A.

Bejan
Assistant Professor,
A Study of Entropy Generation in
Department of Mechanical Engineering,
University of Colorado,
Boulder, CO 80309
Assoc. Mem. ASME
Fundamental Conwectife Heat
Transfer
The second law aspects of heat transfer by forced convection are illustrated in terms of
four fundamental flow configurations: pipe flow, boundary layer over flat plate, single cyl-
inder in cross-flow, flow in the entrance region of a flat rectangular duct. The interplay
between irreversibility due to heat transfer along finite temperature gradients and, on
the other hand, irreversibility due to viscous effects is analyzed in detail. The spatial dis-
tribution of irreversibility, entropy generation profiles or maps, and those flow features
acting as strong sources of irreversibility are presented. It is shown how the flow geometric
parameters may be selected in order to minimize the irreversibility associated with a spe-
cific convective heat transfer process.

Introduction
Our understanding of heat transfer processes is destined to play bated the way it should be, in view of its relevance to the energy
a significant role in the effort toward workable alternatives to the conservation questions facing the engineering profession. By analyzing
growing energy problem. In particular, the task of conserving useful the irreversibility associated with heat transfer we are not including
energy rests heavily on our ability to produce thermodynamically one additional effect in an already complex heat transfer model. On
efficient heat transfer processes and equipment for such processes. the contrary, through irreversibility we are bringing out that feature
Consequently, in recent years we witnessed a growing interest in the of heat transfer present even in the simplest possible heat transfer
thermodynamics of heat transfer and the thermodynamics of heat model.
exchange equipment. This interest will continue to grow in the fu-
ture. Local Rate of Entropy Generation
Heat transfer processes are generally accompanied by thermody- Consider the two-dimensional infinitesimal fluid element dxdy
namic irreversibility or entropy generation. The generation of entropy shown schematically in Fig. 1. The fluid element is part of a consid-
may be due to a variety of sources, primarily heat transfer down erably more complex convective heat transfer picture. However, for
temperature gradients and, characteristic of convective heat transfer, the scope of this presentation, we regard the element as an open
viscous effects. There exists a direct proportionality between the ir- thermodynamic system subjected to mass fluxes, energy transfer and
reversibility (entropy generation) of the process and the amount of entropy transfer interactions through a fixed control surface. The
useful work dissipated in the process [1-3]. This relationship implies element size is small enough so that the thermodynamic state of the
that in cases where the heat transfer process is part of a power cycle, fluid inside the element may be regarded as uniform (independent
the process irreversibility causes a direct drop in the useful power of position). However, the thermodynamic state of the small fluid
output of the cycle. Conversely, should the heat transfer process be element may change with time.
part of a refrigeration cycle, the process irreversibility leads to a direct For this study, we limit our attention to incompressible fluids
increase in the mechanical power input to the cycle. Either way, the without internal heat generation. In such cases, the expression for the
irreversibility brought by the heat transfer process amounts to a volumetric rate of entropy generation reduces to [7]
penalty in useful power.
With a better understanding of how entropy is being generated in
heat transfer processes and engineering components for heat exchange
it is possible to reduce the process irreversibility, thus registering
S'" =
y2 iM- d0\2'
T levy,
dvx dvyy (1)
savings in useful (available) power. With this objective in mind, Bejan dy
showed how the entropy generation rate can systematically be reduced
dxj
in simple components for heat exchange, namely, counterflow gas-
to-gas heat exchangers [4], heat exchangers with prescribed heat flux
distribution [5] and sensible heat units for energy storage [6],
The main objective of the present article is to analyze the mecha-
nism of entropy generation in basic configurations encountered in
convective heat transfer. Unlike the earlier papers which addressed
the subject of irreversibility reduction in the design of engineering y + dy
components for heat exchange [4-6], the present work is fundamental.
In this article we seek to identify the origin of entropy production and
its distribution through fluid flows most commonly found in con-
vective heat transfer situations. In addition, we discuss the engi-
neering implications of this study, specifically, the manner in which
a basic flow geometry may be selected in order to minimize the rate
of entropy generation associated with the convective heat transfer
process.
A second objective of this study is to illustrate, in a very modest way,
the place thermodynamics duly occupies in heat transfer. It is un-
J_
fortunate that the link which exists between the two fields is not de-
x x+dx
Contributed by the Heat Transfer Division for publication in the JOURNAL
OF HEAT TRANSFER. Manuscript received by the Heat Transfer Division Fig. 1 Entropy generation analysis for an infinitesimal element dxdy in
March 8,1979. convective heat transfer

718 / VOL. 101, NOVEMBER 1979 Copyright 1979 by ASME Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/28/2013 Terms of Use: http://asme.org/terms

 As we might have expected, the irreversibility indicator S'" contains
two additive parts, one due to conduction in the presence on non-zero
temperature gradients, the other accounting for viscous dissipation
of mechanical power throughout the flow. In the second term of
equation (1), the factor in brackets represents the viscous dissipation
function $ for two-dimensional incompressible flow. The other
symbols appearing in equation (1) have been defined in the Nomen-
clature.
The local entropy generation rate formula (1) can be derived in
straightforward fashion by performing an entropy transfer accounting
around the infinitesimally small element illustrated in Fig. 1. The
entropy transfer to and from the (dxdy) system is associated with heat
transfer, qx and qy, as well as*with mass fluxes, pvx and pvy. In the
interest of brevity we do not show this derivation, urging the reader
to consult any irreversible thermodynamics book for additional de-
tails. Alternate versions of formula (1) corresponding to coordinate
systems other than the two-dimensional cartesian system of Fig. 1 may
be found in [8].
Conductive Versus Viscous Effects. In many heat transfer
problems it is often possible and convenient to neglect the viscous
dissipation term p.9 in the equation for energy conservation.
at cryogenic temperatures, the temperature difference number r is
generally much smaller than unity, T 1.
It is now clear that the energy argument by which viscous dissipa-
tion is neglected in (2); i.e., Ec Pr 1, has no bearing on the question
of negligible viscous contribution to the local rate of irreversibility
production. It is then possible to encounter situations where, although
the energy equation can be simplified according to Ec Pr 1, S'" is
in fact dominated by viscous effects. This is the limit in which T is very
small, small enough so that Ec P r / r > 1.
Below we examine a series of important convective heat transfer
configurations in an effort to illustrate the coupling of viscous and
conductive effects in the makeup of S'". In the process we will study
the spatial distribution of irreversibility, pointing out those flow
features which act as concentrators (sources) of entropy generation
S'".
Forced Convection in a Round Tube
Laminar Flow. Consider the Poiseuille flow through a round tube
with uniform heat flux q" around its circumference (see insert in the
left side of Fig. 2). The velocity and temperature profiles for this flow
are particularly simple [8]:

d8 dff d2e d26>' (5)

PCp + vv + M*- (2)
dx " dyl dx2 dy2
9"ro
4M V-
This is particularly the case in heat transfer through gases at subsonic (6)
velocities. The dimensionless group which expresses the magnitude k
of the dissipation energy term relative to the conduction energy term with xo Vol 4 \r0l \
in (2) is [9]
dP Xo rox,i ;
=Pe. (7,8)
EcPr; (3)
' 4n dx ro
30*J U The equation for S'" in the cylindrical geometry of Fig. 2 is [8]
where v* and 0* are the characteristic fluid velocity and temperature ^(M\2 2
s,,, = + (ziy] + EiduA
difference for the convective heat transfer problem. Thus, in many (9)
engineering problems, we find Ec Pr 1. T2[\dx/ Idr/J T\drj
Consider now the same question relative to expression (1). Under which, in combination with relations (5-8), yields
what conditions is the viscous dissipation contribution to S'" negli-
q"<- 16 ^tyiU a, max p22 r
gible? If we regard expression (1) as the sum S'" = S"'conductive + (2R - Ra)2 + R, R
S '"viscous, then, in an order of magnitude sense, "kT 2 Pe2 Tr02 ro
\ EcPr (10)
0 (4) Equation (10) is the entropy generation profile in the pipe cross-
, conductive/ section. Together with the velocity and temperature profiles, the
Here, r = 6*/T*, where T* is the absolute temperature characteristic entropy generation profile completes the thermodynamic description
to the problem at hand. The dimensionless temperature difference, of the convective heat transfer phenomenon.
T, is always an important dimensionless parameter in second law It is convenient to nondimensionalize expression (10) and define
analyses of heat transfer problems. With the exception of applications the local entropy generation number Ns

.Nomenclature.
a = half-thickness of flat duct, Fig. 5 Pr = Prandtl number x* = dimensionless coordinate, equation
So, B = duty parameters, equations (17) and q", q', q = heat transfer interaction, [WVm2], (51)
(37) [W/m], [W] y = vertical coordinate
cp = specific heat at constant pressure r = radial position a = thermal diffusivity
Co = drag coefficient, equation (44) ro = tube radius, Fig. 2 & = velocity boundary layer thickness
Cfx = local skin friction coefficient, equation R = dimensionless radial position &T = thermal boundary layer thickness
'(34) Re = Reynolds number f = dimensionless coordinate across flat
D = hydraulic diameter s = specific entropy duct
Ec = Eckert number, equation (3) S'", S", S' = rate of entropy generation, ?) = similarity variable in boundary layer flow
/ = function, equation (26); friction factor, [W/msK], [W/m2K], [W/mK] over flat plate
equation (14) S"FD rate of entropy generation in the 6 = temperature difference, T To
FD = drag force fully-developed region 6 = extreme tremperature difference, T
hx = local heat transfer coefficient t = time To
k = thermal conductivity T = absolute temperature fi = viscosity
LE = entrance region length To = reference temperature v = kinematic viscosity
Lpo = length of irreversibility-equivalent u = specific internal energy p = fluid density
fully-developed section, equation (58) vx, Vy = velocity components r = ratio of characteristic temperature dif-
m = mass flow rate Vo = entrance velocity, Fig. 5 fererite divided by the absolute tempera-
Ns", Ns", Ns' = entropy generation Vi = centerline velocity, Fig. 5 ture
number Vi* = dimensionless centerline velocity, T 0 = wall shear stress, equation (34)
Nu = Nusselt number equation (50) $ = viscous dissipation function, equation
P = pressure x = horizontal coordinate (2)

Journal of Heat Transfer NOVEMBER 1979, VOL. 101 / 719

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/28/2013 Terms of Use: http://asme.org/terms

 where S and &T are the velocity and temperature boundary layer yields the minimum rate of entropy generation in a heat transfer
thicknesses. In writing (31) we made use of the approximation 8/ST application in which the uniform flow velocity vx> and the total heat
Pr 1/3 . The significance of (31) is that when the two thicknesses are flux q' = foL q" dx are specified. We do this by first substituting
not equal, the thinner layer exhibits larger gradients thereby en- appropriate correlations for hx and CfiX into expression (32) and
hancing its contribution to the total S'" figure. According to (31), solving the equation Z>S'/dL = 0.
viscous effects are more likely to play a role in low Prandtl number Optimum Plate Length for Minimum Irreversibility. For
fluids where 8 5T- laminar flow, the local skin friction coefficient is C/,* = 0.664 Rex ~~1/2
Laminar and Turbulent Flow over Constant Heat Flux Plate. while the local heat transfer coefficient is given by hx x/k = 0.332 Pr1/3
To study the generation of entropy in a turbulent boundary layer one Re;t1/2 [14]. Writing q' = q" L for the total heat transfer rate from
has to rely on an integral method which takes into account the heat plate to fluid over the plate length L, the entropy generation number
transfer and fluid friction characteristics of the flow in an overall Ns' becomes
manner, as Nusselt number and friction factor information. Consider,
for example, a flat plate of negligible thickness suspended in a uniform NS' = S' ^- = 2.008 P r ^ 3 Rei,-1/2 + 0.664 Rex,1'2 B" 2 . (35)
i
flow field, parallel to the flow velocity vXi,. The heat flux q" over the q
plate surface is uniform. Consider also a control surface which sur- Here, Rez, and B are the Reynolds number based on L and the "duty"
rounds the plate of finite length L at a large enough distance through parameter, respectively,
portions of external flow in which the fluid motion is nearly uniform
and the temperature nearly constant, T. Regardless of whether the R e L = ^ , B- (36, 37)
boundary layer is turbulent or laminar, the entropy generation rate v vx,(nkT)W
in one half of the control volume (i.e., for one side of the plate) is given The optimum plate length Rez,,opt yielding the minimum rate of en-
by tropy generation at constant q' and vx is
ReL,0pt = 3.024 Pr-!/ 3 S 2 , 2.309 Pr- 1 / 6 ^- 1 (38,39)
S' dx, (32)
J o hx 2 r J o '
where hx and CfiX are the local heat transfer coefficient and skin For turbulent flow we use a similar set of correlations for friction
friction coefficient, and heat transfer, CfiX = 0.0576 Re*- 1/5 and hx x/k = 0.0296 Pr 1/3
Re*0-8 [15]. We also assume that the laminar layer which precedes the
2 TQ turbulent boundary layer is much shorter than the plate length L.
(33, 34)
T0(x) - T- pvx2
- / , * Substituting these correlations into the entropy generation result (32)
yields
In the above definitions TQ(X) is the wall temperature, TO the wall
shear stress and q" the uniform heat flux. Note that unlike equation feT2
Ns> = S' 28.15 Pr- 1/3 ReL 8/6 + 0.036 RezT8/6 B~2, (40)
(29) in the preceding sub-section where we considered the laminar
boundary layer over an isothermal flat plate, the S' expression (32) Rei.opt = 64.31 Pr- 6 ' 24 B6'\ iVsmin = 2.013 Pr- 1 ' 6 B~\ (41, 42)
refers to a uniform heat flux situation.
Expression (32) is the result of an entropy flux accounting around The optimum plate length and the resulting minimum entropy
the control volume, analysis omitted here due to space limitations. generation number prescribed by equations (38,39) and (41,42) are
Like all entropy generation results for forced convection heat transfer, shown on the left side of Fig. 4. The discontinuity illustrated by
S' consists of two additive parts, one due to heat transfer across the dashed lines corresponds to the transition region, cases in which the
[To(x) T J temperature difference, the other being associated with laminar and turbulent portions of the boundary layer are of compa-
the total friction drag force exerted by the fluid on the plate. Below, rable lengths. The trends are similar to those presented in the right
we use result (32) to determine the optimum plate length L which graph of Fig. 2: the higher the duty parameter B, the higher the op-

-,min '.:..
N
S
-

Pr-I^v^ -

10 -
10

Re L,opt -

106

R e
i .
L,opt
10*

/ / i i i

10 10* 10

Fig. 4. Optimum Reynolds number and corresponding minimum entropy generation number. Left: boundary layer flow over a flat plate. Right: single cylinder
in gaseous cross-flow

722 / VOL. 101, NOVEMBER 1979 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/28/2013 Terms of Use: http://asme.org/terms

 For laminar flow, using Nu = 48/11 and / = 16/Re in expression
(15), the minimization procedure yields

Re op t = 0; Ns>' 1 (18,19)
In engineering terms, this result implies that the selected tube radius
ro must be large enough so that the rate of entropy generation is
strongly dominated by the contribution due to heat transfer across
a finite temperature difference. In other words, based on expression
(15),
24 7T4
Re4B(T2l. (20)
11
For turbulent flow, the Ns' expression (15) has a unique minimum.
Substituting Nu = 0.023 Pr0-4 Re0-8 and / = 0.046 Re" 0 - 2 into (15) and
differentiating with respect to Re yields
Re o p t = 2.023 P r - 0 - 0 7 1 S 0 - 0 - 3 (21)
and
solid
NS-' 126 Pr" 0 - 343 B 0 (22) wall

Expressions (21, 22) have been summarized in the right hand side of
Fig. 2 for two discrete values of Prandtl number. As the aggregate duty
parameter Bo increases, we see that the optimum tube radius de-
creases (Re op t increases) and the minimum entropy generation
number Nsmm decreases also.
As a numerical example, consider the heat transfer to an air stream
at atmospheric pressure and an average temperature of 1100 K. In flow direction
order to use realistic values for q' and m in the Bo formula (17), it is
helpful to replace q' by mcp(dT/dx). For a longitudinal temperature
gradient of the order of 10 K/m, and for a mass flow of 100 Kg/hr, we
obtain B 0 = (2.6) 10 10 and, from (21), Re o p t = (1.11)104. Finally, ex- Fig. 3. Entropy generation surface lor laminar boundary layer flow and heat
transfer along a flat plate
pression (16) yields an optimum tube radius ro l0p t = 3.6 cm. Calcu-
lations of this type are relevant to the optimum thermodynamic design The complex dependence of S'" on both x and y is shown in Fig.
of heat exchanger passages with prescribed heat transfer distribution. 3. The three-dimensional display was done in terms of (x vXi/v) and
Examples of such heat exchangers are the core of a nuclear reactor (y vx^h) in the horizontal plane, and Ns/(1 + Ec P r / r ) in the ver-
and, from a recently expanding technology, a superconducting cable tical direction. It is evident that the irreversibility effects are limited
cooled with liquid helium by forced convection [10]. to the boundary layer. Regarding the y dependence oiNs", the en-
tropy generation rate is highest reaching a peak at the solid wall. The
Boundary Layer Over Flat Plate longitudinal variation of S'" is as l/x, indicating that like all gradients
Laminar Flow over Isothermal Plate. Consider now the de- in this boundary layer solution S'" blows up at the origin. The viscous
velopment of laminar momentum and thermal boundary layers along effect again scales up as Ec P r / r .
a flat plate. The situation is shown schematically in the horizontal Integrating (25) across the boundary layer we can calculate the rate
plane of the isometric drawing of Fig. 3. At some distance from the of entropy generation per unit area of flat plate,
solid wall the fluid velocity and temperature are uniform, ux? and
T. The wall temperature is constant, To.
The study of the velocity and temperature fields in the vicinity of
S">
I s . dy . 0 . 26 *!i4=![ 1 + )
a Re*- 1 / 2 (27)

the plate constituted the subject of numerous investigations [11]. The

with the corresponding entropy generation number defined as
purpose of this section is to examine the distribution of entropy
generation in the boundary layer. For this we rely on solutions avail-
Ns = S" = 0.25 (l + ^ ^ W -i/2. (28)
able in the literature for vx(x, y) and d(x, y) in laminar flow. kvx,62
The task of evaluating the entropy generation profile S'" is sim-
Finally, by analogy with the Nusselt number nomenclature for
plified greatly if we restrict the discussion to the case Pr = 1 for which
boundary layer heat transfer, we integrate (27) in the x direction to
the Blasius-Pohlhausen solution [12] reduces to
find the total rate of entropy generation produced by boundary layer
flow and heat transfer over a length L
(23, 24)

The similarity variable r\ equals y[vXia/(vx)]1/2,

dr)
while /(?)) is the
function tabulated by Howarth [13]. Combining now solution (23, 24)
with the S'" expression (1) and neglecting the irreversibility terms
*-r S *d, = 0.50 ^ ( l + ^ ) R e L

where Ret = vX: LI v. The overall entropy generation number based

I/*, (29)

on S' is
associated with velocity and temperature gradients in x direction, we
find
2
Ns> , n2 0.50 | 1 + ^ W I/2. (30)
. k 0 - ^ , ( / ) 2 + /"V 2 kdJ-
S" if") (25)
T02vx " ' T0vx It is worth mentioning that in general the Prandtl number will have
an additional effect on the relative importance of viscous and con-
The local entropy generation number is
ductive effects in the constitution of S'", S" and S'. It is easy to show

NS'
. s '" i vT \2=i t { EcPrj(/")2
(26)
that when Pr 7^ 1 the viscous effects scale up as
' k \evxJ \ T / Re* E c P r / M a _ Ec Pi1/3
(31)
where Re* is defined as vx *,xjv. T \8j T

Journal of Heat Transfer NOVEMBER 1979, VOL. 101 / 721

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/28/2013 Terms of Use: http://asme.org/terms

 where S and &T are the velocity and temperature boundary layer yields the minimum rate of entropy generation in a heat transfer
thicknesses. In writing (31) we made use of the approximation 8/ST application in which the uniform flow velocity vx> and the total heat
Pr 1/3 . The significance of (31) is that when the two thicknesses are flux q' = foL q" dx are specified. We do this by first substituting
not equal, the thinner layer exhibits larger gradients thereby en- appropriate correlations for hx and CfiX into expression (32) and
hancing its contribution to the total S'" figure. According to (31), solving the equation Z>S'/dL = 0.
viscous effects are more likely to play a role in low Prandtl number Optimum Plate Length for Minimum Irreversibility. For
fluids where 8 5T- laminar flow, the local skin friction coefficient is C/,* = 0.664 Rex ~~1/2
Laminar and Turbulent Flow over Constant Heat Flux Plate. while the local heat transfer coefficient is given by hx x/k = 0.332 Pr1/3
To study the generation of entropy in a turbulent boundary layer one Re;t1/2 [14]. Writing q' = q" L for the total heat transfer rate from
has to rely on an integral method which takes into account the heat plate to fluid over the plate length L, the entropy generation number
transfer and fluid friction characteristics of the flow in an overall Ns' becomes
manner, as Nusselt number and friction factor information. Consider,
for example, a flat plate of negligible thickness suspended in a uniform NS' = S' ^- = 2.008 P r ^ 3 Rei,-1/2 + 0.664 Rex,1'2 B" 2 . (35)
i
flow field, parallel to the flow velocity vXi,. The heat flux q" over the q
plate surface is uniform. Consider also a control surface which sur- Here, Rez, and B are the Reynolds number based on L and the "duty"
rounds the plate of finite length L at a large enough distance through parameter, respectively,
portions of external flow in which the fluid motion is nearly uniform
and the temperature nearly constant, T. Regardless of whether the R e L = ^ , B- (36, 37)
boundary layer is turbulent or laminar, the entropy generation rate v vx,(nkT)W
in one half of the control volume (i.e., for one side of the plate) is given The optimum plate length Rez,,opt yielding the minimum rate of en-
by tropy generation at constant q' and vx is
ReL,0pt = 3.024 Pr-!/ 3 S 2 , 2.309 Pr- 1 / 6 ^- 1 (38,39)
S' dx, (32)
J o hx 2 r J o '
where hx and CfiX are the local heat transfer coefficient and skin For turbulent flow we use a similar set of correlations for friction
friction coefficient, and heat transfer, CfiX = 0.0576 Re*- 1/5 and hx x/k = 0.0296 Pr 1/3
Re*0-8 [15]. We also assume that the laminar layer which precedes the
2 TQ turbulent boundary layer is much shorter than the plate length L.
(33, 34)
T0(x) - T- pvx2
- / , * Substituting these correlations into the entropy generation result (32)
yields
In the above definitions TQ(X) is the wall temperature, TO the wall
shear stress and q" the uniform heat flux. Note that unlike equation feT2
Ns> = S' 28.15 Pr- 1/3 ReL 8/6 + 0.036 RezT8/6 B~2, (40)
(29) in the preceding sub-section where we considered the laminar
boundary layer over an isothermal flat plate, the S' expression (32) Rei.opt = 64.31 Pr- 6 ' 24 B6'\ iVsmin = 2.013 Pr- 1 ' 6 B~\ (41, 42)
refers to a uniform heat flux situation.
Expression (32) is the result of an entropy flux accounting around The optimum plate length and the resulting minimum entropy
the control volume, analysis omitted here due to space limitations. generation number prescribed by equations (38,39) and (41,42) are
Like all entropy generation results for forced convection heat transfer, shown on the left side of Fig. 4. The discontinuity illustrated by
S' consists of two additive parts, one due to heat transfer across the dashed lines corresponds to the transition region, cases in which the
[To(x) T J temperature difference, the other being associated with laminar and turbulent portions of the boundary layer are of compa-
the total friction drag force exerted by the fluid on the plate. Below, rable lengths. The trends are similar to those presented in the right
we use result (32) to determine the optimum plate length L which graph of Fig. 2: the higher the duty parameter B, the higher the op-

-,min '.:..
N
S
-

Pr-I^v^ -

10 -
10

Re L,opt -

106

R e
i .
L,opt
10*

/ / i i i

10 10* 10

Fig. 4. Optimum Reynolds number and corresponding minimum entropy generation number. Left: boundary layer flow over a flat plate. Right: single cylinder
in gaseous cross-flow

722 / VOL. 101, NOVEMBER 1979 Transactions of the ASME

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/28/2013 Terms of Use: http://asme.org/terms

 timum Reynolds number (plate length) and the lower the minimum <?' 2 M i,
entropy generation number. In other words, if the total heat transfer + -CDRe: (45)
2
rate q' is constant and the flow velocity increases, the optimum plate -feT Nu 2 T
length decreases and the minimum attainable entropy generation rate In writing (45), we used the drag coefficient to replace FD in equation
increases. (43). We also wrote q' = irD q" for the total heat transfer rate per unit
These conclusions are applicable to the optimization of local flow length of cylinders. Finally, expression (45) is put in dimensionless
geometry in plate-fin surfaces for compact heat exchangers. The chief form defining the entropy generation number
conclusion is that for a plate-fin with q' and uX: fixed, there exists
k T 2 1 1
an optimum fin dimension in the flow direction which, if selected, Ns- = S' 2_ = _ + - CD Re B~* (46)
yields the minimum rate of entropy generation. q' TNU 2
where B is a duty parameter which has the same form as in equation
Single Cylinder in Cross-Flow (37).
The heat transfer between a cylindrical surface and a fluid flow Optimum Cylinder Diameter for Minimum Irreversibility.
normal to the cylinder axis is one of the most frequent heat transfer As in the heat transfer configurations examined earlier, we can use
configurations encountered in actual engineering equipment for heat the entropy generation number formula (46) to determine which flow
exchange processes. It is appropriate to examine here the thermo- geometry (cylinder diameter D) is best for minimizing the thermo-
dynamic irreversibility introduced by this configuration. Due to in- dynamic losses associated with the heat transfer process. For an ap-
herent similarities with the plate of finite length in parallel boundary plication in which q', uXi and the fluid are known, the optimization
layer flow, example concluded in the preceding section, the case of procedure amounts to minimizing expression (46) with respect to Re,
a single cylinder in cross-flow will be summarized very briefly and only subject to constant B. The results of this optimization procedure are
the key analytical results will be given. shown as Re op t and iVs' min on the right side of Fig. 4, a plot qualita-
Consider a cylinder of diameter D with uniform surface heat flux tively similar to the left graph obtained for a flat plate. The right graph
q" in a cross-flow of uniform velocity vx and temperature T . An of Fig. 4 was constructed based on equation (46) coupled with Hil-
entropy generation analysis of the flow region affected by heat and pert's average Nusselt number correlation for gas flow [16] and with
momentum transfer from the cylinder yields the total entropy gen- Eisner's presentation of drag coefficient [17].
eration rate . From a practical viewpoint, expression (46) and the optimum design
summarized on the right side of Fig. 4. can be used to calculate and
S' ;f[T()-T]d+- (43) minimize the thermodynamic losses exhibited by engineering com-
ponents which embody the simple heat transfer configuration con-
In this expression, is the curvilinear coordinate around the cylin- sidered in this section. These results can be applied not only to single
drical surface and T() is the local surface temperature. It is again cylinders in cross-flow, but also to arrays of cylinders (tube banks)
assumed that the temperature differences are much smaller than spaced sufficiently far apart so that their mutual influence is not a
characteristic absolute temperature of the medium [T() T J major effect in the heat transfer and fluid friction characteristics of
T. FD is the force (drag) per unit length exerted by flow on cylinder, the arrangement. In the case of arrays of cylinders in cross-flow in
force calculated from drag coefficient experimental information, which the spacing between neighboring cylinders plays an important
CD, role [18] it is possible to reconstruct the above analytical procedure
based on heat transfer and fluid friction information for tube banks
FD/D
2 (44) in cross-flow [19]. Further, one should be able to combine the external
pvx /2 flow configuration (flow normal to tube banks) with the internal flow
Replacing the line integral in (43) with the average wall-fluid configuration studied in the first example of this paper (pipe flow)
temperature difference times irD, and expressing the average tem- to establish the least irreversible combined geometry for two-fluid
perature difference in terms of the average Nusselt number, T()all heat exchangers using one fluid in the tubes and the other normal to
- T = q" D/(k Nu), yields, the tubes. The question of optimum pin-fin surface geometry can be

S"/S" T
2a
'i
.entrance fully-developed
region region

end of
entrance
region

_i_ _L _L
0 0.05 0.1

Fig. 5. Distribution of the entropy generation rate in the entrance region of a flat rectangular duct in laminar flow

Journal of Heat Transfer NOVEMBER 1979, VOL. 101 / 723

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/28/2013 Terms of Use: http://asme.org/terms

 addressed in a similar manner. For this one would have to use Nusselt
number and friction factor data obtained specifically for pin-fin
surfaces where, as shown recently by Sparrow and Ramsey [20], the
pin-wall attachment makes the flow strongly three-dimensional,
unlike the two-dimensional configuration considered in this sec-
tion.
L a m i n a r F o r c e d C o n v e c t i o n i n t h e E n t r a n c e R e g i o n of
a Flat Rectangular Duct
A frequent flow configuration which embodies the characteristics
of the first two flow examples discussed here is sketched in the insert
of Fig. 5. In the entrance region of a parallel-plate duct temperatjrre
and velocity boundary layers develop simultaneously along both walls,
gradually filling the duct and leading to the well-known fully-devel-
oped laminar regime. There have been many studies of the heat
transfer and velocity problems associated with this basic configura-
tion, as indicated recently by Bhatti and Savery [21]. In this section
we address the irreversibility problem, once again relying on published
solutions for the temperature and velocity field in the duct.
Sparrow [22] analyzed the boundary layer development in the en-
trance region using the Karman-Pohlhausen integral technique. The
Sparrow solution is unique, considering its simplicity versus the good
agreement between its predictions and experimental findings. In what
follows we use this solution in combination and the S'" formula (1)
to illustrate the distribution of irreversibility in the channel.
The ultimate goal of the irreversibility analysis is to discover how
the thermodynamic losses due to entrance effects compare with
similar losses in the fully-developed section of the duct. In other
words, what length of fully-developed flow is guilty of generating as
much entropy as the entire entrance region? The answer to this
question is simplified substantially if, in addition to using Sparrow's
analytical solution, we focus on the r -* 0 limit in which the viscous
effects totally dominate the rate of entropy generation in the channel
(see the discussion following equation (4)). Consequently, to calculate
S'" we need only the velocity solution which is (22)
^ = V0yx*(2f-f2) (47)
with
f = ylb, 8/a = 3(1 - Vi*" 1 ) (48, 49)
V1* = Vi/Vo,x* = J~~ (50,51)
a 2 Vo
As shown on Fig. 5, V0 is the uniform fluid velocity at x = 0 , while Vi
and 5 are the centerline velocity and boundary layer thickness, re-
spectively. For the function Vi* (x*), Sparrow reports a differential
eguation which, integrated, yields
x*= ( 9 V i * - 2 - - 1 6 ^ n V i * | (52)
101 Vi* I seen what features of the flow act as concentrators of entropy gener-
The entrance region extends to x * = 0.1038, corresponding to Vi* =
3/2 as for plane Poiseuille flow.
Using equation (1) and leaving out the conductive contribution we
obtain
S ' " = - ^ f IVMl-f)2 (53)
and, integrating across the duct,
*o 8 u Vn2
J
S'"dy = -^-!-^-V1*2 (54)
o 3 T5
From (54) we find that in the fully-developed section of the duct (FD)
the rate of entropy generation per unit area of duct wall is
= 6
S"FD (55)
Ta
Fig. 5 shows the variation of S"/S"FD with axial position along the
duct. As expected, the irreversibility effects are most intense near x
- 0. However, as x* increases, S" rapidly approaches the fully-de-
veloped entropy generation level S"PD a good distance before the end
724 / VOL. 101, NOVEMBER 1979
Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/28/2013 Terms of Use: http://asme.org/terms
of the entrance region, x* = 0.1038.
The total irreversibility associated with entrance effects is esti-
mated integrating expression (54) from x* = 0 to x* = 0.1038. The
result of this operation is
S' = ^ ^ l (56)
15 T
where p is the fluid density. Similarly, the total rate of entropy gen-
eration over a fully-developed stretch of length LFD is, from (55),
H Vn2
S'FD=6~^-LFD. (57)
Ta
Comparing results (56) and (57), we conclude that the fully-developed
equivalent of the entire entrance region is a fully-developed section
of length LFD given by
LFD 11
-ii = R e1 = 0.00764 Re f l . (58)
D 1440
where D is the hydraulic diameter 4a, while Reo = VQD/V. It is time
now to compare the irreversibility-equivalent length LFD with the
physical extent of the entrance region, LE. Setting x* = 0.1038 and
x = LB in expression (51) we find that familiar result
= 0.00649 Re B . (59)
Therefore, from (58,59), the irreversibility contributed by the en-
trance region is roughly equal to (actually, only about 18 percent
higher than) the irreversibility estimated assuming fully-developed
flow over a duct length equal to the entrance length. This conclusion
is important in practical irreversibility calculations, as it is consid-
erably easier to treat the entire duct as in the fully-developed re-
gime.
In a manner similar to the analysis presented in the previous three
examples; future research should address the more general case when
conductive effects are sizeable. For this extension of the present ex-
ample, the Sparrow solution [22] will again be the place to start in view
of its closed-form presentation of the mechanics of the flow. Overall,
the entrance flow example is relevant to the engineering task of
minimizing the thermodynamic irreversibility of strip fin heat ex-
changer surfaces where, to some extent, the strip fins act as a suc-
cession of entrance regions to flat ducts. In addition, the analysis
developed in this section suggests simpler means of evaluating the
entropy generation rate in the complicated entrance region flow,
namely, by considering the fully-developed entropy generation rate
(57) as a first approximation.
Concluding Remarks >
We presented a study of four basic convective heat transfer phe-
nomena from the unique point of view of entropy generation. We have
ation. In the process we developed analytical means for estimating
and minimizing the irreversibility associated with these heat transfer
configurations. The analytical results presented in this article, to-
gether with similar results which must be derived for other pedes-
tal-type convective heat transfer configurations, constitute the fun-
damental building blocks for calculating and minimizing the irre-
versibility production rate in heat transfer and thermal design cal-
culations pertaining to energy conservation.
The important engineering conclusion to be drawn from this study
is that, while seeking to minimize the destruction of available work
in complex heat transfer equipment, it is necessary to start with op-
timizing the simplest, most elementary, design features such as the
geometry of internal and external surfaces engaged in convective heat
transfer.
Acknowledgment
This work was supported by the Office of Naval Research, Grant
No. N00014-79-C-0006. The many comments received by this author
from Mr. Keith Ellingsworth of the Power Program, Office of Naval
Research, are gratefully acknowledged. In addition, this paper reflects
Transactions of the ASME
 a n u m b e r of c o n s t r u c t i v e s u g g e s t i o n s offered b y t h r e e a n o n y m o u s
r e v i e w e r s of t h e original m a n u s c r i p t . T h i s a u t h o r feels i n d e b t e d t o
t h e s e i n d i v i d u a l s as well a s t o P r o f e s s o r E . M . S p a r r o w whose com-
m e n t s p r o v i d e d a v a l u a b l e s t i m u l u s for t h e p r e s e n t w o r k .
T h i s line of r e s e a r c h is r o o t e d i n t o t h e a u t h o r ' s s t u d i e s a t t h e
M a s s a c h u s e t t s I n s t i t u t e of T e c h n o l o g y , w h e r e P r o f e s s o r s E . G. C r a - Costs of Supercritical Helium Cooled Cable Superconductors," IEEE Trans-
valho a n d J . L. Smith, J r . have created a unique thermodynamics
m e t h o d [2]. D u r i n g h i s p o s t d o c t o r a l y e a r s a t t h e U n i v e r s i t y of Cali-
fornia, B e r k e l e y , t h i s a u t h o r b e n e f i t e d from t h e advice of P r o f e s s o r
C. L . T i e n , w h o e n c o u r a g e d h i m t o go p u b l i c w i t h h i s i n t e r e s t in t h e Mechanik, Vol. 1, 1921, p. 115.
s e c o n d law a s p e c t of h e a t t r a n s f e r a n d t h e r m a l design.
References
1 Keenan, J. H., Thermodynamics, Wiley, New York, 1941, Chapter
XVII.
2 Cravalho, E. G. and Smith, J. L., Jr., ThermodynamicsAn Introduc-
tion, Preliminary Edition, Holt, Rinehart and Winston, 1971, Chapter 8.
3 Van Wylen, G. J., and Sonntag, R. E., Fundamentals of Classical
Thermodynamics, Wiley, New York, 1973, pp. 276, 277.
4 Bejan, A., "The Concept of Irreversibility in Heat Exchanger Design:
Counterflow Heat Exchangers for Gas-to-Gas Applications," ASME JOURNAL
O F H E A T T R A N S F E R , Vol. 99, No. 3, Aug. 1977, p. 374.
5 Bejan, A., "General Criterion for Rating Heat-Exchanger Performance,"
Int. J. Heat Mass Transfer, Vol. 21, 1978. p. 655.
6 Bejan, A., "Two Thermodynamic Optima in the Design of Sensible Heat
Units for Energy Storage," ASME JOURNAL OF HEAT TRANSFER, Vol. 100,
Nov. 1978, p. 708.
7 Kestin, J,, A Course in Thermodynamics, Vol. II, Blaisdell, Waltham,
Massachusetts, 1968, p. 433.
8 Bird, R. B., Stewart, W. E. and Lightfoot, E. N., Transport Phenomena,
Wiley, New York, 1960.
9 Kestin, J. and Richardson, P. D., Heat Transfer Across Turbulent, In-
compressible Boundary Layers, Int. J. Heat Mass Transfer, Vol. 6, 1963, p.
147.
10 Bejan, A. and Hoenig, M. 0., "Method for Estimating the Refrigeration
actions on Magnetics, Vol. MAG-13,1977, p. 686.
11 Schlichting, H., Boundary Layer Theory, 6th Edition, McGraw-Hill,
New York, 1968.
12 Pohlhausen, E. %., Zeitschrifl fur angewandte Mathematik und
13 Howarth, L., "On the Solution of the Laminar Boundary Layer Equa-
tions," Proceedings of the Royal Society of London, Series A. Vol. 164,1938,
p. 547.
14 Rohsenow, W. M. and Choi, H. Y., Heat, Mass and Momentum Transfer,
Prentice-Hall, Englewood Cliffs, 1961, pp. 39, 148.
15 Rohsenow, W. M. and Choi, H. Y., Op. Cit., pp. 76, 198.
16 Hilpert, R., "Warmeabgabe von geheizten Drahten und Rohren im
Luftstrom," Forsch. Geb. Ingenieur., Vol. 4, 1933, p. 215.
17 Eisner, F., 3rd Int. Cong. App. Mech., Stockholm, 1930, also in Rohse-
now, W. M. and Choi, H. Y., Op. Cit., p. 79.
18 Zukauskas, A., "Heat Transfer from Tubes in Crossflow," Advances in
Heat Transfer, Vol. 8,1972, p. 93.
19 Kays, W. M. and London, A. L. Compact Heat Exchangers, McGraw-
Hill, New York, 1964.
20 Sparrow, E. M. and Ramsey, J. W., "Heat Transfer and Pressure Drop
for a Staggered Wall-Attached Array of Cylinders with Tip Clearance", Int.
J. Heat Mass Transfer, Vol. 21, 1978, p. 1369.
21 Bhatti, M. S. and Savery, C. W., "Heat Transfer in the Entrance Region
of a Straight Channel: Laminar Flow with Uniform Wall Temperature," ASME
J O U R N A L O F H E A T T R A N S F E R , Vol. 100, Aug. 1978, p. 539.
22 Sparrow, E. M., "Analysis of Laminar Forced Convection Heat Transfer
in the Entrance Region of Flat Rectangular Ducts," NACA T N 3331,1955.

U.. FDBTALIinVICI

8TATEMENT OF OWNERSHIP, MANAGEMENT AND CIRCULATION

(Rtifutndby 39 U.S.C. 36Si)

JOURNAL OF HEAT TRANSFER i 1 1 1 i 1 ! 9-27-79

<-"*"
AHMUAl l T l c
quarterly j -- F o u r j " " S 5 000

345 East / . 7 t h S t r e e t , New Y o r k , New Y o r k 10017

345 E a s t 4 7 t h S t r e e t , New Y o r k , New Y o r k 1 0 0 1 7

B. NAMES AND COMPLETE ADDRESSES OF PUBLISHE R, EDITOR AMD MAN AQ INO EDITOR

The American S o c i e t y o f Mechanical E n g i n e e r s , 345 Eaift 4 7 t h S t r e e t , New Y o r k , NY 10017

J.J. Jaklltsch, J r . , ASME, 345 E a s t 4 7 t h S t r e e t , New Y o r k , NY 10017

1 OWNER lit MN.n,ri h, B mrpwtlnn. 11. n.m, an<i cdd,r mu h, ,t,d n<l ,f, MmHI.Kb lh,r,un*rr ,h, ,<! M. .( ,f
. .Mr* .

p.. .v ":-_

The A f r i c a n Society F ' " 345' E a s t " 4 7 t h " S

= J
Mechanical Engineers " New" Y o r k , " New Yc rk 10017

.. N O . N . O D H O L . ; ^ Of

NAME AOORIII

None None

. . PO> COMPLETION . V MONFKOFIT OHO AN 11ATIONB AUTHOIIIIED TO MAIL AT IPECIAL - A T E . [ . . . . I l l I I I . > !

AVERAGE NO. C DPI SB EACH

11 MONTHS - - E ' . ^

4,777 5,180

"' "iVS^b'NV.v:Ar.."""'" "*

2,592 3,174

,.,.L,,.I.t,..,,,,., 2,592 3,174

- :sisr^^is;;^^vx:.^V's,sm 496 4 96

3,087 3,670

'"""-"'-""""' 1,690 1,510

In At 4,777 5,180

, >m MIN;;JP?OVO^"'!'"''U"L'"""'.'"'1"""
"' ib"l tV.n."l,T,'''''' ' ' ^
II. FO COMPLETION .V PU.LI.HE-. MAILING AT , E O U l . ^ . , y l l . l l l . , W . . - . ! , . ^ ,

\=
I.NATUP,. .NO TITL. Op , . ! , . . , .U.L,.~... . . , . . . . , , ,

K , 3626 (Pagt I) (See Initiuctloni on reverie)

Journal of Heat Transfer NOVEMBER 1979, VOL. 101 / 725

Downloaded From: http://heattransfer.asmedigitalcollection.asme.org/ on 08/28/2013 Terms of Use: http://asme.org/terms