See

discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/229709417

Forced Convection Heat Transfer Correlations for

Flow in Pipes, Past Flat Plates, Single Cylinders,
Single Spheres, and for Flow in Packed Bed and
Tube Bundles

ARTICLE in AICHE JOURNAL · MARCH 1972

Impact Factor: 2.75 · DOI: 10.1002/aic.690180219

CITATIONS READS

388 1,618

1 AUTHOR:

Stephen Whitaker
University of California, Davis
219 PUBLICATIONS 8,551 CITATIONS

SEE PROFILE

Available from: Stephen Whitaker

Retrieved on: 06 October 2015
 Reprillted from AIChE JOURNAL, March 1972

Forced Convection Heat Transfer Correlations

for Flow In Pipes, Past Flat Plates, Single
e Cylinders, Single Spheres, and for Flow
In Packed Beds and Tube Bundles
STEPHEN WHITAKER
Department of Chemical Engineering
University of California, Davis, Calif. 95616

Previously obtained experimental heat transfer data have been collected and are illustrated
along with minor variations of the standard correlations. Analysis of data for heat transfer in
randomly packed beds and compact (void fraction less than 0.65) staggered tube bundles
indicates that the Nusselt number for a wide range of packing materials and tube arrange-
ments is given by
NNu= (0.5 NRe1!2+ 0.2 NRe 2/ 3 ) Npr l/3 (lLb/ILO)O.14
provided NRe ~ SO. The correlations presented in this paper are not necessarily the most
accurate available; however, they have wide application, are easy to use, and are quite
satisfactory for most design calculations.

For the past forty years there has been a steady effort following equations:
to improve our knowledge of forced convection heat trans- equations of motion
fer rates for a variety of important process configurations.
It is of value to periodically review the experimental data
and construct new correlations on the basis of new data
P ( aa: + v·;z. v) = - V P + pg + ~"t (1)

and theoretical advances. McAdams (23) contributed constitutive equation for a Newtonian fluid
much along these lines in 1954; however, enough new
data have been obtained so that these correlations deserve -c=JL(Vv+Vv+) (2)
a second look.
continuity equation
All the experimental data that we wish to consider in
(3)
this paper can be satisfactorily described in terms of the

AIChE Journal (Vol. 18, No.2) March, 1972 Page 361

thermal energy equation (J == tu" /L", dimensionless time
8 == (T - TO)/(To - TO), dimensionless temperature.
pCp ( :~ + (v' ~T) ) = k'V 2T (4)
TO and To are often chosen
so that 0 ~ 8 ~ 1
N pr = cp p./k
The form of these equations indicates that we are consider-
ing the density and thermal conductivity p and k to be where it is understood that all physical properties are eval-
constant. In addition we will take Cp to be a constant while uated at the reference temperature. At this point we can
allowing the viscosity p. to be a function of temperature. specify the functional dependence of the dimensionless
Substitution of Equation (2) into Equation (1) leads to temperature as
((J,X,Y,Z),
~p+
8=F{
p ( ::- + v . 2. v ) = - pg + p.'V 2v
independent parameters parameters
+ ('V p.) • ('V v + 'V v+) (5) variables appearing in appearing
the differential in the
equations boundary
Since we are taking the viscosity to be a function of tem- conditions
perature alone, we can write (12)

'VP,=(!f-) ('VT) (6) We will not discuss the parameters which appear in the
..,.. iJT """ dimensionless form of the boundary conditions; however,
we will carry the term N Be as a reminder that one must
In addition we can write a Taylor series expansion for p.
always carefully consider the boundary condition.
about the reference temperature T".
CHOICE OF REFERENCE TEMPERATURE

In the analysis of heat transfer processes the bulk or

"cup mixing" temperature is the most convenient tempera-
(7) ture to use in the macroscopic balances. This temperature
will be denoted by T b and is defined by
Here the asterisk superscript indicates that the function
is evaluated at the reference temperature T". We can fA T v'ndA
now express Equation (5) as (13)

p (:: + v ~ . v ) = - ~ p + pg
i v'ndA

+p." [1+ p.l" (:;r (T-TO)] 'V v 2

For the case of flow past bodies immersed in an infinite
fluid the bulk temperature becomes the free stream tem-
perature
T b = T., for bodies immersed in an
+ ( -
iJpo )0 ('VT)' ('Vv+ 'Vv+) (8) infinite fluid (14)
aT "" ..,. ..,.
In many of the early heat transfer studies the film tem-
The higher order terms in the Taylor series expansion have perature T f was taken as the reference temperature. It is
been dropped, thus we are restricting our analysis to the defined as
first order effect of temperature on the viscosity. Choosing 1
the characteristic length as L ° and the characteristic veloc- Tf = -2 (Tb + To) (15)
ity as UO we can put Equations (3), (4), and (8) in
dimensionless form to obtain where To is the surface or wall temperature. We should be
very careful to understand at this point that 8 depends on
( aaU(J) + U . 'V
.....
U = - 'V P + -1- [1 + A8] 'V 2 U
-.,.. N Re
N Re, N Pr, and A regardless of the choice of the reference
temperature. It is true that one or the other of these tem-
1 peratures may represent a "better" choice of a reference
+ - [A('V 8) . ('V U + 'V U+)] (9) temperature, thus minimizing the functional dependence of
N Re ..,. '" ."..
8 on A. Indeed, in the early work on forced convection
'V'U=O (10) heat transfer it was assumed that use of the film tempera-
ture as a reference temperature would essentially eliminate
any dependence on A; however, a number of recent studies
( 11) using highly viscous oils indicate that use of the bulk tem-
perature and a parameter comparable to A is the preferable
method of correlating the data.
Here we have used the following definitions
In order to illustrate how one makes use of the parame-
U = v / u 0 , dimensionless velocity vector ter A, let us consider the problem of heat transfer in a pipe
'V = L ° 'V, dimensionless "del" operator for the following conditions:
p = p - pO + -±-dimensionless pressure which 1. wall temperature is constant at To
pU0 2 U0 2 '
includes the body force 2. bulk temperature at the entrance is Tbl
N Re = puG L °/ p., Reynolds number 3. bulk temperature at the exit is T b2

A = - 1 ( - apo) , d'ImenSlOn
. Iess VISCOSIty
. . parameter We choose the reference temperature to be the average
po a8 bulk temperature

Page 362 March, 1972 AIChE Journal (Vol. 18, No.2)

 case where the viscosity is a function of temperature.
(16)
NNu = 1.86 N Re 1/3 N pr1/3 (L/D)-1/3 (JLb/JLO)0.14 (20)
and note that an order of magnitude estimate of A is In this expression the physical properties of the fluid are
given by evaluated at the reference temperature, that is, the mean
or average bulk temperature. The ratio (I'b/I'O) represents
(.!!:..)
o
A=-
1
= ~ [ JL - JLo ] = [ JL~ _ 1] the ratio of the viscosity evaluated at the mean bulk tem-
JL ae JL 0-1 JL perature to the viscosity evaluated at the mean wall tem-
(17) perature. The solution to the Graetz problem indicates that
In arriving at this result we have used the fact that e is for very long tubes
given by
T- TO
NNu = 3.66, LID ~ 00 (21)
8= (18) When the viscosity is a function of temperature this result
To - TO
can be approached asymptotically from either above or
and thus varies from 1 at the wall where T Toto ap-
proximately zero at the center of the pipe where T = 100 r----r---,---~-,......,.---..,._--__,-__,

1h (T b1 + T b2)' We now see that the effect of variable 80

viscosity can be correlated by the ratioO of viscosities o Sieder and Tate, Heating and
Cooling Oil, L/ D = 82
JLb/ JLO, but we must remember that this result is based on 60
A Drew, Heating Glycerol, L! D = 220
retaining only the first term in a Taylor series expansion [J Colburn and Coghlan, Heating
Hydrogen - Nitrogen Mixtures, L/ D =97
of 1', and the order of magnitude estimate given by Equa- 40
tion (17). For steady state processes we can now express
Curve is Given by:
the functional dependence of the Nusselt number as
NNu = hLo/k 20

(19)

Discussion 10 ~-------~~---+-----------l

Having decided upon the functional dependence of the

Nusselt number for the case where only the viscosity is al-
lowed to vary with temperature, we are in a position to
consider the experimental data.

LAMINAR FLOW IN A PIPE Lower So... d hlr Long Tubes

When the restriction N ReNPr > 100 is satisfied (35) the

heat transfer process is adequately described by the Graetz
solution (18). An empirical modification of the Graetz
solution has been given by Sieder and Tate (37) for the 1~1---~--~-~-~~--~---~-~60

o Here po represents the viscosity evaluated at the average bulk tem-

perature, thus fLb = p. O • Fig. 1. Heat transfer to fluids in laminar flow in tubes.

TABLE 1.

Definition of Definition of Range of Range of

Reynolds Nusselt Definition Reference Reynolds Prandtl Range of
Process number number ofh temp., TO number number (JLb/ JLO)

Laminar How 1
in a pipe <Vz>D/Vb hD/kb Q/1TDLt:.Tln 2" (Tb1 + Tb2) 13-2,030 0.48-16,700 0.0044-9.75
Turbulent How 1 2.3 X 103
in a pipe
Flow past a
<Vz>D/vb hD/kb Q/1TDLt:.T l n _ (Tb1 + Tb2) - 1 X 105 0.48-592 0.44-2.5
2 1 X 105
Hat plate It~L/v., hL/k., Q/2wL (To - T.,) T., - 5.5 X 106 0.70-380 0.26-3.5
Flow past a
single sphere u,D/v., hD/k., Q/41TD2 (To - T.,) T., 3.5 - 7.6 X 104 0.71-380 1.0-3.2
Flow past a
single cylinder u.,D/v., Q/1TDL (To - T.,)
hD/k., T., 1.0 - 1 X 105 0.67-300 0.25-5.2
Flow in • ( hDp)
.. ,1
packed beds DpG/JLf (1 -.) T:;" 1 _ • Q/av V t:.Tln - (Tfl + Tn) 22-8 X 103 0.7 1
Flow in 2

e stag~ered tube
bun les
DpG/JLb (1 - .)0 ( -
hDp) e
kb
.
- - - Q/a v V t:.Tln
1-.

6V. 3
.:. (Tb1 + Tb2) 1.0-105
2
0.7-760 0.18-4.3

• D. for tube bundles is given by D. =-- =- D.

A. 2

AIChE Journal (Vol. 18, No.2) March, 1972 Page 363

 10 3 .-----,----,----,--,-,------,r----.,----,-----r-,---,----,----;,...,-.---...,
8
o Lawrence and Sherwood, Heating Water, L / D ::; 59
"Morris and Whitman, Heating and Cooling Oil, LID =196
o Sherwood and Petrie, Heating Acetone, L / D = 96
• Deissler and Eian, Heating Air, L/ D = 100
.. Colburn and Coghlan, Heating Hydrogen - Nitrogen
Mixtures, L/ D =97
102 • Friend and Metzner, Heating Corn Syrup, L-
8
Curve is Given by:
4 NNu!N 0p~2( fLb /lLo) 0.14::; 0.015 N°R~3
10 ~-----~-~~--+------+--~
8 The range of variables represented in Figure 2 and the
Curves are Given by:
N N./No;~,( 1',/1',1°.14 = 1.86 N!~; (L/ Dr!
the coefficient 0.015 and the Reynolds number depen-
dence are in general agreement with a more thorough
survey of the turbulent Graetz problem by Notter and
Sleicher (27). The data of Friend and Metzner (15) for
corn syrup, and the data of Deissler and Eian for air
(10), represent the most recent and most accurate experi-
mental values; however, one should note that the pioneer-
ing work of Morris and Whitman, performed prior to 1927,
has rather nicely withstood the test of time. Once again
one must remember that the logarithmic scale tends to
hide the experimental error which is on the order of ±
15 %. For a more accurate correlation the interested reader
is referred to the recent work of Notter and Sleicher (27).
definitions of N Re and N Nu are again listed in Table 1.
'3

FLOW PAST A FLAT PLATE

46810 5
For Prandtl numbers ranging from 0.6 to 10, a reason-
able expression for the local Nusselt number in the laminar
Fig. 2. Heat transfer to fluids in turbulent flow in tubes. boundary layer (34) is
NNu.x = 0.332 NRe.xl/Z N p ,1I3,
laminar boundary layer (23)
below (47). In Figure 1 we have compared Equation
(20) with the experimental data of three investigations On the basis of the experimental work of Zhukauskas and
(7, 15, 37). The agreement obtained by three different Ambrazyavichyus (45) we can modify Colburn's (8) em-
investigators for a fairly wide range of the dimensionless pirical expression for the local Nusselt number in a tur-
parameters is certainly very pleasing and indicates that bulent boundary layer to obtain
the Sieder-Tate correlation is quite satisfactory. This par- NNu.x = 0.029 NRe.xo,s N pr O.43,
ticular process has been analyzed in some detail by a
number of investigators (4, 5, 25) and without question turbulent boundary layer ( 24 )
there are better expressions available for the Nusselt num-
ber; however, Equation (20) and other correlations pre-
If we assume that the transition takes place at a length
Reynolds number of 2 X 105 , and if we assume that
sented in this paper are given with the idea that in the
Equation (23) is valid before the transition and Equation
course of the design, construction, and operation of most
heat transfer equipment there will be considerable uncer- (24) is valid after the transition we can determine the
tainty and there is little need for using correlations which average Nusselt number to be
introduce far less uncertainty than that produced by the NNu = 0.036 (NRe.LO,S N pr°.43 - 17,400) + 297 N prl/3
design, construction, and operation. The range of the di- (25)
mensionless parameters for the data shown in Figure 1 and For Prandtl numbers near unity we can write
the definitions of N Re and NNu are given in Table 1. In
plotting the Nusselt number against (NReN pr DIL) 113 we 297 N Pr 1l3 >= 297 N Pr°.43 (26)
may be guilty of what Colburn (8) referred to as plotting and Equation (30) can be Simplified to
a function against itself, and there are many who would
suggest at this point that the use of a i-factor or Stanton NNu = 0.036 N pr O.43 (NRe.Lo,s - 9,200) (27)
number would be more appropriate. A detailed discussion The data of Zhukauskas and Ambrazyavichyus indicate
of this point is given in the appendix, and there it is shown that the dependence of the Nusselt number on the viscosity
that there is no preference for the i-factor or Stanton num- should be of the form (JL~I JLo)1/4 thus Equation (27) can
ber over the Nusselt number. One must always be careful
be expressed as
in viewing plots such as the one shown in Figure 1 to re-
member that the scale is logarithmic, and experimental NNu = 0.036 N prO.43 (NRe.Lo.s - 9,200) (JLooIJLO)1/4
error tends to be obscured. The error in the experimentally (28)
determined Nusselt numbers is on the order of ± 25%.
If the Reynolds number is very large relative to the criti-
cal Reynolds number Equation (28) takes the simplified
form
TURBULENT FLOW IN A PIPE (29)
The experimental data of a number of investigators (7, Both Equations (28) and (29) are compared with experi-
10, 15, 22, 26, 28) for turbulent pipe How are shown in mental data in Figure 3. Although a comparison is not
Figure 2. The Nusselt number is quite nicely correlated shown, the data of Parmalee and Huebscher (28) and the
by the expression data of Edwards and Furber (12) for negligible free
NNu = 0.015 N Reo,S3 N pr°.42 (JL bIJLo)O,14 (22)
stream turbulence are in excellent agreement with the a.
average Nusselt number obtained from Equation (23) •
The Prandtl number dependence is based on the work of provided the Reynolds number is less than 2 X 105 • For
Friend and Metzner (15), the functional dependence of values of N Re > 2 X 105 the data of Parmalee and Hueb-
(JLbl JLo) is from the work of Sieder and Tate (37), and scher show a sudden increase in N Nu with the data being

Page 364 March, 1972 AIChE Journal (Vol. 18, No.2)

well represented by Equation (28). When great care is provided the sphere is immersed in an infinite media. In
taken to eliminate free stream turbulence the transition is choosing a functional dependence on N Re, N PT and (/Lbl/Lo)
delayed and the Nusselt number is lower than that pre- we begin with the form
dicted by Equation (28). This is the case of the negligible
free stream turbulence data of Edwards and Furber. When (NNu - 2) = F(NRe, N pT, /LbIJLo) (31)
high intensity turbulence is generated in the free stream subject to the restriction that F (N Re) ~ 0 as N Re ~ O.
the transition to turbulence takes place ata lower Reyn- In considering the functional dependence for the Reynolds
olds number and the Nusselt number is larger than that number we refer to the work of Richardson (33) who
predicted by Equation (28). This is the case for the 5 % recommended that the heat transfer from the sphere be
free stream turbulence data of Edwards and Furber, which considered as two parallel processes. In the laminar bound-
lie very close to the line of given by Equation (29). Al- ary layer region the contribution to the Nusselt number
though the level of turbulence was not specified by should be of the form N Re 1l2 N PT l/3 , while in the wake
Zhukauskas and Ambrazyavichyus it is clear from their region Richardson argues that the contribution should be
results that a high level of turbulence was present in the of the form N Re2l3 N PT l/3 , so that Equation (31) could
free stream. The effect of free stream turbulence illustrated be expressed as
in these heat transfer data is in good agreement with the
experimental studies on laminar boundary layer stability (NNu - 2) = (aN Re 1l2 + bNRe2l3) N (/LbIJLo)c
pT 1I3

performed by Schubauer and Skramstad (36). The func- (32)

tional dependence of the Nusselt number on the Prandtl
number is essentially identical to that found for turbulent
pipe flow; however, the dependence on the viscosity ratio
(/Lbl/Lo) or (/L~//Lo) is considerably different. In considering
this point one must remember that pinning down the best
exponent for the viscosity ratio is indeed a very difficult
experimental task, and one could easily look upon the
difference between 0.14 and 0.25 as being due to experi-
mental uncertainty. In addition we must keep in mind that
the Nusselt number is influenced by the temperature de-
pendence of the viscosity in some complex manner, thus,
much of the experimental uncertainty could result from
trying to force a simple functional relation to describe a
complex phenomena. The range of variables for the data
presented in Figure 3 are listed in Table l.
On the basis of the work of Vliet and Leppert (40) and
the supporting data of Kramers (21) the exponent on the
Prandtl number is chosen to be 0.4 and the exponent on
the viscosity ratio is taken to be 1/4. The constants a and
b are chosen to obtain a good fit with the experimental
data, and the final form of Equation (32) is
(NNu - 2) = (0.4 NRe l/2 + 0.06 N Re 2/3) N pT°.4 (JLbl/LO) 114
(33)
This expression is rearranged slightly and compared with
experimental data in Figure 4. The agreement is quite
good, the scatter of the data around the correlation being
± 30% at the very worst. Once again the range of values
of the parameters are listed in Table l.

FLOW PAST A SINGLE SPHERE FLOW PAST A SINGLE CYLINDER

In treating flow past a single sphere we must make use In developing a correlation for heat transfer from a
of the fact that a steady state conduction solution exists single cylinder we note that no steady state conduction
(3) which yields solutions exists, thus we require that
N Nu = 2, steady state conduction solution (30) NNu~ 0 as NRe~ 0
The success of the form used for correlating the sphere
data motivates us to propose a correlation of the form
SOlid Curve is Given by:
NNu = (0.4 NR.IIZ + 0.06 N R.2/3) N prO.4 (JL bl/Lo)1/4
N NU/N np;3 (p.=/ 1'0) 1/4= 0.036 N~:
(34)
Dashed Curve is Given by:
The comparison of Equation (34) with the data of a
NNJNOp;3(p..,/1'0)"4 = 0.036 (NOR! - 9200)
number of investigators (6,9,14,19,29,31,46) is shown
in Figure 5. The agreement is generally within ± 25% ex-
cept in the region of low Reynolds numbers where the data
of Hilpert and King are considerably higher than that pre-
dicted by Equation (34). More recent data obtained by
Collis and Williams (48) for air are in reasonable agree-
ment with the data of King and Hilpert. For Reynolds
o Parmalse and Huebscher, Healing Air,
Unspecified Free Stream Turbulence
numbers in the range of 0.02 to 44 and small temperature
z
o Edwards and Furber, Heatina; Air,
difference Collis and Williams were able to correlate their
~ Neglia:ible Free Stream Turbulence data with the expression
z
A Edwards and Fuber, Heatina: Air,
5% Free Stream Turbulence NNu = 0.24 + 0.56 N Re°.45, for air with negligible
• Reynolds, Kays, and Kline, Heating Air, variation in the viscosity
Unspecified Free Stream Turbulence

.. Zhukauskas and Ambrazyavichyus, Heatina:
and Coolinl Air, Water, and Oil, Unspecified
Free Stream Turbulence
Fig. 3. Heat transfer to fluids flowing past a flat plate.
For this case Equation (34) reduces to
NNu = 0.35 NRe l/2 + 0.052 N Re2/3
for air with negligible variation in the viscosity
Taking N Re =
1 the Collis-Williams correlation predicts
a value of 0.80 for the Nusselt number, while the correla-

AIChE Journal (Vol. 18, No.2) March,1972 Page 365

tion given in this paper yields a value of 0.40. One of the The surface area per unit volume av can be related to
drawbacks to the Collis-Williams correlation is that it the void fraction e which is defined as
does not satisfy the theoretical restriction that N Nu ~ 0 as
void volume
N Re ~ 0; nevertheless, it represents the results of some
in the bed V void
very careful experimental work and is in good agreement e= ---- (36)
with the data of King and Hilpert. Clearly there is a need total volume V
for more theoretical and experimental studies for Reynolds of the bed
numbers less than 10. Consider a bed of volume V packed with N particles which
It is convenient that we can correlate both cylinder have a volume Vp and a surface area Ap. If the void frac-
and sphere data with essentially the same equation; how- tion in the bed is e, the volume occupied by the particles
ever, one must keep in mind that the coefficients and is V( 1 - e), thus we can write
powers on the Reynolds number are not sacrosanct. We
are attempting to model the functional dependence of (37)
NNu on N Re with a function of the form: aNReb + cNRed,
and with four adjustable constants one could find other This result can be rearranged to give the number of parti-
choices which might fit the data better than Equation (34). cles per unit volume as
In any event there is some theoretical justification of the N/V = (1 - e)/Vp (38)
form of both Equations (33) and (34) and they do fit the
data reasonably well. The range of variables illustrated and the surface area per unit volume is
in Figure 5 is listed in Table 1. au = Ap (N/V) = (Ap/Vp ) (1 - e) (39)
Knowing the geometry of the particles and the void frac-
FLOW IN PACKED BEDS tion in a bed allows us to determine avo Measurements of
Qand t.T ln then allow one to compute h1n from Equation
In general the log-mean film heat transfer coefficient (35) .
for a packed bed is defined as follows In order to correlate heat transfer data for packed beds
we must decide upon a characteristic length and a charac-
(35)
teristic velocity. The quantities should be chosen with the
Here Q is the total rate of heat transfer from the packing idea that there should be a strong correlation between the
to the fluid, h 1n is the film heat transfer coefficient defined heat transfer rate and the characteristic length and velocity.
by Equation (35), av is the packing surface area per unit A little thought should lead one to the conclusion that the
volume, V is the total volume of the packed bed, and t.T 1n characteristic length should be a measure of either the size
is the log-mean temperature difference. of the packing or the size of the void spaces through which
the fluid flows. We shall see shortly that the hydraulic
radius Hn is related both to the packing size and the void
space, and is therefore an especially suitable choice for a
characteristic length. The hydraulic radius is traditionally
used to characterize turbulent flow in noncircular ducts and
is defined by (41)
Curve is Given by

H _ ( cross section available for flow ) ( 40)

n- wetted perimeter of the cross section
For straight channels or ducts the hydraulic radius is eaSily
--- -------+----"''---.<i'''i'''---I------l- - - - - ,
- 10
8 computed; however, for a packed bed this definition must
be extended as follows (3)
Hn = ( cross section available for flow )
wetted perimeter of the cross section

Fig. 4. Heat transfer to fluids flowing past a single sphere.

=( volume of the bed available for flow ) (41 )
wetted surface within the bed
void volume/total volume )
10 3 (
,
8
o Perkins, H.C., and G. Leppert, Water and Ethylene Glycol - wetted surface/total volume
to Brier, J.C., and S.W. Churchill, Nitroeen
o Piret, E.L., 't! James, and M. Stacy, Water
• Fand, R.M., Water
= elav
"'Kine, L.V" Air Note that in this definition of Rn the wetted surface of
101 -Hilpert, R., Air
8
,. Davis, A.H., Water, Parafin, and Transformer Oil
the walls of the bed are neglected relative to the wetted
;: CUfve is Given by:
surface of the packing. We may now use our previously
derived expression for au to express Rn as
~~
dZoi:

(~)
, 10 1
f ,
8 Rn= (42)
Ap 1- e

.'
Rather than use Rn as the characteristic length we will use a.
1• six times the hydraulic radius so that L" is given by ,.,
4 G 8103 2 • 58104 2 4 68 5
10

·R. (43)
Fig. S. Heat transfer to fluids flowing normal to a single cylinder.

Page 366 March, 1972 AIChE Journal (Vol. 18, No.2)

We now define the particle diameter Dp as We should remember that our dimensional analysis in-
dicated that
Dp == 6Vp lAp (44)
NNu = F(N Re , N pr, p.,b/p.,o, NBC) (53)
so that our expression for the characteristic length becomes
where NBC represented all the dimensionless parameters
L" = Dp (_E
1-
) E
(45)
appearing in the boundary conditions. Up to this point we
avoided the problem of parameters appearing in the bound-
ary conditions, for aside from a possible effect of wall
The definition for Dp given by Equation (44) was chosen roughness on heat transfer in a pipe, no parameters have
so that the particle diameter for a sphere is the actual arisen from the boundary conditions other than the effect
diameter of the sphere. of LID contained in Equation (20). Now, however, we
In choosing a characteristic velocity, it seems logical are confronted with very complex boundary conditions and
that the average velocity of the fluid flowing in the void we might well expect that other factors besides N Re, N PI
space <v z > would be most suitable. This velocity is de- and I-'b/l-'O would affect the Nusselt number.
fined by All of the experimental data for heat transfer in packed
<v z > = _1_
Avoid
f Avoid
Vz dA (46)
beds are for air (16, 17, 24, 39, 43) with the exception
of the results of Glaser and Thodos (17) who worked with
nitrogen. The data are presented in Figure 6 and compare
Here Avoid is the void area contained within any cross quite favorably with the equation
section of the bed. If the bed is uniform, Avoid and <v z>
are constant throughout the bed. The volumetric flow rate NNu = (0.5 NRe + 0.2 N Re 1l2 2/3) N pr 1l3 (54)
Q is given by The functional dependence on the Reynolds number
(47) clearly was motivated by the success of this form for fitting
Nusselt numbers for single spheres and cylinders. There
are several important points about this correlation which
and the void area is related to the cross sectional area of must be noted:
the bed A by the expression (42) 1. The Prandtl number dependence was chosen by the
Avoid = E A (48) original investigators, but it remains unsubstantiated since
there was negligible variation of the Prandtl number in
We can now express our characteristic velocity as the experimental studies. In the next section on staggered
u" = <vz> = Q/EA (49) tube bundles there are results which indicate that the
one-third power is reasonable.
The volumetric flow rate divided by the cross sectional 2. The data in Figure 6 represent an enormous range
area is referred to as the superficial velocity of packings primarily because of the extensive study by
Taecker and Hougen. They performed experiments for
superficial = QI A (50) three sizes of Raschig Rings, one size Partition Ring, and
velocity two sizes of Berl Saddles. Their range of particle geome-
and is often used as a characteristic velocity for fixed bed tries is certainly extensive and yet the results can be
catalytic chemical reactors. The superficial velocity is of correlated with a single equation prOVided one uses the
course completely insensitive to variations in void fraction appropriate Nusselt number and Reynolds number. This
and is therefore judged to be a less suitable characteristic is shown more clearly in Figure 7 where all the data ob-
velocity. tained by Taecker and Hougen are plotted along with all
At this point we have chosen our characteristic length the data of Glaser and Thodos for spheres and cylinders.
and velocity, 3. The data for cubes obtained by Glaser and Thodos
lie significantly below the other data. This is presumably
L"=Dp (_E
1-
) due to the reduction of the effective surface area that can
E occur when two or more cubes become stacked.
Dp = 6VplAp 4. The high void fraction data (. = 0.78) of McCon-
nachie and Thodos are significantly higher than the other
u" = Q/EA
and we can now express the Reynolds number and Nusselt
number as
o C. R. Wilke and 0 A. Houren, €= 0.34-0.44,
Cyliooers witf1 LlQ:::::l
• R. G. Taect" and 0 A. HOUlen, 0.34-0.74,
= pQDplp.,A(l - E)
€=

(51) Commercial Packinl!

c J. T. McConnachie and G. Thodos, ~"0.42,

= DpGlp.,(l - E) 10 2
Sphares.
• M. B. Glaser and G. Thodos, €= 0.44- 0.48,
8 Spheres and Cylinders wilh LID =0: 1
NNu = hL"lk .& B. W. Gamson, G. Thodos, and O. A. Houpn, €
Spheres and Cylinders WIIh L! 0 ~ I
(52) CURe is Given by

Here G is the mass velocity equal to pQI A. Our choice of

6 Me Connachie and Thod05,
characteristic length and velocity have led us to a Reynolds f=O,78,Sph,res

number and Nusselt number having the term (1 - E) in • Glaser and Thodos,
e=0.42, Cubes
the denominator. We must therefore expect that as the
l.OL-+-+-,!-+L---!-_+-J,.-!-.J,.--+_...L.+-!+----!,..-++.~-
void fraction becomes large our characteristic length and 1.0 4 68 10 4 6 8102 4 6 BU3 4 68104

velocity will become unsuitable for (1 - E) -1 varies ex- 'R.

ceedingly rapidly as E approaches unity. Fig. 6. Heat transfer to gases flowing in packed beds.

AIChE Journal (Vol. 18, No.2) March, 1972 Page 367

 Sr = SliD (56)
Experimental Oata of Taecker and Houlen
l:! liZ - IN Raschig Rings, E::: 0.63
5t = stl D (57)
• 1- IN Raschig Rings, 2::: 0.73
o 2 _ IN Raschig Rings, E = 0.74 Drawing upon our experience with packed beds we pro-
[] 2 -IN Partition Rings, 2::: 0.62 pose essentially the same correlation for staggered tube
• 1/2 - IN Raschig Rines, E::: 0.63
o 1/4-IN Berl Saddles, 2::: 0.60 bundles, that is
• 112 - IN Berl Saddles, e:::: 0.63

Curve is Given By:

10
EKperimental Data of Glaser and Thodas
ySpheres, 2=0.44
... Cylinders With LI 0:::: 1,
Fig. 7. Heat transfer to gases flowing in packed beds.
data. This would seem to indicate that the correlation may
not be valid for large values of the void fraction; however,
the data of Taecker and Hougen contain void fractions
which are nearly as large as the value of 0.78. Another
NNu = (0.5 NRe + 0.2 N Re l/2 2/3) N pr l/3 (lLbIILO)O.14
(58)
This correlation is compared with a wide range of ex-
perimental data (1, 2, 20, 30) in Figure 9. There are a
number of important points to be noted regarding Figure
9:
1. Provided N Re ~ 102 and E ~ 0.65 the same correla-
tion, corrected for the effect of variable viscosity, that we
used to predict heat transfer rates in packed beds can be
used for staggered tube bundles.
2= 0.48
2. The data in Figure 9 represent Prandtl numbers rang-
ing from approximately 0.7 to 30 with the exception of
the low Reynolds number data of Bergelin, Colburn, and
Hull. Prandtl numbers for that data are on the order of
290 to 760. Because of the limited Prandtl number varia-
tion in the range of Reynolds numbers from 102 to 104 one
cannot consider the one-third power on the Prandtl num-
ber to be experimentally confirmed. In view of the fact
that the Prandtl number exponent ranged from 0.4 to 0.43
for turbulent flow in pipes, past flat plates, spheres and
cylinders, we must continue to speculate that the N pr l/3

j
DIRECTION
explanation is that all the data shown in Figure 6 are for OF FLOW
randomly packed beds with the exception of the data of
McConnachie and Thodos who constructed body-centered
cubic arrangements of spheres for their packed beds. With

r- --11 I
this type of regular geometry it seems likely that parame-
St D
ters describing the geometry of the bed would be needed
to complete the correlation. Data for tube bundles seem
to indicate the geometrical effects become more pro-
nounced as the arrangement of the heat transfer surface
becomes more regular, that is, having less of a random
_ -T - S!
nature.
If we discard the high void fraction data of McConnachie
and Thodos and the data of Glaser and Thodos for cubes
_ _ _ 1_
we can conclude that Equation (54) represents a satis-
factory correlation for heat transfer in randomly packed
beds of spheres, cylinders, Raschig rings, partition rings,
and Berl saddles. For these cases the correlation can be
considered accurate to better than ± 25 %.

Fig. 8. Geometry for a staggered tube bundle.

FLOW IN STAGGERED TUBE BUNDLES
Because of the success obtained in correlating heat
transfer data for a wide variety of packed beds, it seemed • O.P Berlsiln, M.O.Leithlon. W.LLafierty. and R.L Pi,IOI"d t=O.42-0.60 011

reasonable to apply the correlation to heat transfer in tube ;'1 : ~:~: ~:~~: ::~~! Ten Rows 01 Tubes
oO.P.Berlehn, A.P.Colburn and H,LHull.t=O.42-0.60. Oil
bundles. In this section heat transfer data for staggered ~: : ~:~: i:~: 1:~~ I Ten Rows QI Tubes
tube bundles will be considered. The geometrical con- 10' oW.M. Kays and R. K La: t=O.48-0.65. Au

figuration is illustrated in Figure 8. The geometry of the -,;. ~~ : ~.~~: ~:~~: !:~: !:~~: ~:;~: ~:~! FIfteen Rows of Tubes
_0, LPierson t=O.50,065,Air
tube bundle can be specified in terms of D, St, and Sr pro-
vided the length of the tubes is much larger than the diam-
:
~; ~.~: ::~~! Ten Rows af Tubes

eter, and provided the tube bundle contains at least ten ~ 10

"OlPierson· t =091,S,;300,St;3.00
.... O.LPlerson e=081,S, = 2.00, SI =3.00

rows of tubes. Under these circumstances entrance and

edge effects are negligible and the functional dependence
of the Nusselt number can be expressed as
NNu = F(N Re , N pr, ILblILo, Sr, St) (55)
N••
Here we have replaced N Be in Equation (53) with the
dimensionless parameters Sr and St which are defined by Fig. 9. Heat transfer to fluids flowing past staggered tube banks.

Page 368 March, 1972 AIChE Journal (Vol. 18, No.2)

dependence is something less than the best functional de-
pendence, but that it is satisfactory provided N Pr "'" 30.
3. For Reynolds numbers less than 10 2 the high Prandtl
number data of Bergelin, Colburn, and Hull begins to
deviate from the correlation expressed by Equation (58).
It seems possible that this is caused by the interaction of
the boundary layers with the neighboring tubes. We can
estimate the thickness of the hydrodynamic boundary layer
as
8= yvxlu"
where x is the distance from the stagnation point at the
front of the tube. One-quarter of the way around the tube
we could estimate the boundary layer thickness as
8 = yv(rrD/4)lu"
= DINRe1!2
The boundary layers from adjacent tubes in the same row
would interact when
D
8 ="2 (St - 1), for interaction
In terms of the Reynolds number, this means that interac-
tion occurs when
in Figure 11 and compared with the correlation used suc-
cessfully for staggered tube bundles. Clearly the correlation
fails and the effect of Sl and St must be included in any
correlation for heat transfer in in-line tube bundles.
CONCLUSIONS
Recent and not so recent forced convection heat transfer
data have been collected and used to develop some minor
variations on the traditional correlations. The parallel heat
transfer process suggested by Richardson for bluff bodies
appears to provide a satisfactory approach to correlating
data for spheres, cylinders, packed beds and tube bundles.
A proper choice of the characteristic length and velocity
for packed beds and tube bundles has led to a single cor-
relation which satisfactorily predicts heat transfer rates in
randomly packed beds and staggered tube bundles.
ACKNOWLEDGMENT
In order to assemble the data presented in this paper, many
investigators were asked to diligently search their files for the
required tabulated data. They are too many to list here; they
have my thanks.

tI
NRe~y2/(St-1) =2
DIRECTION
This crude order of magnitude estimate indicates that for
OF FLOW
low Reynolds numbers we can expect deviations from the
Reynolds number dependence given by Equation (58),
and the low Reynolds number data have been fit with the
expression
NNu =2 N Re l/3 N pr1l3 (/Lbl/Lo)O.14 (59)
This Reynolds number dependence was chosen on the
basis of the Sieder-Tate correlation for heat transfer to
laminar How in a tube.
4. The data of Pierson for E = 0.87 and E = 0.91 clearly
lie above the correlation. Although it is not obvious in
•
• • •
Figure 9 these results indicate that the measured Nusselt
numbers are about twice the value predicted by Equation
(58), thus the correlation does not hold for large values
of the void fraction. We should also note that the data of
Pierson for E = 0.50 and E = 0.65 are in excellent agree-
ment with the correlation. The failure of the correlation at
high void fractions is in agreement with the packed bed

• • • •
data where the results of McConnachie and Thodos for
E = 0.78 were higher than the value predicted by the cor-

relation.
5. If one examines the work of an individual group of
investigators one will generally find a correlation for each Fig. 10. Geometry for an in-line tube bundle.
tube bundle configuration. This dependence on geometry
is not completely removed by the proper choice of charac-
teristic length and velocity; h::Jwever, the variation of NNu (> D.P. Bellelin, A. P. COUlli'll, aJIII H. L. Hull. .~ 0.50 ~ 0.65, 0,1
detected by the same investigator for different geometries :;': ~:~!: ::~ f Ten Rows 01 Tubes

is comparable to the variation detected by different investi- • Ber,.lin, L.llhton, Lafferty, and PI,iord <=0.50-0.65,011 .0

gators for the same geometry, thus it would appear to be ~: ~:~: ~.: f Ten Rows 01 Tubes

10' eH.N.Falrchild and C.P.'elch <=0.48-0.69, Au

satisfactory at this time to use a single correlation for all
-. ~ =:.~: !::: ~::: ~:~: ~:~: ::~ t Tel
the staggered tube bundle geometries proVided E ~ 0.65. o Fairchild and .elch • = 0.91, SI' ~ 1.50,'; = 600
Rows 01 Tubes

--:
." .:
4Fairchild and lelch .• ~ 0.90, Sf ~ 2.00, St = 4.00

~~
! 10
.Rr;' • •

FLOW IN IN·LlNE TUBE BUNDLES .,. •• Solid Curve IS Given By

N.jNlp,1tPbi".)O u= O.5Nk~ + 0 2N~.l
The geometrical configuration for in-line tube bundles Dashed CUlve is Given By
is illustrated in Figure 10, and once again our dimensional N• .IHlp,l~/p.)ou=2NIR:

analysis would indicate that

NNu = F(NRe, N pr, /Lbl/LO' Sb St) N••

The data for several investigators (1, 12, 13) are plotted Fig. 11. Heat transfer to fluids flowing past in-line tube banks.

AIChE Journal (Vol. 18, No.2) March, 1972 Page 369

NOTATION .p = gravitational potential energy function, sq.cm./
sec. 2
av = surface area per unit volume, sq.cm./cu.cm. 'Ii = (T T It;, Tin, measured temperature function
Ap = surface area of a particle or tube, sq.cm. b2 - bl)

A = cross sectional area, sq.cm. Subscripts

A" = characteristic area, sq.cm. o = evaluated at the wall or surface
Avoid= void area, sq.cm.
= constant pressure heat capacity, cal.I g._ C. /) = evaluated at the bulk temperature
Cp 0
% = evaluated at the free stream temperature
D = diameter, cm. f = evaluated at the film temperature
Dp = 6 Vpl Ap, particle diameter
G = mass velocity, g. cm./sec. Superscripts
g = gravity vector, sq.cm./sec. = characteristic value
h = film heat transfer coefficient, cal.! (sq.cm.) (sec.)
( °C.)
h = log mean film heat transfer coefficient, cal.! (sq.
ln
LITERATURE CITED
cm.) (sec.) (0 C.) I. Bergelin, P. 0" A. P. Colburn, and H. L. Hull, 1\'0. 2,
i = (hlc G)N p dimensionless parameter
pr2/3, Univ. Delaware Eng. Expt. Station (June, 1950).
k = thermal conductivity, cal./(cm.) (sec.) (OC.) 2. Bergelin, O. P., M. D. Leighton, W. L. Lafferty, and R. L.
L = length, cm. Pigford, Bull. No.4, Univ. Delaware Eng. Expt. Station,
L" = characteristic length, cm. (April, 1958).
n = unit outwardly directed normal vector :3. Bird, R. B., W. E. Stewart, and E. N. Lightfoot, "Trans-
port Phenomena," Wiley, New York (1960).
N = number of particles 4. Christiansen, E. B., and S. E. Craig, AIChE j., 8, 154
N Re = u"L" Iv, Reynolds number ( 1962).
NRe,L = uxLlv, length Reynolds number .5. Christiansen, E. B., C. E. Jensen, and Fan-Sheng Tao,
NRe,x = u.xlv, length Reynolds number ihid., 12, 1196 (1966).
N = cp/Llk, Prandtl number
Pr G. Churchill, S. W., and J. C. Brier, Chem. Eng. Progr.
N = hL" Ik, Nusselt number
Nn
Symp. Sel'. No. 51, 57 (1955).
7. Colburn, A. P., and C. A. Coghlan, Trans. ASME, 63, 5Gl
N Be = dimensionless parameters appearing in the bound- (1941).
ary conditions 8. Colburn, A. P., Trans. Am. Inst. Chem. Engrs., 29, 174
11 = fluid pressure, dynes/sq.cm. ( J93:3).
p = [(p - p") I pU"2] + </>IU"2, dimensionless pressure 9. Davis, A. H., Phil. Mag., 47, 1057 ( 1924).
Q = overall heat transfer rate, cal./sec. 10. Deissler, R. G., and C. S. Eran, NACA Tech. Note 2G29
Q = volumetric flow rate, cu.cm./sec. ( 1952).
11. Drew, T. B., Ind. Eng. Chem., 24, 152 (19:32).
R" = hydraulic radius, cm. 12. Edwards, A., and B. N. Furber, Proc. Inst. Mech. Eng!"s.
St = transverse pitch, cm. (London) 170, 941 (1956).
S! = longitudinal pitch, cm. 13. Fairchild, H. N., and C. P. Welch, Paper No. 61-WA-250,
St = stlD, dimensionless transverse pitch presented at AS ME Ann. Mtg (1961).
S! = sr/D, dimensionless longitudinal pitch 14. Fand, R. M., Intern. J. Heat Mass Transfer, 8, 995 (1965).
15. Friend, W. L., and A. B. Metzner, AIChE J., 4, 393
T = temperature, °C.
T" = characteristic temperature, DC. (1958).
16. Gamson, B. W., George Thodos, and O. A. Hougen, "Heat,
T. = temperature of the free stream flow past a bluff Mass, and Momentum Transfer in the Flow of Gases
body, DC. Through Granular Solids," Trans. Am. Inst. Chem. Engrs.,
Tb = bulk or "cup mixing" temperature, dc. 39, 1 (1943).
To = wall or surface temperature, DC. 17. Glaser, M. B. and George Thodos, AIChE J., 4, 63 (1958).
T! = (Tb + To)/2, film temperature, DC. 18. Graetz, L., Ann. Physik, 25, 337 (1885).
t = time, sec. 19. Hilpert, Von R., Forsch. Gebiete Ingenieurw., 4, 215
u" = characteristic velocity, cm./see. (1933).
20. Kavs, W. M., A. L. London, and R. K. Lo, Trans ASME,
U = vlu", dimensionless velocity vector 76; 387 (1954).
v = fluid velOCity vector, cm./sec. 21. Kramers, H., Physica, 12, 61 (1946).
Vz = z-component of the fluid velocity vector, cm./sec. 22. Lawrence, A. E., and T. K. Sherwood, Ind. Eng. Chem.,
<vz > = average value of V z , cm./sec. 23,301 (1931).
V = volume of a tube bundle or packed bed, cu.cm. 23. McAdams, W. R., "Heat Transmission," 3rd edn, McGraw-
V = void volume of a tube bundle or packed bed, cu.
void
Hill, New York (1954).
24. McConnachie, J. T. L., and George Thodos, AIChE ]., 9,
cm.
Vp = volume of a packing particle or tube, cU.cm. 60 (1963).
25. Metzner, A. B., in Advances in Heat Transfer, J. P. Hart-
w = width of a flat plate, cm. net and" T. F. Irvine, Jr., (eds. Academic Press, New York
X, Y, Z = dimensionless rectangular Cartesian coordinates (1965).
26. Morris, F. II., and W. G. Whitman, Ind. Eng. Chem., 20,
Greek Letters 234 (1928).
27. Notter, R. H., and C. A. Sleicher, unpublished manuscript
t;,T" = characteristic temperature difference, °e. (April, 1971).
t;,T 1n =
log-mean temperature difference, DC. 28. Parmelee, G. V., and R. G. Huebsch"r, Heating, Piping,
S = hydrodynamic boundary layer thickness, cm. and Air Conditioning, 19, 8, 115, (1947).
=V IV, void fraction
void 29. Perkins, H. C., and G. Leppert, J. Heat Transfer, Tmns.
ASME, Series C, 84, 257 (1962).
() = tu" I L", dimensionless time
El = (T - T")/(To - T"), dimensioflless temperature 30. Pierson, O. L., Trans. ASME, 59, 563 (1931).
31. Pi ret, E. L., W. James, and M. Stacy, Ind. Eng. Chem., 39,
A = (ap.,1 aEl) .. I p.,", dimensionless viscosity parameter 1098 (1947).
p., = fluid viscosity, (dyne) (sec.)/sq.cm. 32. Revnolds, W. C., W. M. Kays, and J. S. Kline, Dept. of
p = fluid density, g'/cu.cm. M~ch. Engr., Stanford Univ., NACA rept. for contract
'I: = viscous stress tensor, dynelsq.cm. NAW-6494 (1957).

Page 370 March, 1972 AIChE Journal (Vol. 18, No.2)

33. Richardson, P. D., WADD TN-59-1 (1968).
34. Schlichting, H., "Boundary Layer Theory," 6th ed., Mc- ='l'=------ (A7)
Craw-Hill, New York (1960). NRe N pr(D/4L)
35. Schneider, P. J., Trans. AS ME, 79, 765 (1957). against the Reynolds number. This argument led to the lise
36. Schubauer, C. B., anu H. K. Skramstad, NACA Rept. 909, of the i-factor
1949.
37. Sieder, E. N., and C. E. Tate, Ind. Eng. Chern., 28, 1429
(1936).
38. Sherwood, T. K., and J. M. Petrie, Ind. Eng. Chern., 24,
,=
. (Tb2 - Tbl
1!.Tln
(A8)

736 (1932). as a correlating variable. If we examine Equation (A6) we see

39. Taecker, R. C., and O. A. Hougen, Chern. Eng. Progr., 45, the motivation for Colburn's warning against "plotting a func-
188 (1949). otin against itself"; however, a little analysis will show that
40. Vliet, C. C., and C. Leppert, J. Ileat Trall.I'fer, Trans. plotting the Nusselt number against the Reynolds number has
ASME, Series C, 83, 163 (1961). no disadvantages.
41. Whitaker, S., Introductioll to Fluid Mechanics, Prentice- We begin by writing Equation (A6) as
Hall, Englewood Cliffs, N. J. (1968).
42. - - , Illd. Eng. Chern., 61,14 (1969). NNu = NReNpr (D/4L) 'l' (A9)
43. Wilke, C. R., and O ..A. Hougen, Trails. Am. IlISt. Chern.
Engrs., 41, 445 (1945). where 'l' is essentially equivalent to the i-factor. The true
44. Yuge, T., J. Ileat Transfer, Trails. ASME, Series C, 82, 214 functional relation between the Nusselt number and the Reyn-
(1960). olds number will be denoted by
45. Zhukauskas, A. A., and A. B. Ambrazyavichyus, Intern. J.
Heat Mass Transfer, 3, 305 (1961). (AlO)
46. King, L. V., Phil. Trans. Roy. Soc. (Lond.), 214, 373 and from Equation (A9) we obtain the true functional relation
(1914). . between 'l' and NRe.
47. Kelsey, S. J., and E. B. Christiansen, paper presented at
the AIChE Nat. Mtg., Cincinnati (May, 1971). (All)
48. Collis, D. C., and M. J. Williams, J. Fluid Mech., 6, 357
( 1959). Let the error in the measured value of 'l' be Ii'l' so that

(AI2)
The error in the measured Nusselt number is therefore"
APPENDIX: J.FACTOR VERSUS NUSSELT NUMBER (A13)

In 1933 A. P. Colburn (8) suggested a new method of plot- and the fractional error in the Nusselt number is
ting heat transfer data so that some correspondence between
energy and momentum transfer was illustrated. This method
NRe N pr (D/4L) Ii'l'
(AI4)
consisted of plotting the now famous i-factor versus the Reyn- C(NRe)
olds so that an analogy between the friction factor and the
i-factor existed. Making use of Equation (All) we can put Equation (AI4)
FolloWing the development presented by Colburn we define in the form
the Nusselt number for turbulent flow in a pipe as IiNNu Ii'l'
(AI5)
NNu = hD/k (AI) NNu 'l'

where the film heat transfer coefficient is related to the overall
heat transfer rate by the definition
(A2)
Here 1!.To is some characteristic temperature difference and A"
is some characteristic area. These are traditionally taken to be
indicating that the fractional error in N Nu is the same as the
fractional error in 'l'. Thus there would appear to be no clear
cut advantage in plotting i-factors instead of Nusselt numbers.
This is illustrated in Figure Al where we have plotted (N NlI
- 2)/N Re Npr°.4 (/lb//lQ)1/4 against N Re • Clearly there is no
preference for this form of the correlation over that given in
Figure 4.

!!.To = !!.Tln = . .(TbZ

:. . . .:.:__ :.:..:._....:....
- Toz) - (Tbl ___ . . :. . _
- Tod
(A3)
In ( TbZ - Toz )
Tbl - TOI

A"=ltDL (A4) o VUET AND LEPPERT WATER

o KRAMERS' AIR, WATER AND OIL

If the axial transport of energy can be entirely attributed to the "- YUGE' AIR

time averaged convective transport, that is, axial conduction

and dispersion are negligible, the macroscopic thermal energy
balance for an incompressible flow yields

Q = p<vz> ( 1t~2) cp (Tb2 - Tbl) (A5)

CURVE IS GIVEN BY
Making use of Equations (A2) through (A5) allows us to (NN"-2)/NR.N~.. 4(,u.b/#O)14 ~ O.4N;/2. O.06N~.lJ
express the Nusselt number as
p<V.>(ltD2/4) c D (TbZ - Tbd p
N Nu = ....:.....:-=.:........:.......-....:..........::.-....:.......:.:--.:..:.... No.
1!.Tln (ltDL) k
(AB) Fig. AI. Heat transfer to fluids flowing past a single sphere.
=NRe Npr (D/4L) [(Tb2 - Tbd/!!.T 1n ]
At this point Colburn notes that it is the functional dependence o Here we have assumed negligible error in NRe and Np,. relative to
of (TbZ - Tbd/!!.Tln that the experimentalist seeks to deter- the error in qr.
mine and that one should plot the measured temperature func- Manuscript received June 15, 1971; revision received September 15,
tion 1971; paper accepted September 16, 1971.

AIChE Journal (Vol. 18, No.2) March,1972 Page 371