure of the potential of a substance to impact the environment. In practice,
the authors feel that a thorough understanding of what exergy is, and how it provides
insights into the efficiency and perfonnance of energy systems, are required for the
engineer or scientist working in the area of energy systems and the environment. Based on
Bejan [6] approach combining thennodynamics, heat transfer and fluid mechanics as a
distinct discipline, entropy generation, here, we make a similar approach for exergy,
because of its interdisciplinary character as the confluence of energy, environment and
sustainable development (Fig. 1).
221
A. Bejan and E. Mamut (eds.), Thermodynamic Optimization o/Complex Energy Systems, 221-230.
© 1999 Kluwer Academic Publishers.
 222

policies increasingly play an important

central role in relating to a broad range of
local, regional and global environmental
concerns. Given the complexity of these
problems, there has been a growing need
to understand the link between exergy
SUSTAINABLE utilization and environmental impacts.
DEVELOPMENT The main objective of this paper is to
Figure 1. The interdisciplinary triangle covered
by the field of exergy analysis.
present such a study, and to provide useful
insights and direction to those involved in
Energy production, transfonnation, energy and environment studies for
transport and use have important impacts analyzing and solving environmental
on the earth's environment. Energy problems using the exergy concept.

2. Environmental Problems

Environmental problems span a continuously growing range of pollutants, hazards and

eco-system degradation factors that affect areas ranging from local through regional to
global. Some of these problems may arise from observable, chronic effects on, for
instance, human health, while others may stem from the perceived risk of a possible
accidental release of hazardous materials. A significant number of these environmental
issues are caused by or relate to the production, transfonnation and end-use of energy. For
example, eleven major areas of environmental concern in which energy plays a significant
role have been identified [7, 8] and include major environmental accidents, water
pollution, maritime pollution, land use and siting impact, radiation and radioactivity, solid
waste disposal, hazardous air pollutants, ambient air quality, acid deposition, stratospheric
ozone depletion, and global climate change.
The interface between energy and the environment is complex and constantly
evolving. Generally, the ability of science to identify and quantify the production and
effects of potentially harmful substances has greatly advanced. Throughout the 1970s
most environmental analyses and legal instruments of control focused on conventional
pollutants (i.e., SOx, NOx, CO, and particulates). Recently, concern has been extended
to the control of (i) hazardous air pollutants, which are usually toxic chemical
substances that are harmful in small doses and (ii) globally significant pollutants such as
CO 2 • Developments in industrial processes and structures have led to new
environmental problems. For instance, in the energy sector, major increases in the
transport of industrial goods and people by car has led to an increase in road traffic
which has led to an increase in the attention paid to the effects and sources of NOx and
volatile organic compound (VOC) emissions. Consequently, while concern was given to
energy policy and economic considerations in the 1970s and 1980s, environmental
control and energy efficiency have received increasing attention over the last decade.
An important aspect of environmental impact is that the aesthetics of man and his
machines on the planet. The low cost of power from fossil fuels has made man
increasingly dependent on them, allowed significant pollution increases, and thus
engendered a reduction on the planet's ecological diversity. The topic of planet's
 223

evolution is a subject for researchers in this field, and they can play a vital role in guiding
industrial society toward the use of exergy as an efficient tool to reduce the energy
consumption.

3. Energy Use and Environmental Impact

Energy resources are required to supply the basic human needs of food, water and shelter,
and for improving the quality of life. It was indicated by the United Nations [9] that it is
crucial to consider the energy sector in any broad atmosphere strategy. The need was
identified for two major areas of activity: programs to increase energy efficiency and
programs to encourage the transition to environmentally sound energy systems. The major
program areas accepted for consideration promote: (i) the energy transition; (ii) increased
energy efficiency and consequently, increased exergy efficiency; (iii) renewable energy
sources; and (iv) sustainable transportation systems. Additionally, it was reported that (i) a
major energy efficiency program would provide an important means of reducing CO2
emissions, and (ii) these activities should be accompanied by strong new measures to
reduce the fossil fuel components of the energy mix and develop alternative energy
sources. These ideas have been reflected in many recent energy-related studies, which
have concentrated on the provision of the energy service needed for an activity, with the
lowest environmental impact and cost and with the maximum of energy security possible.

4. Thermodynamics and the Environment

As indicated by several researchers [4, 10, 11], the second law of thermodynamics is
instrumental in providing insights into environmental impact. It is well known that heat
transfer from industrial devices to the environment is of major concern to engineers and
scientists working on energy and environment issues. Irresponsible management of waste
energy can significantly increase the temperature of portions of the environment, resulting
in thermal pollution. If not carefully controlled, thermal pollution can seriously disrupt
marine life in lakes and rivers. However, by careful design and management, the waste
energy dumped into large bodies of water can sometimes be used to improve significantly
the quality of marine life by keeping the local temperature increases within safe and
desirable levels.

Mankind has long been intrigued by the

implications of the laws of thermodynamics
on the environment. One myth speaks of
Ouroboros, a serpent-like creature which
survived and regenerated itself by eating only
its own tail (Fig. 2). By neither taking from
nor adding to its environment, this creature
was said to be completely environmentally
benign and self-sufficient.
Figure 2. The mythological creature
Ouroboros, which survived and regenerated
itself by eating only its own tail [20].
224

It is useful to examine this creature in light of the thennodynamic principles

recognized today. Ouroboros' existence would not have violated the conservation law for
mass. However, it would have violated the frrst law of thennodynamics (which states
energy is conserved), since Ouroboros would have had to expend energy in eating, and
therefore would have required an external energy input to compensate for this
expenditure. This is true, even if it is assumed that Ouroboros was an isolated system (i.e.,
it received no energy from the sun, and emitted no energy during any process). More
importantly, Ouroboros' existence would have violated the second law (which states that
exergy is reduced for all real processes), since Ouroboros would have had to obtain exergy
to regenerate the tail it ate into an equally ordered part of its body (or it would ultimately
have dissipated itself to an unordered lump of mass), and therefore would have required
an external exergy input. Thus, Ouroboros would have to have had an impact on its
environment. Besides showing that within the limits imposed by the laws of
thennodynamics, all real processes must have some impact on the environment, this
example is intended to illustrate the following key point: the second law is instrumental in
providing insights into environmental impact. It is of course true that the principles,
demonstrated here through a historical example, are relevant today. For example,
technologies are still striven for today that have the characteristics of Ouroboros (i.e., that
are environmentally benign and self-sufficient). In fact, researchers at the University of
Minnesota have built an "energy-conserving" house, called Ouroboros [12].

5. Exergy and the Environment

Energy and environment studies which lead to increased energy efficiency can reduce
environmental impact by reducing energy losses. Within the scope of exergy methods,
such activities lead to increased exergy efficiency. In practice, the efficiency
improvements can be identified by means of modeling and computer simulation studies.
Increased efficiency can often contribute in a major way to achieving energy security in an
environmentally acceptable way by the direct reduction of emissions that might otherwise
have occurred. Increased efficiency also reduces the requirement for new facilities for the
production, transportation, transfonnation and distribution of the various energy fonns;
these additional facilities all carry some environmental impacts. To control environmental
pollution, efficiency improvement actions often need to be supported by pollution control
technologies or fuel substitution. It is through regional or national actions, rather than
through individual projects, that improved exergy efficiency can have a major impact on
environmental protection.
Exergy is defmed as the maximum amount of work which can be produced by a
stream of matter, heat or work as it comes to equilibrium with a reference environment.
Exergy is a measure of the potential of a stream to cause change, as a consequence of not
being completely stable relative to the reference environment. For exergy analysis, the
state of the reference environment, or the reference state, must be specified completely.
This is commonly done by specifying the temperature, pressure and chemical composition
of the reference environment. Exergy is not subject to a conservation law. Rather exergy is
consumed or destroyed, due to irreversibilities in any process.
 225

Exergy analysis is a method that uses the conservation of mass and conservation of
energy principles together with the second law of thermodynamics for the design and
analysis of energy systems. The exergy method can be suitable for furthering the goal of
more efficient energy-resource use, for it enables the locations, types, and true magnitudes
of wastes and losses to be determined. Therefore, exergy analysis can reveal whether or
not and by how much it is possible to design more efficient energy systems by reducing
the inefficiencies in existing systems.
Most recently, Cornelissen [13] has conducted an excellent study on the
thermodynamics and sustainable development and pointed out that one of the keystones
for obtaining sustainable development is the use of exergy analysis. Energy can never be
lost as stated in the first law of thermodynamics. Exergy can be lost and this loss, called
irreversibility, created during the use of non-renewables has to be minimized to obtain
sustainable development. He also showed that all environmental effects associated with
emissions and the environmental effect of depletion can be expressed in terms of one
indicator, which is based on physical principles.
Several researchers have suggested that the best way to link the second law and
environmental impact is through exergy because it is a measure of the departure of the
state ofa system from that of the environment [e.g., 2, 11]. The magnitude of the exergy
of a system depends on the states of both the system and the environment. This departure
is zero only when the system is in equilibrium with its environment. The concept of the
environment as it applies to exergy analysis is discussed in detail elsewhere [1, 3, 14].
Detailed discussions of exergy analysis for many processes and systems are also given
elsewhere [14-17]. Tribus and McIrvine [IS] suggest that performing exergy analyses of
the natural processes occurring on the earth could form a foundation for ecologically
sound planning because it would indicate the disturbance caused by large-scale changes.
An understanding of the relations between exergy and the environment may reveal
the underlying fundamental patterns and forces affecting changes in the environment, and
help researchers to deal better with environmental damage. Three relationships between
exergy and environmental impact, which have been introduced previously [4, 5], are
discussed in the following subsections.

5.1. ORDER DESTRUCTION-CHAOS CREATION

The destruction of order, or the creation of chaos, is a form of environmental damage.

Fundamentally, entropy is a measure of chaos. A system of high entropy is more chaotic
or disordered than one of low entropy. For example, a field with papers scattered about
has higher entropy than the field with the papers neatly piled. Conversely, negentropy and
exergy are measures of order. Relative to the same environment, the exergy of an ordered
system is greater than that of a chaotic one. The difference between the exergy values of
the two systems containing papers described above is a measure of the minimum work
required to convert the chaotic system to the ordered one (i.e., in collecting the scattered
papers). In reality, more than this minimum work, which only applies if a reversible
clean-up process is employed, is required. The exergy destroyed when the wind scatters a
stack of papers is a measure of the order destroyed during the process. The ideas relating
exergy and order in the environment become more abstract ifhuman values are considered
[10]. People are distressed when they see a landscape polluted with papers chaotically
226

scattered about, while they value the order of a clean field with the papers neatly piled at
the side. Perhaps human values are related to exergy and order.

5.2. RESOURCE DEGRADATION

The degradation of resources found in nature is a form of environmental damage. Kestin

[19] defines a resource as a material, found in nature or created artificially, which is in a
state of disequilibrium with the environment. Resources have exergy as a consequence of
this disequilibrium. For some resources (e.g., metal ores), it is their composition that is
valued. Many processes exist to increase the value of such resources by purifying them
(Le., by increasing their exergy). This is done at the expense of consuming at least an
equivalent amount of exergy elsewhere (e.g., burning coal to produce process heat for
metal ore refming). For other resources (e.g., fuels), it is normally their reactivity that is
valued (i.e., their potential to cause change, or "drive" a task or process). By preserving
exergy through increased efficiency (i.e., using as little exergy as necessary for a process),
environmental damage is reduced. Increased efficiency also has the effect of reducing
exergy emissions which, as discussed in the next section, also playa role in environmental
damage. The earth is an open system subject to a net influx of exergy from the sun. It is
the exergy (or order states) delivered with solar radiation that is valued; all the energy
received from the sun is ultimately radiated out to the universe. Environmental damage
can be reduced by taking advantage of the openness of the earth and utilizing solar
radiation (instead of degrading resources found in nature to supply exergy demands). This
would not be possible if the earth was a closed system, for it would eventually become
more and more degraded, or "entropic."

5.3. WASTE EXERGY EMISSIONS

The exergy associated with process wastes emitted to the environment can be viewed as a
potential for environmental damage. Typical process wastes have exergy, a potential to
cause change, as a consequence of not being in stable equilibrium with the environment.
When emitted to the environment, this exergy represents a potential to change the
environment. In some cases, this exergy may cause a change perceived to be beneficial
(e.g., the increased rate of growth of fish and plants near the cooling-water outlets from
thermal power plants). More often, however, emitted exergy causes a change which is
damaging to the environment (e.g., the deaths of fish and plants in some lakes due to the
release of specific substances in stack gases as they react and come to equilibrium with the
environment). Emissions of exergy to the environment can also interfere with the net input
of exergy via solar radiation to the earth. The carbon dioxide emitted in stack gases from
many processes changes the atmospheric CO2 content, affecting the receiving and
re-radiating of solar radiation by the earth. The relation between waste exergy emissions
and environmental damage has been recognized by several researchers. By considering
the economic value of exergy in fuels, Reistad [1] developed an air-pollution rating in
which the air-pollution cost for a fuel was estimated as either the cost to remove the
pollutant, or the cost to society of the pollution (Le., the tax which should be levied if
pollutants are not removed from effluent streams). Reistad claimed the rating was
preferable to the mainly empirical ratings then in use.
 227

Unconstrained Exergy
Constrained Exergy
(a potential to cause a ...... (a potential to cause change
chanlje) Emission of exergy to the environment in the environment)

SYSTEM ENVIRONMENT
Figure 3. The comparison of constrained and unconstrained exergy. Exergy constrained in a system
represents a resource. When emitted to the environment, exergy becomes unconstrained and represents a
driving potential for environmental damage.

The reader may at this point detect a paradox. It was pointed out in the second last
section that exergy in the environment, in the form of resources, is of value; while exergy
in the environment, in the form of emissions, is stated in the last section to have a
"negative" value, in that it has a potential to cause environmental damage. This apparent
paradox can be resolved if the word constrained is included in the discussions (see Fig. 3).
Constrained sources of exergy in the environment are of value. Most resources found in
nature are constrained. Unconstrained emissions of exergy to the environment have
negative value. If emissions to the environment are constrained (e.g., separating out sulfur
from stack gases), then there are two potential benefits: (i) the potential for environmental
damage is blocked from entering the environment, and (ii) the now-constrained emission
potentially becomes a valued commodity, a source of exergy.

Figure 4. Qualitative illustration (in • Order destruction

relative units) of how decreasing the (or chaos creation)
exergy efficiency of a process causes
an increase in the related • Resource
environmental impact associated with degradation
the process, through either order • Waste exergy
destruction-chaos creation, or emissions
resource degradation, or waste exergy
emissions.
Process Exergy Efficiency

5.4. QUALITATIVE ILLUSTRATION

The manner by which decreasing the exergy efficiency of a process causes the related
environmental impact (regardless of which of the three measures discussed in Sections
5.1-5.3 is considered) to increase is illustrated a qualitatively and approximately in Fig. 4.

6. Illustrative Example

The ideas discussed in this paper are demonstrated by the authors for the process of
electricity generation, by considering the coal-frred Nanticoke Generating Station (NGS).
The individual units of NGS each have net outputs of approximately 500 MWe. A
substantial base of data has been obtained for NGS, which has been operating since 1981.
Overall balances of exergy, energy and mass for NGS are illustrated in Fig. 5, where the
rectangle in the center of each diagram represents NGS. The overall system consists of
four main sections [4]:
a) Steam Generators: Eight pulverized-coal-frred natural circulation steam generators
each produce 453.6 kg/s steam at 16.89 MPa and 538°C, and 411.3 kg/s of reheat
228

steam at 4.00 MPa and 538°C. Coal is consumed at full load at a rate of 47.9 kg/so Air
is supplied to the furnace by two 1080 kW, 600-rpm motor-driven forced draft fans.
Regenerative air preheaters are used. The flue gas passes through an electrostatic
precipitator rated at 99.5% collection efficiency, and exits the plant at 120°C through
two multi-flued, 198 m high chimneys.

.. _--------------------:--------------------------------'"
:
Preheating :
J ,

A: ... _m .tDltl'lLHor ... d ,.....Wf G: hoI. .11l pv.mp

B: 1I.I,II-,r_lI,. ,.rbial H: Io_pre•• u h••, txchanlus
C: ial.erm.di&Le--pr_,," '.rbl •• I: op•• d....r:u,i •• beat Ixcbailier
D: I_pr_ur. turbia. J: boiltr t.ed pamp
£: , •••r • • • a. traulorllMr K: lIip..preDIlI" h ••, txch.alliln
F:eoad,_

Figure 5. Breakdown of a single unit in the NGS electrical generating station into four main sections.

b) Turbine Generators and Transformers: The steam produced in the steam generators is
passed through a series of turbine generators and a transformer. The net power output
is 505 MW. Each unit has a 3600-rpm, tandem-compound, impUlse-reaction turbine
generator containing one single-flow high-pressure cylinder, one double-flow
intermediate-pressure cylinder and two double-flow low-pressure cylinders. Steam
exhausted from the high-pressure cylinder is reheated in the combustor. Extraction
steam from several points on the turbines preheats feed water in several low- and
high-pressure heat exchangers and one spray-type open deaerating heat exchanger.
The low-pressure turbines exhaust to the condenser at 5 kPa.
c) Condensers: Cooling water condenses the steam exhausted from the turbines. A flow
rate of cooling water of 18.9 mJ/s is used to achieve a temperature rise of 8.3°C in the
cooling water across the condenser.
d) Preheating Heat Exchangers and Pumps: The temperature and pressure of the
condensed steam are increased to 253°C and 16.89 MPa in a series of pumps and heat
ex.changers.
In this example of a conventional coal-fired electrical generating station, each of the
relationships between exergy and environmental impact described in the last is
demonstrated, and described below:
• Waste exergy emissions from the plant occur with (i) the gaseous wastes exiting
through the stack, (ii) the solid wastes exiting the combustor, (iii) and the waste heat
released both to the lake, which supplies condenser cooling water, and to the
atmosphere. The exergy associated with each of these emissions represents a potential
to impact on the environment. The societial consensus regarding emissions of
harmful chemical constituents in stack gases and "thermal pollution" in local bodies
of water indicates that the potential for impact of these emissions is already generally
recognized, but not from an exergy-related perspective.
 229

• During the electricity generation process, a finite resource, coal, is degraded as it is

used to drive the process. Although a degree of degradation is necessary for any real
process, increased process efficiency can reduce this degradation for the same
services/products. In the extreme, if the process considered in the example were made
thermodynamically ideal, the exergy efficiency would increase from 37% to 100%,
and coal use as well as related emissions would decrease by over 60%. These insights
are provided by the exergy, not energy, analysis results.
• The more abstract concepts of order destruction and chaos creation are also
illustrated. Certainly the degradation of coal fuel to stack gases and solid wastes
represents a destruction of order, especially since less exergy is present in the
products of the process. The emissions of wastes to the environment increases the
chaos in the world by allowing the constrained material and heat products of
combustion to be released without constraints into environment.
The authors feel that this example (i) provides some practical illustrations of the
more abstract concepts discussed throughout this paper, and (ii) highlights the importance
of understanding and considering the relations between exergy, rather than energy, and
environmental impact, when addressing environmental challenges and problems. It is
clear that an enhanced understanding of the environmental problems relating to energy
presents a high-priority need and urgent challenge, both to allow the problems to be
addressed and to ensure the solutions are beneficial for the economy and the energy and
energy systems themselves.

7. Conclusions

This paper discusses the relations between environmental impact and thermodynamics, in
general, and the thermodynamic property exergy, in particular. Historical and modem
examples demonstrate these concepts. Three main relations between exergy and
environmental impact are identified: (i) order destruction-chaos creation, (ii) resource
degradation, and (iii) waste-exergy emissions. An apparent paradox is described between
the usefulness of a high-exergy quantity, and the simultaneous potential for environmental
disruption it possesses. It is concluded that the potential usefulness of exergy analysis in
addressing and solving environmental problems is substantial, but that further work is
required before this potential can be properly and fully tapped.

8. Acknowledgements

The support for this work provided by King Fahd University of Petroleum and Minerals, Ryerson
Polytechnic University and the Natural Sciences and Engineering Research Council of Canada is
gratefully acknowledged.

9. References

I. Reistad, G.M. (1970) Availability: Concepts and ApplicatiOns, Ph.D. Thesis, University of
Wisconsin, Madison.
2. Szargut, J. (1980) International progress in second law analysis, Energy 5,709-718.
230

3. Wepfer, W.1. and GaggioIi, RA. (1980) Reference datums for available energy, in
Thermodynamics: Second Law Analysis, ed. RA. Gaggioli, ACS Symposium Series 122,
Washington, DC, 77-92.
4. Rosen, M.A. (1986) The Development and Application ofa Process Analysis Methodology and
Code Based on Exergy, Cost, Energy and Mass, Ph.D. Thesis, University of Toronto, Toronto.
5. Rosen, M.A. and Dincer, I. (1997) On Exergy and environmental impact, Int. J. Energy
Research 21,643-654.
6. Bejan, A. (1996) Entropy Generation Minimization, CRC Press, Boca Raton, FL.
7. Anon. (1989) Energy and the Environment: Policy Overview, OECD-International Energy
Agency, Paris.
8. Dincer, I. (1998) Energy and environmental impacts: present and future perspectives,
Energy Sources 20(4) (in press).
9. Strong, M.F. (1992) Energy, environment and development, Energy Policy 20, 490-494.
10. Hafele, W. (1981) Energy in a Finite World: A Global Systems Analysis, Ballinger, Toronto.
II. Edgerton, RH. (1982) Available Energy and Environmental Economics, D.C. Heath, Toronto.
12. Markovich, SJ. (1978) Autonomous living in the ouroboros house, in Solar Energy Handbook,
Popular Science, 46-48.
13. Cornelissen, RL. (1997) Thermodyanmics and Sustainable Development, Ph.D. Thesis,
University of Twente, the Netherlands.
14. Kotas, T.J. (1985) The Exergy Method of Thermal Plant Analysis, Butterworths, Toronto.
15. Ahrendts, J. (1980) Reference states, Energy 5,667-668.
16. Moran, M.J. (1989) Availability Analysis: A Guide to Efficient Energy Use, ASME, New York.
17. Moran, M.J. and Sciubba, E. (1994) Exergy Analysis: Principles and Practice, J. Engineering
for Gas Turbines and Power 116, 285-290
18. Tribus, M. and Mclrivne, E.C. (1971) Energy and information, Sci. American 225(3), 179- I 88.
19. Kestin, 1.( I 980) Availability: the concept and associated terminology, Energy 5, 679-692.
20. Warchol, K. (1983) Computer News, Computing Services, University of Toronto 207 (cover
page).
EFFECT OF VARIATION OF ENVIRONMENTAL CONDITIONS
ON EXERGY AND ON POWER CONVERSION

Y.A. GOaUS' and O.E. ATAER2

iAeronautical Engineering Dept.
Middle East Technical University,
06531 Ankara, Turkey

2 Mechanical Engineering Dept.

Gazi University, 06570 Ankara, Turkey

1. Introduction

Most thermal engineering problems are solved at sufficient accuracy by second law
analysis of either steady state or control mass models. However problems with transient
or periodic changing conditions pose also challenges. Changes in environmental
conditions may influence thermal performance of energy conversion systems positively
or negatively [1-4]. Shifting greater load of system to times with favourable
environmental conditions, e.g. in power plants to times with low environment
temperature, can increase daily average thermal efficiency. This lecture concentrates on
the effects of variation of environmental conditions and of demand, utilising recently
obtained results [1,2].

2. Momentary Magnitude of Exergy and Exergy Balance

Change of environmental conditions during the process, which is being optimised, will
effect on one hand the momentary value of exergy, on the other hand the magnitude of
exergy destruction. Definition of exergy concept and Gouy-Stodala theorem base upon
existence of a reservoir (environment) so large that its state is not effected by the
process. However the condition of the environment may change due to other reason Po
(t), To (t) and this necessitates a modified exergy balance. Exergy change of a system
consists of 5 groups of terms; three of them are due to heat, work, matter interactions,
fourth due to exergy destruction effect of entropy generation and the last one (~Eo) due
to change of environmental conditions [3, 5]:

(1)

(2), (3)

where Tb(t) is boundary temperature at which heat is transferred to system,

231
A. Bejan and E. Mamut (eds.), Thermodynamic Optimization o/Complex Energy Systems, 231-240.
@ 1999 Kluwer Academic Publishers.
 232

Each group of terms of Eq. (1), except the last one (AEo), can be expressed in terms of
their values for constant environment conditions (index r) and corrections. E.g.:

(6a)

(6b,c)

(6d)

(6e)

For other terms, similar operations can be made. Comparing the new form of Eq. (1)
with exergy balance with reference to fixed environmental conditions:

(7)

one obtains a general expression for the correction term AEo :

(8)

where the first three terms are for control volume and the last for flow substance:

AE or = [U 0 (t 2) - TorSo (t 2) + Por Vo (t 2)]- [u 0 (t l ) - TorSo (t l ) + Por Vo (t l )] (9a)

12 dS dV
1130r = J(To -Tor)-dt; 120r = J(po -Por)-dt (9b,c)
, II dt dt
12 12
AEf,or = J[mex(t)-min(t)J[ho(t)-Torso(t)}it- J [mex(t)-min(t)lh or -Torsor)dt

(9d)
Special cases for control mass and steady state can be obtained easily from these general
expressions.

3. Effect of Periodic Change of Source Temperature and General

Characteristics of the Method

Environment temperature and source temperature will be assumed to change

harmonically, the later with a phase difference 4> ahead of the former;

To = Tom +Toa sin.; Ts =Tsm +Tsa sin(.+w) (10), (11)

 233

where the independent variable or is dimensionless time: or = 21tft = rot, with t: time, ro:
angular speed and f: frequency, w is phase difference between source and environment
temperature. Fig. 1 shows the schema of the engine and temperature variation.

1.5 r-------------------,

qlt) 1.25

W(t)

5 10 IS 20
t (h)

00 00
Figure I. (a) Schema of a reversible engine running between source temperature T. and environment
temperature To; (b) Temperatures and heat supply q. by time for an example case (Table I. first line).

With reference to average source temperature, T sm , these equations can be written in

dimensionless form introducing x = T 0& IT sm , Y= T sa IT sm and z = T om IT sm. Similarly
the instantaneous heat supply, qs, to the reversible harmonic engine can be expressed in
dimensionless form utilising the average heat supply, qsm:

(12), (13)

(14)

Here a = q.Jqsm is the relative magnitude of the amplitude of heat supply. The phase
shift, ell , is the unknown variable which is adjusted to maximise the role played by
changing source temperature. Substituting Eq.s (12) and (13) into Carnot efficiency Eq.
(15) and retaining only the first two terms of the binomial expansion of
[I + Ysin( or + w) r' and rearranging the product of multiplication by neglecting the
second degree terms, the instantaneous thermal efficiency of the heat engine is then:

TIc = 1- To / Ts ; TIc =1- z+ yzsin(or + w)- x sin or (15), (16)

where (l-z) can be interpreted as steady efficiency. The error ofthese approximations is
less than 1 % for T sa <O.3Tsm' The instantaneous power of the engine is:

W' = [I + asin(or +eIl)J[(I -z+yzsin(or+ w)-x sin or] (17)

 234

4r-------------~----------_.
I.ISr-----------:------
"'~,.- ..... ,
w x=/4/360 / ' -"", a=0.6

2 1.1
,I
" "
,
I \
, /4/360 .___ '.
I ,,, ..... ,,. ---..... , 0.3 \
o , 1'* ..... -- .. _ .. - • _ ..... ,
" ,,;.,--
1.05
-. ... .:-, \
:r
II'
7/360 0.3 ""~:'"
.,..
---- - - - - - - - - - -----
0 .

/41360
0.5
-4~-------~---~--~ 0.95 ""0- - - - - - : 2 - - : - - - - - - - : - - - - W - -6:-
-4 -2 0 2 'II 4
Figure 2. (a) Variation of+ along the phase shift of source temperature w for c = x 1 (yz) values 3.0; 0.9; 0.5
and 0.1. (b) Variation of dimensionless work W along w for z = 295/360, Y=20/360 and different a, x values.

Integrating Eq. (I 7) with respect to time over one period and bringing in dimel'..,'onless
form, by means of the work of reversible engine running at constant average heat supply
qsm, between Tsm and Tom, results:

W = 2n; - 2zn; - axn; cos $ + ayzn: cosS$ - w)

(18)
2n;(I- z)
To find the value of ~ which maximises this work, its derivative with respect to ~ will
be taken and solved for $:

cp = arctan [ yzsin w ] + nn; (19)

-x+yzcosw

This is the optimum value of ~ that maximises the work. The value of n is determined
by the second derivative of Eq. 18 with respect to ~. For yz.cos(w)- x = 0 and for
$ = ±n; / 2 , physically meaningful solution should be chosen. As an example, the data of
the first line of Table 1 have been used and variations on Fig. 1b are obtained. Fig. 2a
shows the variation of ~ along w in the range -n; to n; for different values of x/(yz) = c:
3.0; 0.9; 0,5; 0.1.
Figure 2b shows how the dimensionless work of one period is affected by the
amplitudes x and a. As expected, increase of 'a' increases the importance of choosing
proper phase shift ~ for operation. Fig. 3 shows the effects of z and y on dimensionless
work for two different w.
TABLE I. Standard and auxiliary values of the parameters.

x y z w d a P
14/360 20/360 295/360 n;/4 0 0.3 0.4
7/360 o 295/700 n; n;/6 0.5 0.6
 235
1.6 1.8

W
W y=O.20
1.6
1.4

UI5 0./5
1.4

12
1.2

_" -"" 0.10

__ ..... ,,'

---
:.=--=------.... 0.11 z
---- --::--~ ....
0.6 0.8 Z I

(a) (b)

Figure 3. Effect ofz and yon dimensionless work W at a = 0.3, x = 14/360 and (a) w = 1116, (b) w = 51116.

4. Periodic Change of Price and Its Coupling with Source Temperature

Hourly changes in power demand result that utility companies change price P of the
produced power hourly. As a first approximation this change can be expressed by a
harmonic function in dimensionless form:

P/Pm =I+psin('t+d) (20)

Here Pm is the average price and p is the dimensionless amplitude of price variation. The
total monetary value of the produced power over a period, V, can be obtained by
mUltiplying the produced work, Eq. (IS) with price P of Eq. (20). Utilising average
power Wand average price Pm over dimensionless period 21t one can express
dimensionless value of the work produced over one period as:

V = y(1 + asin('t ++»(1- z- xsin 't + yzsin('t + w»(1 + psin('t +d»)d't

qm(1-z)21t·Pm 0

(21)
Derivation of the result of integration with respect to + and solving it gives

... (zysin w +P(I-z)sin d )

'I' = arctan + n1t (22)
zycosw +p(1- z)cosd-x

This is the optimum phase shift of operation. For the simple case of constant source
temperature (y = 0) the effect of price variation on optimum phase shift is similar to the
effect of variation of source temperature as explained in Section 3; the only difference is
that w will be replaced by d and yz will be replaced by p (1-z):

..
'I' = arctan
((1- z)psin d )
P(I - z)cosd - x
+ n1t (23)
236

- Observations made on Fig. 2 on the relation cj>(w) and W(w) with given values of , a'
and y are valid for this case also, however for the latter instead ofyz = 0.0455 with Y =0
and p (l-z) = 0.0455 or z = 2951360 and p = 0.25.

5. Finite Time Effect

The problem of endoreversible engine with external resistances as shown on Fig. 4a.
Instantaneous Novikov-Curzon-Ahlborn relation may be applied [6]:

(24)
Substituting Eq. (10, II) and expanding firstly the denominator in binomial series for
power (-1), multiplying by nominator and expanding in binomial series for power (1/2)
results in:

T} = 1- ZJl2[I_ .!..(ysin( t + w) -
2
I.sin t + xy sin tsin(t +
z z
W»] (25)

The average of this efficiency over one period, assuming that the rate of operation is
approximately constant, results:

T} = 1- ZII2[1 + ; x; COS(W)] (26)

In other words Novikov-Curzon-Ahlborn efficiency need~ a correction factor near 1.0 in

case temperatures of sink and source oscillate.

~ =CQ7Ist

C
T.lt)
W(t)

(a) (b)

Figure 4. Special cases of reversible engine (a) with external irreversibilities, (b) with a storage
on source side where heat flow is uniform.
 237

6. Effect of Storage and Coupling with Price

Utilisation of change of environmental temperature by means of thermal storage at heat

source, which has an average temperature Tsm , base on harmonic operation of the
engine. Energy balance of thermal storage and heat supply expressions are:

(27), (28)

The integral over a period gives: q hm = q sm . Solving this equation one can obtain
(29)

where g' =g + 1t/2 and $' = $ + 1t 12 and average temperature of the storage Tsm is
known. Considering this problem as a special case of harmonic source temperature
variation, one can obtain optimum $ for given g. An important special case is constant
o.
heat supply, i.e. q ha = For this case one obtains

Ts = Tsm + ysin(rot + $'); y= qsa (30a,b)

CsroTsm

As expected, allowed largest q sa improves production, limited by Tsmax ;

qsa =T·
sa , Tsa -< Tsmax - Tsm (3Ia,b)
C sro

From the optimum value of $ (Eq. 19)

$ = arctan ( sin w ) + n1t (32)

cosw -c

With w = $' = $ + 1t 12, one obtains a value for $, as solution of the equation:

'"
'I' = arctan
( sin($+1t/2) ) + n1t (33)
cos($ + 1t I 2) - c

The solution of this equation for optimum $ is

cjl= - arcsin (l/c) and cjl=1t + arcsin(lIc) (34a,b)

For c < 1 there is no horizontal slope, i.e. the function is changing monotonically.
Substituting Tsa / Tsm = Y = (qsa / roC s )(1 I Tsm) one obtains c = (x / z).(roTsm C s I qsa) .
In case of coupling thermal storage effect with the effect of change of price,
optimum phase angle will be:
 238

,j, sin(<I>+nI2)+p'sind
tan 'f' = ---'-'---"'---=----- p'= p(1-z) (35a,b)
cos(<I> + n12) + p' cosd - c

However, for frequently encountered situations either d == 0 or p = O. For d = 0

<I> = - arcsin(1 I(c - p')) (36)

7. A General Thermal Module and Its Applications

Above mentioned specific results about external resistance (Novikov-Curson-Ahlborn

problem) and heat storage cases can be combined in a general thermal module which
consists of: First a series of resistances and storages between a source and a reversible
engine, second a reversible engine and third a series of resistances and storages between
the reversible engine and a sink. Such a set of thermal resistances, storages and a
reversible engine is called thermal module [1].
For perodic change of source temperature (Ts) and environment temperature (To), the
rate of operation of the reversible engine (W' =qh - q, ), working between its supply
temperature Th ( < Ts) and discharge temperature T, (> To) may be changed periodically
at such a suitable phase (<I» and amplitude (qha) that work output will be maximized,
with constraints of given Ts, To and qh,max'
Equations for <I> and qha which maximise work output of a period are obtained by
combining Eq. (19) (between temperatures Thand T,) and relations between: At source
side Ts and Th ; at sink side T, and To respectively. The second group of relations are
obtained using complex variable and matrix method explained in [7] for one
dimensional periodic heat conduction through composite layers.

8. Upper Harmonics of Temperature Variation

Any non harmonic variation of source or environment temperature can be expanded

In Fourier series with parameters Tsa' , Tsa2 , ... ; <l>s' , <l>s2, ... ; Too • " Too • 2 ,.... The
procedure given above can be applied to upper harmonic functions, just in the same way
to find ¢ql' ¢q2, ¢q3'"'' ¢qi . It was shown that qsa = q'im-qsm . The amplitudes qsa2 , qsaJ
, ... have to satisfY the condition that at their own maximum where t = t;!llax , i =
1,2,3, ... (with 2 7r . timax + ¢qi = 7r I 2 + 2n7r , etc.) same condition will be satisfied.

j-I
qlim ;?:qsm +Lqsai sin(Oli t jmax+<I>qi)+qsaj,j=2,3,.... (37)
i=1

9. Other Types of Engines

The extension of the method of rmding optimum operation from a single harmonic
variation of source and environment temperatures to Fourier series expansion of an
arbitrary temperature variation, as described in the previous section is only intuitive. It
needs mathematical proof whether the result is a true optimum.
 239

The general model assumes a work output (or input , which will be indicated by a
negative value). However, this work can be in form of exergies. The engine may
increase exergy , if required also energy of a fluid which flows through. This difference
in type of work exchanged does not need any change in the above theory. To improve
efficiency and economy of solar driven absorption system, one may use weak solution
storage and ammonia receiver as exergy storage. The high pressure side should run
according to variations of source and environment temperatures. The low pressure side
should run according to variations of refrigeration demand and environment
temperature. The only constraint between two engines is the capacity of matter-exergy
storages. Finding the optimum design parameter (characterizing the interface, i.e.
storage capacity) one can optimise the complex system.
The following constraints will be discussed: First, same thermal source will be
used by the engines and the constraint is the given total heat supply to these; i.e.
parallel running. Second, the heat output of one engine may be completely or partly the
heat supply of the other, through storage and resistance; series running. Third, the
exergy output of one engine may be consumed totally or partially by another engine,
having an exergy transfer function in between; i.e. exergy series running. The
study of such complex systems will be in three stages: First, determination of mean
values (qm , T m , ... ) ignoring the fluctuating components. Second, determination of
optimum phase of working for each engine (<I>ql , <l>q2 , ... ). Third, Determination of
optimum amplitude of fluctuating component of heat supplies each engine.
The first step is straight forward; constraints or design values of component will
suffice to find mean values which can be considered also steady state values. The second
and then third stages require application of one of the various methods like circuit
theory, linear programming or sectioning. In sectioning, the interfaces of subsystems
will be characterised by parameters of which the optimum values will be searched for.
The optimum phase and amplitude of working of each engine will be expressed in terms
of these interface values.

10. Applications

Low temperature power production with solar energy driven Stirling engine may be
considered as an example of application of Section 3. Relatively large value of y (=T as 1
SM) ego 0.2 and z (=Tom 1 SM ) e.g. 0.8 is expected to be observed. These values of
parameters with w = 1t/4, x = 14/360 can be used to obtain optimum phase shift <I> from
Fig. 2a and dimensionless work per cycle from Fig. 2b, 3a, 3b.
Another field of application is geothermal power production. Although steady
operation has various advantages, on one hand change of environment temperature on
the other hand change of demand, i.e. price, makes hourly change of power production
rate through natural or engineered thermal storage attractive.
For approximate coincidence of demand function with environment temperature
function, d = 0, for dimensionless thermal storage e.g. CswTsml(qsm a) = [(105 kJ 1 K 360
K) 21t 1(86400 s)]/(333 kW 0.3) = 26.2 and x = 14/360, z = 295/360, P = 0.4 one obtains
c = (xlz)26.2 = 1.24; c - p' = 1.17. This results in <I> = -58.7° and 238.7 0 (i.e. -1.0245 and
 240

4.166). w = 31.3 and 328.7 (i.e. 0.5462 and 5.737). For these solutions from Fig. 2b;
W= 1.045 i.e. 4.53 % more work due to usage of thermal storage and periodic operation.

11. Conclusions

To calculate exergy destruction or find exergy of control mass system or exit exergy of
control volume with timewise variation of environmental condition exergy balance
equation can be written in terms of fixed reference condition and correction terms which
can be calculated easily.
Day and night temperature difference is relatively large for energy systems with low
source temperature. This effect may coincide with variation of source temperature,
variation of price, existence of thermal storage at source or at sink and effect of external
heat transfer resistance. In each of these cases, harmonically changing rate of operation;
with an optimum phase shift with respect to environment temperature improves overall
efficiency. Solar and geothermal energy applications can utilise the results given above.
Increasing amplitude of the operation brings advantage if it does not disturb source
temperature. In case of heat storage, temperature fluctuation depends strongly on rate of
operation and puts limits on it. Optimum phase shift for a practically important case:
heat storage at source and constant heat supply, is an especially simple expression:

cI> = - arcsin(yz / x) and cI> = 1t + arcsin(yz / x) (38a,b)

Further study on different applications of the derived relations is recommended.

Acknowledgement

First author thanks sincerely Alexander von Humboldt Stiftung, German F.R. for
fmancial support during one month visit at Institiut fUr Thermodynamik und
Warmetechnik, Universitaet Stuttgart and Director of Institut Prof. E. Hahne for his
hospitality .

References

I. G()go~, Y., Ataer, O. E., iIeri, A. (1993) A thermal module to study exergetic utilization of environmental
temperature variation in 1.S Szargut (ed.), Proceedings of Energy Systems and Ecology Conference,
Cracow, 857-866.
2. G()go~, Y. A., Ataer, O. E. (1998) Thermal power generation considering periodic change of environment
temperature in M. Feidt (ed), ECOS'98, Efficiency, Costs, Optimization, Simulation and Environmental
Aspects of Energy, Energy Systems and Processes, Nancy
3. Szargut, J., Morris, D. R. and Steward, F.R. (1988) Exergy Analysis of Thermal, Chemical, and
Metallurgical Processes, Hemisphere Publishing Corporation, New York.
4. G()go~, Y., Ataer, O. E., (1998) Time-Dependence Effects on ThermodynamiC Optimization by Second
Law, METU, Ankara.
5. Bejan, A., Tsatsaronis, G., Moran, M., (1996) Thermal Design and Optimization, John Wiley and Sons
Inc., Toronto.
6. Bejan, A. (1996) Entropy Generation Minimization, CRC Press, New York.
7. Carslaw, H.S., Jaeger, J.C. (1959) Conduction of Heat in Solids, Oxford University Press, London.
BOND GRAPHS AND INFLUENCE COEFFICIENTS APPLICATIONS

Y.A.GOOO$
Aeronautical Engineering Dept.
Middle East Technical University
0653 I Ankara, Turkey

1. Introduction

A complex thermodynamic system needs as a first step clarification of components and

their interaction. In many cases the succeeding step which completes the study is
steady state analysis. However when the first step is accomplished i.e. interrelation and
internal behaviour of components are clarified, the unsteady behaviour of the system
also can be studied easily. The aim of this lecture is first to review bond graphs, give
examples and compare with linear graps, second to review the method of influence
coefficients, which is used for steady state processes and show how to apply it to
unsteady processes.

2. Bond Graph

2.1. ELEMENTS, INTERACTIONS AND VARIABLES

Bond graphs have interactions as lines and elements at end points of lines, in
comparison to linear graphs which indicate elements by lines and points of interaction
of these elements with each other by nodes at the end of these lines. Arrows on the
lines show the positive reference direction of flow and energy. Fig. 1 shows an
example. Elements, which undergo only one type of interaction with the rest of system,
are called one-port element although they have in general two terminals. Two basic
types of one-port elements are storage (C) and dissipator (R).

p
Q(t) C ......t--'Q'---II So

(a) (b) (c)

Figure 1. For a high water tank and a positive displacement pump (a) Hardware
(b) Linear graph (c) Bond graph.

241
A. Bejan and E. Mamut (eds.), Thermodynomic OptimiVJtion o/Complex Energy Systems, 241-250.
© 1999 Kluwer Academic Publishers.
 242

In both types of graphs each one-port has two relevant properties, such that during
an interaction the product of these gives the energy transferred. One of them is called
"flow variable", the other "effort varible". Table I shows these variables for different
energy domains.
In complex graphs, each bond has a number and if more than one energy domains
exist, symbols of effort and flow variables are attached to bonds. For heat interaction
the flow variable is taken as heat rate in simple treatment, which is called "pseudo
bond graph". Proper treatment leads to entropy rate.

TABLE I. Variables and their category for different energy domains in bond graphs.

General Flow Effort

(f) (e)
Linear movement Velocity Force
Rotational movement Angular velocity Torque
Hydraulic system Volume flow rate Pressure
Electrical system Current Voltage
Thermal system Entropy flow rate Temperature

2.2. CAUSALITY

There is a short perpendicular line on one end of each bond, which is called "causal
stroke" and indicates the element on which effort (e.g. voltage, pressure, temperature,
force, torque) is imposed by the other element. A source element is starting point of
determining the causal strokes in a system. Se is a source for effort hence always
Se -----t and Sc I , i.e. for a flow source the element with which it interacts,
determines the effort.

2.3. TWO-PORTS AND JOINTS

Some components of engineering systems have two ports, each interacting with a
different part of the graph, even in different energy domains. Energy conserving linear
two-port elements has constitutive equation set offollowing form:

(I)

If C 12 = C2l = 0 and Cll = -lIC22, it has properties of an electrical transformer; it is

called transformer and Cll = TF. On the other hand if Cll = C22 = 0 and C l2 = -lIC2l ,
the two-port element has properties ofa gyroscope, it is called "gyrator", C 12 = GY.
The factors TF and GY are determined by the physical characteristics of these
elements. In some devices TF or GY is a variable of time, the device is called
"modulated" .
 243

For convenience the nodes at which two or more bonds coincide are treated as
"joints" and depending on the connection of bonds such a joint is called zero-junction
or one-junction:

O-junction: parallel; eJ = ~ = e3 fJ+f2+f3=O

l-junction: series ; fJ=f2=f3 eJ + ~ + e3 = 0

Possible flow and causal directions and causal relations of two-port elements and
junctions are shown on Table 2. If energy storage elements are capacitors and causal
t
relations are integral type (i.e. flow is imposed on capacitor, e is function of Jf . dt ,
o
starting with sources like in thermal systems one can determine all causal strokes
without any problem. For special cases references [1] and [2] should be consulted.

TABLE 2. Possible reference flow and causal directions oftwo-ports and junctions.

Transformer 1
2
0- Junction
1
.. 2
1
~ TF 1 ~ ~I 0

2 3
~I TF ~I

2 2
Gyrator ~ GY ~I 1 - Junction ~ 1 I"

1
~IGYI
2
~
}
2.4. CAPACITY AND RESISTANCE FIELDS

The lumped components of a system can have simultaneous interactions with various
energy domains and at the same time energy storage possibility or dissipative function;
so that they are neither pure transformer nor pure capacitance. Such components are
needed especially in thermodynamics. They are called "field" and shown with C or R
with more than one bonds. In such a treatment entropy flow rate and temperature are
flow and effort of thermal energy domains.

2.5. STATE SPACE EQUATIONS

The bond graph is completed when:

a) Nodes (components) are identified (constitutive equations written)
b) Bonds established (numbers and effort and flow variables are indicated)
244

c) Reference flow directions are chosen

d) Causal directions and relations are determined
A completed bond graph is of great help even for steady state analysis of complex
thermal systems. However bond or linear graphs are indispensable for dynamic
analysis.
Because of various advantages, the following form is preferred as the set of
equations, which determine the state of a system.

i =/(X,u,t) (2)
x = X., X2, ... ,Xu and n is the number ofindependent (primary) variables, i.e. order of
system. u Consists ofut(t), U2(t), ... , u.n(t) input variables. The order of the system is
equal to the number of state variables.
Three succeeding steps to establish state--space equations are:
i) Select input, energy and co-energy variables. Input variables have been shown as
t
sources. Energy variables are; h = Jf . dt for capacities with integral causality. Co-
o
energy (or dual) variables are effort for capacities and flow for impedances. They are
secondary variables, which will be eliminated.
ii) Formulate the initial set of system equations
iii) Reduce the initial equations to state-space form

3. Inf1uenee Coeftieients

3.1. STRUCTURAL BONDS AND SINGLE SERIES OF COMPONENTS

Like in investigation of other properties of complex systems, exergy analysis depends

greatly upon behaviour of components and on the relation between component
performance and system performance. Quantitative study of these relations allows
concentrating our efforts mainly on the important components to improve system in
efficient manner. Below explanation is based to a great extend upon the explanation of
the subject by Kotas [4].
Coefficient of Structural Bonds (CSB) are shown by crk,i and defined as follows, by
indicating irreversibility rate by I:

.
(aIr)
Oxi (OIr ) I
,,>, ~ (::) ~ ill, "__ (3)

i.e. change of total irreversibility of system (subscript T) per change of irreversibility of

component k; both due to change of a parameter of system, Xi.
It is assumed that system is in steady state and input parameters are not changed
during the change of the parameter Xi at an amount dxi , except that the exergy input
 245

Em needs to be modified to keep the exergy of product EOlll unchanged. Since Eout is
kept constant, change of system irreversibility is balanced by change of inflow exergy,
all as rates: ~h = ~Ein .
Testing system components k and system parameters Xi with respect to 0k,i will
indicate largest 0k,i , i.e. will show which component's irreversibility should be reduced
more strongly to achieve greatest reduction of system irreversibility. If some 0k,i are
smaller than 1, reasons should be found and remedies implemented.
For a special system consisting of n components connected in series (Fig. 2) and
each with exergy ratio Wk :

(4)

The Ok increases monotonically with increasing k along exergy flow direction. By

. 1
expressing h, and Ij in terms of the fixed Eout and fi w k , taking derivative of these
with respect to w; one can show that

(5)

i.e. if due to change of a parameter Xi the irreversibility of a component j changes, this

change will have greater effect on in-flowing exergy (~Ein 1) as the component is
nearer to exit 01').
On is largest

Em

Figure 2. A system consisting of n components connected in series in exergy flow direction.

3.2. COEFFICIENT OF EXTERNAL BONDS

To study a large system through a study of its subsystems is convenient. However

during optimisation exergy inflow and outflow from subsystems should be variables, in
contrast with the whole system for which Eout is kept constant. Since subsystem has
many components with different irreversibilities, the ratio of change of exergy input of
complete system (~m) to change of an exergy (~Ej) on the boundary of subsystem
(both due to change of x;) is a suitable parameter, which is called coefficient of external
bonds (CEB):
246

x. =
].I
[OEBE.
in ) (6)
J x;=var

Here Ein is for the complete system and Ej is an exergy flow rate on the boundary of
subsystem. This concept is useful because one can show by partial derivation that

(7kS ,I. = X j,1

... (7kSS.,I (7)

Here superscript s means whole system, ss the subsystem under consideration.

Considering irreversibility in a component as difference of incoming and outgoing
exergies, and taking derivative of both sides with respect to variable Xi, comparing the
result with the derivative of total irreversibility of system, which is equal to derivative
of exergy inflow:

(8)

One obtains:
<Jk,i = I_I--I_l-
in Xj,i out Xj,i
(9)

Values of these coefficients for components and subsystems should be found from
numerical simulation at few points of the interesting range.

4. Applications

4.1. STEADY STATE

To model steady state thermal systems, material flow diagrams with indication of
energy transfer, which may be considered as simplified linear graphs, will be
sufficient. Fig. 3a shows [5] the schema of the hardware and Fig. 3b a graph of an
Absorption Heat Transformer (AHT).
The AHT is a device, which can deliver heat at a higher temperature than the
temperature of the fluid by which it is fed, by consuming negligible amount of work.
As shown in figure, the AHT consists of: an evaporator which is at moderately high
temperature, to utilise exergy of the source; a Low Temperature heat Exchanger (LTE)
to gain more energy from the source; a High Temperature heat Exchanger (HTE), to
bring a part of the supply fluid to desired high temperature, to close the
thermodynamic cycle of the actual working substance; and the auxiliary devices like
pumps, valves to maintain steady state.
 247

To identifY the independent parameters of the AHT it is necessary first of all to find
the number of parameters and the total number of constraints; the degrees of freedom
of the AHT is determined as the difference of these numbers. The parameters are
pressure, temperature, mass fraction and mass flow rate at the connection points of the
components, as well as energy and two mass interactions at the components;
altogether 73. The constraints are conservation equations for energy and species in the
components, pressure drop and the phase equilibrium assumptions, altogether 68. The
degrees offreedom ofthe AHT is determined as 5.

4.2. UNSTEADY ELECTROCHEMICAL ENERGY STORAGE

As mentioned above, simultaneous compression, heat interaction and matter or

electron through-flow necessitates to consider thermodynamic systems as an energy
capacitance with more than one port, i.e. capacitance field. In following bond graph
modelling of electrochemical cell is summarised along the approach by Karnopp [6].
Gibbs equation [7]:
dU= TdS-p dV + PI dml + ... + Pm dm m (10)

shows that temperature, pressure and chemical potential are effort variables, entropy
flux, volume change rate and matter flow rates are flows.

_ _-tl_-..,2

PumpP2

- -............-..J3

(b)

Figure 3. Absorption heat transfonner (a) Hardware (b) Simplified linear graph
 248

5.~

v ..
52-.,7
"

(a) (b)

(c)

Figure 4. Bond graph ofan electrochemical cell [6] (a) Basic case Eq. (10), (b) With transformed Gibbs function,
Eq. (11), (c) With chemical reaction.

Fig. 4a shows bond graph of such a system with causal strokes, indicating that S, V,
miare imposed by the environment. However, Legandre transformation gives Gibbs
function with T and p independent variables, (Fig.4b)

dG = -S dT + V dp + III ml + ... + J.1m mm (11)

(To change the sign of pressure to positive (+) a I-juction i.e. series combination with
source of pressure is used). Affinity A and rate of reaction J can represent the chemical
reaction:

A = ~:> viPi (l2a,b)

The reaction will be indicated by transformers of species. Because of its finite rate
also a resistance R is needed. However this resistance of reaction rate can be
considered as function of overvoltage 11 instead of affinity. Because of reaction rate-
temperature relation it is connected to the heat capacity of system. When all these
effects are combined and for this purpose 0- and I-juctions are used, one obtains for
one side of an electrochemical cell the bond graph given in Fig.4c.
 249

Such complex components like electrochemical cell require bond graph analysis.
However very frequently they are used in other complex and large systems, for which
linear graph analysis [8) is better developed.
Koenig et al. [3] describe foundations of the method to transfer information of
subsystem analysis to a complete system. However their book uses linear graph
techniques. Contributions to make the connection of subsystem analysis with complete
system analysis easier are needed, although the basic knowledge already exists.

4.3. INFLUENCE COEFFICIENTS FOR TRANSIENT CASE

Above structural bond coefficient Gk,i was defined for steady state. However in some
cases the thermal system functions in unsteady way. E.g. an aircraft engine always
starts to run from a stopped condition, undergoes a series of changes during the flight
and after landing it stops again.
A thermal system, like aircraft engine, which runs during one flight under
successive different load conditions, should be considered for a complete mission and
instead of structural bond coefficient, structural-functional bond coefficient should be
introduced:

(13)

Here lkj is the rate of irreversibility of component k at time interval j; 't is complete
time interval of a mission and h is total irreversibility of system.
In case of a simple exergy flow system consisting of series connected components,
the irreversibility near the product end has greater importance (Fig. 2). Similarly the
irreversibility a component at a time interval near the end of flight has greater
importance.

5. Conclusions

Bond graphs and influence coefficients are valuable tools for analysis and exergetic
improvement of large systems respectively. However procedures for their application
are not simple. Further studies on these subjects will lead to simpler techniques of their
usage.

Acknowledgment

The author thanks Prof.O.E. Ataer and Prof.Y. Ercan for valuable discussions on these
subjects.
250

References

l. Rowell, D., Wormley, D.N. (1997) System Dynamics: An Introduction, Prentice Hall, New Jersey.
2. Kamopp, N.C., Margolis, D.L., Rosenberg, R.C., (1990) System Dynamics: A Unified Approach, John Wiley
Sons, New York.
3. Koening, H.E., Tokad, Y., Kesavan, K.H., Hedges, G.H. (1980) Analysis of Discrete Physical Systems, Mc
Graw Hill, New York.
4. Kotas, T.J. (1985) The Exergy Method of Thermal Plant Analysis, Butterworths, London.
5. Ataer, O.E., Go~s, YA (1995) Simulation of Closed Absorption Heat transformer Utilizing Bond Graph, in
YA Go~s, A OztiIrk, G. Tsatsaronis (eds.) ECOS '95 Efficiency, Costs, Optimisation, Simulation and
Environmental Impact of Energy Systems, tstanbul.
6. Kamopp, D. (1990) Bond Graph Models for Electromechanical Energy Storage, Electrical Chemical and
Thermal Effects, J. the Franklin Institute, 327, 983-992.
7. Bejan A (1997) Advanced Engineering Thermodynamics, John Wiley Sons, New York.
8. Chandrashekar, M., Wong, F.C. (1982) Thermodynamic Systems Analysis-I. A Graph Theoretic Approach,
Energy, 7, 539-566.
REPOWERING OPTIONS FOR EXISTING POWER PLANTS

P.F. MATHIEU
University of Liege - lnstitut de Mecanique
Department ofNuclear Engineering and Power Plants
Rue Ernest Solvay, 21 .- Bat. C3
4000 Liege, Belgium

1. Why Repower an Existing Power Plant?

There are mainly two repowering options which can be summarized as follows:
• The« Substitute Heat Recovery Steam Generator(HRSG») option which
consists in the replacement of the existing combustion boiler by a heat recovery
steam generator behind a Gas Turbine (GT), making so a combined cycle (CCGT)
«Fig. 1».
• The« Hot Wind box » option which consists in the addition of a GT to the existing
boiler in 3 possible ways:
the air forced draft fan and the air heater are replaced by a GT whose exhaust gas
feeds the existing burners «Fig. 1».
the GT exhaust gas is only used to reheat partly or totally the feedwater as well as
producing low pressure superheated steam in a separate heat recovery boiler; in this
case, it is a parallel powered CC configuration.
When the repowered plant is coal fired, both the GT and the air forced draft fan
feed the boiler's burners; the boiler exhaust gas is then used both to reheat partially
the feedwater.

steam tI.rblll88

condenser

FIgure 1. Substitute HRSG repowering schematic diagram.

251
A. Bejan and E. Mamut (eds.), Thermodynamic Optimization a/Complex Energy Systems, 251-260.
@ 1999 Kluwer Academic Publishers.
 252

new gll8 tllblne

stack
coolers

flue gas

Figure 2. Hot windbox rq:lowering sdll:malic diagJllm.

1.1. THE MAIN REASONS FOR A REPOWERlNG ARE :

I. I. I. Energy/Environment
• Increase of the installed power by addition of the GT power output to the existing
steam cycle.
• Increase of the global plant efficiency with or without an increase of the total power
output. In the Netherlands, in particular, the hot windbox repowering is considered
as a means of burning high grade fuels, like natural gas, at high efficiency.
• Decrease of the emissions, mainly of NOx and CO2 , The GT and boiler are in series
and the system looks like a staged combustion.
• Increase of the share of of natural gas in the fuel basket and so increase of the
flexibility of the repowered plant with respect to the fuel.
• Increase of the operational flexibility, the combination of a GT with a Rankine
cycle allowing the load following transients, the covering of the peak and
intermediate loads, the operation at part load.
• The possibility of phasing the construction, the GT being able to be installed frrst
with a by-pass stack.

1.1.2. Site related reasons

• Lack of new available sites in the country
• Lack of heat sinks (rivers, sea, ... ) and recuperation of the existing cold source
(condenser, cooling tower, pipes, pumps, ... )
• No place for the erection of a new cooling tower
• The installation of an additional GT is easy and rapid, not requiring heavy civil
engineering works.
 253

1.1.3. Economy
• Recovering of all the existing components with a minimum of adaptations
• Low investment and O&M costs of the GT and so decrease of the generated kWh
cost
• Industrial maturity of all the components
• Short pay-back time and low financial risks
In order to give orders of magnitude in a repowering, a typical rise of the installed
power is 30%, a relative improvement of the efficiency is 10% (4-5% pts) and the
reduction of NOx emissions can reach 70%. As a general rule, a repowering, like a
cogeneration, is a case by case study ; even if there are some common guidelines, it is
very dependent on the state of the exisiting installations and of the site.

1.2. COAL VERSUS GAS

The repowering of a given power plant with a given GT (hot windbox option) efficiency
is about 2% pts lower when the existing boiler is fueled with coal instead of natural gas
due to some significant differences of the plant when it is fired with coal or with gas.
The fuel LHV are different (42-45MJlkg for natural gas and 20-25MJlkg for coal)
and the needed excess air is about 15% higher with coal. The sizes of the burners
windboxes and of the air pipes and of exhaust gas pipes are larger. Hence the size
of the boiler is 15 to 20% larger.
The distribution of radiation and convection heat transfer is different so that the
steam properties at the superheaters outlet are different
A coal plant needs more auxiliaries for the coal transport, grinding, for the flue gas
desulfuration and for the electrostatic extraction of the particles. The electricity
consumption is consequently higher.

2. Comparison of Two Repowering Options

Amongst the many ways in which a reheat steam cycle may be repowered using one or
more GTs, two basically different repowering schemes are considered here.

2.1. THE SUBSTITUTE HEAT RECOVERY BOILER REPOWERING OPTION

As above mentioned, in this scheme, the existing boiler is replaced with one or more
HRSGs. Feedwater heaters are usually removed, and the feedwater is used to cool the
stack down to the lowest temperature which is compatible with the fuel (sulfur content,
dew point). The steam mass flow requirement of the steam set detennines the number
and size of the GTs. Figure I shows a schematic diagram for such a repowered
configuration. This repowering method is the only one available if the existing boiler is
close to the end of its lifetime.
254

2.2. THE HOT WINDBOX REPOWERING OPTION

In this scheme (fig. 2), the existing boiler is kept in service. It should therefore have a
proper remaining life and a good availability. The GT replaces the forced draft fan and
exhausts directly into the boiler windbox. This latter needs to be modified because the
GT exhaust temperature is higher than the preheated air temperature. However, the
oxygen content of this exhaust gas is still 16% and is sufficient for the combustion to
take place. Of course, the combustion process (adiabatic flame temperature, radiation
and convection share in the flame zone) is affected by the change in the oxydiser
temperature and composition. Also the evaporator and superheater surfaces could be no
longer adapted to the new temperature profile. The question is : is it necessary or not to
modify the heat exchangers surfaces and if yes, how? Generally, only the economizer
surface is changed and is increased to extract the required total heat from the flue gas.
The oxygen requirement for the boiler combustion provides the necessary guidance for
the GTs to be installed. Figure 2 shows a schematic diagram for such a repowered
cycle. It can be seen that some of the feedwater heaters are kept in service and operate
in parallel with the new stack cooler.

3. Specific Technical Issues

The choice between the substitute HRSG and the hot windbox options requires the
analysis of the following technical issues which are specific to repowering.

3.1. STEAM BLEEDS

In the substitute HRSG scheme, the best design point efficiency is always achieved
when all the steam bleeds from the steam turbine set are closed. In the hot windbox
scheme, the maximum design point efficiency is obtained with an optimum bleeding
flow from the existing steam turbines. This latter flow is usually much lower than the
original one. Hence, most of the steam bleeds, but not all of them must be closed. In
the case of a hot windbox configuration, the global efficiency exhibits an optimum with
respect to the total bleeding flow.

3.2. BACK-END LOAD AND DERATING

When the steam bleeds are closed, the steam flow within the LP stages of the steam
turbine set is increased. Typically, closing all the bleeds would give a 25 to 30%
increase of the low pressure stages mass flow. This is mechanically unacceptable for the
bearings and for some of the blades themselves. It would be very difficult and costly to
modify the LP stages and bearings to accommodate this increased flow and the
associated mechanical load. Moreover, the condenser and boiler feedpumps would need
some modifications as well as the heat load is increased.
The solution is then to derate the steam turbine so that its power output is left
unchanged when all the bleeds are closed. This only asks for minor changes to the inlet
valves to the HP turbine section. If the utility is not looking for additional capacity but
only for fuel economy (as in the Netherlands), the steam set may be further derated but
 255

with an associated decrease of its effectiveness. Using this technique, the steam turbine
derating may compensate for the GT additional output leaving the total plant output
nearly unchanged after repowering.

3.3. FUEL FLEXIBILITY

After the first oil crisis, Western countries have converted some of their steam plants in
order to burn all the fossil fuels, namely, natural gas, oil and pulverized coal. Clearly, if
the boiler is replaced with a HRSG, coal cannot be used anymore. The GT itself must
be running on natural gas or on a light distillate. If the HRSG is supplementary fired,
this additonal combustion has to run on natural gas or liquid fuel as well.
On the other hand, if the hot windbox option is used, the GT must be running on gas
or on distillate as well. When a coal-frred boiler is repowered, the forced draft fans are
kept in service. The coal mills can then be operated with the required flows either of
preheated air or of a mixture of GT exhaust gas and preheated air. In this latter case, a
small air-air preheater has to be used. The efficiency penalty, mainly due to the
increased auxiliary power requirement, may typically amount to two % pts in that case.
However, because of the increased windbox temperature, gas velocities in the boiler are
increased. This is thought to be unacceptable if coal is used, because these higher
velocities would lead to erosion in the superheater and reheater sections of the boiler.
Hence, it is a great concern that the ability to run on coal could be lost when repowering.
In general, the existing boiler and the GT operate on the same fuel.

3.4. REMAINING LIFE AND AVAILABILITY

The repowering always associates new equipments with existing ones. The remaining
life of existing equipments and the associated availability must be taken into account. If
the boilers are old or if they are used for load following, this mode of operation leading
to frequent temperature changes which shorten the lifetime of the hottest components,
the existing boiler has to be replaced by a new HRSG. It appears to be the most suitable
solution in such a context as it is the case in Belgium where the base load is covered by
nuclear power plants.

3.5. EMISSIONS

Emissions characteristics are different and must be addressed in a different way

depending on the selected repowering option.

3.5.1. Substitute Heat Recovery Steam Generator option

Without any supplementary frring, pollutants of the repowered cycle are those of the GT
which is selected. Only NOx could prove unacceptable. NOx emissions are reduced by
well-known techniques like water or steam injection or dry low NOx combustors. If
some supplementary frring is used, CO, unburned hydrocarbons and NOx emissions now
depend on the supplementary firing combustion. It is impossible to generalize, because
many factors (burner design, flow patterns, ... ) must be taken into account. In many
cases, CO2 and S02 emissions depend only on the fuel that is burnt and on the efficiency
256

of the plant. Because the substitue HRSG repowering increases the efficiency by about
20%, CO 2 and S02 emissions are cut by about 20% as well.

3.5.2. Hoi windbox repowering

In this case, the combustion in the GT is followed by a combustion in the existing boiler.
Because some of the work is extracted by the GT before the boiler combustion is
conducted, the fmal burnt gas temperature is less than for a classical steam cycle running
with the same overall fuel/air ratio. This may lead to large reductions in NOx levels. A
staged combustion could further reduce the NOx content of the GT exhaust. For CO2
and S02 , the same applies as for the subsitute HRSG option, except that the cut in
emissions is about 10% for a given fuel, because of the smaller efficiency improvement.

4. Design Point Performance Analysis

4.1. DATA

The design point performance analysis was carried out using a home made computer
simulation program. The GT size is determined by the program itself, but its
characteristics are those ofa 150MW-class machine (see Table I).
TABLE 1. GT characteristics.
Compressor (expander) isentropic effectiveness: 0.91 (0.86)
TIT: 1ll0DC
Mechanical effectiveness: 0.99
Inlet M> (outlet M» : 2000 Pa (4000 Pal
Inlet conditions: dry air, 20DC, I bar

The pressure ratio and the machine output are optimized to provide the best possible repowered cycle
efficiency. Exhaust mass flow and temperature are dependent on the above mentioned parameters.

The selected steam cycle, which is running on coal, represents what state of the art
was some twenty years ago, and hence could be a good candidate for repowering. Its
main characteristics are summarized in Table 2.
TABLE 2. Steam cycle characteristics.
Drum pressure: III bar
Superheater and reheater outlet temperature: 540DC
Superheater discharge flow: 138 kg/s
Reheater inlet pressure: 36 bar
HP (LP) steam turbine isentropic effectiveness: 0.81 (0.83)
Design point power output: 167MW
Design point efficiency: 35.7% (LHV - natural gas)
Design condenser pressure: 0.07 bar
Condenser cooling water inlet temperature: 20 DC
Pump effectivenesses: 0.75
Economizer inlet pressure: 116.6 bar

2 LP and 2 HP steam extractions (bleeding flows are optimized when repowering). Other characteristics
cannot be detailed here but include realistic values of pressure losses, approach and pinch point temperature
differences, auxiliaries, generator effectiveness, ...
 257

4.2. HOT WINDBOX REPOWERING

The temperature of the exhaust gases entering the windbox is 580°C. Both the GT and
the boiler are running on natural gas and the stack temperature cannot fall below 120°C.
The GT that fulfils the boiler O2 requirement is found to have a power output of 56.7
MW with an exhaust flow of 193.6 kg/so
For a given GT size, the most interesting problem is the feedwater heating. It can
be calculated that, starting with a total bleed flow fraction Y=0.125, characterising the
original steam cycle, closing some of the bleeds until this parameter is 0.08 mUltiplies
the repowered plant efficiency by 1.01. The feedwater heating then needs to be done in
parallel by the new stack gas coolers and the steam feedwater heaters (see Fig. 2). With
the optimised bleeding flows, the net repowered plant efficiency (natural gas - LHV)
reaches 40.6%, which is 4.9 % pts above the original steam cycle. Given the necessary
steam turbine derating, a 21 % increase in output is associated with this repowering
option. The utility could decide to further derate the steam turbine set if no additional
power is required or if there is already an excess electric power in the country.
GT optimisation. When considering a range of GTs, an optimum pressure ratio was
found at 12.5. The influence of the pressure ratio on the plant efficiency is however
quite small, and using a GT with a slightly different pressure ratio does not give any
measurable efficiency penalty. Not surprisingly, the GT TIT has a much stronger effect.
Figure 3 shows that the efficiency increases almost linearly with the GT TIT ; i.e. the
efficiency is increased by 1.1 % for each 100°C in TIT.

4.3. SUBSTITUTE HEAT RECOVERY STEAM GENERATOR REPOWERING

The substitute HRSG repowering is computed using the same GT as for the hot windbox
option. Once again, the size is not predetermined and is computed by the program. The
exhaust temperature (580°C) allows to fulfil the boiler requirement even without
supplementary firing.
The assumptions specific to the HRSG are mentioned in Table 3. For a given GT, the
bleeding level and the supplementary firing in the GT exhaust are the more interesting
parameters.
TABLE 3. HRSG assumptions
Single pressure level - natural circulation
Economizer approach point: 20°C
Superheater pressure loss: 3% of the drum pressure
Reheater pressure loss : 2% of its inlet pressure
Economizer pressure loss: 2% of the drum pressure
Gas side pressure loss: 4000 Pa

Bleeding level. The best efficiency is achieved with all the bleeds closed, that is
with all the feedwater heating done by the HRSG economizer. This situation is similar
to that of a genuine combined cycle. If all the bleeds of the original steam cycle were
kept in operation, this would turn into a 1.2 % point (Le. 2.7%) efficiency penalty.
Supplementary firing. Because the steam cycle is a reheated one, it has a quite
important heat requirement at high temperature. The required heat may be provided by
a supplementary firing. The share of the latter in the total heat provided to the cycle is
258

represented by the ratio QGT / Qtot which is the ratio of the heat input to the GT cycle
over the overall heat input (GT combustor + supplementary firing). This ratio may
therically vary between 0 (steam cycle) and 1 (unftred HRSG). Starting with 1, a HRSG
can cope with values of this parameter down to about 0.7, which corresponds to
temperatures of the gas leaving the supplementary firing around 800°C. Quite
fortunately, the calculations show that the efficiency peaks at 48% when the considered
ratio is 0.75, that is when the GT fuel mass flow is % of the overall fuel flow.
The required GT has a power output of208.8 MW and an exhaust flow of713.27 kg/so
OT optimisation. The same conclusions as for the hot windbox repowering apply.
The efficiency peaks for a GT pressure ratio of 12.5, with a very flat optimum. The
influence of the GT TIT is very strong (even stronger than for the hot windbox option).
Figure 6 shows that the repowered plant efficiency goes up by 3% for each 100°C
increase in GT TIT. Line number two on the same figure is the unftred case which has a
slight efficiency penalty when compared with the optimum supplementary ftring case
because the supplementary ftring achieves a lower stack temperature with the reheated
non-optimized steam cycle under consideration. In the heat transfer diagram, the gas
cooling curve rotates around the pinch point (see Fig. 4).

60

-
U
°
'-'
48
48
2

i
44

42
U

i
40

E-< 38
38

1OC5O 1100 1160 1260

Figure 3. Na cycle efficiency (LHV) versus gas turbine TIT.

IlOO I. Fired HRSG option

100 2. Unfired HRSG option
IlOO 3. Hot windbox option
500 4. Original steam cycle
,.......
400
~
0
'-'
300
I="
200
- unfired HRSG
.00 ...... fired HRSG
0
0 0. , 0.2 0.3 0.4 0.& o.S 0.1 0.8 0.8
TIT (0C)
Figure 4. Heat transfer diagrams.
 259

s. Discussion of the Results

The repowering options illustrated above are now compared against each other.

5.1. UNFIRED AGAINST FIRED HRSG

The unfrred HRSG case is fIrst compared with the fired HRSG case, which proves to be
a better solution for reheat steam cycles. This frred HRSG case is then compared with
the hot windbox option. For decision making, the selected repowering option then
needs to be economically assessed with reference to a new combined cycle. Because the
steam cycle under consideration is a reheat one, the best efficiency is achieved with
some supplementary firing. Figure 3 shows that the efficiency advantage of the frred
case is about 0.5 % point over the unfIred one. This difference is due to the lower stack
temperature achievable through the HRSG heat exchange when supplementary frring is
used. Both heat exchange diagrams are superimposed on figure 4 for comparison
purposes.
Figures 5 and 6 compare the energy fluxes for these two solutions. These figures are
based on the assumption that the heat input to the system is constant, to help the
comparison. This points out the difference in repartition between the heat rejections to
the environment through the stack and the condenser, depending on the option which is
used. For example, the lower stack temperature achieved when the HRSG is
supplementary fIred reduces the stack losses from 22 to 8% of the energy input. For a
given GT size, some more powerful steam cycles may be repowered using the
supplementary fIred option. Steam plants up to 150 MW may be repowered using the
unfrred scheme with two 150 MW-class GTs. A 300 MW steam plant may be
repowered using two 150 MW-class GTs exhausting in a HRSG which is supplementary
frred up to 750°C. For the unfIred HRSG option, the ratio of the GT output to the steam
turbine output is 2 ; this ratio becomes I in the optimum frred case. Supplementary
frring drawbacks are complexity, reliability and emissions. The supplementary fIring is
an additional system in itself which can be a source of unavailability. The
supplementary combustion increases the emissions from the GT combustor.

5.2. HOT WINDBOX OPTION VERSUS THE FIRED HRSG

The frred HRSG option achieves an efficiency improvement much more important than
that of the hot windbox. Figures 6 and 7 compare the energy fluxes for these two
options. However, off design considerations tend to decrease this hot windbox design
point disadvantage. The substitute HRSG with the optimum supplementary frring
roughly doubles the output, while the hot windbox multiplies it by a factor between 1
and 1.30, depending on the chosen steam turbine derating.
When the repowered cycle has to operate on a load following basis, the off design
performances of supplementary fIred HRSG cycles are very suitable for this. In
particular, the upper range of the part load operation covered with the supplementary
frring as the controlling parameter exhibits very good efficiencies. When a country is
faced with the need to install additional power, repowering with a supplementary fired
HRSG is very attractive as it roughly doubles the output.
260

HRSG Stec;k (21.84 ttl

CMlr," OulllUI
(47.52 '61

~LTr,:t-.,
Figure 5. Unfired HRSG option - energy flux diawam.

0IIer,1I Output
148 "'1

Gis Turbine Steam Cycle Condenser

Au.IL I' .) LOR. 10.78 '.II 1"2.2'"
Figure 6. Fired HRSG option - energy flux diawam .
Boiler Stec;k (3.4 ""

o.ar,u Output
(40.5 tt,
Gas Turlllne OutPUt (10

Fuel input
(100 '61

Gas TurDine Sleem Cycle Condenser

Auxil. 10.32"1 Loss.s 10.98 "1 154.8 ",

Figure 7. Hot windbox option - energy flux diagram.

6. Conclusions

This study shows the optimal repowering option. This is not a reason why not
repowering a coal or lignite or other low grade fuels frred power plants; however the
efficiency and output will be less than when natural gas is the only fuel. Also some new
technical issues will have to be solved.
COGENERATION BASED ON GAS TURBINES, GAS ENGINES AND
FUEL CELLS

P.F. MATIllEU
University of Liege - Institut de Mecanique
Department ofNuc/ear Engineering and Power Plants
Rue Ernest Solvay, 21 - Bdt. C3
4000 Liege, Belgium

1. Main Features of Cogeneration

Cogeneration means simultaneous or combined production of heat and electrical

power (CHP). The heat can be used to produce hot water or steam for a system or a
process, orland to feed a refrigeration system. In the case of production of electricity,
heating and chilling, one speaks of trigeneration. Usually cogeneration leads to a better
utilisation of the fuel than a separate generation of the same amounts of electricity and
heat. When using a Rankine cycle either in a fossil fuel fired plant or in a combined
cycle with gas turbine (CCGn, there are two main ways to cogenerate electricity and
heat :
with a back-pressure steam turbine (the condenser operates at the pressure of the
process heat or steam water).
through bleeding from the steam turbine at the process pressure.
We will not examine here theses two options, usually used in large-scale
cogeneration plants (for instance, for district heating) and concentrate on small
cogeneration (less than let's say 40MW).
The assets of cogeneration are :
Cut of primary energy use and preservation of fuel reserves thanks to high energy
efficiency (85-90 %), that means an extraction of up to 2.5 times the amount of
useful heat and power from raw fuels in conventional electricity generation (35 to
40 % for a coal-fired power plant).
Lower emissions-for a small size gas fired CHP, for example:
- the emissions of CO2 are cut by more than 50 % due to the high efficiency
- the emissions of S~, the most responsible of acid rains, are practically zero
due to the use of natural gas, the cleanest of the fossil fuels
- the primary energy use is cut by some 35% ; the fuel saving when compared
to separate production is typically 10 to 15% but can be higher
- the NOx emissions are controlled and brought at levels not higher than with
conventional generation.
Production of electricity and heat close to the users, most of the time on the same
site. This is another way to consider power generation, with decentralized small-
scale power plants delivering either alternating or direct current Until recently, we
261
A. Bejan and E. Mamut (eds.), Thennody1lllmic Optimiv;Uio1l o/Complex Energy Systems, 261-270.
@ 1999 KIlIWer Academic Publishers.
262

were used to power generation by ever larger power plants located far from the
users and delivering synchronized alternating current to high-voltage transmission
lines and to local networks at lower voltages. This makes the industrial society
vulnerable to any significant disruption of electricity. Nowadays, electricity
generation plants move into the liberalization of the market and as a consequence,
the investors turn their money to smaller stations, that can be built more rapidly and
sited more easily, reducing so the risks on investments and giving shorter pay-back
times (less than 5 years against 20 to 30 years to recover the investment).
The cost of CHP plants is a leading factor for the decision-making.
The interest of a CHP is not only based on performance and environmental impact
considerations but mainly on its cost and this latter depends a lot on strategy of the
distributor regarding the price paid on the one hand for the purchase of the surplus
generated electricity sent to the network and on the other hand for the selling of
electricity taken from the network.
Like the CCGT, on-site cogeneration systems use most of the time natural gas as
the primary fuel but unlike the CCGT, they remain available even if the network is
unavailable for one or another reason. At this stage the question of the availability
of natural gas is raised. The combustion of this latter is clean when compared to
coal or oil for the same power output., with no sulphur oxides, controllable NOx
formation by primary (control of flame temperature, use of staged combustion) and
secondary measures (use of selective catalytic reactors) and smaller emissions of
CO2 (higher efficiencies and higher hydrogen to carbon ratio). If the natural gas is
available in the country, it can be transported continuously to the site, eliminating
the need for storage and handling.
Operation and maintenance of on-site generating systems will not require a staff on
the site itself. A remote monitoring will become a standard practice based on
remote real-time data acquisition, analysis and diagnosis. Candidates for
cogeneration will be, on top of major industrial installations, hospitals, universities,
schools, public buildings, hotels, office blocks, airports, shopping centers, ...

1.1. SMALL- SCALE COGENERATION

The best candidates for on-site cogeneration at present and for the near future are :
very small gas tmbines, industrial or aeroderived, with or without supplementary
firing (SF).
advanced diesel, gas and Stirling engines.
fuel cells : like batteries and photovoltaic cells, they generate direct current used in
particular in electronics, computers, TV and videos.
Due to a high utilisation of the fuel and the use of natural gas, cogeneration is one
of the most recommended option to decrease the C~ emissions. After the Kyoto
summit in December '97, most of the western countries have decided to increase the
share of CHP in their electricity generation network. Moreover, systems like fuel cells
do not generate CO2 in operation, however they are still very costly. In a sustainable
energy development., this is a proper way to supplement the increase of the share of
renewables like photovoltaics, windmills, micro hydrostations and the use of biomass.
 263

2. Cogeneration Based on the Use of a Gas Turbine

In the simplest case, this system is similar to a CCGT and electricity is generated
by an industrial or an aeroderived gas turbine (GT). The flue gas at the GT exhaust is at
450-550°C depending on the GT type and its sensible heat is recovered in a heat
recovery boiler with one (or possibly several) pressure level(s) to generate steam.
Unlike in a CCGT, this steam is used for a process and not in a Rankine cycle. The
heat in the flue gas is recovered free of charge, except that investment is needed for the
recovery boiler. The electrical efficiency is the same as that of the GT but the design of
the recovery boiler is simpler. This provides high pressure and temperature steam and
is well adapted to the steam demand of industrial processes, like in refinery, sugar
production, paper mills, breweries,... When the steam demand is not steady, the heat
generation must be adapted by a regulation. There are 3 ways for regulating the heat
produced :
The opening or closing of the variable inlet guide vanes (VJGV) of the GT
compressor, increasing or decreasing the inlet air mass flow rate and consequently
the flue gas mass flow and hence its heat content. However, this regulating mode
has a negative impact on the performance of the GT which operates at part load
with a lower effciency than at the design point (over a range of about 20% of the
nominal power).
The use of a supplementary firing (SF). This system consists in a battery of
burners fueled with gas, installed either in the GT exhaust gas or inside the heat
recovery boiler itself. In this arrangement, the GT is designed on the basis of the
minimum heat demand and the associated investment cost on the GT side is the
lowest. The heat surplus is provided by the SF whose role is to rise the gas
temperature at the boiler inlet, only to provide the additional heat demand without
producing electricity. Consequently, the resulting save in fuel consumption is
decreased with respect to a separate combustion (see the example in § 4). In
addition, the possible conventional steam boilers existing on the site can be used in
back up to cover the peaks of the heat demand.
lf the GT is designed to provide an intermediate heat demand, the 2 types of
regulation may be used. However, the feasibility of this is limited by the restricted
choice in the GT available off-the-shelf on the market. Consequently the selection
of the regulation mode is the result of a techno-economic evaluation taking into
account the investment costs on both the GT and the boiler sides, the heat demand
curve, the fuel cost, the sale price of the generated kWh.

3. Calculation of the Fuel Saving Compared to Separate Production

A CHP plant is characterised by the 3 following parameters: ElH, the cogeneration and
exergetic efficiencies "'CO and "'ex.

3.1. COGENERATION ENERGY ANDEXERGYEFFICIENCIES

The cogeneration efficiency is given by :
Tfco = (w + Q )/( AI f .LHV ) = E + H = H [I + E / H] (1)
264

where E = W I( M f .LHV ) % and H = Qt( M f .LHV ) % are the electricity and useful heat

fractions, LHV is the low heating value (MJlkg), M f represents the fuel mass flow rate (kg/s), Q the

useful heat power (W) given up to the process and W the net electrical power (W).

By definition, the exergetic efficiency is :

"" • [W'Q(E) +Q .1'l2(H)j/[ M f lifV ). E, I'l2(E)+ H, 'Q(H)(in %) (2)

where the quality factor is given by: FQ(H) = J - Tr/l'mi, Tmi being the mean integrated
or the average temperature of the closed loop of the 2 flows, the steam flowing at the
boiler outlet (point I) and the water flowing at the return (point 2) in the boiler. The
amount of heat given up by the hot gases to the steam/water loop is Q = m (hi - h2 )

where m is the steam and the return water mass flow rate in the closed loop. The
exergy flow rate is :
;"11e12 =;"[(hl -hz)-To (sl-S2)]=Q[I-To ITm;]=QxFQ (3)

where Tmi = (hI - hJ/(s\ - S2)'

In the particular case when the water return flow is at the reference temperature To, as in
an open circuit, FQ(H) is the usual quality factor of the steam flow at temperature T,
namely: FQ (H) = J -T/T.

3.2. FUEL SAVING

The efficiency of the separate production is given by :

11 c = ( W + Q) I( Q e + Q b ) (4)

where Q e and Q b represent respectively the amounts of primary energy, namely

M f x UlV ,for the generation of the amounts of electricity W and of heat Q in the
separate production. After some algebraic transformations, Tic writes :

1}c= [-e+l-ej-I
-- (5)
1}e 1}b

where e=E/(E+H) and J-e=H/(E+H) whereas 1} e = WI Q e and 11 b= Q I Q b are

respectively the efficiencies of the reference electricity power plant (here 55%) and of
the reference boiler (here 90%) in separate production.
The cogeneration is interesting compared to separate production if 1}00> 1}c
 265

Fig. I shows a set of CUIVes representing the cogeneration efficiency 1Jco versus the
generated electricity fraction e with the heat quality factor FQ as a parameter. The
lower CUIVe represents the decrease of the separate production efficiency 1Jc when the
electricity fraction varies from 0 to 100% and provides the efficiencies for the reference
cases respectively of the boiler (point A, where e=0%, 1}b=90% whilst 10% of the fuel
energy are the heat losses) and of the power plant (point B, where e=IOO%, 1Je=55o/o,
that is the efficiency of a CCGT operating on natural gas). The cUlVe 1J co(e) moves
upwards in the diagram when FQ decreases, with A and B as fixed points. For each
CUIVe FQ, there is an optimum of fuel saving, that is a function of e. This optimum is
the point of the cUlVe the most remote from the cUlVe 1Jc, hence where the difference
1J co (FQ ) - 1J c is the largest for a given FQ and not the maximum of 1J co. When FQ
is low, this optimum is reached for a low value of e. This region corresponds to the use
of small GTs and gas engines with a recovery boiler. When FQ increases, that is when
the process steam is at higher pressure and temperature, the optimum moves to the right
towards higher electricity fractions e. We cross then regions where more complex
systems are used like successively GT + recovery boiler + supplementary firing, then
GT + recovery boiler and finally, for high values of e, combined cycles with steam
extraction. It is also obseIVed that the optimal zone is fairly flat showing that a rather
large range of GTs size is possible without leading to big efficiency penalties. Of
course, in the final choice of a configuration, a techno-economic compromise solution
has to be found. This illustrates quite clearly that each design of a cogeneration plant is
a case by case study.

100%
.......
95%

90%
?
c 85%
0
v 80%
~., "-
~

75%
~~FQ2 ::...
---~ :
'T)c ' .......;,.. .
70%
(separate : ~ :
65%
production) .
··· ~
...~
60%
:e1 • e2
55%

""
C>

Figure 1. Cogeneration efficiency 'leo as a function of the electricity fraction e, the quality factor FQ being
the parameter. In dotted lines, the iso fuel saving curves.
266

The fuel saving is the difference between the fuel (or primary energy in GJ)
consumption in separate systems and in the cogeneration plant. In percentage of the
fuel consumption (or heat input) in the cogeneration process ,it is given by:

Fuel saving (%)=EIT/e + HlT/b-l (6)

where El71e and H/T/b represent the relative fuel consumptions (or the heat input) in a
CCGT with a 71e=55% and in a boiler with a thermal efficiency T/b=90%. On Fig. 1, the
curves of iso-fuel saving are parallel to the curve 71c and tangent to the curves
corresponding to a given FQ at the optimal point. For example, the curve
corresponding to a fuel saving of 10% compared to the efficiency 71c of separate
production with the same fuel represents all the points where the same amount of E +H
is obtained with only 90% of fuel consumption instead of 100% and the corresponding
71co is 71c10.9. Let us consider the curve with a constant FQ2, a saving of 10% is
obtained for er60%. 71c is there equal to 65% and hence 71co (optimum on the curve
FQ2) is 72%. A saving of 20% can only be obtained with lower values of FQ. SO, for
el=45%, 71c=70% and 71co =95% (optimum and not maximum of the curve with a
constant FQl). This represents a fuel saving of [(100-70)/95)=27%. The line joining
the optima for all the curves FQ (crosses on Fig. 1) can be used as a first guide for the
best cogeneration system. The lower the value of FQ, the lower the value of e
corresponding to a maximum of 71co and to the optimum. In practice, the leading
parameter is generally the heat demand at a given temperature and pressure (FQ is
given). From the corresponding FQ curve on Fig. 1, the optimal electricity generation e
(%) can be derived and the fuel saving as well and the GT can be selected to matcb at
tbe best the heat demand.

4. Applications of a Small Scale Cogeneration System

4.1. COGENERATION BASED ON TIlE USE OF A GAS TURBINE AND A FIRED

RECOVERY BOILER

Fig.2 shows the cogeneration unit ordered by the oil company ESSO and built in
partnership with the private electricity company ELECTRABEL, installed on the site of
a refinery next to the harbour of Antwerpen in Belgium. The refinery asks for a steam
flow at 42 bar and 385°C ranging from 70 to 170 tonlhour. The water is returned at 42
bar, 95°C. What are the energy and exergy efficiencies? Here, the GT has been
selected and is rated at 36MW with 320/0 efficiency (E=32%). Hence the heat power
input in the combustion chamber is :
Qin == WI71e = 36/0.32 = 112 .5MW (7)
and the fuel mass flow is :
M f = Q in I UfV = 112 . S I 4S = 2. S kg Is (8)
the fuel being natural gas with a LHV=45MJlkg.
 267

B
3
9
10

11 14

2 5 12 c
A
1 7

8
A. Gas turbine 1. Air intake 6. Generator 11. Evaporator
B. Recovery boiler 2. Compressor 7. Exhaust gases 12. Superheater
C. Oil refinery 3. Gas supply 8. Supplementary firing 13. Steam pipe
4. Combustion chamber 9. Feed water 14. steam drum
5. Expander 10. Economizer 15. stack
Figure 2. Combined Heat and Power unit (CHP) in a refinery

4.1.1. Production ofthe minimum steam demand: 70tlh (=19.44kgls)

The enthalpies and entropies of the steam and water are :
at 42 bar, 385°C: hJ=3173.8kJ/kg; s)=6.69kJ/kgK
at 42 bar, 95°C: h2=401.1 kJ/kg; s2=1.25kJ/kgK
The heat power given up to the refinery steam/water closed circuit from the one-
pressure level unfired recovery boiler is :
Qlh =19.44 x (3173 .8-401.1)=53914 kW = 53.9MW (9)

the thermal efficiency is the heat power fraction transfered to the process/heat
input in the hot source: H == 53.9/112.5=48%
the energy cogeneration efficiency is :
17 co= (w e+ Qlh }t Q in = (36 + 53.9) II 12.5 = 80%(= 32% + 48%) (10)

the exergy efficiency which is the useful energy you can extract from the fuel
chemical energy taking into account the quality of the heat is given by :

Tfex=Exl+HxFQ (11)

Here the feedwater is returned at 95°C, and not at the reference temperature as it is
often the case in an open circuit. This is taken into account in Tmi through the
enthalpies and the entropies of the steam and water flows. Then one has:
268

7J ex = 0.32 x I + 0.48 x 0.43 = 52 .8% (13)

E / H = 0.32/0.48 = 0.66 (14)

Fuel saving = E /0.55 + H /0.9 - I = 0.584 + 0.53 - I = II .4% (IS)

For a sake of accuracy and correctness, a complete heat transfer balance from the gas
tuIbine exhaust gas to the water/steam process has to be calculated taking into account
the temperature of the feedwater entering the recovery boiler. The effectiveness of this
heat transfer is generally less than the reference case (90%). Hence, when the
feedwater comes back at a higher temperature than To, the fuel saving is lower.

4.1.2. Production ofthe maximum steam demand: 170tlh (=47.22kgls)

Now the supplementary firing (SF) is needed in the GT exhaust gas. The total power
given up to the refinery is: 47.22 x (3173.8 - 401.1) = 130.93MW.
From the previous point, the GT flue gas contribution is 53.9 MW ; as a result, the SF
provides 77.03MW, that is 1.7kgls of natural gas consumption. Consequently, the total
heat input in the system is 1 12.5MW from the GT and 77.03MW from the SF, namely
189.5 MW and the total fuel consumption is 2.5 +1.7=4.2kgls (=189.5/45). Then one
has :

E=36/189.5 =1<]>10, H =1309/189.5 =69%, E / H = 0.27 and 1/ =88% (16)

co

1/ ex= (36+ 56.9)/189.5 = 49% (18)

Fuel saving = 0.343 + 0.77 - 1 = 11.3% (19)

When the SF is in operation, H and 7Jeo increase from 48 to 69% and from 80 to 88%
respectively whereas at the same time E decreases from 32 to 19% because the
cogeneration system operates closer to a boiler. Fig.l shows that, when e decreases
from 40% (32/80) down to 21% (19/88), FQ remaining constant (0.43) since the
temperatures of the process are unchanged (the SF only increases the steam mass flow),
'leo increases by 8% points whilst the fuel saving hardly decreases by 0.1 %point. This
is due to the shape of the iso-fuel saving curves which are very flat around the optimum
on the considered FQ curve where the design point without SF is normally located. On
the other hand, 'lex decreases because the fuel used in the SF does not take part to the
electricity production.

5. The Next Generation of Small-Scale CBP : the Fuel CeU Power Plant ?

Fuel cells are under an increasing development as effiCient (high energy conversion
efficiency), clean (low emissions of NO", SO"' CO, unburnt hydrocarbons and no CO2)
 269

and quiet (very low noise and vibration levels) generation primemovers, for both
stationary and automotive applications. Their modular nature provides the potential to
cover a range of unit size from a few kW. to several MW.. They are fueled with
hydrogen or a syngas (CO+H2) and produce water. Four main types are under
consideration for CHP applications and are characterised by their operating
temperature, their electrolyte, the materials :
Phosphoric Acid Fuel CeU (PAFC): operating temperature around 190°C,
l1co=82% with E.=40%, could operate on steam reformed natural gas (C02+H2).
Available on the market as small scale packaged cogeneration system in the range 50 to
200kWh..
Solid Polymer Fuel CeU (SPFC): operates at low temperature (80°C), solid
polymeric membrane as electrolyte, high power density and good transient response,
l1e=32 to 34%, low O&M costs. It is very well suited for the transportation sector.
Available in 2000.
Molten Carbonate Fuel Cell (MCFC): operates at 650°C, accepts a syngas
(CO+H2), possibility of internal reforming (l1e=55% instead of 45% with external
reforming). It is suited to large scale applications (50kW and more).
Solid Oxide Fuel CeU (SOFC): operates in the range 800 to lOOO°C, with a
ceramic electrolyte. When incorporated in combined cycles, l1e is 65 to 70010. Power
output: a few kWe to several MWe . It will not be commercially available prior to 2010.

6. CUP Criteria

6.1. TEMPERATURE
The P AFC is able to supply low pressure hot water at around 90°C or low pressure
steam at around 120°C. Therefore, it is already used in commerciallbuilding related
installations. The SPFC (80°C) is suited to produce low pressure hot water.

6.2. HEAT TO POWER RATIO

For an overall 1100=90%, 11. of MCFC and SOFC exceeds 50% giving HIE less than I.
They are not adapted to cogen applications. As to the P AFC and SPFC, their 11. is in
the range 30 to 40010 providing HIE around 1.8. This is applicable in the buildings
sector.

6.3. CYCLIC OPERATION

SOFC, MCFC and P AFC are inherently unsuitable for stop/start cyclic operation. In
conclusion, for many CHP applications, the high electric efficiency, which is usually
considered as a target, has a significantly reduced advantage, unless full credits can be
obtained for electricity exports. Consequently, the fuel cells cogeneration plants will
probably not enter the market of large scale, industrial cogeneration sector,
characterised by large HIE. The 2 low temperature fuel cell systems, PAFC and SPFC
are potentially usable in the commercial and building sectors provided the heat demand
is continuous (for the PAFC) and the temperature required in the host sites are low «
80°C for a SPFC). The micro-scale market (less than 5OkWe) represents a large
potential market for fuel cell based CHP units, but it is far from being a certainty. In
270

conclusion, the cost and the inability of the fuel cells to follow the heat demand
transients are challenges for their future development and application in cogenetation.

7. Conclusions

The matching of the heat demand and of a GT taises the question of how to design a
small scale cogenetation plant. If steam is wanted. it is recommended to use a GT and
a simple recovery boiler with one-pressure level and possibly a supplementary firing.
On the other hand, a gas engine is better suited to hot water production. using the heat
of both the exhaust gas and the cooling circuit of the engine. In a project of partnership
with an electricity distributor, the MINIMUM heat demand on the site is the basis of the
design, namely the steam mass flow tate and the exhaust gas tempernture. The next
stage is to design a one-pressure boiler and carry out a calculation of the heat ttansfer
from the gas to the water/steam circuit. The data are :
On the water/steam side: the feedwater tempetature, the steam pressure and
tempetature, the MINIMUM steam mass flow rate, the recovery boiler type (I pressure
level), the approach tempetature.
On the GT side: the power output, the open cycle efficiency, the pressure tatio, the
turbine inlet tempetature (TIT), the exhaust gas temperature and mass flow rate. From
heat balances on the heat ttansfers above and under the pinch point, the stack
tempetature and the pinch point difference tempetature are derived. Now the heat
ttansfer surfaces can be calculated, namely of the economizer, the evaporator and of the
superheater. Consequently, the cost of the boiler can be estimated. It has now to be
checked, from the global heat balance, whether or not the selected GT is fitting the
required steam mass flow M sl. The same procedure is remade with different
available GTs and the one giving the closest M sl to the required one is selected. If the
M sl provided by this GT is higher than the required value, then the GT has to be
operated at partial load by closing the variable inlet guide vanes (VIGV) accordingly.
. .
Roughly speaking, M sl and M air are more or less proportional. For example, if the
tatio of the calculated and required steam mass flows is 0.8, the corresponding tatio of
air mass flows at the compressor inlet is also roughly 0.8 and the VIGVs have to be
opened at 80% of the full opening. Of course, running at partial load, both the power
output and efficiency of the GT decrease, increasing consequently the fuel consumption
and decreasing the fuel saving compared to a separate production. However, the tange
of action of the VIVGs is limited, typically to a reduction of the air mass flow by 20%
from its design value when they are fully closed. Consequently, the appropriate GT has
to be found. In the opposite situation. when the M sl provided by the GT is lower
than the minimum required value, a supplementary firing has to be installed and
designed to follow the heat demand.
Amongst the different options obtained according to this procedure, one selects the
most flexible one (generally a GT with supplementary firing) giving the highest E and
T/co for a given H.
GAS DYNAMICS CYCLES OF THERMAL AND REFRIGERATING
MACHINES

A.1. LEONTlEV
Academician of the Russian Academy of Sciences, Professor
Moscow State Technical University named after N.E. Bauman
2-nd Baumanskaya, Moscow 107005 Russia

1. Introduction

On 26 July 1824 year, at a session of section on general physics of French Academy of

sciences, the academician P. Gerard has stated the contents of the book of the young
French engineer Sadie Carnot "Reflections about driving force of a fire and machines,
capable to develop this force". On this session there were practically all scientific stars
of France: Navier, Laplace, Fourier, Gay-Lussac, Poisson and other, which,
unfortunately, could not estimate ingenious ideas, incorporated in work Sadie Carnot.
The scientific and engineering world of France has not displayed in any form of interest
to work S. Carnot. And only through a few tens of years, due to work Clapeyron,
Thomson, Clausius it has become clear, that Carnot has for the first time formulated the
basic laws of thermodynamics and under the right it is considered founder of this
classical discipline. The great physicist Thomson (lord Kelvin) has named the book
Carnot "as the largest achievement of a science XIX of century".
S. Carnot has for the first time entered concept about convertible thermal processes,
that has allowed him to offer and to analyse a cycle of the ideal thermal machine of a
maximum (peak) efficiency.
On Figure I the Carnot cycles are represented in T-S and P- V coordinates.
S. Carnot has shown, what the convertible cycle of the ideal thermal machine, consisting
from two isotherms with heat addition and heat rejection (1-2 and 3-4) and two adiabats
of expansion and compression (2-3 and 4-\) is the most effective of all possible and
efficiency of such cycle depends only on the relation of the maximum and minimum
temperatures ( 11c = 1- T2 / TJ ). The similar conclusions were made and for a return
cycle of the refrigerating machine.

2. Analysis of Offered Ideal Cycles

As is known, one of lacks of a Carnot cycle is a significant difference of specific

volumes of a working body, that results in the large size of a flowing part of blade
machines or length of the cylinder in a piston machines. This lack can be removed by
replacement of adiabatic processes of expansion and compression on isochoric ( cycle of
Stirling) or isobaric processes (cycle of Ericsson) with intermediate regeneration of a
271
A. Bejan and E. Mamut (eds.), Thermodynamic Optimi1.lllion o/Complex Energy Systems, 271-278.
@ 1999 Kluwer Academic Publishers.
272

heat. In this case, at identical with a Carnot cycle profitability, it is possible essentially
to reduce the sizes ofa working part of the thermal machine.
1
p
T
,....-_ _Z-...._ _...., 2

(a) 2 (b)

4 3
3

a b s v
Figure I. Carnot Cycle in T-S and P- V coordinates.

In the proposed Figure I a gas dynamics cycles of the thermal and refrigerating
machine idea of internal regeneration of a heat is in essence used.

2.1. CYCLE OF THE THERMAL MACHINE

We shall consider a cycle ofthe ideal thermal machine in T-S coordinates (Figure 2, a).

T
2
T 5

(a) (b)
2
5
3
3 4
4

a b s a b S
Figure 2. Gas dynamics cycles thermal (a) and refrigerating (b) machines

During (1-2) comes true isothermal addition of a heat to a working body and the whole
added heat QI will be transformed to useful work. Further, as against a Camot cycle, we
 273

carry out adiabatic expansion of gas in supersonic nozzle from pressure P2 up to

pressure P3. During 3-4 isothermal rejection of a heat Q2 from a supersonic gas flow
comes true. The isothermal rejection of a heat can be carried out changing the area of
the channel on length. In this case, as it follows from the theory ideal thermal nozzle,
there will be braking a flow of gas with growth of static and full pressure. During (4-5)
comes true adiabatic braking of a supersonic flow in diffuser up to pressure P 5.
Stagnation temperature in a point 5 (T5*) will be less than stagnation temperature in a
point I (TI*) as during (3-4) we have rejected from gas quantity of a heat, determined by
the area a-4-3-b-a. Hence, to close a cycle, it is necessary from a point 5 up to a point I
compress gas in the compressor, by spending capacity

(I)

Thus, efficiency of a considered convertible ideal cycle will be equal to efficiency of

a Camot cycle:

(2)

As it is visible from Figure 2, a the sizes of the thermal machine, and also the sizes
of the compressor in an offered cycle tum out much less, than in a Camot cycle, at the
same efficiency. As it can see from Figure 2, a and with increase of a degree of
expansion in nozzle, i.e. with reduction of temperature T3, the work, spent on the
compressor, decreases.
On Figure 3, a is shown ideal gas dynamics cycle of the thermal machine without the
compressor. After isothermal expansion of a working body in the cylinder (or blade of
the gas turbine) (1-2), where the whole added heat QI will be transformed to useful

T T
2 3
TI

(a)
(b)

T4 3
5 4
Q2

S s
Figure 3. Gas dynamics cycles thennal (a) and refrigerating (b) machines
without compressor and expander
274

work, the gas acts in the subsonic channel, where comes true isothermal addition of a
heat QI'. In this case, according to the theory thermal nozzle, static pressure and
stagnation pressure fall, and stagnation temperature is increased by QI '/Cp . The gas is
accelerated up to speed of a sound in a point 3 and further extends on adiabat (3-4) in
supersonic nozzle. The rejection of a heat from a supersonic gas flow comes true in
isothermal process 4-5. For closing of a cycle without the compressor it is necessary to
maintain a condition Q2 = QI'. Then stagnation temperature of gas in a point 5 will be
equal to temperature a point I and the cycle becomes closed by adiabat of compression
(5-1) in supersonic diffuser. It is possible to show, that stagnation pressure in a point 5
under accepted conditions will be equal stagnation pressure in a point I.
In isothermal process 1-2 available work is equal

fdPp
2
,~
L= - = RT; In-. (3)
I 1;
,
~
As PI = PI and P 2= P 2 and L = QI : -, = 1/ exp(QI I RT; * ).
* * , I

1;
During process 2-3

M P,' K T.'
~= -In -~ + - - I n _3_. (4)
R R'2 K -1 T:I

~. M 2_1 K 7;'
From here: In _. " = - ---" + - - I n - .
R'2 R K -1 T:I
Ps' M 4_ 5 K 'Fs'
On the other hand for process 4-5 we have In -. = --- + ----In -, .
~ R K-l 7;

2.2. CYCLE OF THE REFRIGERATING MACHINE

Similarly it is possible to present gas dynamics cycle of the ideal refrigerating machine
(Figure 2, b). The cycle begins with adiabatic expansion of gas in the thermal machine
(1-2) with fulfilment of work. In process (2-3) of adiabatic expansion of gas in
supersonic nozzle there is the fall of static pressure and static temperature at the constant
stagnation parameters cr and p\ In isothermal process (3-4) comes true addition of a
heat, in result stagnation temperature (up to initial meaning T4' = TI ') is increased, for
 275

this purpose it is necessary, that Q1 I Cp = T; •- •

T2 , and stagnation pressure P 4
*
<
P j • falls.The process of adiabatic braking of gas (4-5) comes true in supersonic ideal
diffuser. The cycle becomes closed by isothermal compression of gas (6-1) with
rejection of a heat Q2 and expense of capacity on the compressor, determined area (l-a-
b-5-1). The part of this capacity takes from the turbine.
As it is visible from Figure 2, b refrigerating factor of an offered ideal cycle is equal
to refrigerating factor of a Carnot cycle:

(6)

By analogy to a cycle of the thermal machine it is possible to offer a cycle of the

refrigerating machine without expander (Figure3, b).
In this case, after adiabatic expansion of gas in nozzle (1-2), isothermal addition of a
heat (2-3) and adiabatic compression of gas in diffuser (3-4) isothermal rejection of a
heat in the subsonic channel (4-5) in quantity Q2' = Ql comes true. The cycle becomes
closed by isothermal compression of gas in the compressor (5-1), with rejection of a heat
Q2. It is visible, that refrigerating factor of such cycle is equal to refrigerating factor of a
Carnot cycle:

From Figure 3, b follows, that in an offered cycle the sizes of the compressor of a
refrigerating machinery in comparison with a Carnot cycle are much reduced and is
eliminated expander. Does not represent complexities to carry out all thermodynamic
calculations offered ideal gas dynamics cycles, using the theory ideal thermal nozzle.
In an ideal cycle of the thermal engine the processes (2-3) and (4-5) proceed without
irreversible losses of energy because of work of friction forces. Actually dissipation
effects render essential influence to gas flow, especially at supersonic speeds.
It is known, that at supersonic speeds of gas flow temperature thermal isolated wall
(equilibrium temperature) is defined (determined) from the formula:

Tw• = To + r - k -1
W}- = T" ( 1 + r . - -M 2) , (8)
2Cp 2

where r - recovery factor, in a general case dependent from a longitudinal gradient of

pressure and cross flow of substance. For a flow of a flat impenetrable plate r = Pr1l2 -
for a laminar boundary layer and r = Pr 1l3 - for a turbulent boundary layer.
For the majority of gases the criterion Prandtl (Pr) is close to unit and equilibrium
temperature poorly differs from temperature of adiabatic braking
 276

To. = To (k -M 2)
1 - -2-I o • (9)

It is known, that the Reynolds analogy, establishing connection between heat transfer
coefficient and friction coefficient, is fair and for supersonic speeds of gas flow, if heat
transfer coefficient to define as

(10)

3. Analysis of Offered Real Cycles

Thus, in a general case of rejection of a heat in process (4-5) or addition of a heat in

process (2-3) (Figure 3, b) temperature of a wall will differ from thermodynamic
temperature of gas, that will result in irreversible losses of energy in comparison with
ideal convertible processes. For reduction of these irreversible losses it is necessary to
reduce a difference of temperatures between gas and environment. It is possible in a
limiting cycle at the decreasing of recovery factors. As already was marked, by use as a
working body of gases, a recovery factor poorly differs from unit, however for some
mixes of gases (He-Xe or H2 - Xe) the Prandtl number, and consequently and r, it can be
essential less unit.
Other opportunity of reduction of recovery factor as the result of creation of a
positive gradient of pressure along a flow of gas. As is known, [3] in area of separation
of a boundary layer, at a cross flow of the cylinder very low, even negative meanings of
recovery factors were measured. During (4-5) supersonic gas flow comes true with a
positive gradient of pressure and it is possible to expect favourable influence of this
gradient to recovery factor. However it is necessary to note, that in a cycle of the
refrigerating machine the process of addition ofa heat (2-3, Figure 3, b) comes true with
a negative gradient of pressure. In work [I,2] influence of a cross flow of substance to
recovery factor is shown, however use of this effect at realisation of thermodynamics
cycles is represented inconvenient. The interesting opportunity of reduction of recovery
factor opens at organisation of jump of condensation on an exit from supersonic nozzle.
In this case the condensate can drop out on a wall of the channel and recovery factor
aspires to zero, since temperature of liquid drops is close to thermodynamic temperature
of gas. Especially perspective use of this effect in process (2-3, Figure 3, b) cycle of the
refrigerating machine is represented.
The basic question, arising at practical realisation of offered cycles, consists in an
opportunity of restoration of stagnation pressure in cooled channels (process (4-5,
Figure 3, a) in a cycle of the thermal machine and process (4-5, Figure 3, b) in a cycle
of the refrigerating machine). In work Gukhman [4] was shown, that at flow of ideal gas
on a cylindrical pipe it is impossible to realise thermal supersonic nozzle, if the heat
transfer comes true through a wall of a pipe.
From generalisation of Reynolds analogy follows
2St f (11)
C, =Pr~'
 277

where the function f takes into account influence of a longitudinal gradient of pressure
to Reynolds analogy.
For isothermal flow of gas in the cooled channel

(12)

For a case interesting us Tw= To and from the equation (12) we receive

r·f
(13)
2Pr7)' .

From the formula (13) follows, that at f=1 and Pr=I, M R= -0.5M/% i.e. it is
impossible to realise process of cooling of gas in the channel with reduction of entropy,
since always MJ.R> MR.
The change of stagnation pressure of gas is connected to change of entropy by the
formula
dP' k dT' dS
p' = k-I'Y-R (14)

As

(15)

and we have

(16)

hence

c!~* = _k_. dT* [T;~ _1+ 2 PrY, ] (17)

P k-l T()
T r·f
or

d~* =_~.dT*[2pr7)' _k-l').}]. (18)

P k-l T;, r·f k+l

For the cooled channel dT < 0, hence a condition of increase of stagnation pressure
(dP* > 0) is following
2Pr7)' k-l
--<--')..}. (19)
r·f k+l
278

From the fonnula (19) it is visible, that at Pr =1 andf=l this condition is impracticable,
as the maximum meaning of the right part of an inequality (19)
,,? (k - 1) / (k + 1) = 1. However if to take into account influence of Prandtl number
and longitudinal gradient of pressure on Reynolds analogy and recovery factor, the
inequality (19) can be maintained, that will result in growth of stagnation pressure in the
cooled channel. In work [5] it is underlined an opportunity of cooling of gas by injection
of water. In this case the Reynolds analogy is not restriction for growth of stagnation
pressure, though there are the additional losses, connected to interaction of a gas flow to
drops. Resulted in work [5] calculations of parameters of a supersonic gas flow, cooled
by an injection ofa liquid, specifY essential increase of stagnation pressure (up to 40 %)
at the rather small additives of water. The essential irreversible losses can take place at
braking a flow in supersonic diffuser. It is known [6], that in this case practically it is
impossible to carry out braking of a flow without shocks.
The analysis of cycles of thennal machines, realising a cycle offered is brought in
work [7].

4. Conclusion

In this paper ideal cycles of thennal and refrigerating machines are analysed, in
which for closing are used gas dynamics processes of gas flow in channels with heat
addition and heat rejection . Is shown, that for ideal processes the profitability of such
cycles is equal to profitability of Carnot cycles. At the same time basic defect of Carnot
cycles - large degrees of expansion in the turbine is eliminated. Moreover in an ideal
cycle the process of compression of gas comes true in the cooled channel.
There are open the questions of start of machines, working on these cycles, and also
problem of adjustment of capacity. It is a lot of unsolved problems, connected to use of
phase of transitions (evaporation and condensation).
However for practical realisation of offered cycles more thorough theoretical and
experimental researches is necessary.

References

l. Kutateladze, S.S., Leontiev, A.I. (1990) Heat Transfer, Mass Transfer and Friction in
Turbulent Boundary Layuers, Hemisphere Publishing Corporation, New-York.
2. Hartnett, J.P. (1985) Mass Transfer Cooling, Handbook of Heat Transfer Application,
McGraw-Hili Book Company, New-York.
3. Eckert E, Weise W. (1942) Forsch. Gebiete !ngenieurw, J3, 246 (in German)
4. Gukhman AA (1939) To Theory of Limited States of Gas Flow, Journal Tekhnicheskoy
Fisiki, 9, Ng6. (in Russian).
5. Vulis L.A (1950) Termodynamics of Gas Flow, Gosenergoizdat, Moscow (in Russian).
6. Abramovich G. N. (1976) Applied Gas Dynamics, Nauka, Moscow (in Russian).
7. Leontiev AI., Shmidt K.L. (1997) Ideal Cycle of Closed Gas Turbine Installation without
Compressor, !zvestiya Akademie Nauk, Seria Energetica, 3, 132-140 (in Russian).
ENERGY SAVING TECHNIQUES IN DISTILLATION
Thermodynamic Efficiency and Energy Conservation

z. FONYO and E. REV

Technical University o..f"Budapest,
Department of Chemical Engineering.
H-1521 Budapest, Hungary

1. Introduction

When considering a separation process there are two entirely different

interpretations of the word "efficiency". The first refers to product purity (i.e.
separation efficiency) and can be regarded as a pointer to the theoretical possible
improvements of separation in the given process, considering the separation unit
as a set of theoretical (equilibrium) stages. The second refers to energy
consumption (i.e. thermodynamic efficiency) and can be regarded an indicator to
possible reductions of energy consumption in the industrial separation. The first
aim of the present paper is to interpret the thermodynamic efficiency theoretically
and expound a comprehensive vie\v on heat degradation, losses and the efficiency
of thermodynamic coupled processes. After the theoretical foundation, the
various representations of thermodynamic efficiency (TE) are examined by
application to several examples.
Various alternative definitions of thermodynamic efficiency were previously
suggested in literature and recommended in precedence to others. However, these
definitions are open to different interpretations as far as separation is concerned
[1]. In addition, most seem to share the features that different phenomena such as
heat losses, heat degradation and coupling of the heat and component transports
are not treated separately in the theoretical explanation and the different (energy,
entropy and exergy) representations are not distinguished unambiguously at the
practical investigations.
The industrial separation processes (such as distillation or gas absorption
with stripping or extraction with solvent recovery) can be treated as steady
system of the first order, where the coupling of heat and component transports
can be regarded as the essence of the processes. By separating the entropy
consuming and entropy producing processes, the coupling thermodynamic
279
A. Bejan and E. Mamut (eds.), Thermodynomic Optimization a/Complex Energy Systems, 279--296.
@ 1999 Kluwer Academic Publishers.
280

efficiency can be defined, which expresses the effectiveness of the coupling.

From thermodynamic point of view, in the case of separation our goal is to utilize
the entropy produced by heat transport as completely as possible by the entropy
consumption of component separation. In addition to this coupling
thermodynamic efficiency (CTE), so called overall thermodynamic efficiency
(OTE) was also defined taking the additional losses and entropy producing
effects into account, inherent with all thermal processes.
The uniform energetic interpretation of distillation and other separations
based on the thermodynamics will be of assistance, in the evaluation and
practical implementation of energy conservation techniques of these processes.

2. Separation Processes as First Order Steady Systems

Although the linearity of the integral expressions for the transfer processes is
questionable in the irreversible themlOdynamics, the linearity of the
phenomenological equations can be assumed and proved to be appropriate in the
mathematical treatment. The entropy production <D of a system in this theOI)' is
given by:
<D=2:JkX k;:::O (1)
k
where J k is the thermodynamic flux k, and X k is the conjugate thermodynamic
force or driving force. If each flux depends only on the respecting conjugate
driving force, i.e.:
(2)
only diagonal diagonal Lkk phenomenological coefficients occur in the expression
(1):

<D=I JkXk=I LkkX;=I <Dk . (3)

k k k

In this case each process occurs spontaneously and independently inherent with
certain entropy production.
In general each flux depends on all driving forces and so cross-coefficients
also occur in the entropy production:
<D=I I LnkXnXk=I <Dk , (4)
k n k
thus entropy consuming processes can also proceed in addition to entropy
producing processes. When considering, e.g. two fluxes the entropy production is
given by:
(5)
 281

In this case the positive sign of each <f>k is not required, the only restriction may
be formulated as:
(6)

an one of the two fluxes can be an entropy consuming one forced by the other
flux owing to cross-effects. Such a process is called a coupled process
thermodynamically when entropy consuming subprocess utilizes the entropy
production of the other subprocess to a certain extent.
Every system if left to itself tends to change toward a final state of rest or
equilibrium. In the equilibrium state, the entropy production is zero, because
macroscopic fluxes do not exist. In a system left to itself, the driving forces and
drived fluxes decrease gradually and the system strives after zero entropy
production.
In steady (or stationary) states, the state parameters are independent of
time, but the entropy production is not zero, because finite driving forces and
fluxes do exist. Stationary non-equilibrium states have the important property
that under certain conditions (of which the most important are the supposition of
constancy of phenomenological coefficients and the validity of the Onsager
reciprocal relations), they are characterized by a minimum of entropy
production, compatible with the external constraints imposed on the system. This
principle is called the Glansdorf-Prigogine theorem in literature [2,3].
According to the number of external constraints imposed on the system, the
steady systems with coupled processes can be categorized as follows [4]:

- zero order, if not a single variable is fixed, all properties are time independent
and the system is in a state of rest or state of equilibrium;
- first order, if one single driving force fixes one property (e.g. temperature and
composition) to a constant value;
- second order, if two driving forces fix two properties (e.g. temperature and
composition) to constant values; etc.

The criterion of minimum entropy production for a first order steady system with
two fluxes, ,,,hen XI is fixed, can be formulated as:
8<1>
--=(
8X 2 "4~ T ~ +L~I )XI +2Ln X 2 =0
~ --
(7)

From this expression it follows that:

X2 ~2+L2IX (8)
I
2L22
and substituting this fonnula for Xl into (5), the entropy production may be
rewritten in the form:
 282

Cj)=L X2[1 (l'12+ L 21 )~ ]=L X~(I-p2) (9)

II I 4L L II I
II 22

where p is the so called coupling factor, i.e. the measure of the coupling between
entropy producing and entropy consuming processes.
In the case of thermal separation processes, the heat flow is flux 1,
maintained by an XI driving force of constant value due to an external constraint,
and flux 2 is the resulting component flow. Thus during the separation process,
one part of the entropy produced by heat transport can be utilized by the entropy
consuming process of the component separation. The range of the coupling factor
IS:

Ospsl (10)
It is clear that at p = 0 the system is completely irreversible, since Cj)=L II X I2 i.e.
only an entropy producing process takes place (coupling does not exist). If p = 1,
Cj) = 0 i.e. the system is reversible, the entropy produced by heat flow is
completely utilized by the entropy consuming component separation (total is
completely utilized by the entropy consuming component separation (total
coupling exists).

3. Coupling Thermodynamic Efficiency

In general it can be said that the entropy production may be calculated for the
first order steady systems as the difference of the diagonal elements, without
considering the off-diagonal elements:

(11)

In the case of separation, the first term would be the entropy production of the
compatible independent heat flux and the second term would be the entropy
consumption of the compatible independent component flux. The ratio of these
consumed and produced quantities of entropy was defined as thermodynamic
efficiency '1 by Odum and Pinkerton [5]:
~

c L"X:; (l2)
TJ s L-- X;
II I
(Index S refers to entropy representation; superscript C refers to coupling.) It
may be remarked that this kind of efficiency is equal to the square of the coupling
factor defined by Kedem and Caplan [6], i.e. using definition (12) and Eq.(8) we
obtain:
 283

L22 (L 12 +L21 )2 X l2 (L 12 +L21 )2 p2

(13)
LII X~ 4L;2 4LII L22
This ratio can be called the entropy representation of the coupling
thermodynamic efficiency (CTE) and serves as a measure of coupling between
heat and component transports or a measure of the utilization of the entropy
produced. The efficiency reflects the essence of the process, which is the
coupling, but does not take the other entropy producing effects into account.
Since the measure of coupling is expressed by the ratio of utilized and produced
entropy, this CTE value is defined here in entropy representation. Later on,
energy and exergy representation ofTE will also be distinguished.

4. Illustrative Separation Processes

Some of the most common separation processes can be interpreted as first order
steady systems. This interpretation is rather plausible for the processes where
heat is used as the separating agent (e.g. distillation), but the comprehension is
somewhat more complicated for the processes when the separating agent is a
stream of mass (e.g. sorption and extraction). The later ones should be
considered with the recovery systems of the separating agent together. The
energy supplied to a separation process may be in the form of thermal,
mechanical, electrical or chemical energy, depending upon the type of process.

4.1. DISTILLATION

An ordinary distillation column is a good example of a separation process driven

by heat input. From an energy point of view our goal is to utilize the entropy
produced by heat transport as completely as possible by the entropy consumption
of the component separation.
The minimum isothermal "work of separation" (Wmin,r) represents a lower
bound on energy that must be consumed by a separation process [7]. For a
multicomponent liquid mixture being separated isothermally into two products
we have:

Wmin.r=-RTo(DL xwlnYwxiD+WI;xiw InYixiw -FI; XiFlnYiXiF) (14)

I I I

which corresponds to the utilized entropy consumption of the compatible

independent component flux in the numerator of the CTE:
~2X; =-R(DL I
xiDlnYwxiD+WI;xiw InYixiF -FI; XiFlnyixiF )
I I
(15)
284

(R is the gas constant; T is temperature; index 0 refers to ambient state; F, D.

and Ware feed, distillate, and weight product flow rates, respectively; these letter
as indices refer to these streams, x is component mole fraction, r is activity
coefficient. )
The "net work consumption" w,. of the process is defined as the difference
between the work that could have been obtained from the heat Q entering the
system and the work that could be obtained from the heat leaving the system. If
no mechanical \vork is involved and the enthalpy difference between products and
feed is negligible compared to the heat input, the net work consumption is:

Wn=QTo( )D - ~r ) (16)

which corresponds to the entropy production of the compatible independent heat

flux in the denominator of the CTE:

LllX~=Q(_1__
1 ) (17)
TD Tw
Thus the expression of CTE may be given as:

q; -R( Df xwlny,xw+Wfx"lny,x,,-Ff X"lnY,x,,)

(18)
Q(_l__1)
TD Tw
when distillation is treated as a first order steady system.

4.2. ABSORPTION AND STRIPPING

In simple gas absorption, a solvent is used to separate a soluble from an

insoluble gas by passing the mixture through a multistage contractor
countercurrent to a solvent. Accordingly, an absorption column is a typical
example of a separation process, where a stream of mass is used as the
separating agent instead of heat [8]. In such cases, however, the stripping, the
recovery of the separating agent, should also be considered as part of the system.
Stripping is the inverse of absorption and involves the removal of a volatile
component from a liquid into a gas phase. The energy supplied to the stripper
may be in the form of thermal or mechanical energy, but generally heat is used.
Consequently, the absorber-stripper combination can also be considered an
example of a separation process driven by heat input.
 285

The minimum isothennal work of separation and the net \vork consumption
of the process can be defined and interpreted identically with distillation. Thus
the coupling thennodynamic efficiency can be defined as follows:

-R(GI x iG InYiG +VI XiV InYixw -FI X iF Inyix iF )

T/: I I I (19)

Q()r )B)
when the absorption-stripper combination is considered as a first order steady
system.

4.3. EXTRACTION AND SOLVENT REMOVAL

Liquid-liquid extraction also employs liquid separating agent or solvent (s), but
this is used to separate a liquid instead of a gaseous mixture. Solvent extraction
processes may be classified as broadly single and double solvent operations,
according to the number of solvents used to effect the separation. In the case of
liquid-liquid extraction, such as in gas absorption, the first order steady system
includes the solvent removal units, in addition to the extractor. In solvent removal
units, heat is generally used to recover the solvents.
The minimum isothennal work of separation is the same as previously, but
the net work consumption is the sum of the work consumed in two solvent
removal units. Accordingly the expression of CTE may be fonnulated as (A and
B are the extructs):

R(A~ XiA InYixiA +B~ XiB InYixjB -F~ XiFlnYjXjF)

T/: (20)

when the extraction-solvent removal combination is treated as a first order steady

system.

4.4. OTHER SEPARATIONS

The previous two examples are suitable to demonstrate that the separating agent
perfonns a function similar to that of heat in the case of distillation and the
recovery of the separating agent requires energy supplied to the system. Certain
further separation processes employing a liquid separating agent, such as solid-
liquid extraction or extractive and azeotropic distillation, and other ones
employing solid separating agent, such as gas adsorption and ion exchange. can
286

all be interpreted as first order steady systems in a thermodynamic sense. In

certain processes, however, the energy required by the removal units of the
separating agents should be supplied to the process in other forms than heat. E.g.
the production of reflux in the ion exchange process is effected by means of
chemical energy or in vapor recompression distillation a heat pump is used,
requiring mechanical energy, to transfer the heat removed from the condenser to
the reboiler by raising its temperature level. Mechanical energy is also used in the
case of heatless adsorption, with regeneration accomplished by means of frequent
lowering of the pressure on the adsorbent beds.
Although the calculation of the minimum isothermal work of separation is
similar in all cases, the calculation of the net work consumption should be
replaced with appropriate formulas ,,,,hen using other fonns of energy than heat.

5. Heat Losses and Heat Degradation of Uncoupled Process

In any uncoupled stationary process without losses, the energy added to the
system is equal to the energy withdra\\TI from the system. In the case of any kind
of losses, the efficiency of energy conservation in a system can be measured by:

VE=L E out (21)

L E in
,,,,here L E in is the sum of energy added to, and L E out is the sum of energy
withdrawn from the system. This kind of efficiency is defined on the First Law
basis and may be called First Law ejjiciency.
However, this definition of efficiency does not take into account the
temperature at which heat is added to or withdrawn from the system, so the
efficiency on the Second Law basis has to be defined additionally. The maximum
portion of heat, which can be converted into other kinds of energy, can be
determined by the Carnot efficiency:

(22)

when heat enters the system at a temperature T2 and leaves at a temperature TI .

Generally the available energy or exergy B of an amount of heat Q at a
temperature T is limited by:

B=( 1- ~o )Q (23)
 287

where To is the ambient temperature and heat losses are assumed to be zero. The
change in available energy of a system without heat losses is a measure of the
heat degradation. The exergy conservation of a system can be expressed by:
LBolI1
v B= (24)
LB in

where L B in is the sum of exergy added to, L BOllI is the sum of exergy
withdra\\TI from the system. This kind of efficiency is defined on the Second Law
basis and may be called the Second Law efficiency.
In the case of real stationary heat transportation processes, the heat losses
and heat degradation occur simultaneously, thus the definition (24) can serve as a
combination of the First and Second Law efficiencies by measuring the overall
exergy changes of the streams through the system. It should be pointed out that
this definition of efficiency cannot be applied to coupled processes where the
exergy conservation is not an indicator of effectiveness.

6. Overall Thermodynamic Efficiency of Coupled Processes

In the case of thermal separation processes, in addition to coupling the other

entropy producing effects should be considered. The goal of these coupled
processes is to utilize the entropy produced as completely as possible by the
entropy consumption of the separation. This is why the entropy consumption of
the separation should be retained in the numerator of the overall thermodynamic
efficiency for coupled processes. But in this case, beside the main entropy
production caused by heat transport, the other entropy producing effects, such as
hydraulic resistance and complementary heat exchanges of the system should
also be added to the denominator of the overall efficiency:

(25)

where L22 Xg is the entropy consumption of the compatible independent

component separation LII X l2 is the entropy production of the compatible
independent heat transport, and (LiS)L is the additional entropy production caused
by hydraulic resistances, and supplementary heat exchanges, etc.
In addition to the entropy representation above, exergy representation of
OTE can be defined when the exergy increased by separation is compared to the
sum of the exergy decrease:
288

(26)

where BSEP is the exergy increase of the separation BH is the exergy decrease of
heat transport, BL is the sum of exergy decrease of the other entropy producing
effects. This definition of OTE can be rewritten by using the exergy terms of the
Eq. (24), i.e. the exergy changes of the output and input streams.
B SEP
(27)

The term BSEP is added to the denominator, because the total exergy expense
should be considered here and the difference 6:
Bin - LBo /I t) contains only
the resulted exergy decrease reduced by the exergy increase of the separations.

7. The Energy Representation of the Thermodynamic Efficiency

The defmition of the energy representation of the thermod}namic efficiency for

separation processes is somewhat more complicated than entropy or exergy
representation. In the case of an ideal mixture, namely when no energy change
exists, but entropy (or exergy) consumption and production does exist. The
energy representation fthe thermodynamic efficiency can be defined, ho\vever, by
the thermal efficiency of a corresponding heat engine with the same temperature
levels and coupling efficiency.
The thermal efficiency of a heat engine receiving heat Q2 at temperature T2
releasing heat Q, at temperature T, and produced shaft work:
(28)
is defined by:
W W
'h:=-=--- (29)
Q2 Q,+W
The shaft work can be obtained from (29):
W=O 3.L . (30)
-, I -11£
This work is originated from heat at temperature TJ, so the consumed entropy is:

S =W =0 (_I)3L
cons T2 -, T2 l-I]E .
(31)

The entropy production of heat transfer Q, from T2 to T, can be expressed by:

 289

S -0
prod - - j
(~ __
I )
T 7~ ,
(32)
j

consequently the coupling efficiency is:

c Scans '1 E ( 1 ) I (33)

'ls=Sprod =1-'1£ T2 (T j - 1 -T 2- j )

where 'lCI is the Carnot efficiency (see Eg. 22). The energy representation of TE
from Eq.(33) can be written as:
C
'lS'lCI
(34)

so the energy representation of TE can be calculated from the coupling

efficiency, Carnot efficiency and temperatures. Eq. (34) enables us to compare
the effectiveness of a separation process and a heat engine.
Finally it may be remarked that the coupling efficiency theoretically, in
extreme reversible case, can be equal to unity:
n C .
·/S,m<l.,\ -
-1 (35)

but the energy representation of the efficiency IS limited by the value of the
Carnot efficiency:

'1 C 1
'1 £.ma~ = T I IT 2 +'7CI '7 C I . (36)

8. Systematization of the Various Definitions

The systematized tabulation of the various definitions of thermodynamic

efficiency is shown in Table I. Two kinds of processes can be distinguished:
uncoupled and coupled ones. From a thermodynamic point of view, the goal of
uncoupled processes is the energy, or more generally, exergy conservation, while
in case of coupled processes, like separation, the grmvth of relative exergy
production or that of the entropy consumption of the forced process (here
separation) can be regarded as the object.
To drive this forced process, exergy consumption or entropy production is
required and the thelmodynamic efficiency must reflect the ratio of the
appropriate thermod)namic properties (entropy or exergy) of the forced and
driving processes. The definitions of the thermod)namic efficiency for coupled
processes can be divided further into coupling and overall efficiencies according
290

to the system investigated. The former expresses the effectiveness of the coupling
alone, the latter, in addition to coupling, also takes the supplementary entrop)'
producing effects into account. On the other hand. three different representations
of TE are distinguished, i.e. energy, entropy and exergy representations, which
can indicate the effectiveness of the same process from different points of view.
For uncoupled processes, the energy representation of TE reflects the
energy conservation [Eq. (21], the exergy representation reflects the conservation
of available energy [Eq. (24)]. The fomler is a measure of heat losses alone, the
latter takes both the heat degradation and losses into account. The entropy
representation of TE for uncoupled processes cannot be defined because the loss
of entropy is meaningless.
The fundamental definition of TE for coupled processes is based on entropy
consideration. The coupling thermodynamic efficiency expresses the effectiveness
of the coupling alone (Eq. 12, 13), while the supplementary entropy producing
effects are also considered by the overall thermod~namic efficiency (Eq. 25).

TABLE I. Definitions of thennodmamic efiiciency

uncoupled
TE type: Coupled processes
processes

Represent- Coupling TE Overall TE

ation (CTE) ,F (OTE) ,.,0

C SSEP SSEP
Entropy (S) - I]s =-s'- 1]f
, H SH +SL

1] 0 - B SEP
I BOllt B -
VB= C B SEP BH +BL
Exergy (8) I Bin '7B =--- B SEP
BH
(I Bin -I Bout)+BSEP

I E out C C
°- ° °
= 1]C- I]s1]o 1] I] S I] Cl
Energy (E)
VE
I E in E-TC /Tc
I
C C
2 +1]s I]Ct
E-TO/To
I
° °
2 +1]s 1]0
 291

The exergy representation ofthe overall TE is defined by Eq. 26 and 27 and

the coupling TE can be defined similarly (neglecting BL from Eq. 26):
C B sEP
178=-- (37)
BH
where BSEP is the exergy increase of the separation, BH is the exergy decrease of
the heat transport. It should be remarked that the value of the TE in exergy
representation is the same as the value of entropy representation, but the former
is usually more useful for practical investigations.
The energy representation of the TE is interpreted by the thermal efficiency
of the corresponding heat engine with the same coupling efficiency and
temperatures (Eq. 34). The coupling and overall definitions differ only in the
temperature levels and the Carnot efficiency. Accordingly the thermal efficiency
can be regarded as the energy representation of TE of a heat engine, and by using
Eq. (33), its value can be converted into entropy (or exergy) representation.

9. Empirical Definitions

A particular advantage often claimed for exergy is that it allows the definition of
efficiencies in Second Law terms. This topic has been discussed extensively in
German literature, especially in the work of Fratzscher [9], when he defined two
alternative empirical definitions in addition to the fundamental indicator of
exergy conservation for uncoupled processes (Eq. 24). The first definition of
efficiency is given by:

C; I = (I B oul I B in ) purpose
-
(38)
(I B,n - I B oul )driVing
Here the indices purpuse and driving refer to the (exergy) increments or
decrements happen in the purpose and the driving media. If all the purpose
streams are in reference state and the heat of mixing can be neglected we can
write:

(I Boul - I Brn ) purpose =BSEP . (39)

In this particular case, definition (38) is identical with Eq. (27). Substituting
relation (39) into definition (38) it follows namely that:
 292

B SEP
=T-----------~------~~~----------~~----
I Bin - I Bout driving +B SEP -(I Bout - L: Bin) purpose (40)

This proof also shows that our definition of OTE [Eq. (27)] is in accordance with
the so called "exergetic efficiency of distillation" interpreted by Linnhoff and
Smith [1].
The second empirical definition of Fratzscher:

(;1 (I Bout-I BiJpurpose (41)

-. IBm

as well as its modification defined by Linnhoff and Smith:

~ = B SEP (42)
':>3 "B
~ In
'

can be regarded as approximations of OTE constructed by neglecting the exergy

values of the output streams in the denominator. However, definitions (41) and
(42) cannot be proposed for indicating the thermodynamic efficiency of
separation processes, because the exergy of purpose streams is included in the
term I Bin ,in spite of the fact that it is not consumed by the process.

10. Localization of Entropy Production

Before categorizing the energy saving techniques let us analyze the origins of the
entropy production (Fig. I). According to the reversible model concept the
minimum isothermal "work of separation" represents a lower bound on energy
that must be consumed by a separation process [10]. The corresponding net
entropy consumption is marked as AB. Additional entropy is consumed by the
irreversible mixing effects (e.g. non-equilibrium feed conditions) inside the
column (Be). Thus instead of the net requirement a greater overall entropy
consumption (AC) must be compensated by the entropy producing subprocesses
(heat transport).
 293

Because of finite driving forces (adiabatic distillation) anel hydraulic

resistance the internal entropy production is greater than the entropy
consumption to be compensated (CD). In addition to the internal requirements
(AD), external entropy production also occur (DE) caused by heat exchangers
and imperfect heat conservation.

EXTERNAL

D ---------- D

FINITE DRIVING
FORCES +
HIDRALIC
RESISTANCES INTERNAL
JI___________ _
C --------------- C ----------C
IRREVERSIBLE TO BE
MIXING COMPENSED
B NETTO--CON5UMrlON- B
----------8
OF SEPARATION A ~~-----A ..J:...::.a...J._ _ _ _ _ _ _ __
-A
EN·TROPY INERNAL OVERALL
CONSUMPTION ENTROPY ENTROPY
PRODUCTION PRODUC TION

Figl/re 1. Origins of entropy production

11. Energy Saving Policies

The energy saving concepts can be categorized into three basic policies according
to the deciding origins of the entropy production.

The methods of the first group aim to reduce the irreversible mixing effects
(BC) inside the column, e.g.:

• energy conservation by means of tray retrofit (optimization of the changes in

heat duty associated with changes in number of trays)
• optimal feed location
 294

• energy conservation by means of tray retrofit (optimization of the changes in

heat duty associated with changes in number of trays)
• optimal feed location
• pressure reduction to decrease heat duty
• thermal coupling between two towers
• application of several feeds and seyeral sidestreams
• use of divided columns
• use of preflash to\ver.

The techniques of the second group strive after less irreversible heat turnover or
more uniform driving force distribution to reduce the internal entropy production
(CD), e.g.:

• use of interreboilers and intercondensers

• use of internlediate vapor recompression
• non-adiabatic distillation with special countercurrent devices.

The methods of the third group aim to improve energy conservation and reduce
external entropy production (DE), e.g.:

• use of energy integration concept by multiple-effect distillation, (heat

cascading the columns by using the condensation of the overhead vapors
from the second column to provide boilup for the first column)
• optimization of feed preheat networks to capture pump around heat or
product heat wasted to air/water
• direct feed to d0\\11stream units
• use of direct or indirect vapor recompression.

It is to be noted that certain methods are multifunctional. E.g. some of the first
class methods, like pressure reduction, has a positive effect, the thermal coupling
behveen two towers has an adverse effect from the second point of view.
Similarly the second class intermediate vapor recompression method has a
positive effect from the third point of view etc.

12. Effective Concepts for Energy Efficient Process Design

In this section some chemical engineering techniques are discussed which are
being applied to distillation sequences of existing plants in refineries and light-
 295

hydrocarbon plants to reduce energy consumption, while maintaining high levels

of flexibility and operability [12-14].
Based on our theoretical and economic considerations, computer aided
design and plant measurements, the superiority of generating interreboilers and
intercondenders as well as the heat integration and heat pumping concept can
be stated among the energy conservation methods. In the industrial distillation
plants the largest internal irreversibilities occur due to the nongradual addition
and withdrawal of heat and the largest external losses occur due to insufficient
energy conservation.
By applying interreboilers and intercondensers the gain in reversibility is no
attested as a reduced total heat duty but instead as a lesser degradation of the
heat energy passing through the column. In a single reboiler system, the entire
heat load must be applied to the base of the column at a high (most expensive)
temperature. When the same amount of energy is divided up and added to several
intermediate points the temperature levels are progressively lower (cheaper) as
the feed tray is approached. The same concepts can be applied to the rectification
section, \vhere heat can be removed at successively warmer heat sinks (cheaper)
as the feed tray is approached [10].
A new project is underway which \vill develop and test a design for the
optimal thermal control of a distillation process. The control in question
minimizes the total costs and is implemented by thermostatted trays maintained
by heat pumps or Brayton refrigeration cycles. The main objective of the project
is to demonstrate experimentally the potential energy savings in a cost effective
design for a distillation column based on the concept of equal thermodynamic
distance [I5, 16].
According to our experience the energy integration concept and generally
the methods of the third policy offers the highest affectivity to reduce energy
consumption. By applying the multiple effect distillation, rearrangement of feed
preheat networks and direct feed to dO\\TIstream units can yield 20 to 60 %
energy saving in some distillation systems.
This lecture provides also orientation for process engineers on the current
application of heat integration and heat pumping techniques for large
industrial separation plants. Certain well designed heat pump options can show
better economic figures than the base case separation system which uses only
steam and cooling water. Thus, in order to realize enhanced integration at the
process design the consideration of the different heat pumping techniques is
highly recommended. The presented results show that
(I) larger heat load and smaller process temperature difference provide
shorter pay back time for heat pumping,
 296

(2) the absorption heat transfonnation cycles have even chance for
implementation at larger process ~T, when the other configurations are
to be discarded [17].

References
1. Linnhoff, B. and Smith, R. (1972) The thennodynamic efficiency of distillation, l. Chem. E.
Symposium Series 56, 2.1147-2.1/73.
2. Glansdorf, P. and Prigogine, 1.(1954) Sur les proprietes differentielles de la production
d'entropie, Phisica 20,773-780.
3. De Groot, S.R. and Mazur, P.(1962) Non-Equilibrium Thermodynamics, North-Holland
Publ. Co., Amsterdam.
4. Szolcsanyi, P.(1969) Uber die Entropie-Wirkungsgrade von Verfahrenseinheiten,
Wissenschaftliche Zeitschrift der TH fur Chemie "Carl Schorlemmer", Leune-Merseburg
11, 351-354.
5. Odum, HT. and Pinkerton, RC.(l955), American Scientist, 43, 331. Referred to in [6].
6. Kedem, O. and Caplan, S.R (1965) Degree of coupling and its relation to efficiency of
energy conversion, Trans. Faraday Soc. 61, 1897-1911.
7. King, C.J.(l971) Separation Processes, McGraw-Hill Book Co, New York.
8. Pratt, H.RC.(1967) Countercurrent Separation Processes, Elsevier Publ. Co., Amsterdam.
9. Fratzscher, W.(1961) Zum Begriffdes exergetischen Wirkungsgrads, BrennstojJ-Waemle-
Kraft 13,486-493.
10. Fonyo, Z. (1974) Thennodynamic Analysis of Rectification I and II. Reversible Model of
Rectification, Finite Cascade Models, Int, Chemical Engng. 14, (1) 18-27, and (2) 203-210.
11. Fonyo, Z. and Foldes, P. (1974) Optimization of a stepwise realization of heat turnover in
distillation, Acta Chimica Acad. Sci. Hung. 81, 103-126.
12. Fonyo, Z. and Rev, E. (1982) General Interpretation of the Thennodynamic Efficiency for
Separation Process, Hung. J. Industrial ChemIstry 10, 89-106.
13. Mizsey, P. and Fonyo, Z. (1990) Toward a More Realistic Overall Process Synthesis: The
Combined Approach, Compo andChem. Engng. 14.1213-1236.
14. Rev, E. and Fonyo, Z. (1991) Diverse Pinch Concept for Heat Exchange Network
Synthesis: The Case of Different Heat Transfer Conditions, Chem. Eng. Sci. 46 1623-1634.
15. Salamon, P. and Berry, R.S. (1983) Thennodynamic Length and Dissipated Availability,
Physical Review Letters, 51 1127-1132.
16. Salamon, P. Nulton, J. and Ihrig, E. (1984) On the Relation between Energy and Entropy
Versions of Thennodynamic Length, Journal of Chemical Physics 80, 436-437.
17. Fonyo, Z. and Benko, N. (1998) Comparison of Various Heat Pump Assisted Distillation
Configurations, Chemical Engineering Research and Design, 76, 348-360.
SECOND LAW BASED OPTIMIZATION OF SYSTEMS WITH
THERMOMECHANICAL DISSIPATIVE PROCESSES

E.MAMUT
Ovidius University of Constanta
Department of Mechanical Engineering
124, Mamaia Avenue
8700 Constanta, ROMANIA

1. Dissipative Processes in Mechanical Systems

A mechanical system is designed to perform static, kinematic or dynamic functions.

The evolution of a system that involves the change of the value of at least one
mechanical parameter is called mechanical process. Every mechanical process
involves an exchange of mechanical energy. Because of friction and damping,
energy is dissipated in all real mechanical processes. Sometimes the dissipated
part of energy is insignificant, and sometimes it can be very important, as in
damping systems.
Having as function the damping of mechanical vibrations, the damping system
is a good candidate for the study of the influence of thermomechanical processes on
dissipative systems. Usually called vibration isolation systems (VIS), these
systems are divided into active (AVIS) and passive (PVIS) systems. An AVIS is a
complicated system connecting mechanical and electronic sub-systems in so-called
mecatronic systems. Passive vibration isolation systems (also called mounting
systems) are common, and it is difficult to find an industrial or domestic machine,
equipment or installation that is not mounted on, or contains a mounting system.
More than 60% of all mounting systems contain rubber mounts. The most usual
rubber mount consists of two steel plates bounded by a bulk mass of vibration
isolation rubber. There is a wide diversity of shapes and forms of rubber mounts.
The problem that has stimulated the present study was the demand for the
improvement of the reliability of rubber mounts. The aim of the study was to
analyze the physical aspects of the failure process. Experimental work has been
performed at ICECON research institute in Bucharest. Finite-elements (FEM)
analysis has been done at Arcelik R&D Center in Istanbul, with the collaboration of
Bosforus University and Istanbul Technical University.

2. Influence of Dissipative Processes on the Behavior of Rubber Mounts

The behavior of rubber mounts has been studied on an experimental stand, which
allows quasistatic and dynamic loadings. The quasistatic loading is a deformation
based on loading with a speed of 5 mm/minute and incremental steps of 1 mm. The
related values of the load are indicated on a dynamometer. In order to plot the
dependence between load and deformation, the maximum deformation of the mount
should be 15% of its height. The stand has been designed to work with sets of 8
297
A. Bejan and E. Mamut (eds.), Thermodynamic Optimization o/Complex Energy Systems. 297-312.
© 1999 Kluwer Academic Publishers.
298

mounts during each testing cycle. The dynamic loading of the mounts is done using
an unidirectional variable load with 30,000 N amplitude, and 50 Hz frequency.
The experimental research carried out at ICECON under the guidance of Prof.
Polidor Bratu has confirmed the results obtained by other authors, e.g., Poturaev,
Dirda and Krush [1, 2]. When the tests were started, the first observation was the
heating undergone by the rubber mounts. For this reason the stand was equipped
with a temperature measurement system. Continuing the experimental tests, it
became obvious that there had to be a relation between the heating process and the
failure mechanism of rubber mounts. Comparing the results obtained with different
types of rubber (e.g., natural rubber, styrene - butadiene, isoprene rubber) we can
identify three types of failure mechanisms of rubber mounts differentiated by the
dependence of internal damages on local temperature [1, 2]:
Thermomechanic Failure Mechanism-characterized by a nonsteady regime of
temperatures with the tendency of continuous temperature increase until the
complete failure of the rubber mount; usual for vibration insulation materials based
on natural rubbers, highly crosslinked, and with low contents of zinc oxide.
Fatigue Failure Mechanism-characterized by a steady regime of temperatures,
and damage being produced by a fatigue development mechanisms of cracks
starting from microscopic structure defaults; could be observed in natural rubber
based mixtures with high contents of zinc oxide.
Combined Failure Mechanism-the most common, characterized by a complex
influence of temperature on crack development kinetics. Low crosslinked rubber
mixtures with low contents of zinc oxide.
In conclusion, we consider the failure process of rubber mounts as a complex
process coupling a thermal component, a micro and macrostructure mechanical
component, and also some physical and chemical components. In the present study
we take into account only the coupled thermal and mechanical components of the
process.
The experimental observations show that during the damping process the rubber
mount failure starts with a heating process. The distribution of temperature is
usually as shown in Fig. 1. The peak temperature could be stabilized after a
number of cycles, or could continue to rise. If the temperature is stabilized at low
values (e.g., 40 - 50"C), the failure process will be mainly a fatigue crack

-
&;
I-~_ _"O
i

o
&;

Figure J. Temperature distribution in a rubber mount.

 299

.160
~
~
~
f20 /'
I
80

I' 7 . . . . . ~[
40

o 8 '2 t, hours

Figure 2. Evolution of temperature and damage in a rubber mount.

occurrence and development type process. Consequently, when the temperature

field is steady, the damping process will cause an accumulation of micro-faults
until the occurrence of cracks: after that it will continue with a micro-mechanical
crack development.
If the temperature rises to high values (e.g., 120 - 160·C), the failure process
will be caused by physical and chemical reactions characterized by mixture
softening, plastic deformation or rubber-metal bond damage. A generic evolution of
temperature and damage in a rubber mount is shown in Fig. 2.
After this initial stage of experimental research the study continued with the
analysis of rubber properties, theoretical modeling and the development of a
comprehensive theory to explain the failure process of rubber mounts. It was clear
that the complexity of the phenomenon could not allow only a mechanical
approach followed by an after thought of thermodynamic considerations. As noted
by Truesdell [3], it was necessary to develop a new approach.

3. Rubber Properties

The notion "rubber" covers a group of materials, which can be identified by the
property of large deformations and almost complete and instantaneous recovery in
the absence of deformation load. The large elasticity property (also called
hyperelasticity [4]) is not a consequence of a special chemical composition, but a
special disposal of the molecular structure components. The concept of "rubber-
like" material, covers a set of materials called polymers, which contain long
macromolecular chains obtained from repeated units of atoms called monomers.
For this reason, rubber-like materials are also called elastomers. The character-
istics of an elastomer are [5]:
the macromolecular chains have to be long and capable of free rotation
around the links between the neighboring molecular units;
300

the macromolecules have to be liked in a number of positions, in order to

constitute three-dimensional networks, either by chemical links or by
mechanical crosslinks;
even with the above mentioned links, the molecules have to be capable of
executing with ease relative motion: in other words, the intermolecular
attractions have to be weak.
From the mechanical energy dissipation point of view, the main
thermomechanical properties of elastomers are [6]: creep, stress relaxation,
mechanical hysteresis, heat transfer properties, and temperature dependence.
Creep After Sudden Stress-the property of a material to respond to a sudden
stress by a time dependent deformation. Figure 3 shows the time profile of a shear
creep experiment [6].
Stress Relaxation After Sudden Strain-the property of a material to respond to a
sudden strain, which is subsequently maintained constant by a relaxation in time of
the resistive stress. Figure 4 illustrates a generic geometry and time profile of a
simple shear stress relaxation experiment following sudden strain [6].
Mechanical Hysteresis-a consequence of the above mentioned properties,
which characterize nonlinear materials. Suppose that we strain very slowly a
sample of material and then we recover, with the same speed, the initial shape of
the sample. If the strain-stress curve in the loading process is different than the
strain-stress curve in the unloading process, we obtain a static hysteresis loop
(Fig. Sa). If the strain is harmonic, the shape of the loop will be as in Fig. Sb, and
is called dynamic hysteresis loop. From the mechanical energy dissipation point of
view, the hysteresis loop measures the energy dissipated per cycle.
Heat Transfer Properties-the most significant properties for vulcanizing bulky
elastomers are [7]: thermal conductivity, specific heat at constant pressure and

0"21 ,
I
I
I
I
I
I
0
to Time

o ~--~/~.-----------------------------
to Time

Figure 3. Time profile of a shear creep experiment.

 301

o~~~----------- _____
to

Figure 4. Geometry and time profile of a simple shear stress relaxation experiment following sudden strain.

coefficient of thermal expansion. The thermal conductivity takes values between

0.21 W/mK and 0.3 W/mK. The specific heat at constant pressure takes values
between 1.4 kJ/kgK and 1.8 kJ/kgK. The coefficient of thermal expansion is of the
order of 5 x 10-4 K- 1• The specific heat depends on temperature [7].
Temperature Dependence of the Properties of Elastomers-as in different other
materials, between the properties of elastomers there are many and complex
relations. For the purpose of the present study, we shall take into account only the
dependence of shear modulus on temperature.

4. Finite Element Formulation

A theoretical approach of an engineering problem starts with the modeling phase.

Starting from the purpose of the study, several hypotheses are made. The model is
conceived as a system of equations and constraints that result from the studied
problem.

4.1. CONSTITUTIVE EQUATIONS FOR RUBBER-LIKE THERMO-RHEOLOGIC

MATERIALS

The modeling of the thermomechanical processes from metal-rubber vibration

302

g'"
CIl

Strain Strain
a) b)

Figure 5. (a) Static hysteresis loop; (b) Dynamic hysteresis loop.

insulators can be developed by two methods, or by a combination of the two. The

first method is the statistical or molecular method which starts with a hypothesis
about the molecular chains, and describes their behavior based on statistical laws.
The second method is the phenomenologic or macroscopic method, which describes
the influence of macroscopic parameters on thermoviscoelastic continua. Both
methods set up an equation that describes the dependence between the stress
tensor, characteristic parameters of the studied continuum thermal field, strain
tensor and time (if there exists a time dependence). This equation is called
constitutive equation.
The theory of continua introduces a number of axioms that have to be applied
to constitutive equations [8]. A number of constitutive equations have been
established based on the statistical method, and are known by the names of their
authors. The most used in the literature are Flory and Rehner (1943), Treloar
(1946), Wang and Guth (1952) and Flory and Erman (1982). Based on the
phenomenological method are the relations of Mooney (1940), Rivlin (1948),
Valanis and Landel (1967) and Ogden (1972). These equations have been
determined at thermodynamic equilibrium or for quasistatic transformations, based
on the principles of mechanics and sometimes of thermodynamics. For nonsteady
or time-dependent processes, some empirical corrections are usually made.
Given the Mooney-Rivlin equations

(1)

where C IO and CO) are parameters of the constitutive equation, the terms I) and 12
have the following form

(2)

In these expressions /;)' /;2 and /;3 are the principal specific deformations. For
example, in the case of simple strain, we have the following forms
 303

(3)

where I; is the specific deformation measured in the strain direction. Using the
relations (3) we obtain the stress expression for simple strain

(4)

and using the reduced form for stress, we have

[f*] = cr / (I; - 1;-2) = 2C lO + 2C OI 1;-1 (5)

Taking into account only the equilibrium states, the temperature dependence is
linear [9].
The behavior of several other elastomers is similar. For quasistatic processes
the temperature dependence of rubber-like materials is related to CIO.~ and COI,~
as follows

(6)

where KI and K2 are constant parameters depending on the type of material.

In order to complete the model we have to add to the obtained equations the
laws that describe the evolution of the process, and the appropriate mechanics and
thermodynamics principles. In the end we obtain a system of partial differential
equations, which has to be solved using numerical methods.

4.2. THE FINITE ELEMENT METHOD

Between several numerical methods, the Finite Element Method (FEM) proved
that it can offer good approximation of the results. In variational formulation [10],
the method consists of the maximization or minimization in a given domain of a
scalar quantity depending on one or more unknown functions. This quantity is
named functional. The problem is stated as follows: determine the approximation
of a function such that the functional

II = J
A
f({U}, :x {U}, ... )ctA + Jr
g({u}, :x {u}, ... )dT, (7)

defined in the domain A and on its boundary r, reaches an extremum. In order to

obtain the extreme approximate value of the functional II, the domain A is divided
into sub-domains of finite dimensions, which are called elements. The contribution
of each element in setting up the functional is estimated based on a simplified
model of the unknown function u. The approximate value of the functional is given
by the relation

II = I,II(e) (8)

where the index (e) refers to an undefined element. The unknown function u is
304

expressed approximately for each element by a relation of type

{u}(e) = [N]{V}(e) (9)

where {V} (e) is a column vector containing the values of the function in the
characteristic points of the element, and [N] is the shape function that depends on
the finite element type and on the coordinates of the internal point of the element
where the value of the function u is calculated. The functional rr depends on the

r
all parameters {V} of the domain A, and the extreme condition of the functional
has the following form of the system of equations:

an I a{U} =
g;gg; } (10)

In this way one arrives at the construction rule for the general system of equations
that result from each element,
arr arr
a{V} = L a{V}(e) = 0 (11 )

The solution to the system (II) consists of the values of the parameters V of all the
finite elements of domain A, which offer the possibility of calculating an
approximate value of function u. The numerical method for this solution depends
on the shape of expression (9).
The problem of the thermomechanical processes in rubber mounts is a second
order problem of thermoviscoelasticity. To solve this problem and to determine the
stress, strain and temperature distribution, in the first phase, we start with the
uncoupled problem and finally we take into account the couplings.

4.3. FORMVLATION OF THE MECHANICAL PROBLEM

There is an extensive literature on the modeling of rubber-like materials. For the

purpose of the present study we shall present some considerations, following the
demonstrations of Ogden [11], and Jankovich et al. [12]. Given the displacement
vector 0 of a point in a given body [11] in RN in a deformation process which has as
components Uj (i = 1, N; N = 2, 3) and Vo = b jj + Ujj displacement gradients.
Furthermore b ij are Kronecker's coefficient, and Ujj = aU i I ax). The notation Xj
indicates the local coordinates of the deformed body. The deformation tensor
Green-Lagrange is defined by the following relations

(12)

tij is divided into linear and nonlinear components [12]:

eij = (u,j + u,d/2 (13)

 305

where £ij = 0.5 (Uj + Ui) is the linear component, and llij is the nonlinear
component. The stress tensor is defined in the sense of Piola Kirchhoff,

where W = W(IJ, 12 ) is the deformation energy function.

We are looking for a field of displacements u which minimizes the total
potential energy functional,

ll(u) = f W(ll' 12 )dX - g(u) (15)

n
The field of displacements has to satisfy the following incompatibility constraints

det Vu = I~/2 = I (16)

or 13 = 1. Given a function f(.) continuous such that f(l) = 0; f(X) ::1= 0 Vx ::1= 1, we
can write

(17)

In this way we can have for the deformation energy function the following form:

(18)

where p is a Lagrange multiplier. The problem is reduced to determining the

stationary point (u, p) of the potential energy functional,

ll(u,p) = f W(u,p)dX - g(u) (19)

n
The associated variational principle is

oW ex = f Sijo£ijdQ 0 = f f(I3)Op (20, 21)

n n
where oW ex is the virtual work of the external forces and Sij is the stress tensor,

(22)

Equation (20) is a nonlinear functional equation, and is solved by the Newton-

Ralphson method. For external conservative forces one derives the second-order
variation of the deformation energy functional, by taking into account a given
displacement u O and the pO, £0 and SO values for the Lagrange multiplier,
deformation and related stress. Fixing £ij = £ij + L1£ij' and considering in Eq. (20)
the Taylor series limited to the second-order terms, we obtain finally:

ow ex = f·
n
S IJ OL1£··dQ
IJ
+ f··C
n
&
rs
1Jrs 0L1£··dQ
IJ
+ f
n
af(I3)
L1p--M£
a£.. .. dQ
IJ
IJ
(23)
306

(24)

where

c ijrs = 02W + pO o2f(I3) (25)

OCijOC rs OCijOC rs

We consider the case of a bilinear isoparametric element for displacement, and

define the following matrices

~u TKNL~U = fSij~llijdV (26, 27)

vE
We also define the vectors

(28)

(29)

For elastomers we define

(30)

Given the load vector at element level, g, the relations (23) and (24) take the form

(31 )

Following [12], a perturbation of order C is introduced to obtain

(32)

where M is the matrix of volumes. Finally, the relation (32) becomes

(33a)

(33b)
 307

In the modeling of viscoelastic damage [24] the deformation energy nas the form

W(t) = W~ + I,Wnexp (-tn"n) (18')

n=t
where W~ is the deformation energy in the static process, W n is the part added for
dynamic processes, and ').. is a time constant. Solving Eqs. (33a, b) we obtain, u,
p, Eij' and Sij for each element. With these data we can calculate the dissipated
energy on each element, per cycle:

qv = f SijdEij (34)
cycle

4.4. HEAT TRANSFER PROBLEM FORMULA nON

The conductive heat transfer equation has the following form [13]

div(')..VT) - pCvT + qv + r = 0 (35)

where T is the time derivative of temperature, VT is the temperature gradient, p is

the density, Cv is the specific heat at constant volume, qv is the dissipated energy
per cycle, and r are the internal heat sources. The conditions on the boundary r
are defined in two sub-domains, r T and r q : on r T we have Dirichlet type
conditions, T = To, and on r q Neumann type conditions, ')..(aT / an) = qo. Using
the weighted residual method [14] we can write:

-f qBVTdQ - f PCv TBTdQ + f(q v + r) BTdQ + fqoBTBr + 0 (36)

n n n r
For each element we have

where Nt(x) is the shape function and T\e)(t) is the nodal temp-literature. From
Eqs. (36, 37) we obtain the system of ordinary differential equations,

(38)

For each element we have

F(e)in = - f~~ qdQ - f NqvdQ - f NqodTq (39)

n n r

F(e)ex = f NrdQ
n
308

and, by their assembly,

(40)

In the steady state ,1t = ct, where c is a constant, therefore

(41)

And finally, from Eq. (38) we obtain the system for the temperature distribution:

(42)

The algorithm presented in section 4.3 determines the stress-strain field. For
each cycle it is possible to calculate the dissipated energy per element. These data
are used in the algorithm presented in section 4.4, in order to obtain the potential
field of temperature. Based on the considerations presented in section 4.1, and
using the temperature field, we updated the values of COl. Figure 6 shows the
iterative cycle of calculations needed to obtain the temperature field after a number
of cycles. The cycle was simulated using the Mark K6 code, in order to validate
the experimental results and constants used in several formulas. Using the
coefficients for the model C) 0 = 550 kPa and COl = 380 kPa the differences
between the experimental and calculated hysteresis loops are less than 2%.

5. Thermodynamic Model

In the classical engineering approach reliability analysis is usually a test that

follows the series manufacturing phase. Modern mechanical engineering [15]
requires the reliability principles to be taken into account in the conception phase
of the products. A concurrent engineering project, as most of the present
engineering projects are, requires a reliability oriented design. Studying the
literature we have seen that the reliability analysis results reported by numerous
authors can be summarized as follows:
the strength of rubber mounts does not depend particularly on one or
several specific mechanical parameters, but on the combined result of a
group of parameters [2];
the type of loading is very important in the failure analysis, and each type
of mount will have a different behavior a different type of load;
in the tearing process, more appropriate for describing the rubber behavior
seems to be the mechanical work (i.e., an energetic parameter) rather than
the stress [1].
The reliability oriented design concept introduces in the design phase of a
product or system an optimization problem, in order to increase the lifetime and to
reduce the maintenance costs. For this purpose a reliability model is required. The
most appropriate type of reliability model for rubber mounts is the model derived
from the physics of the failure mechanism [17]. With regard to the physics of the
rubber mounts failure mechanism, the mechanical approach does not satisfy the
 309

------+ Mechanical Analysis

J
, °ij'£ij

qv calculation

Thermal Analysis

T =T(x,y,z)

elf) updating Calculation of the

I I
'---
generated entropy

Figure 6. The iterative cycle for calculating the temperature after a number of loading cycles.

requirements of a reliability model. As shown on the left side of Fig. 7, in a

mechanical approach we introduce in the system mechanical energy as mechanical
work (W 10) the system experiences a change in internal energy (AU), and the
output is mechanical work (Wout ) and heat (Qout). Every relation obtained in this
manner follows from the energy conservation law. In this approach we can define
the mechanical response characteristics of the system, but it is not possible to draw
any conclusions about the lifetime of the system.
Thermodynamics introduces a new dimension in the analysis of processes:
entropy. Introduced by R. Clausius and developed by Maxwell, Boltzmann and
Gibbs in a statistical approach, entropy is a concept that continues to be studied.
The motivation for the studies is to clarify the concept, as in the case of Carnap
[18] and others, or to redefine the foundation of thermodynamics in a quantum
mechanics approach, as in the case of Gyftopoulos and his school [19-21].In the
case of the rubber mount, the thermodynamic model is conceived as an nonisolated
closed system. The boundary allows the exchange of work and heat. The metal
plates are made out of mild steel, and the rubber is a Mooney-Rivlin temperature-
dependent material. For the purpose of the present study the load is assumed
compressive and harmonic. The connection between the thermomechanical process
and the reliability model is entropy (Fig. 7, right side). The generated entropy per
cycle is given by the following equation [22]:
(q·VT) Au
As = T2 +T (43)

where Au is the work lost per cycle (the area of the hysteresis loop):
310

~u
~u DOC=>

Fi/?ure 7. Thermodynamic model of the damping process.

0)
Llu = 21t f .
21t/ro

(Jijfijd't (44)
o
In Eq. (43) the generated entropy has two components: the first is from heat
transfer, Lls T , and the second is from hysteresis work loss, Lls H . The generated
entropy is a criterion for the failure accumulation in a rubber mount [1]. The
maximum value of Lls depends on the type of rubber mixture. It can be determined
from samples. The optimization steps are:
1. Static and dynamic calculation of the system that should be mounted.
2. Selection of the mounting system architecture and types of mounts.
3. Finite element analysis of every type of mount, or of mounts with every
type of loading. Identification of the areas with highest values of Lls
per cycle, and concentration of the analysis on those areas.
4. Forecasting of the lifetime in number of cycles.
5. If the forecasting is shorter than the required lifetime, usually we are
looking for other types of mounts. If there are constrains, three types of
changes are available: the shape of the mount, the improvement of the

¢/9NlJ
1''1/:1'

Fi/?ure 8. The rubber mount type subjected to the optimization process.

 311

heat transfer properties, and the reduction of the hysteresis loop area.
6. The analysis of the entire system is repeated.
7. Validation of the results by experimental measurements.
The rubber mount shown in Fig. 8 was optimized using the algorithm presented
above. A significant reduction of the temperature level was obtained by improving
the heat transfer properties of the rubber mixture. Before the optimization, we had
a mixture of rubber SAB4a; the optimized mixture was SAB31 (Fig. 9). The
mounts have been tested for several types of loadings: a = 0·, 30', 45' and 90·.

F = 1000 sin(314t) daN

D!= (0-900 )

AD! = 15 0

150
erC)
125 _-----------n
100
,
75
,,
, I

I
I
I
so , I

/
/
/
/

25 -SAB31
o{}r SAB4a
o~------------~------------~----~----~
o 5 10 15
t (min.)

Figure 9. Experimental results obtained after heat transfer optimization: SAB4a rubber before
optimization. versus SAB3l rubber after optimization.
312

6. Conclusions
Entropy generation minimization [23] is a method coming from thermal engineering.
The present study shows that it is worth extending the method toward nonthermal
areas of engineering. The entropy concept could be a powerful tool in the physical
reliability modeling of nonlinear behavior materials. Some research has been
conducted in the early sixties, but he modern CAD tools open new possibilities in
this area.
7. References
I. Poturaev, V.N. and Dirda, V.1. (1977) Rezinovie Detalii Mashin, Mashinostroenie, Moscow.
2. Poturaev, V.I., Dirda, V.1. and Krush, 1.1. (1975) Prikladnaia Mehanika Rezinii, Nauka Dumka, Kiev.
3. Truesdell, C. (1969) Rational Thermodynamics, McGraw-Hill Inc. New York.
4. Naunton, W.J.S. (1961) The Applied Science ()lRubber, Edward Arnold Publishers, London.
5. Hepburn, C. and Reynolds, R.J.W. (1975) Elastomers. Criteria .Ic)r Engineering Design, Applied
Science Publishers, London.
6. Ferry, J.D. (1970) Viscoelastic Properties (!l Polymers, Wiley, New York.
7. Wood, L.A. (1973) Physical Constants (!l Different Rubbers, National Bureau of Standards,
Washington D.C.
8. Eringen, C. (1980) Mechanics ()l Continua, Krieger, Malabar, FL.
9. Becker, R.H., Chin, U.Y. and Mark, J.E. (1975) Thermoelastic studies of diene polymers in elongation
and compression, Polymer J. 7-1, 234-240.
10. Popa, 8., Madarashan, T., Batzaga, N. and Adameshteanu, I. (1978) Thermal Stresses in Mechanical
Engineering, Editura Tehnica, Bucharest.
II. Ogden, IT. (1972) Finite Elements (!lNonlinear Continua, McGraw-Hili, New York.
12. Jankovich, E., Leblanc, F., Durand, M. and Bercovier, M. (1981) A finite element method for the
analysis of rubber parts, experimental and analytical assessment, Computers & Structures 14-(5-6),
385-391.
13. Zhu, Y.Y. and Cescotto, S. (1994) Transient thermal and thermomechanical analysis by mixed FEM,
Computers & Structures 53-2,275-304.
14. Zienkiewich, O.C. (1977) The Finite Element Method, McGraw-Hill, New York.
15. Kutz, M., ed. (1998) Mechanical Engineers' Handbook, Wiley, New York.
16. Freakley, P.K. and Payne, A.R. (1978) Theory and Practice (!l Engineering with Rubber, Applied
Science Publishers, London.
17. Ozekici, S., ed. (1996) Reliability and Maintenance of Complex Systems, Springer-Verlag, Berlin.
18. Camap, R. (1977) Two Essays on Entropy, University of Cali fomi a Press, Los Angeles.
19. Gyftopoulos, E.P. (1998) Maxwell's and Boltzmann's triumphant contributions to and misconceived
interpretations of thermodynamics, Int. J. Appl. Thermodynamics 1-(1-4), 9-20.
20. Gyftopoulos, E.P. (1997) Fundamentals of analysis of processes, Energy Converso Mgmt. 38-(15-17),
1525-1533.
21. Gyftopoulos, E.P. and Cubukcu, E. (1997) Entropy: thermodynamic definition and quantum
expression, Physical Review E55-4, 3851-3858.
22. Arpaci, V.S. (1966) Conduction Heat Transfer, Addison-Wesley, Reading, MA.
23. Bejan, A. (1995) Entropy Generation Minimization, CRC Press, Boca Raton, FL.
24. Simo, J.C. (1987) On a fully three-dimensional finite-strain viscoelastic damage model: formulation
and computational aspects, Compo Math. Appl. Mech. Eng. 60, 153-173.
SOLAR ENERGY CONVERSION INTO WORK:
SIMPLE UPPER BOUND EFFICIENCIES

V. BADESCU
Candida Oancea Institute of Solar Energy,
Faculty of Mechanical Engineering,
Polytechnic University of Bucharest,
Spl. Independentei 313, Bucharest 79590, Romania

1, Introduction
In this work we present simple upper bound formulae for the efficiency of
solar energy conversion into mechanical work. They are upper bounds as
they refer to a large class of systems. They are simple as they are functions
of two parameters only, namely the temperatures of the energy sources (Le.
the sun and the ambient). The models we develop here can be classified into
two categories, as they deal with enclosed or free (transferred) radiation,
respectively.

2, Enclosed Radiation
The main components of any solar power system are a solar converter
(which transforms solar energy into heat) and a thermal engine (which
uses that heat to generate work). The models of this section do not enter
the details of solar converter operation but focus on work production. Two
different sorts of thermal engines will be analyzed. First, we shall treat
the ideal case of a reversible (Carnot) engine. Second, an endoreversible
thermal engine will be considered.

2.1. REVERSIBLE THERMAL ENGINE

In a paper concerning the ideal conversion of enclosed radiation Bejan [1]
constructed a model which allowed unification of three existing theories.
The model was developed in [2]. Shortly. there is a reversible three-part
process performed by the isotropic blackbody radiation contained by a de-
formable reflecting enclosure (Fig. 1). In the first process (0 ----t 1) a volume
VI is filled with radiation at temperature T I , while in thermal contact

313
A. Bejan and E. Mamut (eels.), Thermodynamic OptimiZIJtion o/Complex Energy Systems, 313-322.
© 1999 Kluwer Academic Publishers.
314

·i· · ·
.........
.. .. .. . . ....

::m~j:m·
~:~:~:~::.::;:~:::

Figure 1. Conversion of enclosed radiation energy into work.

with a hot temperature reservoir consisting in isotropic blackbody radia-

tion (medium 1). Available work is produced during the constant-volume
cooling process (1 ---+ 2), in which the temperature gap between system
and the second (cold) temperature reservoir (medium 2) is bridged by a
Carnot engine. The (0 ---+ 1) scenario goes in the reverse direction in the
last process (2 ---+ 3), the system volume decreasing from VI to zero. As the
volume in states 0 and 3 is zero, the system allows cyclic operation.
In the states 1 and 2 the system parameters are, respectively

(1),(1')

where p stands for radiation pressure and a = 7.565 x 10- 16 Jm- 3 K- 4 •

During the process (0 ---+ 1) the volume VI is filled reversibly with
isotropic blackbody radiation of temperature T I . The work and heat inter-
actions are, respectively:

(2),(2')
During the process (1 ---+ 2) the radiation is cooled at constant volume
down to the temperature T2. Consequently, the internal energy U drops
from a Vi Tt to a Vi Tj. Any infinitesimal energy drop (-dU) acts as a pos-
itive heat input to the high temperature end of a reversible Carnot cycle,
whose instantaneous high temperature is T (this temperature drops grad-
ually from Tl to T2 ). The work extracted is therefore

7~ "'engine ( -dU) (3)

U=aV1Tt

where "'engine == 1 - T2/T is the instantaneous efficiency of the Carnot

engine, and where dU = d(aVi~). Performing the above integral leads to:

4T21T24
WI2 = aViTI4 [1- - - + -(-) 1 (4)
3TI 3 TI
 315

Let us assume the second medium consists of blackbody isotropic radia-

tion. In this case, during the process (2 - t 3) the interaction work required
to decrease the system volume from VIto zero is:

W23 = -P2 Vl (5)

The heat engine efficiency of the cycle is defined as:

_ i 6W
'f}---=
WOl + W12 + W23 (6)
- QOl QOl
However, additional assumptions yield different upper bound formulae as
follows.

2.1.1. Jeter's Efficiency

By using Eqs. (1)-(6) one finds:

'f}J = 1 - T2/T l (7)

This is the upper bound proposed by Jeter [3] (see Fig. 2).

2.1.2. Spanner's Efficiency

If we replace the reversible process (0 - t 1) with a spontaneous (irreversible)
process in which the system does not deliver any work, we have:

WOlirrev = 0 (8)
By using Eqs. (1'),(2'), (3)-(5), (6) and (8) one obtains:

4T2
'f}S = 1--- (9)
3T1
This is Spanner's efficiency [4] shown in Fig. 2, too.

2.1.3. Landsberg-Petela-Press Efficiency

Let us assume the second medium does not consist of blackbody radiation.
Then, we can replace the reversible processes (0 -t 1) and (2-t 3) with
spontaneous (irreversible) processes in which the system does not deliver
any work. Consequently: '

WOlirrev = W23irrev = 0 (10)

In this case Eqs. (1'), (2'),(4),(6),(7) and (10) yield the Landsberg - Petela
- Press (LPP) efficiency (Fig. 2):

4T2 1 T2 4
'f}LPP = 1- - - + -(-) (11)
3T1 3 Tl
316

J_.
/ Lanclaberg-

I 0.4

o:l

0 Eq.(16)
.().2
Eq.(1S)
-0.4 0
TzfTt

Figure 2. Upper bound efficiencies. Tl and T2 - temperatures of hot and cold reservoirs.

The LPP efficiency was first proposed by Petela in 1964 [5] and was re-
derived independently twelve years later by Landsberg [6] and Press [7]
under different theoretical approaches.

2.2. ENDOREVERSIBLE THERMAL ENGINE

The analysis above can be repeated in case the process (1 - ? 2) represents

work generation using an endoreversible thermal engine. For the sake of
generality "'engine in Eq. (3) is written as

"'engine = 1- ( T2)n
T (12)

where n is a positive underunitary number. If n=1/2 one has the usual

case of a Chambadal- Novikov - Curzon - Ahlborn (CNCA) efficiency (for
various endoreversible engines see Bejan [8]). Use of Eqs. (3) and (12) yields:

4 4 T2 n n T2 4
W12endorev = aV1T1 [1 - -4- ( T ) + -4- ( -T ) ] (13)
-n 1 -n 1
This equation may be used to derive three new efficiencies, by repeating
the procedure given in section 2.1. A single case is presented here.

2.2.1. Lower Upper Bound Efficiency

One replaces the reversible processes (0 - ? 1) and (2 - ? 3) with irreversible
processes in which the system does not deliver any work (as in section
2.1.3). Use of Eqs. (2'),(6),(10) and (13) yield:

ij = 1- _4_(T2)n + _n_(T2)4 (14)

4-n Tl 4-n Tl
Of course, when n=1 the Eq. (14) yields the LPP efficiency [Eq. (11)]. In the
case n=1/2 (associated with an endoreversible thermal engine) one obtains
 317

(15)

This last upper bound efficiency was derived for the first time in [9J. It is
lower than other efficiencies (see Fig. 2) and it is slightly higher than

",sup = 1_ ~(T2)~ + !(T2)2 (16)

3 Tl 3 Tl
which is the lowest simple upper bound efficiency we know [1OJ.

3. Transferred Radiation
In this section we present a more detailed model based on energy and en-
tropy fluxes [11]. This model allows a simple interpretation of the LPP and
Spanner efficiencies. The two heat reservoirs will be called pump (p) and
sink (s). The later consists of a fluid transparent for the pump radiation.
A solar converter (c) interacts with the pump by interchange of isotropic
radiation and with the sink by this and also by thermal convective and/or
conductive transfer. In the converter the energy of the blackbody radia-
tion from the pump (temperature Tp) is transformed into heat which is
transferred to the thermal engine where it is partially transformed into

s;nv,
mechanical work (see Fig. 3 below).
Including the entropy-generation rate the rates of energy and
entropy accumulation per unit converter area (Eeonv and Seonv ) are re-
spectively:

. .,
'Ppe + a'Pse - 'Pes - 'Pcp - Qcon - Q (17)
'l/Jpe + a'l/Jse - 'l/Jes - 'l/JCP - (Qeon + Q')/Te + Sr;mv (18)

In Eqs. (17)-(18) Te is the converter temperature while 'P and 'l/J refer to
energy and entropy fluxes, respectively. The suffix pc denotes fluxes from
the pump to the converter and cs fluxes from the converter to the sink.
Also, we considered the return fluxes from sink towards converter (sc) and
from converter towards the pump (cp). The introduction of the parameter a
in Eqs. (17)-(18) must be considered only as a mathematical artifice which
will be useful during the discussion of the results. Its value will normally be
unity. This means that the ambient is a radiative medium. However, in some
situations a could be zero. For example, we could imagine the converter
as being enveloped by a selective material which is permeable to the pump
radiation but is not permeable to the thermal radiation of the ambient.
Also, in case the ambient is a transparent non-radiative medium (i.e. a
318

rarefied gas) and we neglect the radiation received by the converter from
the more distant bodies we could again put a = O. The converter supplies
the heat flux Q' to the engine and the heat flux Qeon to the ambient. These
two heat fluxes are accompanied by the entropy fluxes Q' / Te and Qean/Te,
respectively.
Two balance equations may be written for the accumulation of energy
and entropy in the thermal engine (per unit converter area):

Eengine = Q' - Q - W , (19)

Sengine = Q' /Te - Q/Ts + s~ngine (20)

Here Ts is sink temperature, s~ngine is the entropy-generation rate during

work production, while the rate of work and the heat flux transferred from
the engine to the sink are denoted by vir and Q, respectively.
Simple algebra allows to derive the work rate W. First, we eliminate
Q' and Q' /Te between Eqs. (17) and (19). and (18) and (20), respectively.
Then, we eliminate Q between the two equations we just obtained. Finally
we derive:

w (21)
-('Pep - Ts'l/Jcp) - CEeonv - TsSeonv) - CEengine - TsSengine)
-Q'eon (1 - Ts/T.)e - T Sg(sconv + sengine)
9

In case of steady-state CEeonv = Eengine = Seonv = Sengine = 0) the Eq.

(21) becomes:

'Ppc - Ts'l/Jpe - ('Pes - Ts'l/Jes) + a( 'Pse - Ts'l/Jse) (22)

-('Pep - Ts'l/Jep) - Qeon(1- Ts/Te) - Ts(s~onv + s~ngine)

The above results apply to any kind of radiation. In the following only
steady-state blackbody radiation is considered.

3.1. BLACKBODY RADIATION

When isotropic blackbody radiation is considered the energy flux of par-
tially polarized radiation emitted by component i towards component j
(ij=p,c,s) is [11]:

2 - I{ B T4 (. . )
'Pij = ;:-2-
(J
ij i 1,J=P,S,C (23)
 319

....... , -, ....
. .. ... ,.. , .

sink· 1;

Figure 3. Solar power generation. Pump near (a) and far from (b) the converter.

where Bij are geometric factors while Pi E [O,lJ is the coefficient of polar-
ization. A similar procedure leads to the entropy fluxes:

4 2 - Pi B T3 (. . )
01.
'f/ij = '3;:--2-
(J'
ij i 1,J=p,S,c (24)

The reversibility of the individual" rays" has as a consequence the following

relations and notation [12J

Bck = Bkc(k = s,p) , BT == Bcp + Bcs (25),(25')

The energy flux emitted by a fluid layer depends on its thickness. Thus
two possible situations may be considered:
(a) the pump is near the converter (i.e. between them there is a thin
layer of radiative ambient medium). Example: the mirror and the receiver
of a concentrating system.
(b) the pump is far from the converter (i.e. between them there is a thick
layer of radiative ambient medium). Example: the sun and the absorber of
a flat-plate collector.
In the case (a) we may neglect the radiation received by the absorber
from that (thin) part of the ambient placed in front of the pump (see Fig.
3a). Consequently, we can write:

(26),(26')

The energy flux emitted by the converter in the total solid angle is:

2 - Pc BT rn4
'Pcp + 'Pcs = --2--;-CTJ. c- (27)

In the case (b) the converter receives radiation from the direction pump
=? converter from both the pump and the ambient (Fig. 3b). Consequently:
320

2 - Pp Bpc 4 2 - Ps BT 4
'Ppc = -2--:;ruTp , 'Psc = -2--;-uTs (28),(28')
Again, as in Eq. (27), the energy flux emitted by the converter depends on
the total accessible solid angle only.
At this stage notice that the two different physical situations mentioned
above can be represented by just one unified formalism if we recast Eqs.
(26),(26'),(28) and(28') in the following form:

_ 2 - PpBpc T.4 _ 2 - Ps BT - 8Bpc T4 (29)

'Ppc - 2 7r up, 'Psc - 2 7r U s

Here 8 = 1 or 0, depending on whether the pump is near or far from the

converter, respectively.
Similar equations may be written for the entropy fluxes

.1. _ i 2 - Pp Bpc T.3 .1. _ i 2 - Ps BT - 8Bpc TS

'f'pc - 3 2 7r up, 'f'sc - 3 2 7r U s (30),(30')

42 - Pc BT 3
'l/Jcp + 'l/Jcs = 3"-2--;-uTc (31)

We shall use the following notations

r4 Bpcl BT , a == Ts/Tp, b == TclTp (32),(32'), (32")

(0 < r:S 1 ; O:S a :S b :S 1) (33)

By replacing the Eqs. (27)-(31) in Eq. (22) and using Eqs. (32)-(32") we
obtain the steady-state work rate under the form:

W = BT 2 - Pp uT.4[ (1 _ ia + ~o:8 2 - P s a4)r4 _ 2 - Pc (b4 _ iab3)

7r 2 p 3 3 2 - Pp 2 - Pp 3

=
-~o:~ ~s a4 ] -
p
[Qcon(l - ~) + Ts(S~onv + s~ngine)] (34)

3.2. CONVERSION EFFICIENCIES

The efficiency of work production is defined as

'fJ == W/'Ppc (35)

which is the ratio of two fluxes. Use of Eqs. (29), (32)-(32"), (34) and (35)
yields:
 321

4 1 2 - Ps 4 1 2 - Pc 4 4 3 1 2 - Ps 4
"., 1 - -3 a + "3 a8 2 _ P. a - r4 [2 _ P. (b - "3 ab ) + "3 a 2 _ p. a ]
p p P
211" [Q. (1 a) T. (sconv sengine)] (36)
(2 - Pp)BTr 4 aTt con - b + S 9 + 9

Equation (36) may be used for a simple interpretation of the LPP and
Spanner efficiencies (taking into account t.hat sgonv + sgngine ~ 0). Three
particular cases are discussed in the following. Unpolarized radiation is
considered only (Pi = 0; i=p,c,s).

3.2.1. Non-radiative Ambient (a = 0).

Equation (36) reduces to:

4 1 4 4 3
1 - -a - -(b - -ab ) (37)
3 r4 3
11" [Q. (1 a)
-BTr 4 a T.4p con - -b
+ T. (s·9conv + s·9engine)]
S

We note that

"., < 1 - ~a == ".,S subject to b3 (b - ~a) > 0 (38),(38')

3.2.2. Radiative Ambient (a = 1) and Pump far from Converter (8 = 0)

Equation (36) becomes:

The terms in the two square brackets of Eq. (39) are non-negative for a ::; b.
Consequently, "., < 1 -~a == ".,S without the constraint Eq. (38').

3.2.3. Radiative Ambient (a = 1) and Pump near the Converter (8 = 1)

Equation (36) becomes:
322

-We see that

(41)

Note that 'f/LPP may be used as an upper bound efficiency in the other two
cases (sections 3.2.1 and 3.2.2).

4. Conclusions
Three well-known upper bound efficiencies for solar energy conversion into
work are derived in section 2 by using the unifying formalism developed
in [1]'[2]. The additional assumption of endoreversible power generation
yields a new, lower upper bound efficiency [Eq. (15)]. Lower but more com-
plex upper bound formulae are treated in other papers [10], [13] - they are
more complex as they depend on the heat source temperatures and other
parameters.

References
1. Bejan, A. (1987) Unification of three different theories concerning the ideal conver-
sion of enclosed radiation, J. Solar Energy Eng. 109, 46-51.
2. Badescu, V. (1988) Discussion on the unification of three different theories concern-
ing the ideal conversion of enclosed radiation, J. Solar Energy Eng. 110, 349-350.
3. Jeter, J. (1981) Maximum conversion efficiency for the utilization of direct solar
radiation, Solar Energy 26, 231-236.
4. Spanner, D. C. (1964) Introduction to thermodynamics, Academic Press, London,
p.218.
5. Petela, R. (1964) Exergy of heat radiation, J. Heat Transfer 86, 187-192.
6. Landsberg, P T. and Mallinson, J R. (1976) Thermodynamic constraints, effective
temperatures and solar cells, in ColI. Int. sur l'Electricite Solaire, CNES, Toulouse,
pp 27-35.
7. Press, W.H. (1976) Theoretical maximum for energy from direct and diffuse sun-
light, Nature 264, 734-735.
8. Bejan, A. (1996) Entropy generation minimization: The new thermodynamics of
finite-size devices and finite-time processes, J. Appl. Phys. 79, 1191-1218.
9. Badescu, V. Simple upper bound efficiency for endoreversible conversion of enclosed
radiation, to be published.
10. Badescu, V. (1998) Accurate upper bound for the efficiency of converting solar
energy into work, J. Phys. D: Appl. Phys. 31,820-825.
11. Badescu, V. (1992) On the thermodynamics of the conversion of partially polarized
black-body radiation, J. Phys. III Prance 2, 1925-1941.
12. Landsberg, P.T. and Baruch, P. (1989) The thermodynamics of the conversion of
radiation energy for photovoltaics, J. Phys. A: Math. Gen. 22, 1911-1926.
13. Badescu, V. (1990) On the thermodynamics of the conversion of diluted radiation,
J. Phys. D : Appl. Phys. 23, 289-292.
THERMODYNAMIC APPROACH TO THE OPTIMIZATION OF
CENTRAL SOLAR ENERGY SYSTEMS

A. SEGAL
Solar Research Facilities Unit, Weizmann Institute o/Science
76100 Rehovot, Israel

1. Introduction

The sun is equivalent to a 5800 K thermal reservoir. Ifwe should be capable to

concentrate the sun light on the earth, close to the thermodynamic limit and to
use its power almost without losses, we should obtain the most powerful
thermal engine with a fantastic efficiency of more than 90%. Unfortunately this
seems to be impossible and we must be content with much less. Only a minute
process of optimization of all the system components that take the solar energy
and transform it in a transportable energy form (i.e. electricity) will increase the
actual (rather) low efficiency and will transform it in a competitive efficiency
comparative with the non-renewable energy process efficiencies. The central
solar energy systems seem to be unique qualified of efficiently producing of
high temperatures capable to achieve high thermodynamic efficiencies and
therefore the future looks to be more promising for the central receiver (or solar
tower) concept. This work is dedicated to an attempt to optimize a central solar
energy system from the thermodynamical and optical points of view.
The overall efficiency of a such system is the product of its subsystem
efficiencies' components. A central receiver system is composed of a few
subsystems: optical collector subsystem, thermal receiver (mounted in a tower
or, its challenger, the ground receiver that uses a tower reflector [1, 2]) and
final conversion subsystem. Design of this system essentially involves
determining the sizes, types and configurations of each of the subsystem's
components that would maximize a certain performance criterion (and/or
minimize a certain cost criterion) while satisfying the technical specifications
and performance requirements. For this central solar receiver system, the
computationally feasible approach is to decompose the system and to design
each subsystem separately. In that sense the overall system efficiency lloverall is
the product of three subsystem efficiencies:
323
A. Bejan and E. Mamut (eds.), Thennodynmnic Optimization o/Complex Energy Systems, 323-334.
@ 1999 Kluwer Academic Publishers.
324

17m·erall = 170pl (C) '17ree (C , Tree) '17 ph (Tree' PT ) (1)

where l10pl is the efficiency of the reflecting and concentrating optical
components (primary and secondary; llree is the efficiency of the conversion of
sunlight to heat in a receiver and llph is the efficiency of the conversion of heat
to electricity (or mechanical work) in the power block. As is seen in the
previous relation, the subsystem efficiencies are interdependent on three main
design parameters: the concentration ratio C, the receiver average temperature
Tree, and gas turbine pressure ratio Pr (when the power block is based on a
gas-turbine as will be discussed further). In a first approximation, as the
concentration C is increased, l10pl decreases and llree increases; as the receiver
average temperature Tree is increased, llrec decreases and llph increases. The
efficiency of the optical subsystem, l10pl commonly decreases with
concentration, since achieving a high concentration usually requires special
devices (secondary concentrators) which will introduce their proper losses due
to the imperfections of these devices, inherent absorption, back reflections and
so one. Ignoring these losses and considering an ideal machine which converts
the heat in the mechanical work, (i.e. l10pl = I, llph = llcamo,) Fletcher [3]
represented the combined effect of temperature and concentration on the overall
system efficiency. He found that this efficiency increases with concentration
ratio, with an optimal temperature for each concentration.
Further we will define and will analyze the possibilities (and the limits) for
the optimization of the three principal subsystems: the primary concentrator
(heliostat field), the receiver together its concentrator (secondary/receiver
concentrator) and power block system (from the thermodynamic point of view
only).

2. The Optimization of the Collector Field Layout

From all subsystems component a central solar plant, the most expensive item
is the collector subsystem. It composed of an individually sun tracking mirrors
(heliostats) that continuously tracks the sun during daylight hours so that the
specular component of incident direct beam solar radiation is reflected to a
common single aim-point or to multiple aim-point array. In the aim-point, either
cavity or external receivers can be placed. An advanced configuration, having a
hyperboloidal mirror with one of its foci in the aim point of the collector
subsystem to reflect the concentrated radiation to the ground, is also a possible
optical configuration for a central receiver system [1, 2, 4]. It is therefore
economically important to optimize this subsystem and minimize the number of
heliostats.
 325

Many attempts have been done to establish a theory that gives a

mathematical description of heliostats subsystem [5-7]. However since a field
may have thousands of units it is not practical to handle each one separately. It
is therefore necessary to simplify the problem and represent the field layout
with the least possible number of spacing parameters (preserving however the
accuracy of the calculus). Lipps and Vant-Hull [5], adapted this approach. They
have divided the field into a number of cells and analyzed the optimum relative
position between heliostats in each cell.
Segal and Epstein [8] proposed a new method of heliostat field layout
optimization starting with a general accepted algorithm for designing a heliostat
field and including actual optimized values for radial and azimuthal separation
between the heliostats.

2.1. BACKGROUND

The optimal design of a collector heliostat field involves determining the

optimal location of each heliostat in the field and the boundary contour of this
field so'that the daily, monthly or annually performances of the entire field will
optimally meet the requirements.
In general, the figure of merit used for this optimization is the ratio of the
total system cost to the total energy delivered by the whole system. In this case,
the optimization will consists in finding the minimum for this figure of merit.
Referring to the optimization of the collector field subsystem, the previous
figure of merit is equivalent to the maximization of the another figure of merit
which is the average energy delivered by a single heliostat from the field. (This
approach ignores in the first approximation the cost of the land and the wiring).
In order to calculate this average energy per heliostat, the coordinates of each
heliostat in the field must be determined. At the same time, this optimization
cannot be independent of the optimization of the balance of the central solar
plant, especially the height of the tower/aim- point. This is important data input
for the optimization of the collector subsystem. Therefore, we will express all
the lengths in the field in units of tower/aim-point height (in other words the
tower height is equal to one unit length).
A field of heliostats can be arranged in two fashions. One is cornfield array
and the second one is radial stagger array. In the classical cornfield the
heliostats are disposed in parallel and equidistant rows. Because of this
equidistant positioning of the heliostats, those far of the tower bring about a
serious percentage of blocking, so that it is by large agreed that the radial
staggered array is a more feasible arrangement. So far, there is only one
algorithm for building a precise heliostat field using radial staggered
arrangement. In conformity with this algorithm, the heliostats are disposed in
326

concentric circles, the pOSItioning starting with the circle farthest from the
tower. Here the heliostats are disposed at equal azimuthal distance starting with
the north position. In a inner circle, having a smaller radius from the previous
circle with a radial separation, the heliostats are arranged in a staggered array,
and so one. When the distance between two neighboring heliostats decreases
below a prefixed limit (in general equal to the heliostat diagonal length plus an
additional safety distance), one of n+ I heliostats is deleted from the rowand, in
the next row, the azimuthal separation is increased by a uniform repartition of
the remaining heliostats. The present code, named RAST, part of the program
package WISDOM [9], is capable of calculating the position of each heliostat in
the field, based on the prefixed correlations for the radial separation and for the
azimuthal separation in the farthest row. The azimuthal distance and radial
separations can be determined by various empirical correlations given by
Kistler [10] or Winter et al.[ 11]. These correlations, although mathematically
are not likewise, give very closed results. For a given field (i.e. the exterior
boundary is fixed at rmax) we found that both these correlations leads ·to
remarkable linearity (more than 98%, as can be seen in Fig. 1) between the
radial separation and the distance to the tower:
M = (u· r + v)· hi (2)
where u and v are functions of rmax and hi is the typical heliostat dimension. A
graphical dependence of these functions vs. r max is shown in Fig. 2.

070 070

SJ ---- afkrKistl ... [IO) ----v

- 0 - afkrWrft'retai. (11) 065 0 --<J- U 065

4)
\ 0
4) OaJ OaJ

·r
;. 055 055
3)
::f

OSJ OSJ

045 045

ro ro 04)
\.<r'" 04)
o laJ an :m 4ll 3D 0 2 4 6

Fiddmius(m) Mlximmfield raiLS (torB lrits)

Figure 1. Correlations for the radial separation Figure 2. Parameters for linear separation
vs. field maximum radius. vs. field maximum radius.
 327

2.2. BUILDING OF AN OPTIMUM FIELD OF HELIOSTATS

In view of the above arguments, we believe that an optimal field, arranged

radial staggered array, built on basis of the above-described algorithm, can be
achieved using the following mathematical formulation:
For a given tower (with a unit height) and a field of NH heliostats the
optimization of the layout of the heliostats in the field (with prefixed
boundaries) means the maximization of the ratio:
F = E,/N H (3)
Here E, is the total yearly energy delivered to the target. Initially the field must
be sized so that the power delivered at the design point to be larger than the
required nominal power P,IO/II which we have to deliver to the target at the
design point. The target is considered, for the first iteration, as a horizontal disk
situated at the aim-point and having a diameter sufficiently large to intercept all
the energy reflected by the field. This initial guess will be improved in the
subsequent iterations connected with the following components of the central
solar plant.
This simplified model takes into account only the contribution of the
heliostat to the cost of energy (the tower height and the target size being
prefixed). As a result, it is necessary to couple this model with other computer
code (for example DELSOL [10]) which optimize the tower height when the
receiver characteristics are given as input.
Therefore, for finding the optimal array of heliostats, we consider a field
having the row separation given by Eq.(2) and a fixed initial separation between
two heliostats from the same row (Lla l = w·h,).
This formulation constitutes a problem of optimization in three dimensions
(u, v, w) which can be easily solved, assuming that the field was built based on
the above described algorithm. After solving this problem, we have an optimal
configuration of the field, with maximum average contribution of each heliostat
integrated over a year and the power at the design point larger than the required
nominal power P,IO/II . We can draw on this field a map of isoenergetic heliostats
[8J. At this point, we will select from this layout, only those heliostats ordered
according to their annual energetical contribution to the receiver energy, until
we achieve the desired power at the design point. The list of heliostats selected
in this mode will be the optimal field for the given condition.
This method for building an optimum field of heliostats was exercised in the
design of a 30MWth industrial central tower plant for solar cracking of liquid
petroleum gas [4].
The optical efficiency can be defined as the ratio of energy delivered at the
receiver aperture plane to the total collectible solar energy. The general
328

acceptance of the term collectible refers to the amount of radiation that primary
collectors (heliostats) would have intercepted in some layout of those heliostats.
An upper limit of the amount of energy that can be collected from a large field
of heliostats is the radiation falling within the boundary of the field. This
amount of energy can be collected by a tightly packed collector array arranged
so that no ground area is ever exposed to the sun. Of course, such type of field
is completely unrealistic from the economic point of view, but this packed field
can serve as an idea on the upper limit. Realistic, the collectible field is that
field that is beforehand optimized in the sense described in this section. Depend
of the field dimension, in that case, the optimum optical efficiency varies in the
range 0.65 - 0.72.

3. The Receiver Efficiency

The average primary concentration obtained with a conventional heliostat field

and a central tower is not sufficient to reach high temperatures (above 1100K)
with maximum receiver efficiency. Therefore, a secondary stage of concen-
tration with a Receiver Concentrator (RC), usually Compound Parabolic
Concentrator (CPC), [12], is used to further concentrate the primary level. The
receiver and the RC then become a combined unit where both the thermal and
the additional optical losses have to be analyzed and considered. The receiver
and the RC unit can be placed on a Tower-Top (as in the conventional central
solar plants) aimed either straight downwards or tilted towards the center of a
heliostat field section, or it can be placed on the ground, aiming upwards as in
the case of a Tower-Reflector [1, 2]. In a large plant it is likely that there will be
an array or a cluster of receiver/RC units either in the Tower-Top optics as well
as in Tower-Reflector optics.
There is a need to redefine the efficiency of the combined receiver/RC unit.
The receiver efficiency, TJrec, is defined as the net power absorbed in receiver
reported at the total power arrived at the RC's entrance plane. Therefore, in this
definition is included both the optical and thermal efficiencies of the ensemble
consisting of the receiver and the RC in front of the receiver.
It is assumed that the RC has a circular entrance with a radius R that
intercepts an amount Pap from the total power PI reflected from the heliostats
and reaches the plane of the RC's aperture. The difference between Pap and PI is
the spillage around the RC. The previous definition of the receiver efficiency,
7], can be formulated as:

17ret = (a~u .(Pap - Pre; - puh.,) - Prod - p,,,,)/p, (4)

 329

where a~tr is the effective receiver absorptance, Pap is the power intercepted by
the RC's aperture, Pr~i is the amount of power rejected from the RC, Pahs is the
power absorbed by the RC surface, Prad is the amount of power lost by
reradiation from the cavity, P nc is the power lost by natural convection and PI is
the total power arriving to the plane of the RC's aperture. It is assumed that the
receiver is well insulated and the losses by conduction can be neglected. An
ideally insulated, isothermal receiver, without a transparent window at the
receiver entrance, can be considered as a blackbody with the effective
emissivity Eeff:::::l. When a window is added, a spectral discrimination must be
performed and Eeff «1) will be a function of the average temperature of the
receiver.
Studying the dependence between the previous components from the Eq.
(4) we obtained [1], for a receiver without window, the efficiency given by:
_ r (R) _ (J"T~e7rR2 sin 2 0 CHAr {Tree -Ta)1.25 (S)
'f/rec - a elf J ap 'f/(,I'C P, P,
where: !ap(R) is the fraction of P, enclosed within a circle having radius R with
!ap(RmwJ = 1, Rmax being the radius in this plane which collects all the power P"
llcpc is the RC's efficiency (it is assumed that the concentrator is a perfect CPC
having the acceptance angle ,the entrance radius R and the exit radius R sinO
[12]), a is the Stefan - Boltzmann constant, Tree is the representative (average)
receiver working temperature, Ta being the ambient temperature, CH is a
coefficient depending of geometrical receiver parameters and physical air
characteristics and Ar is the typical area of the receiver implied in the natural
convection process [13]. When a window protects the receiver, the natural
convection losses can be reduced substantially although some window loss is
introduced.

4. Thermodynamical Optimization of the Receiver and Power Block

Subsystems

The existing central solar plants (as Luz SEGS plants operating at 300-400°C,
or Solar Two from Barstow, CA, operating at S6S°C) are primarily based on
Rankine-cycle steam turbines. Gas turbines (Brayton-cycle engines) are most
commonly used for peak power production only. The efficiency and power of
gas-turbine cycle can be noticeably improved by recovering the heat from the
exhaust gases to vaporize water and run steam turbines, usually referred to as
combine cycles. Combined-cycle plants retain the gas-turbine advantages of
quick starting and flexible operation [14, IS]. The relatively high efficiency of
these plants (over SO%) stems, in part, from the high hot source temperatures
330

(in our case, TreJ that can be exploited, typically in the range 1200-1400 K. The
new generation of solar volumetric receivers [16] can achieve easily these
temperatures.
We apply basic thermodynamic principles toward evaluating the efficiency
of realistic power-block deserving a solar receiver and based on combined
cycles. There is also the analytic prediction of optimal operating points
(especially the working receiver temperature Tre<.).

4.1. GAS-TURBINES

The classic Brayton cycle gas-turbine is shown schematically in Fig. 3. The

cycle is comprised of: adiabatic compression (l ~2), constant pressure heat
input (2 ~3), adiabatic expansion (3 ~4), and constant pressure heat rejection
(4 ~l).

T,

T,

Gas
Com pressor/-oll--------i turbine

To T.

Cooler

Figure 3. Schematically representation of a Brayton cycle.

The heat received by the gas is:

Q= me p (T3 - T2 ) (6)
where m is the mass flow rate of the air heated between the temperatures T3
and T4. In this case, the Brayton cycle efficiency is [14]:
1] -1-[1-
Ii -
(}'lr(Pr
Pr(O-I)
-1)]/[1_ -1 ] Pc
'lc(O-I)
(7)

<r-Il! r ()(r-Il! r
where: () = ( TJ IT, ) = ( Tmax ITmin ) , Pl' = ( P3 IP4 ) ,Pc = P2 / PI ,
y=cp/c y is the specific heat ratio, Pi (j ~1,2,3,4) is the air pressure after each stage j,
11T, 11c being the isentropic efficiencies for gas turbine and compressor
respectively. If P2 = P3 and PI = P4, Pc = PI' = P and the gas turbine efficiency
can be written:
 331

(p-l)(X-p)
(8)
T/B = p(~_ p)
where X = T/rT/J) and ~ = 1+ T/c(B-l). The maximum efficiency will be when:
X-[x(l- ~)(X- ~)r2
p= ~)
1+ X-~
The compression work We is given by:
We = me pr.. (Pc -1) 177e (10)
The expansion work We is given by:
We = me pT31'Jr (1- p~l) (11)
Whereas the temperatures T/ and T3 are usually known, the intermediate
temperatures T} and T.J can be calculated as:
We We
7; = I; +-.- T4 = T) --.- (12)
mcl' mcl'
The overall gas turbine efficiency is:
(13)
where 'lleonv is the conversion efficiency from mechanical to electrical energy.

Solar receiver

Air in

Steam
turbire
block

Figure 4. Combined cycle with gas turbine and a steam cycle; temperatures diagram.
332

4.2. COMBINED CYCLES

A combined cycle is illustrated schematically in Fig. 4. The cycle configuration

is one topping cycle (the gas turbine) and one or more (practically, no more
than two) bottoming cycles (the steam turbines). A heat exchange couples the
gas turbine to the steam cycle(s). Following [15], we will maximize the work
output for each cycle individually and then combining the cycles is equivalent
to maximizing total work output. From the FigA results that the turbine exhaust
is at temperature T./ and the input for the steam turbine is at temperature T/
(which is the subject of the optimization). For a single bottoming cycle, the
work output is:
W,eam
(I) _ • /. •
-17"eam17co"ymc,,(T4 T4 -1)(T4 -J;) (14)
where T4" is the steam temperature at entrance into turbine and Tl is the
temperature of the cold external reservoir. The maximum cycle work output is
obtained when [17]:
, 1/2
~,O"'im"l = (~J;) (15)
so that the maximum cycle work will be:
(I) • (T1I2 T.1I2)2 (16)
w,\'leam,max = T}sleam 17t'Ollv mc p 4 - I

When two steam bottoming cycles are introduced, the corresponding results
are [15, 17]:
(17)
with:
(2T.
W(2) -
steam,max. -
.
lJ,\·team 1Jco""me p 1 + T4 -
3T.2f3TI/3)
1 4 (18)
As a result, the work output for the optimized combined cycles W max:
Wmax = 17m"y (w;, - WJ + W,~;;'~,max (19)
where N=I, rarely 2. The maximum efficiency of the combined cycle (power
block) is then:
Wmax Wmax
(20)
17"h = -Q = mc, ,(T3 - T2 )

The realistic partial efficiencies presented in previous relations are [15]: 1]c =
0.88, 1]8 = 1].I/eam = 0,85, 1]conv=0,95, Also other values are: y = lA, Pc = 1.03, PT'

5. Observations and Conclusions

We indicated in a previous section that solar collection efficiency decreases

with working receiver temperature. But the efficiency of the combined cycles
heat engines, however, increases with this operating temperature. Therefore the
 333

solar-driven combined-cycle power plants require an additional optimization

step to determine the solar working receiver temperature that maximizes the net
plant efficiency.
The optimization process for the entire solar system can be broken into three
subsystem optimizations: -- optimization of the power block combine cycle for
the pressure ratio obtaining 17pb.oplim(Tre,); -- optimization of the solar receiver
for the temperature Tree obtaining 17ree.oplim(C) (we also included here the optimi-
zation of the RC); -- finally, the optimization of the solar primary collector in
order to obtain the optimum concentration at the receiver entrance at the
nominal power required.
The Eq. (5) presents the efficiency of the solar receiver function of the RC
efficiency and the receiver temperature Tree. In a first approximation we will put
T3 = Tree. Following the strategy for optimization described here we have
obtained the results presented in Fig. 5 where we considered 17opt,max=0.7. In
[18] is demonstrated that the receiver efficiency can be improved by a partition

y . . . . ---------. .' . . . .
...................
.... - .. Uall turbine only
__ -OIlS turbine + one steam cycle
0.30
• ..... ·Improvment due 10 receiver parlionin@.

0.25
.,' ."
c 0.20 ............................
u
;
~ . .. -'"
~
...... .. ,
~ 0.15 ~.. ..,
"5 ~.. ~,
c "

'" .
0.10

0.05

0.00
800 1000 1200 1400 1600 1800
Receh'cr temperature, 1' ... (K)

Figure 5. Overall efficiency for a solar-driven combined cycle system.

in the receiver aperture so that the central part which receives the highest flux
will have a high temperature Tree while the outer ring receiving lower flux will
have a lower temperature Tper (in our calculations Tper = 0.65 Tree) but this pro
portion depending of the initial partition is subject of a further optimization).
This improvement contributes with 1-2% at the overall efficiency as can be
seen from the upper curve from Fig.S. The optimal working temperature
(central receiver) is in the range 1100 - 1300 K and the overall solar plant
efficiencies are in 26-28% range.
334

The central solar energy system is a very complex system. The ultimate foe
of the extensive utilization of the solar energy is economic. Therefore,
optimization of this system is of paramount importance and the present work
represents an attempt in this direction.

6. References

1. Segal, A and Epstein, M. (1997) Comparative performances of Tower -Top and Tower-
Ref/ector central solar receivers, submitted to Solar Energy.
2 Segal, A, Epstein, M.(l998) The reflective tower as an option for high temperature central
receivers, Solar Thermal Concentrating Technologies, Proc. of the 9 th Intn'l Symposium,
Odeillo - Font-Romeu, France.
3. Fletcher, E and Moen, R.L (1977) Hydrogen and oxygen from water- The use of solar energy
in one-step effusional process is considered, Science 197, 1050-1056.
4. Segal, A and Epstein, M. (1997) Modeling of solar receiver for cracking of liquid petroleum
gas, J. o/Solar Energy Engineering 119, 48-51.
5. Lipps, F.W. and Vant-Hull, L.L. (1978) A cellwise method for the optimization of large
central receiver systems, Solar Energy 20, 505-516.
6. Collado, F.I. and Turegano. I.A (1989) Calculation of the annual thermal energy supplied by
a defined heliostat field, Solar Energy 42, 149-165.
7. Elsayed, M. and Fathalah. K.A (1996) Solar flux-density distribution due to partially
shaded/blocked mirrors using the separation of variables superposition technique with
polynomial and Gaussian Sunshapes, J. a/Solar Energy Engineering 118, 107-114.
8. Segal, A and Epstein, M. (1996) A model for optimization of a heliostat field layout, in
M.Becker and M.Bohmer (eds.), Solar Thermal Concentrating Technologies, Proc. of the 8th
Int'l Symposium, Koln, 2, 989-998.
9. Segal, A (1996) WISDOM - Weizmann Institute Solar Dedicated cOmprehensive Mastercode
in Campbell-Howe R. and Wilkins-Crowder B. (eds.) The Proceedings 0/Solar'96, The 1996
American Solar Energy Society Annual Conference., Asheville, NC, 308-313.
10. Kistler, B.L. (1986) A user's manual for DELSOL3: A computer code for calculating the
optical performance and optimal system design for solar thermal central receiver plants,
Sandia National Laboratories, SAND86- 8018, Albuquerque, NM.
II. Winter, C. 1., Sizmann, R.L. and Vant-Hull, L.L. (Eds.) (1991) Solar Power Plants:
Fundamentals, Technology, Systems, Economics, Springer- Verlag, Berlin.
12. Welford, W.T. and Winston, R. (1989) High Collection Nonimaging Optics, Ch.4, Academic
Press, San Diego.
13. Bejan, A, (1995) Convection Heat Transfer, Ch.5, 2 nd Ed., Wiley, New York.
14. Horlock, I.H. (1992) Combined Power Plants. Including Combined Cycle Gas Turbine
(CCGT) Plants, Pergamon Press, Oxford.
15. Fraidenraich, N., Gordon, I.M., and Tiba, C. (1992) Optimization of a gas- turbine combined
cycles for solar energy and alternative- fuel power generation, Solar Energy 48,301-307.
16. Karni, I., Kribus, A, Doron, P., Rubin, R., Fiterman, A and Sagie, D. (1997) The DIAPR - a
High-Pressure, High-Temperature Solar Receiver, J. a/Solar Energy Engineering 119, 74-78.
17. Bejan, A (1996) Method of entropy generation minimization, or modeling and optimization
based on combined heat transfer and thermodynamics, Rev. Gen. Therm. 35, 637-646.
18. Ries, H., Kribus, A and Karni, I. (1995) Nonisothermal Receivers, J. 0/ Solar Energy
Engineering 117, 259-261.
THERMODYNAMIC OPTIMIZATION IN OCEAN THERMAL
ENERGY CONVERSION

y. lKEGAMI and H. UEHARA

OTEC Laboratory, Faculty o/Science and Eng., Saga University
1 Honjyo-machi, Saga-shi, Saga 840-8502, Japan

1. Introduction

As alternative energy sources to oil and uranium, we can consider well known
alternative sources such as solar power, geothermal power and wind power.
However when we consider the 21st century energy sources, ocean energy is also
an important potential energy source which must not be overlooked.
Among the various forms of ocean energy, the ocean thermal energy
associated with the vertical temperature gradient of the ocean is plentiful and very
stable. Power generation from ocean thermal energy is referred to as Ocean
Thermal Energy Conversion (OTEC). Namely, OTEC is a system of converting
heat energy into electricity by using the temperature difference between water at the
surface of the ocean and cold water of the depth (600 m"-' 1000 m).
Therefore, due to the small temperature difference (1 5 "-'26°C) that is available
for the OTEC plant, the Rankine cycle efficiency is only 3 - 4 percent [1]. Since the
efficiency of an OTEC plant is low, the performance of the whole electric power
generating system must be improved in order to reduce the busbar electric cost.
This entails accurate estimation of the dimensions of the major components
(evaporator, condenser, turbine, pump, valve, etc.) of the electric power plant.
Empirical equations for estimating the performance of each component are likewise
needed in the optimum design. The authors and other researchers have done
numerous theoretical and experimental researches about major components of the
OTEC system [1, 2]. Results of these studies show that ammonia and freon 22 are
suitable working fluids for a closed cycle OTEC plant. It was also found that the
plate-type heat exchangers are best fitted for OTEC. The authors have developed
the optimization method for OTEC using Rankine cycle. This method has been
used in actual design.
On the other hand, recently the attention has been focused on new cycles
invented by Kalina and Uehara, which employs a mixture of ammonia and water as
the working fluid [1, 3]. A new stage on the research ofOTEC has begun.
In this paper, the principle of thermodynamic design for OTEC is explained
and basically the optimum design which is most important for OTEC is presented.
335
A. Bejan and E. Mamut (eds.), Thermodynamic Optimization o/Complex Energy Systems, 335--344.
© 1999 Kluwer Academic Publishers.
 336

2. Principles of Thermodynamic Design for OTEC

Figure 1 is a schematic of the closed-cycle OTEC plant that shows the general
arrangement of heat exchangers, pumps, pipings, and the turbine generator. The
corresponding T-s diagram is shown in Fig. 2.
Ocean thermal energy has the following characteristics.
(1) Ocean thermal energy, just like solar energy, is a natural energy and is
renewable.
(2) Ocean thermal energy is a clean energy which does not pollute the natural
environment.
(3) The size of the energy resource is large but the energy density is low.
As a consequence of these characteristics, power generation from ocean thermal
energy requires an approach which is fundamentally different to that employed in
the present mainstream resources such as oil, coal and uranium.
Figure 3 shows the flow chart of principles of thermodynamic design for
OTEC. As the first step, the optimum design of the OTEC plant is carried out
under the given conditions of warm sea water temperature and cold sea water
temperature, considering as the design point [4]. The warm sea water temperature
fluctuates by the hour, day and month and also on depends on the site. When the
sea water temperature is deviated from its design point, the performance of the
OTEC plant also differs. As the second step, it is important to predict the optimum
operating conditions such as the flow rates of warm sea water, cold sea water and
working fluid, when the sea water temperature is deviated from its design point
[5, 6]. The annual performance of the OTEC plant which is operated on optimum
conditions, depends on the given design point. As the third step, it is valuable to
decide the optimum design point for each sea water temperature considering the
annual temperature variation of seawater [7]. In this paper the first step of the
thermodynamic optimization for the OTEC is presented.

Surface warm Turbine Generator ~~~ Lamp

--''-------1,;;:::=====1 Jf!1f@ ~~~ .::~

'j
Pump

Pump

Depth of ,
coldwater

Figure 1. Schematic of the closed-cycle OTEC.

 337

s
Figure 2. T-s diagram of the closed Rankine cycle.

Site Conditions tor OTEC

Optimum Design under Given Design Point

Optimum Operation under the Various Conditions

Optimum Design Point for Annual Conditions

Figure 3. Flow chart of principles oftherrnodynarnic design for the OTEC.

3. Analysis

3.1. OBJECTNE FUNCTION

The busbar electricity cost is shown in the following equation,

y = (total cost)/(net power per unit) (1)
It is desirable to use y as the objective function in the optimum design of the electric
power system, but the total cost of the electric power plant will vary depending on
the time at which the plant is constructed and on the designer/constructor. This
makes it difficult to optimize the OTEC plant using the value of y.
Since the heat exchanger cost (about 25 ~ 50 percent of the plant cost) is one
of the major costs of the OTEC plant [1], it is better to use r as the objective
338

function in the optimum design. The objective function r IS defined by the

following equation
(2)
where AT is the total heat transfer surface area of the evaporator and condenser and
PN is the net power. The net power is obtained by subtracting the power consumed
by the warm sea water pump, cold sea water pump, and working fluid pump from
the generated power. The total surface area is calculated using the heat-transfer
coefficients of the sea water and working fluid side.
The net power will be slightly lesser than the one in Eq. (2) because auxiliary
power is used in the plant for lighting, etc. This power is omitted in the present
analysis because it is relatively small and difficult to estimate exactly.

3.2. NET POWER

The net power PN is defined as

PN =Pa -(Pws + Pes + Pm;-) (3)
where PG is the turbine generate power, Pws is the warm sea water pumping power,
Pes is the cold sea water pumping power, and PWF is the working fluid pumping
power.

3.2.1. Turbine Generator Power

The turbine generator power, PG, is obtained from the product of the working fluid
mass flow rate, mm;-, and the adiabatic heat drop between the evaporator and the
condenser, and is given as
(4)
where TIr is the turbine efficiency and TJa is the generator efficiency. In this study,
the configurations of a single stage reaction axial flow turbine are used [4]. TJT IS
given as
TJT = TJ m • TJe (5)
where TJ m is the mechanical efficiency and TJe is the theoretical efficiency, defined
in reference [4] as
Had -(MN +MR +MEX +MD +MWEI')
TJ e = H (6)
ad
where Had is the theoretical adiabatic heat drop, LJ.hN is the nozzle loss, LJ.hR is the
rotor loss, LJ.hEX is the exhaust loss, LJ.hD is the rotary disc loss, andLJ.hWET is the
wetness loss. LJ.hN , LJ.hR , LJ.hEX, LJ.hD and LJ.h WET are calculated as reference [4].
 339

3.2.2. Cold Sea Water Pumping Power

The cold sea water pumping power Pes is given as
Pes ="'esMfesgl'lesp (7)
where, TJ csp is the cold sea water pump efficiency. LlHes is the total pressure
difference of the cold sea water piping, defined as
Mfes = (Mfes)p + (Mfes)e + (Mfes)d (8)
and (LlHes)p is the pressure difference of the cold sea water pipe, defined as
(Mf es)p = (Mf es )sp + (Mf eS)8 (9)
(LlHcs)sp (reference [4]) is the friction loss of the straight pipe, defined as
(Mfes)sp=6.82Iesl des l.17 {Ves ICes)1.85 (10)
Ces = 100 (11)
where les is the length of the cold sea water pipe, des is the diameter of the cold
sea water pipe, and Ves is the velocity of the cold sea water. (LlHcs)8 (reference [4])
is the bending loss of the cold sea water pipe line, defined as
2
(Mf es) B = (AI + Av + As + Aj + Ao + Ae + Ad + Ar )Ves 12g (12)
where :l I is the inlet loss coefficient,:l v is the valve loss coefficient,:l s is the
separating loss coefficient,:l j is the joint loss coefficient, :l 0 is the outlet loss
coefficient, :l e is the elbow loss coefficient, :l d is the diffuser loss coefficient, and
:l r is the reducer loss coefficient. These loss coefficients are taken from [4].
( LlHCS)d is the pressure difference caused by the density difference between
the warm and the cold sea water, defmed as
1 1
(Mfes)d =les ---(-(Pws +Pes)les) (13)
Pws 2
( LlHes)e is the cold sea water pressure difference in the condenser defined as
2
Ves Ie
(Mf es)e = Ae (14)
2g (Deq)cs
where Ie is the length of the condenser plate. The value of :l e is from [4].

3.2.3. Warm Sea Water and Working Fluid Pumping Power

The warm sea water pumping power Pws and the working fluid pumping power PWF
are given as
Pws = "'wsMfwsg l'Iwsp (15)
PWF = "'WF Mf WFg
I'IWFP (16)
where TJ wsp and TJ WFP are the warm sea water pump efficiency and the working fluid
pump efficiency. LlHws and LlHWF are the total pressure difference of the warm sea
 340

water plpmg and the total pressure difference of the working fluid plpmg,
respectively and defined similar to LlHcs .

3.3. HEAT TRANSFER SURFACE AREA

The total heat transfer surface area, AT, is given as

AT == AE +Ac (17)
AE and Ac are the heat transfer areas of the evaporator and the condenser. In this
paper, the shell-and-plate-type heat exchanger is used as the evaporator and the
condenser. The heat exchanger configuration and the manifold system are similar to
the one reported in reference [1]. The heat transfer enhancement of the evaporator is
achieved by using aluminum powder on the plate and providing flutes and
drainages. In the condenser only flutes and drainages are provided to enhance the
heat transfer. Details are reported in references [S, 9]. Respectively, AE and Ac are
defined as
AE == QE /(UE(ATm)E) == rhwscpws(TwSJ - Twso)/(U E(ATm)E) (IS)
Ac == Qc /(UC(ATm)c) == rhcscpcs(Tcso - TcSJ )/(Uc(ATm)c), (19)
where TWSI and Twso are the inlet and the outlet temperature of warm sea water. TcSJ
and Tcso are the inlet and the outlet temperature of cold sea water. QE and Qc is the
heat transfer rate of the evaporator and the condenser, respectively, defined as
QE == rhWF(h] -h4 ) (20)
Qc == rhWF (h2 - h3) (21)
hI, h2' h3' and 14 are the enthalpies indicated by the four points in Fig. 2,
respectively. (Ll Tm)E and (Ll Tm)c are the logarithmic mean temperature differences
of the evaporator and the condenser, respectively, defined as
(ATm)E ==[(TwSJ -TE)-(Twso -TE)]/ln[(TwSJ -TE)/(Twso -TE )] (22)

(ATm)c ==[(Tc -TcSJ)-(Tc -Tcso)]/ln[(Tc -TcSJ)/(Tc -Tcso)] (23)

where TE is the evaporation temperature and Tc is the condensate temperature.
rhWF is the working fluid flow rate, given as
(24)
UE and Uc are the overall heat-transfer coefficients of the evaporator and the
condenser, respectively. UE can be calculated by using the heat transfer coefficient
on the warm sea water side, hws, the boiling heat-transfer coefficient, hB' and
thermal conductivity of the heat transfer surface, (k..)E' Uc can be calculated by
using the heat transfer coefficient on the cold sea water side hcs, the condensation
heat transfer coefficient, hc, and thermal conductivity of the heat transfer surface,
(k",b
The boiling heat-transfer coefficient, hB' (reference [S]) is calculated from the
following equation:
 341

Y = 0.716(fpX)0919 H- 0834 ·(PL / P V r.j)448 ,(fpX ~ 14.9) (25)

Y = 2.218(fpX)05 H-0 834 • (PL / Pv )-0448 ,(fpX > 14.9) (26)
The heat transfer coefficient on the warm sea water side, hws [8] is given as
0.8 1/3
Nu ws = 0.047 Re ws Prws (27)
The condensation heat transfer coefficient, he [9] is calculated from
"-0.1 114
NU J = l.77Bo (GrJ PrJ / H) (28)
The heat transfer coefficient on the cold sea water side, hes [9] is given as
0.8 1/3
Nu es = 0.047 Rees Pres (29)

4. Optimization Method

4.1. OBJECTIVE FUNCTION AND VARIABLE

Substitution of Eqs. (3) - (29) into (2) gives the objective function y as a function
of the design parameter C, the geometric parameter of the plate-type heat exchanger
G, the state parameter of the working fluid S, the design parameter of components
of the power station D, the piping parameter of the plant P, and the design
parameter of the turbine T shown as follows:
Y= f(C,G,S,D,P,T) (30)
where C, G, S, D, P, and T are defined by the following equations:
C = f(Pa ,TWS] ,Tes] ,KW ,Prop) (31)

G = fUE,le, WE' we,~Yws,~Yes'(~YWF)E'(~YWF)e,J"E,J"e) (32)

S = f(TE,Te) (33)
D = f(Vws' Ves , '1wsp, '1esP' '1WFP' 'IT' '10) (34)
P = f(dws,lws,des,les,dWF,lWF) (35)
T = f(ZN,ZR,Ns,Ds,hl D,a 2 ) (36)
where Prop is the property of working fluid. IE and Ie are the lengths of the
evaporator and the condenser plate, WE and We are the widths of the evaporator and
the condenser plates, Llyw.s and LlYes are the clearances of warm sea water side in
the evaporator and cold sea water side in the condenser, (LlYWF)E and (LlYWF)e are the
clearances of working fluid side in the evaporator and the condenser, 0 E and 0 e
are thickness of the evaporator and the condenser plates, respectively. Vws and Ves
are the velocities of the warm sea water in the evaporator and the cold sea water in
the condenser. ZN is the number of nozzle blades, ZR is the number of rotor blades,
hiD is the ratio of blade height to diameter, and a 2 is the nozzle outlet angle. The
specific speed (Ns) and the specific diameter (Ds) are usually used as the design
parameters of a turbine.
342

4.2. CALCULATION METHOD

If the variables of parameter C are given, a minimum value of the objective function
can be calculated by using Eqs. (3) - (29). Furthermore, since IE' Ie, WE' We'
Ayws, AYes, (AYWF)E' (AYWF)e, 8E' 8 e , kw, d ws ' lws' des' les, d WF , IWF'
llwSP, lleSP' llWFP' lla, Z N' Z R and a 2 are given as input data Eq. (30) can be
rewritten as follows:
(37)

The objective function becomes a function of only seven independent variables.

Calculation is done as follows. First, the objective function (y ) at a certain point is
calculated. Again the objective
function (y 1) is calculated for a small
change of one variable [for example
TE is changed by L1 TE ] and keeping
the other six variables fixed and the
value of (y - y 1)/ L1 TE is obtained.
The new value of TE ' at the next step
is obtained by multiplying an
arbitrary small constant 0 with (y -
y 1)/ L1 TE and adding this value to the
former TE (TE' = TE + 6 (y - y 1)/ L1 TE). -g
The same procedure is done for the :5
Q)

other six variables in Eq. (37). E

The objective function IS
calculated by the combination of
these new variables and the
minimum value of the objective
function can then be obtained. The
minimum value of the objective
function, y mi", can be calculated by
the method of Powell [4].
Figure 4 shows the flow chart
of the calculation of optimum design
of the OTEC under the given
conditions of warm sea water inlet
temperature TWSb cold sea water inlet
temperature Tcs], generated turbine
power Pa. length and diameter of the
cold sea water pipe les and des. Figure 4. Flow chart of the calculation of
optimwn design.
 343

5. Results

Optimum design for any OTEC plant can be done using the above method. For
example, the optimization results for the OTEC plant at Ho-Ping in Taiwan is
shown in Table 1 [4]. The warm sea water temperature varies from 22 to 28.5°C at
Ho-Ping annually. In this paper, the warm sea water inlet temperature of 28°C is
used. Temperature at the 1000-m depth is about 4°C. The length of the cold sea
water pipe is 1000 m and the diameter is 11.2 m. Ammonia is used as working
fluid. The plates of the heat exchangers are made of titanium. Calculations were
done for a 100-MW OTEC plant. The evaporator and the condenser were divided
into four units of same size. The piping diagram of designed OTEC is shown in
reference [4].

Table 1 Optimization results.

Warm sea water inlet temperature TWSI [0C] 28.0
Cold sea water inlet temperature Tesl [0C] 4.0
Minimum objective fWlction Y min [m2/kW] 3.75
Rankine cycle efficiency lJRan [%] 3.98
Heat transfer area of evaporator AE [X 10 5m 21 1.38
Heat transfer area of condenser Ac [X 10 5m 21 1.30
Net power PN [MW] 71.4
Warm sea water pumping power P ws [MW] ILl
Cold sea water pumping power Pes [MW] 15.6
Working fluid pumping power P WF [MW] 1.97
Warm sea water flow rate rnws [tis] 206
Cold sea water flow rate rnes [tis] 207
Working fluid now rate rn WF [tIsl 2.27
Evaporation temperature TE [0C] 22.3
Condensate temperature Tc rOC] 10.2
Warm sea water velocity Vws [m1s] 1.009
Cold sea water velocity Ves [m1s] 1.068
Specific speed Ns [r.p.m., m3 /s, m] 54.6
Specific diameter Ds [m, m3/s, m] 1.67
Ratio of blade to diameter hiD [-] 0.244
Nozzle outlet angle l¥2 [deg.] 15.0
Turbine efficiency lJT [-] 0.905
Rotational speed N [X 103 r.p.m]. 3.135
Nozzle loss LlhN [kJ/kg] 0.317
Rotor loss LlhR [kJ/kg] 1.749
Exhaust loss LlhEX [kJ/kgl 0.016
Rotary disc loss LlhD [kJ/kg] 1.534
Wetness loss LlhWET [kJ/kg] 0.264
Over all heat transfer coefficient of VE [kW/m2J<] 5.39
evaporator
Over all heat transfer coefficient of Vc [kW/m2J<] 4.74
condenser
 344

In Table I the values of the Rankine cycle efficiency, the net power, the
pumping power, the optimum heat transfer area, the optimum sea water velocity in
heat exchanger, the optimum flow rates of warm sea water, cold sea water and
working fluid, and the pressure loss, etc., which correspond to the objective
function are obtained. Although it is not shown in the table, the optimum heat
exchanger specifications in more detail and the optimum turbine shapes, etc., are
obtained by this optimum design.

6. Conclusion

The principles of thermodynamic design for the OTEC was discussed using the
optimum design method under the given conditions and how the optimum design
point for each sea water temperature considering the annual temperature variation
of seawater. In this paper, as the first step of the thermodynamic optimization for
the OTEC, the optimum design method for the OTEC plant under the given
conditions of warm sea water temperature and cold sea water temperature is shown.
The method and concept shown in this paper can be applied to other conditions and
various energy systems using natural energy and lower temperature heat sources
(for example, refrigerators, geothermal power plants, power plant using wasted
water, etc).

Acknowledgments. The financial support for this study provided by the Japanese
Ministry of Education, Science and Culture is gratefully acknowledged.

7. References
1. Uehara, H (1995) The present status and future of ocean thermal energy conversion,
Intemational Joumal of Solar Energy, 16,217-231.
2. Avery, W. and Wu, C. (1994) Renewable Energy from the Ocean, Oxford University Press.
3. Uehara, H, Ikegami, Y, Nishida, T., Kilalchi, S. and Tsuboi, K. (1997) OTEC system using
Uehara cycle, IOA97 Intemational OTECIDOWA Association Coni Proceedings, 29-38.
4. Uehara, Hand Ikegami, Y (1990) Optimization of a closed-cycle OTEC system, Trans. of
the ASME Joumai ofSolar Energy Engineering, 112-4,247-256.
5. Ikegami, Y and Uehara, H (1992) Performance analysis of OTEC plants at ofT-design
conditions: ammonia as working fluid, ASJvIE. Solar Engineering 1992, 633-638.
6. Ikegami, Y and Uehara, H (1992) Optimum operating condition for a closed-cycle OTEC
system, Proc. 2nd JSlvIE-KSME Thermal Engineering Conference, 2·313-2·316.
7. Ikegami, Y and Uehara, H. (1994) Optimum design point for a closed-cycle OTEC system,
Proc. Fourth Intemational Offshore and Polar Engineering Coni Osaka, 383-389.
8. Nakaoka, T. and Uehara, H (1988) Performance test of a shell-and-plate type evaporator for
OTEC, Experimental Thermal and Fluid Science, 1-3,283-291.
9. Nakaoka, T. and Uehara, H. (1988) Performance test of a shell-and-plate type condenser for
OTEC, Experimental Thermal and Fluid Science, 1-3,275-281.
HOW TO UNIFY SOLAR ENERGY CONVERTERS AND CARNOT
ENGINES

A. DE VOS
vakgroep elektronika en informatiesystemen
Universiteit Gent
Sint Pietersnieuwstraat 41
B - 9000 Gent, Belgium

1. Introduction
There exist three ways of converting solar energy into useful work:

• The photothermal way absorbs solar energy as heat and subsequently con-
verts it into work by means of a heat engine.
• The photovoltaic way converts solar energy directly into electrical work.
• The hybrid way combines the photovoltaic and photothermal conversion.
Usually the description of photothermal and photovoltaic conversion is very dis-
similar. Photothermal converters are described in terms of absorptive coatings,
temperatures, insulation, heat circulation circuits, cyclic engines, etc. Photovoltaic
converters are described in terms of semiconductor bandgaps, electron and hole
diffusion lengths, electric voltages, etc. Through the wealth of technological de-
tails, it is very difficult to see any relationship between the fundamental principles
of the two converters.
Nevertheless, one can suspect that both are in some way or another related to
the most fundamental energy converter, i.e. the Carnot engine. Indeed, any solar
energy converter is a thermodynamic engine working between two heat reservoirs,
one at high temperature Tl and one at low temperature T 2 :

Tl = temperature of the Sun = 5762 K

T2 = temperature of the Earth = 288 K .

But the complicated technological description of solar energy converters stands

in strong contrast to the transparency of the Carnot theory. Indeed, let us for a
moment assume we could connect a reversible heat engine simultaneously to the
Sun's surface and to the Earth's surface: see Fig. 1. Let Ql be the heat we extract
from the Sun; let Q2 be the (waste) heat we deposit in the Earth. The produced
work is W. From the first law of thermodynamics (i.e. conservation of energy):

the second law (Le. conservation of entropy):

Ql Q2
Tl T2
345
A. Bejan and E. Mamut (eds.), Thermodynmnic OptimiZlltion o/Complex Energy Systems, 345-362.
© 1999 KIIIWer Academic Publishers.
346

t----t~ W

Figure 1: The Carnot engine.

and the definition of efficiency TJ = WI Qll we immediately find the Carnot formula
TJ = 1 - TalTl, yielding 95 %. Two questions arise here:

• Why does solar energy conversion, in contrast to Carnot theory, need two
complicated descriptions?
• What is the relationship between solar energy conversion efficiency and the
Carnot efficiency ?

2. Endoreversible Thermal Engines

The above puzzles are illuminated by application of endoreversible thermodynam-
ics [1, 2]. Both Novikov [3] and Curzon and Ahlborn [4] proposed a simple model,
where an ideal Carnot engine is combined with a linear heat resistor. The latter
has a heat-temperature characteristic

(1)

See Fig. 2a. Here, 9 is a constant, called the thermal conductance.

Taking into account the Carnot theorem, we have

(2)

Thus:
(3)

Now, the crucial step is to consider T3 in (1) and (3) as a parameter, we can
vary freely. Fig. 3a shows the functions Ql(T3 ) and W(Ta). Note that Ql(Ta) is
a straight line, whereas W(Ta) is a hyperbola. We will choose Ta such that we
 347

obtain maximum power output:

dW
-=0.
dT3
This leads to the maximum-power condition T; - TIT2 = 0, a quadratic equation
in T3 , yielding immediately
(T3 )opt = JTIT2 .
Substitution subsequently into (2) and (1) immediately leads to

( 1- /¥; ) Ql((T3 )opt}

9( v'1l - JT; )2 .
The factor 1 - JT2/T1 is known as the Gurzon-Ahlborn-Novikov efficiency. One
can easily demonstrate that (for any Tl and T2 obeying 0 < T2 < Td this number
is smaller than the Garnot efficiency 1 - T2 /T1 • This property is in agreement
with the fact that part of the engine is irreversible and with Garnot's law which
states that any irreversible engine is less efficient than the reversible engine working
between the same two temperatures.
We now relax the condition (1). This means we do not restrict ourselves
anymore to linear heat transport laws. We assume that the heat flux is given by

Ql = J(Td - J(T3) , (4)

with J( z) some mathematical function, not yet specified. This expression auto-
matically guarantees that no heat is exchanged if no temperature difference exists
(i.e. Ql = 0 whenever Tl = T3)' We assume J(z) monotonously increasing (in
order to guarantee that heat always flows from high to low temperature). The
special case J(z) = gz, of course, recovers the Gurzon-Ahlborn-Novikov case.
If we assume that heat is exchanged not by conduction but instead by radiation,
we have to apply the appropriate law. The simplest model is the case of black
bodies. The heat radiated by a black body is proportional to the fourth power
of its temperature, according to the Stefan-Boltzmann law. We therefore apply
J(z) = gz\ such that the net exchanged heat is
Ql = 9( Tt - Ti ) .
Thus the characteristic Ql(T3 ) is a fourth-power curve, replacing the straight line
of Fig. 3a. The corresponding W(T3) characteristic is a fifth-power curve:

W = 9 ( 1 - T2)(
T3 Tl4 - T34 ) .

The corresponding endoreversible model is the Muser model [5]. Maximum power
is extracted, if we satisfy

d~3 [ (1 - ~: )(Tt - Ti) ] = 0 ,

348

1---' W I---.W

Figure 2: Endoreversible engines: (a) thermal engine, (b) chemical engine, and
(c) thermochemical engine.

leading to the fifth-degree equation

4T; - 3T2Ti - T{T2 = 0 .

Such a polynomial equation cannot be solved analytically. We thus have to recur
to numerical calculations. For Tl = 5762 K and T2 = 288 K, we find T3 = 2443 K.
If we calculate the efficiency TJ ~ = 1- = ¥.-'
we find 88.2 %. However, in solar
energy conversion, the tradition exists to define the efficiency as the ratio of the
produced power to the incoming solar radiation alone:

w=-.
w
Qin
Whereas the denominator Ql of TJ is g(Tt - Ti), here the denominator Qin of w
is merely gT{, i.e. a constant. Therefore w is like a dimensionless W. This solar
conversion efficiency w amounts to
(1 ~) g(71- T.4) T.4
- T3 1 - 3 = (1 _ .7',
~ )(1 _ ~) .
gTt T3 Tt
Substitution of the Tl , T 2, and (T3)opt values leads to an efficiency w of 85.4 %.
 349

~3

o o

w W
W max W max

T3 ~3

0 T2 (T 3 )opt T1 0 ~2 (~3)OPt ~1

I \ \
0 ®
Figure 3: Characteristics of endoreversible engines: (a) heat-temperature and
work-temperature characteristics of a thermal engine, and (b) particle-potential
and work-potential characteristics of a chemical engine.

We now go one step further: instead of black-body radiation, we assume selec-

tive black-body radiation:

f(T) = ['JO E N(E, T) dE ,

lEg
where E is the photon energy and where the spectral density N(E, T) is assumed
to obey Planck's law:

211" E2
N(E, T) = c2h3 exp( k~) - 1 ' (5)

with c the speed of light, h Planck's constant, and k Boltzmann's constant. This
constitutes a generalization of the black-body case, as the Stefan-Boltzmann law
350

corresponds to the special case Eg = o. Indeed, because of

71"4

exp(x) - 1 15
we have

where a = ;;c'ha
5/c'
is called the Stefan-Boltzmann constant. For Eg > 0, we have

Ql =9 [ 100

Eg exp(
E3
E
/cT,) - 1
dE -
1E.
00
E3
E
exp( /cTa) - 1
dE ]

and thus

T2 [
W = 9 (1 - -)
T3
1 00

E.
E3
E
exp( leT, ) - 1
dE - 1 00

E.
E3
E
exp( /CTa) - 1
dE ] .

Again we consider T3 (which appears twice in the right-hand side of the latter
equation) as a free parameter. The maximum power condition dWjdT3 = 0 now
leads to a transcendental equation in T3 • Numerical calculations (for the above
values of Tl and T 2 ) lead to Fig. 4a, showing the efficiency w as a function of the
material's bandgap E g • We see that the optimum bandgap is zero. The maximum
efficiency is thus the black-body efficiency found earlier, i.e. 85.4 %. A finite
bandgap Eg is not advantageous for a photothermal solar converter. Nevertheless,
the introduction of a bandgap Eg will turn out to be fruitful, in later sections.
To model a photovoltaic solar converter is somewhat more complicated. Indeed,
a solar cell is not a thermal engine but a thermochemical engine, i.e. a combination
of a thermal engine and a chemical engine. Thus we first have to introduce the
endoreversible chemical engine.
3. Endoreversible Chemical Engines
In a chemical engine [6] the intensive variable is the chemical potential p" playing
a role similar to temperature T in the thermal engine. The extensive variable
exchanged between reservoirs is the particle flow N, playing a role similar to heat
flow Q in the thermal engine. See Fig. 2b. We assume the following type of
transport law holds:
Nl = h(p,t) - h(P,3) ,
where h( x) is some (monotonously increasing) mathematical function, not yet
specified. This equation replaces Eq. (4), the law of conservation of energy being
replaced by the law of conservation of particle number:

Nl - N2 = 0,
and the Carnot formula (2) being replaced by Gibbs's law
 351

100-+----~--~--~--~--~----+-

w [%] t
I 50

O~~--_r----~--~----_r----~--~--

o 1 2 3

Eg [eV]

Figure 4: Efficiency of solar energy converters: (a) photothermal converter,

(b) photovoltaic converter.

Thus:

In the same way as T3 was the free parameter in the endoreversible thermal engine,
now 1-'3 is the parameter of the end or eversible chemical engine. See Fig. 3b.
Maximum power output is found for

dW
-=0.
dl-'3

Applications are e.g. the linear law h(:z;) = g:z; and the exponential law h(:z;)
9 exp( .\:';.). But we will not go into the details of such models. Instead, we will, in
the next section, immediately combine the chemical engine with a thermal engine.
4. Endoreversible Thermochemical Engines
In a thermochemical system, both energy and particles are exchanged between the
two reservoirs # 1 and # 3. The energy current Ui and the particle current Ni
are converted into a work flow W, with conservation of energy, i.e.

conservation of particles, i.e.

352

and conservation of entropy (in the reversible part):

We easily deduce the unified Carnot-Gibbs law, i.e.

T2 T2
W = (1- -T3) U1 + (-J.£3
T3
- J.£2) Nl .

Here the input currents U1 and Nl still have to be replaced by their appropriate
transport expressions

U1 f(T1 , J.£d - f(T3 , J.£3)

Nl h(T1 , J.£l) - h(T3, J.£3) ,

such that W is a function of two variables, i.e. T3 and J.£3. See Fig. 2c.
Maximum power is guaranteed if

o.

We now investigate the case where the exchanged particles are photons. For
this purpose, we apply a generalization [7, 8, 9) of Planck's spectrum (5):

N(E T ) _ ~ E2
, , J.£ - C
2h3 exp (E-Il-)
kT -
l'

such that
h(T, J.£) =9 [00 EE2 dE .
JE g exp( k71J.) - 1
For the exchanged energy, we need to multiply the number of photons by the
photon energy E, such that the E2 factor becomes an E3 one:

f(T, J.£) =9 [00 EE3 dE .

lEg exp( k71J.)-1

Thus

W= 9 ( 1- T2)
T3
[1 00
E 3 dE
(E-Il-l) 1
_ 1 00
E 3 dE
(E-1l-3)_1
1

1 1
Eg exp kT, - Eg exp kT3
T2 00
E2 dE 00
E2 dE
+ 9(-TJ.£3-J.£2) [
3 Eg exp
(E-ll-l)_l-
kT, Eg exp
(~)
kT3 -
1]'

The function W(T3, J.£3) is a surface, displayed in Fig. 5.

 353

1-13

Figure 5: Solar energy conversion: (a) photothermal maximum-power point,

(b) photovoltaic maximum-power point, (c) hybrid photothermal-photovoltaic
maximum-power point.

For (Tl' J-£1) = (5762 K,0) and (T2' J-£2) = (288 K, 0), this yields a maximum
w of 86.6 per cent at (T3, J-£3) = (3890 K, - 0.75 eV). Because we have here two
degrees of freedom (T3 and J-£3), it is no surprise we attain higher efficiencies than
the 85.4 % we can obtain with a single degree of freedom (T3), as in Section 2.
The hybrid cell functions at a high temperature T 3 , but thanks to the negative
potential J-£3 the radiation of the semiconductor is limited [10, 11] and thus also
the (negative) terms of W containing the factor ( s 1"3 ) .
exp "T3 -1

5. Solar Cells
In the previous section, we have investigated the optimum performance of the
thermochemical solar energy converter, by imposing

8W
o
8T3
8W
O. (6)
8J-£3
354

This results in a maximum efficiency w of 86.6 % for this so-called hybrid de-
vice. We will now investigate the performance of purely thermal as well as purely
photovoltaic solar energy converters.
In a purely photothermal converter, we do not use the jJ, degree of freedom.
We choose jJ,3 equal to the terrestrial value 11-2 = o. Thus the set of conditions (6)
is replaced by the 'less optimal' conditions

We thus intersect the W(T3,11-3)-hill of Fig. 5 with a plane jJ,3 = 11-2, i.e. 11-3 = 0,
and search for the maximum on the resulting W(T3 )-curve. It turns out to amount
to 85.4 %, in accordance with Section 2.
In a purely photovoltaic converter, we do not use the T degree of freedom. We
choose T3 equal to the terrestrial value T2 = 288 K. Thus the set of conditions (6)
is replaced by the 'less optimal' conditions
T3
BW
o.
BI1-3
We thus intersect the W(T3, 11-3)-hill of Fig. 5 with a plane T3 = T2 and search for
the maximum on the resulting W(11-3)-curve:

W = g jJ,3 [ ("Xl ~2 dE _ [00 E2 dE 1 (7)

1E. exp( kT,) - 1 1E. exp( Ek"l3) - 1
with dWI dl1-3 = o. The results of the calculations are displayed in Fig. 4b. We see
that, in contrast to photot.hermal conversion, the photovoltaic conversion really
needs a non-zero bandgap value in order to yield a non-zero power W. The optimal
bandgap is 1.15 eV and the corresponding maximum efficiency is 40.8 %, i.e. far
less than the photothermal converter.
We now remark that the chemical potential 11-3 of the emitted photons is
changed by varying the bias voltage V of the semiconductor, according to the
simple relation jJ,3 = qV. This means that by an electric voltage V we influence jJ,3
and thus both intensity and spectral distribution of the emitted radiation. This is
what we call electroluminescence. Applying 11-3 = qV allows us to rewrite Eq. (7)
as W(V) = V x I(V) with

I(V) = qg [ [ex> --(-;E;-O-)

lE. exp
E2
kT, -
-1
dE -1 00
E. exp
E2
(E-qV) _ 1
kT,
dE 1. (8)

This current-voltage characteristic leads to the classical I(V) characteristic of a

solar cell by replacing 'Bose statistics with chemical potential' by 'Bose statistics
without chemical potential multiplied by Boltzmann statistics':
1 qV
exp( Ek"fV) _ 1 ~ exp( kT)
 355

1
Indeed, this approximation converts (8) into

I(V) ~ qg [ 1 00

Eg exp
E2
( E )
leT, - 1
qV
dE - exp( kT- )
2
00

Eg exp(-)
leT~
E2
E
-1
dE],

or
qV
I(V) ~ 1/ - I. [exp( kT2 ) - 1 ] ,
where 1/ is called the light current and I. the diode saturation current.
6. Tandem Solar Cells
We recall that the above calculations demonstrate that photovoltaic conversion is
far less efficient than photothermal conversion. There is however a way of boosting
the photovoltaic result w = 40.8 % of Section 5 towards higher efficiencies. This
happens by introducing more than one bandgap. This is realized by combining
different .semiconductors, each having its own bandgap. This leads to multi-gap
solar cells or 'tandem solar cells' [12, 13].
We assume that the n semiconductors are arranged in order of bandgap, the
largest bandgap at the sun-side:
Egl > Eg2 > ... > Eg(i-l) > Egi > Eg(i+l) > ... > Egn .
See Fig. 6. The most energetic photons (E > E gl ) are absorbed in the upper-
most semiconductor. Less energetic photons are absorbed in subsequent ma-
terials underneath. The last semic.onductor absorbs photons with little energy
(Egn < E < Eg(n-l)).
If each subcell ofthe tandem has its own voltage Vi, then the i th semiconductor
emits a spectrum

for E > Egi .

9 exp( E1:i{') - 1
We remark that the subcell absorbs two spectra:
• a part of the solar spectrum that transverses all semiconductors 1 to i-I:
E2
9 for Egi < E < Eg(i-l)
exp( Ie~,) - 1

• a part of the spectrum emitted by the layer beneath:

E2
for E> Egi .
9 exp( E-qV.±, ) - 1
leT,

Each sub cell not only has its own electric voltage Vi but also its own electric
current Ii and thus produces its own electric power Wi = ViIi :

Wi = 9 qVi [ l
Eg.
Eg ('-'l E2dE
E
exp( leT, ) - 1
+
1 00

Eg. exp(
E
E 2dE
qY·±,
leT; ) - 1
-
100

Eg.
E 2 dE
E- y.
exp( Ie.f~·) - 1
] .

Note that this expression is similar to Eq. (7) with two modifications:
356

i=1

i=2

i=n-1 E g (n-1 )

Figure 6: Tandem solar cell.

• a different upper bound of the first integral, and

• an additional term (i.e. the spectrum incident from the layer underneath).
The total work delivered per unit time is W = Ei Wi. We will not go into the
calculation details. Suffice it to give the results [13]. Fig. 7 shows the results for
n = 2: the efficiency as a function of the two band gaps Egl and E g2 . We find a
maximum of 55.7 % for Egl = 1.7 eV and Eg2 = 0.8 eV. The maximum deliverable
work is a strongly increasing function of n, the number of layers in the system:

• w = 40.8 per cent for n = 1,

• w = 55.7 per cent for n = 2,
• w = 63.9 per cent for n = 3,
• w = 68.9 per cent for n = 4, etc.
In the limit n -+ + 00, all bandgap increments Egi - Eg(i-l) become infinites-
imally small. And so do the work contributions Wi. After some manipulations,
we find that each subsystem acts as a monochromatic solar cell:
E2 E2
dW = V dI = gq V [ E9
exp(.::::.1L) - 1
9
( E.-qV)_1
]dE
g.
leT, exp leT2

Eventually, the sum Ei Wi becomes an integral:

W = roo
JEg=O
dW = qg 10
00
V [ E;
exp(h)-1 exp (
E2
9
E.-QV)_1
leT2
jdE
g'
(9)
 357

/.
/.
/.
/.
/.
/.
/.
2 /.
/.

r
/.
/.
/.
/.
/.
Eg2 [eVI /.
/.
/.
/.
/.
/.
/.

1
/.

/.
/.
/.
/.
/.
/.
/.
/.
/.
/.
/.
/.
0
0 1 2 3
•
Eg1 reV]

Figure 7: Efficiency of a tandem photovoltaic converter.

The maximum of the functional W is obtained for the optimal choice of the func-
tion V(Eg ). One finds 86.8 %.
We conclude this section with the formula for the upper bound of photovoltaic
energy conversion for an arbitrary temperature ratio t = T 2 /T1 • By merely in-
troducing the dimensionless bandgap u = E g /kT2 and the dimensionless voltage
:r:: = qV/kT2 , formula (9) becomes

15t4 1 1
w=-- -----:-----:--- ] u 2 du
11"4 :r:: [ exp(tu) _ 1 exp(u -:r::) - 1

where :r::(u) satisfies

1
~ {:r::[ 1 --:-------:-- ]} = 0 ,
d:r:: exp(tu) - 1 exp(u -:r::)-1

i.e. where :r::(u) is the solution of

(1 + :r::) exp( u -
:r::) - 1 1
[exp(u-:r::)-1]2 exp(tu) - 1
358

,,
,,
,,
,,
,,
,,
,,
,,
0.5
,,
,,
,,
,,
,,
,,
,
o -+--.--.---.--.--.--.--r--~-T~'~
o 0.5 1
t

Figure 8: Efficiency of solar energy conversion.

Fig. 8 displays this function w(t) and compares it with the Carnot function 1- t.
The curve w(t) is tangent to the ordinate axis for t = 0 and tangent to the abscissa
for t = 1. A good approximation [14, 15] is given by:

w(t) ~ (1- t)2 [ 1 + 0.3tlog(t)] .

E.g. for t = 288 K/5762 K, we find w ~ 0.862, close to the exact value 0.868.
7. Tandem Photothermal Converters
Not only photovoltaic converters can gain efficiency by introduction of more than
one bandgap. Also photothermal converters gain from additional bandgaps. Fig. 9
shows the computational results [17] for n = 2: the efficiency as a function of the
two bandgaps Egl and E g2 • We find a maximum of 86.1 % for Egl = 1.7 eV and
Eg2 = 0.0 eV. The maximum deliverable work is a weakly increasing function of
n, the number of layers in the system:

• w = 85.4 per cent for n = 1,

• w = 86.1 per cent for n = 2,
• w = 86.3 per cent for n = 3,
• w =
86.4 per cent for n =
4, etc.
 359

/
/
/
30%
2 /
/
/

[eVll
/
/
/

Eg2 /
/
/
60%
/
/
/
/

1
/
/
/
/
/

o 1 2 3

Eg1 [eV]

Figure 9: Efficiency of a tandem photothermal converter.

Thus the benefits from large n are not as important as in the photovoltaic case.
The case n --+ 00 results in an efficiency of 86.8 %, i.e. the same number as for
solar cells. This is no surprise. Indeed, making all bandgap increments infinitely
small results in a total power of

W = 100
E.=O
dW =9
100 (1 -
0
To
~) [
T3
E3
E9
exp( f.t,) - 1
-
exp(
E3
w.) - 1 1dE
E9 g •

In dimensionless form (with y = T2/T3 ) this becomes

15t4
w=--
11"4
100 o (
1- y
1
-
1
) [ exp(tu) - 1 exp(yu) - 1 1u
3 du

where y( u) satisfies
d 1 1
dy { (1 - y) [ exp(tu) - 1 exp(yu) _ 1 1} = 0 ,
i.e. where y( u) is the solution of
(1 +u- yu) exp(yu) - 1 1
[ exp(yu) - 1 J2 exp(tu) - 1
 360

It is clear that this expression is mathematically equivalent to its photovoltaic

counterpart in the previous section. This is verified by applying the substitution
=
:I: (1 - y)u or y = 1 - ~. Physically this means the following: each partial con-
verter converts a monochromatic slice of incident solar energy, where the argument
of the Bose factor exp(.~.)-l is
E
kTl .
For the monochromatic light radiated back this factor is

E-J1.3 E
either --- or --
kT2 kT3 '
either by applying a chemical potential J1.3 or by changing the temperature from
T2 to T3 , respectively. The effect can be made identical. It suffices to choose J1.3
such that J1.3 = (1 - ¥.)E or to choose T3 such that T3 = E~P.3 T 2. Such a choice
is possible, as E is a number (not an interval) in the monochromatic case.
As there are only two parameters (J1. and T) for influencing the Bose statistics,
the above proofs that the 86.8 % result is not only the common photovoltaic
and photothermal upper bound, but actually constitutes the upper bound for any

;::2
conceivable solar energy converter (working between the temperatures of 5762 K
and 288 K). The fact we cannot get the Carnot efficiency 1- = 95.0 % itself,
is caused by the fact that absorption of radiation without simultaneous emission
of radiation with the same spectrum, is inevitably an irreversible process [16].
8. Conclusion
In the present chapter, we have revealed the relationship between a solar energy
converter and a Carnot engine. We have demonstrated that a solar energy con-
verter can be modelled as an endoreversible thermochemical engine. The latter
consists of two separate parts:

• an irreversible part taking care of the transport of particles (i.e. photons)

and energy, by means of black body radiation and
• a reversible part taking care of the proper energy conversion.

We have found fundamental upper bounds for solar energy conversion efficiency:

• if we use a single semiconductor material:

40.8 % in case of photovoltaic conversion
85.4 % in case of photothermal conversion
86.6 % in case of hybrid conversion;
• if we use many semiconductor materials: 86.8 % for any of the three ways
of conversion.

It is necessary to stress that all theory in the present chapter was developed
for sunlight combined with an ideal light concentrator, consisting either of ideal
 361

mirrors, ideallenzes or a combination of both. For natural sunlight and for moder-
ately concentrated sunlight, the mathematics are somewhat less transparent. For
details the reader is refered to Reference [1). Suffice it here to give the results for
natural, i.e. unconcentrated, sunlight. The theoretical upper bounds are:

• if we use a single semiconductor material:

31.0 % in case of photovoltaic conversion

53.6 % in case of photothermal conversion
67.5 % in case of hybrid conversion;
• if we use many semiconductor materials: 68.2 % for any of the three ways
of conversion.

We finally like to stress that endoreversible thermodynamics not only is useful

for describing solar energy conversion, but also has been successfully applied in
numerous other fields, like chemistry [6, 18]' climatology [19, 20], economics [21,
22,23]' computing [24, 25, 26], etc.

References
[1] De Vos, A. (1992) Endoreversible Thermodynamics of Solar Energy Conver-
sion, Oxford University Press, Oxford.
[2) De Vos, A. (1993) The endoreversible theory of solar energy conversion: a
tutorial, Solar Energy Materials and Solar Cells 31, 75-93.
[3) Novikov, 1. (1957) Effektivyj koefficient poleznovo deystvia atomnoy ener-
geticeskoj ustanovki, Atomnaya Energiya 3, 409-412; in English translation
(1958) The efficiency of atomic power stations (a review), Journal of Nuclear
Energy II 7, 125-128.
[4) Curzon, F. and Ahlborn, B. (1975) Efficiency of a Carnot engine at maximum
power output, American Journal of Physics 43,22-24.
[5] Muser, H. (1957) Behandlung von Elektronenprozessen in Halbleiter-Rand-
schichten, Zeitschrijt fUr Physik 148, 380-390.
[6) De Vos, A. (1991) Endoreversible thermodynamics and chemical reactions,
Journal of Physical Chemistry 95, 4534-4540.
[7] De Vos, A. and Pauwels, H. (1981) On the thermodynamic limit of photo-
voltaic energy conversion, Applied Physics 25, 119-125.
[8) Landsberg, P. (1981) Photons at non-zero chemical potential, Journal of
Physics C: Solid State Physics 14, L 1025-1027.
[9] Wiirfel, P. (1982) The chemical potential of radiation, Journal of Physics C:
Solid State Physics 15, 3967-3985.
[10) De Vos, A. and Landries, J. (1992) Endoreversible thermodynamics of the hy-
brid photothermal-photovoltaic converter, 11 th European Photovoltaic Solar
Energy Conference, Montreux, 363-366.
362

[11) Spirkl, W. and Ries, H. (1995) Luminescence and efficiency of an ideal photo-
voltaic cell with charge carrier multiplication, Physical Review B 52, 11,319-
11,325.
[12) Jackson, E. (1955) Areas for improvement of the semiconductor solar energy
converter, Conference on the Use of Solar Energy, Tucson, 122-126.
[13] De Vos, A. (1980) Detailed balance limit of the efficiency of tandem solar
cells" Journal of Physics D: Applied Physics 13, 839-846.
[14] Grosjean, C. and De Vos, A. (1981) On the upper limit of the energy conver-
sion efficiency in tandem solar cells, Journal of Physics D: Applied Physics
14, 883-894.
[15) De Vos, A., Grosjean, C., and Pauwels, H. (1982) On the formula for the upper
limit of photovoltaic solar energy conversion efficiency, Journal of Physics D:
Applied Physics 15, 2003-2015.
[16) De Vos, A. and Pauwels, H. (1983) Comment on a thermodynamical paradox
presented by Wiirfel, Journal of Physics C: Solid State Physics 16, 6897-6909.

[17] De Vos, A. and Vyncke, D. (1983) Solar energy conversion: photovoltaic ver-
sus photothermal conversion, 5 th E. C. Photovoltaic Solar Energy Conference,
A'!?~lI(H, 186-190.
[18] De Vos, A. (1995) Thermodynamics of photochemical solar energy conversion,
Solar Energy Materials and Solar Cells 38, 11-22.
[19] De Vos, A. and Flater, G. (1991) The maximum efficiency of the conversion
of solar energy into wind energy, American Journal of Physics 59, 751-754.
[20] De Vos, A. and van der WeI, P. (1993) The efficiency of the conversion of solar
energy into wind energy by means of Hadley cells, Theoretical and Applied
Climatology 46, 193-202.
[21] De Vos, A. (1995) Endoreversible thermoeconomics, Energy Conversion and
Management 36, 1-5.
[22] De Vos, A. (1997) Endoreversible economics, Energy Conversion and Man-
agement 38, 311-317.
[23) De Vos, A. (1998) Endoreversible thermodynamics versus economics, Energy
Conversion and Management, to be published.
[24] De Vos, A. (1995) Reversible and endoreversible computing, International
Journal of Theoretical Physics 34, 2251-2266.
[25) De Vos, A. (1996) Introduction to r-MOS systems, 4 th Workshop on Physics
and Computation, Boston, 92-96.
[26] De Vos, A. (1997) Towards reversible digital computers, European Conference
on Circuit Theory and Design, Budapest, 923-931.
OPTIMAL CONTROL FOR MULTISTAGE ENDOREVERSffiLE ENGINES
WITH HEAT AND MASS TRANSFER

S. SIENIUTYCZ
Faculty of Chemical Engineering, Warsaw University of Technology
00-645 Warsaw, 1 Warynskiego Street, Poland

1. Multistage Endoreversible Process with Coupled Heat and Mass Transfer

This work makes an effort to synthesize a number of recent results obtained within
the theory of active (work-producing) heat processes and to generalize these results
to the so-far unsolved case of active simultaneous heat and mass transfer. Reviews on
the theory of engines with pure heat transfer are available [1-3] which display various
contributions to the classical theory of the Novikov-Curzon-Ahlborn process [4. 5]. A
description of an engine process with active simultaneous transfer of heat and mass
is. however, more complicated than that for an engine with pure heat transfer. mainly
due to the complexity of the relation describing the continuity of the entropy flux
through the internal reversible part of the engine. This complexity makes derivation of
the power generation function difficult. and, even if the function is finally found. is so
complex that an analytical solution to a work maximization problem cannot be
obtained. Thus certain effective numerical approaches need to be sought.
Consequently, the main task of this work is twofold: i) to find complex analytical
characteristics of a single-stage system ~nd ii) to develop effective numerical
algorithms for optimization of multistage cascades. which allow to exploit available
energy contained in finite resources more efficiently than in the single-stage objects.
In this paper we accomplish these goals under the assumption of the Lewis analogy
between the heat transfer and the mass transfer [6, 7]. Since. to our knowledge. most
of the currently-published results on FfT of chemical engines are restricted to
isothermal transfer [3. 8. 9], we believe that a successful approach to nonisothermal
transfer of energy and matter should be an important step towards the development of
a realistic theory of nonisothermal chemical engines.
As previously [1] we are dealing with a finite-time extension of the classical
problem of maximizing the work produced by a nonequilibrium engine system
operating with two flowing fluids or a fluid and a bath, i.e. an infinite reservoir. The
fluids have finite thermal conductivity and diffusivity, and hence nonvanishing
resistances in their boundary layers and the irreversible properties of any such heat-
transfer system. The process operates at a steady state. as represented in the
multistage system shown schematically in Fig. 1 for the case of a fluid and a bath.
For the purpose of optimization. we work with a standard representation of this
multistage process as a discrete optimal control problem in terms of a set of
difference equations. The multistage process is, in fact. a sequence of generalized
Novikov-Curzon-Ahlborn processes (NCA processes; [3-5]). These repeat in all the
stages of the cascade; the only differences among them are the temperatures of each
363
A. 8ejan and E. Mamut (eds.), Thermodynomic Optimization of Complex Energy Systems, 363-384.
© 1999 Kluwer Academic Publishers.
364

stage, which must be consistent with energy balance. The multistage syster
contains: the driving fluid, which flows with temperatures Tl .. TN and concentration
of a single active component Xl .. XN through the stages; the environment at th
constant temperature Te and concentration Xe ; the boundary layers which act as th
thermal conductances; and the set of the reversible engines, R l .. RN, generalizel
Carnot machines which generate the mechanical work at each stage n.
At each stage the bulk of the fluid is well mixed, and a mixing scale is assigned tl
the fluid flow. The fluxes of the work produced in stages are summed up, so that
cumulative average power output W is obtained at the last stage. The power output pe
unit flux of the driving fluid, W = WIG, has units of work per unit mole of the iner
fluid, and describes the molar work associated with the steady-flow process of powe
production. In classical thermodynamics such processes involve fluids and Carno
engines only; no boundary layers are considered. Their continuous quasi static limi
leads to the classical exergy for a reversible process; it has no rate term. It is th
irreversible multistage process with sequence of endoreversible machines of NCi
type which leads to a finite-time counterpart of the available energy (exergy) of th
driving fluid.
By definition, the work released in the engine mode is positive. An optimizatiol
problem for the cascade can be stated for the extremum work W= WIG, consisten
with extremum of power W. For the cascade of engines, where the work is released
the work W has to be maximized. For the cascade of heat pumps, where the work i
supplied to the system, the minimization of the negative of the work, (-W), i
required. When one of the boundary states of the resource fluid is the state 0
equilibrium with the environment (e.g. TN = ~, XN = X e), the maximal specifil
work represents an extended or finite-time exergy of the flowing fluid. We shall se,
that this extended exergy simplifies to the classical thermal exergy in the reversibl,
limit when the overall-stage conductances. gil, tend to infinity.
An active, work-producing, transfer process between two fluids causes destructiol
of initial non-equilibrium structure and approach of the resource fluid to equilibriun
with the reservoir fluid. For this process the problem of the maximum work deliverel
in a finite time is a relevant optimization problem. This (first) version of the proces
is called the engine mode. In the second version, called the heat-pump mode, th
system leaves thermodynamic equilibrium and work is added. The system works thel
as a heat pump, and the associated optimization problem is that of minimizing th,
work supplied in a finite time.
In classical, reversible thermodynamics, the associated magnitudes of th
mechanical work are the same for the two operations. Finite-time thermodynamics.
theory for irreversible processes, is a schema showing that (and why) the magnitud,
of the heat-pump-mode work is always larger than that of the engine-mode work.
We assume a steady-state process in the cascade. Each stage n, which is describel
by its difference equations (Sec.6) as a distinct "black-box" unit. has nonetheless it
own internal structure shown in Fig. 1 characteristic of that stage. This structure i
composed of the following elements: a well-mixed cell of bulk fluid (the shadowel
upper part of the stage n, with the inlet temperature TIl-l and outlet temperature Til
the thermal resistances of the two boundary layers. a reversible engine RIl, and.
contact zone with an isothermal bath (the shadowed lower part of the stage). Th,
bath is assumed to be thermally homogeneous at the outset and retains it
homogeneity for all time because it is infinite. The mixing is assumed ideal withil
 365

each nth fluid cell, consistent with the thermal homogeneity for pipeline fluid within
each stage and at the outlet from the stage

...
G
1 n-l

Rn- 1 Rn ~~
reversible n
machine W

. L..!2 2

o environment
Ilill X
~
Figure 1. Scheme of work production when a moist gas exchanges the heat and matter with an
infinite environment in a sequence of endoreversible thermal machines.

2. An Analysis of a Single-Stage Process to Obtain the Power Formula

We consider balances of energy, entropy and mass assuming that a single active
component (e.g. moisture) and the sensible heat q can be transferred in an inert gas
whose molar flow is G. The energy balance,

tl = t2 + w, (1)

and the mass balance in terms of molar fluxes,

nl = n2, (2)
are combined with an equation describing the continuity of the entropy flux in the
reversible part of the system

(El - ~l'nd/Tt' (3)

This yields the power expression

366

In terms of the heat flux q, the energy flux in the resistive parts of the system has the
continuity property, for example,

101 = ql + hlnl' = ql' + hl'nl' = 101' • (5)

Using the primed part of this equation in the power production formula one finds

(6)
We conclude that the combination of the second law and the reversible balance of
the entropy yields the power expression (6) in which the thermal efficiency of an
endoreversible process
(7)
is the component of a two-dimensional vector of efficiency. The second component of
this vector is the exergy-like function of the active component evaluated for the
primed state 2 as the reference state

(8)

The primed quantities and equations will further be used in a transformed form
expressinfg all physical quantities in terms of the bulk state variables of both fluids
and certain control variables, The latter will be related to fluxes of heat and matter in
a simple way as shown in the following section,
Using in Eq. (6) the energy flux continuity (5) we tend to obtain the power formula
in a 'flux representation'

(9)

rather than in an 'efficiency representation'

(10)

The energy flux continuity (5) is applied in Eq. (6) to eliminate the flux ql' on
account ql. This yields

(11)
where /3 is defined as

and, as 11, it should be expressed in terms of ql and n, By this the power w can finally
be expressed in terms of ql and n. These variables are good indicators of the process
intensity; their simple rate counterparts will be accepted as control variables in all
power optimization problems formulated in this work.
 367

3. Description of Transfer Processes with Rates as Control Variables

When considering the simultaneous heat and mass transfer we use mainly the transfer
conductances in units of transferred mass, g rather than the heat transfer
conductances, g', used frequently in descriptions of pure heat transfer. They both are
connected by the relationship g'= cg. In the formulae below we exploit the analogy
between the heat and mass transfer which yields for the conductances
. , ,
dy = !!.dA = kdA = ~vFdx = kavFdx = ~vFvdt (13)
c c c
We also define dy' = cdy, where dy' is the cumulative conductance in units of heat.
Now we introduce the so-called 'height of the transfer unit'

Gdx = GdV = GcdV = GdV = ~ = -.!:L = HTU (14)

dy Fdy Fa.'dA FkdA a.'avF kavF

and the differential of a nondimensional length called the 'number of transfer units'

d '
d't == dx = -.:L = a. avF dx = kavF dx = kavFv dt . (15)
Hn; G Gc G G

Note that a finite increment of 't, designated bye, equals gIG. From the mass
balance, the driving mass flux, N, satisfies dN = - GdX, where dX is the differential
change of the moisture concentration in the first fluid along the length dx. Hence the
following equations describe the change of the cumulative heat and mass fluxes, dQl
and dN 1 along the coordinates of cumulative conductances, 1 and y

dQ = _ GcdT =- GcdT (16)

dy' a.'dA a.'avFdx
and
dN = _ GdX = _ GdX = _ dX == _ v (17)
dY kdA kavFdx d't

Subscript 1 refers to the first fluid (a finite resource fluid). In Section 6 reference is
made to such formulae for cascades composed of finite number of finite-size stages.
The above equations prove that the rate of change of Tl and Xl in nondimensional
time 't may be chosen as the process control variables. The control u has the
temperature units, and the control v has nondimensional units of mole of the active
component per mole of inert gas, as the concentration X. They both are components
of the control vector u = (u, v).
In terms of the mass conductance gl an equation for the first intermediate
temperature T}' can be derived. Each of the primed parameters should be now
expressed in terms of the state variables of both fluids and the accepted control
variables. The operations use the continuity of the energy flux and yield successively
368

(18)

(19)

whence the heat flux ql follows as

(20;

and
(21)

For a process of pure heat transfer (n=O) one has

T 1' = T 1 - --.9L = T 1 - dOl = T 1 _ (dOl) ( d Y )

Cgl cdYl cdy dYl (22)
= Tl + (GcdTl )(~) = Tl + ..B...{dTl)
a'avFdx gl gl d't

which means that

ql = - cgdTdd't = - cgu == - g'u , (23)

A derivation of the related formula for T2' is more difficult. Here only an outline of
the derivation is given and a detailed derivation will be presented elsewhere. This
temperature follows from the reversible balance of the entropy through the internal
engine
If Newton's law of the heat exchange holds, the temperature Tl' in terms of the heat
flux ql is given above by Eq. (21). For the primed concentrations analogous relations
follow from the Fick law of mass transfer. They are

(24)
and
(25)

To obtain the second intermediate temperature T2' and concentration X2' one
substitutes Eqs. (21), (24) and (25) into the reversible entropy balance

(26)

in which the mass balance (2) was taken into account. For the accepted equations of
heat and mass transfer, the equation of the continuity for the entropy flux through the
engine becomes

(27)

Using the primed flux ql' as the process variable this equation can be written as
 369

(28)

Yet, this formula should be expressed in terms of the heat q 1 which is a direct
indicator of temperature changes in the fluid bulk. The heat q 1 is related to the heat
q I' driving the engine by the formula

(29)

the formula describing the continuity of entropy flux is

ql + Cp(TI - T!,)n + sl'(TI _ ql , Xl - n/gl)n

TI - (Clgl - cpnrlql Clgl - cpn
(30)
:: g2c(T2' - T2)!T2' + sz'(Tz', X2 + n/g2)n .

This equation along with Eqs. (18) and (29) defines the temperature T2' as an implicit
function of the fluxes ql and n. To work with this equation we assume an isobaric
entropy function for a humid gas in the form f7]

S(T, X) :: cln(T!T*) + R<I>(X), (31)

where
<I>(X) == (1+ X)ln (1+X) - XlnX . (32)

For this structure of the entropy we derive an equation for the temperature T2' (ql, n)
in the form

The associated thermal efficiency follows in the form

370

Our approach takes also into account the primed quantities which characterize the
mass transfer. In view of the equality n 1 = n2, we have

Xl' = Xl -
!!.L = Xl _ dNl = Xl _ (dNl)( dy)
gl dYI dy dYI
= XI + ( GdXI )Jt = XI + ( GdXl )Jt = XI + JtcdXl ) (35)
kavFdx gl kavFdx gl gl d't
and

Therefore the continuity equation for the transferred active component

(37)

is identically satisfied by the formulae for X I' and X2' as shown by equations

(38)

(39)

Because of coupling between the heat transfer and the transfer of mass, an equation
for the temperature Tl' takes into account the effect of transfer of moisture vapor

Tl' = T1 - _-"q~l_ = Tl ___d;..!Q:;,:l__

Cgl - cpn CdYl - cpdN
(40:
=TI _ (dy/dy))(dQI / cdy) =TI + (g/gddTt/ d't
1 - (cp/c)(dN/dy)(dy/dy)) 1 + (cp/c)(g/gl)(dX/d't)

This equation holds along with the mass transfer equation above. Hence, in terms of
the accepted controls, u and v, we obtain the following set of equations for the
primed quantities
 371

TI{U, v) = TI + ug/gl =TI + u (41)

1 + (cp/c)(g/gdv gl/g + (cp/c)v

Moreover, for the primed concentrations

(43,44)

From the last two equations we conclude that whenever process satisfies the equality
Xl' = X2' then the concentration change v = dX l/d't satisfies

- v = XI - X2 . (45)

Along with the equality Tl' = T2' this defines the so-called 'short-circuit point', the
point at which the production of the power in the system equals zero. This production
vanishes also at the point of vanishing fluxes. Thus, for some moderate values of the
fluxes, there exists a point in the system at which the power is maximum.

4. Flux and Rate Representations of Power and Related Quantities

The formulae for the temperature T2', given above, correspond to the second heat flux
at the interface contacting the engine

(46)

Absence of phase changes is assumed here. We should work, however, with the flux
q2 to describe an effect of the partial enthalpy of vapor of the active component

q2 = q2' + Cp(T2' - T2)n . (47)

Hence, using the expression for T2', one obtains the second heat flux in the form
372

(49)
follows in the form

where T2' is defined by Eq. (33). This serves to find the power produced. Since

(51;
and
(52:

the power output can be obtained as

From this equation and Eq. (50) the power expression in terms of fluxes follows
for T2' defined by Eq. (33).

T ( rl
W(ql. n) = ql + cITln + g2C2T2 - (g2C2 + cpn)Tz{ ( 1 - ql Clgl - cpn )
Tl - (gl/g)(Clgl - cpnr1ql
(1 +X) (I+XIJ(X n)(XI-.!L) (1 X)(1+X2)(X n)(X2+..!L) _D
1 1- - gl glR + 2 2 +=- ~ (gz.o')
( gl n ig2 2)(
C g2 C
n ) - g2 2 } (54)
XX'(1
1
X
+ 1- -
n )(1+ XI - -J
gl XX2(1
2
X
+ 2+-
n )(1 + X2 + -)
~ .
gl gz
The negative ratio of the power wand the conductance g defines the Lagrangian L 01
the problem in terms of the controls u and v. It has units of the energy per mole. We
also use the quantity fo = - L i.e. the power output per unit molar conductance g = GS

dW/d't = wig = w/(SG) == fo = - L . (55

To obtain the work produced per unit mole of the fluid. we integrate over 't the
function fO expressed in terms of controls and state. Substituting q = -gcu and n = -g~
to the power formula and using the non dimensional conductance S = giG yields

(56)
 373

This equation allows us to find a maximum for the cumulative mechanical work W
when a finite-resource fluid changes its thermodynamic parameters in a finite time
between two assumed states. However, we will first show that the first power formula
reduces exactly to that of pure heat transfer when n = O.

5. Correspondence with Pure Heat Transfer

For the vanishing flows ql and n the power w equals zero. In absence of mass transfer
(n = 0) one obtains from the above equation

(57)

or, equivalently

W(ql, 0) = ql - g2C2 T2{ ql }. (58)

{Tl - (gtlg)(Clgl rlqt}/{ (gtlg - l)(Clgl rl)

By transforming the denominator a simpler form of the above result can be obtained

(59)

Consistently, an equation describing the pure heat flux q2 =q} - w takes the form

q2(ql, 0) = g2C2(T2' - T2) = g2C2T2 ql (60)

g2C2(Tl - qtl(glCt» - ql

This was obtained previously in the context of thermal multistage NCA processes [1].
It can also be shown that, in absence of the mass transfer flux, n = 0, the analytical
solution with respect to the temperature T2' is obtained in the form

T2' = T2 (61)
1 - ql[g2C2(Tl - qtl(Clgt»r 1

which is consistent with the above equation for the heat flux q2.
The above power formula can also be transformed into the form

w = ql(1 - g2C2T2 ) = ql(l - -----=Tc.=.2- 1

- - --)
1
, (62)
g2C2(Tl qtl(glCl» - ql Tl - ql«glCt) - + (g2C2)" )

where the last expression contains the overall conductance of standard inactive heat
transfer. This conductance is defined in the standard way as
374

(63)

Consequently, in absence of mass transfer, one arrives at a simple equation linking

the power w with the driving heat flux ql = -gcu

(64)

and the corresponding first-law efficiency of the stage (g' = gc)

= l-~ (65)
T, +u

and the heat q 1 follows from this equation in the form

(66)

In the efficiency representation, the power produced, w = 'l")ql(ll), becomes

(67)

in agreement with ref. [3]. The Lagrangian of the thermal process in an infinite
cascade of NCA stages follows as the negative power w per unit conductance g, or
the work obtained from the one mole of the resource fluid at flow

L == - fo = - w = - 'l!..(l - T2 ) = cO - ~)u . (68)

g g T, - gcq j' T, + u

The efficiency of an engine stage deviates from the Carnot efficiency with the finite
heat flux q I. For a quasistatic transfer, the efficiency becomes the Carnot efficiency.
Otherwise, for a sufficiently large ql, equal to g'(TI - T2), which corresponds with the
pure heat conduction (no power) the efficiency vanishes. The efficiency formula
shows that, in a finite rate process, the efficiency is Carnot-like, with an effective
temperature of the upper source T' = TI + u. In the engine mode, where ql is positive
(the resource fluid is cooled), the quantity T' is reduced due to the finite flux q I, and
the efficiency decreases with ql. The power expression, Eq. (68), shows the deviation
from the Carnot theory caused by the nonvanishing heat current. The Carnot
efficiency is achieved when the effect of the overall resistance (g't 1 is negligible or
the flux ql is very low. The maximum power at the stage corresponds to ql satisfying

ow = (T, - q"g,)2 - T,T2 = 0 . (69)

oq, (T, _ q"g,)2

Whence, the driving heat flux at the maximum power conditions (subscript 0)
 375

(70)

It may be verified that the second derivative of w is negative at the extremum, so

that the extremum is a maximum. Note that the extremum qI is lower than the purely
conductive heat flux qI = g'(TI - T2) which takes place at the zero efficiency T\.
When the result described by Eq. (70) is substituted into the efficiency formula, Eq.
(65), the well-known efficiency of the Novikov engine at maximum power,

(71)

is obtained. Equation (69) leads to the following equation for the maximum power

Wo = g'[(T 1 )112 - T 2 )112]2 (72)

in agreement with the standard theory of thermal machines which uses the heat
transfer conductance g' = gc. When mass transfer plays a role, the results for optimal
values for the fluxes n, qI, q2 and w should be found numerically.
Thus we have proven the correspondence of our theory of active simultaneous heat
and mass transfer with the classical theory of the Novikov-Curzon-Ahlborn processes.
Change of control variable yields an equivalent description. Thus it is permissible to
choose a convenient control variable, typically one preferred for technical reasons.
Use of the driving heat as a control has some virtues for making physical
interpretation simple: with this choice, the reversible point is the point at which the
control variable qI is zero, and the irreversibility description in terms of qI leads to
formulae similar to those used frequently for describing 'inactive' transfer. On the
other hand, transition from one control to another is possible.

6. Applying Single-Stage Formulae to a Multistage Process

Now we return to the process of simultaneous heat and mass transfer to consider
cascade processes with finite number of stages. The state changes between the
boundary states of a stage are no longer infinitesimal. We call such processes
genuine discrete processes. We add the stage number superscript, n, to all symbols in
the above formulae when we apply them to a cascade. Yet, since in the infinite
reservoir case T2 and X2 are only parameters and not state variables, we can also
simplify our designation for the state variables T\', Xr and the heat qf by rejecting the
subscript I, unnecessary at the present time. Thus we will use the symbols Tn and qn
for the resource fluid temperature and driving heat flux at the stage n. Also we will
use the symbols Xn and nn for the resource fluid concentration and mass flux at the
stage n. We shall also designate the constant thermal reservoir temperature as Te
rather than 12 and its concentration as xe rather than xt
In genuine discrete processes we use the cumulative power per unit mass flow, wn
== ~wn/G, a quantity whose units are those of work per mole. When an equation
describing the power is applied at the stage n with the identifications T2 = ~, TI =
Tn, X2 = x e and Xl = xn and we introduce a nondimensional conductance en at the
stage n, such that
 376

en == gn/G, (73)

then un = - (g'n)-lqn = - (Gcen)-lqn satisfies

(74)

Similarly as for continuous processes, we also define a quantity vn as the mass flux
per unit mass flow G and unit nondimensional time en by the equation

(75)

We also obtain a state equation for the molar work in terms of the controls un and vn

(76)

where fo is the function appearing in Eq. (56).

Modeling of cascades with finite number of stages is precisely a discrete
counterpart of that for continuous processes, described earlier. The total work is the
sum of the expressions foen over the stages; this sum replaces the work integral of a
continuous process, which is the integral over fodt. Again, taking into account that
the conductance gn may be expressed as the product of an overall coefficient of
mass transfer and a corresponding overall transfer area, we identify the
nondimensional time interval en as the number of transfer units at the stage n, en =
Atn. The ratio of the stage length fi to en is a quantity with units of length; it is
identified with the height of the transfer unit at the stage n, H¥U. From its definition
and Eq. (73), H¥U equals Gfi/gn. The smallness of this quantity is a good indicator of
the quality of transfer at the stage; the quantities H¥U exhibit very small values for
large molar conductances gn. However it will be sufficient to use the dimensionless
variable en only. One may note that the variable, en, appears linearly in our discrete
model of optimization, a fact we will use in optimizing the model. To do this, we first
need the discrete equation of state, which describes the driving heat.
To find the equation of state for the temperature Tn, we define the cumulative
driving heat Qn for the first n stages of the cascade: Qn = I:qk, where k = 1, 2 .. n. This
quantity satisfies the equality Qn - Qn-l = - Gc(Tn - TD-l ) = qn, where, because of Eq.
(74), the heat qn is the product of - g,n (= - Gce n) = a'an and un. The comparison of
these two expressions yields a state equation for the temperature of the resource fluid

(77)

This equation tells us that un = - qn/g,n, i.e. the negative of the heat received by the
nth reversible engine (in units of temperature), at the same time plays the role of the
of the discrete rate of the temperature change, ATn/Atn .
To find the equation of state for the concentration xn an analogous reasoning holds.
The cumulative molar flux Nn for the first n stages of the cascade: Nn =I:nk, where k
= 1, 2.. n. This quantity satisfies the equality Nn - Nn-l = - Gexn - X n- 1) = nn, where,
 377

because of Eq. (75), the mass flux nn is the product of gn = Ge n and vn . The
comparison of these two expressions yields the equation of state. Thus an equation of
state for the concentration changes follows in the form

(78)

where vn is the rate change of the concentration per unit of the nondimensional time
or. To close the model, we recall that en is the increment of the independent variable
-rn. This variable is nondimensional, of definite (positive) sign, and measures the
cumulative number of heat transfer units. The variable -rn satisfies the definition -rn =
Lek for k = 1, 2 .. n. Thus last discrete equation of state is simply

(79)

Equation (79) is essential in problems with the constrained sum of en. This
corresponds to the constrained total transfer area, because the single-stage transfer
areas are contained in gn and hence in en. This also corresponds to a constraint on the
total length. On comparing these results with those for continuous processes, we
conclude that, in finite-n cascades, the discrete state equations follow from a simple
discretizing operation applied for the derivatives dT/d-r = u, dXld-r = v and d-r/d-r == 1.
In accordance with the terminology used in optimization, Eqs. (76) - (79) are the
discrete equations of state for the cascade. They contain on their right-hand sides the
state variables (in $, Eq. (56» and controls. An optimizing procedure for the cascade
requires specification of an optimization criterion, a complete set of the state
equations (as above), and possibly some local constraints at the stage.

7. General Optimal Control for Discrete Systems Linear in Holdup Times

For the endoreversible process considered here, we now show the application of two
efficient algorithms, of which the first is based on the dynamic programming and the
second on the so-called discrete maximum principle with a constant Hamiltonian [l0-
13]. Each can be applied to endoreversible cascades with complex function fo. The
existence of the maximum principle has been shown for discrete models which are
linear and homogeneous with respect to a particular unconstrained control variable.
This control may be an unconstrained interval of one of the state variables, such as
our interval of the dimensionless time, en, or an unconstrained interval of an
extensive parameter. The model's linearity with respect to en is crucial for the formal
similarity of the necessary conditions for optimality in discrete and continuous
processes because, broadly speaking, this linearity causes vanishing the second order
and higher terms in Taylor expansions of characteristic functions for discrete
processes. Consequently discrete formulae acquire the structures already known for
continuous processes.
Our aim here is a brief analysis of these properties for discrete (multistage)
processes with finite numbers of stages. For discrete processes, the standard discrete
theory of optimal control [14, 15] does not predict any special similarity of discrete
and continuous descriptions, e.g. such characteristic features of the continuous theory
as constancy of an autonomous Hamiltonian or the Hamilton-Jacobi equation [l6, 17].
378

This is because discrete descriptions are generally not reducible to continuous ones in
the limit of an infinite number of discrete units. To assure that a discrete model
converges to the continuous limit and symplectic properties, one must restrict oneself
to a special class of discrete processes in which finite intervals of state variables are
not constrained explicitly, and the allowed constraints can affect only ratios of
differences of the state variables at any stage. To satisfy these requirements, a
discrete set of the state equations and an optimization algorithm must be linear with
respect to a particular decision, en. Whenever a discrete model has a structure linear
in en, a remarkable similarity emerges between necessary optimality conditions in
both continuous and discrete cases. In particular, a discrete maximum principle
emerges in in a form analogous to that known for continuous systems. In the case of
endoreversible cascades, it is our optimization model, Eqs. (76)-(79) which shows the
similarity discussed. The outline of an abstract analytical description with occasional
references to the endoreversible cascades is given below.
Applying the usual description based on ordinary difference equations, the
optimization theory for Eqs. (76) - (79) of the endoreversible cascade can be
represented by a general set set of equations

(80)

Here i = 0, 1...s, s+ 1 and f.:+l = 1 for the (s+ 1)-th coordinate of the state vector xn,
consistent with Eqs. (76)- (79) for en which is the unconstrained interval of the
variable 't == x~+l' In the case of our endoreversible cascade, for which s=2, the state
vector xn is four-dimensional: x8 == wn, xf == Tn, x2 = X~ and x~ == 'tn, for each stage
n. The set (80) is now considered as a proper vector representation of the original
model of our endoreversible cascade. For this set the optimization problem is stated
as maximizing the zero-th state coordinate for n = N when the initial states of xn are
fixed. To derive necessary optimality conditions for for Eqs. (76)-(74) or general Eq.
(50), Bellman's method of dynamic programming (DP) can be applied [17-19]. This
allows one to pass from the DP algorithm to the algorithm of discrete maximum
principle, both being effective computational algorithms.
Let us define the optimal performance function for a general n-stage process

ID(xD) == max L° fa(x D, uD)eD = - min L -t8(XD, uD)e D,

D
(81)
1 1
where fa is the work generation function described by Eq. (56) with added stage index
n. In Eq. (81) one assumes that x = (XI, X2""X s +l= 't) i.e. that the performance
coordinate x8 has been excluded from the coordinates of the state vector. It may be
shown that the function IN incorporates the minimum entropy production with a
negative sign. and . for fixed end states. the work maximization is equivalent with
minimization of the entropy production. This is in agreement with general
thermodynamic principles [2, 21, 22].

For a general n-stage process, in the stages 1, 2, ... ,n described by Eqs. (80) and
(81), Bellman's recurrence equation is
 379

This corresponds with the so-called forward algorithm of dynamic programming where
the characteristic function 1 is generated in terms of the final states. This algorithm is
illustrated in Figure 2 which explains the principle of generation of optimal profits.

Figure 2. Application of Bellman's principle of optimality for the forward algorithm of the
dynamic programming method. Elipse-shaped balance areas refer to sequential computational
subprocesses, which evolve by inclusion of remaining stages.

Starting with 10=0, the sequence of the optimal functions II, I2, .. .I n.. .IN can be
obtained with a computer by a well-established recurrence procedure [18, 19] in
which extremizations on the right sides are carried out with respect to controls at
constant coordinate values of the final state xn. The tables describing the computer
output contain the numerical values of In, un, vn and en, in terms of Tn, xn and 'tn.
Dimensionality reduction is possible [1] for 'tn, which improves the data accuracy.
Now we outline the passage to the second solving method, a maximum principle.
To pass to a maximum principle one write~ Eq. (82) in the form

which enables us to include variations of final coordinates of state [11, 20]. For an
unconstrained en and possibly constrained un, the set equivalent to Eq. (83) has the
form of the three equations

(84)

(85)
380

max{f8(xD, uD)e D- (ID(XD) _ ID-l(XD - fl(x D, uD)e D)} = O. (86)

uD

In Eq. (85) the stationarity condition for the extremal intervals en was applied.
Whenever en is finite and positive, then, after using Eqs. (84) and (85) in (86), the
last equation can be transformed to

which is the strong maximum principle for the so-called enlarged Hamiltonian

HD-l = ..DC °
- 10 X , U
D) aI D-I L"il(
- - - . 1 x, U
°
D) -_ ..LO0 - £.. al D-l ~ -aI
~ --li --
D-I
(88)
ax D-1 i=1 axp-I a't D-1

with respect to the controls un. As fs+ I (=1) is u-independent, this optimality
condition can also be expressed in terms of the energylike Hamiltonian which does
not contain the partial derivative of In-1 with respect to the time,

s
H D- I = f8 - L (aID-1/axp-I)f;'. (89)
1=1

Equation (89) represents the Hamiltonian (88) without the u-independent term al n -
I/~n-l. Next the so-called adjoint variables are defined which are the partial
derivatives of In with respect to the state coordinates xp,
(90)

(i = 1...s, s+ 1.) The Hamiltonian (89) must be a maximum with respect to the controls
un which maximize the work production, WN =Lfae n, whose optimal function IN is
defined by Eq. (81) in which the work intensity fa is described by Eq. (56).
Differentiating the bracketed expression in Eq. (83) to determine its stationarity
conditions with respect to the final space and time coordinates, we obtain an optimal
difference set which is canollical with respect to two sort of equations, one defining
the changes of state and one the changes of the adjoints. Using the energylike
Hamiltonian, Eq. (89), expressed in terms of the adjoint variables,

L zp-1tf(X
S

HD-I(XD, zD-I, u D, 'tD) = f8(x D, u D, 'to) + O, u D, 'to) (91 )

i=1

the computational algorithm of maximum principle is obtained in the canonical form

= JH - zp - zp-I __ aH D- 1
n n1 o 1
XI - XI- (92, 93)
eD azp-I eO axp
 381

(94)

(95)

(n=l, .. ,N ; i=l, ... s and I =1. ... r.) Equation (92) constitutes the Hamiltonian form of the
state equations, and Eq. (93) is its adjoint equation. Equation (94) sets the
Hamiltonian interval at the stage n, whereas Eq. (95) states that the enlarged
Hamiltonian Hn-l = Hn-l + Z~-l of the extremal process is always constant and equal
zero. Equation (95) describes the discrete maximum principle in the allowable range
U. This is a very powerful computational algorithm which can effectively be used
even in case of very complex functions fO, such as that given by Eq. (56).

8. Work Maximizing in Endoreversible Cascades and Finite-Time Exergy

Since we possess an analytical expression for the power generation, fo, either
Bellman's Eq. (82) or discrete maximum principle, Eqs. (92) - (95), can be used to
numerically generate optimal controls and optimal profit functions. When one of the
end states is that of ambient a finite-time exergy, A. is generated

A(T, X) = (Cg + XCp)[(T - -re) _ Teln-I] + R-re {Xln[X(1 + Xe)]

~ (1 +X)X e (96)
+ In[ 1 + xe]} ± -reS(J(T. X. -reo Xe, 'tf) "
l+X

Of the two parts of the finite-time exergy the classical one is known from reversible
thermodynamics, thus one may use the tables of A to evaluate also the associated
mimimal entropy production for the process. Equation (96) refers to a sufficiently
large N. approximating well the continuous limit. Its last term contains the minimum
entropy production Sa as a function of the end states and the non-dimensional
duration, 'tf. which is the number of mass transfer units. This last term is the
nonclassical or path-dependent term which vanishes for infinite durations and which
should be distinguished from the first or classical term that has potential properties.
The upper (plus) sign at the last term refers to processes departing from the
equilibrium and the lower (minus) sign to the proceses approaching to the
equilibrium. With the knowledge of the classical exergy, shown in the above
equation, our numerical procedure can generate both data for A and Sa. The
enhanced bounds resulting from the finite-time exergy are illustrated in Fig. 3.

9. Concluding Remarks

Enhanced thermodynamic bounds limiting the work delivered or consumed in a finite

time constitute the main benefit reSUlting from generalized exergies. Because they
include the role of dissipative factors and the dynamic nature of bounds, these are
stronger, more useful and informative than those stemming from the classical exergy.
 382
exergy
A J=Tf
, / real equilibrium'
/T~ k ,uppl1'd
e
for vo r

o
minimum entropy production

Figure 3. Finite-time exergy A of a limiting continuous processes (N ~OO) prohibits processes

from operating below the heat-pump mode line which is the lower bound for work supplied
(lbws) and or above the engine mode line which is the upper bound for work produced (ubwp).
These bounds, more realistic than those based on the classical, reversible definition of exergy,
are the result of finite rates consistent with finite duration of the process.

To understand the problem of bounds and their distinction for the work production
and consumption, recall that the work-producing process is the inverse of the work-
consuming process (the final state of the second process is the initial state of the
first, and conversely), when we have fixed the durations of the two processes to be
the same. In thermostatics the two bounds on the work, the bound on the work
produced and that on the work consumed, coincide. However the static limits are
often too far from reality to be really useful. The generalized exergy provides bounds
stronger than those predicted by the classical exergy. They do not coincide for
processes of work production and work consumption, and they are 'thermokinetic'
rather than 'thermostatic' bounds. Only for infinitely long durations or for processes
with excellent transfer (an infinite number of transfer units) do the thermo kinetic
bounds reduce to their classical thermostatic limits. The hysteresis, or divergence of
the bounds (Fig. 3), proves that many idealized processes allowed by classical
thermostatics are prohibited by the more severe and realistic constraints 0 f
thermokinetics. The hysteretic effect, caused by the dissipation and the associated
increase of the exergy supplied in the pump mode (and decrease of exergy released
in the engine mode), reveals the extent to which thermokinetic bounds are stricter
than thermostatic bounds. The greater the mean rate we demand, the greater is the
range of performance excluded by limits predicted by the generalized exergy.
A real process which does not apply the optimal protocol but has the same
boundary states and duration as the optimal sequence, requires a real work supply
that can only be larger than the finite-rate limit obtained by optimization. Similarly,
the real work delivered from a nonequilibrium system (with the same boundary states
and duration but with a suboptimal control) can only be lower that the corresponding
finite-rate limit. Thus the two bounds for a process and its inverse, which coincide in
thermostatics, diverge in thermodynamics, at a rate that grows with any of the indices
of deviation from static behavior, fo, Sa, L or H. For sufficiently high values of rate
indices, the work consumed may far exceed the classical work; the work produced
 383

can even vanish. Thus we can confront and surmount the limitations of applying
classical thermodynamic bounds to real processes. This is a direction with open
opportunities, especially for chemical or electrochemical systems.

Acknowledgments. The author acknowledges an invitation and involvement of Profs.

A. Bejan and R.S. Berry, and a support from NATO ASI and the EC program 'Carnet'.

Glossary of Symbols

A - exergy; a v - specific area; c- molar heat capacity; fo- work generation function;
g I, g2 - partial conductances; g- overall conductance; G- molar flux of resource fluid;
H- hamiltonian; HTU- height of transfer unit; h- partial molar enthalpy of component;
1- optimal performance function; k-molar mass transfer coefficient; n- molar flux of
active component; q I-driving heat; Q-cumulative heat; t-usual time; TI, T2
-temperatures of fluid and reservoir (usually T2 =Te); ~ -constant temperature of
infinite reservoir; TI', T2' -upper and lower temperatures of engine fluid; Tn-l -
temperature of fluid entering the stage n; u n_ control vector; un =ATn/At n - rate of
temperature change; v n =AXn/At n - rate of concentration change; w n - power
delivered at stage n; W-total molar work or power per unit molar flux; X-
concentration (inert basis); Xe constant reference concentration; xn -state vector of a
general process; x- length coordinate; zn - adjoint vector; z- temperature adjoint; a.' -
heat transfer coefficient; 'Y- cumulative conductance; TI = thermal efficiency; t
-dimensionless time (xIHTU); en_ nondimensional conductance.

References

1. Sieniutycz. S. and Berry. R. S. (1999) Discrete Hamiltonian analysis of endoreversible thermal

cascades. in S. Sieniutycz and A. de Vos (eds.). Thermodynamics of Energy Conversion and
Transport, Springer, New York, Chap. 6.
2. Bejan, A. (1995) Entropy Generation Minimization: The Method of Thermodynamic
Optimnization of Finite-Size Systems and Finite-time Processes, CRC Press, Boca Raton.
3. De Vos, A. (1992) Endoreversible Thermodynamics of Solar Energy Conversion, Clarendon
Press, Oxford, pp. 29-51.
4. Novikov. I. I. (1958) The efficiency of atomic power stations (a review), Journal of Nuclear
Energy II 7. 125-128, English translation from At. Energ. (1957) 3,409-412.
5. Curzon. F. L. and Ahlborn. B. (1975) Efficiency of Carnot engine at maximum power output,
Amer. J. Phys.43 , 22-24.
6. Spalding, D. B. (1963) Convective Mass Transfer. Edward Arnold. London.
7. Ciborowski, 1. (1973) Process Engineering, Wydawnictwa Naukowo Techniczne, Warszawa.
8. Gordon 1. M. and Orlov, V. (1993) Performance characteristics of endoreversible chemical
engines. J. Appl. Phys. 74, 5303-5309.
9. Chen, L., Sun, F., Wu Ch., and Wu. 1. (1997) Performance characteristic of endoreversible
chemical engines, Energy Converso Mgmt 38, 1841-1846.
10. Sieniutycz, S. (1973) The constancy of hamiltonian in a discrete optimal process, Reports of
Insf. ofChem. Engng. Warsaw Tech. Univ. 2.399-429.
11. Sieniutycz. S. (1978) Optimization in Process Engineering. I-st edn., Wydawnictwa Naukowo
Techniczne, Warszawa.
12. Szwast. Z. (1979) Discrete Algorithms of Maximum Principle with Constant Hamiltonian and
their Selected Applications in Chemical Engineering, PhD Thesis, Institute of Chemical
Engineering at the Warsaw University of Technology, Warsaw.
384

13. Sieniutycz, S. and Szwast, Z. (1983) A discrete algorithm for optimization with a constall
hamiltonian and its application to chemical engineering", Intern. J. Chem. Engng. 23, 155-166
14. Fan, L. T. and Wang, C. S. (1964) The Discrete Maximum PrincipLe, A Study of Multistag
System Optimization, Wiley, New York. .
15. Boltyanski, V. G. (1973) Optimal Control of Discrete Systems, Nauka, Moscow.
16. Fan, L. T. (1966) The Continuous Maximum PrincipLe, A Study of Complex Systefl
Optimization, Wiley. New York.
17. Findeisen, W., Szymanowski. 1., and Wierzbicki. A. (1980) Theory and Computationt
Methods of Optimization , Panstwowe Wydawnictwa Naukowe, Warsaw.
18. Bellman. R. E. (1961) Adaptive Control Processes: a Guided Tour, Princeton University Pres!
Princeton.
19. Aris. R. (1964) Discrete Dynamic Programming, Blaisdell, New York.
20. Boltyanski, V. G. (1971) Maximum Principle and Optimal ControL of Continuous Systemj
Nauka, Moscow.
21. Bejan. A. (1996) Models of power plants that generate minimum entropy while operating!
maximum power, Am. J. Phys. 64. 1054-1059.
22. Bejan, A. (1996) Entropy generation minimization: The new thermodynamics of finite-siz
devices and finite-time processes. 1. AppL. Phys. 79, 1191-1218.
23. Sieniutycz, S. (1997) Hamilton-1acobi-Bellman theory of dissipative thermal availabilit}
Phys. Rev. E 56,5051-5064.
THERMODYNAMICS AND OPTIMIZATION OF REVERSE CYCLE
MACHINES

M.L. FEIDT
G.E.S.P.E., L.E.M. T.A., University Henri Poincare of Nancy 1
2, avenue de fa Foret de Haye, 54516 Vandoeuvre fes Nancy, France

1. Introduction

The interest in thermodynamic modeling and optimization of the reverse cycle machines
has been recently renewed, due to the C.F.C. replacement. This new consideration had
probably the start in Blanchard's paper [1]. Since that time, the endoreversible models
have been extensively considered, particularly for the two heat-source machines [2-7].
Attempts have been made in order to generalize, more or less, the particular results [8-
11]. Also internal irreversibilities have been considered, as well as heat losses, according
to Bejan's approach on entropy analysis [12-14]. Numerous papers have been published,
namely on Carnot machines (refrigerator or heat pump) [15-28]. Most of these studies
suppose that the source and the sink are isothermal. Some of them consider the effect of
the heat transfer law on the machine performance [24, 26, 27] when the same law form is
supposed at the source and the sink. Performance of more sophisticated configurations
has also been investigated recently: combined refrigeration cycle [29, 30], cascade heat
pump [31-32], Brayton machine [33, 34], solar air conditioning system [35, 36].
We propose here a tentative of model synthesis. In the first paragraph the
methodology is applied to two heat-reservoir machines with perfect regeneration (Carnot,
Ericsson, Stirling). Various optimization cases of these machines are listed, exemplified
and related to existing results. The effect of internal irreversibilities, heat losses and form
of the heat transfer laws is considered. Particular attention is given to the optimal
allocation of the heat transfer areas. Extension to the machines with real internal fluids is
presented. Finally, the case of three heat-source machines is considered showing its
relation with the preceding considerations, and how it can restore the preceding results as
particular limits.

2. Machines Following Carnot, Ericsson, Stirling Cycles

2.1. GENERAL FORMULATION OF THE OPTIMIZATION PROBLEM

The classical cycles mentioned above are shown in Fig. 1. Perfect mechanical or thermal
regeneration is supposed. It results the same basic equations for the three machines. The
First Law of Thermodynamics is written
385
A. Bejan and E. Mamut (eds.), Thermodynamic Optimization o/Complex Energy Systems, 385-401.
© 1999 Kluwer Academic Publishers.
 386
T
v=ct P=ct v=ct P=ct

Tr

Figure I. Endoreversible Carnot, Ericsson, Stirling reverse cycles.

(1)

where qc is the energy delivered at the hot side of the machine (thermostat of
temperature Tsc ; qc < 0) ; q F is the energy received at the cold end of the machine
(thermostat of temperature TSF ; qF> 0) ; W is the energy consumption of the machine
(W < 0). The Second Law of Thermodynamics is expressed by

~+~+S=O (2)
Tc TF
In this algebraic form we notice: T c, temperature of the cycled fluid at the hot end of the
machine; T F, temperature of the cycled fluid at the cold end of the machine; S, entropy
generation due to the internal irreversibility of the machine. As a first step, this last term
will be supposed constant. It can be detailed if necessary by a careful analysis, as
suggested Bejan [12]. The heat transfer description must be added to the preceding
relations, for both ends of the machine. We state that the general form of the heat transfer
laws at the source and the sink is
(3)

where Ti is the cycled medium temperature at the source (or sink) i, TSi is the temperature
of the source i, and qi is the energy transferred to the medium at the source (or sink) i.
For generality, we illustrate the method by the formal description of two heat-source
machine having direct heat leak between the source and the sink. A general form of the
heat loss energy, qJ, transferred directly from the source to the sink is
ql =G(Tsc' TSF ) (4)
The general approach taking account of all the preceding aspects uses the same
method as previously proposed [8,9].
We precise three major points of interest of the proposed method. First, this
approach is valid for reverse cycle machine, as well as for direct ones. Second, this
approach is valid for discontinuous cycle machine (time duration for each
thermodynamic process), but also for continuous ones (i.e., in two heat-source machines,
the contact durations of the cycled medium with the heat sources are replaced by the heat
exchange areas). Third, the model is valid tor any form of the heat transfer law used in
 387

Eg. (3). Unlike the literature, where the heat transfer law is supposed to be the same at
the source and the sink, we note that Eq. (3) permit to use different heat transfer laws in
the models.

2.2. PARTICULAR CASE OF REFRIGERATOR AND HEAT PUMP

The objective function for a refrigerating machine can be (1) the minimum of the
consumed energy (case R I) or (2) the maximum of the useful heat coming out of the cold
source (<I> > 0, case R2). For this refrigerating machine a new constraint has to be added
at the cold source
(5)

The objective function for a heat pump can be (1) the minimum of the consumed
energy (case HI) or (2) the maximum of the useful heat delivered to the hot sink (<I> < 0,
case H2). For the heat pump a new constraint appears also, expressed at the hot sink as
(6)
In addition to the above mentioned constraints, a new one must be considered. It is
given by the imposed <I> (case Rl and HI) or imposed W (case R2 and H2). Generally,
for a refrigerating machine (respectively for a heat pump) TSF (respectively Tsd is
imposed. For the two machines, we can also consider the cases with both TSF and Tsc
parameters. In these cases remain only two variables related to the two constraints (2)
and (4), and there is only a thermodynamic state that can be attained, not an optimum.

TABLE I. The equations for the optimum of the two-source machines. Variables are the two
cycled medium temperatures.

Case Equation

RI -2 1_(TFJ)o.~ - J)o) =- 21_[TeFc.e - Fe( 1+ Fe.s )]

TFJ)o.F TeFe.e Gse

J)o I Fe ( Fe.s ) I
R2
TJI),.F - TF = T2Fc.e 1+ Gse - Te

1
- 2 -- [TFJ)oF- J)o(I-.!h)] =-2-1-(TeFe.e -Fe)
HI TFI),.F Gse TeFe.e

~(I_ FF.S )_....!...=~_....!...

H2 TJI),.F GSF TF T2Fe.e Te

In the general case, applying the Lagrange's multipliers method leads to a system with
three equations having the three cited unknowns (Table 1)

F. = __' aF aF
1.1 aT i
F,.S =--' aTsi
When the heat losses are not considered, all the equations of Table I become
identical.
388

Owing to the presence of Eq. (2) in any cases, it results a unique system of two
equations that yields the optimum values of Tc and TF as functions of the other
parameters. Particularly, one can see that the optimum Tc and TF depend on the internal
irreversibility, Eq. (2). If S vanishes (endoreversible case), the system restores the
endoreversible results currently exposed in the literature. There must be noticed that the
final solution differs from case to case, due to the specificity of the last equations used
«I), (5), or (6». These equations can be solved analytically in some cases, and only
numerically in the remaining ones.

2.3. OPTIMAL ALLOCATION OF HEAT TRANSFER CONDUCTANCES

Following the approach presented in a previous paper [8], the same Lagrange's method
can be applied to solve the cases where the conductances of the heat exchangers (or
contact duration times) are variables. In these cases, one more relation must be added to
the two preceding ones. It will relate the two new variables which are the overall heat
transfer conductances at the source and the sink
or (7,8)

where Kc is the thermal conductance at the hot end, KF is the thermal conductance at the
cold end, tc is the thermal contact duration at the hot end, and tF is the thermal contact
duration at the cold end of the machine.
For the continuous cycles, Eq. (7) expresses the total heat transfer conductance to be
distributed between the hot and the cold end of the machine that is considered in a
stationary state. Eq. (8) is useful for discontinuous cycles and supposes that the duration
of the transfer between the source and the sink is negligible.

TABLE 2. Extended optimization of two heat-source machines. Variables are

the two cycled medium temperatures, and the overall heat transfer
conductances at the source and the sink.

Problem Objective function Constraints

RIE minW (2), (5), (7) or (8)
R2E MAX</> (2), (I), (7) or (8)
R3E min Kort (2), (5)
R4E min K ort (2), (I)
HIE minW (2), (6), (7) or (8)
H2E min</> (2), (I), (7) or (8)
H3E min Kort (2), (6)
H4E min Kort (2), (I)

With the two supplementary variables related by Eq. (7) or (8), the optimization
problem will have two degrees of freedom. The corresponding form of the heat transfer
laws at the source and the sink becomes

(9)
Table 2 shows that the added Eqs. (7) or (8) make possible two new optimization
processes (R2E, R4E for a refrigerating machine, and H3E, H4E for a heat pump) for
each machine, as originally was proposed by Bejan [19]. We notice that the cases R4E
 389

and H4E are formally identical. They differ only by the third variable, (Tsc, TSF), if it is
considered. By derivation and direct elimination, one can derive a general relation
between the two variables T c and T Ffor minimum of K

(10)

For these cases, it appears that the minimum of K depends only on the internal irreversibility,
and not on the heat losses. This general expression restores easily the known particular results.
For the case R3E (H3E), the same approach shows that K is a monotone function of T c (TF),
and these two variables are related by

(11)

or

(12)

2.4. OPTIMAL ALLOCATION OF AREAS

This extension is valid essentially for continuous cycle machines (stationary operating).
It uses a more realistic relation than Eq. (7) given by
(13)

where A is the total heat transfer area to be allocated, Ac is the heat transfer area at the
hot end, and AF is the heat transfer area at the cold end of the machine.
Using these new variables, the generalized overall heat transfer coefficient at the hot
end kc, respectively at the cold end kF' must be known. As a first step, we suppose that
these coefficients are constants. Consequently, the general form of the heat transfer laws
at the source and the sink becomes

(14)

Further, the optimization procedure can be extended taking account of the investment
cost of the heat exchangers. As a first approximation, the investment cost is expressed
with reference to the heat exchanger areas
(15)
where Vc is the cost by unit area at the hot end, VF is the cost by unit area at the cold end
of the machine. Even if detailed cost correlations are available [46], this approach offers
a good overview of the corresponding tendencies.
Eq. (15) will replace Eq.(7) and it can be used also to find the optimal allocation of
the conductances, when the heat transfer coefficients, ki ,are equals to unity. Similarly, if
the unity costs Vc and VF are considered equals to unity, Eq. (15) restore Eq. (13). The
problem considering Vc and VF as parameters has been already studied [2].
Combining Eqs. (15) and (2) results the general expressions of the heat transfer areas
at the optimum
390

(16)

-(vkefe +SveTe)TF
AF =-:...--=--=---=--=...:.--=- (17)
ve TekFfF - v FTFkefe
where the values of T e, T F correspond to those obtained from the system having Eqs. (I),
(5) or (6) as constraints and the objective function owing to the current problem (case I
or 2). If the objective function is the investment cost v (case 3 or 4), the corresponding
expressions are given in Table 3. It results the effect of 4>, G, S variation and of the
cycled medium temperature rise (Tc-TF) on the investment cost in various cases. We
consider this approach particularly important because of its potential to differentiate
wide-spread thermal machines such as air-air or water-water systems.

TABLE 3. Thennoeconomic optimization of two heat-source machines: minimum

of investment cost for heat exchangers.
Problem ExPression 0 the objective function

R3G v=(<I>+G{~_~.Te
kFfF kefe TF
)-s veTe
kefe

R4G v= __I_[~(_W +STF )+ vFTF (w -STe )]

Te-TF kete kFfF

H3G v=(<I>-G{~-~. TF )-S vFTF

kefe kFfF Te kFfF

H4G the same eq. as for R4G

These systems are supposed to use similar heat exchangers technologies at both ends
of the machine, while the air-water or water-air machines will have different specific
costs for the air system, Va, respectively, water system, vw • Studies on this subject are now
under development in our research group [47].

2.5. SUMMARY OF RESULTS

In order to illustrate the general methodology proposed in the preceding paragraphs, we

propose hereafter a tentative synthesis of the available results. Table 4 presents the
optimum values of the four variables for a refrigerating machine, when the heat losses
and internal irreversibilities are considered. The studied cases correspond to a fixed
useful effect and to K p , Tsc, the model parameters.

S
where s=--
AkF
n _ Tse
0e - is a hidden parameter in <po
TSF
 391

TABLE 4. Optimum operating point of a refrigerating machine with

heat losses and internal irreversibilities.

Conductances Surfaces
..!L= -2q>+1 ..!L = kl/2 -q>(1 + k1/2)
TSF I+s
TSF kl/2 +s
l£...=~
Tsc 4q>+s-1 l£... = k1/2[ q>(1 + kI/2)_ kl/2]

Tsc s - klf2 + q>(1 + k 1/2t

KF = q>{I+s) q>(k"2 +s)

K 2q>+s -AF _
-
A q>(I+ k"2)+s
Kc q>{I-s)+s
K 2q>+s Ac = q>{I- s)+ s
A q>(1 + kIl2)+ s

These analytical results show clearly the influence of <p and s on the optimal conductance
distribution. We notice that generally conductance equipartition is not valid. It is true
only for the endoreversible case.
If the optimization is extended to the area distri bution, the influence of the global heat
transfer coefficients appears through a square root (Table 2, 2nd row). In the same
manner, it is easy to deduce the optimum operating point for a refrigerating machine
when W is imposed, the corresponding results for a heat pump, and the associated values
of the COP at optimum. The results are available, but they will not be presented here for
brevity.
Table 5 illustrate the optimization of a reverse cycle machine for two generalized heat
transfer laws. Law I is relative to a generalized radiative law
(18)

Law 2 is relative to a generalized convective law

(19)

where Ei is chosen according to the convention stating that qc remains always negative,
and qF positive, whatever the exponent m values would be, and Xi = TSi - T i .
The most common value of the exponent n is 4, while m could indicate natural
convection (m = 0.24), or pool boiling (m = 3). Also, the value n= - I can be find in the
literature. It corresponds to the conductive phenomenological law of the linear
irreversible thermodynamics. The two forms given by Eqs. (18) and (I9) admit the same
limit, for n = m = I which corresponds to Newton's convection law, well developed and
known case in the literature.
From Table 5 results that the optimization problem can be generalized. The system of
four equations to be solved with respect to the variables Ac, AF, Tc or X, TF or Y is
constituted for any case of the general equation
392

TABLE 5. The system of four equations to be solved for the optimum (Wo. qo. A).

Lawl I Law 2

r r
Refril!eratinl! machine

kF(T;;' - T;' (-I)"k Xn+1 k ym+l

kc(Tsc - Tc + F -0
xi
C
=0 (Tsc- (TSF _ y)2 -
mT;'+l nTc+ 1
(-I)"XnkcAc ymkFAF
k~F(1SF-1;') kcAc(~ -lC) +---+s=o
+ +s=o Tsc- X TSF-y
TF Tc
Ac+AF-A =0
Ac+AF-A =0
Wo-(-I)"kcAcXn +kFAFym = 0
\\f,+kcAc(~-lC)+k~~1SF-1;')=o
or
or
qo-kFAFym =0
qo - kFAF(T;;' - T;) = 0
Heat pump

kF(Ts";, -T;,t kc(T;c-Tcf kFym+l (-I)"kc Xn +1

=0 2 =0
mT;'+l nTc+ J (TSF _ y)2 (Tsc -X)

k~F(1SF -1;') kcAc(~ -lC) k~Fym (-I)"kcAc Xn

++S=O
+ +S=O
TF Tc TSF-y Tsc- X
Ac+AF-A =0 Ac+AF-A=O

W(,+kcAc(~-lC)+k~~~-1;')=O Wi,-(-I)nkcAcXn+kFAFym =0
or or

qo - kcAc(T;c - Tc) = 0 qn -(-I)"kcAcX n = 0

(20)

The second equation is the expression of the second law of thermodynamics. The two
other equations are relative to the total area to be allocated when W or qi is fixed. When
the minimum of A is sought, there is only one constraint (W or qi fixed), as it was told
previously.
Important and interesting results can be deduced from Table 5. When W or qi is
fixed and S = 0, the general distribution of the heat transfer area is given by
1 B
Ac=A-- A =A-- (21,22)
I+B F I+B
where B is a given function of (ke I kF)112, Te, TF and of the heat transfer law form.
It is obvious that the equipartition of the heat transfer conductances is not a optimum
distribution. The expression of the ratio Kc/KF relative to the non symmetrical
generalized radiative laws illustrate this fact
1/2
KF _ ( n T~n-1)12
(23)
Kc -
)
m' T(m-l)/2
F
 393

It appears that, even for endoreversible machine, the equipartition of the heat transfer
conductances is obtained only for the symmetrical case of the linear heat transfer law (n
= m = 1). If the heat transfer law is symmetrical, but not linear, the conductance ratio
depends generally on the temperature rise of the cycled medium.

3. Machines with Real Internal Fluid

3.1. PHYSICAL PROPERTIES OF THE REFRIGERANT FLUID

A new step in the model development consists of considering the real properties of the
refrigeration fluids. This corresponds to a recent work done in collaboration with V.
Radcenco [46], now in press.
Correlations for three fluids (R12, R22, R134a) are proposed. R12 is a C.F.e. not
allowed by now, R22 is a H.C.F.e. for transition, and RI34a is the most common H.F.e.
proposed to replace the preceding ones.
The model requires the form of the latent heat of boiling or condensing, proposed as
(24)

Also, the saturation pressure correlation, with reference to the normal boiling
temperature TN (at PN= 1.03105 Pa) is needed

In(p)=A(~T _TN_
1 )+Bln(~)+C(T-TN)
TN
(25)

where ex, ~, y, A, B, C are the coefficients given in [48].

3.2. REAL MACHINE MODEL WITH VAPOR COMPRESSION

The vapor compression machine is one with regeneration, so that the subcooling of the
liquid refrigerant (3-3") is done by the vapor superheating at the exit of the boiler (1-1 J.
Fig.2 illustrates the complete evolution of the refrigerant during the cycle, where T. is the
ambient temperature, T, is the temperature of the superheated vapor entering the
compressor. The heat exchanged in the boiler and the condenser is expressed according
to Newton's heat transfer law, in term of conductances or areas. The compressor is
supposed adiabatic.

s
Figure 2. Temperature-entropy diagram for a real refrigerating machine.
 394

Introducing the mean entropic temperature

(26)

the entropy balance for a cycle results

.9l:. _qo =_1_ .
Tm 1D CI1()

(27)
where Tr is the temperature of the liquid refrigerant at the exit of the recuperator, H is the
pressure ratio (H = PclPo), and 3 e is the isentropic efficiency of the compressor.
Based on the entropy analysis methodology, an expression of the entropy flux
generation for stationary operating of the machine is deduced in this case. For brevity the
details regarding the nondimensional parameters and variables will be omitted, as well as
the constraints and the intermediate function f, (i = I to 7). The objective function to be
minimized is

P=(l-~).e
e o
m
't-q () (28)

It depends on the three reduced variables

e = Tf-To e = Tc -Ta
o Tf ' C Ta '

with the following explicit parameters

't=-
Ta qo
T, ' qo = KT
f

Actually the optimization problem has five variables if we also add 1<0, Kc and twelve
parameters. Two of the parameters are related to the application ('t, qo), other two to the
machine ( 't s = LlTs / T f ,3 c ), and eight are related to the refrigerant fluid. The variables
K, can be eliminated as it was indicated previously. and there remain only two equations
relating the three temperatures

c(f7 -1)+ f6
(29)
cln(f7 )+f6
 395

From Eq. (29), em can be expressed as a function of ec, eo, with c = CrfC. It remains
only one objective function, Eq. (28), and one constraint, Eq. (30). Due to the non
linearity of the entropy balance, only numerical solution of the two equations system can
be obtained.

3.3. SENSITIVITY ANALYSIS OF THE OPTIMUM

There is well known the temperature Tr range for air conditioning, 283 K-293 K, for
refrigeration, 268 K-278 K, and for freezing, 243 K-263 K. The vapor superheating
varies from 1°C to 10 0c. If the ambient temperature reference is supposed to be 298 K,
we found that 3.6 10.3 < 'ts < 4 10-2 , and 1.07 < 't < 1.35. The isentropic efficiency of the
compressor varies from 0.6 to 1 (the limit case). A realistic range of variation for qo is 10-
3 < qn<IO-I.

Here after the optimum values variation of the Ko/K, en, e c versus qo are presented
when 't = 2981273, 't, = 51273, and ::\ = 0.8 , for the three proposed refrigerant fluids.
Fig. 3 shows that when qo increases, the departure from the equipartition of the heat
transfer conductance increases. The results for R 134a and R22 are very closed, but for
R12 the curve shows that less conductance is needed at the cold end. The pinch is
practically the same for the three refrigerants at the cold end of the machine. It is not the
same at the hot end, where for R22 the pinch values are smaller. In each case, eo, and e c
are increasing function of qo.
The trend of reestablishing the equipartition of the conductances, when ~\ increases
(lowering the internal irreversibility of the machine) has been confirmed.

4. Presentation of Other Machine Configurations

4.1. INVENTORY OF STUDIED MACHINE CONFIGURATIONS

Most of the recent published studies are focused on Carnot refrigerating machine or heat
pump with different driving systems [36-38]. An attempt was also made for a cascade
machine [31], but generally these studies are limited to endoreversible system. An
interesting confrontation between experimental and industrial available results and a
thermodynamic model was developed by Gordon and co workers since 1993 [40-43].
This comparison is relative to the reciprocating chillers, being also extended to
absorption chillers.
Before developing this last machine optimization, just a comment on the other studied
configurations appearing recently in the literature. Chen and al. [34] considered an
endoreversible Brayton refrigerating machine and the relation between the COP or
specific load, and the pressure ratio. The same approach was developed by Erbay and
Yavuz [47, 48], but relative to the Ericsson and Stirling machines. Similar studies were
performed in our research group [49] in order to obtain the minimum W for given qF for
a refrigerating machine or qc for a heat pump. Brayton, Ericsson and Stirling cycles were
considered with perfect heat regeneration, but with global internal irreversibility (S), and
heat leak between the hot end and the cold end of the machine. Due to its growing
practical interest, the case of three heat-source machines was also envisaged, and it is
shortly reported hereafter.
396
49r-----------~------------._------------r_----------~------------~

48.

. . . . . . . . . . . ~.:::~:~J.'.
47. . .................. :,........... ;- .. . . ..":""""';"""

47 ~ ~
. ......... ... ........... .

48.
'.
. ... ""' ..,: .. ~.. - .. - - ..
::

48 .................. ;. ...... ~,:'.~.~ .. " ....

'. ..... :... _ ... -.. _. _ .
45. . .... _ ........ :,:.~

.~
45L-________~~------------L-----------~------------~------------.J
O.OOG 0.01 0.015 0.02 0.03

(a)

0.07

0.08

0.05

0.04

80
0.03

0.02

0.01

00

(b)

0.07 r---------~----------,_--------~----------~--------~~--------~

0.06

0.05 ... . . ... . .....:... -...... -. - ... ~ .

0.04 .,-:- ... .

0.03 .•.......... __ . __ ................ .

0.02

0.01

0~0~~~--~0~.0~0~5~------~0~.0~1------~0~.~0~1~5~----~0~.~0~2~------~0~.0~2==5------~0~.~03

(c)

Figure 3. Evolution of Ko/K (a), So (b), Se (c) versus qo, for R 134a, - - R22, -.- R 12.
 397

4.2. REFRIGERATORS AND HEAT PUMPS IN CONTACT WITH 3 OR 4 HEAT

RESERVOIRS

4.2.1. COEFFICIENT OF PERFORMANCE VERSUS USEFUL EFFECT

Several papers were published on the three heat-reservoir machines [52-54]. Generally
the intermediate contacts are assumed with the same reservoir (reduction from 4 to 3
heat-reservoir machines). This assumption constitutes a good approximation of the
technological configurations. We notice only a paper of Chen and Yan proposing a
unified description of the endoreversible cycles [51]. A recent review paper [54] focuses
on a more general model of the three heat-source machines that can be conceived owing
to the published materials.
The general relation between the coefficient of performance and the useful effect was
derived using the algebraic forms of the first and second law of thermodynamics
(31)

k+~+~+s=o (32)
Tc TI TF
The coefficient of performance. COP. is defined as the ratio of the useful effect and
the energetic cost. It is expressed respectively

for a refrigerating machine 1:=_q_F_ (33)

qc+ W
. -ql
for a heat pump 1: = - - (34)
qc+ W
" qc
for a heat transformer 1: = - (35)
q,

Combining Eqs. (31). (32) with the definitions given by Eqs. (33)-(35) we derive for
real irreversible machines
~ _ TI-I-Tcl [ W-TcS ._T_ 1 -]
(36)
"'I - _I _I I +---=--
TF -TI W+qc Tc -TI

(37)

(38)

Complementary to the proposition of Chen et al and Gordon et al. the COP's appears
to be naturally expressed in term of the heat input. It is an interesting result. especially
when the heat input is fixed (constraint). Chen et al and Gordon et al consider the useful
effect as constraint. We propose here the corresponding relation between the COP and
the useful effect
 398

1 W-Tc S ._'[_1_
-=- (39)
1: I '"
qF Tc -TI
1 W-TcS TF
(40)
~= ",' - qI Tc - TF
1 _ 1 W-TIS TF
?;"" -7 - ----q;;-. TI - Tc (41)

where "'. ",'. "," correspond to the three thermal machines COP. generally indicated in the
literature (endoreversible machines with W contribution neglected).

4.2.2. OPTIMAL AREA DISTRIBUTION OF THE MACHINE

The three heat transfer quantities are given in a general form by Eq. (3)

It is obvious that in this case the size of the system is finite (or the duration of the
machine cycles is finite. the formalism being strictly identical. so A; becomes t;)

(45)

Also we notice that Eqs. (31). (32). and (45) constitute a system of equations. with
three unknowns. Ae. A.. AF. These variables are expressed in function of the temperatures
and the case parameters

A c =-
.1
1[
Akfk f -
I IFF T -T
- +W(1 1) (k
FfF
T ---
I fl ) (
T - -Sk Ff F -kf
I I
)] k (46)
F I F I

I .1
1[F FCC Tc TF
(1 I)
A = - Ak f k f - - - + W(kcfc
----
Tc
kFfF)
- -S ( k f - k f )]
TF C C FF
(47)

A F=
.1
I[
- Ak Cf C
TI
( I 1)
k IfI - -T- +W(klfl
-T- -kcfc)
- ( f )]
T - -S k IfI -k CC (48)
C I C

The above equations show that the area distribution depends on the three variables
(Te• T I • TF) and on the model parameters (Tse• Ts.. TSF• kc. k.. kF• W. S).We renew that the
heat transfer laws are not necessary the same at the three thermal contacts. even if the
assumption of similarity is always used in the literature. The influence of the heat transfer
law form appears through the functions fe• fl. fF in Eqs (46)-(48). It remains to precise the
objective function and the constraint(s). or eventually the COP. in order to optimize the
machine.
 399

Eqs. (46)-(48) represents the generalized formulas of the area distribution in the
machine, indicating the influence of Wand S on that. If Wand S are equal to zero, the
particular results published previously [54] are obtained.

4.2.3. PARTICULAR CONFIGURATIONS

As it was demonstrated in a recent paper, it is easy to restore the form of Eq. (16), (17)
corresponding to the limit of two heat-reservoir refrigerating machine or heat pump. In
these relations W vanishes, Tsc and Tc tends to infinity, qH becomes pure mechanical
energy. The corresponding positive value of W results from the finite value of the
undetermined form Ac (or tc) ~ 0, multiplied by Kcfc that both tends to infinity.
In the same manner, it is possible to show that the two heat-reservoir heat engine
corresponds to the limit of a heat pump, when Ts, tends to infinity [54].

4.2.4. ILLUSTRATION OF SOME OBTAINED RESULTS

Based on similar, but particular approach, there have been studied more sophisticated
machines, as heat pumps driven by thermal machines or solar three heat-reservoir
machines. We focus here on the finite thermal capacities machines. The preceding
analysis of a solar system indicated that generally, the heated fluid in the solar collector
is not subject to phase change. It means that it cannot be considered as isothermal
(Tc~t). This fact has been considered [51] taking account of the finite thermal capacity
of a system under transient conditions.
The same approach was adopted for a stationary system [4], where it was taking
account of the finite mass fluxes of the external fluids. The corresponding results are
available [53].

5. Conclusion

The proposed paper intended to achieve three goals. First, it proposes a review of the
major fields of interest in recent thermodynamics approach and optimization of the
reverse cycle machine. This interest is probably due to the replacement of C.F.C. that
requires the reconsideration of the conception of these machines. It is noticed that
thermodynamics was often limited to endoreversible consideration until now.
Second, this paper aims to offer a general methodology. The exposed general model
is able to take account of (1) the internal dissipation of the energy, by a global or refined
form (paragraph 4.2), (2) the heat losses between the source and the sink, by a general
linear form, (3) the general heat transfer laws that can be different between the hot
(intermediate) and the cold end, when in the literature there is explored only the
symmetrical case.
Third, there have been developed some examples in order to enlighten the proposed
method. These examples have been particularly focused on the optimal allocation of the
heat transfer areas, subject to various objectives and constraints. It has been shown that
this approach can be easily transposed to the contact duration or heat transfer
conductances, and that the methodology is also valid for thermoeconomic analysis of
these machines.
400

6. References

I. Blanchard, C.H. (1980) Coefficient of perfonnance for finite speed heat pumps, J. Appl. Phys.,
51, 2471-2472.
2. Goth, Y. and Feidt, M. (1986) Conditions optimales de fonctionnement des pompes Ii chaleur ou
machines Ii froid associes a un cycle de Camot endoreversible, C.R.A.S., T. 303, serle 11,19-24.
3. Feidt, M. (1986) Un outil d'aide Ii la conception des machines Ii froid et pompes de chaleur Ii partir
de la modelisation des cycles de machines a deux et trois sources de chaleur. Optimisation
resultante, compte tenu des contraintes en temps de cycle ou en surface d'echange, Rencontre
S. F. T., Marseille, France.
4. Feidt, M. (1988) Finite time thermodynamics applied to optimization of heat pumps and
refrigerating machine cycles, 12th lMACS World Congress on Scientific Computation, Paris,
France.
5. Philippi, I. and Feidt, M. (1991) Optimisation des machines thenniques Ii deux sources, cas des
faibles pincements, Entropie, 164, 113-114.
6. Bejan, A. (1989) Theory of heat transfer irreversible refrigeration plants, lnt. J. Heat and Mass
Tran.~'-er, 32, 1631-1638.
7. Gordon, I.M. (1990) The impact of heat transfer mode on the efficiency of heat engine operating
at maximum power, 9th. Int. Heat Tran.~'-er Co"f"erence.
8. Feidt, M. and Philippi, I. (1992) Basis of a general approach for finite time thennodynamics
applied to two heat-source machines, Proc. of ECOS'92, Zaragoza, Spain.
9. Feidt, M. (1993) Thennodynamique en temps fini des cycles moteurs ou generateurs de froid et de
chaleur, Proc. '?f"Thermodynamics National Co"f"erence, Bucharest, Romania.
10. Philippi, I. and Feidt, M. (1991) Finite time thennodynamics applied to inverse cycle machine,
XVIII-erne C,mgres Inernational du Froid, Montreal, Canada.
II. Feidt, M. and al. (1994) Synthesis unification and generalization of preceding finite time
thennodynamics studied devoted to Camot cycles, Proc. of FLOWERS '94, 429-441.
12. Bejan, A. (1994) Entropy Generation Through Heat and Fluid Flow, Wiley & Sons, New York.
13. Feidt, M. (1996) Thermodynamique et Optimi.wtion Energitique des Systemes et Procedes, Tech
& Doc, Paris.
14. Feidt, M. (1997) Sur une systematique des cycles imparfaits, Entropie, 205, 53-61.
15. Petrescu, S. and al. (1993) The study for optimization of the Camot cycle which develop with
finite speed, Proc. o'-the Inter. Sympos. ECOS'93, Kracow, Poland.
16. Crazzini, G. (1993) Irreversible refrigerators with isothennal heat exchangers, Rev. Int. Froid, 15,
101-106.
17. Yan, Z. and Chen, 1. (1990) Coefficient of perfonnance for an irreversible Camot heat pump,
Proc. (?,-FWWERS'90, 193-200.
18. Klein, S.A. (1992) Design considerations for refrigeration cycles, Rev. Int. Froid, 15, 181-185.
19. Bejan, A. (1993) Power and refrigeration plants for minimum heat exchangers inventory, J. (!,-
Energy Resources Technology, 115, 148-150.
20 Yan, Z. and Chen, 1. (1990) A class of irreversible Camot refrigeration cycles with general heat
transfer law, J.(!f" Appl. Phys., 23, 136-141.
21 Agrawal, D.C. and Meson, V.I. (1990) Perfonnance of a Camot refrigerator at maximum cooling
power, J. of Appl. Phys., 23, 136-141.
22 Agrawal, D.C. and Meson, V.I. (1993) Finite Camot refrigeration with wall gain and products
loads, J. ofAppl. Phys.. 74, 2153-2158.
23 Chen, L. and al. (1997) A generalized model of a real refrigerator and its perfonnance, Appl.
Therm. Eng., 17,401-412.
24 Chen, L. and al. (1996) The influence of a heat transfer law on the endoreversible Camot
refrigerator, J. Int. Energy. 69, 96-100.
25 Wu, C. and al. (1996) General perfonnance characteristics of a finite speed Camot refrigerator,
App/. Therm. Eng., 16, 299-303.
26 Chen, L. and al. (1996) Heat transfer effect on the specific cooling load and refrigerators,
App/.Therm. Eng., 16,989-997.
27 Chen, L. and al. (1996) Heat transfer effect on the specific cooling load and refrigerators,
App/.Therm. Eng., 17,103-110.
28 Sun, F. and al. (1997) Optimal perfonnance of an endoreversible Camot heat pump, Energy
Converso Mgmt., 38, 1439-1443.
 401

29 Chen, L. and al. (1997) Steady flow combined refrigeration cycle performance with heat leak,
Appl.Therm. Eng., 17, 639-645.
30 Chen, L. and aI. (1995) Optimization of the specific rate of refrigeration in combined refrigeration
cycle, Energy., 20, 1049-1053.
31 Chen, L. and Wu, C. (1995) Performance of a cascade endoreversible heat pump system, Int. J. (It
Ambient Energy., 68,137-141.
32 Wu, C. and al. (1997) Finite time thermodynamic performance for a class of irreversible heat
pumps, ASME, ASIA '97 Conterence, Singapore.
33 Wu, C. and al. (1996) Optimization of a steady flow refrigeration cycles, Int. J. (II Ambient
Energy., 17, 199-206.
34 Chen, L. and al. (1997) Finite time thermodynamic performance of an isentropic closed
regenerated Brayton refrigeration cycle, Int. J. of Energy Environment Economics, 4, 261-274.
35 Wu, C. (1997) Specific heating load of an endoreversible Carnot heat pump, Int. J. (II Ambient
Energy., 14, 25-28.
36 Wu, c. (1993) Performance of a solar engine driven air conditioning system, Int. J. of Ambient
Energy., 14, 77-82.
37 Davis, G.W. (1995) Finite time thermodynamic analysis of an internally irreversible heat engine
driven heat pump system utilizing low grade energy, Proc. of Thermodynamics and Design,
Analysis and Improvement 01 Energy Systems, AES-35, ASME, 93-98.
38 D'Aponte, F. and al. (1995) Influence of irreversible heat transfer on heat engine driven heat pump
performance, Pmc. of Thermodynamils and Desixn, Analy.,is and Improvement of Enerxy
Systems, AES-35, ASME, 29-34.
39 Lu, P.c. (1995) Thermodynamics with finite heat transfer area of finite surface, Proc. (If
Thermodynamics and Design, Analysis and Improvement of Energy Systems, AES-35, ASME, 51-
60.
40 Gordon, J.M. and al. (1997) Optimizing chiller operation based on finite time thermodynamics:
universal modeling and experimental confirmation, Int. 1. Refrig., 20, 191-200.
41 Chua, H.T. and al. (1996) Experimental study of the fundamental properties of reciprocating
chillers and their relation to thermodynamic modeling and chiller design, Int. 1. 01 Heat and Mass
Tran.'fer, 39, 2195-2204.
42 Gordon, I.M., NG, K.c. (1995) Predictive and diagnostic aspects of a universal thermodynamic
model for chillers, Int. J. of Heat and Mass Transfer, 38, 807 -818.
43 Gordon, J.M., NG, K.C. (1993) Thermodynamic modeling of reciprocating chillers, J. olAppl.
Phys .. 75, 2769-2774.
44 Grosu, L. and al. (1997) Contribution a I'etude exergoeconomique des systemes de refrigeration,
Pmc. ofBIRAC'97 Cont:, Bucharest, Romania.
45 Grosu, L. (1998) Contribution to exergo economy of refrigerating machine, Pmc. of ECOS '98,
Nancy, France.
46 Metrot, A. (1996) Etude comparative de I'optimisation des machines a cycle inverses, Engineering
student report, ENSEM, Nancy. France.
47 Erbay, L. and Yavuz, H. (1998) Maximum heating density of a Stirling heat pump, Proc. of
ECOS'98, Nancy, France.
48 Yavuz, H. and Erbay, L. (1998) General performance characteristics of an Ericsson refrigerator,
Proc. olECOS'98, Nancy, France.
49 Feidt, M. and al. (1995) Optimisation des machines a froid et pompe a chaleur fonctionnant selon
des cycles varies (Carnot, Stirling, Ericsson, Brayton ; machines a 3 sources), Internal report,
ENSEM, Nancy, France
50 Diny, M. and al. (1995) Optimisation thermodynamique des machines 11 froid trithermes, Proc. of
the 19th 1.1.R.Congre.<.<, Hague, Netherlands, 147-154.
51 Chen, J. and Van, Z. (1989) Unified description of endoreversible cycles, Phys. Rev., 39, 4140-
4147.
52 Feidt, M. and Benelmir, R. (J 992) Premiere approche de I'optimisation des echangeurs de
machines trithermique, Journees S.F. T., Paris, France
53 Diny, M. and al. (1998) Optimal performance of three heat-source machines, Cryoxenics, in press.
54 Feidt, M. (1998) Tentative unified description of engines, refrigerators and heat pumps in contact
with two or three heat reservoirs, 1. of Energy, Environment, Economics, in press.
SYNTHESIS ON STIRLING ENGINE OPTIMIZATION

M. COSTEA I, M. FEIDT2 and S. PETRESCU3

I Department of Applied Thermodynamics

Polytechnic University of Bucharest
Splaiullndependentei 313, Bucharest, Romania
2 L.E.M. T.A., University Henri Poincare of Nancy 1
2, avenue de la Foret de Haye, 54516 Vandoeuvre, France
3 Department of Mechanical Engineering
Bucknell University, Lewisburg, PA 17837, USA

1. Introduction

The Stirling cycle analysis is a topic of interest to those concerned with the performance
of thermal machines. The current trend in this analysis is given by the evaluation of the
losses due to the irreversibilities in the cycle and their effect on the machine
performance.
Recent books had major contributions to the Stirling machine analysis [1-4]. Others
presented optimization models of endoreversible and exoirreversible cycles [5-8]. The
effects of the non linear form of the heat transfer law (i.e., radiative) related to the solar
applications [9-12] were also investigated.
Our analysis of a Stirling engine irreversible cycle has followed several steps [13-17]
aiming at the optimization of the engine performance, particularly of the heat transfer
surface conductance or area distribution among the machine heat exchangers. The
optimization criterion was the maximum engine power output. A synthesis of some
results produced by the method is presented in this paper.

2. General Presentation of the Optimization Method

2.1. THE STIRLING ENGINE CYCLE WITH IRREVERSIBILITIES

The Stirling engine cycle is shown on T-S diagram in Fig. 1. The diagram is modified to
include the effects of heat transfer through a temperature difference at the source and
sink, machine thermal losses and for incomplete regeneration. The finite thermal
capacities of the source and the sink are suggested by their continuos temperature
variation. Additional heat from the external source is shown to be needed in the process,
Qx-2, due to incomplete regeneratio:J. The H; notation on the source temperature
variation identifies the separation between the two heat exchangers.
403
A. 8ejan and E. Mamul (eds.J, Thermodynamic Optimization of Complex Energy Systems, 403-410.
© 1999 Kluwer Academic Publishers.
404

Tu

Figure I. Stirling engine cycle with irreversibilities.

Similarly. the unregenerated heat is shown being rejected in the process. Qy-4. and
the Lj notation corresponds to the separation between the two heat exchangers at the sink.
The heat losses of the engine are shown between the extreme temperature levels of the
source and sink. TH/ and TL/. Note that when TSI = TS2 • the source and the sink become
thermostats.
The heat transfer rate available at the source replaces in our model the source of
constant temperature. usually considered. Thus. the source temperature can also be
optimized.

2.2. MATHEMATICAL MODEL

The Stirling engine is analyzed using a mathematical model based on the First and
Second Law of Thermodynamics applied to the cycle. and the energy balance at the
source. By hypothesis. the regenerated heat corresponds to the temperature difference
( T.. - Tv or Tx - Tc)' Thus. the heat regeneration factor introduced in the model is
defined [18] as
T -T T.. -T,.
11 H =_"--" =---'- (1)
TK - T.. T... -T..
Note that this definition is the same as the heat exchanger effectiveness E.
Generally. the Stirling machine regenerator can be considered as a counterflow heat
exchanger. In order to make the approach more general. a parallel-flow heat exchanger
 405

was also considered. Taking account of the fluid flow in the machine regenerator, it
results that Cmat = Cmin • Hence, adjusting the effectiveness expressions of a counterflow
and a parallel-flow heat exchanger [19] yields

1
£ =1----- (2)
R.cf 1 + NTU R

1- exp[ -2NTU R]
cR.pj = 2 (3)

where NTUR is the number of heat transfer units corresponding to the regenerator.
Between the extreme temperature levels of the source and the sink, the heat losses are
accounted for by an additional equivalent heat transfer conductance:

(4)

To make the results more general, the model was conceived to investigate the
influence of the heat transfer law form on the engine performance. The two nonlinear
forms of the heat transfer law at the source and sink are given by

(5)

(6)

where the values of ni and Hj corresponding to the physical cases are: nj = 114 for natural
convection; nj = 3 for bo~ing ; ni = I for Newton's law of cooling; n i = 4 for radiative
heat transfer. We notice h - heat transfer coefficient, A - heat transfer surface area, Ts -
temperature of the heat reservoirs, and '0 - working fluid temperature.
The resulting nonlinear system of equations is :

First Law of Thermodynamics w= 02-3 -104-11 (7)

Second Law of Thermodynamics

02-3
-----=0
104-11 (8)
Tw Tc

Energy balance at the source

mc v (Tw -T,)(1-1lR)+02_3+(hA) /oss,eq (TH-Td=~ (9)

where mc,. is the flow capacity rate of the working fluid, and (p is heat transfer rate
available at the source. The mathematical modeling also includes other equations,
especially for the heat transfer in the heat exchangers that permit to express the
intermediate model variables. Hence, it results a reduced number of constraints and
variables to be optimized.
406

In order to render the model more general and to facilitate the extension of the
analysis, all calculations have been made in non-dimensional form of the equations.
Thus, the temperatures were divided by the sink temperature, TL , the heat transfer rates
and the engine power output, by the product mcvh, respectively. The term "reduced"
were added to each non-dimensional value.
The optimization analysis of the Stirling engine results in a system of nonlinear
equations that are solved using the method of Lagrangian undetermined mUltipliers. The
non-dimensional system of equations consists generally from an objective function and
two constraints:

W,. =Q2-3r Q -I 4-lrl (10)

Q2-3r IQ4-lrl o (11)

(12)

where r is the ratio of specific heats, c/c,. The general variables of the model are the
reduced temperatures : ~, e., and 8" For source and sink of finite thermal capacity,
instead of 8H" the outlet temperature of the source, 8m" is considered. Also, depending
on the constraint, in this case one or two variables are added to the model. These
variables are reduced heat transfer conductance or area.
The model general parameters are (1) the reduced heat transfer rate available at the
source, (2) the NTU,,,.,., corresponding to the heat losses of the engine. Also depending on
the case, the NTUT corresponding to the total heat transfer surface conductance or area of
the engine heat exchangers and the ratio of heat transfer coefficients, rhi, are added.
The optimization method aims at indicating the optimum state and optimum distribution
of the heat transfer surface conductance or area for maximum engine power output.

2.3. STUDIED CASES INVENTORY

The optimization models were developed for several cases characterized by

(i) source and sink of constant temperature and finite thermal capacity;
(ii) linear and nonlinear form of the heat transfer law in the heat exchangers at the
source and sink ;
(iii) regenerator conductance or area included or not in the fixed corresponding
parameter of the engine.
By combining the above system configurations with the constraints of fixed (1) heat
transfer surface conductance or area at the source and sink and (2) total heat transfer
surface conductance or area of the engine, it results a significant ramification of the cases
tree. Only some simple cases permitted analytical solution to the optimization problem,
the others being solved numerically.
 407

3, Results

The preceding optimization method of the Stirling engine provides the optimum state and
distribution of the heat transfer surface conductance or area among the engine heat
exchangers. Calculations have been made for illustration using helium as the working
fluid and with the following variation range of fixed parameters: ¢r =0.1-:-10, NTU T
=0-:-5, NTU/o.<., =0-:-5, rhi =0.1-:-10. From the results we have obtained, the effect of the
type of heat reservoirs and that of the heat transfer law form on the engine performance
was selected for presentation in the paper.

3.1. EFFECT OF THE FINITE THERMAL CAPACITY OF THE HEAT

RESERVOIRS COMPARATIVELY TO THERMOSTATS

The analysis of the effect of the finite thermal capacity of the source and the sink on the
optimum state point, maximum power output of the engine and cycle efficiency is shown
in Table 1. The temperatures are not substantially affected. It can be seen that all the
temperatures increase for the finite thermal capacity, especially the source temperature. It
is quite normal because it represents the inlet source temperature, OHl"

TABLE I. State point variables, maximum engine power output and cycle thermal efficiency
for the case with regenerator included in the model and total heat transfer area fixed.

Counterflow Regenerator Parallel-flow Regenerator

Thermostat CF Finite Capacity CF Thermostat PF Finite Capacity PF
0, ' 1,170 1,240 1,150 1,220
Ow' 1,460 1,500 1,430 \,450
OH' 1,580 1,830 1,530 1,800

.
0,065 0,030 0,060 0,030
IWri.nax
11. 0,140 0,100 0,130 0,085

The maximum power output and the thermal efficiency of the cycle are mostly affected.
It is registered a reduction of 40-50% of the power and of 25-30% of the cycle efficiency
for the case of finite thermal capacity.
Other results of the calculations are shown plotted in Fig. 2. The heat transfer area is
mostly concentrated in the second heat exchanger of the sink, the regenerator and the
first heat exchanger of the source. The comparison of the distributions indicates that the
effect of the finite thermal capacity consists of the reduction of the heat transfer area of
both second heat exchangers of the source and sink, while the heat transfer area of the
other engine heat exchangers increases. Also we note that the regenerator heat transfer
area is important, mostly for the finite thermal capacity case.
 408

0.4 y-----------------,
0.35 +-----------.f1-----/
0.3
_... -

~
0.25 • Thermostat CF
0.2 +--____- - - - - - . _ • Finite Capacity CF;
[J Thermostat PF
0.15
IIFilli!~~apacity PF
0.1
0.05
o
arHI* arH2* arLl* arL2* arR*

Figure 2. Comparison of the optimum distribution of the heat transfer area among the engine heat
exchangers corresponding to the cases with constant or variable source and sink temperature.

3.2. EFFECT OF THE FORM OF THE HEAT TRANSFER LAW

The values corresponding to the optimum state point, maximum reduced power output
and thermal efficiency of the cycle for symmetrical and non symmetrical combinations of
the linear and radiative nonlinear form of the heat transfer law are presented in Table 2.
One can se that the minimum temperature difference at the thennal reservoirs (B H' -B w',
Be' -I) is registered for the radiative nonlinear form of the heat transfer law in

TABLE 2. Optimum state point. maximum power output and the corresponding thermal efficiency of
the cycle for the total heat transfer area fixed and regenerator included in the model.

Linear Law Combination of Radiative Nonlinear Law - Linear Law

CF PF CF PF
nIl = I nH = I nH =4 nH=4 nH = I nH =4 nH=4 nH = 1
nL= 1 nL= I nL= 1 nL=4 nL=4 nL = 1 nL=4 nL=4
e/ l.l70 1,150 1,150 1,055 1,070 1,130 1,030 1,050
Sw' 1.460 1,430 1,500 1,450 1,400 1,430 1,360 1.350
SH' 1,580 1,530 1,520 1,460 1,480 1,450 1,373 1,420

Iw,l ... o,
0,065 0,060 0,078 0,104 0,090 0.074 0,093 0.082

11, 0,140 0,130 0,165 0,200 0,175 0,142 0,165 0,150

any combination. It means that the irreversibilities due to the heat transfer at temperature
difference are minimum in these cases. The maximum power output and cycle efficiency
of the cycle are attained in the case of symmetrical combination of the radiative nonlinear
form of the heat transfer law and for the convective nonlinear form, that of the
combination nH = 3 ; nL = 1. It is obvious that the engine performance in the case of
symmetrical combination of the linear form of the heat transfer law usually adopted in
thermal machine analysis are low. Generally, the engine performance decreases in this
 409

order: (I) combinations of radiative nonlinear - linear form, (2) symmetrical combination
of linear form and (3) combinations of the linear - convective nonlinear form.
The heat transfer area distribution among the engine heat exchangers for three
combinations of the convective nonlinear form with the linear form of the heat transfer
law is shown in Fig. 3. This figure indicates that the heat transfer distribution changes
substantially for the last combination (nH = I ; nL = 3) comparing with those
corresponding to the others two, especially for the source heat exchangers. It appears that
the heat transfer area must be greather for the heat exchangers with linear heat transfer
law form. The same conclusion results for the combinations of the radiative nonlinear -
linear form of the heat transfer law.
The regenerator heat transfer area is important comparatively to the area distribution
in the other heat exchanger. Thus, for some combinations of the heat transfer law form, it
may have the greatest value among the engine heat exchangers.

nH = 3; nf. = 1 nH = 3; nL = 3

arL2*

nil =1; nf. =3

arHI *

FiKlIre 3. Heat transfer area distribution among the engine heat exchangers for three combinations
of the convective nonlinear form with the linear form of the heat transfer law.

The regenerator type does not influence substantially the engine performance.
Although the counterflow heat exchanger modeling the regenerator is preferred due to
the higher power output and cycle efficiency performed.

4. Conclusion

A synthesis of the results of an optImIzation method of the Stirling engine with

irreversibilities is presented that makes possible the prediction and comparison of the
performance of solar Stirling engines operating under the same conditions [20]. It has
shown how the state parameters and the distribution of the heat transfer area among the
engine heat exchangers are affected by the finite thermal capacity of the heat reservoirs
and the heat transfer law form in the heat exchangers.
410

5. References

I. Walker, G., Reader, G., Fauvel, O.R. and Bingham. E.R. (1994) The Stirling Alternative. Gordon
and Breach Science Publishers. Amsterdam.
2. Organ. I.A. (1992) Thermodynamic., and Gas Dynamics (It Stirling Cycle Machine, Cambridge
University Press. Cambridge.
3. Walker. G. (1983) Cryocoolers - Part I.' Fundamentals. Plenum Press. New York.
4. Tew. R.C. (1986) Progress of Stirling cycle analysis and loss mechanism characterisation. NASA
TM-8889/.
5. Feidt, M. (1996) Thermodynamique et Optimi.<ation Energetique des Sy.'tetnes et Procides,
TECH & DOC. Paris.
6. Radcenco. V. (1994) Generalized Thermodynamic,. Tehnical Publisher. Bucharest.
7. Bejan, A. (1988) Theory of heat transfer - irreversible power plants, Int. J. Heat Mas.' Tran~t'er.
31,6. 1211-1219.
8. Bejan, A. (1988) Advanced Engineering Thermodynamics, Wiley & Sons. New York.
9. De Vas, A. (1992) Endoreversible Thermodynamics otS()lar Energy Conver.,ion. Oxford Science
Publishers, Oxford.
10. Gordon. I.M. (1990) Nonequilibrium thermodynamics for solar energy applications. Advanced in
Thennodynamic,. 4. 95-120.
I I. Chen, F. and Yan, Z. (1989) The effect of heat transfer law on performance of a two-heat source
endoreversible cycle, Journal Chell!. Ph),s .• 90. 7. 3740-3743.
12. Blank. D. A.• Davis. G.W. and Wu. C. (1994) Power optimization of an endoreversible Stirling
cycle with regeneration. Energ),. 19. I. 125-133.
13. Feidt, M., Costea. M. and Popescu. G. (1995) Optimal design of Stirling engine using a
generalized heat transfer law model. Workshop 011 Second Law Analysis ot Energy Systems .'
towards the 21-st century. Rome. Italy. 431-441.
14. Costea, M .• Feidt. M. and Belabbas, B. (1995) Optimal distribution of the heat transfer
conductances or areas for Stirling machines. Proc ot WAM otASME, San Francisco. USA, 35, 35-
44.
15. Feidt. M .. Costea. M. and Belabbas. B. (1996) Influence of the regenerator on the optimal
distribution of the heat transfer surface conductance in a Stirling engine. Proc. of the Inter.
Sympos. ECOS'96, Stockholm. Sweden, 99-103.
16. Costea. M .• Feidt, M. and Belabbas. B. (1996) Comparison of the optimal heat transfer surface
area distribution in a Stirling engine with source and sink of finite thermal capacity, Proc ot the
Inter. Sympo.,. ESDA'96. Montpellier. France. 207-214.
17. Popescu. G., Radcenco, Y., Costen. M. and Feidt. M. (1996) Optimisation thermodynamique en
temps fini du moteur de Stirling endo- et cxo-irrcversible, Rev. Gell. Therm .• 35, 656-661.
18. Costea, M. (1997) Improvement of heat exchangers performance in view of the thermodynamic
optimization of the Stirling machine; Unsteady-state heat transfer in porous media, Ph.D. Thesis.
P. U. Bucharest & V.H.P. Nancy I.
19. Incropera. F.P. and DeWitt. D.P. (1996) Fundamentals (It Heat and Mass Trall.<ter. Wiley&Sons,
New York.
20 Stine. W. and Diver. R. (1994) A compendium of solar dish/Stirling technology, Sandia National
Laboratories, Albuquerque. NM. USA. Contraet No. 67-3678.
MINIMIZING LOSSES - TOOLS OF FINITE-TIME
THERMODYNAMICS

B. ANDRESEN
{Jrsted Laboratory, University o/Copenhagen
Universitetsparken 5, DK-2100 Copenhagen {J
Denmark

1. Introduction

Finite-time thermodynamics was 'invented' in 1975 by R. S. Berry, P. Salamon, and

myself as a consequence of the first world oil crisis. It simply dawned on us that all the
existing criteria of merit were based on reversible processes and therefore were totally
unrealistic for most real processes. That made an evaluation of the potential for
improvement, of minimizing the losses, for a given process quite difficult [1].
Finite-time thermodynamics is developed from a macroscopic point of view with
heat conductances, friction coefficients, overall reaction rates, etc. rather than based on a
microscopic knowledge of the processes involved. Consequently most of the ideas of
traditional thermodynamics have been assimilated, e.g. the notions of thermodynamic
potential (Sect. 2.1) and availability (Sect. 2.2). At the same time we have seen new
concepts emerge, e.g. the non-equivalence of well-honored criteria of merit (Sect. 3.2),
the importance of power as the objective, the generality of the endoreversible engine,
and in particular thermodynamic length (Sects. 2.3 and 3.3). Several of these abstract
concepts have been successfully applied to practical optimizations [2 - 9]. We will pre-
sent some of these tools below, roughly in order of increasing complexity.
The model of Curzon and Ahlborn [10] has evolved into almost a classic paradigm
of systems operating in finite time. This is a Carnot engine with the simple constraint
that it be linked to its surroundings through finite heat conductances. Figure 1 illustrates
the slightly more general endoreversible system with the triangle signifying any revers-
ible engine. (The term endoreversible means 'internally reversible' , i.e. all irreversibili-
ties reside in the coupling of flows to the surroundings. In this case that means resist-
ance to heat transfer and possibly friction.) It turns out that the results derived by
Curzon and Ahlborn (and by Novikov [11] and Chambadal [12] for simpler systems)
explicitly for an interior Carnot engine are equally valid for a general endoreversible
system. The maximum efficiency of their engine is of course llc = 1 - T LIT H'
obtained at zero rate so that losses across the resistors vanish, but these authors showed
that, when the system operates to produce maximum power, the efficiency of the engine
is only

(1)

411
A. Bejan and E. Mamut (eds.), Thermodynamic OptimiZlJlion o/Complex Energy Systems, 411-420.
@ 1999 Kluwer Academic Publishers.
412

Besides the simplicity of the expression it is remarkable that it does not contain the
value of the heat conductances.

TH Figure 1. An endoreversible engine has all

its losses associated with its coupling to the
environment. there are no internal irrevers-
ibilities. This is illustrated here as resistances
in the flows of heat to and from the working
device indicated by a triangle. These un-
avoidable resistances cause the engine
proper to work across a smaller temperature
interval. [Th;T[l than that between the reser-
TI voirs. [TH;TLl. one which depends on the
rate of operation.

TL ////////////~

2. Performance Bound without Path

The smallest amount of information one can ask for concerning the performance of a
system is a single number, e.g. the work or heat exchanged during the process, its effi-
ciency, or any other figure of merit. In most cases this can be calculated without knowl-
edge of the detailed path followed and is then computationally much simpler to obtain.

2.1. GENERALIZED POTENTIALS

In traditional thermodynamics potentials are used to describe the ability of a system to

perform some kind of work under given constraints. These constraints are usually the
constancy of some state variables like pressure, volume, temperature, entropy, chemical
potential, particle number, etc. Under such conditions the decrease in thermodynamic
potential P from state i to state f is equal to the amount of work that is produced when a
reversible process carries out the transition, and hence is the upper bound to the amount
of work produced by any other process,

(2)
In this section we will show that the constraints need not simply be the constancy of
some state variable, and that the potentials may be generalized to contain constraints
involving time [l3]. The procedure will be a straight forward extension of the Legendre
transformations [14] used in traditional thermodynamics [15, 16], and we will start with
such an example.
In a reversible process heat and work can be expressed as inexact differentials,
dQ =TdS, dW =PdV, i.e. they cannot by themselves be integrated, further constraints
defining the integration path are required. Such a constraint could be that the process is
 413

isobaric, dP = O. One can then add a suitable integrating zero-term, xdP to make dW an
exact differential. The obvious choice is x = V,

dW = PdV = PdV + VdP =d(PV), (3)

such that the isobaric work potential becomes P = PV.
Now, the constraints need not be the constancy of one of the state variables. Consi-
der a baBoon with constant surface tension a. In equilibrium with an external pressure
P ex such a sphere of radius r has an internal pressure P = Pex + 2a/r, which can be
rearranged into

(4)
Since the right hand side of this equation is a constant, this means that (P - Pex)V 1/3 is
an integral of motion for the fluid inside the baBoon. We can then add a suitable amount
of d[(P - Pex)V 1/3] (equal to zero) to dW to make it exact,

dW = PdV = PdV + ~ V 2/3 d[(P - Pex )V1/3] =d[i V(3V - Pex)]. (5)

Thus the work done by the coupled system, surface plus fluid, is given by the decrease
in the potential P = ~ V(3V - Pex)' regardless of path foBowed.
This procedure of adding a suitable amount of a 'zero-quantity' can be generalized to
any differential constraint, even including time explicitly [13].

2.2. FINITE-TIME AVAILABILITY

One of the more powerful results in finite-time thermodynamics is the definition of a

finite-time availability [17]. The traditional availability A of a system in contact with
given surroundings is a state function with the quality that the decrease in its value in
going from state i to state f is the maximum (and hence reversible) work that can be
extracted during that process. The finite-time availability A retains this property and
simply adds that the process is restricted to operate (go to completion) during time
't = t f - t i . Then

(6)

where the last equality uses the Tolman-Fine form of the Second Law of thermodyna-
mics [18].
The maximum search in Eq. (6) can either be constrained to exactly reach a given
final state at time 4 (the initial state is always considered known), in which case !l.A is
fixed, and the optimization becomes one of minimizing the entropy production, or also
the final state may be included in the optimization, in which case A must be evaluated
by optimal control. If the final state is specified, a solution may not exist if 't is too
short, since only a certain set of states can be reached from a given initial state in time
414

'to In addition, the finite-time availability does not necessarily have !:iA as its limit for
very long times, because the system may contain internal relaxation processes which
remain irreversible even for very slow operation. If there is a direct heat leak from the
system reservoir to the surroundings, then a long process time may even reduce A to
zero. On the other hand, losses are not always detrimental to the performance of a
system if they open up new pathways - actually some processes depend on irreversibil-
ities for their very existence (e.g. [19]).

2.3. THERMODYNAMIC LENGTH

For static purposes Weinhold [20, 21] defined a metric on the abstract space of equilib-
rium states of a system represented by all its extensive variables Xi as

(7)

where U is the internal energy. Based on this metric Salamon and Berry [22] found a
connection between the thermodynamic length along a process path and the (reversible)
availability lost in the process. Specifically, if the system moves via states of local
thermodynamic equilibrium from an initial equilibrium state i to a final equilibrium
state f in time 't, then the dissipated availability -!:iA is bounded from below by the
square of the distance (i.e. length of the shortest path) from i to f times £/'t, where £ is a
mean relaxation time of the system. If the process is endoreversible, the bound can be
strengthened to

(8)
where L is the length of the traversed path from i to f. Equality is achieved at constant
thermodynamic speed v = dLidt corresponding to a temperature evolution given by [23]

cIT = _ vT/£(C)112 (9)

dt '
where C is the heat capacity of the system. For comparison, the bound from traditional
thermodynamics is only -!:iA ;::: O.
An analogous expression exists for the total entropy production during the process:

(10)

The length L is then calculated relative to the entropy metric

(11)

In statistical mechanics, where entropy takes the form S( {Pi} ) = - Li Pi In Pi' the metric
Ms is particularly simple, being the diagonal matrix [24] Ms = -{ UPi}·
 415

The same procedure of calculating metric bounds for dynamic systems has been
applied to coding of messages [25] and to economics [26].
More recently [27] we have relaxed a number of the assumptions in the original
work, primarily those restricting the system to be close to equilibrium at all times and
the average form of the relaxation time c. The more general bound replacing Eq. (10)
then becomes

with :=: = Sf - Si being the total duration of the process in natural dimensionless time
units, dS = dtlc(T), and where we have defined S(T) = 1 + (T/2C)(ClC/ClT).
The equality (lower bound) in Eq. (12) is achieved when the integrand is a constant,
i.e. when dSu/dS = constant. Consequently, constant rate of entropy production, when
expressed in terms of natural time, is the path or operating strategy which produces the
least overall entropy. One can express the optimal path in a form similar to Eq. (9):

cIT
dt
[1 +
S(T) E (dT/dt)
T + ...
]
In T
= constant X c(C)1I2 . (13)

The constant thermodynamic speed algorithm, Eq. (9), is thus the leading term of the
general solution in an expansion about equilibrium behavior.

3. Optimal Path

3.1. OPTIMAL PATH CALCULATIONS

A knowledge of the maximum work that can be extracted during a given process, e.g.
calculated by one of the procedures described in the previous section, may not by itself
be sufficient. One may also want to know how this maximum work can be achieved,
i.e. the time path of the thermodynamic variables of the system. The primary tool for
obtaining this path is optimal control theory.
This is not the place to repeat the mechanics of optimal control calculations (see
e.g. [28 - 30]). Let it suffice to point out that in order to set up the optimal control
problem one must specify
the controls, i.e. the variables that can be manipulated by the operator (they may be
a volume, rate, voltage, heat conductance, etc.);
limits on the controls and on the state variables, if any (in order to avoid unphysical
situations such as negative temperatures and infinite speeds);
the equations that govern the time evolution of the system (they will usually be
differential equations describing heat transfer rates, chemical reaction rates, friction,
and other loss mechanisms);
 416

the constraints that are imposed on the system (e.g. conserved quantities, the quanti-
ties held constant, or any requirements on reversibility. The constraints may either
be differential, instantaneous i.e. algebraic, or integral i.e. not obeyed at each point
but over the entire interval);
• the desired quantity to be maximized, called the objective function (usually expres-
sed as an integral); and finally
• whether the duration of the process is fixed or part of the optimization.
Typical manipulation usually leads to a set of coupled, non-linear differential equa-
tions for which a qualitative analysis and a numerical solution are the only hope. Thus
answering the more demanding question about the optimal time path rather than the
standard question about maximum performance requires a considerably larger computa-
tional effort. On the other hand, once the time path is calculated, all other thermodyna-
mic quantities may be calculated from it, much like the wave function is the basis of all
information in quantum mechanics.

3.2. CRITERIA OF PERFORMANCE

Efficiency, the earliest criterion of performance for engines, measured how much water
could be pumped out of a mine by burning a ton of coal. Other familiar criteria include
effectiveness (efficiency relative to the Carnot efficiency), change of thermodynamic
potential, and loss of availability, all of which are measures of work. Potentials for heat
can also be defined but are less common. In addition total entropy production is indica-
tive of inefficient operation [31]. The Curzon-Ahlborn analysis and many of our own
analyses use a quite different criterion, that of power. This quantity is of course zero for
any reversible system, and maximizing power forces us to deal with systems operating
at finite rates. Other criteria of performance are the rate of entropy production and the
rate of loss of availability. Under some circumstances, optimizing one of these quanti-
ties is equivalent to optimizing another. For example, minimizing the entropy produc-
tion is equivalent to minimizing the loss of availability.
Salamon and Nitzan [32] have optimized the Curzon-Ahlborn engine for a number
of these objective functions. Assuming the working fluid to be in contact with the hot
reservoir for the period 1:1 and the cold reservoir for the period 1:2' the optimal time dis-
tributions are shown in Fig. 2. The diagonal 1: = 1: 1 + 1:2 indicates fixed total cycle
time, and only processes above the curve labeled 'zero power' produce positive power. It
is quite obvious that different criteria of merit dictate different operating conditions for
the process. Even when not knowing the precise objective function but only that it
belongs to a specified class, one can sometimes say a good deal about the possible
optimal behavior of the system. If one considers the Curzon-Ahlborn engine to be a
model of a power plant which buys heat q (coal) at the unit price <X and sells work w
(electricity) at the unit price /3, its net revenue is n = /3w - <xq. All solutions to the
problem of maximizing this revenue are bounded on one side by the solutions to the
maximum power problem (when <X is significant compared to /3) and on the other side
by the solutions corresponding to minimum loss of availability (when coal and electric-
ity are priced according to their availability contents). While this is a very simple
 417

example, this approach has far reaching possibilities for describing biological, ecologi-
cal, and economic systems.

!z
{ min, entropy production
min loss of availability
max, revenue
{ max, efficiency
max, effectiveness
max, power

_,""-_-=::zero power

"[1
Figure 2, If an endoreversible engine (Fig, I) spends time tJ in contact with the hot reservoir and 't2 in
contact with the cold reservoir, the optimal proportioning hetween 'tJ and 't2 depends on what one chooses
to optimize, as indicated, The locus of maximum revenue for a power producing system falls in the shaded
area for any choice of prices, as described in the text. Only contact times above the hyperbola marked
'zero power' actually correspond to positive power production,

3.3. DESIGN OF DISTILLATION COLUMN BY EQUAL THERMODYNAMIC

DISTANCE

In Sect. 2.3 we saw that the concept of thermodynamic length not only provided a lower
bound on dissipation (Eqs, (8), (10), and (12», it also predicted which path would
achieve that bound, namely operation at constant thermodynamic speed. We will use
that line of thought to improve the performance of a conventional distillation column.
The internal distribution of temperature and mole fractions of light and heavy com-
ponents in a conventional distillation column where heat is added and withdrawn only in
the reboiler and the condenser, respectively, is fixed exclusively by the laws of energy
and mass conservation. In most columns that leads to an S-shaped curve of temperature
versus plate number with most of the variation occurring near the end points of the
column, connected with a flat stretch around the feed point (see Fig. 3). A qualitatively
similar picture emerges for the mole fractions. This temperature profile implies that the
major part of the entropy production in the distillation process occurs near the ends of
the column and is thus not uniformly distributed. Even more importantly, if the number
of plates is increased, essentially only the middle flat section is extended while the seg-
ments of rapid variation are unchanged. This means that dissipation does not approach
zero as the number of plates goes to infinity.
418

-... 350
4>

=
...
ell
4>

-
C.
e
4>

340

o 12 14 16 18
plate number
Figure 3. The temperature profile as a function of plate number, counted from the condenser, for conven-
tional (solid) and equal-thermodynamic-distance (dashed) separation of methanol and isopropanol.

The principle of constant thermodynamic speed for continuous systems, or equiva-

lently equal thermodynamic distance for discrete systems, aims at minimizing total
dissipation in the column by distributing it evenly among the plates whatever their
number [33]. In [34] a general quasistatic step process was optimized, i.e. a process
composed of N discrete steps where the system equilibrates fully after each step. The
standard description of a distillation column is exactly such a process where it is as-
sumed that gas and liquid come to equilibrium at a particular temperature on each plate.
Entropy is produced when the up- and down-moving flows encounter liquid on the next
plate at slightly different temperature and composition.
The result of the optimization [34], for processes not too far from equilibrium, was
that the minimum of the entropy production

(14)

is achieved when the thermodynamic distance between each pair of plates is kept con-
stant, i.e. they are equidistant in that measure. L is the total thermodynamic distance
from one end of the column to the other. For any other distribution of the total length
along the column dissipation is larger.
Using the metric Eq. (11), the thermodynamic distance DSk from plate k-l to plate
k is equal to

Dl = (L1Xk Msk L1Xk) 112 (15)

where L1Xk = X k - X k- 1 is the difference in extensities from plate k-l to plate k. The
total thermodynamic length of the column is thus Ls = L~l DSk.
Two consequences of this general result are immediate: dissipation (here entropy
production) must be equally distributed along the column; and the total dissipation
approaches zero as N, the number of plates, goes to infinity (Eq. (14)), i.e. the separa-
tion becomes reversible.
In binary distillation 8 extensive quantities are involved on each plate, besides the
objective function entropy, namely enthalpy, volume, mole number of light
 419

component, and mole number of heavy component, each one for both gas and liquid,
leading to an 8x8 metric matrix. Fortunately a number of relations allows to reduce the
dimensionality of the problem dramatically. First of all, the two components are usually
considered non-interacting, at once making Ms block-diagonal. Next, constant pressure
in the column, energy and mass conservation, and the gas-liquid equilibrium equation
reduce the problem to just one free variable which may conveniently be taken to be
temperature. After a lot of algebra the plate-to-plate distance then becomes simply

DSk = [(Ch/2rrk] ~T (16)

where Cr is an effective heat capacity involving all the constraints mentioned above in a
complicated expression.
The computational procedure is to integrate Eq. (16) from the distillate temperature
T D to the reboiler temperature T B (both of course given by the required product purities)
to obtain the total thermodynamic column length Ls. The distance from one plate to the
next must then be fixed at Ds =LslN for optimal performance by adjusting the plate
temperatures appropriately according to Eq. (24).
Obviously such freedom of adjustment does not exist in a conventional adiabatic
column. Rather, it is necessary to allow individual heat exchange with each plate to
maintain it at the desired temperature. This heat addition/removal is of course part of the
energy balance used above. The result of the whole calculation is either a graph like
Fig. 3 specifying the temperature of each plate in the column or a graph of the amount
of heat added/removed at each plate. It should be emphasized that the total amount of
heat used to perform a certain equal-thermodynamic-distance separation is only margi-
nally different from that required by a conventional column, but a large part of it is used
over a much smaller temperature difference than TB to T D leading to a correspondingly
smaller entropy production.

4. Acknowledgments

Much of the work reported in this review has been done in collaboration with colleagues
near and far. In particular I want to express my gratitude to profs. R. Stephen Berry,
Peter Salamon, Jim Nulton, and Jeff Gordon.

References

I. Andresen. B. (1983) Finite·Time ThermodYlUlmics, Physics Laboratory II. University of Copenhagen.

2. Hoffmann. K. H., Burzler, J., and Schubert. S. (1997) Endoreversible systems. 1. Non-Equil. Therm.
22.311.
3. Hoffmann. K. H., Watowitch, S., and Berry, R. S. (1985) Optimal paths for thermodynamic systems:
The ideal Diesel cycle, 1. Appl. Phys 58, 2125 - 2134.
4. Mozurkewich, M. and Berry. R. s. (1981) Finite-time thermodynamics: Engine performance im-
proved by optimized piston motion, Proc. NaIl. Acad. Sci. USA 78,1986.
5. Mozurkewich, M. and Berry, R. S. (1982) Optimal paths for thermodynamic systems: The ideal Otto
cycle. 1. Appl. Phys 53. 34 - 42.
6. Gordon, J. M. and Zarmi, Y. (1989) Wind energy as a solar-driven heat engine: A thermodynamic
approach, Am. 1. Phys. 57,995 - 998.
420

7. Andresen, B. and Gordon, J. M. (1992) Optimal heating and cooling strategies for heat exchanger
design, J. Appl. Phys. 71, 76 -79.
8. Andresen, B. and Gordon, J. M. (1992) Optimal paths for minimizing entropy generation in a
common class of finite-time heating and cooling processes, Int. 1. Heat and Fluid Flow 13, 294 - 298.
9. Gordon, J. M. and Ng, K. C. (1994) Thermodynamic modeling of reciprocating chillers, 1. Appl.
Phys. 57,2769 - 2774.
10. Curzon, F. L. and Ahlborn, B. (1975) Efficiency of a Carnot engine at maximum power output, Am.
1. Phys. 43, 22 - 24.
II. Novikov, I. I. (1957) The efficiency of atomic power stations (in Russian), At. Enertg. 3,409. English
translation in (1958) 1. Nucl. Energy 117, 125.
12. Chambadal, P. (1957) Les centrales Iluc/eaires, Armand Colin, Paris., p. 41.
13. Salamon, P., Andresen, B., and Berry, R. S. (1977) Thermodynamics in finite time. II. Potentials for
finite-time processes, Phys. Rev. A 15, 2094 - 2102.
14. Hermann, R. (1973) Geometry, Physics, and Systems, Marcel Dekker, New York.
15. Callen, H. B. (1985) Thermodynamics, Wiley, New York.
16. Tisza, L. (1966) Generalized Thermodynamics, MIT Press, Cambridge, Massachusetts.
17. Andresen, B., Rubin, M. H., and Berry, R. S. (1983) Availability for finite-time processes: General
theory and a model, 1. Chern. Phys. 87, 2704 - 2713.
18. Tolman, R. C. and Fine, P. C. (1948) On the irreversible production of entropy, Rev. Mod. Phys.20,
51-77.
19. Wheatley, J., Hofler, T., Swift, G. W., and Migliori, A. (1983) Experiments with an intrinsically irre-
versible acoustic heat engine, Phys. Rev. Lett. 50, 499.
20. Weinhold, F. (1975) Metric geometry of equilibrium thermodynamics, 1. Chern. Phys. 63, 2479-
2482.
21. Weinhold, F. (1978) Geometrical aspects of equilibrium thermodynamics, in D. Henderson and H.
Eyring (eds.), Theoretical Chemistry, Advances and Perspectives, Vol. 3, Academic Press, New
York, p. 15 - 54.
22. Salamon, P. and Berry, R. S. (1983) Thermodynamic length and dissipated availability, Phys. Rev.
Lett. 51,1127 - 1130.
23. Nulton, J. and Salamon, P. (1988) Statistical mechanics and combinatorial optimization, Phys. Rev. A
37,1351 - 1356.
24. Feldmann, T., Andresen, B., Qi, A., and Salamon, P. (1985) Thermodynamic lengths and intrinsic
time scales in molecular relaxation, 1. Chern. Phys. 83, 5849 - 5853.
25. Flick, J. D., Salamon, P., and Andresen, B. (1987) Metric bounds on losses in adaptive coding, Info.
Sci. 42,239 - 253.
26. Salamon, P., Komlos, J., Andresen, B., and Nulton J. (1987) A geometric view of welfare gains with
non-instantaneous adjustment, Math. Soc. Sci. 13, 153 - 163.
27. Andresen, B. and Gordon, J. M. (1994) On constant thermodynamic speed for minimizing entropy
production in thermodynamic processes and simulated annealing, Phys. Rev. E. 50, 4346 - 4351.
28. Boltyanskii, V. G. (1971) Mathematical Methods of Optimal Control, Holt, Reinhart, and Winston,
New York.
29. Tolle, H. (1975) Optimization Methods, Springer, New York.
30. Leondes, C. T. (ed.) (1964 and onward), Advances in Control Systems, Academic Press, New York
[especially articles in the first volumes].
31. Salamon, P., Nitzan, A., Andresen, 8., and Berry, R. S. (1980) Minimum entropy production and the
optimization of heat engines, Phys. Rev. A 21,2115 - 2129.
32. Salamon, P. and Nitzan, A. (1981), Finite-time optimization of a Newton's law Carnot cycle, 1.
Chem. Phys. 74, 3546 - 3560.
33. Salamon, P., Andresen, B., Nulton, J., and Siragusa, G. (1998), The horse-carrot theorem and geo-
metric optimization of distillation (in preparation).
34. Nulton, J., Salamon, P., Andresen, B., and Anmin, Q. (1985) Quasistatic processes as step equilibra-
tions, 1. Chern. Phys. 83, 334 - 338.
PHYSICS VERSUS ENGINEERING OF FINITE-TIME THERMODYNAMIC
MODELS AND OPTIMIZATIONS

P. SALAMON
Department ofMathematical and Computer Sciences
.s'an Diego .s'tate University
San Diego, CA 92182-7720, USA

1. Introduction

Finite-time thermodynamics [I [ is an extension of traditional thermodynamics that seeks

to characterize in-principle lilnits to the performance of thermodynamic processes given
the constraint that such processes take place in a finite time. This constraint forces any
process that involves transport through a finite conductance to produce entropy and thus
leads to strengthened versions of the second law and provides positive lower bounds on
the associated entropy production.
The present paper differentiates between the physics and the engineering perspectives
regarding the value of such studies and gives some of the highlights of what can be
gained from physical (as opposed to engineering) studies of such processes. The
difference can be characterized as follows: physical studies seek in-principle lilnits to
what is possible while engineering studies seek to characterize the best possible process
design given current technology and current prices. Most of the papers at tins conference
describe the engineering perspective; the present paper describes the physics perspective.

2. Bounds and the Role of Ol)timization

One important reason for studying tIlennodynannc processes in finite time is to pursue tile
quest for understanding tile limits of what can be achieved in such processes quite aside
from any direct implications for engineering. The simplest analyses of this kind reveal
how different sources of irreversibility individually influence the limits to power
generation and entropy production. In such analyses there is no pretense of counting all
sources of irreversibility -- one merely counts tIlOse sources one is focusing on and
tIlereby obtains bounds on what can happen in real processes.
The underlying matIlematical fact that allows this is a basic lemma of optimization
theory: solving an optimization problem with fewer constraints (additional controls)
bounds the value of the objective for the original problem. Analyzing the optimal
performance of a system when we improve the control that we are able to exercise. can
only improve the optimal value of our objective and thereby gives a bound on what is
physically possible.
One important example of tIns lemma in finite-time thermodynamics is the theorem of
Orlov and Berry [2) which is striking in its simplicity and power. They begin by writing
down the full dynamical equations for tile working fluid, the heat baths, and the power
source coupled to a heat engine complete with partial differential equations for the
421
A. 8ejan and E. Mamut (eds.), Thermodynamic Optimization o/Complex Energy Systems, 421-424.
© 1999 Kluwer Academic Publishers.
422

distributed subsystems. The heat transfer to the working fluid is proportional to the
temperature difference across a surface. They complete the careful formulation of their
control problem by specifying an objective function. They then switch to a problem in
which the temperature at each position inside the working fluid and the heat exchangers is
controllable. This discards the partial differential equation constraints describing the
dynamics of the temperature in the regions inside the engine and heat exchangers, thereby
changing these temperatures from state variables to control variables. The optimum in the
simplified problem bounds the optimum in the real problem since the simplified problem
results by effecting better controls. Ignoring a great number of the dynamical equations
converts state variables to control variables and enlarges the set of allowed processes. Of
course tillS must be done carefully respecting the laws of thermodynamics -- these are
constraints tIlat we must not ignore.
Physically it is easy to ensure that the laws of thermodynamics are not violated. We
simply take the additional (fictitious) controls on our process to be reversible, i.e. effected
by reversible processes using additional work and heat sources as needed. Thus we
assume reversible control over all but a few of the degrees of freedom in our system. We
then ask what is tile least entropy production. or maximum power output. or the optimal
value of various other possible objectives. when we control the rest of the universe in eUl
ideal way -- namely reversibly. This focuses the analysis on the limitations to energy
conversion inherent in that part of the system or device which we did not assume to be
reversibly controllable. The approach is not limited to heat engines and the results bound
the possible values of the objective in a more memllngful way than the bounds provided
by classical thermodynamics. While the additional controls may be unrealistic from an
engineering point of view. they do not violate physical laws and can help us extract
information regarding the limitations to real processes imposed by specific modes of
irreversibility which are the focus of a specific study. Since typical applications of the
Orlov-Berry theorem require the controls to maintain uniform intensities inside a system.
they can even be realized by real processes in which the transport coefficients inside a
system are large compared to the transport coefficients across the boundary. TillS is
usually referred to as endoreversibility which is often useful in simplifying the mlalysis of
limitations to tile behavior of thennodynamic systems I. Note that tIllS is not always
needed for our analysis but only when the dissipation inside a system is not part of the
irreversible modes which are the focus of our study. Note further that the term "endo" is
slightly misleading since this argument can be equally well applied to subsystems such as
tIlermal reservoirs which form part of the "environment" and are thus more aptly labeled
"exo".
A second tIleorem due to A. Tsirlin [4] serves to simplify finite-time thermodynanlic
studies even further. TillS theorem states that provided the constraints eUld objective of a
problem depend only on net (or average) process quantities, i.e. on integrals over time,
the optimal way to control the system is to take our controls to be piecewise constant in
time. These two theorems make simple models interesting and able to give useful.
rigorous bounds regarding what can be achieved during the control of real processes. The
simplified problems often lead to solutions which tell us to run the process infinitely fast
or infinitesimally slow. The interesting domain is in between: well posed problems which

1 Prior to the Orlov-Berry theorem the argument required an assumption of the separability of the time scales for
the transport processes within a system and across system boundaries 13).
 423

ignore enough irreversibility to be tractable and yet count enough irreversibilities to keep
the solution finite. The value of the objective obtained in the simpler optimization bounds
the value of the objective in real processes in a more realistic way than a reversible
bound, although it is still often very far from what can be achieved in reality.
The results must be used with great care however. Other features of the optimal solution
obtained cannot be directly compared with similar features of real processes. A simple
and familiar example is provided by the well known equation of Curzon-Ahlborn-
Novikov (CAN) [5.61 for the efficiency at maximum power

(I)

The striking simplicity of this equation. as well as its similarity to the well known
efficiency of Camot. has brought it rapid acceptance. Nonetlleless. as pointed out in [71.
this efficiency is surpassed in real power plants! The explanation is that this equation does
not represent tlle maximum efficiency but rather the efficiency at maximum power. The
associated maximum power

(2)
must be more than the power delivered in any process which corresponds to the CAN
calculation augmented by additional dynamical equations. This much is mathematical
fact. Note that there is no implication regarding a comparison between the efficiency
obtained at this power and efficiencies in real processes -- even when the real processes
are also designed to maximize power. albeit with additional constraints!

3. Mathematical Structure
As the above discussion illustrates. understanding tlle physics is different from
understanding the engineering. Physicists are not directly interested in how to design or
operate a specific device or process but rather in understanding what limitations are set
for the process by the thennodynamics laws. In mathematical physics, such understanding
usually comes by studying the mathematical structure of tlle set of possibilities. Tlus is
not a problem tllat engineers have been concerned witll, even when very similar
optimizations were performed for very similar systems. It is a problem worth studying.
As an example of the type of insight tlmt can be gained from an examination of the
mathematical structure, we cite a result that has become known as the horse-carrot
theorem [8. 9). This tlleorem quantifies the nunimmn entropy production associated with
bringing a system along a specific sequence of states, i.e. with making it traverse a certain
path in its state space. The theorem concerns a Riemannian metric structure on this state
space 2 and says that the minimum entropy produced in the process is the square of the
length of the process divided by twice the number of relaxations. A recent application of
the theorem says something of interest to engineering by specifying the optimal
temperature profile for a distillation colmrm [10 I.

2 The Riemannian metric is given by the second derivative matrix of the entropy ofthe sy~tem as a nmction of
the extensive variables.
424

4. Difficulties of Communication

There exist nontrivial barriers to communication between physicists and engineers

working in this area. To make this point I cite an anecdote regarding the use of the phrase
"industrial example". In theorist-speak, tIlis phrase means that some industry might care
about it some day. In a referee report. the esteemed orgaIlizer of our conference, AdriaII
Bejan, dismissed geothennal power plants as "hardly an industrial example" since "its
share of the global energy industry is meaningless. ,,3
As a second example I consider the word "applied". In a matlIematics department. tlIis
word is used to mean that the arguments and results relate to sometIling outside tlIe sphere
of matlIematics, even when this "use" is extremely esoteric, e.g. characterizing certain
cross sections of a fiber bundle over a seventeen dimensional space-time. Note that tlIis
use is very far from what "applied" typically means in a physics department which is still
dramatically different from what "applied" means in an engineering department.
Other examples of words that meaII very different tllings to engineers. physicists, and
mathematicians include "realistic." "general." "optimal." and even "understanding."

5. Conclusions
In our modem multicultural and multidisciplinary world it is important to adopt a wider
perspective. Wllile tIle type of physical studies outlined above may be of interest to
engineering design only in its earliest stages. there is a long tradition of physics ultimatelv
leading to principles important for engineering. It is important to accept the fact that our
field is interdisciplinary and to try to move beyond the perspectives offered by our own.
narrowly defined disciplines.

References
1. Sieniutycz. S. and Salamon. P. (1990) FlI1ile-Tllne ThermodynamiCs and ThermoeconOlnics. Taylor and
Francis. New York.
2. Orlov. V. and Berry. R. S. (1990) Power output trom an irreversible heat engine with a nonuniform working
fluid. [,hYSical Review A 42. 7230-7235.
3. Salamon. P.. Nitzan. A.. Andresen. B. and Berry. R. S. (1980) Minimum entropy generation and the
optimization of heat engines. [,hYSical Review A 21. 2115-2129.
4. Tsirlin. A. M. (1986) ()plilnal ('onlrol o(Technologlcal ['rocesses. Energoatomizdat. Moscow.
5. Novikov. 1. I. (1958) The dliciency of nuclear powa stations. Journal oj Nuclear Energy !I (1 I.S.S.R.) 7.
125-128.
6. Curzon. F. L. and Ahlborn. B. (1975) Ellicicncv ofa Carnot engine at maximum power output. American
Journal o(Physlcs 43. 22-24.
7. Gyfiopoulos. E. (1997) Fundamentals of analysis of processes. Energy ('ol/verslOn and Managemenl 311.
1525-1533.
8. Salamon. P. and Berry, R. S. (1983) Thermodynamic length and dissipated availability. PhYSical ReView
Leiters St. 1127 - 1130.
9. Nulton. J .. Salamon, P.. Andresen. B. and Anmin. Q. (1985) Quasistatic processes as step equilibrations.
Journal o(Chemical Physics 83, 334-338.
10. Salamon. P. and Nulton. J. D. (1998). The geomdry of separation processes: a horse-carrot theorem tor
steady flow systems. Europhyslcs Leiters 42. 571-576.

3 We remark that in Iceland. the overwhelming majority of the energy industry is geothermal. Thus the
perspectives offered by different cultural and ethnic affiliations also create barriers to communications.
OPTIMIZATION AND SIMULATION OF TIME DEPENDENT HEAT DRIVEN
REFRIGERATORS WITH CONTINUOUS TEMPERATURE CONTROL

J.V.c. V ARGAS
Department ofMechanical Engineering
Universidade Federal do Parana. Curitiba. PRo 81531-990. Brazil

1. Introduction

Economic and environmental considerations have brought new interest to heat driven
refrigerators. A large research effort has been brought to bear on refrigeration systems of
this kind in recent years [I, 2]. These systems can make use of waste heat and renewable
energy sources such as solar energy [2, 3]. In countries in which electricity and fossil
fuels are available only intermittently or are very expensive, solar refrigeration may be
an alternative for the preservation of foodstuffs and medical supplies [4], since a
"reservoir of cold" can also be generated through the formation of subcooled ice for use
when the sun is obscured. However, when uninterrupted precisely controlled cooling is
required, heat driven systems powered by gas, electricity or oil are needed either as an
alternative to solar thermal energy and waste heat or as a supplementary thermal energy
source. The improvement and optimization of the design and control of heat driven
refrigerators is a crucial issue regardless of the energy sources used. Heat driven
refrigerators, with an effective control system, may well be able to combine economy
with a low environmental impact.
To be effective, a theoretical analysis of a thermal system must be capable of
capturing "realistic" features of the heat transfer processes, and the unit. Several studies
have dealt with these issues by using the method of entropy generation minimization [I,
2, 5-7] (exergetic optimization), which pursues realistic models and takes into account
the inherent irreversibility of heat, mass and fluid-flow, this is also known as the method
of finite-time thermodynamics.
In this paper, a mathematical model is introduced to simulate the transient
behavior of a heat driven refrigerator that uses a fuel burning heater as the high
temperature source. External irreversibilities are accounted for. A generalized version of
a power law control system, first introduced by Vargas and Parise [8], is then applied to
control the fuel flow rate to the system in closed loop operation. The power law
approach replaces the well known proportional-integral-derivative [9] combined
actions. An experimental validation of the steady state behavior of the model is
425
A. Bejan and E. Mamut (eds.). Thermodynamic Optimization o/Complex Energy Systems. 425-435.
@ 1999 Kluwer Academic Publishers.
426

conducted through direct comparison of the theoretical results with experimental data
obtained from an existing absorption refrigeration unit [10].
The objectives of this study, therefore, were (1) to examine how the thermal
conductance allocation should be optimized for maximum refrigeration rate and
minimum "pull-down" time, (2) to investigate the fuel consumption of the system under
"on-off' and closed loop operation, and (3) to present a model for the transient
simulation of heat driven refrigerators for design and optimization purposes.

2. Mathematical Model

The block diagram of Fig. I shows the main features of the system with continuous
temperature control (closed loop). From left to right: center, summer (circle), power law
control action, net gain, fuel pump gain, absorption refrigerator and ambient temperature
blocks; top, integration block; bottom, amplifier gain and thermal sensitivity blocks.
The sole difference between the two systems is in the nature of the control action.
In the "on-off' system the fuel flow rate can only have one of two values: At or zero. In
the system with "feed-back", the value of At varies continuously in a manner
determined by the control system.
The general features of the absorption refrigerator - refrigerated space model are
shown in Fig. 2, which are represented by elements in the block diagram shown in Fig.
1.

Tout

Figure 1. Closed loop block diagram.

The cycle requires negligible work input. The cycle is driven by the heat transfer
rate QH received from the hot source temperature at TH, which is determined by the
operating temperature of the generator. The refrigeration load QL is removed from the
refrigerated space, i.e., the evaporator drum shown in Figure 3, at TL , and the heat
transfer rate Qo is rejected to the ambient, at To. The label reversible compartment
means that no entropy is being generated between the heat exchangers temperatures seen
by the working fluid, according to the endoreversible model.
The thermal inertia of the chilled two-fluid side of the evaporator is large enough
such that the time to achieve the desired water level in the evaporator drum can
 427

QH 0
mm!!l
generator
T HC ,..----------,
reversible
compartiment
(C)

ToC
condenser Q 0 .u-

TO ~

TL ,""lI!iYlik
evaporator QL.u-

T LC .--------'

Figure 2. Absorption refrigerator model with three heat transfer irreversibilities.

be neglected when compared to the time evolution of the temperature inside the
refrigerated space.This justifies the use of a quasi-steady model [1, 2]. Additionally, it is
assumed that the water and trapped air in the evaporator drum are in thermal
equilibrium, such that Tw = I'air = TL .
Due to heat transfer, mixing and throttling, the refrigerator operation is
irreversible. The endoreversible model considers the refrigerator to be internally
reversible, such that only the external irreversibilities associated with the heat transfer
interactions are taken into account, namely QH' QL and Qo. Since the endoreversible
transient model was developed based on previous steady-state models [1, 2], only the
new equations introduced in the model for the present work are shown, for the sake of
brevity.
The heat input rate, QH is proportional to the fuel mass flow rate, where, without
loss of generality, we assume stoichiometric combustion, in an adiabatic chamber, and
negligible heat loss to the ambient through the exhaust gases, as follows:

(1)
The refrigerated space model was based on an evaporator coil surrounded by a drum
where chilling water circulates, as shown in Fig. 3. This configuration is part of an
experimental study of a gas-fired absorption unit [10]. Therefore, the model allows for
the validation of steady-state theoretical results against steady-state experimental data.
The transient cooling of the refrigerated space is accounted for by writing the first-law
of thermodynamics, as follows:
428

dTw dT;.ir. ( ).
mwcw - - + mairCv air - - = mwcw 1in - Toul
- QL + q (2)
dt . dt
where q accounts for the heat leakage from the ambient across the walls of the
evaporator drum.

Ammonia e<>
on
on

in
Chilled
water ....-~
out
- -
----
..:"""...;.. - - -"

Evaporator drum

Figure 3. The refrigerated space of the experimental unit.

2.00 . , . . - - - - - - - - - - - - - - - - . . ,

I.SO
Qi xlOO
tout
1.00
'" •

0.50 +-__--r--.-..--.--__-.,...--.-..--.---.--.,...--.-....--t
0.020 0.025 0.030 0.035

Figure 4. Comparison with experimental data (symbols)

and theory (lines).

The factors (UA) H' (UA) L and (UA)o represent the overall thermal conductances
of the heat exchangers. The overall thermal conductance of the walls of the refrigerated
 429

space is given by (UA)w' Since (UA) H' (UA) Land (UA)o are not known, the
following constraint is introduced at this stage:

(3)

In addition, defme the ratio w that accounts for the thermal conductance of the heat
transfer area of the vessel walls,

w=(UA)w /UA (4)

The complete set of nondimensional equations for the model, including the absorption
refrigerator steady state equations [1, 2], is

Q~=MC Q~ = Y{TH - THC) (5,6)

Q; =(1-y-z)(roc-1) Q~ =Z(TL -TLc) (7,8)

Q~ + Q~ = Q;- (9,10)
'r HC 'r LC 'roc

(11,12)

where 'r L is the arithmetic mean of the chilled water inlet and outlet temperatures. The
nondimensional groups have been defmed as follows:

'r H = TH / To •'r L = TL / To •Tin = Tin / TO •'r out = Tout / To (13)

'r HC = THC / To •TLC = TLC / To •'r OC = Toc / To (14)
. • Q; . • Q~ . * Q; . • Q~ (15)
Q, = UAT. .QH = UAT. .Qo = UAT. .QL = UAT.
o 0 0 0

a =
mairCvair
. (16)
mwcw

(17)

The conductance allocation ratios are y = (UA)H / UA.z = (UA)L / UA.

According to the UA constraint in Eq. (3), the conductance allocated to the
condenser is (UA)o / UA = 1- y - Z .
During off-cycle periods, the fuel mass flow rate drops to zero instantaneously and
the water is kept circulating in the evaporator drum, thus, the model accounts for both on
and off operation.
430

A complete treatment on these topics is available in the recent work of Vargas et

al. [12]. Starting from Vs = K~KI!;.Z:OO",rol+K2) where Kc' K J and K2 are constants to be
selected according to the system. the voltage input to the net gain block, Va and,
consequently, the voltage input to the fuel pump block is determined by Va = V s ' if
V, < Vmax ' and Va = Vmax ' if Vs ~ Vmax .

3. Experimental Validation

The validation of the absorption refrigerator model presented in this study was
performed by direct comparison of the model's steady state results with experimental
data [10]. No available experimental data has so far been found in the literature with
measurements of the transient behavior of an absorption unit.
The procedure consisted of evaluating the nondimensional parameters defmed by
the theoretical model for the particular unit tested by Horuz [10). The unit was a 10 kW
gas-fired vapor absorption refrigerator water chiller using aqua-ammonia solution, with
an evaporator configuration as shown in Fig. 3. The air cooled absorber and condenser
heat exchangers account for the heat rejection of the cycle.
The unit uses a water-ethylene glycol solution, an antifreeze for applications in
process cooling at lower temperatures. The composition of the solution used in the
experiments was 30 % by weight ethylene glycol and 70 % by weight water, giving [11]
C w = 3.77 kJ I (kg. °C) and Pw = 1040 kg I m 3 •

Through the available experimental data, thermal conductances of the heat

exchangers were computed as (UA) L = 0.87 kWloC, (UA)o = 0.89 kWloC and
(UA) H = 0.34 kWlo C, therefore (UA) = 2.1 kWlo C. The other necessary data were
mw =0.52kgls, H w =0.29m, H air =0.26m, TH =130°C, QH=23.2kW and
(UA)w = 1.32 Wlo C. The resulting dimensionless parameters, initially kept fixed are
r H = 1.4, Y = 0.16, Z = 0.41 , w = 0.0007, r = 0.934 and a = maircv,air / mwcw =
HairPaircv,air / H wPwcw = 1.64E -4, and Q~ = 0.038.
The experimental work of Horuz [10] analizes the effect of variations in several
parameters on the performance of the unit. In this work, theoretical and experimental
results are compared for the effect of the variation of chilled water inlet temperature,
chilled water flow rate and heat input to the system.
The experimental measurements shown in Fig. 4 were obtained with To = 19 °C .
The water level in the evaporator drum was H w = 0.29 m . The experimental data are
presented discretely with symbols (triangles and squares), whereas the model's results
are presented with continuous lines.
In Fig. 4, it is seen that the increase in heat input provokes an increase in cooling
capacity and a decrease in the water outlet temperature, in the tested heat input range,
with r in = 0.990 and r = 0.934. However, the decreasing rate of the water outlet
temperature is considerably smaller than the increasing rate of heat input. Therefore, the
decrease in water outlet temperature does not justify the increase in heat input. A
 431

transient analysis, such as the present model provides, to predict the ''pull-down''
capacity of the unity is mandatory for the designer, in order to achieve the project
demands. The theoretical model predicts a plausible "pull-down" trend, but as yet no
experimental data is available on which to base a comparison.
The analysis of Fig. 4 therefore validates the steady state results of the theoretical
model against experimental data. Based on these observations, the next section uses the
theoretical model to predict the "pull-down" capacity of an absorption unit, and to
perfonn a simulation of its transient behavior during a specified time of operation.

4. Results and Discussion

The model uses a total thermal conductance constraint, UA, as shown in Eq. (3).
Nevertheless, it is interesting to observe that, from a practical standpoint, for heat
exchangers with the same geometrical operating characteristics, i.e., similar global heat
transfer coefficients, the UA constraint turns into an area constraint. Therefore, for a
given total thermal conductance inventory, the designer could access the optimal thermal
conductance (or heat transfer area) allocation, such that (i) the cooling capacity is
maximized, and (ii) the ''pull-down'' time is minimized, by the use of the model.
The first fundamental question, concerning the optimal fraction allocated to the
condenser, has been answered in previous studies [1,2], where it was shown theoretically
that (UA) 0 / UA is insensitive to the external temperature levels Cr H' 'l'L) and equal to
half of the total thennal conductance for maximum refrigeration rate. In general, the
other half must be equally divided between generator and evaporator. To verify this
trend, by means of a numerical example, Fig. 5 shows the effect of the variation of the
total thermal conductance allocation on the steady state results for

2.00 , - - - - - - - - - - - - - - - - - - ,

QL xlOO
1.50

'tout
1.00

0.50T-~~_r~~~~~_r~~~~_r~~

0.10 0.15 0.20 0.25 Y 0.30

Figure 5. Maximum refrigeration rate.

432

Q~ and rout, using the experimental parameters of the tested unit. It has been further
assumed that (UA)o / UA = 0.5, and once Y is specified, Z results, and vice-versa, the two
collapsing in just one degree of freedom.

2.00 -r-------------------,

1.50

1.00

0.50 +-~-r-.-..,_.....-~r-T.....,.......,...........__r__"1.....,.~_r__r_"T"""""I

0.10 0.15 0.20 0.25 0.30

y

Figure 6. Minimum "pull-down" time.

The results show there is an optimal thermal conductance allocation to attain

maximum refrigeration, i.e., Yopt = Zopt ~ 0.25. Therefore, by computing

Q~,max = 0.0188 and rL,min = 0.967 with the model, and comparing with Q~ = 0.017
and r L = 0.970, obtained experimentally, the existence of a better design for that unit
may be inferred from the model, with a possible gain of 10.6 % in refrigeration rate and
0.3 % in chilled water outlet temperature.
The answer to the second fundamental question is given by the results shown in
Fig. 6, where a desired chilled water outlet temperature, rout ,I = 0.98, was established. It
verifies the existence of a minimum "pull-down" time, () p' when the thermal
conductance is allocated in an optimal way (Yopt = Zopt ~ 0.25). Using the model to
compute ()p for the tested unit, it is found that ()p = 1.2 , for y = 0.16 and z = 0.41 , and
() p = 1.0 for Yopt = Z opt = 0.25, i.e., with the optimal thermal conductance distribution,
a gain of 17 % is expected in the "pull-down" time. The importance of the transient
analysis to identify the best design is thus stressed.
Next, using the optimal thermal conductance allocation, a simulation of the unit,
operating with closed loop and "on-off" control, is performed for a specified period of
 433

time, Of = 4. The following parameters were adopted in the simulation: C = 0.15,

r = 0.630, y = z =0.25, w= 0.0007, a = 1.64E -4, Vmax = 4.8 V, K net = 25 ,K2 = 0,
Ka=4000, K,=40pV/oC, Kc=lOV, K1=lOoC- 1 , Kpump=2.1E-3V- 1 ,
To=17°C, 1in=20°C, TH=130°C, Tout,max=20°C, Tout,min=-lOoC and
Tout = 1in = 20 °C , at 0 = 0 (initial condition). The temperature setpoint in closed loop
was T,et = 13 °C, and the "on-off' control higher- and lower-temperature were
Tout,h = 14 °C and Tout'! = 12 °C , respectively.
The comparison of fuel consumption between the "on-off' and closed loop
systems, operating with the same parameters is depicted in Fig. 7. The results show that
the closed loop system requires approximately half of the fuel the on-off system does,
during the simulation time ( Of = 4 ).

0.8,--------------------,
M

0.6

0.4

0.2

O~~~~~~~~_r~~~~~~~~~~

o 1 2 9 3 4

Figure 7. Fuel consumption comparison.

This reduction increases as the unit works longer, since the slope of the "on-off'
fuel consumption curve is bigger than the closed loop one. The use of closed loop
control in absorption refrigerators is therefore very promising. Since a simple control
system was used and no additional cost is required with respect to the "on-off' system,
the same electrical fuel pump can be used in both systems, with obvious economic
advantages.
Next, the issue of how sensitive the closed loop control system is to changes in the
chilled water inlet temperature is addressed. The total simulation time was increased to
Of = 20, and the water inlet temperature started as 20 °C up to e = 4 , then increased
to 22 0 C up to e = 10, when it was set to 18°C, for the remainder of the simulation
434

period. The chilled water outlet temperature does not vary at all, in spite of the
variations in the water inlet temperature, as it is shown by the results of Fig. 8.

25
M

20 -

15
\ 'tout

10

5
MxlOO

o
o 5 10 e 15 20

Figure 8. Closed loop transient simulation.

When the water inlet temperature increases, there is only a slight increase in the
controlled temperature, which in this simulation was less than 0.05 0 C . The fuel flow
rate is automatically adjusted by the control action, according to the variations in the
water inlet temperature, such that the water outlet temperature does not vary.

Finally, it is important to observe that there are three degrees of freedom in the
control action that have to be selected appropriately. The controller parameters,
Kc = 10 V, Kl = 10 oC- 1 and K2 = 0 were selected for this system by trial and error,
such that liTcontrol = 0.05 0 C was the maximum deviation observed in the simulations
performed, after stabilization at the temperature setpoint was achieved.

5. Conclusions

In this paper, a transient model of a heat driven refrigeration system, operating with
continuous temperature control, was presented to investigate the behavior of the system
under "on-off' and closed loop operation, for design and optimization purposes. The
absorption refrigerator model was validated for steady state operation by direct
comparison between theoretical results and experimental data, with good qualitative and
quantitative agreement. Appropriate dimensionless groups were identified and the
generalized results reported in charts using dimensionless variables.
The model was then used to identify the optimal allocation of a fixed thermal
conductance inventory for minimum "pull-down" time combined with the maximum
 435

refrigeration rate. The most important conclusion of this second part of the study is that
the gain in the reduction of the refrigerated space temperature is much smaller than the
gain in the "pull-down" time. The importance of the transient analysis to identify the
best design is thus clearly seen.

A simulation of the system operating with "on-off''' and closed loop control was
performed, for a specified period of time, in the third part of this study. The continuous
temperature control eliminates undesirable oscillations in the controlled temperature,
observed with the traditional "on-off" control. The most promising result is that a large
reduction in fuel consumption, of about 50 %, was obtained with the closed loop system,
in comparison with the "on-off" control, even for a short simulation time (Of = 4 ).
Note that this result is plausible because, with the "on-off" control there are only two
fuel flow rate settings, maximum (on) or zero (off) and the ontime ratio is about 70 to 80
%, whereas the closed loop system works at a well reduced fuel flow rate after the
temperature setpoint is achieved. Since the same electrical fuel pump could be used in
both systems, no changes in hardware are required, except for those required for the
implementation of a simple control system. Hence, an obvious reduction, both in
operational cost and in environmental impact, will be achieved, if a continuous
temperature control is implemented on heat driven refrigerators. The response of the
power law control system was tested by varying the thermal load during a longer period
of simulation (Of = 20), and no oscillations were observed on the controlled
temperature, whereas fuel consumption was automatically kept to a minimum.

6. References

1. Bejan, A, Vargas, 1. V. C. and Sokolov, M. (1995) Optimal allocation of a heat exchanger inventory in
heat driven refrigerators, International Journal of Heat and Mass Transfer 38,2997-3005.
2. Vargas, J. V. C., Sokolov, M. and Bejan, A (1996) Thermodynamic optimization of solar-driven
refrigerators, Journal of Solar Energy Engineering 118, 130-135.
3. Sokolov, M. and Hershgal, D. (1993) Optimal coupling and feasibility of a solar-powered year-round
ejector air conditioner Solar Energy 50, 507-516.
4. Worsfe-Schmidt, P. and Holm, F. (1989) Development field testing of a solar-powered block-ice plant,
Jutland Technological Institute Teknologiparken DK-8000 Aarhus C Denmark FHONIM 1-17.
5. Bejan, A (1989) Theory of heat transfer-irreversible refrigeration plants, International Journal of Heat
and Mass Transfer 32, 1631-1639.
6. Bejan, A (1995) Entropy Generation Minimization CRC Press, Boca Raton, FL.
7. Bejan, A. (1988) Theory of heat transfer-irreversible power plants, International Journal of Heat and
Mass Transfer 31, 1211-1219.
8. Vargas,1. V. C. and Parise, J. A R. (1995) Simulation in transient regime of a heat pump with closed
loop and on-off control, International Journal of Refrigeration 18:4, 235-243
9. Kuo, B. C. (1987) Automatic Control Systems Prentice-Halllnternational Editions.
10. Horuz, I. (1994) An experimental study of the use of vapor absorption refrigeration in road transport
vehicles PhD thesis University of Strathclyde, Glasgow, UK.
II. Perry, R. H. and Chilton, C. H. (1974) Chemical Engineering Handbook International Student Edition,
McGraw-Hili.
12. Vargas, J. V. C., Horuz, I., Callander, T. M. S., Fleming, J. S. and Parise, 1. A. R (1998) Simulation of
the transient response of heat driven refrigerators with continuous temperature control, International
Journal of Refrigeration, in press.
INTELLIGENT COMPUTER AIDED DESIGN, ANALYSIS,
OPTIMIZATION AND IMPROVEMENT OF THERMODYNAMIC
SYSTEMS

c. WU
Department ofMechanical Engineering
u.s. Naval Academy
Annapolis, MD 2 J402, USA

1. Introduction

Research in Cognitive Science potentially offers tremendous benefits in education.

Psychological studies of human learning yields insights as to what forms of instruction
might be most beneficial for particular purposes and what mental models and
misconceptions students might have. Artificial Intelligence (AI) research provides formal
representations and reasoning techniques that can be used in new kinds of educational
software which itself contains a deep understanding of the material being taught. These
scientific advances, coupled with dramatic changes in computing technology provide new
opportunities for exploring how intelligent software can be used in education.
A new kind of educational software called articulate virtual laboratories (A VL) has been
implemented into the engineering curriculum at the US Naval Academy since 1996. An
articulate virtual laboratory is designed to help students learn thermodynamics by
scaffolding them in conceptual design tasks, using a simulated environment to allow them
to create and analyze designs for complex thermodynamic systems and using software
coaches to provide guidance.

2. Articulate Virtual Laboratory

The rapid increase in the power of microcomputers combined with rapidly falling prices,
provides the opportunity to apply recent advances in artificial intelligence technology to
new kinds of educational software. By exploiting advances in qualitative physics,
educational software can use methods and concepts similar to those used by domain
experts, including the tacit knowledge of the domain. By exploiting advances in reasoning
technology, particularly truth-maintenance systems, educational software can be built that
explains the principles it uses. The ability of qualitative physics to provide
computer-assisted modeling capabilities combined with efficient reasoning techniques
means that educational software can work with new problems in the domain, without
reprogramming. Such generality will empower students and instructors by enabling them
437
A. Bejan and E. Mamut (ells.), Thermodynamic Optimivltion o/Comp/a Energy Systems, 437-444.
© 1999 Kluwer Academic Publishers.
 438

to pose their own problems, and will thus provide a more motivating instructional context.
By exploiting advances in analogical reasoning, we can use libraries of examples and
previously-solved problems to provide a foundation for coaches that introduce students to
appropriate problem-solving methods in a domain.

3. The Role of Design

Design is the core activity of engineering. The ability to develop useful artifacts by
understanding how to apply fundamental principles to applications is a major reason why
engineering education is so important in today's highly competitive environment. Providing
design experience is clearly essential for training engineers. However, design activities
also provide powerful motivation for learning fundamental physical principles. Design
requires using knowledge in an integrated fashion rather than memorizing isolated facts.
Therefore, design provides a meaningful context for learning physical principles. Getting
students to think in design terms leads naturally to building a strong interest in
understanding complex, real-world relationships.
The most important class of educational software for engineering education is oriented
around conceptual design activities. In engineering, conceptual design is the early phase
which identifies the basic features of the design. This includes specifying the physical
processes and principles by which the design achieves its goals, as well as performance
estimates and rough cost estimates.
Conceptual design is most appropriate for engineering education because it focuses on
the fundamentals of a domain. On the other hand, detailed design is concerned with
fleshing out the results of conceptual design, including figuring out the geometry of the
system, detailed parts costs, operating and maintenance procedures for the resulting
artifact, and all of the other issues which must be settled in order to build and operate the
newly designed artifact. Engineering students need detailed design experience, of course.
However, we believe that the largest pedagogical payoff will be in computer technology
that simplifies broad-scale incorporation of conceptual design in engineering curricula.
Conceptual design forces the designer to think directly about many of the major themes.
Despite the centrality of design in engineering, today's engineering curricula generally
offer few design experiences. Some of the reasons come from the cost of, and sometimes
danger in, building physical artifacts. Additional reasons that design exposure is limited
include (I) the time-consuming nature ofdetailed design which strains already-overloaded
curricula, and (2) the complexity of design which requires additional supervision to
minimize unenlightening aspects of student explorations.
Virtual laboratories provide a partial solution to these dilemmas. A virtual laboratory
is a software environment consisting of "parts" corresponding to physical parts, tools for
assembling collections of these parts into systems, and facilities for analyzing and testing
the systems. These software environments allow students to "build" their designs and try
them out without expense or danger. Thus, these programs solve the cost and safety
problems associated with design. However, they provide no coaching. They do not explain
their results in qualitative terms that students can easily understand. They cannot teach
students how to choose an appropriate model or how to build new models. They have no
representation of the student's goals so they cannot assist the student in evaluating his or
her design. Finally. they cannot help the student understand the connections between the
 439

simulation results and the fundamental principles used to generate those results. To
overcome these I imitations will require synthesizing artificial intelligence technology to
develop articulate virtual laboratories, that is, intelligent learning environments with a deep
understanding of engineering principles.

4. CyciePad

CyclePad is the conceptual CAD component of the engineering thermodynamics A VL

developed by Professor K.D. Forbus [I] of Northwestern University and evaluated by the
author [2]. CyclePad scaffolds the activities of students who are learning how to design
and analyze thermodynamic cycles. An inventory shop of parts (e.g., compressors,
turbines, pumps, heat exchangers, etc.) are combined into networks, potentially generating
an unlimited set of designs of thermodynamic systems. CyciePad performances the
operational characteristics and sensitivity analysis of the system.
The design of thermodynamic systems is a significant element of engineering education
that requires a broad and deep understanding of the fundamentals of thermodynamics.
Most introductory engineering thermodynamics textbooks devote several chapters to
system analysis. In some advanced books, system analysis is the main concentration.
Developing software architecture that scaffold students in conceptual design problems,
allows professors to bring advanced concepts into the curriculum at an earlier time.
CyclePad is a conceptual CAD tool for visualizing and analyzing thermodynamic
systems. Students construct their design by "wiring up" components and assigning
numerical values for their parameters.
Due to the substantial amount of numerical calculation involved, students typically
avoid exploring more than one or two alternatives and avoid performing advantageous
trade-off studies. Since trade-off studies are one of the most effective means to building
intuitions about thermodynamic systems, CyciePad's value lies in the ability of the
students to incrementally derive the consequences of assumptions and automating the
work, thus relieving the student of the tedium of numerical calculations. Another major
stumbling block for students is learning how to model, that is, how to make the simplifYing
assumptions needed to derive reasonable answers. CyclePad helps by making the choices
for modeling assumptions explicit in its representation and depiction of the design.
Therefore, the students are encouraged to consider what reasonable assumptions might be
applicable. One benefit of Articulate Virtual Laboratories is to decrease students' primary
difficulties and intuitive but faulty conceptions within the domain, based on empirical
evaluations of their performance using A VLs.
CyclePad' s knowledge base captures a substantial fraction of the knowledge in
introductory thermodynamics text. The knowledge base consists of representations for the
components of thermodynamic systems and qualitative representations of the physical
processes inside them.

5. Using Articulate Virtual Laboratory in the Classroom

Undergraduate students in mechanical engineering atthe U.S. Naval Academy are required
to take two courses in thermodynamics. Since thermodynamics is such a problem-oriented
course, students must learn to work cycles to understand thermodynamics. To advance and
440

challenge the intellect, students must be able to calculate the problem in a reasonable
amount of time by hand. Extraneous or more tedious work will only tend to frustrate and
dissuade the students from contemplating further designs.
Normally, calculating system design is a lengthy, complex and tedious for a student.
Analyzing cycles is more meaningful, time saving and fun for students using quick-change
simulation software such as CyclePad in design. Students are able to change any parameter
in the cycle and see the effect of the parameter on the performance of the cycle
instantaneously. Design is an iterative learning process. By trial and error, students are
able to choose various components in the design of power plants or choose different
working fluids in the plants using CyclePad as a tool. They gained a tremendous amount
of experience, confidence, and knowledge in a relative short time. Due to the quick
iterative design abilities of CyclePad, instructors were able to assign more cycle analysis
and design problems to their students.

6. Example: Improvement of a Brayton Cycle

Starting with the simplest possible Brayton cycle, improvements can be progressing
through a number of modifications of the basic configuration. These modifications, each
made at some additional components from the inventory shop of CyclePad, are typically
evaluated for their contribution to performance criteria such as efficiency and power
output. The modifications include the addition of a reheater, the addition of a regenerator,
the addition ofan inter-cooler, the addition of any combination of inter-coolers, reheaters
and regenerators, etc. This gradual approach introduces to the students how Brayton
cycles operate and how to improve plant performance.
An innovative Brayton cycle for potential naval shipboard power plant application is
used as an example. The cycle consists offour compressors, one combustion chamber, two

• .... !
turbines, two coolers, one source and one sink as illustrated in Fig. I.

• 85 fiR

TUR2 CMP2

TUR1
'~'I 87

'~'T SB

~~'
I"~
82

•• •
S1 81NK1 CMP4

1" ~ . 810 89
SOURCE1 CMP1

Figure I. Example cycle.

In the cycle, air enters from the atmospheric source to an isentropic compressor at 293
K and I bar (state 1) and leaves at 8 bar (state 2); air enters the combustion chamber
(heater) and leaves at 1373 K (state 3); air enters the high-pressure isentropic turbine and
 441

leaves at 1 bar (state 4); air enters the low-pressure isentropic turbine and leaves at 0.04
bar (state 5); air enters the first-stage isentropic compressor and leaves at 0.2 bar (state 6);
air enters the isobaric low-pressure inter-cooler and leaves at 293 K (state 7); air enters the
second-stage isentropic compressor and leaves at 0.6 bar (state 8);
air enters the isobaric high-pressure inter-cooler and leaves at 293 K (state 9); and air
enters the third-stage isentropic compressor and is discharged to the atmospheric sink at
1 bar (state 10). Once the cycle is built, the mode is changed to analysis.
The user can analyze a cycle the same way as traditionally done by defining each device
(process) and inputting numerical values as shown in Fig. 2 and Fig. 3, respectively. The
properties of the working fluid at each state, and the performance of each device and the
cycle are quickly calculated by the software. The results ofthese calculation are displayed
as shown in Fig. 4. The thermodynamic efficiency and the net power output of the cycle
are 59.67% and 504.2 kW, respectively.

-"""1+
86 86
t -• •
. . . . . . . . . . . . . . . . . .I....

CMP2 ~
II"

t
CLR1

87

r: 'CLR2~
' J.,
L.-:t-..-l
810 89

Figure 2. Processes of cycle.

Substance: AIR
Pha•• : GAS
T • ZO.OO·C

[ 8:
,..~ .........S.2.......
SOURCE1 CMP1
...----
810
.. ....
CMP4

...-
59
--'

Figure 3. Input parameters of cycle

442

. ,.r. .
85

CMP2
t -
86
..........---.,

CLR1
...
~

t
--.-.-1
Modeled as: not HEAT-PUMP 87
Hodeled as: not REFRIGERATOR
Modeled as: HEAT-ENGINE
eta-Camot CMP3'f'

i
= 78.6S~
eta-themal .. 59.67%
'l'max
= 1,100'C
Tmin= 20.00·C S8
Pm.ax
.. 8.00 bar.
Pmin= 0.0400 bal: 81NK1
Power. in'" -570.4 kl))' CLR21

~~.f-
Power: out.. 1,075 kTJ
net-power = 504. Z ktJ
back-work-racio = 53.08%
TJOI:k-r.atio = 46.92% .......
a-dot in • 845.0 kTJ
a-dot out - -294.6 kUJ 810 89
net a-dot • 550.4 klJ ~ •..:

Figure 4. Results of cycle.

In order to find the effect of an input parameter to the perfonnance of the combined
cycle, one may do a sensitivity analysis. For example, one may choose the cycle efficiency
(TI) to be a dependent parameter, and the outlet pressure of the second-stage compressor
(Ps) to be an independent parameter. Selecting a range for the pressure (Ps), a graphical (TI-
Ps) sensitivity analysis is immediately displayed as shown in Fig. 5. It is seen that the
maximum cycle efficiency occurs at Ps equals to 0.45 bar.

59.60
59.40
59.20
59.00
58.80
58.60
58.40
58.20
58.00
57.80
57. 60+---II---+-+--I----+--+--t-~
0.20 0.40 0.60 0.80 1.00
P (bar)

Figure 5. Graphical (TJ-P.) sensitivity diagram.

 443

Design of different arranged Brayton cycles can be easily perfonned by the usage of the
computer software. Refinements of the three-stage compressor cycle could include more
reheaters and more turbiness added, and more compressors and more inter-coolers added,
etc.

7. Discussion

In the realm ofthennodynamics, CyclePad is to a thennodynamist what a word processor

is to a journalist. The benefits of using this software are numerous. The first is that
significantly less time will be spent doing numerical analysis. As an engineer, this is much
appreciated because computational work that would have taken hours before, can now be
done in seconds. Second, as this paper has demonstrated, Cycle Pad is capable of
analyzing thennodynamic systems with various working fluids. Third, due to its
computer-assisted modeling capabilities, the software allows for individuals to
immediately view the effects of varying input parameters, either through calculated results
or in the fonn of graphs and diagrams, giving the user a greater appreciation of how a
system actually works. More specifically, there is a feature that provides the user the
opportunity to optimize a specified cycle parameter. This has been proven to be incredibly
useful. Last, and most important, is the built-in coaching facility that provides definitions
of tenns and descriptions of calculations. CyclePad goes a step further by infonning the
user if a contradiction or an incompatibility exists between inputted parameters within a
cycle and why.
When viewing the applicability of software of this nature on the drawing board, users
could reap the benefits at all stages in an engineering career. For the young engineer just
beginning the learning process, less time is spent doing iterations resulting in more time
dedicated to reinforcing the fundamentals and gaining valuable experience. In the case
ofthe seasoned engineer, who is well indoctrinated in the principles and has gained an
"engineer's intuition", can augment his/her abilities by becoming more computer literate.

8. Conclusion

The intelligent computer software has been demonstrated to be a very effective tool in
analyzing thennodynamic cycles. Any complicated cycle can be analyzed easily using
CyclePad. The advantage of using CyclePad is not only does it provide quick and accurate
cycle analysis predictions, but that it also forces the user to define each process in the
cycle and to make reasonable input infonnation to the cycle. Users using CyclePad in
designing cycles could be freed from tedious table searching and hand calculations. They
could spend more time doing optimization of operational parameters and selection of
different combinations of working fluids of the thennodynamic systems.

Acknowledgement. This paper is a result of a research contract sponsored by ONR

(Office of Naval Research) under the award number NOOOI498AF00002.
444

References

I. Forbus, K. (1994) Self-explanatory simulators: making computers partners in the modeling process,
Mathematics and Computers in Simulation 36, 91-97.
2. Wu, C. (1998) Using articulate virtual laboratories in teaching energy conversion at the U.S. Naval
Academy, Journal {,{Educational Technology Systems 26,127-136.
A STUDY OF THE LARGE OSCILLATIONS OF THE THERMODYNAMIC
PENDULLUM BY THE METHOD OF CUBICATION

G. STANESCU
Federal University of Parana. Polytechnic Center. Mechanical
Engineering Department. Box 190 II, 81531 - 990 Curitiba. PR, Brazil

I. Introduction

Study of mechanically oscillating thermodynamic systems (thermodynamic pendullums

or oscillators) has been neglected for a long time, partly because no exact analytic
solution of the corresponding system of differential equations modeling the processes is
known [I], and partly because of lack of practical interest. During the last two decades,
space, industrial and domestic applications offree piston thermal machines (FPTMs) and
progress in numerical methods have stimulated scientific interest in this field since the
thermodynamic pendullum represents an appropriate physical model to study FPTMs
functioning.
Study of small vibrations (perturbations) of thermodynamic oscillators [2 - 3]
contributed to shape this new research area. It is very attractive because of simple
analytic solutions, but did not provide general results. By considering the speed of
processes to be infinitesimally small, Keller [2] determined the relaxation of small
oscillations of a thermodynamic pendullum with friction. He used the classical
thermodynamics laws, without conceptual difficulties generated by the finite non-
vanishing speed of processes. and neglected simultaneously the inel1ia force of the
moving piston mass. Due to the last hypothesis Keller's results are not suitable to model
large oscillations and the FPTMs functioning. Brening [3] studied the dynamics of
thermodynamic oscillators with thermal and mechanical attenuation. He developed a
very laborious matrix method to determine approximately the characteristic frequencies
of "pure thermal" and "pure mechanical" vibrating modes, corresponding to the limiting
case of isothermal and adiabatic oscillator, respectively. Since it takes into account the
heat exchange through the piston between the two gas masses, the Brening's method
represents a better approach in studying FPTMs functioning and it makes possible to
model the influence of heat flow into the regenerator on the FPTMs output.
FPTMs optimization to satisfy the "criteria of L4" (low pollution, low cost. low
loss. long life) emerged new problems related to the resonance of nonlinear oscillators
[4] and to the existence and stability of limit cycles of motion [5]. The present paper
presents a study of free and forced large vibrations of conservative and dissipative
thermodynamic pendullums, based on the "cubication" of nonlinear oscillators method
445
A. Bejan and E. Momut (eds.), Thermodynamic Optimization o/Complex Energy Systems, 445-452.
© 1999 Kluwer Academic Publishers.
446

of Yuste [5]. Similarly to the mathematical pendullum case [6], it is possible to calculate
the period of large oscillations of the thermodynamic pendullum by using the elliptic
integrals. We found that period of large oscillations depends on initial conditions (initial
speed and position ofthe moving piston), contrarily to the small oscillations that develop
during the same period [7].

F,

The pressure forces with which

the gas I
acts on the moving piston V\PPI)

F,

-I o x x
Figure I. The initial out-of-mechanical-equilibrium position ofthe moving piston (top). and forces acting on
the moving piston during oscillations (bottom).

To study forced oscillations we consider a dissipative mechanism that takes into

account effects of mechanical interaction with finite variable non-vanishing speed
between a gas into a closed system and a moving piston [8]. For the large forced
oscillations case we determine the exterior forcing forces to get stable limit cycles of the
movement of thermodynamic pendullum.

2. Physical Model

The thermodynamic pendullum or oscillator is modeled in here using a cylinder that

contains a moving piston (Fig. I). The piston is assumed free to move, with or without
friction, and to provide perfect adiabatic insulation and to provide a seal between the
gases in the cylinder. The two compartments within the cylinder, containing the same
mass of an ideal gas, are each assumed to be closed system. The cylinder walls may be
considered either adiabatic or non-adiabatic.
 447

The piston motion is the result of forces, including externally imposed force
function, continuously acting on it. The external forcing force may for instance be
applied by a linear motor surrounding the cylinder. We assume that both external forcing
force and frictional force depend on the instantaneous piston speed.
Thermodynamic processes occurring into gases during oscillations are assumed to
be either adiabatic or isothermal. Initially, temperature of both gases is the same. For the
sake of simplicity of equations, in Fig. I the zero point of the X-axis coincides with the
initial out-of-mechanical-equilibrium position of the moving piston. The initial piston
speed is assumed to be positive and non-zero. We consider a dissipating mechanism that
takes into account losses generated by the finite speed of mechanical interaction between
gases within the cylinder and the moving piston, as described in [8].

3. Mathematical Model

We consider the system in Fig. I. The initial out-of-mechanical-equilibrium state is thus

characterized by the same values of pressure Po and temperature To of the ideal gas
within the two compartments of the cylinder. For the sake of simplicity of equations we
consider also that gases fill initially equal volumes. The heat produced by the mechanical
friction between the piston and the cylinder does not return to the system, but is
considered lost to the surroundings. Balancing the forces acting on the moving piston we
can write the equation:
mx = P1[ \- ax f (3RT1)1/2]A - P2[ \ + ax f (3RT2 )1/2]A + f<~. + F(t) (I)
where PI' P2 are the instantaneous pressures of gases, m is the mass ofthe moving piston,
A is the cross section area of the piston, Fr=-Il' X is the friction force between the piston
and the cylinder, and F(t)=Tj' x is the exterior forcing force function, with coefficients
11'>0 and Tj'> O.
To consider losses generated by the mechanical interaction with finite speed, we
assume in Eq. (I) that the pressure with which a gas into a closed system acts on a
moving piston is given by Pp=P(\±aw/c) [8]. P is the interior instantaneous gas pressure,
x
w= is the instantaneous speed of the moving piston, c=(3RT)112 is the average speed of
molecules for the gas in the system, and a=(3k)lll is a coefficient that depends on the
nature of the gas, where k=cp/cv i... the ratio between the specific heats of gas at constant
pressure and constant volume. The sign (+) corresponds to the case dV<O, and (-) sign to
the case dV>O. The processes into gases are assumed to be reversible during oscillations,
and therefore the instantaneous pressures and temperatures can be written
Pdx) = Po (\ ± Ax / Vo)-n Tdx) = To (I ± Ax / Vo)l-n (2a, b)
by using a polytropic function with a constant coefficient n. In Eqs. (2a) and (2b) x is the
instantaneous position of the moving piston. We switch n= I to study isothermal
oscillations and n=k to study adiabatic oscillations.
Based on the Keller's idea, we calculate P1T1- 1/2 and P2Til/2 in Eq. (I) and term
into brackets in Eq. (2a) by using Maclaurin series expansions up to the fourth term
448

( I ± Ax)--11 ;:; 1+ n Ax + n(n + I) (AX)2 + n(n + I)(n + 2) (AX)3 (3a)

Vo Vo 2 Vo 6 Vo
11+1
( I±AX)-2 ;:;1+n+IAx+ (n+I)(n+3)(Ax)2 + (n+I)(n+3)(n+5)(Ax)3 (3b)
Vo 2 Vo 8 Vo 48 Vo
To get a higher precision we may consider. at least in principle. in the Maclaurin
series expansions any large number of terms, but we will arrive to deal on more and
more complicated differential equations. With X = Ax / Vo we can rewrite Eq. (I )
X+C1X+C3X3 +E(u+PX 2)X=O (4)
with coefficients cl=2nPoA2/(mVo). c3=n(n+I)(n+2)PoA2/(3mVo), E=aAPo/[m(3RTo)1/2].
a=2 + (f..I' - TJ*)(3RTo)II2/(aAPo) and 13=(n+ I )(n+3)/4. We note in here that -I < X < I.

4. Method of Cubication

Solution of the autonomous non-linear oscillator in Eq. (4) can be approximated by the
solution of associated "cubic oscillator"
(5)
as proved by Yuste [5]. Coefficients c~ • c; and 'A* are determined using the principle of
harmonic balance. With our notation the solution of Eq. (5) is given by
X(t) = Xm cn(oot + e. f..I2) = Xm cos <1> (6a)
<1> = am(oot + e. f..I2) == am(\II. f..I2 ) (6b)
where Xm is the amplitude of oscillations. and
sn(~J. f..I2) = sin <1> = sin [am(\II. f..I2)] (7a)
cn(\II. f..I2) = cos <1> = cos [am(\II, 1l 2)] (7b)
dn(\II, f..I2) = ~<1> = {1-f..I2sn2[am(\II. f..I2)]} 112 (7c)
are the Jacobi elliptic functions [9]. In defining the elliptic integral of the first kind \II
<I> d~
\11(<1>. 1l2) = f (8)
o ~1-1l2sin2~
<1> is the amplitude and f..I is the modulus.
In Eqs. (6a) and (6b) the frequency 00 and the modulus f..I can be written
00 2 = c~ + C;X~, f..I2 = (C;X~n /2) / (c~ + C;X~,) (9a. b)
For the "cubic" oscillator in Eq. (5). Yuste found c~ = CI and c; = C3. In our case
coefficient 'A* is given by
'A' = -E[ a + I3(T /Q)X~,] (10)
2n 27t
where Q = f sin2<t>~<t>d<t> T = f sin 2<t>cos2 <t>~<t>d<t> (Ila. b)
o 0
The method of harmonic balance is especially useful to study the stability of limit
cycles. In Eq. (5) we note the existence of such stable limit cycles for 'A*= O. The trivial
 449

solution of the last equation corresponds to a=/1*=T]*=O. In this case we study the large
oscillations ofa conservative thermodynamic pendullum. When studying the limit cycles
of the large forced oscillations with friction, the trivial solution of equation ",,*=0
corresponds to a=O and /1*=T]*. Requiring a;tO, /1*;t0 and T]*;t0, we determine the non-
trivial solution of equation ",,*=0 for large forced oscillations of a dissipative
thermodynamic pendullum.

4.1. DISSIPATIVE THERMODYNAMIC PENDULLUM

For the thermodynamic pendullum in Fig. I we are interested in determining solutions of

equation ",,' = 0, when a ;t 0, /1* ;t 0 and T]* ;t O. To simplify the computation we set,
without loss of generality, PoA2 1 (mVo) = I S-2. Coefficients in Eq. (5) become now
c; =2n, c; = n(n+I)(n+2)/3, &=aAPoI[m(3 RTo) 112], a=2 + (/1*-T]*)(3RTo)1/2/(aAPo) and
~=(n+ I )(n+ 3)/4. There is no difficulty in obtaining numerical results when considering
any other value of parameter POA21 (mVo). To determine the limit cycles of oscillations
of the dissipative thermodynamic pendullum we rewrite equation ",,'=0 in the equivalent
form
X2 =_~ Q(v) ( 12a)
111 l3(n) T(v)
Eq. ( 12a) combined with the equality
2 c~(n) 6v
XI11 = - . - v = - - - - (12b)
c,(n) (n+1)(n+2)
yield the following system of two algebraic equations
4a Q( v) 3 (n + 3) v Q(v)
(13a, b)
(n + I)(n + 3) T(v) 2 (n+2) a T(v)
where v = c;X~n / c; is the non-linearity factor, En = c~X~n (I + v /2) is the total energy of
oscillation, and V I11 = (c;)2 / (2c;) is the maximum potential, as defined in [5].
It is apparent in Eqs. (lla) and (II b) that Q(v) 1 T(v) > 0, and therefore the system
(I3a), (13b) gets real solutions when
_ 3(n + 3) 1 ::;-5 0< 4aQ(v) IT(v) < I
(14a, b)
2(n+2)a (n+l)(n+3)

or _ 3(n+3) I ~-1.329 0<- 4aQ(v)lT(v) < I (14a', b')

2(n+2)a (n+I)(n+3)
If we arrange Eq. (l3b) as rv = Q(v)/T(v), the two limiting values of slope r=-5 and
r=-1.329 correspond in Fig. 2 to the positions of line rv when it is tangent to the
Q(v)/T(v) curve. The limit cycles can be written
X(t) = Xs cn(\jJ, /1 2) )(t) = -Ol Xs sn(\jJ,/1 2) dn(\jJ, /1 2) (l5a, b)
We label the solutions of system (l2a) and (l2b) by the subscript S, (Xs, vs). For
n=1 or n=k we study a hard oscillator (c; >0 and c; >0). This is the case when 0::;/12::;1/2
and O::;v<oo, with 012 = c; (I + vs) in terms of the non-linearity factor. The period ofX(t)
450

is 4K(J.l2), where K(J.l2) = \jJ(Y:!n, J.l2) is the complete elliptic integral of the first kind. The
analytical expressions ofQ and T functions are given by [5]
Q = (4 / 3J.l2)[(2J.l2 - I )E(J.l2) + (I - J.l2)K(J.l2)] (16a)
T = (4 / 15J.l4)[2(J.l4 + 1-~l2)E(J.l2) + (I - J.l2)(J.l2 - 2)K(J.l2)] (l6b)
where J.l2 ~ vs.' [2( 1 + vs)] and E(J.l2) is the complete elliptic integral of the second kind.

4.2. CONSERVATIVE THERMODYNAMIC PENDULLUM

Equation of large free oscillations (T]' =0) of the conservative thermodynamic pendullum
(J.l' = 0) is written
X=-(c;X+c;X 3 ) (17)
In here we neglect also the losses generated by the mechanical interaction with
finite speed between gases in the cylinder and the moving piston (a=O). We consider the
following initial conditions in Eq. (17): X(O)=O and X(0)= Xo >0. Multiplying Eq. (17)
by 2 Xdt=2dX we have
( 18)
By integrating the left-hand side of the last equation between Xo and X, and the
right-hand side between 0 and X, we obtain
( 19)

6 r-----------~----------~------__,
Q(u) r = -5 : r=4 : :
5
T(u)
4

3
, , ", x ::
' :' r = 2.853
2 ",

.
'I

' .. ••••

", ',::, ~+oo

r=-1329 ,,,,,\' r - v
. ...:1
0
-5 -4 -3 -2 -1 0 2 3 4 5

FiJ{lIre 2. Solutions ofEq. (l3b) for adiabatic oscillations (n = k = 1.4) plot the curve Q(v)rr(v) (heavy line) lor
o~ v ~ 1.332 (r~ 2.853), and tor isothennal oscillations (n=l) for O~ v ~ 0.982 (r~ 4).
or, by deriving the instantaneous speed of the moving piston

dX/dt=Xo[l-x2(c~+c;x2/2)/X~r (20)
We use the (+) sign in front of the square root since we considered before that
Xo >0, and therefore dX/dt>O during the first quarter period, Separating the variables in
Eq. (20) and integrating dt between 0 and t, we find
1 x dy
t= -. J (21)
Xo 0 ~( 1_ u2y2)( 1+ v 2y2)
 451

with (22a, b)

I ux d~
t = -.- f r=========:==== (23)
uXo 0 ~(I_e)(I+w2~2)
when considering ~ = uy, and w 2 = v 2 / u2. Now if we let v 2 / (u 2 + v2) = f.12 and ~ = coss,
Eq. (23) is written
I <I> -ds
t = f (24)
Xo~u2+y2 ,,/2 ~1-~2sin2s Xo~u2+y2
To derive X as an explicit function of time we rewrite the last equation as
\jJ(<I>, f.12) = K(f.12) - Xot( u 2 + y2 )1/2 (25)
and, since f.12 is a constant, we find
<I> = am[K(~l2) - Xot(u 2 + y2 )1/2 ] (26)
Finally, we obtain
X(t) = X Ill cos<I> = XIll cn {[K(f.12) - Xot( u 2 + y2) 1/2 ], f.12} (27)
with the amplitude XIll = I /u and the frequency (J)
(J)= XO(u 2 +y2)1/2 (28)
The instantaneous speed of the moving piston is given by
X(t)= X m XO(u 2 +y2)If2sin<l>~<I> (29)

5. Numerical Results

Fig. 3 shows comparatively the analytic approximate limit cycle of oscillations of an

adiabatic dissipative thermodynamic pendullum, given by Eq. (5), and the results of
numerical integration of Eq. (4). We assumed the ideal gas, filling both compartments
within the cylinder, to be diatomic (n=k= 1.4), and chose a=-0.3 and 1'=0.85. With these
values Eq. (4) is written
X+2.8X +3.808X' + 0.85(2.64X 2 -O.3)X = 0 (30)
The instantaneous position and speed of the moving piston are given by
X(t)=0.647cn(2.129t,0.191) (3Ia)
X(t) = -1.435 sn(2.129t, 0.191 )xdn(2.129t, 0.191) (31 b)
with the frequency (J) = 2.129 S·I, the modulus f.1 = 00437 and the period 4K(0.191 )=6.62.
Solutions of systems (l4a), (14b) and (14a'), (14b') are -0.675:::;a<0 for adiabatic
oscillations, when n=k=lo4, and -0.5:::;a<0 for isothermal oscillations, when n=1. The
parameter a is indicative for the external forcing force F(t). It is apparent from the
mathematical expression of TJ* that F(t) depends also on the values of coefficient a, that
accounts for dissipation. To get stable limit cycles of oscillations, the driving effort
increases for large dissipation.
452

1.5 .-:::-:-:_ _ _ _.,.-_ _ _ _- - ,

1.0

0.5

0.0

-0.5

-1.0

-1.5 '--_ _ _ _ _"'-_ _ _ _---1

-0.67 -0.47 -0.04 0.44 0.67

Figure 3. Numerical (thin line) and approximate analytical (heavy line) limit cycles of oscillations of the
adiabatic dissipative thermodynamicpendullum Eq. (30).

6. Conclusions

The described method provides a tool to study free and forced non-linear oscillations of
thermodynamic pendullums. By modeling actual FPTMs as thermodynamic pendullums
it is possible to investigate. based on this method. important issues of their design.
optimization and functioning. such as the existence and stability of limit cycles of
osci Ilations.
For an external forcing force function depending on the instantaneous speed of the
moving piston. it is also possible to determine its characteristics such that to obtain stable
limit cycles. The method can be easily extended to study resonance in FPTMs
functioning and to minimize the driving effort and maximize the FPTMs ouptut.

7. References

I. Minorsky. N. (1947) Introduction to Non-Linear Mechanic Edwards Brothers. Inc .. Ann Arbor. MI.
2. Keller. J. U. (1977) Thermodynamik der Irreversiblen Pro=esse. Walter de Gruyter & Co .. New York.
3. Brening. W. (1989) Statistical Theory q( Heat. Nonequilibrium Phenomena Springer-Verlag.
Heidelherg.
4. Wargitsch. C. and HUbler. A. (1995) Resonances of non-linear oscillators. Physical Review E 51. 1508-
1519.
5. Yuste. B. S. (1992) Cuhication of non-linear oscillators using the principle of harmonic balance.lnl. 1.
Non-Linear Mechanics27. 347-356.
6. Salvadori. M. G. and Schwarz. R. J. (1956) Differential Equations in Engineering Problem5 Englewood
Clills. Prentice-Hall. Inc.
7. Petrescu. S .. Stanescu. G .. and lordache. R. (1993) Study of reversible theml0dynamicpendullum in the
small oscillations hypothesis. Energetica Review 41. 66-70. Bucharest.
8. Petrescu. S. and Hannan. C. M. (1994) The connection between the first and second laws of
thennodynamics for processes with finite speed. A direct method for approaching and optimization of
irreversible processes. Journal q( Heat Transfer Society ofJapan33. 60-67.
9. Losch. F. (1960) Tables q(Higher Functions. McGraw-Hili Book Company. Inc .. New York.
10. Handhook q(Tablesfor Mathematics(l975). CRC Press. Inc .• Cleveland. Ohio.
Author Index

Abramovich, G.N., 278 Berg, C.A., 58, 60

Aceves-Saborio, S., 210, 219 Berry, R.S., 295,296, 363, 379, 383,
Adameshteanu, I., 303, 312 411-414,416,419-424
Agrawal, D.C., 385, 400 Beyer, J., 96, 100
Ahern, J.E., 59 Biermann, A.E., 173, 175
Ahlborn, B., 48, 50, 58, 60,74, 94, Bingham, E.R., 403, 410
346,363,383,411,420,423,424 Bird, R.B., 75, 94
Ahrendts, J., 81, 94, 225, 230 Blanchard, C.H., 385, 400
Andresen, B., 58, 60, 411-416, 418- Blank, D.A., 403, 410
420, 422-424 Boltyanski, V.G., 377, 384, 415, 420
Anmin, Q., 418, 420, 423, 424 Bourdri, J.C., 209, 219
Aoki, S., 175 Brandt, S., 23, 24, 43
Aris, R, 378, 384 Brening, W., 445, 452
Arpaci, V.S., 309, 312 Brush, S.G., 1, 17
Ataer, b.E., 231, 238, 240, 246, 250 Burzler, 1., 411, 419
Avery, W., 335, 344
Ayres, R, 155, 158 Callander, T.M.S., 430, 435
Azzarone, F., 154, 155, 158, 159 Callen, H.B., 34,44,412,420
Cambel, A.B., 155
Badescu, V., 314, 318, 324 Caplan, S.R., 282, 296
Bartolome, J.L., 121-124, 135, 136 Carey, V.P., 89, 94
Baruch, P., 320, 324 Carnot, S.N.L., 9, 17, 187, 194
Batzaga, N., 303, 312 Carnap, R., 309, 312
Becker, RH., 303, 312 Carslaw, H.S., 238, 240
Bejan, A., 1, 17,45-54,56-74,82,83, Cescotto, S., 312
85,93,94,96,97, 100, 101, 103, Chambadal, P., 48, 50, 57, 59, 60, 411,
104, 107, 108, 110, 114, 115, 164, 420
165, 175, 187, 194,209,211,219, Chandrashekar, M., 249, 250
221,230,231,236,240,247,250, Chapman, A.J., 212, 219
312,314,317,324,329,332,334, Cheilyakh, V.T., 48, 59
363, 378, 383-386, 388, 400, 403, Chen, F., 403, 410
410,425,427,429,431,435 Chen, J., 385, 400
Be1abbas, B., 403, 410 Chen, L., 363, 383, 385, 395, 397,
Bellman, R.E., 378, 384 399-401
Benedetti, P.L., 164, 171, 172,175 Chilton, C.H., 430, 435
Benelmir, R, 397, 401 Chin, V.Y., 303, 312
Benko, N., 296 Chua, H.T., 395, 401
Bercovier, M., 304, 306, 312 Ciborowski, J., 363, 383
Beretta, G.P., 2,17,19,21,27-33,39, Clausius, R., 19,43
41,43,44 Collado, F.J., 325, 334
Consoli, F., 209, 219
453
454

Conway, J.B., 23,43 Ferry, J.D., 300, 312

Cornelissen, RL., 195, 207, 209, 213, Feynman, R, 1, 17
219, 225, 230 Findeisen, W., 377, 378, 384
Costea, M., 403, 404, 410 Fine, P.C., 413, 420
Creyts, J.C., 89, 94 Fiterman, A, 330, 334
c;ubuk~u, E., 41, 44, 309,312 Flater, G., 362, 364
Curzon, F.L., 48, 50, 58, 60, 74, 94, Fleming, J.S., 430, 435
346,363,383,411,420,423,424 Fletcher, E., 324, 334
Flick, J.D., 415, 420
D' Aponte, F., 395, 401 Foldes, P., 296
Dahmen, RD., 23, 24, 26,43 Fonyo, Z., 292, 295, 296
Dan, N., 71, 72 Forbus, K.D., 439, 444
Davis, G.W., 395, 401, 403, 410 Fraidenraich, N., 329, 332, 334
Davis, RL., 175 Frangopoulos, C.A., 109, 115
De Groot, S.R, 281, 296 Fratzscher, W., 291, 296
De Swaan, Arons, J., 196,207 Freakley, P.K., 312
De Vos, A, 45, 59, 346, 351, 353, 356,
357,359,361-363,374,383,403, Gaggioli, RA., 155,221,225,230
410 Glansdorff, P., 281, 296
DeWitt, D.P., 410 Gogiis, Y., 231, 238, 240, 246, 250
Denton, J.D., 170, 175 Gong, M., 155, 158
Dincer, I., 89, 94, 221, 222, 225, 230 Gordon, J.M., 329, 332, 334, 363,
Diny, M., 397, 399,401 383,385,395,400,401,403,410,
Dirda, V.I., 298, 308, 310, 312 411, 415, 419, 420
Diver, R, 409, 410 Goth, Y., 385, 389, 400
Dorney, D.J., 175 Gouy, G., 137, 150
Doron, P., 330, 334 Grassmann, P., 48, 59, 139, 150
Drost, M.K., 175 Grazzini, G., 385, 400
Durand, M., 304, 306, 312 Grifioen, A, 195, 207
Grosjean, c., 359, 363
Eckert, E., 276, 278 Grosu, L., 401
Edgerton,R.H., 223, 225, 230 Guggenheim, E.A, 1, 17
Ellenbrock, H.H., Jr., 173, 175 Gukhman, AA, 276, 278
Elsayed, M., 325, 334 Gwozdz, J., 155
Elsayed, Y., 155 Gyftopoulos, E.P., 2, 17, 19,21,26-33,
EI-Wakil, M.M., 48, 60 35,37-41,43,44,49,50,58,60,74
Engelman, M.S., 172, 175 94,309,312,423,424
Epstein, M., 323, 325, 327-329, 334
Erbay, L., 390, 393, 395,401 Hafele, W., 223, 225, 230
Eringen, c., 302, 312 Harman, C.M., 446, 447, 452
Errera, M.R, 53, 54, 60, 71, 72 Hartnett, J.P., 276, 278
Evans, R, 141, 150, 155 Hatsopoulos, G.N., 1, 7, 17, 19,26,28,
30,35,37-41,43,44
Fan, L.T., 377, 384 Haught, A., 172, 175
Fathalah, K.A., 325, 334 Hedges, G.H., 249, 250
Fauvel, O.R., 403, 410 Hedman, B., 195, 207
Feidt, M., 45, 48, 57, 59, 385, 386, 388, Heijungs, R, 209, 219
389,397-401,403,410 Hepburn, C., 299, 312
Feldmann, T., 414, 420 Hermann, R, 412, 420
Feng, X., 96, 100 Hershgal, D., 425, 435
 455

Heyen, G., 129, 136 Komlos, J., 415, 420

Hinderink, AP., 196,207 Kopp, J., 48, 59
Hirs, G.G., 195, 207 Korsch, H.J. 28,44
Hoffmann, K.H., 411, 419 Kotas, T.J., 80, 82,94. 198,207,209,
Hofler, T., 414, 420 219, 225, 230, 244, 250
Holm, F., 425, 435 Krause, A, 108, 109, 115
Holzwarth, A, 170, 171, 175 Kreith, F., 56, 57, 60
Horlock, J.H., 329, 334 Kribus, A., 330, 333, 334
Horuz, I., 426, 427, 430, 435 Krush, 1.1., 298, 308, 312
Hiibler, A, 445, 452 Kuhn, T., 28, 44
Kutateladze, S.S., 276, 278
Iantovski, E., 149, 150 Kuo, B.C., 425, 435
Incropera, F.P., 410 Kutz, M., 308, 312
Ihrig, E., 295, 296 Kvaalen, E., 210, 219
Ikegami, Y., 51-53, 60, 65, 71, 72,
335, 336, 338, 339, 342-344 Landau, L.D., I, 17
lIeri, A., 231, 238, 240, 250 Landries, J., 353, 363
Iordache, R, 446, 452 Landsberg, P.T., 316, 320, 324, 353,
363
Jackson, E., 356, 363 Langer, P., 155, 158, 159
Jaeger, J.C., 238, 240 Lazzaretto, A, 96, 100, 107, 115
Jancel, R, 34, 44 Leblanc, F., 304, 306, 312
Jankovich, E., 304, 306, 312 Ledezma, G.A., 65, 71, 72
Jauch, J.M. 38, 43, 44 Lee, F.L., 187, 194
Jeter, J., 316, 324 Leighton, RB., 23, 43
Jones, T.V., 175 Leondes, C.T., 415, 420
Leontiev, A.I., 276, 278
Kalf, M.A, 209, 219 Lie, A.B.K., 196,207
Kalitventzeff, B., 129, 136 Lifshitz, E.M., 1, 17
Kamlos, J., 415, 420 Lightfoot, E.N., 75, 94
Karni, J., 330, 333, 334 Lin, L., 108, 109, 115
Karnopp, N.C., 243, 247, 248, 250 Lindblad, G., 28, 44
Katz, A, 41, 44 Linnhoff, B., 108, 115,279,292,296
Kays, W.M., 212, 219 Lipps, F.W., 325, 334
Kearney, D.W., 56, 57, 60 London, A.L., 212, 219
Keenan, J.H., 1,7, 17,30,44 Losch, F., 454
Kedem, 0., 282, 296 Lozano, M.A., 96, 100, 109, 115, 118-
Keller, J.U., 445, 452 124, 135, 136, 141, 150, 155
Kerkhof, P.J.M., 196,207 Lucca, G., 193, 194
Kesavan, K.H., 249, 250 Lu, P.c., 401
Kestin, J., 226, 230
Kikuchi, S., 335, 344 Madarashan, T., 303, 312
Kilic, A, 187, 194 Mallinson, J.R., 316, 324
Kindler, H., 214, 219 Majza, E., 146, 150
King, C.J., 283, 296 Marchuk, G.I., 155
Kistler, B.L., 326, 327, 334 Marechal, E., 129, 136
Klein, S.A, 385, 400 Margenau, H., 36-39, 44
Kline, S.J., 19, 43 Margolis, D.L., 243, 250
Knoche, K.F., 97, 100 Mark, J.E., 303, 312
Koenig, H.E., 249, 250 Markarytchev, S.V., 89,94
456

Markovich, S.J., 224, 230 Paoletti, S., 175

Martynovskii, V.S., 48, 59 Paolocci, S., 155
Maxwell, J.C., 34, 44 Park, J.L., 19, 28, 34, 36-39, 43, 44
Mayhew, Y.R., 212, 219 Parise, J.AR, 425, 430, 435
Mazur, P., 281, 296 Pauwels, H., 353, 359, 361, 363
McClintock, F.A., 48, 59 Payne, A.R., 312
McIrvine, E.C., 225, 230 Perry, RH., 430, 435
McQuarrie, D.A, 34, 44 Petela, R, 316, 324
Menon, V.J., 385,400 Petrescu, S., 65, 72, 385, 400, 446,
Mereu, S., 175 447,452
Metrot, A, 389,401 Philippi, I., 385, 386, 388, 400
Migliori, A, 414, 420 Pinkerton, RC., 282, 296
Minorsky, N., 445, 452 Pisa, J., 107, 108, 109, 115, 141, 150
Mizsey, P., 295, 296 Popa, B., 303, 312
Moen, RL., 324, 334 Popescu, G., 403, 410
Moore, J., 175 Poturaev, V.N., 298, 308, 310, 312
Moore, J.G., 175 Pratt, H.R.C., 284, 296
Moran, M.J., 45, 46, 57-60, 63-65, 71, Press, W.H., 316, 324
73, 74, 77, 80-83, 85, 93, 94, 96, 97, Presz, K., 155
100, 101, 103, 104, 107, 108, 110, Prigogine, I., 281, 296
114, 115,225,230,231,240
Morris, D.R, 81, 94, 142, 150, 155, Qi, A, 414, 420
195, 196,207,209,219
Mozes, E., 195, 207 Radcenco, V., 45,59, 393,403,410
Mozurkewich, M., 411, 419 Ranasinghe, J., 210, 219
Munoz, M., 120, 135, 155 Rant, Z., 137, 150
Miiser, H., 48, 59, 348, 363 Reader, G., 403, 410
Reini, M., 121-123, 135, 136
Nakaoka, T., 340, 341, 344 Reistad, G.M., 210, 219, 221, 225, 226
Natalini, G., 164, 166, 167, 169, 170, Rev, E., 295, 296
174, 175 Reynolds, RJ.W., 299, 312
Naunton, W.J.S., 299, 312 Richter, H., 97, 100
Nelson, RA, Jr., 71, 72 Riekert, L., 137, 150
Ng, K.C., 395,401,411,420 Ries, H., 333, 334, 353, 363
Nikles, A., 214, 219 Rispoli, F., 175
Nishida, T., 335, 344 Rogers, G.F.C., 212, 219
Nitzan, A, 416, 420, 422, 424 Rosen, M.A, 89, 94, 221, 223, 225,
Novikov, 1.1., 48, 50, 57, 60, 346, 363, 227, 230
383,411,420,423,424 Rosenberg, RC., 243, 250
Nulton, J., 295, 296, 414, 415, 418, Rowell, D., 243, 250
420,423,424 Rubin, M.H., 413, 420
Rubin, R., 330, 334
Obert, E.F., 19,43, 187, 194 Russell, C.M.B., 175
Odum, H.T., 282, 296
Ogden, J.T., 304, 312 Sagie, D., 330, 334
Oni, A.A, 155 Salamon, P., 45, 59, 295, 296, 412-416,
Organ, J.A., 403, 410 418,420-424
Orlov, V., 363, 383,421,424 Sa1vadori, M.G., 446, 452
Ozekici, S., 308, 312 Sarna, D.A., 87, 94, 96, 100, 108, 115,
Oztiirk, A, 187, 188, 192, 194 139, 140, 150
 457

Sato, T., 175 Swift, G.W., 414, 420

Schroedinger, E., 34,44 Szargut, J., 81,94,96, 100, 137, 140-
Schubert, S., 411, 419 142, 144, 146-150, 155, 158, 195,
Schwarcz, P., 122, 123, 136 196,207, 209, 219, 221,225,229,
Schwarz, RL., 446, 452 231,240
Sciubba, E., 65, 72, 91, 94, 151 152, Szolcsanyi, P., 281, 296
154, 155, 158, 164, 166, 167, 169- Szwast, Z., 377, 383, 384
175, 225, 230 Szymanowski, J., 377, 378, 384
Sears, F.W., 187, 194
Segal, A, 323, 325-329, 334 Takeishi, K., 175
Sekulic, D.P., 58, 60, 74, 94 Tawfik, T., 108, 109, 115
Serra, L., 109, 115, 120, 121, 135, 136, Tew, RC., 403, 410
141, 150 Thompson, D' A, W., 70, 72
Shankar, R., 21, 43 Tiba, c., 329, 332, 334
Shapiro, AH., 1, 17 Timoshenko, S., 2, 17,27,44
Shapiro, H.N., 77, 80, 83,94 Tisza, L., 1, 17,412,420
Shewchuk, C.F., 129, 136 Tokad, Y., 249, 250
Shiner, I.S., 57, 60 Tolle, H., 415, 420
Shmidt, K.L., 278 Tolman, RC., 40, 44, 413, 420
Shnaid, T.N., 48, 59 Tondeur, D., 71, 72, 210, 219
Sieniutycz, S., 45, 57, 59, 60, 363, Torres, C., 109, 115, 120, 135, 141, 150
377, 379, 383, 384,421,424 Tribus, M., 137, 150,225,230
Simmons, R.F., 28, 38, 44 Truesdell, c., 299, 312
Simo, I.C., 307, 312 Tsirlin, A, 422, 424
Siragusa, G., 418, 420 Tsatsaronis, G., 45, 46, 59, 63-65, 71,
Sizmann, RL., 326, 334 73, 74, 82, 83, 85, 93, 94, 96- 98,
Slater, J., 23, 43 100, 101, 103, 104, 107-110, 114,
Smith, J.L., Jr., 48, 59 115, 141, 150,231,240,
Smith, R, 279, 292, 296 Tsuboi, K., 335, 344
Sokolov, M., 425, 427, 429, 431, 435 Tsukagoshi, K., 175
Sonntag, RE., 187, 194 Turegano, I.A, 325, 334
Spalding, D.B., 363, 383
Spanner, D.C., 316, 324 Uehara, H., 335-344
Spindler, K., 97, 100
Spirkl, W., 353, 363 Valero, A, 96, 100, 109, 115, 118-124,
Stanek, W., 149, 150 135, 136, 141, 150, 155
Stanescu., G., 446, 452 Van der Kooi, HJ., 196, 207
Steffen, H., 28, 44 van der WeI, P., 362, 364
Stephenson, G.R, 129, 136 Van Wylen, G.J., 187, 194
Steward, F.R, 81,94, 155, 195, 196, Vant-Hull, L.L., 325, 326, 334
207, 209, 219, 231, 240 Vargas, J.V.C., 425, 427, 429-431, 435
Stewart, W.E., 75, 94 von Neumann, I., 22, 34, 38, 43, 44
Stine, W., 409, 410 von Spakovsky, M.R, 109, 115, 122,
Stodola, A, 137, 150 123, 136
Stojanoff, C.G., 103, 104, 115 Voskresenskii, K.D., 48, 60
Stoppato, A, 96, 100 Vukalovich, M.P., 48, 60
Stroh, T., 26, 43 Vulis, L.A, 278
Strong, M.F., 223, 230 Vyncke, D., 359, 364
Sullivan, TJ., 170, 175
Sun, F., 363, 383, 385, 400 Walker, G., 403, 410
458

Wall, G., 155, 158 Wong, F., 249, 250

Wang, C.S., 377, 384 Wood, L.A, 300, 301, 312
Warchol, K., 223, 230 Wormley, D.N., 243, 250
Wargitsch, C., 445, 452 Worsfe-Schmidt, P., 425, 435
Wark, K., 187, 194 Wu, C., 335, 344, 363, 383, 385, 395,
Watowitch, S., 411, 419 400,401,403,410,439,444
Wehr!, A, 19, 28, 43 Wu, J., 363, 383
Weinhold, F., 414, 420 Wilrfel, P., 353, 363
Weise, W., 276, 278
Welford, W.T., 328, 329, 334 Yan, Z., 385, 397, 399, 400,403, 410
Wepfer, W.J., 221, 225, 230 Yavuz, H., 390, 393, 395,401
Wheatley, J., 414, 420 Young, D.H., 2, 17,27,44
White, M.D., 175 Yuste, B.S., 445, 448, 450, 452
Wick, G.c., 37, 44
Wierzbicki, A, 377, 378, 384 Zarmi, Y., 411, 419
Wightman, AS., 37,44 Zheng, J.P., 96, 100
Wigner, E.P., 37, 44 Zhu, X.x., 96, 100
Winhold, M., 103, 104, 108, 115 Zhu, Y.Y., 312
Winter, C.J., 326, 334 Ziebik, J., 155
Winston, R., 328, 329, 334 Zienkiewich, O.C., 312
Subject Index

Absorption, 284 Bose statistics, 355

heat trans fonner, 246 Bounds, 48, 313, 317, 318, 381, 412,
refrigerator, 427 415, 421
Additive property, 5, 29 Breathing, 54-56
Aesthetics, 222 Brinkman number, 173
Affinity, 248 Bulk flow, 33
Air conditioning, 385
Air preheater, 111 CAD; see Design, computer-aided
Air separation, 195-208 Canonical equations, 380
Aircraft engine, 249 Capital recovery factor, 108
Algorithms, 363 Causal structure, 120
Allometric laws, 54-56 Chance, 70
Animals, 45 Chambadal-Novikov-Curzon-Ahlborn
Animate systems, 45-60 engine model (CNCA), 50, 317, 346,
Area-to-point flow, 65-72 363,411,423
Argon purifying unit, 206 Chaos creation, 225, 226
Auxiliary costing equations, 104 Chemical engines, 351, 352
Availability, 191, 192 Chemical equilibrium, 78
adiabatic, 8, 30 Chemical potential, 13,247
finite-time, 413 Circulatory system, 70
generalized adiabatic, 8, 30 Clausius inequality, 187, 188, 191, 194
CNCA model; see Chambadal-Novikov-
Bejan number, 65, 168, 170 Curzon-Ahlborn engine model
Benard convection, 71 Coal, 253
Bifurcation, 69 Coefficients of net consumption, 145
Blackbody radiation, 314, 347 Cogeneration systems, 97, 110, 213,
selective, 350 261-270
Blade, rotor, 169-273 Collector, solar, 56, 57
Blast furnace process, 195 Combustion, 88, 111, 427
Blood flow, 70 Complexity increase, 71
Body size effect, 56 Compressor, 112, 203, 394
Boiler, 253, 266 Concentration ratio, 324
Boltzmann statistics, 355 Condenser, 228, 431
Bond graphs, 241-250 Constitutive equation, 302
storage, 241 Constituent, 3
dissipator,241 Constraints, 48, 51, 54, 55, 159, 164,
causal stroke, 242 227, 247, 340, 388, 398, 412, 416,
coefficient of structural bonds, 244 429
coefficient of external bonds, 245 Constructal law, 71
structural-functional bond coefficient, Constructal theory, 45, 57, 59,61,65-72
249 Consumption sites, 152-154
459
460

Control variable, 375 Uehara, 335

Control volume, 77, 79 vapor compression, 393
Convection, 61-65
Conversion efficiency , solar, 313 Data processing utility, 126
Cornfield array, 325 Dead state, 78, 196
Correspondence postulate, 37 restricted, 79
Cost, 74, 118-124, 151-162 Decision variables, 101
analysis, exergetic, 118-124 Degree of thermodynamic perfection, 140;
associated with exergy destruction, see also Entropy generation number,
107 and Exergetic efficiency
balance. 104 Dendrites; see Constructal theory
busbar electricity, 337 Density operator, 39
capital investment, 104, 107, 108 Depletion of unrestorable natural
direct, 110 resources, 147
distance, 158 Derating, 254
distribution, 157 Design
ecological, 147 computer-aided, 151-186,437-444
friction, 158 constructal, 61-72
indirect, 110 cost-optimal, 93
functions, 157 energy-efficient process, 294
levelized, 110 for the environment, 73-94
operating and maintenance, 104 optimization, 10 1
per exergy unit, 106 OTEC, 336
per unit of fuel exergy, 107 performance analysis, 256-258
per unit of product exergy, 107 thermal, 73-94
periodic change, 235 thermodynamic; see Thermodynamic
product, 102 optimization
purchased-equipment, 102 Dichotomy, 69
relative difference, 107 Diffusion, 33
specific cost of heat, 142 Distillation, 196-208, 279-296
transportation, 157 Dissipative systems, 297
Counterflow heat exchanger, 54, 404 Dynamic programming, 377, 378
Coupling, 282 Dynamics, 4, 40
Cracks, 71
Creep, 300 Economic analysis, 110, 140-142
Criteria of merit, 416 Eddies, 70
Cryogenics, 47, 48, 195-208 Efficiency, 50, 411, 416
Cubication method, 445, 448 empirical definitions, 291, 292
CyclePad, 439 gas turbine, 331
Cycles instantaneous, 233
Brayton, 329, 385,440-443 isentropic, 110, 330, 394
Carnot, 188,314,345-364 overall, 324
combined, 332, 385 solar receiver, 328
Ericsson, 385 thermodynamic, 279, 287-292
ideal,271 EGM method; see Entropy generation
Kalina, 335 minimization
limit, 445 Elastomers, 301
Rankine, 329, 335 ELCA method; see Exergetic life-cycle
real, 276, 393 analysis
Stirling, 385, 403 Electrical motor, 49
 461

Electroluminescence, 355 temperature variations, 238

Electronic packages, 62-72 Equation of motion, 4, 28
Elemental volume, 67 Equilibrium state, 6, 29, 78
EIliptic integrals, 446 Equipartition, 67
Emissions, 226, 227, 255, 256 Evaporator, 428
End-of-the-pipe approach, 89 Exergetic
Endoreversible; see Thermodynamic opti- cost analysis, 118-124, 131
mization cost theory, 120
questionable modeling, 49, 50, 73-75, efficiency, 52, 83, 87, 94, 96, 104,
98 106, 107, 138, 198-203,287
questionable claims of novelty, 57-59, life-cycle analysis, 209-220
73-75,98 units, 151
End states, 4 Exergoeconomics, 97, 10 1-116
Energy, 5,29 Exergoeconomic factor, 107, 109
available, 10, 187, 189, 191, 193 Exergy, 80, 137, 364
balance,S, 29, 75 analysis, 45-47, 73-94, 93-100, 137-
conservation,S, 29 162, 195-208
generalized available, 10, 30 balance, 79, 138, 139, 199,231
gravitational potential, 76, 191 chemical, 104
internal, 13 costing, 74, 103
kinetic, 76, 191 cummulative consumption, 142
saving techniques, 279-296 economic applications, 140-142
Energy relation, 12 extended, 364
Energy versus entropy graph, 14 input, 244
Energy and the environment, 221-230, mechanical, 104
252 physical, 80
Entropy, 11,31,76, 193 product, 244
Entropy balance, 12, 32 thermal, 104
Entropy change, 76 potential, 80
Entropy generated by irreversibility, 11, waste product, 138,226,227
32, 61, 76, 79, 188, 221, 292, 414, Exergy-aided thermal design, 73-94
417 Exergy and the environment, 221-230
Entropy generation minimization (EGM), Exergy destruction, 79, 93, 95, 96, 106
45,47,48,54,58,61,62,73, 163- Exergy loss, 93, 137
186,210-215,221,425-436 partial, 144
Entropy generation number, N s , 62 local, net, 145
Entropy transfer, 76 Exergy reference environment, 78
Environment, 3, 28, 46, 78, 80, 138, Exergy transfer, 79
221-231 Exoirreversible, 45; see Thermodynamic
Environmental optimization
cost, 158 Experimental validation, 430, 431
impact, 89, 147, 221-230 Extended surfaces, 62, 173
problems, 222, 223 Extraction, 285
Environmental conditions
change of, 231 Failure mechanism, 298
correction factor, 236 Free piston thermal machines, 445
finite time effect, 236 Field,243
general thermal module, 238 Fins, 62, 173
phase difference, 233 Finite elements method, 301-308
phase shift, 234 Finite-resources, 45; see Constraints
462

Finite-size, 45; see Constraints Heat engine, 45

Finite-time; see Thermodynamic opti- Heat exchangers, 48, 54, 62, 65, 171-
mization 173,204,210-215,228,404
questionable modeling, 49, 50, 73-75, Heat-exchanger network, 97
98 Heat exchanger passage, 61
questionable claims of novelty, 57-59, Heat integration, 295
73-75,98 Heat interaction, 16
First construct, 68 Heat leak, loss, 47, 57, 286
First law efficiency, 286 Heat pumping, 295, 385, 387
First law of thermodynamics, 5, 29, 46, Heat recovery, 112
75, 188, 191, 194 steam generator, 251
Flow reference direction, 241 Heat to power ratio, 269
Flow work, 76 Heat transfer, 46, 47, 61-72, 300, 313-
Fluid mechanics, 46, 47,61-65 324,340,363,408
Forced convection, 61-65 Heliostats, optimum field, 327, 333
Forces Highest entropy physics, 15
external, 3, 27 Highest entropy principle, 12
internal, 3 History, 45-60
internal intermolecular, 27 Homogeneous ensemble, 36, 41
Fractal geometry, 66 Horse-carrot theorem, 423
Fuel, 84, 104, 105, 253 Hot windbox, 251
analysis, 122, 123, 131 Human labor, 142
cells, 261-270 Hysteresis, 300
flexibility, 255
Fundamental relation, 12,33 Inanimate systems, 45-60
Independent properties, 3, 28
Gas, 253 Influence coefficients, 241-250
Gas dynamics cycles, 271-278 Insulation, 47, 48
Gas engines, 261-270 Interactions, 4, 15
Gas turbines, 112, 166-173, 251, 258, Interboilers, 295
261-270, 324, 329-331 Intercoolers, 295
Geographical allocation, 71, 151-162 Internal spacings, 62-65
Geometry, 45 Irreversible, 31
Geometric optimization; see Constructal Irreversibility, 188, 244
theory associated with the use of material,
Geometric view factors, 320 213
Geothermal power, 52, 239 combustion, 88
Gibbs' law, 351 external, 74
Glansdorff-Prigogine theorem, 281 fluid flow, 61,165, 173, 179
Gouy-Stodola theorem, 46, 79, 93, 137, heat transfer, 61,165, 173, 180
231 internal, 82
Growth, as mechanism of resistance life cycle, 217
minimization, 68 local, 165
Growth, urban, 70, 71 manufacture, 214
Gyrator, 242 minimization; see EGM
operation, 213
Hamiltonian, 377, 380
Heart beating, 54-56 Jeter's efficiency, 316
Heat, 16,33,76 Junctions, 243
Heat degradation, 286
 463

Kelvin-Planck statement, 190 Nernst principle, 15

Kinematics, 3, 36, 70 Nervous system, 70
Kinetic exergy, 80 Neural dendrites, 70
New inequality, 187, 189, 191
Lagrange multiplier, 48, 387, 406 Nonequilibrium state, 6, 29
Lagrangian, 372, 374 Nozzle, 166-173
Landsberg-Petela-Press efficiency, 316,
323 Observable, 37
Legendre transformation, 248 Ocean thermal energy conversion, 52,
Life-cycle analysis, 209-220 335-344
Life-cycle design process, 90 On-line diagnosis, power plants, 117-136
Lightning, 70, 71 Optimal control theory, 415, 421, 425
Limited dynamical postulate, 40 Optimization; see Thermodynamic
Limits; see Bounds optimization, and Entropy
Linear graph, 241, 246 generation minimization
Liquefaction unit, 205 Orderdestruction,225,226
Load Orlov-Berry theorem, 422
back-end,254 Oscillators, 445
refrigeration, 47 OTEC; see Ocean thermal energy
Lost power, 46, 47 conversion
Lost work, 188 Ouroboros, 223, 224
Lowest energy physics, 15
Lungs, 54-56, 70 Pairing, 69
Parameters, 27
Malfunction, 122 Path, 414, 415
Mass balance, 6, 75 Pendullum, thermodynamic, 445-454
Mass transfer, 55, 363 Performance test codes, 129
Mathematical structure, 423 Perpetual motion machine of the second
Maximum power, 49, 51, 411 kind, 7, 30
Mean-value postulate, 38 Perturbations, 445
Mean-value theorem, 38 Photothermal converters, 359
Measurement act, 38 Physics, 59
Measurement result theorem, 39 Physiological processes, 54-56, 70
Mechanical equilibrium, 78 Pinch analysis, 97
Mechanical support, 47, 48 Planck's law, 350
Metabolic rate, 56 Plants, 45
Metric, 414, 418 Polarized radiation, 320
Mirror, hyperboloidal, 324 Pollutants, 222
Modeling, 45, 49 Porous media, 71
Monetary index of harm, 148 Power demand, changes in, 235
Mutual stable equilibrium, 9 Power block, 333
Miiser model, 48, 348 Power plants, 117-136,249,251-278,
403-410, 445
Ns; see Entropy generation number Pressure, 13, 33
National gross product, 144 Pressure drop number, 65
Natural convection, 62-65 Primitive problem, 90
Natural flow structures, 45 Principle
Natural frequencies, 54-56 avoided costs, 141
Natural phenomenon, 45 avoided expenditures, 141
Necessity, 70 complexity increase, 71
464

conservation of energy, 75 Reversible, 31, 46, 188

constructal,71 Reynolds number, 62
equipartition of resistance, 67 River basins, 70, 71
maximum, 377, 380 Romania, 159, 160
minimum entropy production, 281 Rubber-like material, 299
nondecrease of entropy, 11,32 Rubber mount, 297
resistance minimization, 71
stable equilibrium state, 6, 30 Second construct, 69
Probability theorem, 38 Second law analysis, 73, 297
Process, 4 Second law efficiency; see Exergetic
adiabatic, 31 efficiency
coupled, 281, 287 Second law of thermodynamics, 7, 30,
dissipative, 297-312 46, 76, 187, 191, 223
irreversible, 31 Separations, 279, 285
multistage, 363 Sequence method, 143
on-off, 54-56, 425-436 Set of balance equations, 143
reversible, 31 Solar cells, 353-359
separation, 279, 285 Solar power, 48,56-57,239,313-334,
uncoupled, 286 345-364
Process development, 98-100 Solar refrigeration, 385, 425-436
Process industry, 195-208,295 Solvent removal, 285
Product, 84, 104, 105, 141 Spacings, 62-65, 173
Production center, 152-154 Spanner's efficiency, 316, 323
Productive structure, 120 Stable eqUilibrium state principle, 12
Property, 3, 27, 300 Stable equilibrium state, 6, 30
Pump, 52, 53, 228, 339 Stagnation temperature, 57
Standard chemical exergy, 80
Radial stagger array, 325, 326 Standard environment, 80
Radiation, thermal, 313-324, 347 State, 3,27
Rational efficiency; see Exergetic State of reference model, 130
efficiency State principle, 12
Rayleigh number, 64 State space equations, 243
Receiver, solar, 323, 324, 329, 333 Steam bleeds, 253
Reference state, 196 Steam generator, 112, 227, 228
Refrigerating machines, 47, 48,271-278, Steam turbine, 329
385-402, 425-436 Steel making, 195
Reliability model, 308 State space equation, 243
Remaining life, 255 Steady state, 6, 29, 77
Repowering options, 251-260 Steady systems, first order, 280
Reservoir, 9, 31 Stefan-Boltzmann law, 347, 350
temperature, 46, 49, 52 Stirling engine, 239
pressure, 49 Storage,
voltage, 49 thermal, 237,407
Resistance electrochemical,247
electrical, 50 Stress relaxation, 300
fluid flow, 49,55 Street patterns, 70
thermal, 50, 52 Stripping, 284
Resistance minimization, 61-71 Subsystems, 245
Resource degradation, 226 Sustainable development, 221, 222
Resonance, 445 System, 3, 27
 465

System parameters, 245 Tsirlin theorem, 422

Systematization, 289 Turbomachinery, 49, 112, 163-186,203,
228, 258, 338
Temperature, 13, 33 Turbulent flows, 70, 71
Thermal equilibrium, 78 Two-ports, 242
Thermal machines, 271-278
Thermal systems design, 73-94 Unambiguous ensemble, 36
Thermochemical engines, 352, 353 Unstable equilibrium state, 6, 30
Thermodynamic design; see Unsteady state, 6, 29
Thermodynamic optimization Urban growth, 70, 71
Thermodynamic inefficiency; see Exergy
destruction Value of the property, 3, 28
Thermodynamic length, 414, 417 Variable
Thermodynamic optimization, 45-72, 93, co-energy, 244
97,98, 163-186,210-215,221,297- flow, 242
312, 317, 323-344, 346-353, 363- effort, 242
402,411- 444 energy, 244
Thermodynamic potential, 412, 416 independent, 244
Thermodynamic speed, 414, 417, 418 input, 244
Thermodynamics and the environment, Variational calculus, 48, 54
223, 224 Virtual laboratory, 437
Thermoeconomics, 45, 73, 93, 117-136, Volume-to-point flow, 65-72
140
Thermomechanical systems, 297-312 Waste, 138,226,227
Third law of thermodynamics, 15 Work, 16,33,76
Total potential, 33 Work interaction, 16
Tower reflector, 323, 328 Work maximization, 363, 378
Tower top, 328 Workable design, 108
Transformer, 228, 242, 246
modulated, 242 Zero-entropy physics, 15
Tree networks; see Constructal theory