Professional Documents
Culture Documents
A Critical Review of Extended Surface Heat Transfer
A Critical Review of Extended Surface Heat Transfer
Panagiotis Razelos
To cite this article: Panagiotis Razelos (2003) A Critical Review of Extended Surface Heat
Transfer, Heat Transfer Engineering, 24:6, 11-28, DOI: 10.1080/714044411
A Critical Review of
Extended Surface
Heat Transfer
PANAGIOTIS RAZELOS
Professor Emeritus, College of Staten Island, CUNY, New York
This article is devoted to the re-examination of extended surface heat transfer basic treatment that
appears in heat transfer textbooks. It demonstrates that—unlike what is suggested in textbooks—the
emphasis on efficiency should be shifted to effectiveness instead. Graphs display how the heat
dissipation of a given fin is directly obtained without any reference to the efficiency, which is now the
standard approach, and total effectiveness is proposed to replace the total efficiency. It is further
shown that the effectiveness approach aids in verifying some of the simplifying assumptions and
shows that extended surfaces must be thermally and geometrically thin in order to fulfill their
function. The results of a brief excursion to the optimum fin problem will help students become
involved in a preliminary fin design, a topic conspicuously absent from many textbooks. Three
postulates extend certain results of the unidirectional analysis of the constant thickness fins to other
shapes and two-dimensional problems.
11
the objective of this article is to present a simple method
that will permit students to evaluate the performance of
an extended surface and above all to perform at least a
rudimentary design analysis. In order to accomplish this
goal, we have included several graphs and correlations
of certain results.
To introduce the readers to this subject, it is helpful
to first present a brief historical background.
HISTORICAL BACKGROUND
q f = ηh̄ S f T (1b)
Figure 2 Heat dissipation versus u in Gardner’s straight fins. Figure 4 Heat dissipation in Gardner’s radial fin versus u.
heat transfer engineering vol. 24 no. 6 2003 17
Figure 6 Heat dissipation in Gardner’s radial fin versus reduced
height.
Figure 5 Heat dissipation Gardner’s radial fin versus u.
these two figures reveals that the second postulate is not dissipation by changing the fin parameters instead of
confirmed. This is an example of the inconsistent results obtaining it from the efficiency.
that have been obtained by choosing inappropriate non- In the next section, we discuss the optimization of
dimensional parameters. To explain this discrepancy, let fins and present a method that permits the students to be-
us examine the parameter u: come familiar with preliminary fin design. This method,
which considerably simplifies the design process, al-
lows the practicing engineers to determine the influence
u = L h/wk = (β − 1) hrb2 wk = v(β − 1) (17)
of the thermal parameters h and k upon the fin’s opti-
mum dimensions. Before embarking on the assessment
If one uses u as a parameter, then the ratio β, which is of the simplifying assumptions and optimization, how-
a given parameter, also appears in the abscissa. Razelos ever, we develop some interesting relationships that will
and Imre [30] analyzed the circular fins using the in- enable anyone to evaluate the performance of a given fin
dependent dimensionless variables β and v. Note that from its geometry only. Considering Eq. (3) and taking
v 2 = hrb2 /kw = Br2 /Bi, where Br = hrb /k, the surface
√
Biot number. Substituting the parameter Bi = Br /v
in Eq. (16), we obtain the following dimensionless heat
dissipation:
qf
Qr = (18)
4π hrb2 T
Irey [31] and Lau and Tan [32] investigated the cri-
teria for the validity of the one-dimensional conduction
for the three types of fins considered here. Irey exam-
ined the cylindrical spine using the parameters Bi and
L/w, with their values ranging between 0.04 and 10 and
0.01 and 100, respectively, while Lau and Tan exam-
ined the straight and annular type fins, using the same
parameters with values of 0.01 ≤ L/w ≤ 500 and
0.01 ≤ Bi ≤ 10. In both papers, the conclusions are
confusing, due to the incorrect dimensionless param-
eters L/w and Bi and the wide range of their values.
Despite their correct analytical solutions, the authors ne-
glected to consider the practical aspects of the problem.
For example, have the authors ever seen in any practical
application the height of the fin to be much smaller then
its thickness? On the other hand, they should have been
aware of Schneider’s [27] mild restriction on the values
Figure 8 Efficiency of annular fin versus reduced height. of the Biot number. However, their graphs that are con-
fined in the range of L/w and Bi encountered in practical
applications, as discussed in the previous section, reveal
into account that the efficiency is always less than one,
that the resulting error by using the unidirectional con-
we obtain the following inequality for the effectiveness:
duction is less than 1 percent. Both graphs confirm the
first postulate. Since we have already established that for
Sf
εf < (19) fins (which act as good heat transfer-enhancing devices)
Sb the Biot number must be Bi 1, the one-dimensional
conduction is valid. Despite the previous results, sev-
For the fins considered here, we readily obtain the fol- eral authors have attempted to solve fin problems us-
lowing expressions: ing the two-dimensional approach, either analytically
or numerically, and compare their results with those ob-
L tained using with the one-dimensional model. However,
εf < longitudinal fins (20)
w they never took into consideration the reasons for using
fins—to enhance the heat transfer from the primary sur-
ε f < 2L/w pin fins (21) face, which is expressed by the effectiveness—or all
previously derived constraints regarding the appropri-
L(β + 1)
εf = radial fins (22) ate dimensionless variables, small values of Bi, and the
2w restrictions on the fin geometry expressed by Eqs. (20)–
(22). In every case, the one-dimensional analysis was in-
Therefore, one can now readily estimate the limits of valid, having used Bi values larger than one, very short,
the fin’s effectiveness without getting involved with any thick fins, and a L/w much smaller than one—which
mathematical analysis. Note that the above expressions means that these authors did not really analyze fins but
and those obtained earlier show that fins must always perfectly insulated surfaces, which resulted in erroneous
be thermally and geometrically thin. conclusions.
With all due respect to the authors, we are compelled
to cite the following typical examples of such fin analy-
ASSESSMENT OF THE SIMPLIFYING ses: Aziz and Nguyen [33] employed the finite element
ASSUMPTIONS method to obtain the two-dimensional effects in a tri-
angular convecting fin. They use the same parameters
In this section, we briefly discuss the validity of the and same range of values as in Lau and Tan [32]. We
simplifying assumptions, starting with the comparison believe that their results are incorrect, and even in the
of assumptions 1 (one-dimensional conduction) and 11 practical range of values 5 ≤ L/w ≤ 15 and Bi = 0.01,
(length of arc idealization) that both depend on the Biot the authors have obtained very large errors. They have
number. also concluded that for certain values of Bi > 1, the
heat transfer engineering vol. 24 no. 6 2003 19
heat dissipation obtained using the two-dimensional lutions,” which is also discussed by Huang and Shah [8].
method can be larger than the one determined by the They use the dimensionless parameters 1 ≤ L/w ≤ 5
one-dimensional conduction, and that for given ratios and values of Bi = 0.01, 0.1, 1, and 10. They also con-
L/w and Bi, the triangular profile fin dissipates more sidered a tip Biot number Bi∗ = h e w/k = Bi. They jus-
heat than its rectangular counterpart. These results vi- tify their analysis with the erroneous statement: “Appli-
olate both the first and second postulates. In another cation of one-dimensional classical approach is severely
paper, Aziz and Tesarik [34] consider a constant thick- restricted to situations with Biot numbers much smaller
ness longitudinal fin with asymmetric boundary condi- than unity.” Apparently, the authors were not even aware
tions, where the base and the tip were held at constant of Schneider’s [27] mild criterion on the value of Bi.
temperatures. In their analysis, they use the parameters Their solution is an infinite series, the eigenvalues of
0.1 ≤ L/w ≤ 100, Bi = 0.005, 0.01, 0, 1, 0.5, 1, 2, 5, which are determined from the roots of a certain tran-
and 10, and dimensionless temperature ratio θ∗ = scendental equation. It has been shown, however, in
(Te −Ta )/(Tb −Ta ) = 0.00, 0.5, 0.75, and 1. The authors Razelos and Georgiou [26] that due to the nature of the
not only used inappropriate dimensionless parameters, roots of the pertinent transcendental equations and the
but instead of subtracting the two heat fluxes from the restriction of using very small values Biot numbers,
ends of the fin, they added them; as a result, they have the solutions for the three types of fins considered are
obtained erroneous results that show that when θ∗ in- represented adequately by the first term of the series;
creases, the total heat transfer increases instead of de- hence the two-dimensional solution is no more com-
creasing. (Razelos and Krikkis [35] have obtained the plicated than the one-dimensional one. For example,
correct solution of this problem.) in the case of a constant thickness longitudinal fin, the
Also, Aziz [36] examined the effects of internal two-dimensional expression for the heat dissipation is
heat generation, anisotropy, and base temperature non- reduced to
uniformity on the performance of a rectangular longitu- √
dinal fin. He used a finite-element program to solve this qf √ C h Bi + α1 tanh(u)
problem, considering it to be two-dimensional. How- Q 2−D = = 2 Bi
2kT α1 1 + Bi + α21
ever, his analysis is based on many mathematical errors
and employed several inappropriate dimensionless pa- (23)
rameters. Although a thorough discussion of this work
is not possible here, we will only point out that due to In Eq. (23) we use the symbols C h = h e / h, and α1 =
his incorrect Eq. (11), he obtained the fallacious result 1 − Bi/6. For very small values of Bi, we may neglect
where the fin heat dissipation increases by increasing Bi in comparison to one; thus, the above expression be-
the internal heat generation, which violates the basic comes identical to the one-dimensional solution Eq. (8).
energy equation. In another paper, Aziz and Nguyen Finally, certain comments on the latest treatise Ex-
[37] attempted to analyze the two-dimensional perfor- tended Surface Heat Transfer by Kraus et al. [41] are in
mance of convecting-radiating fin of different profiles. order, because the purpose of this book is to guide stu-
The authors again are using inappropriate dimensionless dents and practicing engineers interested in the analysis
variables, and their Eq. (2b) is incorrect. As shown by and design of fins. In their preface, the authors correctly
Razelos and Kakatsios [38], the appropriate dimension-√ state: “Extended surfaces, in the forms of longitudinal
less parameters for convecting-radiating fins are Bi, u, or radial or spines are ubiquitous in applications where
and the ratio h r / h, with h r being the radiation heat trans- there is a need to enhance the heat transfer between a
fer coefficient εσ Tb3 . Consequently, Aziz and Nguyen surface and the adjacent fluid.” Discussing further the
[37] have obtained several incorrect results. For exam- usefulness criteria for longitudinal fins of rectangular
ple, in their analysis of the four profiles they consid- profile, they rederive Schneider’s [27] argument that
ered, they have determined that when Bi = 10, the heat fins must be designed with values of Bi < 1; other-
dissipation obtained by employing the one-dimensional wise, the heat flow would be retracted, as the fin will act
model is smaller to the one predicted using the two- as an insulation surface. However, in their discussion
dimensional analysis. The same authors have made sim- of the multidimensional heat conduction, they point out
ilar mistakes in another paper [39]. that according to Lou and Tans [32], “as Bi increases
Thomas [19] states: “For values of Bi much greater the error (meaning the difference between one and two-
than 0.1 multidimensionality must be accounted for in dimensional solutions) is as high as 60% for Bi = 10,
the fin analysis.” This statement contradicts the first and L/w > 1”. Unfortunately, this book is swarming
postulate, which indicates that the two-dimensional ap- with discussions of examples that contain many erro-
proach is not going to be of any help. Aparecido and neous results, such as those mentioned earlier in this
Cotta [40] present “an improved one-dimensional fin so- section.
20 heat transfer engineering vol. 24 no. 6 2003
The Length of Arc Idealization (LAI) solution, provided that the criterion of validity of one-
dimensional conduction is retained. We should mention
Razelos and Georgiou [42] have examined this prob- here that for a fin to be effective, the above require-
lem with regard to the forms of the fins considered here. ment is mandatory. Now consider a constant thickness
They have assumed a general fin profile of the following rectangular fin of thickness 2wb and height L. The heat
form: transfer coefficients at the top and bottom of the fin are
h + and h − , respectively. The coordinates are the same
y as shown in Figure 1a, except that the y-axis is now
f (ξ) = = λ + (1 − λ)ξn (24)
wb located at the distance of some equivalent distance weq
from the bottom. The semithickness weq is defined as
where in Eq. (24) ξ = x/L and λ = we /wb . In this follows:
case, the lateral surface without the LAI becomes
2wb
1/2 we = where C h = h − / h + (26)
dS Bi (1 + C h )
= 1 + 2 (1 − λ2 )n 2 ξ2(n−1) (25)
dξ u
Introducing the usual dimensionless
parameters Bieq =
Evidently, the maximum error is introduced when ξ = 1 h + weq /k, u eq = (L/weq ) Bieq and ξ = x/L, we ob-
and λ = 0, while there is no error for λ = 1, the constant tain the solution of the fin problem that was previously
thickness fin. Notice that all the profiles considered by discussed. Note that in this case, the adiabatic surface
Gardner [3], with the exception of the constant thickness is located at y = weq . For h − = h + , Eq. (26) yields
fins, have tip to base surface ratio λ = 0. The error in the weq = wb , while as h − decreases, the adiabatic surface
heat dissipation, which is introduced by approximating moves towards the bottom and for h − = 0, weq = 2wb ;
the above equation with d S/dξ = 1, has been evaluated thus the bottom surface is now adiabatic. Details of the
in [42], and it was found that values of Bi and u, which solution to this problem are given in Razelos [44]. We
are compatible with the effectiveness argument given should point out here that the preceding analysis is not
earlier, are less than 1%. limited to constant thickness fins.
For radial fins, we can replace x with the more fa- Recall that,√as shown in pin fins, in Table 1 the parameter
miliar parameter r, rb ≤ r ≤ re , L = re − rb , y, which u = (L/w) 2hw/k.
is exactly as in the longitudinal fins, while the dif- In the literatures, standard one-dimensional fin treat-
ferential surface element here is d A = 2πr d y. Fol- ment, which is based on the simplifying assumption
lowing the same approach as in the longitudinal 9, the quantity ε has been neglected in Eqs. (A6, A7,
fins, we readily obtain the following one-dimensional and A10). Therefore, because the surface temperature
equation: θ̄ − ε < θ̄, we can conclude that the one-dimensional
approach always overestimates the heat dissipation.
d 2 θ̄ 1 d θ̄ Q. E. D.
2
+ − u 2 (θ̄(ξ) − ε(ξ)) = 0 (A7)
dξ ξ dξ
Panagiotis Razelos is a Professor Emeritus of the
City University of New York. He has a Diploma in
In the above equation, ξ = r/L; however, it was pointed Mechanical and Electrical Engineering from the
out earlier that it is more advantageous in the above National Technical University, Athens, Greece,
equation to use the parameters v = u/(re /rb − 1) = and M.S. and Ph.D. degrees from Columbia Uni-
u/(β − 1) and ς = r/rb to obtain the following new versity. Formerly he was director of the Heat
and Mass Transfer Laboratory at Columbia Uni-
expression: versity. Dr. Razelos’s research interests are heat
transfer in regenerators, extended surfaces, con-
d 2 θ̄ 1 d θ̄ vective heat transfer, computational methods,
+ − v 2 (θ̄ − ε(ς )) = 0 (A8) simulation, and heat exchangers. He has pub-
dς 2 ς dς lished more than 60 articles in these areas.