Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Energy 203 (2020) 117806

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Thermal calculations of plateefineand-tube heat exchangers with


different heat transfer coefficients on each tube row
Dawid Taler , Jan Taler *, Marcin Trojan
Cracow University of Technology, Faculty of Environmental and Power Engineering, ul. Warszawska 24, 31-155, Cracow, Poland

a r t i c l e i n f o a b s t r a c t

Article history: In finned tube heat exchangers, the first rows of tubes are most effective when the air velocity in front of
Received 9 January 2020 the exchanger is less than about 3.5 m/s. The heat transfer coefficient (HTC) decreases with each sub-
Received in revised form sequent row and stabilizes only from the fifth row onwards. The paper examines a two-pass double-row
6 April 2020
plate-fin and tube heat exchanger (PFTHE) made of circular or oval pipes. A method for determining the
Accepted 5 May 2020
air side Nusselt number on individual pipe row was developed, using the results of CFD (Computational
Available online 10 May 2020
Fluid Dynamics) modelling of the heat exchanger. Also, a heat transfer correlation was found for the
mean HTC for the whole PFTHE using CFD modelling. The correlations based on the results of CFD
Keywords:
Tubular heat exchanger with plate-fins
modelling match very well the empirical correlations based on experimental tests of two car radiators.
CFD modelling The analysis shows that the first rows of PFTHE are the most effective in terms of thermal effectiveness.
Air-side nusselt number By building a heat exchanger with two rows of pipes, it is possible to reduce investment expenditures at
Different heat transfer correlation on the same thermal output significantly. Heating systems with PFTHE air heaters with fewer rows of pipes
specific pipe row are characterized by lower fuel consumption due to their higher efficiency. The investment costs for
Empirical heat transfer correlation automotive radiators can also be reduced. The introduction of PFTHE heat exchangers with fewer pipe
rows has a positive impact on the environment. Both material expenditures on the construction of
PFTHE, energy consumption by the air fan, and fuel consumption for heating purposes are reduced.
© 2020 Elsevier Ltd. All rights reserved.

1. Introduction the calculation of the Nusselt number on the air side and the inner
surface of the pipe [6].
Finned tube heat exchangers are extensively used as water Most often, for a given heat exchanger, the experimental cor-
heaters in steam boilers, evaporators and condensers in air- relations are determined for the Nusselt number on the air side,
conditioning systems [1], air heaters of the water, heat recovery while for the calculation of HTCs on the internal surfaces of the
steam generators (HRSG) after gas turbines [2], water coolers in dry pipes, the relationships available in the literature are adopted.
cooling systems in power plants, and evaporators in air-source heat Frequently used correlations are those of the power type: Dittus-
pumps [3]. They are also commonly used as the finned air coolers, Boelter and Sieder and Tate [7], Taler [8] and the Gnielinski corre-
car coolers, and air heaters in vehicles [4]. Finned heat exchangers lation [9] which is a modification of Kirillov and Petukhov rela-
or PFTHEs are usually made by the expansion of openings in plate tionship [10].
fins. In the case of aluminium fins, they can be extruded directly It should be stressed, however, that the liquid side correlation
from the aluminium pipe [5]. These fins are very thermally efficient significantly influences the air side correlation. Choosing different
because there is no thermal resistance between the outer surface of correlations on the liquid side, one obtains different relationships
the pipe and the fin base. Design calculations, as well as the on the air side [11]. This defect in the experimental method is
simulation of the steady-state and transient operation of PFTHEs because the same overall HTC, which is determined experimentally,
with numerical models, require knowledge of the relationships for can be obtained for different HTCs at the pipe inner and outer
surfaces [6]. If the HTC at the tube inner surface is too large, the HTC
on the pipe outer surface must be lower, and conversely, if the HTC
* Corresponding author. Energy Department Faculty of Environmental and Power at the inner surface is less, then the HTC at the outer surface must
Engineering Cracow University of Technology, Al. Jana Pawła II 37, 31-864, Cracow, be higher [11]. Taler [11] suggested that the correlations on the
Poland. internal and external surfaces of pipes should be determined in
E-mail address: jtaler@pk.edu.pl (J. Taler).

https://doi.org/10.1016/j.energy.2020.117806
0360-5442/© 2020 Elsevier Ltd. All rights reserved.
2 D. Taler et al. / Energy 203 (2020) 117806

parallel (simultaneously). However, the experimental determina- the air flow velocity at the exchanger inlet was between 1.5 and
tion of correlations in heat exchangers is costly. It requires the 3.5 m/s and a turbulent regime when the air velocity was higher
construction of a sophisticated research stand and time-consuming than 3.5 m/s. The results of the computer simulation were
flow and heat measurements. confirmed by means of experimental tests. The correlation was
Chen and Lai [12] determined the average HTC experimentally derived for the calculation of the average HTC in the multi-row
in the double-row PFTHE with a staggered pipe arrangement. The PFTHE. Chen et al. [18] simulated the heat exchange under natu-
method of least squares was used to determine the distribution of ral convection conditions using ANSYS FLUENT v15.0. Two pipes
HTC on the surface of the continuous fin so that the calculated fin placed in one horizontal plane were modelled. The pipes were
temperatures at several selected points of the fin were as close as interconnected with plate fins. A correlation was proposed to
possible to the measured. The fin surface temperature was calcu- calculate the average Nusselt number on the air side as a function of
lated using the finite difference method. Correlations for the Rayleigh’s number. Bezaatpour and Rostamzadeh [19] analysed the
calculation of the average Nusselt number in the entire heat possibility of improving heat transfer in PFTHE by using nanofluid
exchanger were developed, which are well in line with the corre- Fe3O4/deionised water and by applying an external magnetic field.
lations determined by other authors using the naphthalene subli- The nanofluid flowed perpendicularly to the axis of three 2 mm
mation method. Gonz alez et al. [13] offered an experimental diameter cylinders, which were situated in-line. CFD modelling was
procedure to determine the average HTC in the entire heat carried out under the assumption of a laminar flow regime as the
exchanger. A double-row PFTHE made of round pipes arranged in- Reynolds number based on cylinder diameter was less than 30. An
line was examined. The heat exchanger pipes were replaced by full increase in PFTHE’s output was achieved by using magnetite
cylinders, which were electrically heated. The surface temperature nanofluid, in particular by the combined application of the nano-
of all pipes was assumed to be constant. The authors developed the fluid and the external magnetic field in comparison with the ca-
correlation for the average Nusselt number for the whole heat pacity of the PFTHE when the working medium was deionised
exchanger, which is in good accordance with other correlations water. The heat flow rates exchanged by finned tubes located in the
available in the literature. first, second and third row were not indicated.
Unger et al. [14] tested a finned tube under natural convection In recent years, direct numerical simulation (DNS) has also been
conditions. The spacing of fins was much larger compared to the applied to determine the air side correlation for the Nusselt number
exchangers where the air flow through the heat exchanger is pro- in PFTHs [20]. DNS calculations require long calculation times. The
vided by a fan. Thanks to the large spacing of the fins, air flows simulation of heat transfer and heat transfer in a small part of a heat
through the heat exchanger as the air side pressure drop is small. exchanger takes more than a month.
The object of the experimental research was a pipe with oval fins The literature review carried out shows that in most experi-
electrically heated from the inside. The effect of the pipe’s angle of mental PFTHEs studies, the liquid side correlations known from the
inclination in relation to the horizontal plane on the average HTC literature were used, which were derived for straight pipes. CFD
was investigated. Heat transfer correlation was proposed to calcu- simulations, on the other hand, are limited to determining the HTC
late the mean air side Nusselt number as a function of Rayleigh’s relationships only on the air side. CFD modelling was usually per-
number. formed on a small piece of heat exchanger, assuming a constant,
Markovi c et al. [15] developed a new correlation for the Darcy- arbitrarily adopted temperature of the external surface of the pipe
Weisbach friction factor in the multi-row PFTHE. The proposed and even the same temperature of the external surface of the pipe
correlation is of great practical importance as it can be used to and fin. The influence of the liquid side HTC on the temperature of
calculate the power of the motor driving the fan, which causes the the pipe and fins was omitted, which in turn reduced the accuracy
air flow through PFTHE. of determining the air side HTC. It should also be emphasized that
CFD modelling is often used to check whether the proposed in both experimental studies and CFD modelling the average HTC
modifications to the shape of the fins improve the heat transfer on was determined for the whole exchanger and not for each pipe row.
the air side without significantly increasing the pressure drop in the The obtained Nusselt number correlations based on CFD modelling
heat exchanger. can, therefore, be applied for the heat exchanger of a specific design
CFD simulation is also increasingly applied to determine the that has a defined number of pipe rows.
mean HTCs or heat flow rates in PFTHEs. The computing power of Only Rich [21] and Halici et al. [22] showed experimentally by
modern computers is too small to model the entire PFTHE. Usually, investigating the four-row PFTHE with a staggered pipe array that
only the flow and heat transfer on the air side are simulated for a the Colburn parameter and heat transfer coefficient are the highest
little repetitive fragment of the PFTHE. The velocity of the air in on the first row of pipes when the air velocity before PFTHE is less
front of the PFTHE is in real systems, frequently lower than 3.5 m/s. than about 3.5 m/s. The HTC decreases in successive rows of pipes
For this reason, the air side heat transfer can be modelled, assuming up to the fourth row. In the fifth and further rows of pipes, the HTC
that the air flow in the heat exchanger is laminar if the air velocity is the same. The differences in HTC on the first four rows of pipes
before the PFTHE is less than approximately 3.5 m/s. Sun et al. [16] are higher the lower is the air velocity. In Rich’s research, the
proposed a new fin design. Air vortexes forming behind the pipe in Reynolds number, in which the distance between successive rows
the area of the rear stagnation point are eliminated by directing a of tubes in the direction of air flow was adopted as a hydraulic
more significant airflow through channels to the area behind the diameter, corresponding to an air velocity of 3.5 m/s, was 15,000. In
pipe. In this way, a dead zone in which the heat transfer between the study of Halici et al. [22], the hydraulic diameter was calculated
the whirling air and the fin surface is almost zero is eliminated. The using the formula proposed by Kays and London [23]. The limit
effectiveness of the proposed new fin configuration was tested by value of the Reynolds number below which a decrease in the HTC
CFD modelling using ANSYS FLUENT v18.1. The air flow was was observed on the following four rows of pipes was about 1500.
modelled as laminar as the air velocity in front of the test tube The paper presents a novel method to determine individual heat
changed from 1 to 3 m/s. The purpose of the CFD simulation carried transfer correlations for each pipe row in the tubular cross-flow
out by Li et al. [17] was to assess the effectiveness of radial winged heat exchangers. The heat exchangers can be made of individu-
inlets used to enhance the heat transfer on the air side in the ally finned tubes or plate-fins and tubes. The HTC on a given pipe
examined PFTHE. A laminar heat flow regime was adopted when row is determined from the nonlinear algebraic equation obtained
by comparing the air temperature increase on a given pipe row
D. Taler et al. / Energy 203 (2020) 117806 3

calculated analytically and using CFD modelling. Using the same


approach, the average air side HTC was determined for the entire
two-pass double-row-PFTHEs made of round and oval pipes. The
agreement between the CFD based correlation and the experi-
mental relationship is very satisfactory both for round and oval
pipes. The method developed can be used for determining air side
HTCs in tubular heat exchangers of different designs, e.g., PFTHEs in
air conditioning systems, car radiators, air-heaters, superheaters in
steam boilers and individually finned tube heat exchangers used in
various industries.

2. Method of determining HTC on the specified tube row

A method for determining the mean HTC on a selected tube row


on the air side will be proposed, which is consistent with the
experimental procedure for the determination of HTC. Determi-
nation of HTCs by CFD simulation can be made more straightfor-
ward in the first, second, and subsequent rows of pipes by assuming
uniform liquid temperature along the length of the pipe.
This assumption influences the heat flow rate transferred from
the liquid to the air. However, the influence of this assumption on
the searched air-side HTC value is negligibly small. Such an
assumption is possible because the HTC on the air side depends
mainly on the air velocity. The effect of changes in the tube surface
and fins temperature along the tube length on the average HTC on a Fig. 1. Repeatable fragment of a one-pass double-row heat exchanger.
given row of tubes is small. A constant HTC value was adopted on
the liquid side, i.e., on the tube internal surface.
temperatures of both pipes differ little. The HTC values on the water
In the case of experimental determination of the water-air
side are assumed to be the same in the first and second pipe. They
Nusselt correlation on the basis of the indirectly measured heat
can be determined using the correlations available in the literature,
flow rate, unknown parameters occurring in the water and air
e.g., the Dittus-Boelter, Gnielinski, or Taler formulas. The water
correlations should be determined simultaneously (in parallel). If,
temperature in both pipes was assumed to be Tw . The governing
for example, we take from the literature the heat transfer correla-
differential equations for the air temperature in the first and the
tion for calculating the water-side heat transfer coefficient for
second pipe row are as follows
straight pipes, it may turn out that the heat transfer coefficient
determined on the air side may be too large or too small, despite
  h
the same overall heat transfer coefficient in a given measurement d TaI yþ  i
series. þ
1
¼ NaI Tw  TaI yþ
1 0  yþ
1 1 (1)
d y1
The situation is different with the proposed method for deter-
mining the heat transfer coefficient of the first and second row of
  h
pipes. The air-side heat transfer coefficient depends mainly on the
d TaII yþ  i
air velocity and to a very small extent on the temperature of the fins þ
2
¼ NaII Tw  TaII yþ
2 0  yþ
2 1 (2)
and the outer surface of the pipe. Due to the high value of the heat
d y2
transfer coefficient, the temperature of the outer surface of the pipe
is similar to that of water. Thus, we can set the heat transfer coef- where: yþ þ
1 ¼ y1 =p2 y2 ¼ y2 =p2 - local non-dimensional coordinates
ficient on the inner surface of the pipe, e.g. 1000 or 2000 W/(m2K) within the first and second row of pipes, TaI , TaII - temperature dis-
without much influence on the determined air-side heat transfer tribution of the air over the width of the first and second row of
coefficient. The overall heat transfer coefficient and the heat flow pipes, Tw - water temperature.
rate transferred from water to air are different in both cases, but Local coordinatesy1 and y2 are shown in Fig. 1. The NTU -
this does not have a greater impact on the determined air-side heat numbers (number of transfer units) NaI and NaII are given by
transfer coefficient.
In the case of the experimental determination of the water and .  . 
air-side heat transfer coefficient for a given set of measurement NaI ¼ U I Ao m_ a cIpa ; NaII ¼ U II Ao m_ a cIIpa (3)
data, the heat flow rate and the overall heat transfer coefficient are
specified. where: U I and U II - overall HTC for the pipe situated in the first and
The proposed procedure will be presented on the example of a second row of pipes related to the surface area Ao ¼ Po L of the plain
two-row heat exchanger (Fig. 1). Extending it to more rows of pipes pipe without fins, m_ a - air mass flow per pipe, i.e., flowing through a
is easy. cross-section with a surface area equal to p1  L , Po - outer
The water temperature at the inlet to the first and second row perimeter of the plain pipe without fins (for a circular pipe Po ¼
pipes is the same. The water temperatures at the outlets of the first pdo ), L - pipe length, p1 - pitch of pipe spacing perpendicular to the
and second pipes do not differ much. Also, the water temperature
direction of air flow, cIpa - air average specific heat at constant
drop over the entire length of the pipes is small. It is, therefore,
pressure on the thickness of the first row of pipes, i.e., in the
possible to assume that the water temperatures in the first and 00
second pipes are uniform and the same. This is possible because the temperature range from Ta0 to Ta , cIIpa - air average specific heat at
temperature of the external surface of the pipes has little influence constant pressure on the thickness of the second row of pipes, i.e.,
00 000
on the physical properties of air and HTC, primarily when the in the temperature range from Ta to Ta .
4 D. Taler et al. / Energy 203 (2020) 117806

conduction. Despite the efforts of many researchers, it has not been


possible to develop a method to determine the thermal resistance
between the pipe and the fin base that has a satisfactory accuracy.
There is also a lack of data on the value of this resistance in the
literature, due to the different technologies of fin attachment to the
pipe. Therefore, both the design and performance calculations of
plate-fin and tube heat exchangers do not include this resistance.
The effective heat transfer coefficients hIoe and hIIoe are defined as
follows
   
Abf Af  I  Abf Af  II 
hIoe ¼ hIa þ h h hIIoe ¼ hIIa þ h h (5)
Ao Ao f a Ao Ao f a

where: Af , Abf - the area of fins and the outer surface area of the
plain pipe between fins, hf - efficiency of the fin.
The air temperature Ta0 in front of the first pipe row and the
00
temperature Ta in front of the second row of pipes are known. The
boundary conditions for equations (1) and (2) are therefore as
follows

TaI j yþ ¼0 ¼ Ta’ (6)


1
Fig. 2. Flow layout of the two-pass engine radiator with two rows of tubes; 1 and 3
designates the first tube row in the first and second pass, 2 and 4 stand for the second
tube row in the first and second pass. TaII jyþ ¼0 ¼ Ta} (7)
2

The solution of equation (1) with boundary condition (6), ob-


tained by separating the variables, has the following form
1 Ao 1 Ao dw 1 1 Ao 1 Ao dw 1  
¼ þ þ ¼ þ þ    0 
U I Ain hw Awm kw hIoe U II Ain hw Awm kw hIIoe TaI yþ I þ
1 ¼ Tw þ Ta  Tw exp  Na y1 0  yþ
1 1 (8)
(4)
By substituting yþ
1 ¼ 1 into equation (8) one obtains the air
00
where: Ain - pipe inner surface area, Ao - pipe outer surface area, temperature Ta behind the first row of pipes, which is also the inlet
Awm ¼ ðAin þAo Þ=2 - mean surface area, hw - HTC on the internal temperature to the second row of pipes
surface of the pipe, hIoe , hIIoe - effective HTC on the external pipe    0   
00
surface considering the presence of fins for the pipe within the first Ta ¼ TaI yþ I
1 ¼ 1 ¼ Tw þ Ta  Tw exp Na (9)
and second row of pipes, dw - pipe wall thickness, kw - pipe material
thermal conductivity. The solution of Equation (2) with boundary condition (7), ob-
Considering that the pipe and fins are made of non-corrosive tained by separating variables, has the following form
aluminium, it was assumed that the contact between the pipe    0  h  i
and the fin is perfect. The contact resistance between the tube and TaII yþ I II þ
2 ¼ Tw þ Ta  Tw exp  Na þ Na y2 0  yþ
2 1
fin base was neglected in Eq. (4).
(10)
There are also reasons of a more general nature that make it
difficult to take such thermal resistance into account, even if it By substituting yþ
2 ¼ 1 into the expression (10) one obtains the
occurs. The thermal resistance between the outer surface of the 000
air temperature Ta behind the second row of pipes (the air tem-
pipe and the fin base cannot be included in Eq. (4) since the tem- perature behind the heat exchanger)
perature distribution in the pipe underneath the fin base is two-
   0  h  i
dimensional and cannot be taken into account in Eq. (4), which 000
Ta ¼ TaII yþ I II
2 ¼ 1 ¼ Tw þ Ta  Tw exp  Na þ Na (11)
was derived with the assumption of one-dimensional heat
The air temperature increases on the first DTI and second DTII
row of pipes are given by the following expressions

Fig. 3. Heat exchanger section modelled using ANSYS-CFX 19.1 program; a) geometry,
b) boundary conditions types. Fig. 4. The finite element mesh for the modelled area.
D. Taler et al. / Energy 203 (2020) 117806 5

Table 1
Specifications of the finite element grid for the area depicted in Fig. 3.

Number of the finite elements 456,474

Number of nodes 320,131


The minimal dimension of the element, mm 0.1
The maximal dimension of the element, mm 0.4
Boundary layer First layer thickness: 0.17 (yþ ¼ 1)
Growth Rate: 1.2
Number of layers: 16
Minimum Orthogonality Angle Air: 21.2
Fin: 82.1
Pipe1: 86.6
Pipe2: 86.6
Mesh expansion factor Air: 22.9
Fin: 4
Pipe1: 2
Pipe2: 2
Maximum aspect ratio Air: 257
Fin: 40
Pipe1: 24
Pipe2: 24

at the outlet and the inlet temperature from a repeatable heat


00  h  i
DTI ¼ Ta  Ta0 ¼ Tw  Ta0 1  exp NaI (12) exchanger fragment.
The air temperature rise DTt; i ð ha; i Þwas calculated using the
00
  h  i following Equation (Fig. 1)
DTII ¼ Ta 0 0 0  Ta ¼ ðTw  Ta 0 Þ exp NaI 1  exp NaII
  0 000
(13) DTt;i ha;i ¼ DTI;i þ DTII;i ¼ Ta;i  Ta;i (17)
Equations (12) and (13) were derived using the relationships (9) Taking into account that the air side HTC ha; i is uniform for both
and (11), respectively. The HTC hIa; i on the first row of pipes for the I ¼ N II , then the expression for the overall
pipe rows, e.g. Na;i ¼ Na;i a;i
i-th measurement series was found solving the following non- rise in the air temperatureDTt;i ¼ DTI;i þ DTII;i over the two-row
linear algebraic equation heat exchanger reduces to
 
DTI; i hIa; i ¼ DT I; i i ¼ 1; …; n (14)    
DTt;i ¼ Tw  Ta0 1  exp 2 Na;i (18)

where DTI; i ðhIa; i Þ is calculated using the relationship (12), and DT I; i The temperature differences DTI;i and DTII;i are defined by
is the value of the mass average temperature rise over the first row Equations (12) and (13).
of pipes, which was calculated using ANSYS-CFX software for the The mass average air temperatures behind the first and second
assumed air velocity w0;i before the heat exchanger. The symbol n row of pipes are determined by CFD modelling.
The air temperature behind the first row of pipes can also be
in Eq. (14) stands for the number of measurement data sets.
calculated from formula (9) after determining the heat transfer
The HTC hIIa; i on the second pipe row was found similarly solving
coefficient from the non-linear algebraic Eq. (14). Once the alge-
the following algebraic equation for hIIa; i braic Eq. (15) has been solved and the heat transfer coefficient on
  the second pipe row has been determined, the air temperature
DTII; i hIIa; i ¼ DT II; i i ¼ 1; …; n (15) behind the second pipe row can be calculated using formula (11).
After determining the HTCs ha; i i ¼ 1; …; n by solving Eq. (16), the

where DTII; i ðhIIa; i Þ is calculated from the formula (13) after deter- Nusselt numbers NuIa; i NuIIa; i Nua;i and Reynolds numbers
Rea; i ¼ ðwmax; i dha Þ=na; i were determined for i ¼ 1;…;n, where n is
mining hIa; i . The symbol DT II; i denotes the value of the mass
average temperature rise in the second pipe row, which is calcu-
lated using the ANSYS-CFX software for the assumed air velocity Table 2
w0;i before the heat exchanger. The procedure described above was Temperature increases on the 1st and 2nd pipe row and the whole exchanger for 11
different air velocities before the heat exchanger determined by ANSYS CFX 19.1
used as well to determine the average HTC in the entire heat software.
exchanger. The value of the average HTC ha; i in the complete heat
i w0;i , m/s DTI; i , K DTII; i , K DTt; i , K
exchanger for i-th input data and mass-average air temperature rise
1 0.5 30.54 4.68 35.22
DT t;i in the heat exchanger was selected so that the total air tem-
2 0.7 26.90 6.08 32.98
perature rise DTt; i ð ha; i Þcalculated from the analytical formula was 3 0.9 23.86 6.91 30.77
equal to the mass-average temperature increase DT t;i determined 4 1.1 21.32 7.29 28.61
5 1.3 19.22 7.42 26.65
using CFD simulation 6 1.5 17.48 7.39 24.88
  7 1.7 16.02 7.27 23.29
DTt; i ha; i ¼ DT t; i i ¼ 1; …; n (16) 8 1.9 14.79 7.09 21.88
9 2.1 13.73 6.88 20.61
The CFD-based mass-average temperature difference DT t;i was 10 2.3 12.81 6.66 19.47
11 2.5 12.00 6.44 18.44
calculated as the difference between the mass average temperature
6 D. Taler et al. / Energy 203 (2020) 117806

Table 3
HTCs for 1st and 2nd row of circular pipes and the whole heat exchanger with corresponding Colburn factors.

i w0;i , m/s Rea hIa; i hIIa; i ha;i jIa; i jIIa; i ja;i


W/(m2K) W/(m2K) W/(m2K)

1 0.5 102.23 60.92 22.74 39.58 0.048159 0.017977 0.031289


2 0.7 143.51 68.96 30.78 47.72 0.038947 0.017384 0.026952
3 0.9 185.02 74.04 36.8 53.46 0.032531 0.016169 0.023489
4 1.1 226.73 77.12 40.94 57.22 0.027729 0.014721 0.020574
5 1.3 268.61 79.26 44.06 59.96 0.024119 0.013408 0.018246
6 1.5 310.61 80.84 46.4 62.02 0.021323 0.012239 0.01636
7 1.7 352.72 82.08 48.22 63.62 0.019106 0.011225 0.01481
8 1.9 394.91 83.2 49.78 65 0.017331 0.01037 0.01354
9 2.1 437.17 84.12 50.96 66.08 0.015855 0.009606 0.012456
10 2.3 479.49 84.94 51.92 66.98 0.014619 0.008937 0.011529
11 2.5 521.85 85.66 52.68 67.74 0.013565 0.008343 0.010728

the number of data sets. The air-side Nusselt numbers NuIa; i , NuIIa; i the oval finned pipe on the air side and na; i is the kinematic vis-
cosity of air at mean temperature. The maximum air velocity wmax;i
and Nua;i are given by the following expressions: NuIa; i ¼ hIa; i dh a =
appears in the narrowest passage between adjacent pipes. The
ka;i , NuIIa; i ¼ hIIa; i dh a =ka;i and Nua;i ¼ ha;i dh a =ka;i where the symbol diameter dha on the air side [23] and the maximum velocity wmax;i
ka;i stands for the air thermal conductivity. were calculated using the following expressions
The symbol dha designates the equivalent hydraulic diameter of

Fig. 5. Colburn factors ja versus the air side Reynolds number Rea for 1st and 2nd pipe row, and the whole heat exchanger; a) 1st pipe row, b) 2nd pipe row, c) complete heat
exchanger.
D. Taler et al. / Energy 203 (2020) 117806 7

1=3 1=3
NuIa; i =ðRea; i Pra; i Þ jIIa; i ¼ NuIIa; i =ðRea; i Pra; i Þ, and
4A p p s T a;i þ 273:15 1=3
dha ¼ 0 min 0 2 wmax;i ¼  1 0 w0;i ja;i ¼ Nua;i =ðRea; i Pra; i Þ for i ¼ 1, …,n were calculated. Then, the
Af þ Abf s  df ðp1  dmin Þ Ta;i þ 273:15 calculated Colburn factors jIa;i jIIa;i , and ja;i were approximated by the
(19) power-type relationships as functions of Rea using the least-
squares method. Nusselt number correlations NuIa NuIIa , and Nua
where: Amin ¼ ðs df Þðp1 dmin Þ - the narrowest section of flow were obtained by transforming the relevant expressions for the
area between two adjacent fins, A0f ¼ 2ðp1 p2 Ah Þ - surface area of Colburn factor. Thermal and flow simulations were carried out for
the fin with a hole for the pipe, Ah - surface area of a circular or oval two PFTHEs, one of which was made of round pipes and the other of
cut-out for a pipe in the fin (for round pipe Ah ¼ p d2o = 4), A0bf ¼ oval pipes.
Po ðs df Þ - the area of the pipe outer surface between two adjacent
fins, p1 - pipe spacing pitch transverse to the direction of the air 2.1. CFD simulation of a PFTHE made of round pipes
flow, s- the pitch of fins, df - fin thickness, dmin - smaller diameter of
The studied PFTHE was a car radiator designed to cool a 1580
the oval pipe, T a;i - mean temperature of the air in the entire heat
0 - air temperature before the PFTHE.
cubic centimetre internal combustion engine. Three forms of liquid
exchanger, Ta;i
flow inside the pipes were present during the studies: laminar,
For round pipes, the minimum diameter of the pipe dmin should transition, and turbulent. The car cooler has two passes and two
be replaced by the outer pipe diameter do. The hydraulic diame- rows of pipes (Fig. 2). There are 20 pipes in the upper pass and 18 in
terdha, as defined by the expression in Eq. (19) proposed by Kays the lower pass. In the first and second row of exchangers, there are
and London [19], is a modification of the classical formula for an 19 pipes each. Circular pipes are arranged inline [24]. Both rows of
equivalent diameter of a channel with a cross-section different pipes in the first pass are fed in parallel. Water from the first and
from a circular one. The definition of hydraulic diameter takes into second row, after cooling down, flows into the manifold, where it is
account that in a rectangular duct formed by two adjacent fins mixed. Both rows of pipes in the second pass are supplied with
there are pipes in which water flows. At first, the values of Colburn’s water mixed. At the outlet of the radiator, the water from the first
parameters for the first, second, and complete heat exchanger jIa; i ¼ and second row of pipes of the second pass is mixed again in the

Fig. 6. Nusselt number Nua versus the air side Reynolds number Rea for 1st and 2nd round pipe row, and the whole heat exchanger; a) 1st pipe row, b) 2nd pipe row, c) complete heat
exchanger.
8 D. Taler et al. / Energy 203 (2020) 117806

Fig. 7. Repeatable fragment of PFTHE made of oval pipes used in CFD modelling.

header. The spacing between aluminium pipes of external diameter


Table 4
dout ¼ 7:2 mm and wall thickness dt ¼ 0:5 mm are as follows: Characteristics of the finite element grid of PFTHE made of oval tubes for the
pitch perpendicular to the direction of air flow p1 ¼ 18:5 mm, and calculation using the ANSYS-CFX software.
pitch parallel to the direction of air flow p2 ¼ 12 mm.
Number of elements 1,380,300
The spacing of the plate fins is s ¼ 1:5 mm, and their thickness
Number of nodes 1,551,152
df ¼ 0:08 mm. Continuous fins (plate fins) are also made of
Maximum size of the element, mm 0.4
Boundary layer First layer thickness: 0.017 mm (yþ ¼ 1)
Growth Rate: 1.2
Number of layers: 6
Minimum Orthogonality Angle Air: 65.6
Fin: 74.4
Pipe1: 89.3
Pipe2: 84.0
Mesh expansion factor Air: 4
Fin: 3
Pipe1: 2
Pipe2: 2
Maximum aspect ratio Air: 16
Fin: 15
Pipe1: 7
Pipe2: 7

Table 5
Data for CFD simulation of the PFTHE made of oval pipes.

i w0;i , m/s Tw;i , oC 0 , oC


Ta;i hw; i , W/(m2$K)

1 0.96 85.3 0.73 1512


2 1.21 83.9 10.54 2121
3 1.45 87.5 10.49 2169
4 1.61 82.5 10.47 2076
5 1.77 79.7 12.41 3559
6 2.12 78.1 13.81 3549
7 2.12 78.8 12.63 3556
Fig. 8. Division of the modelled section of the PFTHE made of oval pipes into finite
8 2.11 85.0 11.04 2084
elements, a) longitudinal section of the flow passage between two adjacent fins, b) an
9 2.13 83.0 11.04 2047
enlarged fragment of the finite element mesh in the passage between fins.
D. Taler et al. / Energy 203 (2020) 117806 9

Table 6 shown above. Mesh number 2 was made of 532,950 elements and
Temperature increases on the 1st and 2nd pipe row and the whole exchanger for 9 373,765 nodes. The boundary layer consisted of 8 layers, and the
different air velocities before the heat exchanger determined by ANSYS CFX
software.
thickness of the first layer was 0.017 mm for the parameter yþ ¼ 1.
Mesh number 3 was made of 392,968 elements and 274,675 nodes.
i Rea DTI; i , K DTII; i , K DTt; i , K The boundary layer consisted of 8 layers, and the thickness of the
1 152.79 55.883 12.614 68.497 first layer also was 0.017 mm for the parameter yþ ¼ 1. Relative
2 183.62 46.406 12.612 59.018 differences in temperature increments calculated on each pipe row
3 220.28 44.214 14.169 58.383
did not differ more than 0.7% for all three analysed grids.
4 246.58 38.367 13.47 51.837
5 268,02 38.22 1344 51.66 The results of calculations of mass average temperature in-
6 320.51 33.05 13.122 46.172 creases on the 1st, 2nd row of pipes, and over the whole heat
7 322.62 33.309 13.219 46.528 exchanger for 11 different air velocities w0;i before the heat
8 324.02 33.014 13.882 46.896
exchanger are listed in Table 2.
9 327.78 31.737 13.441 45.178
Based on the determined temperature differences, the HCTs for
the 1st and 2nd pipe row, and the whole heat exchanger were
calculated. Also, the corresponding Colburn factors were deter-
aluminium. The length of the pipes in the exchanger is 520 mm,
mined (Table 3).
and the active height of the radiator is 359 mm. The thickness of the
By the approximation of the Colburn factors using the method of
radiator in the direction of air flow is 2p2 ¼ 24 mm. The air side
the least squares, the Colburn factor relationships versus the Rey-
hydraulic diameter is equal to dha ¼ 1:95 mm. The CFD modelling
nolds number Rea were obtained for circular pipes (Fig. 5).
of the PFTHE on the air side was carried out with the software
The approximation function for all three Colburn parameters is
ANSYS-CFX 19.1. Only the repetitive fragment of the heat exchanger
as follows
shown in Fig. 3 was modelled due to the limited computing power
of the computer. Fig. 3 also shows the applied boundary conditions ja ¼ x1c Rexa2c 100  Rea  525 (20)
types which were used for the CFD computations.
Dimensions of the modelled heat exchanger section are (Fig. 3) The values of parameters x1c and x2c for 1st and 2nd pipe rows
s  p1  2p2 . The finite element mesh for the area to be modelled is and the entire heat exchanger are as follows.
depicted in Fig. 4.
Specifications of the finite element mesh for the area depicted in  1st pipe row, x1c ¼ 1:6502±0:32714, x2c ¼ 
Fig. 3 are listed in Table 1. 0:7586±0:03786
In the air domain discreet model, “Air at 25  C” for Reference  2nd pipe row, x1c ¼ 0:1569±0:07882, x2c ¼ 
pressure 1atm was used as a material. The heat transfer was 0:4501±0:09283
modelled using the “Total Energy” option. Laminar flow in the  Complete heat exchanger, x1c ¼ 0:6070±0; 17896, x2c ¼ 
PFTHE was modelled because the air velocity in front of the 0:6322±0:05556
exchanger is low and does not exceed 2.5 m/s. The Reynolds Rea
number based on the hydraulic diameter on the air side and the The transformation of the expression (20) for the Colburn factor
maximum air velocity in the tested radiator does not exceed 522. gives
The inlet velocity was set to be the same throughout the inlet cross-
ð1þx2c Þ 1=3
section. The air temperature at the inlet was assumed to be 20  C, Nua ¼ x1c Rea Pra 100  Rea  525 (21)
and the air velocity at the inlet (before the PFTHE) changed be-
The following correlations for the Nusselt numbers on the air
tween 0.5 and 2 m/s. The outlet was defined using the Opening
side for the individual pipe rows and complete PFTHE were ob-
option with Relative Pressure ¼ 0 Pa and Opening Temperature
tained using Eq. (21)
equal to the mass average air temperature at the outlet from the
modelled element. For the Pipe1, Pipe2, and Fin domains, the CFX ð1=3Þ
library material Aluminium was adopted as the material from NuIa ¼ 1:6502Re0:2414
a Pra 100  Rea  525 (22)
which the pipes and fins are made. The heat transfer was modelled
ð1=3Þ
using the Thermal Energy option. NuIIa ¼ 0:1569Re0:5499
a Pra 100  Rea  525 (23)
The boundary conditions of the third kind were set on the inner
surfaces of the pipes (Fig. 3). The bulk temperature of the water was ð1=3Þ
60  C and the HTC on the pipe inner surface was 2000 W/(m2$K). A Nua ¼ 0:6070Re0:3678
a Pra 100  Rea  525 (24)
mesh independent study of the CFD results was carried out. Three By the approximation of the Nusselt number using the method
different calculation meshes were tested. Mesh number 1 was of the least squares, the relationships for the Nusselt number Nua as

Table 7
HTCs for 1st and 2nd row of oval pipes and the whole heat exchanger with corresponding Colburn factors.

i w0;i , m/s Rea hIa; i W/(m2K) hIIa; i ha; i jIa; i jIIa; i ja;i
W/(m2K) W/(m2K)

1 0.96 152.79 156.739 44.406 80.756 0.064146 0.015725 0.031003


2 1.21 183.62 151.634 51.127 86.238 0.0425 0.017075 0.026995
3 1.45 220.28 145.152 58.147 91.035 0.03638 0.01584 0.024011
4 1.61 246.58 140.245 62.512 93.453 0.033801 0.01472 0.022246
5 1.77 268.02 137.970 66.552 96.152 0.027446 0.01626 0.021244
6 2.12 320.51 131.689 75.243 101.033 0.023475 0.014 0.018223
7 2.12 322.62 130.998 75316 100.858 0.023565 0.014043 0.018282
8 2.11 324.02 130.380 75.267 100.615 0.026849 0.012429 0.018188
9 2.13 326.45 130.019 75.884 100.961 0.026788 0.012268 0.018039
10 D. Taler et al. / Energy 203 (2020) 117806

Fig. 9. Colburn factors ja versus the air-side Reynolds number Rea for 1st and 2nd oval pipe row, and the whole heat exchanger; a) 1st pipe row, b) 2nd pipe row, c) complete heat
exchanger.

a function of the Reynolds number Rea were obtained for round which the air flows are as follows: width 520 mm, height 359 mm,
pipes (Fig. 6). The values of the Nusselt numbers were calculated and thickness 34 mm. The wall thickness of the oval aluminium
using the HTCs determined from the CFD simulation. Correlations pipes is dt ¼ 0:4 mm. The smaller diameter of the outer surface of
(22), (23) and (24) differ slightly from the corresponding correla- the oval pipe is do; min ¼ 6:35 mm and the larger diameter
tions shown in Fig. 6a, b and 6c. This is due to a different form of do; max ¼ 11:82 mm. The equivalent hydraulic diameter of the pipe
function that approximates the data received by means of CFD on the water side is dh w ¼ 7:06 mmand on the air side dh a ¼
simulation. For correlations (22)e(24) the function ja ¼ ja ðRea Þ was 1:41 mm. The number of continuous fins with the thickness
used, and for correlations in Fig. 6 the function Nua ¼ Nua ðRea Þ was df ¼ 0:08 mm on the length of the radiator is 520, and the fin pitch
applied. is 1 mm. The water stream to the radiator is evenly distributed over
The determined correlations (22)-(23) can be used in PFTHE 20 first pipes of the first pass. The cooled water from the outlets of
calculations taking into consideration the various correlations on the first and second row of pipes is mixed in the intermediate
each pipe row. Equation (24) will be compared with an appropriate manifold and then uniformly distributed between the 18 pipes of
correlation determined experimentally. the second pass. The water at the outlet from the second pass is
mixed in the outlet manifold, from which it flows into the cooled
2.2. CFD simulation of a PFTHE made of oval pipes water tank.
The modelling of the flow and heat transfer in the PFTHE was
The car cooler is used to cool down the coolant that is heated carried out with the software ANSYS-CFX 19.1. Only a repeatable
from the cylinder block and engine head. As a result of engine part of the heat exchanger illustrated in Fig. 7 was modelled due to
cooling, the engine oil has excellent lubricating properties and the the limited computing power of the computer.
thermal stresses in the engine head are significantly reduced. The Dimensions of the modelled exchanger fragment s  p1  2p2
flow system of a radiator made of oval pipes is identical to that of a are depicted in Fig. 7. The mesh of the finite elements used in
radiator built of round pipes, as described in subsection 3.1. The modelling the repeatable fragment of the PFTHE illustrates Fig. 8.
radiator is a two-pass PFTHE, and the water flow path is in the letter The mesh of the finite elements used in modelling the repeat-
U shape. The car radiator is double-row. The number of pipes in able fragment of the PFTHE illustrates Fig. 7. A mesh with the
each row is 19, with 10 pipes in the first pass and 9 in the second following parameters was used for the calculation using the ANSYS-
pass. The dimensions of the active part of the radiator through CFX software (Table 4):
D. Taler et al. / Energy 203 (2020) 117806 11

Fig. 10. Nusselt number Nua versus the air side Reynolds number Rea for 1st and 2nd oval pipe row, and the whole heat exchanger; a) 1st pipe row, b) 2nd pipe row, c) complete heat
exchanger.

Also, a verification calculation was carried out for another mesh. The values of parameters x1c and x2c for 1st and 2nd pipe rows
Mesh number 2 was made of 7,701,300 elements and 8,520,176 and the complete heat exchanger are as follows.
nodes. The boundary layer consisted of 6 layers, and the thickness
of the first layer was 0.017 mm for the parameter yþ ¼ 1. The  1st tube row, x1c ¼ 30:7105±42:7230, x2c ¼  1:2400±0:2601
compatibility of the results obtained for two different grids was  2nd tube row, x1c ¼ 0:0744±0:0937, x2c ¼ 0:2931±0:2288
very good. The maximum difference between the determined  entire heat exchanger, x1c ¼ 1:0605±0:1361, x2c ¼ 
temperatures after the 1st and 2nd pipe row was equal 0.011  C. 0:7026±0:0236‘
The experimental data used in computer simulation are listed in
Table 5. The liquid side HTC at the pipe inner surface was deter- Based on the estimated relationships for the Colburn factor, the
mined by the semi-empirical correlation proposed in Ref. [25]. Nusselt numbers for each pipe row and the entire PFTHE were
The results of calculations of mass average temperature in- obtained
creases on the 1st and 2nd row of pipes, and over the whole heat
ð1=3Þ
exchanger for 9 different air velocities w0;i before the heat NuIa ¼ 30:7105Re0:24
a Pra 150  Rea  330 (26)
exchanger are listed in Table 6.
Based on the determined temperature differences, the HTCs for
ð1=3Þ
the 1st and 2nd pipe row, and the whole heat exchanger were NuIIa ¼ 0:0744Re0:7069
a Pra 150  Rea  330 (27)
calculated, and the corresponding Colburn factors were determined
(Table 7). ð1=3Þ
By the approximation of the Colburn factors using the method of Nua ¼ 1:0605Re0:2974
a Pra 150  Rea  330 (28)
the least squares, the relationships for the Colburn factors as a By the approximation of the Nusselt number using the method
function of the Reynolds number Rea were obtained for oval pipes of the least squares, the relationships for the Nusselt number Nua as
(Fig. 9). a function of the Reynolds number Rea were obtained for oval pipes
The approximation function for all three Colburn parameters is (Fig. 10). The values of the Nusselt numbers were calculated using
as follows the HTCs determined from the CFD simulation. Correlations (26),
(27) and (28) differ a little from the corresponding correlations
ja ¼ x1c Rexa2c 150  Rea  330 (25)
depicted in Fig. 10a, b and 10c. This is caused by a different form of
12 D. Taler et al. / Energy 203 (2020) 117806

Fig. 12. Comparison of heat transfer correlation based on CFD modelling with
empirical correlation for PFTHE manufactured from oval pipes.
Fig. 11. Comparison of the air side heat transfer correlation based on CFD modelling
with empirical correlation for the entire PFTHE manufactured from round pipes.

If 3; 000  Rew , the liquid flow in the pipe is turbulent. The


Darcy-Weisbach friction factor xw was calculated using the rela-
function that fits the data obtained by CFD simulation. For heat
tionship proposed by Taler [27].
transfer correlations (26)e(28) the function ja ¼ ja ðRea Þ was used,
and for correlations in Fig. 10 the function Nua ¼ Nua ðRea Þ was
applied.
xw ¼ ð1:2776 logRew  0:406Þ2:246 3; 000  Rew  106
The determined correlations (26)e(27) can be used in heat
exchanger calculations with different correlations for each row of (31)
pipes. Equation (28) will be compared with the heat transfer cor- The air side Nusselt number Nua was approximated by the
relation established experimentally. following relationship

3. Comparison of correlations based on CFD simulations with


1=3
experimental relationships Nua ¼ x1 Rexa2 Pra 225  Rea  560 (32)

The comparison of the heat correlations determined by CFD where x1 and x2 are the unknown parameters to be determined
modelling with the correlations determined based on experimental using the experimental data. The physical properties of the air were
tests was carried out for two different car radiators. The flow 000
evaluated at the mean temperature T a ¼ ðTa0 þ Ta Þ =2. The
arrangement of both radiators is the same. The first radiator is following values of the parameters x1 , x2 , and x3 were obtained by
made of round pipes [24] and the second one of the oval pipes [26]. the method of least squares [24].

3.1. PFTHE made of round pipes


x1 ¼ 0:5446±0:1506 x2 ¼ 0:4000±0:0455 x3
The experimental relationships for the Nusselt number on the ¼ 10:5357±0:5995 (33)
water and air side were estimated by Taler and Taler [24] based on
70 data sets The numbers following the plus/minus symbol are half of the
95% confidence interval.
xw " 2=3 # A comparison of the correlation (24) obtained by CFD simulation
ðRew  2; 100Þ Pr1:008
w din
Nuw ¼ x3 þ 8 rffiffiffiffiffiffi   1þ ; with the empirical correlation (32) with parameters (33) is shown
xw  L in Fig. 11.
2=3
1:084 þ 12:4 Prw  1
8 The relative difference ε between empirical and CFD-based
correlation was calculated using the formula
din 6:2
2; 100  Rew  18; 000; 2:6  Prw  3:9; ¼
L 520
emp
(29) Nua  NuCFD
a
ε¼ emp ,100; % (34)
Nua
xw ¼ 0:030476 þ 1:45216 , 105 ðRew  2; 100Þ 2; 100  Rew
Analysis of the Nusselt numbers depicted in Fig. 11 shows that
 3; 000 there is a very good correspondence between the Nusselt number
(30) determined empirically and the Nusselt number determined by
CFD modelling. The maximum relative error ε does not exceed 9%.
D. Taler et al. / Energy 203 (2020) 117806 13

3.2. PFTHE made of oval pipes CFD modelling the air side correlations in PFTHEs can be obtained
which are in good compliance with the heat transfer correlations
The relationships for the Nusselt numbers were found using 39 determined based on experimental data. Such an approach would
measurement series. The average value of the water side Prandtl significantly reduce the costs of experimental testing of new heat
number of all 39 data sets was Prw ¼ 3:007. The following rela- exchangers or even eliminate them completely. In the future,
tionship approximated the water side Nusselt number similar studies are planned for turbulent multi-row PFTHEs. The
method developed can be applied to find specific heat transfer
xw " # correlations for each pipe row using the CFD modelling. In practice,
ðRew  2; 300Þ Pr1:008
w dhw 2=3
Nuw ¼ x3 þ 8 r ffiffiffiffiffi
ffi it allows constructing a PFTHE with an optimum pipe row number
xw  2=3  1þ L
;
so that the ratio of the PFTHE output to investment and operating
1:084 þ 12:4 Prw  1 expenditure is as high as possible. PFTHEs with fewer pipe rows do
8
not require high power fans.
dhw 7:06
2; 300  Rew  12; 000; 2:5  Prw  3:3; ¼
L 520 Declaration of competing interest
(35)
The authors declare that they have no known competing
where: Prw ¼ cpw mw =kw , Rew ¼ ww dhw =nw - water side Prandtl and financial interests or personal relationships that could have
Reynolds number, dhw - hydraulic diameter of the pipe, L - length of appeared to influence the work reported in this paper.
the pipe. For circular pipes, the diameter dhw is equal to pipe inner
diameter din.
CRediT authorship contribution statement
The formula developed by Taler and Taler [24] was used to
calculate the Darcy-Weisbach friction xw in the transition flow
Dawid Taler: Methodology, Writing - original draft, Investiga-
region
tion, Data curation. Jan Taler: Conceptualization, Methodology,
Writing - review & editing, Supervision. Marcin Trojan: Software,
xw ¼ 0:02783 þ 2:2457 , 105 ðRew  2; 300Þ 2; 300  Rew
Writing - review & editing, Visualization, Data curation.
 3; 000
(36) References

If the flow regime is turbulent, i.e. 3; 000  Rew , the model (31) [1] Mazzeo D. Solar and wind assisted heat pump to meet the building air con-
proposed by Taler [27] was used to calculate the friction factor xw. ditioning and electric energy demand in the presence of an electric vehicle
charging station and battery storage. J Clean Prod 2019;213:1228e50. https://
The air-side Nusselt number was approximated by the correlation
doi.org/10.1016/j.jclepro.2018.12.212.
[2] Taler D, Pitry R, Taler J. Operation assessment of hybrid heat source for heating
1=3
Nua ¼ x1 Rexa2 Pra 146  Rea  350 (37) the building and preparation of hot water in the fire brigade building. J Clean
Prod 2019;2014:962e74. https://doi:10.1016/j.jclepro.2018.12.259.
The unknown parameters x1 , x2 and x3 were estimated by the [3] Karvonen M, Kapoor R, Uusitalo A, Ojanen V. Technology competition in the
internal combustion engine waste heat recovery: a patent landscape analysis.
method of least squares as described by Taler [6]. The following
J Clean Prod 2016;112(5):3735e43. https://doi.org/10.1016/
values of parameters x1 , x2 and x3 were determined j.jclepro.2015.06.031.
[4] Daly S. Automotive air-conditioning and climate control systems. Amsterdam:
x1 ¼ 0:5020±0:0091 x2 ¼ 0:41±0:02661 x3 Elsevier; 2006.
[5] Kuppan T. Heat exchanger design handbook. second ed. Boca Raton: CRC
¼ 12:0607±0:6300 (38) Press-Taylor and Francis Group; 2013.
[6] Taler D. Numerical modelling and experimental testing of heat exchangers.
The numbers after the plus/minus sign indicate half of the 95% Berlin-Heidelberg: Springer; 2019.
[7] Mills AF. Basic heat & mass transfer. Upper Saddle River: Prentice Hall; 1999.
confidence interval. A comparison of the correlation (28) obtained
[8] Taler D. Simple power-type heat transfer correlations for turbulent pipe flow
by CFD simulation with the empirical correlation (37) with pa- in tubes. J Therm Sci 2017;26:339e48. https://doi.org/10.1007/s11630-017-
rameters (38) is shown in Fig. 12. 0947-2.
[9] Gnielinski V. New equations for heat and mass transfer in turbulent pipe and
The relative difference ε between empirical and CFD-based
channel flow. Int Chem Eng 1976;16:352e68.
correlation was calculated using the formula (34). The compari- [10] Petukhov BS. Heat transfer and friction in turbulent pipe flow with variable
son of the results presented in Fig. 12 shows that there is a very physical properties. In: Hartnett JP, Irvine TF, editors. Advances in heat
good agreement between the Nusselt number determined empir- transfer. vol. 6. Academic Press; 1970. p. 503e64.
[11] Taler D. Prediction of heat transfer correlations for compact heat exchangers.
ically and the Nusselt number determined by CFD modelling. The Forsch. Ingenieurwes. 2005;69:137e50. https://doi.org/10.1007/s10010-004-
maximum relative error ε does not exceed 12%. 0148-5.
[12] Chen HT, Lai JR. Study of heat-transfer Characteristics on the fin of two-row
plate finned tube heat exchangers. Int J Heat Mass Tran 2012;55:4088e95.
4. Conclusions https://doi.org/10.1016/j.ijheatmasstransfer.2012.03.050.
[13] Gonza lez AM, Vaz M, Zdanski PSB. A hybrid numerical-experimental analysis
of heat transfer by forced convection in plate-flnned heat exchangers. Appl
A new method of determining the correlation for the Nusselt Therm Eng 2019;148:363e70. https://doi.org/10.1016/
number on each pipe row in PFTHEs using CFD modelling was j.applthermaleng.2018.11.068.
[14] Unger S, Beyer M, Thiele J, Hampel U. Experimental study of the natural
developed. The average HTC on a given row of bare or finned pipes convection heat transfer performance for finned oval tubes at different tube
was determined from the comparison of the mass average air tilt angles. Exp Therm Fluid Sci 2019;105:100e8. https://doi.org/10.1016/
temperature increases, calculated by CFD modelling and obtained j.expthermflusci.2019.03.016.
[15] Markovi c S, Ja
cimovic B, Geni
c S, Mihailovi
c M, Milovan
cevic U, Otovi
c M. Air
by the analytical formula. The correlations for the air side Nusselt
side pressure drop in plate finned tube heat exchangers. Int J Refrig 2019;99:
number were found for the first and second pipe row as well as for 24e9. https://doi.org/10.1016/j.ijrefrig.2018.11.038.
the entire PFTHE using the CFD modelling. Two PFTHEs were [16] Sun Ch, Lewpiriyawong N, Loong KL, Zeng S, Lee PS. Thermal enhancement of
analysed, one of which was constructed from round pipes and the fin and tube heat exchanger with guiding channels and topology optimisation.
Int J Heat Mass Tran 2018;127:1001e13. https://doi.org/10.1016/j.ijheat-
other from oval pipes. masstransfer.2018.08.093. Part C.
One of the aims of the paper was to demonstrate that by using [17] Li MJ, Zhang H, Zhang J, Mu YT, Tian E, Dan D, Zhang XD, Tao WQ.
14 D. Taler et al. / Energy 203 (2020) 117806

Experimental and numerical study and comparison of performance for wavy [22] Halici F, Taymaz I, Gündüz M. The effect of the number of tube rows on heat,
fin and a plain fin with radiantly arranged winglets around each tube in fin- mass and momentum transfer in flat-plate finned tube heat exchangers. En-
and-tube heat exchangers. Appl Therm Eng 2018;133:298e307. https:// ergy 2001;26:963e72.
doi.org/10.1016/j.applthermaleng.2018.01.012. [23] Kays WM, London AL. Compact heat exchangers. third ed. Malabar, Florida:
[18] Chen HT, Ma WX, Lin PY. Natural convection of plate finned tube heat ex- Krieger Pub Co.; 1998.
changers with two horizontal tubes in a chimney: experimental and nu- [24] Taler D, Taler J. Prediction of heat transfer correlations in a low-loaded plate-
merical study. Int J Heat Mass Tran 2020;147. https://doi.org/10.1016/ fin-and-tube heat exchanger based on flow-thermal tests. Appl Therm Eng
j.ijheatmasstransfer.2019.118948. 2019;148:641e9. https://doi.org/10.1016/j.applthermaleng.2018.11.060.
[19] Bezaatpoura M, Rostamzadehb H. Heat transfer enhancement of a fin-and- [25] Taler D. A new heat transfer correlation for transition and turbulent fluid flow
tube compact heat exchanger by employing magnetite ferrofluid flow and in tubes. Int J Therm Sci 2016;108:108e22. https://doi.org/10.1016/
an external magnetic field. Appl Therm Eng 2020;164. https://doi.org/ j.ijthermalsci.2016.04.022.
10.1016/j.applthermaleng.2019.114462. [26] Taler D. Mathematical modelling and experimental study of heat transfer in a
[20] Nagaosa RS. Turbulence model-free approach for predictions of air flow dy- low-duty air-cooled heat exchanger. Energy Convers Manag 2018;159:
namics and heat transfer in a fin-and-tube exchanger. Energy Convers Manag 232e43. https://doi.org/10.1016/j.enconman.2018.01.018.
2017;142:414e25. https://doi.org/10.1016/j.enconman.2017.03.063. [27] Taler D. Determining velocity and friction factor for turbulent flow in smooth
[21] Rich DG. The effect of the number of tube rows on heat transfer performance tubes. Int J Therm Sci 2016;105:109e22. https://doi.org/10.1016/
of smooth plate fin-and-tube heat exchangers. ASHRAE Tran Pt1 1975;81: j.ijthermalsci.2016.02.011.
307e17. Paper no. 2345.

You might also like