Transactions on Modelling and Simulation vol 30, © 2001 WIT Press, www.witpress.

com, ISSN 1743-355X

The overall heat transfer characteristics of a

double pipe heat exchanger: comparison of
experimental data with predictions of standard
correlations
M. A. Mehrabian, M. Hemmat
Department of Mechanical Engineering,
Shahid Bahonar University of Kernlan, Iran

Abstract
T h e single phase flow and thermal performance of a double pipe heat
exchanger are examined by experimental methods. T h e working fluid is
water a t atmospheric pressure. Temperature measurements at the inlet and
outlet of the two streams and also at an intermediate point half way between
the inlet and outlet are made. Heat is supplied to the inner tube stream by an
immersion heater. The overall heat transfer coefficients are inferred from the
measured data. The heat transfer coefficient of the inner tube flow (circular
cross section) is calculated using the standard correlations. The heat transfer
coefficient of the outer tube flow (annular cross section) is then deduced.

1 Introduction
Double pipe heat exchangers are the simplest recuperators in which heat is
transferred from the hot fluid to the cold fluid through a separating
cylindrical wall. It consists of concentric pipes separated by mechanical
closures. Inexpensive, rugged and easily maintained, they are primarily
adapted to high-temperature, high-pressure applications d u e t o their
relatively small diameters.
Double pipe heat exchangers have a simple construction. They are fairly
cheap, but the amount of space they occupy is generally high compared with
the other types. The amount of heat transfer per section is small, that makes
the double pipe heat exchangers a suitable heat transfer device in
applications where a large heat transfer surface is not required.
 Transactions on Modelling and Simulation vol 30, © 2001 WIT Press, www.witpress.com, ISSN 1743-355X

Although the performance and analysis of double pipe heat exchangers

have been established long time ago. Abdelmessih, A. N., and Bell, K. J. [l]
have taken a closer look to these exchangers recently. They have
summarized some of the ex~stinglaminar flow heat transfer correlations in
circular. horizontal, straight tubes. They have studied the effects of natural
convection upstream of the bend and also the effects of secondary flow
downstream the bend.
For a fully (or almost fully) developed velocity profile in the straight tube
(upstream of the bend), where the thermal profile is not fully developed
under any conditions, Abdelmessih and Bell [l] found that both forced and
natural convection contribute to the heat transfer process according to the
following correlation:
Nu = [4.36 + 0 . 3 2 7 ( ~~r r ) l ' 7 ( ~ ) ~ . ~ ~(1)
Pw
where all physical properties (exceptp,) are evaluated at the local bulk
temperature. Nu is the local peripheral Bverage Nussult number. The term
representing the forced convection effect (4.36) will be recognized as the
analytical result for fully developed laminar flow with constant properties
and constant wall heat flux. The data used to generate Eq.1 covered the
following ranges:
120 5 Re 5 2500
3.9 5 Pr 5 110
2500 < Gr < 1130000 (2)
X
275-5171
d,
Eq. 1 has an absolute average percent deviation from the data of 5.98'Y".
Downstream from the bend, in addition to the forced and natural
convection contributions, there is a secondary flow contribution. Adding the
term correlating this effect gives the final equation [2]:

1,955 10-6 ~ , bl ~ , 0 8 ~ 4 . 0 7 2 5 % X l d , ) pb 0.14

l(-) (3)
P,
The data covered the following ranges:
120 5 Re I 2500
 Transactions on Modelling and Simulation vol 30, © 2001 WIT Press, www.witpress.com, ISSN 1743-355X

Eq. 3 has an average absolute deviation of 9.9% in comparison to the data.

Eq. 3 has three limiting cases, as the curvature tends to zero, the Dean
number tends to zero and Eq. 3 reduces to Eq. 1 . The second case is the
absence of natural convection, i. e., the Grashof number reduces to zero.
The third case is for the fully developed velocity and temperature profiles in
a straight tube and the absence of natural convection, i , e.. both the Grashof
number and the Dean number tend to zero, then Eq. 3 reduces to:
NU = 4.36(-)p b 011
(5)
P,,
F o r a nearly uniform wall heat flux, there will always be a natural
convection contribution; i. e. Grashof number will not approach zero, unless
gravity approaches zero.

2 Experimental setup
The experiments on which this work is based, were performed for the case of
nearly uniform heat flux at the surface. This condition of uniform heat flux
is probably closer to representing the practical condition of operating double
pipe heat exchangers in concurrent flow, where the outside fluid heats the
fluid in the pipe, or vice versa. Constant heat flux should not be confused
with constant surface temperature, where the latter is closely approximated
when there is a phase change in one of the fluids, or when the fluids are in
CO-currentflow.
The experimental rig was designed and constructed in the Heat Transfer
Laboratory, Department of Mechanical Engineering. Kerman University. A
schematic diagram of the rig circuit is shown in Fig. 1 . Some geon~etrical
data about the exchanger are listed in Table 1 . The double pipe heat
exchanger is in the vertical position; it is bent 90 degrees twice.

Tank

I Pump
Fig l Schematic diagram of the rig circuit
 Transactions on Modelling and Simulation vol 30, © 2001 WIT Press, www.witpress.com, ISSN 1743-355X

6 10 Con~putariotzalMethods and E.rperimet~a1Measures

Table 1: The exchanger geometrical data

The inner tube inner diameter 16.5 mm

The inner tube outer diameter 21.5 mm
The outer tube inner diameter 27.5 mm
The inner tube height 1 650 mm
The outer tube height 1 600 mm
Total exchanger length 1500 mm
External tube area 0.101316 m2
The tube material steel

Water is heated by an immersion heater. A constant speed pump is used to

p u m p the hot water from the tank into the inner tube. Water returns to the
tank through valve 5 . T h e cold water is supplied through the mains a n d
drains through valves 2 a n d 3 in the counterflow a n d parallel flow
conditions respectively. The counterflow conditions are achieved by shutting
the valves 1 and 3 and opening the valves 2 and 4, while the parallel flow
conditions are obtained by shutting the valves 2 a n d 4 a n d opening the
valves 1 and 3.
Temperatures are measured at the inlet a n d outlet regions of the
exchanger using copper-constantan thermocouple wires. The locations of
thermocouples are so designed that the cold and hot stream temperatures at
each terminal are measured a t the same cross section. The insulating
material covering the outer tube is 1.5 cm thick. It can be assumed that n o
heat from the hot stream dissipates into the atmosphere.

3 Experimental procedure and results

3.1 Experimental procedure
T h e overall characteristics of t h e exchanger unit a r e investigated
experimentally. The steps taken are as follows:
Measuring the temperatures of water at the inlet and outlet sections and
also at a n intermediate point half way between the inlet and outlet for each
stream, using copper-constantan thermocouple wires.
Measuring the water flow rate for each stream using calibrated rotameters.
Rotameters have been tested manually by measuring the amount of fluid
collected in a vessel in a certain amount of time at room temperature.
Rotameters have stainless steel floats.
Calculating the overall rate of heat transfer in the exchanger assuming
heat losses from the outer tube stream t o be negligible. Therefore the
overall rate of heat transfer is equal t o either the heat released from the
hot stream or the heat absorbed by the cold stream, namely:
Q = (WCAT), = (WCAT), (6)
 Transactions on Modelling and Simulation vol 30, © 2001 WIT Press, www.witpress.com, ISSN 1743-355X

Conprtatiorlal Methods and E.~perin~etital

Measures 611

Calculating the log-mean temperature difference between the two streams.

The total heat transfer rate from the hot fluid to the cold fluid in the
exchanger is expressed as:
Q = UA(LMTD) (7)
Calculating the overall heat transfer coefficient a t different operating
conditions assuming to be constant throughout the exchanger, using Eq.
(A.5).
Calculating the film heat transfer coefficient for the inner tube side flow,
using the Dittus-Boelter [3], correlation:
Nu = 0.023~ e ' Pr"
. ~ (8)
The value of n is 0.3 if the inner tube side fluid is being cooled and 0.4
if the inner tube side fluid is being heated. The Dittus-Boelter correlation
is valid for fully developed turbulent flow (Re>10000) in smooth tubes for
fluids with Prandtl numbers ranging from about 0.6 t o 100 and with
moderate temperature differences between the wall a n d the fluid
conditions. The other restriction for Eq. 8 is that it is used when constant
heat flux boundary condition is applied.
Calculating the film heat transfer coefficient for the outer tube side from
Eq. (A.2).

3.2 Experimental results:

T h e experimental procedure for each run was t o set a pre-defined
temperature and flow rate for the hot water stream, set the cold water flow
rate and then wait for the steady state conditions to be reached. Following
steps 1 to 7 for each run provides a value for the heat transfer coefficient of
the outer tube flow. Repeating the experiment for different operating
conditions, results in a set of tabulated data. Table 2 contains a range of 8
operating conditions measured as described in steps 1 and 2. The first 4 rows
correspond to counterflow conditions while the second 4 rows correspond to
parallel flow conditions.
Table 2. Temperature and flow rate measurements
 Transactions on Modelling and Simulation vol 30, © 2001 WIT Press, www.witpress.com, ISSN 1743-355X

Table 3 contains the heat transfer characteristics calculated as described in

steps 3 to 5. Table 3 corresponds to the data listed in Table 2.

Table 3. Heat transfer characteristics of the exchanger

Calculating the film heat transfer coefficient for the tube side flow as
mentioned in step 6, requires one to know viscosity, Reynolds number,
Prandtl number, Nusselt number, and conductivity of water. These data for
operating conditions corresponding to Table 2 are listed in Table 4. The
inner tube side heat transfer coefficients based on Eq. 8 and the outer tube
side heat transfer coefficients based on Eq (A.2) are listed in Tables 5 and 6
respectively.

Table 4. Inner tube side heat transfer coefficient deductions

4 Comparison of data with standard correlations

4.1 The Hausen correlation:
The Hausen correlation [4] may be used for the outer tube flow, in this case
the hydraulic diameter is:
 Transactions on Modelling and Simulation vol 30, © 2001 WIT Press, www.witpress.com, ISSN 1743-355X

Where D, is the outer tube inner diameter and Du is the inner tube outer
diameter. for the case at hand D, is 6 mm. The Hausen correlation reads a s
follows:

where G z is the Graetz number defined as follows:

D
Gz = Re Prf
L
The factor 1.20 appeared in Eq. 10 takes care of the uniform heat flux
boundary condition a t the surface of the existing heat exchanger. This is
because the Hausen correlation is for isothermal wall.
Table 5 includes viscosity. Reynolds number, Prandtl number, Nusselt
number and conductivity of the outer tube stream corresponding to the
operating conditions listed in Table 2.

Table 5. Outer tube heat transfer properties

The outer tube heat transfer coefficients evaluated according to the Hausen
correlation and using the data listed in Table 5 are included in Table 6. The
outer tube heat transfer coefficients can now be compared u i t h those
evaluated experimentally. The comparison is illustrated in Figs. 2 and 3.
 Transactions on Modelling and Simulation vol 30, © 2001 WIT Press, www.witpress.com, ISSN 1743-355X

6 14 Conlyututior~ulMethods and E.\per.in~erlralMeuswes

Table 6. Comparison between the experimental data and the standard

correlations for laminar flow in the outer tube

Gz= Ka s[5]
h0 h, h, an J
[W / m2."c] [W / m 2,'C] [W / ~ ? , O C ] Sellars,
experimental and
data, Eq. (A.5) Eq. 10 Eq. 1 1 Klein[6]
N~~~~~

It should be mentioned that some different combinations of standard

correlations have been recommended to predict the film heat transfer
coefficients of inner tube flow and outer tube flow by different workers. For
example the ESDU [2] and the Kern [7] correlations are used to predict the
inner tube side and the outer tube side heat transfer coefficients, respectively
in the TASC [8] computer program.
It is evident that the results obtained by the independent workers are
different from each other. The Dittus-Boelter [3] and Hausen [4] correlations
have been used to predict the film heat transfer coefficients of inner tube
flow and outer tube flow, respectively, a n d the results have been compared
with those deduced experimentally in Table 6. T w o other correlations are
used to predict the film heat transfer coefficients of the outer tube side in the
same manner as the Hausen correlation was applied and the results will be
compared with those deduced experimentally in Table 6. In these two cases
the Dittus-Boelter [3] correlation is monotonically used to predict the film
heat transfer coefficients of the inner tube side.

4.2 The Sieder-Tate Correlation:

The Sieder-Tate [9] correlations have been used to design the double pipe
heat exchangers since 1950 and they are strongly recommended by Kern [7]
in his old but reliable text. The Sieder-Tate correlations can be used for
predicting the film coefficients of flow in both the inner tube side and the
outer tube side of a double pipe heat exchanger. They can be used for both
heating and cooling of a number of fluids, principally petroleum fractions, in
horizontal and vertical tubes.
 Transactions on Modelling and Simulation vol 30, © 2001 WIT Press, www.witpress.com, ISSN 1743-355X

Conlpirrutionul Methods and E.tper-imerltul Meu~uves 6 15

Nu = l.2O[O.O27R ~ O Pr - (-)O P
' '$1 (13)
P,
The factor 1.20 appeared in Eqs. 12 and 13 takes care of the uniform heat
flux boundary condition at the surface of the existing heat exchanger. This is
because the Sieder-Tate correlations are for isothermal wall. Eq. 12 applies
for laminar flow (Re 2100), while Eq. 13 takes care of turbulent flow. L is
the total heat transfer length. The outer tube heat transfer coefficients are
recalculated based on Eq. 12. The results are included in Table 6 and the
comparison is shown in Figs. 2 and 3.
T h e Heat Exchanger Design H a n d b o o k , H E D H [IO], has also
recommended the Sieder-Tate correlations to be used for predicting the film
coefficients of single-phase flow in both the inner tube and the outer tube of
a double pipe heat exchanger.

4.3 The Kays and Sellars, Tribus and Klein predictions:

Kays [ 5 ] and Sellars, Tribus, and Klein [6] calculated the total and average
Nusselt numbers for laminar entrance regions of circular tubes for the case
of a fully developed velocity profile. The results of these analyses are shown
in Fig. 4 in terms of the inverse of Graetz number [l l]. The outer tube heat
transfer coefficients are recalculated based on the above predictions. The
results are included in Table 6 and the comparison is shown in Figs. 2 and 3.
Fig. 2. The shell side heat transfer
coefficients-Counterflow

h l000
.Test
500 = Eq. 11
Eq. 12
1000 1500 2000 2500

Fig. 3. The shell side heat transfer

coefficients-Parallel flow

h
500
m Eq. 11
0 Eq. l 2
1000 1100 1200 1300 1400 1500 Kays
Re
 Transactions on Modelling and Simulation vol 30, © 2001 WIT Press, www.witpress.com, ISSN 1743-355X

Fig. 4. Average Nusselt numbers for circular tube thermal

entrance regions in fully developed larninar flow [l21

0.0001 0.001 0.01 0.1

Inverse Graetz number

5 Conclusions
The outer tube side heat transfer coefficients deduced from experimental
data are compared with those evaluated based on standard correlations. The
comparison is illustrated schematically in Figs. 2 and 3 for counterflow and
parallel flow conditions. respectively.
In both the counterflow and parallel flow conditions, all three standard
correlations predict lower heat transfer coefficients compared with the
experimental results. The Sieder-Tate [9] correlation predicts the highest
values among the three standard correlations. they are still lower by a factor
of 1.04 to 2.64. That means, the standard correlations for laminar flow in the
outer tube side in which the Nusselt numbers are proportional to Re0j3
underestimate the heat transfer coefficients.
The outer tube side Reynolds numbers are lower by a factor of 6 to 10,
but the hydraulic diameter is lower by a factor of nearly 3. Therefore the
outer tube side heat transfer coefficients would be expected to be similar t o
the tube side heat transfer coefficients. The agreement with predictions is not
bad. The discrepancy may be because of three reasons. Firstly, there was
probably heat transfer in regions between the thermocouple wires and the
exchanger terminals, so that the actual heat transfer area was larger than
calculated. Secondly, there was probably a higher coefficient in certain
regions. such as in turnaround region, than predicted by straight pipe
equation. Thirdly. the effect of natural convection in internal flows,
especially when the forced and free convection currents are in the same
direction (aiding flow), can enhance the heat transfer coefficients by a factor
 Transactions on Modelling and Simulation vol 30, © 2001 WIT Press, www.witpress.com, ISSN 1743-355X

of 1.41 compared with the case when the heat transfer mechanism is
assumed strictly on the basis of laminar forced convection [l l].
Temperatures have been rounded t o the nearest decimal. S t r e a m
temperature differences ranged from 3 to 18 degrees. Heat loads, deduced
from these temperature differences are likely to be in error by 1 t o 2 percent.
With this error in heat load, the error in the measured heat transfer
coefficients is 3-6 percent.
Hewitt et al [l21 and also H E D H [4] have recommended that the Dittus-
Boelter correlation with n=0.4 to be used for predicting the film coefficients
of both the inner tube side and the outer tube side flows for both laminar
and turbulent flows. Using the Dittus-Boelter correlation with Reo"or the
laminar flow in the outer tube side increases the heat transfer coefficients by
a factor 2 to 3. That makes us believe that Hewitt et a1 [6] and also H E D H
[4] are correct.
T h e experimental heat transfer coefficients d o n o t have a direct
relationship with the outer tube side Reynolds numbers. This behaviour is
not repeated by any of the standard correlations. This is because the
experimental heat transfer coefficients are governed by the overall heat
transfer coefficients (Eq. 5), rather than by the Reynolds number.
The purpose of this article is to recognise the mechanisms of heat transfer
that occur in double pipe heat exchangers and to report higher heat transfer
coefficients in the laminar flow regime.

References

[ l ] Abdelmessih, A. N., and Bell, K. J . :Effect of Mixed Convection and U-

Bends on the Design of Double Pipe Heat Exchangers, Heat Transfer
Engineering, vol. 20, no. 3, 1999.
[2] Engineering Science Data Unit, Paper 67016, 1967; Paper 68007, 1968;
Paper 68006 with amendment A. 1970; 4, Hamilton Place.
[3] Dittus. F. W.. and Boelter L. M . K. : Unibersity of California (Berkeley)
Pub. Eng., vol. 2, p. 443, 1930.
[?l Hausen, H . : Darstellung des Warmeuberganges in Rohren durch
verallgemeinerte potenzbeziehungen, VD1 Z., no. 4, p.91, 1943.
[5] Kays, W.M.: Convective heat and Mass Transfer, pp. 187-190. McGraw-
Hill, New York, 1966.
[6] Sellars, J.R., Tribus, M., and Klein J.S.: Heat transfer to laminar flows
in a round tube o r flat conduit: The Graetz problem extended, Trans.
ASME, vo1.78, p.441, 1956.
[7] Kern, D. Q. :Process Heat Transfer, MacGraw-Hill, New York. 1950.
[8] HTFS News Sheet. no.6, December 1985.
[9] Sieder, E. N., and Tate, F. E . : Heat Transfer and Pressure D r o p in
Tubes. Ind. Eng. Chem., vol. 28, pp. 1329-1436, 1936.
[l01 Schlunder, E . U., Editor-in-Chief: Heat Exchanger Design Handbook,
 Transactions on Modelling and Simulation vol 30, © 2001 WIT Press, www.witpress.com, ISSN 1743-355X

6 18 Conzputatiotial Merhod~and E.uperimmta1 M e a ~ i r r v ~

Section 3.2, Hemisphere Publishing Corporation, New York, 1989.

[l l] Holman, J.P.: Heat transfer, eighth edit~on,McGraw-Hill, New York .
(121 Hewitt, G . F., Shires, G. L., and Bott, T . R . :Process Heat Transfer, C R C
Press, London, lSBN 0-8493-991 8-1,1994.
[l31 Lord, R . C., Minton, P. E., and Slusser, R. P,, Union Carbide Corp.:
"Design of Heat Exchangers" in Process Heat Exchange, MacGraw-
Hill, New York, 1970.

Appendix A
The heat transfer rate in a composite cylindr~calwall is expressed in terms of
the total temperature difference and the resistance of different layers:
Q / A = (Th - T c ) /(l / h , A , +(In r, / r,)/(2nkL) + 1/ hoA,) (A.1)
The overall heat transfer coefficient is defined as the inverse of the sum of
resistances to heat flow, so that Eq.(A. I) reduces to:
Q = UA(1, - T,) (A.2)
where U can be defined in terms of either the internal or external tube areas.
In each case Eq. (A.2) applies:

The values of ' I and *o are as follows:

U, =11(1/h, + ( A , l n r , / r , ) l ( 2 n k l ) + A , / h , ~ , )
And
U, = I / ( A , I A , ~+(A,
, 1nr, /r,)/(2rrkL)+ l / h,) (-4.5)
T, -T, is the temperature difference between the hot and cold fluids at a
local point. This temperature difference varies with position along the path
of flow. T o determine the rate of heat transfer between the hot and cold
fluids is therefore a complicated matter. In practice it is convenient to use an
average effective temperature difference for the entire heat exchanger. Only
if the overall heat transfer coefficient, U, is constant, this average effective
temperature difference turns out t o be the logarithmic mean temperature
difference, LMTD, defined as:
LMTD =((Thl -T'2)-(Thl -TLI))/
(In[(T,, - T',)/(Th, - TJI) (A.6)
The subscripts used in Eq.(A.6) are consistent with those used in Table 2 and
therefore apply for parallel flow conditions only. In counterflow conditions,
however, the subscripts 1 and 2 for the cold fluid temperatures must be
replaced.