Entrooy Generation Minimization

Entropy generation minimization: The new thermodynamics of finitesize
devices and finitetime processes

Adrian Bejan
Citation: J. Appl. Phys. 79, 1191 (1996); doi: 10.1063/1.362674
View online: http://dx.doi.org/10.1063/1.362674
View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v79/i3
Published by the American Institute of Physics.
Related Articles
Existence of the thermodynamic limit for disordered quantum Coulomb systems
J. Math. Phys. 53, 095209 (2012)
Visual and quantitative measurement of the temperature distribution of heat conduction process in glass based
on digital holographic interferometry
J. Appl. Phys. 111, 093111 (2012)
Field induced gradient simulations: A high throughput method for computing chemical potentials in
multicomponent systems
J. Chem. Phys. 136, 134108 (2012)
Optimization on a three-level heat engine working with two noninteracting fermions in a one-dimensional box trap
J. Appl. Phys. 111, 043505 (2012)
Robust interpolation between weak- and strong-correlation regimes of quantum systems
J. Chem. Phys. 136, 044109 (2012)
Additional information on J. Appl. Phys.

Journal Homepage: http://jap.aip.org/
Journal Information: http://jap.aip.org/about/about_the_journal
Top downloads: http://jap.aip.org/features/most_downloaded
Information for Authors: http://jap.aip.org/authors
Downloaded 31 Jul 2012 to 132.236.27.111. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
APPLIED PHYSICS REVIEWS
Entropy generation minimization: The new thermodynamics

of finite-size devices and finite-time processes
Adrian Bejan
J. A. Jones Professor of Mechanical Engineering, Department of Mechanical Engineering and Materials
Science, Duke University, Durham, North Carolina 27708-0300
~Received 12 January 1995; accepted for publication 4 October 1995!

Entropy generation minimization ~finite time thermodynamics, or thermodynamic optimization! is
the method that combines into simple models the most basic concepts of heat transfer, fluid
mechanics, and thermodynamics. These simple models are used in the optimization of real
~irreversible! devices and processes, subject to finite-size and finite-time constraints. The review
traces the development and adoption of the method in several sectors of mainstream thermal
engineering and science: cryogenics, heat transfer, education, storage systems, solar power plants,
nuclear and fossil power plants, and refrigerators. Emphasis is placed on the fundamental and
technological importance of the optimization method and its results, the pedagogical merits of the
method, and the chronological development of the field. 1996 American Institute of Physics.
@S0021-8979~96!04103-9#
TABLE OF CONTENTS
I. Objectives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
II. The method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
III. Cryogenics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
IV. Heat transfer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
V. Storage systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
VI. Solar power plants. . . . . . . . . . . . . . . . . . . . . . . . . .
VII. Nuclear and fossil power plants. . . . . . . . . . . . . . .
VIII. Refrigeration plants. . . . . . . . . . . . . . . . . . . . . . . .
IX. Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1191
1192
1193
1197
1201
1203
1206
1212
1214
I. OBJECTIVES
Entropy generation minimization ~EGM! is the method

of modeling and optimization of real devices that owe their
thermodynamic imperfection to heat transfer, mass transfer,
and fluid flow irreversibilities. It is also known as thermodynamic optimization in engineering, where it was first developed, or more recently as finite time thermodynamics
in the physics literature. The method combines from the start
the most basic principles of thermodynamics, heat transfer,
and fluid mechanics, and covers the interdisciplinary domain
pictured in Fig. 1. The most exciting and promising interdisciplinary aspect of the method is that it also combines research interests from engineering and physics.
The objectives of the optimization work may differ from
one application to the next, for example, minimization of
entropy generation in heat exchangers,1 maximization of
power output in power plants,2 8 maximization of an ecological benefit,9 and minimization of cost.10 Common in
these applications is the use of models that feature rate processes ~heat transfer, mass transfer, fluid flow!, the finite
sizes of actual devices, and the finite times or finite speeds of
real processes. The optimization is then carried out subject to
J. Appl. Phys. 79 (3), 1 February 1996
physical ~palpable, visible! constraints that are in fact responsible for the irreversible operation of the device. The
combined heat transfer and thermodynamics model visualizes for the analyst the irreversible nature of the device.
From an educational standpoint, the optimization of such a
model gives us a feel for the otherwise abstract concept of
entropy generation, specifically where and how much of it is
being generated, how it flows, and how it impacts thermodynamic performance.
The emergence of a new field of research is marked by
the appearance of several fundamental results that hold for
entire classes of known and future applications. Although
isolated publications had appeared throughout the 1950s and
1960s, thermodynamic optimization emerged as a selfstanding method and field in the 1970s in engineering, with
applications notably in cryogenics, heat transfer engineering,
solar energy conversion, and education. These first developments were reviewed in Ref. 1, and Refs. 1113.
The field has experienced tremendous growth during the
1980s and 1990s. The objective of this article is to review the
field, and to place its growth in perspective. The explosion of
interest that we are witnessing today is due to three new
developments: the diversification of the problems tackled in
engineering after the energy policies of the 1970s, the lifting
of the Iron Curtain and the absorption of work done by previously unrecognized pioneers, and the contributions that appear in the physics literature. The field today is so large and
active that the author is forced to focus this review on just
three aspects, which were stressed by the editors in their
original invitation:
~1! The fundamental and technological implications of the
results obtained until now.
~2! The pedagogical merits of the method, i.e., how EGM
0021-8979/96/79(3)/1191/28/$6.00
1996 American Institute of Physics
1191
this array of interactions is the heat transfer rate between the

, on
system and the atmospheric temperature reservoir, Q
0
which we focus shortly.
The thermodynamics of the system of Fig. 2 consists of
accounting for the first law and the second law ~e.g., Ref.
14!,
n
dE
2W
1
5
Q
dt i50 i
S gen5
FIG. 1. The interdisciplinary field covered by the method of entropy generation minimization ~from Ref. 1!.
can make thermodynamics and heat transfer easier to

teach, understand, and apply in actual technical work.
~3! The chronological development of the field, based on a
combined survey of the engineering and physics literature.
II. THE METHOD
It is instructive to begin the review with a brief look at

why in EGM we need to rely on heat transfer and fluid mechanics, not just thermodynamics. Consider the general
systemenvironment configuration shown in Fig. 2. The system operates in the unsteady state, and its instantaneous inventories of mass, energy, and entropy are M , E, and S. The
, heat transsystem experiences the net work transfer rate W
fer rates (Q 0 ,Q 1 ,...,Q n ) with n11 temperature reservoirs

in , m
out! through any
(T 0 ,T 1 ,...,T n ), and mass flow rates ~m
number of inlet and outlet ports. For simplicity, only one
inlet and one outlet are illustrated in Fig. 2. Noteworthy in
h,
m
(in m h2 (
out
Q
dS
i
2
2
dt i50 T i
s>0,
m
(in m s1 (
out
~1!
~2!
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 S gen
is simply a definition ~notation! for the entire quantity on the
left-hand side of the inequality sign. Soon, we will recognize
that it is advantageous to decrease S gen , and this can be accomplished only by changing at least one of the quantities
~properties, interactions! specified along the system boundary.
as the interaction that is always allowed to
We select Q
0
float as S gen 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
no or little consequence in the overall cost analysis of the
between Eqs. ~1! and ~2! we obtain
design. Eliminating Q
0
n
52
W
1
d
T0
12
Q
~ E2T 0 S ! 1
dt
Ti i
i51
~ h2T 0 s ! 2T 0 S gen .
m
(in m ~ h2T 0 s ! 2 (
out
~3!
The work transfer rate in the limit of reversible operation

~S gen50! is clearly
FIG. 2. General work transfer, heat transfer, and mass flows between a system and its environment in the unsteady state ~from Ref. 14!.
1192
J. Appl. Phys., Vol. 79, No. 3, 1 February 1996
Appl. Phys. Rev.: Adrian Bejan
52
W
rev
1
d
T0
12
Q
~ E2T 0 S ! 1
dt
Ti i
i51
~ h2T 0 s ! .
m
(in m ~ h2T 0 s ! 2 (
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 ~e.g., Ref. 14, Chap. 3!, and the calculation
of W
rev is known as exergy analysis. Subtracting Eq. ~3!
from Eq. ~4! we arrive at the formula
5T S
2W
W
rev
0 gen
~5!
which for most of this century in engineering has been

known as the GouyStodola theorem.1,15,16
Pure thermodynamics ~e.g., exergy analysis! ends, and
2W
! is alEGM begins, with Eq. ~5!. The lost power ~W
rev
ways positive, regardless of whether the system is a power
producer ~e.g., power plant! or a power user ~e.g., refrigeration plant!. Equation ~5! states that the destroyed power is
proportional to the total rate of entropy generation. If engineering systems and their components are to operate such
that their destruction of work is minimized, then the conceptual design of such systems and components must begin with
the minimization of entropy generation ~Ref. 1, pp. 25,33!.
The critical new aspect of the EGM methodthe aspect
that makes the use of thermodynamics insufficient, and distinguishes it from exergy analysisis 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 relate the degree of thermodynamic nonideality of the design to 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 finite-size and
finite-time constraints.
To explain how the following review was structured, it is
pointed out that the optimization of power and refrigeration
systems has a long and established tradition in engineering.
The portion that is based on exergy analysis is voluminous
and has been reviewed on several occasions.14,1723 Energy
analysis is distinct from EGM, and is not the object of the
present review. In fact, to calculate S gen and minimize it, the
analyst does not need to rely on the concept of exergy ~Ref.
1, pp. 25,33!.
The examples that are presented next illustrate how
EGM blends thermodynamics with heat transfer and fluid
mechanics in the optimization of real systems. The minimum
entropy generation design ~S gen,min! is determined for each
model, as in the case of the optimal diameter of a duct with
flow and heat transfer, or the optimal hot-end temperature of
a power plant with a bypass heat leak to the ambient. The
approach of any other design ~S gen! to the limit of realistic
FIG. 3. Mechanical support with variable heat leak and intermediate cooling
effect ~after Ref. 24!.
thermodynamic ideality represented by the design with minimum entropy generation ~S gen,min! is monitored in terms of
the entropy generation number N S 5S gen/S gen,min>1, or alternatives of the same ratio.1
As a final comment on Fig. 1, note that thermodynamics
is a foundation that should be visible ~i.e., present, and
taught and used! in related disciplines such as heat transfer
and thermodynamics. Historically, however, thermodynamics
was formulated after heat transfer, and long after mechanics
~e.g., Ref. 1!. The interdisciplinary domain that is now being
mapped by the research on EGM or finite-time thermodynamics is finally bridging the gap between thermodynamics
and the other thermofluid engineering disciplines. This is
why the developments reviewed in this article not only have
technological relevance but also fundamental and pedagogical value.
III. CRYOGENICS
The field of low temperature refrigeration was the first

where irreversibility minimization became an established
method of optimization and design. As a special application
of Eq. ~5!, 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 T 0 into the cold space. The
structure of a cryogenic system is in fact dominated by components 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 harmony24 with the
rest of the refrigerator of liquefier.
Figure 3 shows a mechanical support of length L that
connects the cold end of the machine ~T L ! to room temperature ~T H !. The rate of entropy generation inside the support
shown as a vertical column is
S gen5
TH
TL
Q
dT,
T2
~6!
1193
is allowed
where it is important to note that the heat leak Q
to vary with the local temperature T. The origin of the integrand in Eq. ~6! is the infinitesimal element ~shaded in
Fig. 3!, in which the rate of entropy generation
/T1dQ
/T2(Q
1dQ
)/(T1dT)5Q
dT/T 2 , beis dS gen5Q
is removed
cause dT!T. The local heat leak decrement dQ
by the rest of the installation, which is modeled as reversible.
The heat leak is also related to the local temperature gradient
and conduction cross-section A,
dT
,
dx
5kA
Q
~7!
where the thermal conductivity k(T) decreases toward low

temperatures. Rearranged and integrated from end to end,
Eq. ~7! places a size constraint on the unknown function
(T),
Q
L
A
TH
TL
k
dT.
~8!
According to variational calculus ~e.g., Ref. 14!, the heat

leak function that minimizes the S gen integral ~6! subject to
the finite-size constraint ~8! is obtained by finding the extreT
mum of the aggregate integral * T H FdT whose integrand F is
L
a linear combination of the integrands of Eqs. ~6! and ~8!,
, and l is a Lagrange multiplier. The Euler
/T 2 1lk/Q
F5Q
50, which yields
equation reduces in this case to ] F/ ] Q
1/2
Q opt5(lk) T. The Lagrange multiplier is finally deter

mined by substituting Q
opt into the finite-size constraint ~8!.
24
The results are
5
Q
opt
S E
SE
A
L
S gen,min5
TH
TL
A
L
k 1/2
dT k 1/2T,
T
TH
TL
k 1/2
dT
T
~9!
~10!
Equation ~6! was provided by thermodynamics and Eq.

~8! by heat transfer: together they prescribe the optimal design @Eqs. ~9! and ~10!#, which is characterized by a certain
/dT) . Any
distribution of intermediate cooling effect (dQ
opt
(T), will generate more entropy and will reother design, Q
quire more power in order to maintain the cold end of the
support at T L . Quantitative examples are given in Refs. 1,
13, and 24. Together, Eqs. ~6! and ~8! illustrate the method of
thermodynamic optimization subject to a physical constraint,
and, to paraphrase some of the more recent physics terminology, they constitute one of the earliest examples of finitesize thermodynamics, this in 1974 in engineering.
The technological applications of the variable heat leak
optimization principle are numerous and important. In the
case of a mechanical support, the optimal design is approximated in practice by placing a stream of cold helium gas in
counterflow ~and in thermal contact! with the conduction
c p , where
path, Fig. 4. The heat leak varies as dQ/dT5m
c p is the capacity flow rate of the stream. The practical
m
value of the theory @Eqs. ~9! and ~10!# is that it guides the
designer to an optimal flow rate for minimum entropy generation. To illustrate, if the support conductivity is tempera1194
FIG. 4. The intermediate cooling effect provided by a cold stream in counterflow with the conduction current through a structural support ~from
Ref. 25!.
ture independent, then the optimal flow rate is

opt5(Ak/Lc p )ln(T H /T L ). In reality the conductivity of
m
cryogenic structural materials varies strongly with the temperature, and the single-stream intermediate cooling technique can approach S gen,min only approximately. The optimal
flow rates to be used in conjunction with common structural
materials are reported in Ref. 25.
The fabrication of the heat exchanger between the support and the coolant is less difficult and more economic if the
continuous contact ~Fig. 4! is replaced by a succession of
discrete cooling stations ~Fig. 5!. The optimal heat leak function required by Eq. ~9! must be approximated by a stepwisevarying function. The problem consists of determining not
only the optimal cooling rate for each station, but also the
optimal spatial position of each station. Hilal and Boom26
solved this problem using the method of Lagrange multipliers, and reported optimal structural designs for several superconducting coil applications of very large scales.
The optimization of conduction heat leak paths was considered in more general circumstances by Bisio.27,28 An even
more practical technique for optimizing the support of a vessel filled with cryogen is to place the natural stream of vapor
~boil off! in contact with the support at a number of cooling
stations ~Fig. 6!. The optimization problem is also simpler
because it reduces to finding only the optimal positions of
FIG. 5. Structural support cooled at several locations ~from Ref. 26!.

number of shields ~N21! is sufficiently large such that the

net radiation heat transfer between two adjacent shields is
approximately1,14
5 s AF ~ T 4 2T 4 ! >4 s AFT 3
Q
i
i11
i
i
DT i
,
Di
~11!
where DT i 5T i11 2T i , Di5(i11)2i and A, F, and s are

the shield area, effective view factor, and StefanBoltzmann
constant. Integrating Eq. ~11! across the stack we obtain a
finite-size constraint that replaces Eq. ~8!,
N
A
FIG. 6. Structural support with discrete stations for helium boil off cooling
~from Ref. 1!.
the cooling stations. The optimal design ~S gen,min! corresponds to the minimum boil off rate, or the minimum cold
end heat leak into the vessel. This problem was solved by
Lagrange multipliers in Ref. 29, which also reports the optimal positions and boil off rates for designs with up to six
intermediate cooling stations and structural materials with
k/T5constant.
The thermodynamic optimization of a stack of radiation
shields in vacuum, Fig. 7, follows essentially the same route
as in Eqs. ~6!~10!. To see this at a glance, assume that the
FIG. 7. Discrete intermediate cooling of a stack of radiation shields ~from

Ref. 14!.
T0
TL
4 s T 3F
dT,
~12!
where T 0 is the same as the room temperature T H used earlier. The result of minimizing the entropy generation rate
integral ~6! is an optimal radiation heat leak variation
(T), i.e., an optimal way of cooling each shield. The
Q
opt
practical techniques of providing this intermediate effect
~single stream, discrete stations, boil off! have also been applied to stacks of radiation shields. The optimization effort
requires computer assistance when the number of shields is
small and the approximation made in Eq. ~11! does not hold.
Martynovskii et al.30 determined the optimal intermediate
cooling ~and optimal shield temperatures! for minimum entropy generation rate in stacks with one, two, and three
shields. Eyssa and Okasha31 optimized stacks of radiation
shields where the space between shields is filled with superinsulation. Chato and Khodadadi32 minimized the entropy
generation rate in stacks with shield-to-shield spaces occupied by a structural material. Further aspects of the thermodynamic optimization of cryogenic insulation and support
techniques are presented in Chen et al.33
Superconducting devices must communicate with the
room temperature environment not only mechanically but
also electrically. For example, the electric cables ~current
leads! that connect the liquid helium temperature windings to
the room temperature network can be modeled as a heat conducting path of length L, cross-section A, and constant thermal conductivity k. In addition, the cable conducts a total
electric current I, against an electric resistivity re . The fact
that the electric power dissipated via Joule heating in the
cable is inversely proportional to A, and that the conductive
heat leak is proportional to A, guarantees the existence of an
optimal cable cross section for which the two sources of
irreversibility add up to a minimum. Furthermore, when the
cable geometry and material are specified, it is possible to
equip the cable with the optimal intermediate cooling effect
so that the refrigerator power needed to keep the cable cold
is minimized.
The optimization method is the same as in Eqs. ~6!~10!.
The optimal intermediate cooling regime for cryogenic
cables was developed by Agsten34 who minimized the total
power ~refrigerator power1electric power! required to operate the cooled cable. When the single stream cooling technique of Fig. 4 is used, the optimal flow rate of cold helium
gas is1,35
S D
2
optc p L
ILL 1/2
TH
m
1
0
5ln
1
,
kA
T L ln~ T H /T L !
kA
~13!
1195
, and U is
where p is the perimeter of the duct that carries m
the overall heat transfer coefficient based on p. The total
stream-to-stream heat transfer area pL is fixed. We account
for this constraint by eliminating DT between Eqs. ~15! and
~16!, and integrating from x50 to x5L,
pL5
TH
TL
cp
m
dT.
U
Q
~17!
The integral constraint ~17! plays the same role as Eq. ~8!,
and the result of the EGM analysis is again an optimal heat
leak variation,
5
Q
opt
FIG. 8. The longitudinal convective heat leak carried by a counterflow heat

exchanger ~from Ref. 43!.
where L 052.4531028 ~W/A K!2 is the constant in the

WiedemannFranz law, L 0 5k r e /T. An additional advantage
to using the flow rate ~13! is that it guarantees the thermal
stability of the cable.36 The optimal cable cross section for
minimum entropy generation rate is1
A opt5
ILL 1/2
0
k ln~ T H /T L !
~14!
The current cable with Joule heating and longitudinal

heat leak is just one of the optimization application problems
at the interface between cryogenics and electrical engineering. Another wide class of applications deals with the cooling of heat generating electric components that must operate
at a certain temperature. Examples are the superconducting
transition,37 superconducting windings for stationary magnets and rotating machines,38 40 and electronic packages in
computers.41,42
Another class of engineering components that has been
optimized based on the irreversibility minimization principles ~6!~10! is the counterflow heat exchangers that connect the coldest regions of refrigerators and liquefiers to the
room temperature compressor.43 The counterflow sketched in
Fig. 8 was intentionally oriented and labeled in the same way
as the mechanical support of Fig. 3 to stress the analogy
between the two devices. The entropy generation rate associated with the two streams and the space between them ~Fig.
is now the longitudinal
8! is given by Eq. ~6! in which Q
convective heat leak
5m
c p DT
Q
~15!
and DT is the transversal ~stream-to-stream! temperature difference. The equivalent of the finite-size constraint ~8! is
obtained by writing that the enthalpy gained by the colder
stream is equal to the local stream-to-stream heat transfer
rate,
c p dT5 ~ pdx ! UDT,
m
1196
~16!
cp
TH
m
c p T,
ln
m
UpL
TL
S gen,min5
cp
TH
m
ln
UpL
TL
~18!
~19!
UA
/dT) .
or an optimal regime of intermediate cooling, (dQ
opt
The way this cooling effect is built into the practical
design of the counterflow heat exchanger is by bleeding a
e ) of the high pressure stream, expanding it in a
fraction (m
work producing device ~cylinder and piston, or turbine!, and
using this cold stream to cool the counterflow heat exchanger, Fig. 9. The optimal flow rate through the expander
/dT) . When
e,optc p 5 (dQ
is known from Eq. ~18! and m
opt
the pressure ratio P H / P L is not large enough for the ex e to become as cold as the cold end of the
panded fraction m
counterflow (T L ), the engineering solution is to install two or
three expanders along the counterflow. The optimization of
the temperature locations of such a sequence of expanders is
described in Ref. 14. Another fundamental development is
the optimal temperature staging of cryogenic refrigerators.44
One interesting characteristic of the counterflow heat exchanger with optimal intermediate cooling effect is that the
stream-to-stream temperature difference DT opt decreases
proportionally with T, Fig. 9,
S D
DT
T
5
opt
cp TH
m
ln
.
UA
TL
~20!
This rule follows from Eqs. ~15! and ~18! and is widely
recognized in the design of cryogenic counterflow heat
exchangers.45 Another interesting aspect is that the convective heat leak ~18! and entropy generation rate ~19! decrease
when the stream-to-stream thermal conductance UpL increases. In other words, by promoting heat transfer in the
transversal direction, counterflow heat exchangers serve as
effective insulations in the longitudinal direction.
The structure of any low temperature installation contains more than one heat leak path. The intermediate cooling
regime for parallel heat leak paths has been optimized46
based on the EGM approach presented in this section. One
conclusion is that parallel heat leak paths must be fitted individually with optimal continuous distributions of intermediate cooling. References such as 24, 30, and 45 are further
examples of early finite size thermodynamics work in engineering.
FIG. 9. The optimal intermediate cooling of a counterflow heat exchanger ~from Ref. 14!.
IV. HEAT TRANSFER
The field of heat transfer engineering adopted the techniques developed in cryogenics and applied them to many
classes of devices for promoting heat transfer. The optimization was carried out at two levels of complexity: complete
components ~e.g., heat exchangers!, and elemental features
~e.g., fins, ducts!. The field is vast, therefore in this section
we review only some of the most basic examples.
Consider first the flow of a single-phase stream m

through a heat exchanger tube of internal diameter D. The
heat transfer rate per unit tube length, between the tube wall
and the stream, q 8, is given. The entropy generation rate per
unit tube length is47,48
3f
32m
q 82
1
S 8gen5
,
p kT 2 Nu p 2 r 2 TD 5
~21!
where the Nusselt number Nu5hD/k is a nondimensional

way of expressing the wallstream heat transfer coefficient
h. Similarly, the friction factor f accounts for the frictional
pressure drop along the tube, f 5(2d P/dx) r D/(2G 2 ),
/( p D 2 /4). The properties r, T, and k are the
where G5m
bulk density, temperature, and thermal conductivity of the
fluid, and x increases in the downstream direction.
The Nusselt number is a result taken from the field of
0.4
when 2500
heat transfer, e.g., Nu50.023 Re0.8
D Pr
6
2
/( rp D /4). The friction
,ReD ,10 , ReD 5VD/ n and V5m
factor is a result taken from fluid mechanics, namely
when 23104,ReD ,106. In this way Eq.
f 50.046 Re20.2
D
~21! shows at a glance the meaning of Fig. 1, i.e., how thermodynamics is combined from the start with heat transfer
and fluid mechanics in the calculation and minimization of
S 8gen . The first term on the right-hand side is the contribution
8
made by heat transfer, S gen,DT
, while the second term is the
contribution due to fluid friction, S 8gen,D P , in other words
8 5S gen,DT
8
S gen
1S 8gen,D P .
~22!
A characteristic of all heat transfer devices with fluid

8
competes against S 8gen,D P : for example, in
flow is that S gen,DT
, the changes in
the smooth tube with fixed q 8 and m
8
and S 8gen,D P have opposing signs as D changes. The
S gen,DT
optimal tube diameter that minimizes the S 8gen expression
~21! is given by48
20.07
ReD,opt52.023B 0.36
,
0 Pr
~23!
where B 0 is a heat and fluid flow duty parameter that

):
accounts for the constraints (q 8 ,m
B 05
q 8m
.
1/2 5/2
~ kT ! m / r
~24!
8
is obThe minimum entropy generation rate S gen,min
tained by combining Eq. ~23! with Eq. ~21!. The perfor8 ) relative to the optimal
mance of any other design (D,S gen
design (D opt ,S 8gen,min) is described by the entropy generation
number N S ,
S 8gen
ReD
50.856
N S5
ReD,opt
S 8gen,min
20.8
10.144
ReD
ReD,opt
4.8
,
~25!
where ReD /ReD,opt5D opt /D because the mass flow rate is

fixed. This ratio is plotted in Fig. 10, which shows that the
rate of entropy generation increases sharply on either side of
the optimum. The irreversibility distribution ratio plotted on
8
. The entropy genthe curve is defined as f 5 S 8gen,D P /S gen,DT
1197
FIG. 10. Entropy generation number N S , or relative entropy generation rate

in a smooth tube with heat transfer ~from Ref. 48!.
8 /S gen,min
8
eration rate ratio S gen
is used to monitor the approach
of any design relative to the best design that can be conceived subject to the same constraints. This performance criterion was used extensively in the engineering literature
~e.g., Refs. 1 and 14! and, just last year, was also recognized
in the physics literature.49
The minimization of entropy generation in ducts with
heat transfer attracted considerable interest from engineers
working on heat transfer augmentation techniques. In such
techniques the main objective is to increase the wallfluid
heat transfer coefficient relative to the coefficient of the unaugmented ~i.e., untouched! surface. A parallel objective,
however, is to register this improvement without causing a
damaging increase in the pumping power demanded by the
forced-convection arrangement. These two objectives reveal
the conflict that accompanies the implementation of any augmentation technique: a design modification that improves the
thermal contact ~e.g., roughening the heat transfer surface! is
likely to also augment the mechanical pumping power requirement.
The true effect of a proposed augmentation technique on
thermodynamic performance can be evaluated by comparing
the entropy generation rate of the heat-exchange apparatus
before and after the implementation of the augmentation
technique. This method of optimizing augmentation techniques was proposed in Ref. 50, where it was applied to the
design of ducts with sand-grain roughness and transversal rib
roughness. Spiral tubes, twisted tape inserts, propeller inserts
and tubes with internal spiral ribs were optimized in Ref. 51.
Nelson52 used this method to evaluate the effect of heat
transfer augmentation on optimized full-size heat exchangers. Sekulic et al.53 documented the thermodynamic optimization of several ducts ~smooth and enhanced!, and showed
that the minimum entropy generation design differs markedly from the design based on conventional methods.
Perez-Blanco54 integrated the entropy generation rate along
the entire surface of the heat exchanger, and then evaluated
1198
FIG. 11. The optimal size of a plate, cylinder, and sphere for minimum
entropy generation ~from Ref. 64!.
the effect of the heat transfer augmentation technique on the

total entropy generation rate. Zimparov and Vulchanov55 applied the EGM method to assess the merits of using spirally
corrugated tubes. Performance evaluation studies were also
conducted by Chen and Huang56 and Prasad and Shen.57,58
The EGM method was recognized in the most recent review
of heat transfer augmentation techniques.59
The opportunity of minimizing the entropy generation to
determine the optimal sizes of ducts and other heat transfer
devices has been noted by practitioners in the field of chemical process engineering.60 The same topic became a sizable
component in the new edition of a classic undergraduate heat
transfer textbook.61
Another large and diverse group of heat transfer devices
relies on external convection, that is, heat transfer between a
stream and a body immersed in the stream. The minimization
of entropy generation for members of this class has been
performed by adapting the internal flow analyses ~21!~25!
to external flow. The start is the formula for the total entropy
generation rate associated with heat transfer and drag on an
immersed body62
S gen5
~ T 2T ! F U
Q
B
B
`
D `
1
,
T BT `
T`
~26!
, T , T , F , and U are the heat transfer rate,

where Q
B
B
`
D
`
body temperature, free stream temperature, drag force, and
is profree stream velocity. The relation for calculating Q
B
vided by the field of heat transfer. Similarly, the fluid flow
information required for evaluating F D comes from fluid mechanics. The S gen expression has the same two-term structure
as Eq. ~22!. The competition between the two terms points to
an optimal body size for minimum entropy generation rate.
The simplest example is the selection of the swept length
L of a plate immersed in a parallel stream ~Fig. 11 inset!. The
optimization procedure for laminar and turbulent flow was
destroys minimum power while performing its assigned heat

and fluid flow duty. The design is improved and, at the same
time, simplified. Even the teaching of the discipline of heat
transfer benefits from this approach.
Elemental features such as ducts and fins are what heat
exchangers are made of. The EGM method was also applied
to heat exchangers as complete, more complex systems.
Consider for example the counterflow heat exchanger seen in
Fig. 8, where the inlet conditions (T H , P H ) and (T L , P L ) are
c p , and
fixed. The two streams have the same capacity rate m
the fluid is the same ideal gas (R,c p ). Each stream flows
with pressure drop through a space ~passage! of length
(L H ,L L ) and hydraulic diameter (D h,H ,D h,L ). The subscripts H and L refer to the high pressure side and, respectively, low pressure side of the heat transfer surface. It has
been shown that the rate of entropy generation of each passage is minimized if the passage slenderness ratio is selected
optimally66
S D
L
Dh
FIG. 12. Optimal pin fin diameter and length for minimum entropy generation ~from Ref. 62!.
given in Ref. 1, and for turbulent flow in Ref. 63. The results
for ReL,opt5U ` L opt / n are shown in Fig. 11 where B is the
duty parameter
B5
/W
Q
B
U ` ~ k m T ` Pr1/3! 1/2
~27!
and W is the plate dimension perpendicular to Fig. 11. The

same figure shows the corresponding results for the optimal
diameter of a cylinder in cross flow,63,64 where ReD,opt
5 U ` D opt / n , and B is given by Eq. ~27!. The optimal diameter of the sphere1,63,64 is referenced to the sphere duty parameter defined by
B s5
Q
B
.
n ~ k m T ` Pr1/3! 1/2
~28!
The fins planted on the surfaces of heat exchanges act as

bodies with heat transfer in external flow. The size of a fin of
given shape can be optimized by minimizing S gen of Eq. ~26!
while accounting for the internal heat transfer characteristics
~longitudinal conduction!65 of the fin. Figure 12 shows one
example of how to select the optimal length and diameter of
a pin fin ~cylindrical spine!, where ReL,opt5U ` L opt / n ,
2 . The figure was
ReD,opt5U ` D opt / n , and B5 rn 3 kT ` /Q
B
1/2 21/6
drawn for M 5(k/l) Pr 5100 where k and l are the fin
and fluid thermal conductivities. Optimal sizes for other fin
shapes are reported in Ref. 62.
The technological benefit of applying the EGM method
to heat transfer devices is that such devices are notorious for
their large number of dimensions, which have to be selected
by the designer. The thermodynamic optimization method
shows how to select certain dimensions such that the device
5
opt
t
,
4G @~ R/c p ! f St # 1/2
*
~29!
S D SD
~30!
R
S gen,min
52 t
cp
m
cp
1/2
f
* St
1/2
where t5u T H 2T L u /(T H T L ) 1/2, G 5G/(2 r P) 1/2, and G is

*
the mass velocity through the passage flow cross-section A f ,
/A f . As a first approximation, the ratio f /St is
namely G5m
a constant for a given type of heat exchanger surface.66 The
subscripts H and L are dropped from Eqs. ~29! and ~30!
because one such set can be written for either passage. This
further illustrates the power of the method: the same optimization rule applies on both sides of the heat transfer surface.
In Eq. ~29! we see just one of the geometric optima that
have been developed. The balanced counterflow heat exchanger subject to fixed heat transfer area was optimized in
Ref. 66, which also reported the optimization subject to fixed
heat exchanger volume. The doubly constrained optimization
of the counterflow heat exchanger with fixed area and volume was reported in Ref. 14. In Refs. 47 and 66 we see early
examples of finite size thermodynamics applications in heat
transfer. The optimization subject to fixed total number of
heat transfer units ~N tu! was developed in a noteworthy study
by Sekulic and Herman.67 Another line of research focused
on the thermodynamic optimization of several types of heat
exchangers where the fluid friction irreversibility is
negligible.68 71 Heat exchangers with phase change ~boilers,
condensers! were optimized by London and Shah,72 Zubair
et al.,73 and Lau et al.74
Extensions to the study of entropy generation in counterflow heat exchangers66 have been performed by Sarangi
and Chowdhury68 and Huang.75 Grazzini and Gori76 reconsidered the entropy generation analysis, and distinguished
between three separate entropy generation number definitions, two of which are new. They investigated the extrema
of these numbers, and applied their results to an air-to-air
counterflow heat exchanger. In chemical engineering,
Tondeur77 and Tondeur and Kvaalen78 showed that, for a
1199
given duty, the best configuration of a heat and mass process

is that where the entropy generation rate is distributed in the
most uniform way possible.
A study of crossflow heat exchangers was conducted by
Baclic and Sekulic.79 Their study reveals once again the
tradeoff between heat transfer and fluid flow irreversibilities,
as well as the remanent ~imbalance! irreversibility associated
purely with the crossflow arrangement. Several basic studies
of the thermodynamics of forced convection heat transfer in
heat exchangers were undertaken by Soumerai.80,81 The relationship between entropy generation minimization and cost
minimization was illustrated by Wepfer et al.82 in the problem of deciding the optimal size of a steam pipe and its
insulation.
The generation of entropy at the local ~differential! level
in heat exchangers was documented by El-Sayed,83 Liang
and Kuehn,84 Evans and von Spakovsky,85 and Drost and
Zaworski.86 The calculation and display87 of entropy generation rate distributions through heat and fluid flow fields was
recommended by Paoletti et al.,88 White and Drost,89 Cheng
and Huang,90 Drost and White,91,92 Benedetti and Sciubba,93
and Cheng et al.94 Taken together, these studies argue that
commercial computational fluid dynamics packages should
have the capability of displaying entropy generation rate
maps for both laminar and turbulent flows.
For example, Figs. 13~a! and 13~b! show the flow and
temperature fields in the entrance region to a vertical channel
with mixed ~forced1natural! convection. The right wall is
heated ~T 1!, and is fitted with horizontal fins. The left wall is
cooled to the ambient temperature ~T 0!, which is also the
temperature of the fluid that enters through the bottom of the
channel. New in this display of numerical information is Fig.
13~c!, which shows the contour lines of the local ~volumetric! entropy generation rate. For the latter, Paoletti et al.88
proposed plotting contour lines for constant values of the
/(S gen,DT
1S dimensionless group S gen,DT
gen,D P ), to show the
relative importance of heat transfer irreversibility in the local
total entropy generation rate ~see also Refs. 93 and 95!.
The thermodynamic optimization of complete heat exchanger systems was performed by Tapia and Moran,96 Ranasinghe et al.,97 and Li et al.98 Heat exchanger networks
were optimized by Hesselmann,99 Chato and Damianides,100
and Sekulic and Milosevic.101 Regenerators were optimized
by Tsujikawa et al.102 for gas-turbine power plants, by
Hutchinson and Lyke103 for Stirling cycle machines, and
Matsumoto and Shiino104 for cryogenic plants. Heat exchangers with offset strip-fin surfaces were optimized more
recently by Schenone et al.105
The method was also extended in several fundamental
directions of heat transfer science. The generation of entropy
by pure heat conduction, in the absence of fluid flow, has
come under close scrutiny in more recent papers. Bisio106
focused on one-dimensional heat conduction in systems with
time-dependent boundary conditions. The thermal conductivity was a function of temperature and the coordinate of the
heat flux direction. He examined homogeneous as well as
multilayered systems, and pointed out relations between entropy generation extrema and the solution to the heat conduction problem ~the temperature distribution!. Extensions of
1200
FIG. 13. Flow, temperature, and volumetric entropy generation rate in a

vertical duct with mixed convection ~from Ref. 94: Courtesy of Prof. C.-H.
Cheng, Tatung Institute of Technology, Taipei!.
this work cover conduction in a medium with thermal conductivity ~or its derivative! that is a piecewise continuous
function of temperature,107 and thermal conductivity that is
also a function of the temperature gradient in the heat flux
direction.108
An entirely new area of application of entropy generation by conduction is being charted by the work of Kinra and
his associates109113 at Texas A&M University. They used the
entropy generation calculations to explain and predict damping in homogeneous and inhomogeneous elastic systems.
They named their theory elastothermodynamic damping.
Kinras approach begins with the observation that a material that is stressed reversibly and adiabatically always experiences local changes in temperature, however small. This
thermoelastic effect can be predicted using the first law and
the second law. The new observation is that since the temperature field and the stress field are coupled, nonuniformities in stress and materials properties induce nonuniformities
in temperature. As a consequence, heat is conducted locally
from regions of relatively high temperature to regions of low
temperature. The entropy generated by conduction throughout the material is responsible for the damping effect.
Kinra, Bishop, and Milligan have used this approach to
demonstrate that it is now possible to design a composite
material that has certain, desired damping characteristics.

Theirs is an interdisciplinary example of the importance of
the concept of entropy generation in the field of materials
design. This is an important development, because the other
applications collected in this review come strictly from the
mainstream of thermal engineering ~Fig. 1!.
As an analogy to the convective heat transfer irreversibility illustrated in Fig. 10, the competition between mass
transfer and fluid flow irreversibilities was demonstrated by
San et al.114 The irreversibility of combined heat and mass
transfer was minimized in internal flow by San et al.115 and
in external flow by Poulikakos and Johnson.63 Carrington
and Sun116,117 reconsidered the internal and external flow
problems. Sun and Carrington118 developed a general analysis of the flow of a fluid mixture with heat conduction, mass
diffusion, fluid friction, and chemical reactions.
The generation of entropy in reacting flows with radiation has been studied by Arpaci and his associates119124 and
Puri.125 For example, Arpaci and Selamet122 focused on premixed flames stabilized above a flat flame burner, and
showed that the tangency condition ~i.e., the minimum
quench distance! is related to an extremum of the entropy
generation rate, which is inversely proportional to the Peclet
number. Puri125 minimized the entropy generation rate of a
droplet burning in a stream. He showed that the minimum
entropy generation rate corresponds to a tradeoff between
drag and mass loss from the droplet, and to maximum energy
per unit mass flowrate at the combustor exit.
The entropy generation rate in convection through a
saturated porous medium was presented in Ref. 126. The
general form for the local entropy generation rate due to
combined heat, mass, and fluid flow through a porous medium is given in Ref. 14.
V. STORAGE SYSTEMS
The optimization of time-dependent heating and cooling

processes has generated a subfield of its own. The applications are diverse, for example, sensible heating versus latent
heating, or the temporary storage of heating versus the
storage of refrigeration. Common to all these applications
is the search for optimal strategies for executing heating and
cooling processes, i.e., the search for optimal histories, or
optimal evolutions in time.
The earliest work of this type127 focused on the sensible
heating of a storage element ~solid or incompressible liquid!
of mass M and specific heat C, by using a stream of hot gas
,c p ,T ` ) as shown in Fig. 14. Initially, the storage element
(m
is at the ambient temperature T 0 . Gradually, the element temperature T and the gas outlet temperature T out rise and approach T ` . The temperature histories T(t) and T out(t) can be
written analytically as functions of the dimensionless time u
and the number of heat transfer units N tu based on the
stream-element thermal conductance UA ~fixed!,
u 5t
cp
m
,
MC
N tu5
UA
.
cp
m
~31!
Figure 14 shows that the sensible heating process has

two sources of entropy generation, the heat transfer between
FIG. 14. Solid or liquid-bath element for sensible heat storage ~from
Ref. 127!.
the stream and the storage element, and the heat transfer
between the exhaust and the ambient. A third source, neglected here but treated in Ref. 127, is the pressure drop on
the gas side of the UA heat exchanger. It was shown that the
total entropy generated from t50 to t5t is minimum at a
certain time ~duration! of the heating process, t opt or uopt . In
the limit t!t opt , the generated entropy is due mainly to the
internal source, while in the limit t@t opt the dominant source
is the external thermal mixing ~Fig. 14!. Charts for calculating the optimal heating time uopt as a function of N tu and
(T ` 2T 0 )/T 0 are available in Refs. 1 and 127. For
(T ` 2T 0 )!T 0 , the optimal heating time is available in
closed form:
u opt5
1.256
.
12exp~ 2N tu!
~32!
Since in most heat exchangers N tu is of the order of 1 or

greater, Eq. ~32! shows that uopt is of the order of 1, or that
c p ). In conclusion, for engineering design purt opt;M C/(m
poses, the optimal heating strategy is such that the process
must be terminated when the thermal inertia of the hot gas
c p t) becomes of the same order as the thermal inerused (m
tia of the storage element (M C).
The thermodynamic optimization of the element of Fig.
14 during a complete storage and removal cycle ~i.e., heating
followed by cooling back to T 0! was performed by Krane.128
An improvement to the single-element storage method of
Fig. 14 is the use of several elements in series, as proposed in
Ref. 1. This proposal was investigated in great detail by Taylor et al.129 based on a solid distributed-storage-element
model in which the storage material temperature varied continuously along the stream. Taylor et al. showed among other
things that the longitudinal conduction of heat through the
storage material during the periodic operation of the heat
exchanger can have a major impact on the overall irreversibility of the installation. The overall entropy generation is
again a strong function of the time interval required by the
storage part of the cycle: the identification of the optimal
storage time interval is critical. The series of storage units
was optimized further by Sekulic and Krane.130
Closely related to the continuous one-dimensional storage scheme with periodic counterflow circulation is the class
of periodic heat exchangers recognized as regenerators. The
design of this type of heat exchanger was approached on the
basis of EGM by San et al.131 Their model consists of twodimensional parallel-plate channels sandwiched between
1201
This question is essential in the operation of large scale superconducting systems, which must be cooled down before
they can function. Note that to minimize the amount of cooling agent ~cryogen! is equivalent to minimizing the refrigerator work needed to produce the cryogen, or the total entropy generated in the cold space.
This problem was solved by variational calculus for the
opt(t) that minimizes the m inteoptimal flow rate history m
gral ~33! subject to the time constraint t c . The details can be
found in Refs. 1 and 140. The end result is
opt~ t ! 5
m
FIG. 15. Batch cooling and temperature history during a cooldown process
~from Ref. 140!.
slabs of storage material. The longitudinal conduction of heat

through the storage material is neglected. An important difference between this regenerator model and the continuous
storage system analyzed by Taylor et al.129 is that in the regenerator the stream exhausted during the storage phase is
not dumped into the atmosphere.
The minimization of entropy generation in periodic-flow
regenerative heat exchangers was studied also by Hutchinson
and Lyke,103 Shen and Worek,132 Mathiprakasam and
Beeson,133 Das and Sahoo,134 and Sahoo and Das.135 The
thermodynamic arguments of this section were combined
with overall cost minimization arguments into a thermoeconomics extension of the sensible heat storage process by Badar et al.136 and Kotas and Jassim.137
An overview of the thermodynamic optimization of sensible heat storage methods was presented by Rosen et al.138
Treated were the individual storage and removal processes as
well as the complete cycle. The emphasis was on the development of simple and consistent ways ~conventions! to
evaluate and compare the performance of competing designs.
This work was continued by Rosen139 who showed that different efficiency measures are suitable for different applications, and that it is important to agree on a common efficiency definition before comparing designs of the same class.
The thermodynamic optimization of heating and cooling
processes continued in two additional directions. One is the
optimization subject to a finite-time constraint. Figure 15 illustrates the fundamental question of how to cool a mass to a
prescribed temperature level (T L ) during a fixed time interval t c , while using the minimum quantity of coolant140
m5
tc
1202
~ t ! dt.
m
~33!
U~ T !A
C *c p~ T !
1/2
~34!
where T(t) is the companion result for the optimal body

temperature history. The constant C * can be evaluated by
using the time constraint t c . Equation ~34! is a compact result with important engineering implications. Bearing in
mind that the heat transfer coefficient U varies as T decreases, we learn that during poor heat transfer conditions
~low U! the flow rate should be lower. Furthermore, since
during cooldown the coolant c p increases, the flow rate
should decrease even more as the end of the process nears.
The engineering work on the optimization of heating and
cooling subject to time constraints was continued recently in
the physics literature by Andresen and Gordon,141 who minimized the generation of entropy instead of the amount of
heating or cooling agent used. As heat source or heat sink
they assumed a heat reservoir of temperature that can be
(t) of Fig.
varied at will, instead of the coolant flow rate m
15. Between the heat reservoir and the thermal inertia they
assumed several heat transfer rate laws, e.g., convection with
a constant heat transfer coefficient, and radiation with constant ~temperature-independent! emissivities. In a companion
paper, Andresen and Gordon142 considered the related problem where the heating of the body of interest is effected by a
stream, while the heat reservoir temperature may change or
remain constant. The result of the EGM procedure is again
an optimal flow rate of the heating agent. In both papers the
effect of pressure drop was neglected.
The other direction of the work on storage systems is
concerned with phase-change storage elements, e.g., latent
heating instead of sensible heating. This activity began with
the papers of Bjurstrom and Carlsson143 and Adebiyi and
Russell144 who applied to a phase-change storage element the
entropy generation analysis given in Ref. 127 and Fig. 14 for
the single-phase element. These authors showed that the entropy generated during heating ~melting! is minimum when
the melting temperature of the storage material has the optimal value
T m,opt5 ~ T ` T 0 ! 1/2.
~35!
Technologically, this result is extremely valuable because it

guides the designer in the selection of the type of phasechange material.
The details of the real ~time-dependent! melting and solidification processes were accounted for in Refs. 145 and
146, where it was shown that Eq. ~35! is a good approximation for the optimal phase-change temperature. Lim et al.147
showed that Eq. ~35! holds for the entire melting and solidiAppl. Phys. Rev.: Adrian Bejan
FIG. 16. Latent heat storage by using two phase-change materials in series
~from Ref. 147!.
fication cycle when the fluid is modeled as well mixed inside

the storage element. The same authors showed that the thermodynamic performance of the latent heat storage process
can be improved by using two or more different phasechange materials in series. Two such elements are pictured in
Fig. 16, for which there are two optimal melting temperatures ~i.e., two materials! as shown in Fig. 17. The dimensionless optimal temperatures in Fig. 17 are referenced to Eq.
~35!,
namely
t 1,opt5T m,1,opt /(T ` T 0 ) 1/2 and t 2,opt
1/2
5T m,2,opt /(T ` T 0 ) . Lim et al.147 similarly optimized an infinite number of phase-change materials heated in series, and
a single material melted by an unmixed stream.
The thermodynamic optimization of practical storage installations with melting and solidification has attracted a lot
of interest in both engineering and physics. For example,
Adebiyi148 considered a bed with particles with several ratios
of latent heat to sensible heat storage capability. He modeled
the heat conduction as one dimensional in the cylindrical
pellet geometry. For the complete storage and removal cycle,
he found that the optimal phase-change temperature is equal
to the arithmetic average of the heat source and ambient
temperatures.
Charach and Zemel149 optimized the thermodynamic
performance of latent heat storage in the shell of a shell-andtube heat exchanger. The focus was on the effect of twodimensional heat transfer, that is, as a step beyond the onedimensional model employed by De Lucia and Bejan.145
Charach and Zemel149 also considered the effect of pressure
drop on the stream side of the heat exchanger. Their analysis
has been extended by Charach and Zemel150 and Charach151
to the complete melting and solidification cycle, by using a
quasisteady treatment of the phase change process occurring
in the shell. These latest studies showed that the optimal
phase-change temperature is bounded from above and from
below by the arithmetic and, respectively, geometric averages of the source and ambient temperatures.
Another interesting direction has been the study of latent
heat storage units coupled in series with a power plant, and
optimized over the entire storage and removal cycle. Bellecci
and Conti152 showed that minimization of entropy generation
and operation stability are two competing criteria in the optimization of the aggregate installation. The arithmetic average of the extreme temperatures emerges again as the optimal phase-change temperature. Detailed modeling and
numerical simulations of the heat transfer behavior of the

shell-and-tube phase-change heat exchanger were performed
by Bellecci and Conti.153,154
Aceves-Saborio et al.155 developed a systematic way of
modeling phase-change storage systems, by applying the
lumped model to many independent phase-change capsules.
They also considered the more general case where the phasechange material melts over a finite temperature range. In an
earlier study, Aceves-Saborio et al.156 optimized a single
latent-heat cell by first simulating numerically the phasechange process. This is an important and difficult task because the natural convection currents that occur in the liquid
control the shape and movement of the liquidsolid interface
~e.g., Sec. 10.4 in Ref. 157!. It was shown by De Lucia and
Bejan145 that when the melting process is controlled by natural convection the optimal melting temperature is equal to
the geometric average shown in Eq. ~35!.
Adebiyi et al.158 constructed a numerical model for
simulating and then optimizing the performance of storage
systems with multiple phase-change materials. They found
that the second law efficiency of systems with multiple materials can exceed by 13%26% the efficiency of systems
employing a single material. A fundamental study of EGM in
time-dependent unidirectional heat conduction was conducted by Charach and Rubinstein.159
VI. SOLAR POWER PLANTS
The generation of mechanical or electrical power has

been subjected to thermodynamic optimization in many studies that cover a vast territory. Therefore in this section and
Sec. VII we cover only some of the central ideas and the
literature that followed. We do this chronologically, i.e., in
the order in which the use of the method reached critical
mass in the various sectors of the power generation field.
This occurred first in the optimization of solar driven power
plants, as exemplified by the solar theme chosen for a
recent book160 dedicated to the application of the method.
Book chapters on the application of the EGM method to
solar power plants appeared earlier in Refs. 1 and 14.
In a 1957 paper, Muser161 optimized the power produced
by an engine driven by solar heating. Musers model is
equivalent to the upper portion drawn between T s and T L in
Fig. 18. The hot end of the heat engine cycle receives heat
from a solar collector (T H ) that sees not only the solar disc
(T s ) but also the cold universe ~T `!. The heat engine cycle is
modeled as reversible, with the cold end in equilibrium with
the ambient. Muser161 and, independently, Castans,162
Jeter,163 and De Vos and Pauwels164 showed that the power
is maximum when the collector temperature has
output W
the optimal value given by
4T 5H,opt23T L T 4H,opt2T 4s T L 50.
~36!
This was an important first step in the demonstration that

an optimal design can be found based on a model that combines only the thermodynamics of the (T H 2T L ) compartment with the irreversibility of radiation heat transfer in the
three-surface enclosure formed by T H , T s , and T ` . The
same combination of thermodynamics and radiation heat
1203
FIG. 18. Model of solar power plant with heat transfer irreversibilities at the
hot end and the cold end ~after Refs. 14 and 65!.
FIG. 17. The optimal melting temperatures of the phase-change materials in

the series arrangement of Fig. 16 ~From Ref. 147!.
)
2Q
shows that when the external irreversibility due to (Q
is taken into account, the maximization of W is indeed analogous to the minimization of S gen , in agreement with the
Gouy-Stodola theorem ~5!.
De Voss book160 is also a review of the models such as
Fig. 18 that have been used in physics. In this section we
review some of the engineering contributions to the field,
which are not covered in Ref. 160. For example, the model
of Fig. 18 refers to an extraterrestrial solar power plant that
uses a radiator to reject heat to the universe ~T `!. The heat
transfer from T L to T ` is by radiation in a two-surface enclosure. In Refs. 14 and 65 this power plant model was optimized subject to the total area constraint
A H 1A L 5A.
transfer was the method used independently in 1971 by Martynovskii et al.30 in the optimization of radiation shields.
Another important observation is that to maximize the
is equivalent to minimizing the total entropy
power output W
generation rate associated with the solar power plant ~Ref.
14!. Note that as T H is varied during the optimization pro , from T to
cess, the net solar heat input to the collector ~Q
s
T H ! also varies. And if Q is to float freely, then the actual
heat input available from the sun must be greater than any Q
value that might be required in the course of W maximiza be this sufficiently large ~and fixed! heat transfer
tion. Let Q
),
is intercepted by the collector (Q
rate. A portion of Q
while the remainder (Q 2Q ) must necessarily fall on the

ground, i.e., be rejected to the ambient T L . Both portions
vary with the lone degree of freedom in the power plant
, or W
!. The total entropy generation rate
design ~T H , Q
S gen5
1204
2Q
1
Q
Q
Q
W
1
2
1
52 1Q
2
TH Ts
TL
TL
TL Ts
~37!
~38!
The design has two degrees of freedom, and the twicemaximized power output is represented by
A H,opt50.35A,
F Hs
S D
Ts
T H,opt
A L,opt50.65A,
~39!
51.538,
~40!
W
max50.0414s AF Hs T s ,
~41!
where s and F Hs are the StefanBoltzmann constant and the

collector-sun view factor. If we substitute T s 55762 K and
F Hs ;1024 into this solution we obtain T H,opt > 520 K and
T L,opt > 340 K for the recommended extreme temperatures
~boiler, condenser! of the power cycle executed by the working fluid. Another technological aspect of this result is that
the radiator area should be roughly twice as large as the
collector area.
As a way to think, the extraterrestrial solar power plant
model of Fig. 18 is related to the conversion of solar heating
into wind power ~natural convection! on earth. This power is
FIG. 19. Solar power plant model with collector-ambient heat loss and
collector-power cycle heat exchanger ~from Ref. 167!.
destroyed almost totally by friction and heat transfer across

finite temperature differences. The thermodynamic heat engine that drives any natural convection process was also
noted in the field of natural convection heat transfer ~Ref.
126!. The theoretical limits of the conversion of terrestrial
solar heating into wind power were investigated by Gordon
and Zarmi,165 De Vos and Flater,166 and De Vos.160
Another engineering model is the power plant driven by
a solar collector with convective heat leak to the ambient,167
which is shown in Fig. 19. The heat leak was modeled as
proportional to the collectorambient temperature differ 5(UA) (T 2T ). The internal heat exchanger beence, Q
0
c
c
0
tween the collector and the hot end of the power cycle ~the
5(UA) (T 2T ). It was
user! was modeled similarly, Q
i
c
u
found that there is an optimal coupling between the collector
and the power cycle such that the power output is maximum.
This design is represented by the optimal collector temperature
T c,opt u 1/2
max1R u max
,
5
T0
11R
~42!
where R5(UA) c /(UA) i , u max5T c,max /T0, and T c,max is the

maximum ~stagnation! temperature of the collector. The corresponding optimal coupling between a collector with heat
loss and a refrigerator was documented by Sokolov and
Hershgal.168
Radiation heat transfer contributes significantly to the
collectorambient heat loss mechanism when the collector
operating temperature rises above the 100200 C range.
The radiation effect on the heat loss and the optimization of
the solar power plant were analyzed by Howell and
Bannerot.169 A sample of their results for optimal collector
temperature is presented in Fig. 20 for four classes of collector designs ~A, B, C, D!, which are described in Ref. 169.
The dimensionless radiative heat loss parameter a and convective heat loss parameter b are also defined in Ref. 169.
Howell and Bannerot produced similar graphs for the optimal thermodynamic design of solar driven power cycles
~Stirling, Ericsson, Brayton!, heat pumps, and absorption refrigerators.
The solar power plant model with collectorambient
heat loss and heat exchanger between the cold end of the
FIG. 20. Optimal collector temperatures for maximum power generation

~after Ref. 169!.
power cycle and the ambient was optimized in Ref. 170 subject to a total heat transfer area constraint. The same study
presents the area-constrained optimization of a model with
phase-change energy storage at the hot end of the cycle,
between the collector and the working fluid. One useful result is that the melting material must be such that its melting
point is the geometric average of the collector and ambient
temperatures,
T m,opt5 ~ T c T 0 ! 1/2.
~43!
Solar power plant models with nonisothermal collectors

were optimized in Ref. 167. Models with collectors operating under time-varying conditions dictated by the daily insolation cycle were optimized in Refs. 171 and 172. A model
with time-dependent solar heating and variable amount of
fluid in the collector was optimized in Ref. 172. In this
model the collector has the ability to store the solar input as
sensible heat, and to deliver it to the rest of the power plant
when the solar heating effect is less intense.
The common message of these models is that several
extremely basic tradeoffs exist in the thermodynamic optimization of power plants driven by heat transfer from the sun.
The models share the feature that heat loss always occurs
between the collector and the ambient. The thermodynamic
tradeoffs are of two kinds. When the overall size of the installation is constrained, there is an optimal way of allocating
the hardware ~e.g., heat transfer area! between the various
components. When the daily variation of the solar heat input
is known, there is an optimal time-dependent strategy of operating the power plant.
Thermodynamic tradeoffs have been found in the optimization of the direct ~photovoltaic! conversion of solar radiation. This work was reviewed in the book by De Vos160
and continues today. For example, Baruch and Parrott173 examined the possibility of constructing a Carnot cycle for the
conversion of photovoltaic energy by using electronhole
plasma as the working substance.
The basic thermodynamic limits to the conversion of solar radiation have attracted considerable attention. This work
was reviewed in Ref. 174, where it was also shown that the
1205
ture of the stream is equal to T HC . There is only one degree

of freedom in the optimization of the power plant: the hot
end of the inner compartment, or the exhaust temperature
as a function of T and
T HC . It is not difficult to express W
HC
is maximized when
to show that W
T HC,opt5 ~ T H T L ! 1/2.
~44!
Chambadals corresponding energy conversion efficiency at

maximum power is
h 512
FIG. 21. The sources of entropy generation in Chambadals power plant
model ~from Ref. 197!.
different ideal conversion efficiencies reported by Petela,175

Spanner,176 and Jeter163 are complementary, and can be unified into a single theory. Related aspects were considered
more recently by Badescu.177180 For example Badescu178
showed that the maximum efficiency decreases abruptly
when the collector concentration ratio decreases.
Roy and Grasse181 reviewed comparatively the practical
merits of solar power technologies across the entire spectrum, from photovoltaic to solarthermal applications. The
mechanisms of entropy generation in solar collectors were
also investigated by Fujiwara,182 Suzuki,183 and Han and
Lee.184 The thermodynamic optimization of collectors
coupled with power plants was considered further by
Grazzini,185 Borner et al.,186 De Vos et al.,187 and Yan and
Chen.188 Models with radiation-dominated collector and radiator were optimized by De Vos and van der Wel189 and
Gotkun et al.190
A more applied line of inquiry concerns the thermodynamic optimization of specific designs of processes and
power plants driven by solar heating. Recent examples are
the optimization of flat-plate solar air heaters,191,192 phasechange energy storage,193 and solar assisted heat pumps.194
The optimal coupling between solar collectors and Stirling
and Ericsson cycles was also documented by Badescu.195
Another type of power plant driven by a renewable energy source is the power plant with heating from a hot-dryrock system. The optimal time-dependent strategy of running
such a power plant ~e.g., the optimal water flow rate! was
developed in Ref. 196.
VII. NUCLEAR AND FOSSIL POWER PLANTS
In 1957 in nuclear engineering, Chambadal2 and

Novikov5 showed independently that the hot-end temperature of a power plant can be optimized such that the power
output is maximum. Chambadals analytical argument corresponds to the model drawn by the author in Fig. 21. The
power plant is driven by a stream of hot single-phase fluid of
inlet temperature T H and constant specific heat c p . The
power plant model has two compartments. The one sandwiched between the surface of temperature T HC and the ambient (T L ) operates reversibly. The area of the T HC surface is
assumed to be sufficiently large such that the outlet tempera1206
S D
TL
TH
1/2
~45!
It can be shown that the same efficiency formula ~45! holds

when the heat exchanger area is finite and the exhaust temperature is higher than hot-end temperature of the reversible
compartment.197 Equation ~45! also holds when the unmixed
stream of Fig. 21 is replaced with a single-temperature
~mixed! fluid inside that heat exchanger.197
The maximum-power efficiency ~45! can also be derived
by minimizing the total entropy generation rate associated
with the power plant.197 One obvious source of entropy generation in Fig. 21 is the heat exchanger. The other is less
obvious: it is the dumping of the used stream into the ambient. This second source was featured prominently by an entire subfield dedicated to the optimization of storage systems
~Sec. IV, Fig. 14!. The dumping of the T HC -hot exhaust is an
essential part of the optimization process: T HC is a degree of
freedom only when the exhaust (T HC ) is free to float, i.e.,
when it is not required ~used! by someone else downstream.
The external irreversibility indicated in Fig. 21 is an essential
part of the physics of the optimization process: without it the
plant design cannot be optimized. This additional irreversibility is what gives the design room to move, therefore it can
be called the room-to-move irreversibility.
With reference to Fig. 21, the total entropy generation
rate due to the power plant is
~46!
where the stream was treated as an ideal gas at constant
5m
5m
c p (T H 2T HC ) and Q
c p (T HC 2T L ).
pressure, Q
H
e
Equation ~46! shows that S gen has a minimum with respect to
T HC , and that the S gen,min design corresponds to the
maximum-power formulas ~44! and ~45!.
It is worth noting that if we had overlooked the roomto-move irreversibility, that is, if we had written only the
entropy generation associated with the visible confines of the
power plant, then we would have found that S gen has a minimum at a T HC value that differs from the maximum-power
value ~44!. These T HC values differ not because maximum
power and minimum entropy generation rate are two different designs, but because an oversight has occurred in the
evaluation of the total rate of entropy generation. This observation sheds light on the physics literature claim198 that, in
5 (1 1 i)Q
(s d 2 s a ) 5 (1 1 i)(s d,rev 2 s a ), or Q
L
L,rev , where
~11i!>1 and Q L is the heat transferred to the ambient (T L ).

The rest of the power plant operates reversibly. Novikovs
optimal heating temperature and efficiency for maximum
power output,
T HC,opt5 ~ 11i ! 1/2~ T H T L ! 1/2,
h 512 ~ 11i ! 1/2

FIG. 22. Power plant driven by heating from a combustion chamber with
well mixed products of combustion ~from Ref. 14!.
the same power plant, minimum entropy generation and

maximum power are two different design conditions.
A maximum power design similar to Chambadals is the
optimal combustion chamber temperature that was derived
independently in Ref. 14. Figure 22 shows a two-part model
of a power plant with isobaric and isothermal ~well mixed!
combustion chamber. It was shown that when the specific
heats of all the products of combustion are assumed to be
sufficiently constant ~independent of temperature!, the maximum power design corresponds to a hot-end ~flame! temperature (T f ) equal to the geometric average of the adiabatic
flame temperature ~T af! and the ambient temperature ~T 0!,
T f ,opt5 ~ T afT 0 ! 1/2.
~47!
The maximization of power output was also described

by Odum and Pinkerton,199 who gave several examples from
engineering, physics, and biology, without deriving results
such as Eqs. ~44! and ~45!. Equal credit for such results goes
to Novikov5 who, like Chambadal, published them in 1957.
Novikovs analysis was reprinted in English in NorthAmerican engineering textbooks200,201 as well as in Russian
engineering textbooks.7,202,203 His model is shown in Fig. 23.
The hot-end heat exchanger of finite thermal conductance
into the working fluid,
UA drives the heat transfer rate Q
H
which is heated at constant temperature (T HC ) from state (b)
to state (c). The fluid is expanded irreversibly from (c) to
(d): Novikov accounted for this irreversibility by writing
S D
TL
TH
~48!
1/2
~49!
match Chambadals Eqs. ~44! and ~45! in the limit where the
expansion is executed reversibly ~i50!.
The efficiency formula ~45! was rediscovered in 1975 in
the physics literature by Curzon and Ahlborn.8 Their model
differed from Chambadals and Novikovs in two respects.
First, the power plant operated in unsteady fashion, in time.
It executed a four-process ~two-stroke! cycle modeled as in
Sadi Carnots original memoir, however, the piston and cylinder apparatus made contact during finite time intervals
with the two heat reservoirs. This contact occurred across
finite temperature differences. The second new feature in
Curzon and Ahlborns model is the heat transfer irreversibility ~finite thermal conductance! placed at the cold end of the
cycle. In summary, the maximum-power efficiency ~45! may
be called, chronologically, the ChambadalNovikov
CurzonAhlborn efficiency, or the CNCA efficiency for
short.
An entirely different way of modeling the irreversible
operation of a power plant was proposed in 1976 in
engineering.204 The loss of heat from the hot end of the
power plant was modeled as a thermal resistance ~bypass
heat leak! in parallel with an irreversibility free compartment
, Fig. 24. The hotthat produces the actual power output W
end temperature T H could vary. The heat leak was modeled
as proportional to the temperature difference between the hot
5C(T 2T ), where C is the therend and the ambient, Q
C
H
L
mal conductance of the leaky insulation of the power plant.
The power is maximum when the hot-end temperature T H
reaches the optimal level
FIG. 23. Novikovs model for a steady-state power plant with heat transfer and expander irreversibilities ~from Ref. 197!.
1207
FIG. 24. Power plant model with bypass heat leak ~after Ref. 204!.
T H,opt5T L 11
Q
H
CT L
1/2
~50!
The corresponding efficiency ~W

max/Q H ! is expectedly lower
than the Carnot efficiency,
12T L /T H,opt ~ 11r ! 1/221
,
~51!
5
11T L /T H,opt ~ 11r ! 1/211
/(CT ) is the dimensionless thermal resistance
where r5Q
H
L
of the power plant insulation. An optimal hot-end temperature exists because when T H , T H,opt the Carnot efficiency of
the right-hand side of the model of Fig. 24 is too low, while
bywhen T H . T H,opt too much of the fixed heat input Q
H
passes the power producing component.
An interesting aspect of this power plant model is that
when we minimize the entropy generated inside the dashed
box between T H and T L in Fig. 24, we find an optimum that
differs from Eqs. ~50! and ~51! ~Ref. 1!. This paradox is an
indication that the entropy generation calculation is incorrect, as noted earlier under Eq. ~46!. For if T H is a degree of
freedom in the design, then the hot end of the power plant
must be free to float between the ambient and another ~necessary! heat reservoir whose temperature T f ~fixed! must be
always higher than T H . The neglected room-to-move irre
versibility occurs as Q
H crosses the temperature gap
(T f 2T H ). It is easy to show that by minimizing the entropy
generated between T f (.T H ) and T L in Fig. 24, the optimal
hot-end temperature and efficiency match Eqs. ~50! and ~51!
~Ref. 12!.
Curzon and Ahlborns work8 triggered a series of papers
on power maximization in the physics literature. Reviews of
this specialized line of research were published by
Andresen205 and Andresen et al.206 Noteworthy in this series
is the first follow up paper,207 which appeared in 1977 and in
which the power plant was modeled in the steady state. Compared with Novikovs model ~Fig. 23!, the model of Ref. 207
had three new features: ~i! a finite thermal conductance at the
cold end, ~ii! a speed-dependent dissipation of mechanical
power caused by shaft friction, and ~iii! a speed-independent
FIG. 25. Steady-state power plant model with finite hot-end and cold-end
thermal conductances ~from Ref. 1!.
h5
1208
contribution to power dissipation. Feature ~i! makes the

steady state power plant model look like the one drawn in
Fig. 25, which will be discussed shortly. Feature ~ii! illustrates the interplay between friction, thermodynamics, and
heat transfer in the application of the method ~Fig. 1!. Regarding feature ~iii!, to which Andresen et al.207 refer as
heat leak, it must be noted that the origin of that energy
interaction is the work reservoir, or, in the authors own terminology, a reservoir with infinite temperature ~i.e., an
energy interaction accompanied by zero entropy transfer!.
Feature ~iii! is actually power dissipation ~e.g., electric, Joule
heating! or a friction leak as stressed by Grazzini,208 not a
bypass heat leak in the sense of Fig. 24.
The steady-state power plant model with two finite heat
exchangers was proposed independently in Ref. 1, not for
deriving the CNCA efficiency ~45!, but for answering a more
direct technological question: Given the two heat exchangers
~U H A H and U L A L ! and an additional unit of expensive heat
transfer area (DA), should DA be placed at the hot end, or at
the cold end? The finite-area allocation problem was generalized by considering a power plant that makes contact with
an infinity of heat reservoirs distributed over a finite temperature interval, and by deriving based on variational calculus the optimal distribution of area over the same temperature interval ~Ref. 1!.
The optimal allocation of heat exchanger inventory in
the power plant model of Fig. 25 was illustrated in Refs. 10
and 14, where it was assumed that the total thermal conductance is fixed,
U H A H 1U L A L 5UA.
~52!
The power output, which was maximized once to arrive at

Eq. ~45!, can be maximized one more time by choosing the
proper ratio U H A H /(U L A L ) subject to constraint ~52!. The
optimal conductance allocation rule is
~ U H A H ! opt5 ~ U L A L ! opt .
~53!
This optimization result also holds when the conductance

inventory UA is minimized subject to fixed power output.209
If, instead of Eq. ~52!, the optimization is performed subject
to the total area constraint
A H 1A L 5A
~54!
the optimal way to divide the area inventory is210

1
A H,opt
5
.
A
11 ~ U H /U L ! 1/2
~55!
Optimization results such as Eqs. ~45!, ~53!, and ~55! can

also be derived by minimizing the entropy generation rate of
the power plant model of Fig. 25. In this alternate approach
it is essential to note that the optimization requires that Q

H
vary, i.e., that the appropriate room-to-move entropy generation rate197 must be included in the calculation of the total
entropy generation rate. On the other hand, if the heat input
is treated as fixed,197 the only degree of freedom left in
Q
H
the optimization of the model of Fig. 25 is the allocation of
the finite heat exchanger inventory. For example, if the UA
inventory is constrained in accordance with Eq. ~52!, then
the optimal allocation rule continues to be Eq. ~53!, with the
corresponding maximum efficiency197 @see also the discussion under Eq. ~63!#:
W
TL
max
h5
512
Q
T
H
12
4Q
H
UAT H
21
~56!
Rubin and Andresen211 optimized the temperature staging of two power plants fitted with three finite-size heat exchangers, Fig. 26. They showed that the efficiency at maxi 1W
) is given by the same formula as
mum total power (W
1
2
in Eq. ~45!. The power output was maximized further in Ref.
170, where the total thermal conductance inventory was constrained
U H A H 1U M A M 1U L A L 5UA.
FIG. 26. Combined-cycle power plant with three heat exchangers ~from
Ref. 170!.
leading to the CNCA efficiency, Eq. ~45!. Second, the finite

UA inventory ~52! must be divided equally between the two
heat exchangers, cf. Eq. ~53!. Third, there is a tradeoff between investing more in the heat exchanger equipment (UA)
and in the resistance R i to the bypass heat leak
~57!
The optimal way to allocate UA to the three heat exchangers,

and the resulting maximum power are
~ U H A H ! opt5 ~ U M A M ! opt5 ~ U L A L ! opt ,
1W
! 5
~W
1
2 max
F S DG
TL
1
UAT H 12
9
TH
~58!
1/2 2
~59!
A power plant model that combines the irreversibility

mechanisms illustrated in Figs. 24 and 25 is the model with
heat leak and two heat exchangers,10 Fig. 27. The power
output can be maximized in three ways. First, the optimal
temperature ratio across the imagined reversible compartment turns out to be the same as in Ref. 8,
S D S D
T HC
T LC
opt
TH
TL
1/2
~60!
FIG. 27. Power plant model with bypass heat leak and two finite-size heat
exchangers ~from Ref. 10!.
1209
Figs. 2123 and 2527!. However, in the presentation of the

empirical efficiency data T H was erroneously taken as equal
to the highest temperature reached by the working fluid, instead of the equivalent temperature of the combustion chamber as an exergy source ~Ref. 14!. An alternative explanation
for the position of the empirical data is provided by the
model of Fig. 29, in which the bypass heat leak of Fig. 24 is
optimized in harmony with the power-producing compartment, i.e., in accordance with the method used in cryogenics
~Fig. 3!. When the entropy generated inside the insulation of
resistance R i is minimized as in Eqs. ~6!~10!, the power
plant model of Fig. 29~b! delivers maximum power with the
efficiency170
FIG. 28. The efficiencies of: actual power plants, models with finite heat
exchangers, and a model with bypass heat leak optimized in harmony with
the power-producing compartment ~From Ref. 170!.
5(T 2T )/R . It was shown that if the total UA and R

Q
i
H
L
i
i
compete against one another in the cost constraint
p c UA1p r R i 5K,
~61!
where p c and p r are the unit costs of conductance and resistance, then the optimal way of allocating the cost is
p c ~ UA ! opt5p r R i,opt .
~62!
The model of Fig. 27 was used later in the physics literature

by Gordon and Huleihil,212 Gordon and Orlov,213 and Pathria
et al.214
The technological applicability of the modeling features
presented so far in this section varies. The practical implications of results such as Eqs. ~53! and ~62! are clear: the heat
exchanger inventory and the total cost must be divided in
certain ways. The practical meaning of the optimal temperature staging ~60! is less clear. Equation ~60! means that there
is one set of optimal internal temperatures (T HC,opt ,T LC,opt)
for each set of specified heat exchanger sizes. The message
to the designer is that the working fluid must be selected in
such a way that it can be heated while at a certain temperature ~e.g., boil at T HC,opt in a simple ideal Rankine cycle!,
and be cooled while at another optimal temperature ~e.g.,
condense at T LC,opt! for each given pair of heat exchanger
sizes, U H A H and U L A L . The designer is considerably less
free to play around with the fluid type ~e.g., to abandon the
use of water! than to divide the UA inventory. This reality,
the author believes, explains at least in part why
Chambadals2 4 and Novikovs57,200203 optimal boiler temperature ~44! was not noted more in engineering. Another
reason is that large scale power plants are designed ~opti ,
mized! per unit of rate of fuel consumption @e.g., fixed Q
H
Eq. ~56!#, not with a freely varying ~infinitely abundant! heat
input.
Another way to summarize the modeling features reviewed in this section is by comparing the reported efficiencies of actual power plants ~the dots in Fig. 28! with the
CNCA efficiency ~45!. The efficiency data are based on a
compilation started in Ref. 8 and continued in Refs. 10 and
14. The agreement between the data and Eq. ~45! tends to
suggest that the irreversible nature of power plants is captured by simple models with finite heat exchangers ~e.g.,
1210
h 512
TL
TH
11ln
.
TH
TL
~63!
The agreement between Eq. ~63! and the efficiencies reported in Fig. 28 suggests that an actual power plant may
also be viewed as an obstacle to direct heat transfer from the
heat source T H to the heat sink T L , i.e., as an insulation
designed to produce maximum power when its size is constrained. A third alternative to explaining the position of the
reported efficiencies is provided by the fixed-heat-input
model: Equation ~56! agrees well with all the plotted h data
/(UAT ) has a value of order
if the dimensionless group Q
H
H
0.1. The constancy of this group makes sense because both
and UA scale with the overall size of the power plant.
Q
H
Two notes on the history of the discipline of heat transfer
can be made at this point. The assumed proportionality between convective heat transfer rate and temperature difference, e.g.,
5U A ~ T 2T !
Q
H
H H
H
H1
~64!
in Fig. 26, in the physics literature is sometimes referred to

as Newtons law of cooling. A study of the original writings shows that there is no published basis on which to attribute such a formula to Newton ~Ref. 65!. Newton stated
that the cooling rate of a body ~i.e., the derivative dT/dt! is
proportional to the temperature difference between the body
and the ambient. To use an analytical statement such as Eq.
~64!, Newton would have needed the concepts of quantity of
FIG. 29. ~a! Power plant model with bypass heat leak to the ambient, and
~b! the optimization of the heat transfer interaction between the bypass
conductance and the power-producing compartment ~From Ref. 170!.
heat, specific heat and convective heat transfer coefficient,

which were conceived roughly a hundred years later by
Black,14 Wilcke,14 and Fourier,65 respectively. Proportionalities such as Eq. ~64! are the invention of Fourier, who defined in this way the convective ~fluid flow! heat transfer
coefficient as a concept distinct from that of thermal conductivity. To his contemporaries and the subsequent development of heat transfer science, Fouriers heat transfer coefficient was revolutionary: it is an important reason why he,
and not Biot, won the race for the development of a successful analytical theory of conduction heat transfer.
The second observation concerns the nonlinear alternative to Eq. ~64!, which in the physics literature is sometimes
referred to as the Dulong & Petit law. This terminology
belongs to the mid-1800s, when it was known empirically
that the heat transfer relation can be more complicated than
in Eq. ~64!. The readers may be interested to know that modern heat transfer is now a mature science, which is capable of
anticipating the nonlinear heat transfer rates based on highly
successful theories of forced convection, natural convection,
radiation, mixed convection, and conjugate heat transfer. The
predictive powers of heat transfer science were developed
over the past 200 years on the back of analytical advances
made, chronologically, in heat conduction, hydrodynamics,
thermal radiation, aerodynamics, boundary layer theory, and
convection in porous media. The most recent overviews of
the current status of heat transfer science are presented in
Refs. 65, 157, and 215.
The work that has been published on the thermodynamic
optimization of power plant models is sizable. As we saw in
Fig. 23, Novikovs model included the effect of irreversible
expansion through the turbine of the steam cycle. Lu,216
Grazzini,208 Ibrahim et al.,217,218 and Wu and Kiang219 extended this model by also accounting for the irreversibility of
the compression process and for the fact that the temperature
of the working fluid changes along the two heat exchangers.
Petrescu et al.220 generalized the model by including the effect of piston speed and type of working fluid.
The model of Fig. 27 was extended by Swanson221 to
account for the capacity flow rates and effectivenessN tu
relations of the heat exchangers. Swanson showed that in this
case the optimal thermal conductance allocation ratio depends on the total number of heat transfer units of the two
heat exchangers. A similar model was optimized by Ibrahim
and Klein222 and Lee and Kim,223,224 who extended it further
to the optimization of Lorentz cycles, and by Ibrahim
et al.,217 who also considered the life-cycle economic optimization of the power plant.
The four-process ~two-stroke! model of Curzon and
Ahlborn8 and its steady-state counterpart ~Fig. 25! was pursued along several lines. These were reviewed most recently
by Wu et al.,225 Feidt et al.,226 and Arias-Hernandez and
Angulo-Brown,227 to whom the reader is directed. For example, in several studies beginning with Gutkowicz-Krusin
et al.,228 the assumption that the heat transfer rates are proportional to the local temperature differences was replaced
with more general, nonlinear heat transfer models that account for natural convection, radiation, and temperature dependent properties.229231 The maximum power efficiency
formula ~45! does not hold when the heat transfer model is
not linear. Early studies were also contributed by Rubin,232
Lucca,233 Rozonoer and Tsirlin,234 and Mozurkewich and
Berry.235 The effect of speed on optimal performance was
studied by Spence and Harrison236 and Petrescu et al.220
There is an important technological issue to consider in
connection with the two-stroke model inspired by Ref. 8. In
Sadi Carnots 1824 essay we were told of a gas contained in
a cylinder and piston apparatus that underwent a cycle composed of two strokes and four processes: two quasistatic and
isothermal processes interspaced with two quasistatic and
adiabatic processes. Curzon and Ahlborn8 added finite thermal resistances between the cylinder and the respective temperature reservoirs, and, in this way, described and optimized
the time-dependent evolution of the cycle. Although the
cycle described by Curzon and Ahlborn is a good instrument
for teaching, it is a questionable roadmap to improvements in
the thermodynamic performance of real heat engines. Recall
that the maximized power output can be further increased by
increasing the thermal conductances associated with the isothermal processes. Can this be accomplished in a real heat
engine in which the same cylinder wall is asked to be a
perfect insulator during one process and a very good thermal
conductor during the next process in the same stroke? The
engine builders have faced this question early in the development of practical machines. Examples are Watts 1769
separate condenser, Braytons 1873 external combustion
chamber, and Otto and Langens 1876 internal combustion
engine ~Ref. 14!.
Another line of research focused on individual features
of the four-process model. Band et al.237 performed the optimization of the heating process undergone by a fluid in a
piston and cylinder apparatus. Richter and Ross238 and
Fairen and Ross239,240 considered the effect of timedependent operation and inertia. Orlov and Berry241 optimized an engine model where the working fluid is nonisothermal and viscous ~with pressure drop! while in contact
with the heat reservoirs. The mechanical optimization of the
kinematics of engines is an interesting direction defined by
the work of Senft.242244 Related to this is the cylindroids
rotary engine of Vargas and Florea.245
The maximization of work output as opposed to power
output was pursued by Grazzini and Gori246 and Wu et al.225
Subtle differences between the maximum power in timedependent ~reciprocating! versus steady-flow power plant
models were clarified by Kiang and Wu.247 As an ecological
figure of merit in power plant optimization, Angulo-Brown9
2T S , where S is
proposed to maximize the function W
L gen
gen
the entropy generation rate of the power plant and T L is the
heat sink temperature.
Nomenclature innovations included the introduction of
the term endoreversible ~Rubin!232 to describe the reversibility of the innermost compartment such as the one shown
in Fig. 25, or alternatively, the term exoirreversible for the
external irreversibilities that surround the same compartment
~Radcenco!.248 As pointed out by Berg,249 the concept of
internal reversibility ~or external irreversibility! is basically
the same as the local thermodynamic equilibrium model that
serves as foundation for all modern heat transfer and fluid
1211
mechanics. The term thermodynamics in finite time was

introduced in 1977 specifically for the optimization of thermodynamic processes subjected to time constraints.250
Alefeld251 analyzed an entire steam-cycle modern power
station, and showed how the entropy generation rates contributed by the components can be summed up to evaluate
the overall performance of the power plant. A similar procedure was illustrated using the simple Rankine cycle in Ref.
14. The Rankine cycle power plant was also analyzed and
optimized by Wilson and Radwan,252 Roche,253 Habib and
Zubair,254 Smith,255 and Radcenco et al.256 Several of these
studies emphasized the importance of matching the variable
temperature of the working fluid to the temperature of the
heating agent. The graphic presentation and optimization of
complex power cycles was illustrated by Jin and Ishida.257
A comprehensive treatment of the distribution of sources
of entropy generation in a gas-turbine power plant was
present by Denton.258 The effect of turbine blade cooling on
entropy generation was investigated by Farina and
Donatini.259 Fundamental studies of power maximization in
simple Brayton ~Joule! cycles were conducted by Leff,260
Landsberg and Leff,261 Wu,262 as well as in Refs. 10 and 14.
Organ263 presented a detailed analysis of the distribution
of entropy generation in a Stirling-cycle power plant. This
topic and the maximization of power received considerable
attention in subsequent papers by Organ,264 Organ and
Jung,265 Radcenco et al.,256 Ladas and Ibrahim,266 and Blank
et al.267 Reviews of this field were published by Organ,268
Reader269 and, in a recent book, by Organ.270
A systematic treatment of the Otto, Brayton ~Joule!, Diesel, and Atkinson cycles was made by Leff260 and, in a generalized form, by Landsberg and Leff.261 The Diesel cycle
was optimized for maximum power by Hoffman et al.,271
and the distribution of entropy generation was studied by
Primus and Flynn.272 Papers on Otto-cycle power plants
were written by Mozurkewich and Berry235 and AnguloBrown et al.273 Internal combustion power plants were also
optimized by Orlov and Berry:274 potentially important in
practice is the fact that optimized stroke-by-stroke cycles
such as the Otto cycle require optimal time-dependent piston
motions, which, in turn, require optimal kinematics ~linkages, shapes! between piston and crankshaft.
Ocean thermal energy conversion ~OTEC! power plants
were optimized by Johnson275 and Wu.276 The MHD power
cycle was treated by Aydin and Yavuz277 and Human.278 The
on and off operation of power plants that have to be shut
down to have their heat exchangers cleaned ~defouled! was
optimized in Ref. 279. An example of the model of timedependent operation is shown in Fig. 30, where fouling is
assumed to occur on the surfaces of the hot-end heat exchanger. The thickness of the scale increases with time, d(t).
It was shown that the period-averaged power output is maximized when the on time interval has an optimal duration.
The optimal design conditions are reported in dimensionless
charts that also hold for power plants in which fouling occurs
on the surfaces of the cold-end heat exchanger.
1212
FIG. 30. Model of power plant with on and off operation and scale ~fouling!
accumulating on the hot-end heat exchanger surface ~from Ref. 197!.
VIII. REFRIGERATION PLANTS
The modeling features used for power plants ~Secs. VI

and VII! have also been used in the optimization of refrigeration plants. This body of work parallels in several respects
the work done on power plants, therefore its review in this
article will be more brief. It is interesting that by ending this
article with a second look at refrigeration we are in fact
completing a circle that started in cryogenic engineering
~Sec. III!. This is another indicator of how established the
method is, and how well rounded the field has become.
The model that was studied the most is the refrigerator
composed of a cold-end heat exchanger ~e.g., evaporator!, a
reversible compartment that receives power and moves the
entropy stream toward higher temperatures, and a roomtemperature heat exchanger ~e.g., condenser!. This model is
shown in Fig. 31~b!. It was used first by Andresen et al.207
who focused on the optimal temperature staging of the three
FIG. 31. ~a! Actual steady-state refrigeration plant, and ~b! model with two
finite size heat exchangers ~from Refs. 1 and 209!.
FIG. 32. Refrigerator model with bypass heat leak and two finite-size heat
exchangers ~from Ref. 281!.
compartments: unlike in the power-plant equivalent ~Fig.

25!, in a refrigerator there is no optimal T HC /T LC ratio for
minimum power input.
The model of Fig. 31~b! was proposed independently in
Ref. 1 for determining the optimal allocation of a finite heat
exchanger inventory. This topic was pursued along several
lines in the literature. In Ref. 1, the initial question was
where to invest an additional unit of heat transfer area (DA)
when the cold-end conductance U L A L and room temperature
conductance U H A H are given. Goth and Feidt280 assumed
that the total heat transfer area A is constrained according to
Eq. ~54!, and showed that the power input is minimum when
A is divided as shown in Eq. ~55!. An independent study281
was based on the optimization based on the UA constraint
~52!, and led to the conclusion that the thermal conductance
inventory must be split evenly between the two heat exchangers, cf. Eq. ~53!. The same optimization rule applies
when the UA inventory is minimized subject to fixed power
input.209,282
Of technological importance is that compact design rules
such as Eqs. ~53! and ~54! lead to savings in power input,
and that they apply to power plants as well. The model of
Fig. 31~b! was also analyzed by Yan and Chen283,284 and
Grazzini.285 The optimal allocation of either A or UA in a
modern defrosting refrigerator based on the vapor compression cycle with fluids R-12 or R-134a is documented numerically in Ref. 286.
A model that combines the heat exchanger irreversibilities of Fig. 31~b! with the heat leak irreversibility of Fig. 3
was proposed in Ref. 281 ~Fig. 32!. The bypass heat leak was
modeled as proportional to the room-load temperature differ 5(T 2T )/R . There are two degrees of freedom
ence, Q
i
H
L
i
, the alloin the maximization of the refrigeration effect Q
L
cation of the heat exchanger inventory between the two ends
of the machine, and the allocation of cost between the total
heat exchanger inventory and the thermal insulation. Specifically, when the heat transfer rates are modeled as
5U A (T 2T ), and
5U A (T 2T ) and Q
Q
HC
H H
HC
H
LC
L L
L
LC
the heat exchanger inventory is constrained according to Eq.
~52!, the optimal UA allocation rule is Eq. ~53!. If the thermal conductance inventory UA and the thermal resistance of
the insulation R i compete in a cost formula that is the same

as Eq. ~61!, then the optimal cost allocation rule is given by
Eq. ~62!. We see again the power of compact design optimization results such as Eqs. ~53! and ~62!: they apply across
the board, to power plants and refrigeration plants. The
model of Fig. 32 was used subsequently by Agrawal and
Menon287 and Chen.288 The optimal allocation of heat exchanger inventory subject to a constraint that is qualitatively
related to Eq. ~52! was documented by Klein.289
Another practical merit of this modeling method is with
respect to correlating and explaining the trends in the data
reported on the performance of existing refrigeration plants.
To correlate the data is important because correlations are
needed for making projections on the refrigeration needs of
future large scale projects ~e.g., Ref. 40!. This line or work
was pursued in Refs. 14 and 281 for low temperature refrigerators, and in Ref. 290 for near ambient refrigerators and
heat pumps ~chillers!.
The power of the method can be seen by examining the
cloud of performance data of actual refrigerators,291 Fig.
33. The data cover refrigerators in the temperature range 1.8
,106 W.
K,T L ,90 K and capacity ~size! range 0.1 W,Q
L
The ordinate shows the reported second-law efficiencies
. A three-line analysis
hII5COP/~COP!rev where COP5Q L /W
of the model of Fig. 34 predicts that hII should vary as14,281
C iT H
h II5 11
QL
12
TL
TH
DG
21
~65!
where C i is the thermal conductance of the leaky insulation,

5C (T 2T ). A comparison with Fig. 33 shows that Eq.
Q
L
i
H
L
;5 is an adequate curve fit for the re~65! with C i T H /Q
L
ported performance data. Equation ~65! predicts several of
the observed trends. First, hII decreases monotonically as C i
increases, which should be expected because leakier refrigerated enclosures make less efficient refrigeration plants.
Second, hII decreases as T L /T H decreases. Third, the second increases, which
law efficiency ~or the COP! increases as Q
L
means that larger machines are more efficient. Finally, the
fact that the empirical data of Fig. 33 are fitted by
;5 means that in modern low temperature refrigC i T H /Q
L
erators the heat leak is of the order of five times the actual
refrigeration load. Some of these trends were also anticipated
by Gordon and Ng290 and Ref. 281 based on more complex
models such as Fig. 32.
Other aspects of improving the thermodynamic performance of refrigeration machines are discussed by
Alefeld.292,293 In contrast to using exergy analysis, Alefeld
describes the effectiveness of components in terms of their
respective contributions to the total entropy generation rate.
The optimal distribution of a finite amount of insulation over
a cold enclosure of nonuniform temperature was developed
in Ref. 294. The application of this insulation allocation principle to a modern domestic refrigerator with two temperature
compartments ~fridge and freezer! is described in Ref. 295.
Even closer to modern refrigeration technology is the optimization of time-dependent refrigerators that must be defrosted periodically. The model of Fig. 35 was sufficient for
pinpointing an optimal rhythm of on and off operation when
the growth of the frost layer thickness d(t) is known. OptiAppl. Phys. Rev.: Adrian Bejan
1213
FIG. 33. Compilation of the second-law efficiencies of existing refrigerators and liquifiers ~from Ref. 291!.
mization results that are directly applicable to current design

and manufacture are presented in Ref. 286 for refrigerators
based on the vapor compression cycle. These results cover
not only the optimal control of the defrosting cycle but also
the optimal allocation of a finite heat transfer area to the
evaporator and the condenser.
The formation of frost on cold heat exchangers is detrimental to refrigerator thermodynamic performance, which is
why the optimization of the freezing and frost removal cycle
is an important technological issue. Most interesting is that
the same phenomenonthe same on and off cyclecan be
viewed and optimized as something beneficial, namely, the
maximization of ice production in industrial ice making installations. Models of the same class as Fig. 35 have been
used to determine the optimal freezing and removal fre-
quency for the manufacture of ice on the outside of horizontal tubes296 and on the inside or outside of vertical surfaces
of several shapes.297 The analyses also require the use of
contact melting theory.298 This optimization principle is relevant to the production by intermittent solidification of other
materials, not just ice.
FIG. 34. Model of refrigeration plant with heat leak irreversibility ~from
Refs. 14 and 281!.
FIG. 35. Model of refrigerator with unsteady operation and frost accumulation on the evaporator surface ~from Ref. 197!.
1214
IX. CONCLUSION
A revolution is taking place in thermodynamics, and it

amounts to the bridging of the gaps between thermodynamics, heat transfer and fluid mechanics. Today a common
methodology unites these seemingly separate disciplines,
and, when viewed as self-standing, the method and its various models are surprisingly simple. This review showed that
new contributions are being added at an accelerated pace in
opportunity to distribute the hardware optimally in the design of actual installations. The contribution made by simple
models and the method of entropy generation minimization
is to show the way, i.e., to uncover new opportunities for the
work that will follow in industrial research and development.
ACKNOWLEDGMENTS
I want to thank the Editors, Dr. J. M. Poate, and Dr. S. J.

Rothman, for inviting me to prepare this review article. I also
thank the National Science Foundation for supporting my
current research on basic thermodynamics, heat transfer, and
fluid mechanics.
A. Bejan, Entropy Generation through Heat and Fluid Flow ~Wiley, New
York, 1982!.
2
P. Chambadal, Les Centrales Nucleaires ~Armand Colin, Paris, 1957!, pp.
4158.
3
P. Chambadal, Rev. Gen. Elect. 67, 332 ~1958!.
4
P. Chambadal, Evolution et Applications du Concept dEntropie ~Dunod,
Paris, 1963!, Sec. 30.
5
I. I. Novikov, Atomnaya Energiya 3, 409 ~1957!.
6
I. I. Novikov, J. Nucl. Energy II 7, 125 ~1958!.
7
M. P. Vukalovich and I. I. Novikov, Thermodynamics ~Mashinostroenie,
Moscow, 1972!.
8
F. L. Curzon and B. Ahlborn, Am. J. Phys. 43, 22 ~1975!.
9
F. Angulo-Brown, J. Appl. Phys. 69, 7465 ~1991!.
10
A. Bejan, Int. J. Heat Mass Transfer 31, 1211 ~1988!.
11
A. Bejan, Thermodynamic Design, Course ME 551 ~University of Colorado, Boulder, 1979!.
12
A. Bejan, Solutions Manual for Entropy Generation through Heat and
Fluid Flow ~Wiley, New York, 1984!.
13
A. Bejan, Adv. Heat Transfer 15, 1 ~1982!.
14
A. Bejan, Advanced Engineering Thermodynamics ~Wiley, New York,
1988!.
15
G. Gouy, J. Phys. 8, 501 ~1889!.
16
A. Stodola, Steam and Gas Turbines ~McGraw-Hill, New York, 1910!.
17
M. J. Moran, Availability Analysis: A Guide to Efficient Energy Use, 2nd
ed. ~ASME, New York, 1989!.
18
A Future for Energy, edited by S. S. Stecco and M. J. Moran ~Pergamon,
Oxford, UK, 1990!.
19
Analysis of Thermal and Energy Systems, Proceedings of the International
Conference, edited by D. A. Kouremenos, G. Tsatsaronis, and C. D. Rakopoulos ~Greg. Foundas, Athens, Greece, 1991!.
20
Energy for the Transition Age, edited by S. S. Stecco and M. J. Moran
~Nova Science, New York, 1992!.
21
ECOS 92, Proceedings of the International Symposium on Efficiency,
Costs, Optimization and Simulation of Energy Systems, Zaragoza, Spain,
edited by A. Valero and G. Tsatsaronis ~ASME, New York, 1992!.
22
M. J. Moran and E. Sciubba, J. Eng. Gas Turbines Power 116, 285 ~1994!.
23
A. Bejan, G. Tsatsaronis, and M. Moran, Thermal Design and Optimization ~Wiley, New York, 1996!.
24
A. Bejan and J. L. Smith, Jr., Cryogenics 14, 158 ~1974!.
25
A. Bejan and J. L. Smith, Jr., Adv. Cryogen. Eng. 21, 247 ~1975!.
26
M. A. Hilal and R. W. Boom, Adv. Cryogen. Eng. 22, 224 ~1977!.
27
G. Bisio, Proc. 23rd Intersoc. Energy Conv. Eng. Conf. 4, 67 ~1988!.
28
G. Bisio, Proc. 24th Intersoc. Energy Conv. Eng. Conf. 1713 ~1989!.
29
A. Bejan, Cryogenics 15, 290 ~1975!.
30
V. S. Martynovskii, V. T. Cheilyakh, and T. N. Shnaid, Energ. Transp. 2
~1971!.
31
Y. M. Eyssa and O. Okasha, Cryogenics 18, 305 ~1978!.
32
J. C. Chato and J. M. Khodadadi, J. Heat Transfer 106, 871 ~1984!.
33
W. E. W. Chen, J. R. Turner, and T. H. K. Frederking, AIChE Symp. Ser.
251 82, 101 ~1986!.
34
R. Agsten, Cryogenics 13, 141 ~1973!.
35
A. Bejan, Ph.D. thesis, Massachusetts Institute of Technology, Cambridge,
MA, 1975.
36
A. Bejan and E. M. Cluss, Jr., Cryogenics 16, 515 ~1976!.
37
F. Angulo-Brown, E. Yepez, and R. Zamorano-Ulloa, Phys. Lett. A 183,
431 ~1993!.
38
1
FIG. 36. The diversity and structure of the field of entropy generation minimization ~finite time thermodynamics, or thermodynamic optimization!
~from Ref. 197!.
both engineering and physics. The contributors are almost as

diverse as the topics: engineers, physicists, chemists, and optimal control scientists. My hope299 is that this review will
lead to more effective interactions between the various sectors of the field, and, certainly, to a broader and more accurate view of the field.
A birds-eye-view of the territory we just covered is
shown in Fig. 36. The field has a two-dimensional structure
with specific applications added steadily on the vertical, and
with the optimization level indicated on the horizontal. The
optimization can be approached from left, by focusing from
the start on the total system and dividing it into compartments that account for identifiable irreversibilities. In this
case, success depends greatly on the analysts intuition ~feel
for irreversibility sources!: this stresses the importance of
teaching the entropy generation minimization method as an
integral part of thermodynamics. The optimization can also
be approached from the right in Fig. 36, by recognizing that
systems are made of actual components, that each component may contain a large number of one or more elemental
features, and that each elemental feature owes its irreversibility to processes that occur at the differential level.
The present review emphasized the fundamental and
technological implications of the optimal results delivered by
the method. At a fundamental and pedagogical level, the
method EGM is changing thermodynamics itself, i.e., the
way in which we think and apply thermodynamic principles.
For additional review material, the reader is directed to the
volume of papers edited by Sieniutycz and Salamon,300 the
books of De Vos,160 Feidt,301 and Radcenco,302 as well as to
the books 1, 14, 23, and 197.
On the practical side, we learned repeatedly that a fixed
inventory of hardware can be divided optimally among two
or more components of an installation. If this can be done in
models as simple as Figs. 27 and 32, then there is a real
1215

A. Bejan and M. O. Hoenig, IEEE Trans. Magn. MAG-13, 686 ~1977!.
41
J. Anderson, R. J. Krane, J. R. Parsons, and A. Bar-Cohen, Proc. Intersoc.
Thermal Phenom. Electronic Syst., I-THERM III, Austin, TX, 5 8 February 1992 ~unpublished!, p. 85.
42
A. Bejan and G. A. Ledezma, Int. J. Heat Mass Transfer ~in press!.
43
44
S. Jeong and J. L. Smith, Jr., Cryogenics 34, 929 ~1994!.
45
P. Grassmann and J. Kopp, Kaltetechnik 9, 306 ~1957!.
46
W. Schultz and A. Bejan, Int. J. Heat Mass Transfer 26, 335 ~1983!.
47
48
A. Bejan, Energy 5, 721 ~1980!.
49
V. A. Mironova, A. M. Tsirlin, V. A. Kazakov, and R. S. Berry, J. Appl.
Phys. 76, 629 ~1994!.
50
A. Bejan and P. A. Pfister, Jr., Lett. Heat Mass Transfer 7, 97 ~1980!.
51
W. R. Ouellette and A. Bejan, Energy 5, 587 ~1980!.
52
R. M. Nelson, ASME AES 7, 59 ~1988!.
53
D. Sekulic, A. Campo, and J. C. Morales, ASME paper no. 94-WA/HT-24
~1994!.
54
H. Perez-Blanco, ASME HTD 33, 19 ~1984!.
55
V. D. Zimparov and N. L. Vulchanov, Int. J. Heat Mass Transfer 37, 1807
~1994!.
56
B. H. Chen and W. H. Huang, Int. Commun. Heat Mass Transfer 15, 59
~1988!.
57
R. C. Prasad and J. Shen, Int. J. Heat Mass Transfer 36, 4193 ~1993!.
58
R. C. Prasad and J. Shen, Int. J. Heat Mass Transfer 37, 2297 ~1994!.
59
C. A. Balaras, Energy 15, 899 ~1990!.
60
W. F. Kenney, Chem. Eng. Progr. February, 57 ~1989!.
61
F. Kreith and M. Bohn, Principles of Heat Transfer ~Harper and Row,
New York, 1986!.
62
D. Poulikakos and A. Bejan, J. Heat Transfer 104, 616 ~1982!.
63
D. Poulikakos and J. M. Johnson, Energy 14, 67 ~1989!.
64
A. J. Fowler and A. Bejan, Int. Commun. Heat Mass Transfer 21, 17
~1994!.
65
A. Bejan, Heat Transfer ~Wiley, New York, 1993!.
66
A. Bejan, J. Heat Transfer 99, 374 ~1977!.
67
D. P. Sekulic and C. V. Herman, Int. Commun. Heat Mass Transfer 13, 23
~1986!.
68
S. Sarangi and K. Chowdhury, Cryogenics 22, 63 ~1982!.
69
C. E. S. M. da Costa and F. E. M. Saboya, Proceedings of the Eighth
Brazilian Congress on Mechanical Engineering, 1985 ~unpublished!, p.
185.
70
D. P. Sekulic, Faculty of Technical Sciences 16, University of Novi Sad,
Yugoslavia, 19851986.
71
D. P. Sekulic, Heat Transfer Eng. 7, 83 ~1986!.
72
A. L. London and R. K. Shah, Heat Transfer Eng. 4, 59 ~1983!.
73
S. M. Zubair, P. V. Kadaba, and R. B. Evans, J. Heat Transfer 109, 287
~1987!.
74
S. C. Lau, K. Annamalai, and S. V. Shelton, J. Energy Resources Technol.
109, 90 ~1987!.
75
S. Y. Huang, VDI-Ber. 539, 623 ~1984!.
76
G. Grazzini and F. Gori, Int. J. Heat Mass Transfer 31, 2547 ~1988!.
77
D. Tondeur, in Ref. 300, pp. 175208.
78
D. Tondeur and E. Kvaalen, Ind. Eng. Chem. Res. 26, 50 ~1987!.
79
B. Baclic and D. Sekulic, Termotechnika 4, 34 ~1978!.
80
H. Soumerai, ASME HTD 33, 11 ~1984!.
81
H. Soumerai, Practical Thermodynamic Tools for Heat Exchangers Design Engineers ~Wiley, New York, 1987!.
82
W. J. Wepfer, R. A. Gaggioli, and E. F. Obert, J. Eng. Ind. 101, 427
~1979!.
83
Y. El-Sayed, ASME HTD 97, 19 ~1988!.
84
H. Liang and T. H. Kuehn, ASME AES 6, 31 ~1988!.
85
R. B. Evans and M. R. von Spakovsky, ASME AES 7, 1 ~1988!.
86
M. K. Drost and J. R. Zaworski, ASME AES 4, 7 ~1988!.
87
88
S. Paoletti, F. Rispoli, and E. Sciubba, ASME AES 10-2, 21 ~1989!.
89
M. D. White and M. K. Drost, ASME AES 10-2, 13 ~1989!.
90
C.-H. Cheng and W.-H. Huang, Appl. Energy 32, 241 ~1989!.
91
M. K. Drost and M. D. White, Report No. PNL-7519/UC-363, Battelle,
Pacific Northwest Lab., Richland, WA, 1991.
92
M. K. Drost and M. D. White, J. Heat Transfer 113, 823 ~1991!.
93
P. Benedetti and E. Sciubba, ASME HTD 266, 81 ~1993!.
94
C.-H. Cheng, W.-P. Ma, and W.-H. Huang, Int. Commun. Heat Mass
Transfer 21, 519 ~1994!.
S. Petrescu, Int. J. Heat Mass Transfer 37, 1283 ~1994!.

C. F. Tapia and M. J. Moran, ASME AES 2-1, 93 ~1986!.
97
J. Ranasinghe, S. Aceves-Saborio, and G. M. Reistad, in Second Law
Analysis of Thermal Systems, edited by M. J. Moran and E. Sciubba
~ASME, New York, 1987!, pp. 2938.
98
D. Li, D. S. Scott, and P. Crane, ASME HTD 266, 93 ~1993!.
99
K. Hesselmann, ASME HTD 33, 95 ~1984!.
100
J. C. Chato and C. Damianides, Int. J. Heat Mass Transfer 29, 1079
~1986!.
101
D. P. Sekulic and Z. S. Milosevic, ASME HTD 97, 49 ~1988!.
102
Y. Tsujikawa, T. Sawada, T. Morimoto, and K. Murata, Heat Recovery
Syst. 3, 245 ~1986!.
103
R. A. Hutchinson and S. E. Lyke, Proc. 1987 ASME/JSME Thermal Eng.
Joint Conf. 2, 653 ~1987!.
104
K. Matsumoto and M. Shiino, Cryogenics 29, 888 ~1989!.
105
C. Schenone, L. Tagliafico, and G. Tanda, Heat Transfer Eng. 12, 19
~1991!.
106
G. Bisio, Warme-und Stoffubertragung 23, 143 ~1988!.
107
G. Bisio, Warme-und Stoffubertragung 25, 117 ~1990!.
108
G. Bisio, J. Non-Equilib. Thermodyn. 17, 11 ~1992!.
109
J. Bishop and V. Kinra, J. Reinforced Plastics Composites 12, 210 ~1993!.
110
J. E. Bishop and V. K. Kinra, ASME HTD 266, 127 ~1993!.
111
K. Milligan and V. Kinra, ASME HTD 266, 139 ~1993!.
112
K. Milligan and V. Kinra, Mech. Res. Commun. 20, 137 ~1993!.
113
V. Kinra and K. Milligan, J. Appl. Mech. 61, 71 ~1994!.
114
J. Y. San, W. M. Worek, and Z. Lavan, J. Heat Transfer 109, 647 ~1987!.
115
J. Y. San, W. M. Worek, and Z. Lavan, Int. J. Heat Mass Transfer 30,
1359 ~1987!.
116
C. G. Carrington and Z. F. Sun, Int. J. Heat Mass Transfer 34, 2767
~1991!.
117
C. G. Carrington and Z. F. Sun, Int. J. Heat Fluid Flow 13, 65 ~1992!.
118
Z. F. Sun and C. G. Carrington, J. Energy Res. Technol. 113, 33 ~1991!.
119
V. S. Arpaci, AIAA J. 24, 1859 ~1986!.
120
V. S. Arpaci, Int. J. Heat Mass Transfer 30, 2115 ~1987!.
121
V. S. Arpaci, Adv. Heat Transfer 21, 239 ~1991!.
122
V. S. Arpaci and A. Selamet, Combust. Flame 73, 251 ~1988!.
123
V. S. Arpaci and A. Selamet, Prog. Energy Combust. Sci. 18, 429 ~1992!.
124
A. Selamet and V. S. Arpaci, J. Thermophys. Heat Transfer 4, 404 ~1990!.
125
I. K. Puri, Int. J. Heat Mass Transfer 35, 2571 ~1992!.
126
A. Bejan, Convection Heat Transfer ~Wiley, New York, 1984!.
127
128
R. J. Krane, Int. J. Heat Mass Transfer 30, 43 ~1987!.
129
M. J. Taylor, R. J. Krane, and J. R. Parsons, in Ref. 18, pp. 885908.
130
D. P. Sekulic and R. J. Krane, in Ref. 21, pp. 6172.
131
J. Y. San, W. M. Worek, and Z. Lavan, Energy 12, 485 ~1987!.
132
C. M. Shen and W. M. Worek, Energy 18, 355 ~1993!.
133
B. Mathiprakasam and J. Beeson, Proc. AIChE Symp. Ser., National Heat
Transfer Conf. ~Seattle, WA, 1983!.
134
S. K. Das and R. K. Sahoo, Cryogenics 31, 862 ~1991!.
135
R. K. Sahoo and S. K. Das, Cryogenics 34, 475 ~1994!.
136
M. A. Badar, S. M. Zubair, and A. A. Al-Farayedhi, Energy 18, 641
~1993!.
137
T. J. Kotas and R. K. Jassim, Proceedings of the International Conference
on Energy Systems and Ecology ~ENSEC93!, Cracow, 1993 ~unpublished!, p. 313.
138
M. A. Rosen, F. C. Hooper, and L. N. Barbaris, J. Solar Energy Eng. 110,
255 ~1988!.
139
M. A. Rosen, J. Solar Energy Eng. 114, 100 ~1992!.
140
A. Bejan and W. Schultz, Int. J. Heat Mass Transfer 25, 1087 ~1982!.
141
B. Andresen and J. M. Gordon, Int. J. Heat Fluid Flow 13, 294 ~1992!.
142
B. Andresen and J. M. Gordon, J. Appl. Phys. 71, 76 ~1992!.
143
H. Bjurstrom and B. Carlsson, Heat Recovery Syst. 5, 233 ~1985!.
144
G. Adebiyi and L. D. Russell, ASME HTD 80, 9 ~1987!.
145
M. De Lucia and A. Bejan, J. Solar Energy Eng. 112, 110 ~1990!.
146
M. De Lucia and A. Bejan, J. Solar Energy Eng. 113, 2 ~1991!.
147
J. S. Lim, A. Bejan, and J. H. Kim, J. Energy Resources Technol. 114, 84
~1992!.
148
G. Adebiyi, J. Solar Energy Eng. 113, 146 ~1991!.
149
Ch. Charach and A. Zemel, J. Solar Energy Eng. 114, 93 ~1992!.
150
Ch. Charach and A. Zemel, Open Systems Information Dynamics 1, 423
~1992!.
151
Ch. Charach, J. Solar Energy Eng. 115, 240 ~1993!.
152
C. Bellecci and M. Conti, Solar Energy 53, 163 ~1994!.
39
95
40
96
1216
C. Bellecci and M. Conti, Solar Energy 51, 169 ~1993!.

C. Bellecci and M. Conti, Int. J. Heat Mass Transfer 36, 2157 ~1993!.
155
S. Aceves-Saborio, H. Nakamura, and G. M. Reistad, J. Energy Resources Technol. 116, 79 ~1994!.
156
S. Aceves-Saborio, A. Hernandez-Guerrero, and H. Nakamura, ASME
HTD 266, 73 ~1993!.
157
A. Bejan, Convection Heat Transfer, 2nd ed. ~Wiley, New York, 1995!.
158
G. A. Adebiyi, B. K. Hodge, W. G. Steele, A. Jalalzadeh, and E. C.
Nsofor, ASME AES 27, 1 ~1992!.
159
Ch. Charach and I. Rubinstein, J. Appl. Phys. 66, 4053 ~1989!.
160
A. De Vos, Endoreversible Thermodynamics of Solar Energy Conversion
~Oxford University Press, Oxford, 1992!.
161
H. Muser, Z. Phys. 148, 380 ~1957!.
162
M. Castans, Revista Geofis. 35, 227 ~1976!.
163
S. Jeter, Solar Energy 26, 231 ~1981!.
164
A. De Vos and H. Pauwels, Appl. Phys. 25, 119 ~1981!.
165
J. M. Gordon and Y. Zarmi, Am. J. Phys. 57, 995 ~1989!.
166
A. De Vos and G. Flater, Am. J. Phys. 59, 751 ~1991!.
167
A. Bejan, D. W. Kearney, and F. Kreith, J. Solar Energy Eng. 103, 23
~1981!.
168
M. Sokolov and D. Hershgal, Solar Energy 50, 507 ~1993!.
169
J. R. Howell and R. B. Bannerot, Solar Energy 19, 149 ~1977!.
170
171
A. Bejan, Int. J. Heat Fluid Flow 3, 67 ~1982!.
172
D. E. Chelghoum and A. Bejan, J. Solar Energy Eng. 107, 244 ~1985!.
173
P. Baruch and J. E. Parrott, J. Phys. D: Appl. Phys. 23, 739 ~1990!.
174
A. Bejan, J. Solar Energy Eng. 109, 46 ~1987!.
175
R. Petela, J. Heat Transfer 86, 187 ~1964!.
176
D. C. Spanner, Introduction to Thermodynamics ~Academic, London,
1964!.
177
V. Badescu, Entropie 145, 41 ~1988!.
178
V. Badescu, Energy 14, 237 ~1989!.
179
180
181
Proceedings of the International Conference on Comparative Assessments of Solar Power Technologies ~SOLCOM-1! ~Jerusalem, 14 18
February 1994! edited by A. Roy and W. Grasse ~unpublished!.
182
M. Fujiwara, J. Solar Energy Eng. 105, 163 ~1983!.
183
A. Suzuki, Doctoral thesis, Sophia University, Japan, 1989.
184
K. W. Han and C. C. Lee, Energy 16, 859 ~1991!.
185
G. Grazzini, Energie Alternative HTE 12, 325 ~1990!.
186
R. Borner, M. Feidt, and P. Ramany Bala, in Ref. 137, pp. 839 846.
187
A. De Vos, P. T. Landsberg, P. Baruch, and J. E. Parrott, J. Appl. Phys. 74,
3631 ~1993!.
188
Z. Yan and J. Chen, in FLOWERS 94, Proceedings of the Florence World
Energy Research Symposium, edited by E. Carnevalle, G. Manfrida, and
F. Martelli ~SGEditoriali, Padova, 1994!, pp. 10511057.
189
A. De Vos and P. van der Wel, J. Non-Equil. Thermodyn. 17, 77 ~1992!.
190
S. Gotkun, S. Ozkaynak, and H. Yavuz, Energy 18, 651 ~1993!.
191
K. Altfeld, W. Leiner, and M. Fiebig, Solar Energy 41, 127 ~1988!.
192
K. Altfeld, W. Leiner, and M. Fiebig, Solar Energy 41, 309 ~1988!.
193
C. K. Hsieh and C. Y. Choi, J. Solar Energy Eng. 114, 203 ~1992!.
194
S. K. Chaturvedi, T. O. Mohieldin, and D. T. Chen, Energy 16, 941
~1991!.
195
196
J. S. Lim, A. Bejan, and J. H. Kim, Int. J. Heat Fluid Flow 13, 71 ~1992!.
197
A. Bejan, Entropy Generation Minimization ~CRC, Boca Raton, FL,
1996!.
198
P. Salamon and A. Nitzan, J. Chem. Phys. 74, 3546 ~1981!.
199
H. T. Odum and R. C. Pinkerton, Am. Sci. 43, 331 ~1955!.
200
M. M. El-Wakil, Nuclear Power Engineering ~McGraw-Hill, New York,
1962!, pp. 162165.
201
M. M. El-Wakil, Nuclear Energy Conversion ~International Textbook
Company, Scranton, PA, 1971!, pp. 3135.
202
I. I. Novikov and K. D. Voskresenskii, Thermodynamics and Heat
Transfer ~Atomizdat, Moscow, 1977!.
203
I. I. Novikov, Thermodynamics ~Mashinostroenie, Moscow, 1984!.
204
A. Bejan and H. M. Paynter, Solved Problems in Thermodynamics ~Dept.
Mech. Eng., Massachusetts Institute of Technology Press, Cambridge,
MA, 1976!, Problem VII-D.
205
B. Andresen, Doctoral thesis, Physics Laboratory II, University of
Copenhagen, 1984.
206
B. Andresen, P. Salamon, and R. S. Berry, Phys. Today, September, 62
~1984!.
153
154
207
B. Andresen, P. Salamon, and R. S. Berry, J. Chem. Phys. 66, 1571

~1977!.
208
G. Grazzini, Energy 16, 747 ~1991!.
209
A. Bejan, J. Energy Res. Technol. 115, 148 ~1993!.
210
A. Bejan, Thermal Sci. Eng. ~J. Heat Transfer Soc. Japan! 33, 68 ~1994!.
211
M. H. Rubin and B. Andresen, J. Appl. Phys. 53, 1 ~1982!.
212
J. M. Gordon and M. Huleihil, J. Appl. Phys. 72, 829 ~1992!.
213
J. M. Gordon and V. N. Orlov, J. Appl. Phys. 74, 5303 ~1993!.
214
R. K. Pathria, J. D. Nulton, and P. Salamon, Am. J. Phys. 61, 916 ~1993!.
215
D. A. Nield and A. Bejan, Convection in Porous Media ~Springer, New
York, 1992!.
216
P.-C. Lu, Energy 5, 993 ~1980!.
217
O. M. Ibrahim, S. A. Klein, and J. W. Mitchell, J. Eng. Gas Turbines
Power 113, 514 ~1991!.
218
O. M. Ibrahim, S. A. Klein, and J. W. Mitchell, J. Solar Energy Eng. 114,
267 ~1992!.
219
C. Wu and R. L. Kiang, Energy 17, 1173 ~1992!.
220
S. Petrescu, C. Harman, and A. Bejan, in Ref. 188, pp. 441 452.
221
L. W. Swanson, ASME HTD 125, 31 ~1989!.
222
O. M. Ibrahim and S. A. Klein, ASME HTD 124, 91 ~1989!.
223
W. Y. Lee and S. S. Kim, Energy 16, 1051 ~1991!.
224
W. Y. Lee and S. S. Kim, Energy 17, 275 ~1992!.
225
C. Wu, R. L. Kiang, V. J. Lopardo, and G. N. Karpouzian, Int. J. Mech.
Eng. Ed. 21, 337 ~1993!.
226
M. Feidt, P. Ramany-Bala, and R. Benelmir, in Ref. 188, pp. 429 439.
227
L. A. Arias-Hernandez and F. Angulo-Brown, Rev. Mex. Fis. ~1995!.
228
D. Gutkowicz-Krusin, I. Procaccia, and J. Ross, J. Chem. Phys. 69, 3898
~1978!.
229
A. De Vos, Am. J. Phys. 53, 570 ~1985!.
230
L. Chen and Z. Yan, J. Chem. Phys. 90, 3740 ~1989!.
231
F. Angulo-Brown and R. Paez-Hernandez, J. Appl. Phys. 74, 2216 ~1993!.
232
M. H. Rubin, Phys. Rev. A 19, 1272 ~1979!.
233
G. Lucca, Cond. Aria Risc. Refrig. January, 19 ~1981!.
234
L. I. Rozonoer and A. M. Tsirlin, Avtomatika i Telemekanika 44, 50
~1983!.
235
M. Mozurkewicz and R. S. Berry, J. Appl. Phys. 53, 34 ~1982!.
236
R. D. Spence and M. J. Harrison, Am. J. Phys. 53, 890 ~1985!.
237
Y. B. Band, O. Kafri, and P. Salamon, J. Appl. Phys. 52, 3745 ~1981!.
238
P. H. Richter and J. Ross, J. Chem. Phys. 69, 5521 ~1978!.
239
V. Fairen and J. Ross, J. Chem. Phys. 75, 5485 ~1981!.
240
V. Fairen and J. Ross, J. Chem. Phys. 75, 5490 ~1981!.
241
V. N. Orlov and R. S. Berry, Phys. Rev. A 42, 7230 ~1991!.
242
J. R. Senft, J. Franklin Inst. 324, 273 ~1987!.
243
244
245
J. V. C. Vargas and R. Florea, ASME ICE 22, 215 ~1994!.
246
G. Grazzini and F. Gori, Rev. Gen. Therm. 312, 637 ~1987!.
247
R. L. Kiang and C. Wu, Int. J. Power Energy Systems 14, 68 ~1994!.
248
V. Radcenco, Rev. Chimie 41, 40 ~1990!.
249
C. Berg, in Ref. 21, pp. 716.
250
B. Andresen, R. S. Berry, A. Nitzan, and P. Salamon, Phys. Rev. A 15,
2086 ~1977!.
251
G. Alefeld, ASME AES 10-2, 61 ~1989!.
252
S. S. Wilson and M. S. Radwan, Int. J. Mech. Eng. Ed. 5, 68 ~1977!.
253
M. Roche, Rev. Gen. Thermique Fr. 260261, 561 ~1983!.
254
M. A. Habib and S. M. Zubair, Energy 17, 295 ~1992!.
255
J. L. Smith, Jr., ASME HTD 266, 181 ~1993!.
256
V. Radcenco, G. Popescu, V. Apostol, and M. Feidt, Rev Gen. Thermique
Fr. 382, 509 ~1993!.
257
H. Jin and M. Ishida, Energy 18, 615 ~1993!.
258
J. D. Denton, J. Turbomach. 115, 621 ~1993!.
259
F. Farina and F. Donatini, ASME paper no. 93-JPGC-GT 7 ~1993!.
260
H. S. Leff, Am. J. Phys. 55, 602 ~1987!.
261
P. T. Landsberg and H. S. Leff, J. Phys. A Math. Gen. 22, 4019 ~1989!.
262
C. Wu, Int. J. Energy Environment Economics 2, 57 ~1992!.
263
A. J. Organ, Proc. Inst. Mech. Eng. 201, 107 ~1987!.
264
265
A. J. Organ and P. S. Jung, Proc. Inst. Mech. Eng. 203, 301 ~1989!.
266
H. G. Ladas and O. M. Ibrahim, Energy 19, 837 ~1994!.
267
D. A. Blank, G. W. Davis, and C. Wu, Energy 19, 125 ~1994!.
268
269
G. T. Reader, ISEC, 5th Int. Stirling Engine Conf., Dubrovnik, pp. 353
359.
270
A. J. Organ, Thermodynamics and Gas Dynamics of the Stirling Cycle
1217
Machine ~Cambridge University Press, Cambridge, 1992!.

K. H. Hoffman, S. J. Watowich, and R. S. Berry, J. Appl. Phys. 58, 2125
~1985!.
272
R. J. Primus and P. F. Flynn, ASME AES 23, 61 ~1986!.
273
F. Angulo-Brown, J. Fernandez-Betanzos, and C. A. Diaz-Pico, Eur. J.
Phys. 15, 38 ~1994!.
274
V. N. Orlov and R. S. Berry, J. Appl. Phys. 74, 4317 ~1993!.
275
D. H. Johnson, Energy 8, 927 ~1983!.
276
C. Wu, Ocean Eng. 14, 349 ~1987!.
277
M. Aydin and H. Yavuz, Energy 18, 907 ~1993!.
278
M. Human, ASME AES 33, 113 ~1994!.
279
A. Bejan, J. V. C. Vargas, and J. S. Lim, Int. J. Heat Mass Transfer 37,
523 ~1994!.
280
Y. Goth and M. Feidt, C. R. Acad. Sci. Paris 303, 19 ~1986!.
281
282
O. M. Ibrahim, S. A. Klein, and J. W. Mitchell, ASME AES 24, 15
~1991!.
283
Z. Yan and J. Chen, J. Phys. D: Appl. Phys. 23, 136 ~1990!.
284
Z. Yan and J. Chen, Int. J. Energy Environment Economics 2, 63 ~1992!.
285
G. Grazzini, Int. J. Refrig. 16, 101 ~1993!.
286
V. Radcenco, J. V. C. Vargas, A. Bejan, and J. S. Lim, Int. J. Refrig. 18,
76 ~1995!.
271
1218
D. C. Agrawal and V. J. Menon, J. Appl. Phys. 74, 2153 ~1993!.

J. Chen, J. Phys. A: Math. Gen. 27, 6395 ~1994!.
289
S. A. Klein, Int. J. Refrig. 15, 181 ~1992!.
290
J. M. Gordon and K. C. Ng, J. Appl. Phys. 75, 2769 ~1994!.
291
T. R. Strobridge, NBS TN 655 ~Washington, DC, June 1974!.
292
G. Alefeld, Int. J. Refrig. 10, 331 ~1987!.
293
G. Alefeld, Newsletter IEA Heat Pump Ctr. 7, 54 ~1989!.
294
295
J. S. Lim and A. Bejan, Heat Transfer Eng. 15, 35 ~1994!.
296
J. V. C. Vargas, A. Bejan, and A. Dobrovicescu, J. Heat Transfer 116, 702
~1994!.
297
J. V. C. Vargas and A. Bejan, Int. J. Heat Mass Transfer 38, 2833 ~1995!.
298
A. Bejan, Adv. Heat Transfer 24, 1 ~1994!.
299
A. Bejan, Am. J. Phys. 62, 11 ~1994!.
300
Finite-Time Thermodynamics and Thermoeconomics, edited by S. Sieniutycz and P. Salamon ~Taylor & Francis, New York, 1990!.
301
M. Feidt, Thermodynamique et Optimisation Energetique des Systemes et
Procedes ~Technique et Documentation, Lavoisier, Paris, 1987!.
302
V. Radcenco, Generalized Thermodynamics ~Editura Tehnica, Bucharest,
1994!, in English.
287
288

Entrooy Generation Minimization

Uploaded by

Copyright:

Available Formats

Entrooy Generation Minimization

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Entrooy Generation Minimization

Uploaded by

Copyright:

Available Formats

Entropy generation minimization: The new thermodynamics of finitesize

devices and finitetime processes

Additional information on J. Appl. Phys.

APPLIED PHYSICS REVIEWS

Entropy generation minimization: The new thermodynamics

~Received 12 January 1995; accepted for publication 4 October 1995!

Entropy generation minimization ~EGM! is the method

1996 American Institute of Physics

this array of interactions is the heat transfer rate between the

can make thermodynamics and heat transfer easier to

II. THE METHOD

It is instructive to begin the review with a brief look at

fer rates (Q 0 ,Q 1 ,...,Q n ) with n11 temperature reservoirs

float as S gen varies. Historically, this choice was inspired ~and

The work transfer rate in the limit of reversible operation

J. Appl. Phys., Vol. 79, No. 3, 1 February 1996

Appl. Phys. Rev.: Adrian Bejan

In engineering thermodynamics, each of the terms on the

which for most of this century in engineering has been

The field of low temperature refrigeration was the first

where the thermal conductivity k(T) decreases toward low

According to variational calculus ~e.g., Ref. 14!, the heat

Q opt5(lk) T. The Lagrange multiplier is finally deter

Equation ~6! was provided by thermodynamics and Eq.

J. Appl. Phys., Vol. 79, No. 3, 1 February 1996

ture independent, then the optimal flow rate is

FIG. 5. Structural support cooled at several locations ~from Ref. 26!.

number of shields ~N21! is sufficiently large such that the

where DT i 5T i11 2T i , Di5(i11)2i and A, F, and s are

FIG. 7. Discrete intermediate cooling of a stack of radiation shields ~from

Appl. Phys. Rev.: Adrian Bejan

FIG. 8. The longitudinal convective heat leak carried by a counterflow heat

where L 052.4531028 ~W/A K!2 is the constant in the

The current cable with Joule heating and longitudinal

J. Appl. Phys., Vol. 79, No. 3, 1 February 1996

IV. HEAT TRANSFER

Consider first the flow of a single-phase stream m

where the Nusselt number Nu5hD/k is a nondimensional

A characteristic of all heat transfer devices with fluid

where B 0 is a heat and fluid flow duty parameter that

where ReD /ReD,opt5D opt /D because the mass flow rate is

FIG. 10. Entropy generation number N S , or relative entropy generation rate

J. Appl. Phys., Vol. 79, No. 3, 1 February 1996

the effect of the heat transfer augmentation technique on the

, T , T , F , and U are the heat transfer rate,

destroys minimum power while performing its assigned heat

and W is the plate dimension perpendicular to Fig. 11. The

The fins planted on the surfaces of heat exchanges act as

where t5u T H 2T L u /(T H T L ) 1/2, G 5G/(2 r P) 1/2, and G is

given duty, the best configuration of a heat and mass process

J. Appl. Phys., Vol. 79, No. 3, 1 February 1996

FIG. 13. Flow, temperature, and volumetric entropy generation rate in a

material that has certain, desired damping characteristics.

The optimization of time-dependent heating and cooling

Figure 14 shows that the sensible heating process has

Since in most heat exchangers N tu is of the order of 1 or

slabs of storage material. The longitudinal conduction of heat

J. Appl. Phys., Vol. 79, No. 3, 1 February 1996

where T(t) is the companion result for the optimal body

Technologically, this result is extremely valuable because it