Вы находитесь на странице: 1из 13

International Journal of Thermal Sciences 75 (2014) 127e139

Contents lists available at ScienceDirect

International Journal of Thermal Sciences


journal homepage: www.elsevier.com/locate/ijts

Numerical investigation of the turbulent cross ow and heat transfer


in a wall bounded tube bundle
Xiaowei Li*, Xinxin Wu, Shuyan He
Institute of Nuclear and New Energy Technology, Tsinghua University, Beijing 100084, PR China

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 12 January 2013
Received in revised form
10 July 2013
Accepted 1 August 2013
Available online 4 September 2013

Tube bundles are usually used for heat transfer in heat exchangers. The walls bounding the tube bundles
will inuence the ow and heat transfer in the tube bundles. This becomes critical for the once through
steam generators of gas cooled reactors due to it is very compact and sensitive. URANS simulation of the
cross ow and heat transfer in a wall bounded inline tube bundle is presented in this paper. The numerical method was veried with experimental measurements. The local and average ow and heat
transfer characteristics were analyzed. The ow has the intrinsic characteristics of unsteadiness as that in
a free tube bundle. Bounding walls modify the ow and heat transfer signicantly. Near wall ow
passages have lower ow resistances due to the walls suppress wakes after the tubes. The uid velocities
in the near wall passages are larger and the temperatures are higher. The wall effects depress turbulence
intensities of the ow in the near wall ow passages, so the heat transfer coefcients of the near wall
tubes are 10% lower than those of the tubes in the middle of the bundle.
2013 Elsevier Masson SAS. All rights reserved.

Keywords:
Tube bundle
Turbulent cross ow
Heat transfer
Wall effect
Steam generator

1. Introduction
High temperature gas-cooled reactor (HTGR) is especially
excellent in safety due to its inherent safety design [1]. The helical
tube once through steam generator is an important equipment in
HTGR. It should be compact and very reliable. The primary side
helium absorbs heat from the reactor core, and then transfers it to
the secondary side water through the helical tube bundles in the
once through steam generator. The shell or bounding walls over the
tube bundles will certainly inuence the ow and heat transfer in
the tube bundles. Understanding the cross ow and heat transfer
characteristics in wall bounded tube bundles is very important for
the design of the HTGR steam generator.
Turbulent cross ow heat transfer in tube bundles is a basic ow
pattern in tube and shell type heat exchangers. There are many
investigations on the turbulent cross ow and heat transfer over a
single tube or a free (big enough) tube bundle [2]. Experimental
correlations of heat transfer and pressure drop have been established [3,4]. Coutanceau and Defaye [5] did a detailed survey of the
ow visualization works of the cross ow over a single circular
cylinder. Niemann and Holscher [6] reviewed the experiments
about drag and Strouhal number of cross ow over tube bundles.

* Corresponding author. Tel.: 86 10 62784825; fax: 86 10 62797136.


E-mail addresses: lixiaowei@tsinghua.edu.cn, lixiaowei99@gmail.com (X. Li).
1290-0729/$ e see front matter 2013 Elsevier Masson SAS. All rights reserved.
http://dx.doi.org/10.1016/j.ijthermalsci.2013.08.001

Paul et al. [7] measured the velocity eld of turbulent cross ow


over tube bundles. Simonin and Barcouda [8] detailedly measured
the ow velocity distribution in a staggered tube bank, the results
showed that after four rows of tubes, the ow will be fully developed. Weavera and Elkashlana [9] recommended six rows of tubes
should be used for studying the cross ow induced vibrations in
tube banks. Zukauskas and Ulinskas [3] also recommended 4 rows
for the heat transfer to be fully developed in a free tube bank.
There are walls bounding the tube bundles in all shell and tube
heat exchangers used in thermal power and chemical engineering.
The calculation of heat transfer coefcients do not need to be so
accurate in ordinary industries. The development of HTGR steam
generator requires more accurate estimation of the heat transfer
coefcients of the wall bounded tube bundles due to it requires
high compactness, high reliability and high heat transfer rate. The
investigations of wall inuences on cross ow and heat transfer
over tube bundles are limited. Many works have been done to
investigate walls inuences on the cross ow over a single cylinder.
One of the early works was done by Bearman and Zdravkovich [10].
They experimentally investigated the ow around a cylinder placed
at various heights above a plane boundary. The results showed that
when the gaps between the wall and cylinder are less than 0.3
cylinder diameters, the regular vortex shedding was suppressed.
For dimensionless gaps greater than 0.3, the Strouhal number was
constant. Their Reynolds number was 45,000. Angrilli et al. [11]
experimentally investigated the wall induced modication to

128

X. Li et al. / International Journal of Thermal Sciences 75 (2014) 127e139

vortex shedding from a circular cylinder. The working Reynolds


numbers were 2,860, 3820 and 7,640, which was in the low
subcritical region. They found that the proximity of the wall induces a slight increase of vortex shedding frequency. Their results
were in contrast with that of Bearman and Zdravkovich [10]. They
attribute this disagreement to the different Reynolds numbers.
Buresti and Lanciotti [12] measured the mean and uctuating
forces on a circular cylinder in cross-ow near a plane surface and
found that the mean lift coefcient decreases when increasing the
gap size, while the mean drag coefcient seamed to increase. The
Reynolds number was from 86,000 to 277,000. They also found that
the Strouhal number decreased a little when decrease the gap size
for the cases with relatively thick boundary layers. Zhan et al. [13]
measured the drag and lift forces acting on a near wall circular
cylinder. The results were that the drag coefcient becomes smaller
and lift coefcient becomes greater when the gap ratio, H/D, becomes smaller. Forces acting on the cylinder tend to be constant
when the gap ratio reaches 1.0. Nishino et al. [14] detailedly
measured the vortex shedding from a circular cylinder near a
moving ground. Their measurements also showed that the drag
coefcient of the cylinder near the wall is 0.85e0.95 of that for a
free cylinder when H/D is from 2.0 to 0.05. Duan and Jiang [15]
numerically simulated the ow distribution in a wall bounded 3D
helical tube bundle using steady Reynolds averaged NaviereStokes
(SRANS), there are 5 rows of tubes in the bundle. The results
showed that the wall effect will decrease the ow rate in the ow
passages near the wall and the clearance between tube and wall
should be increased to make the ow rate distribution in the tube
bundle uniform. Aiba [16] measured the heat transfer around a
circular cylinder near a plane surface. The results showed that the
Nusselt number of the cylinder rst increases with increasing gaps,
and then decreases with increasing gaps. Miskinis et al. [17]
measured the heat transfer from a rough cylinder in cross ow
near a plane wall and found that the heat transfer rate has two
maximum values when increasing gaps between the plane and the
cylinder. Katinas and Tumosa [18] experimentally measured the
heat transfer coefcient of the tubes near the wall in a tube bundle.
The results showed that when the dimensionless clearance, H/
(S1  D), between the wall and tube bundle is below 2, the heat
transfer coefcient of the tubes near the wall was lower than that in
the center of the tube bundle, the maximum decrement can be 20%.
Numerical simulation helps a lot in understanding the ow
and heat transfer characteristics. There are many numerical
works about turbulent cross ow and heat transfer in tube
bundles, including direction numerical simulation (DNS) [19],
large eddy simulation (LES) [20e22] and Reynolds averaged
NaviereStokes (RANS). Benhamadouche and Laurence [23] and
Rollet-Miet et al. [24] reported the numerical simulation results
of tube bundles using LES and RANS. They concluded that LES
gave better results than RANS. Paul et al. [25] simulated the
turbulent cross ow in a tube bundle using commercial software
CFX. They compared the numerical results with their experiments. The results showed that the two equation turbulence
models can give good results, especially when the ow becomes
spatially periodic. Watterson et al. [26] simulated the turbulent
cross ow over a staggered tube bundle with a periodic boundary
condition using a low-Reynolds number ke3 model and
compared their results with a test case. They found that the ke3
model can give surprisingly good results and save a lot of time.
Wang et al. [27,28] simulated turbulent cross ow in a staggered
tube bundle using commercial software, Fluent. The heat transfer
prediction was found to be in reasonable agreement with
experimental data and empirical correlation. Ridluan and Tokuhiro [29,30] did a benchmark simulation of the turbulent ow
through a staggered tube bundle in the lower plenum of a very

Fig. 1. Numerical model.

high temperature reactor. The conclusions were that steady


Reynolds averaged NaviereStokes (SRANS) give poor results,
unsteady Reynolds averaged NaviereStokes (URANS) can give
acceptable results. URANS was recommended for the simulation
of cross ow over tube bundle. Bouhairie and Chu [31] introduced
their numerical work about 2D cross ow over circular cylinder.
They reviewed the 2D and 3D numerical simulation of cross ow
over a circular cylinder. Their conclusion was that although twodimensional simulations are constrained, it can in some cases
give important physical insight.
There are many investigations about cross ow and heat
transfer over free tube bundles, while the inuences of the wall
effect on the cross ow and heat transfer over a tube bundle is
limited and not consistent. From the review of the previous numerical works, two dimensional URANS is capable to predict the
cross ow and heat transfer in tube bundles. The cross ow and
heat transfer characteristics in a wall bounded tube bundle will be
simulated using two dimensional URANS. The numerical method
will be rst veried with experimental measurements and
experimental correlations.

Fig. 2. Grid distribution.

X. Li et al. / International Journal of Thermal Sciences 75 (2014) 127e139

129

Fig. 4. Grid size inuence on local velocity distribution. (a) Overall. (b) Enlarged.

Fig. 3. Tube bank of Simonin and Barcouda.

2. Numerical method
2.1. Geometry model and boundary conditions
The numerical model is shown in Fig. 1. It is a two dimensional
wall bounded tube bundle. The tubes are in inline arrangement.
There are 7 rows and 5 columns of tubes in the bundle. Column No.
1 and column No. 5 tubes are adjacent to the walls. The tube
diameter is 19 mm. The tube pitches in the ow direction are
S2 25 mm, the tube pitches in the transverse direction are
S1 30 mm. The distances between the wall and the centers of wall
adjacent column tubes are S1/2. The computational area is
175 mm  150 mm. Flow enters the tube bundles from the left. Grid
distribution is shown in Fig. 2. The grid is intensied near the wall
regions. The smallest element dimension is 0.01 mm to ensure that
y is well below 1. The total element number is 222,000. Constant
temperature and non-slip wall boundary conditions were set for
the tube walls. Adiabatic and non-slip wall boundary conditions
were used for the bounding walls. Periodically fully developed
boundary condition was assumed for the inlet and outlet. For more
detailed description of the periodic boundary condition and the
numerical methods please refer to Li et al. [32,33] and reference
[34]. The working uid was helium with constant properties. The

operating pressure is 7 MPa. The Reynolds number based on the


maximum average velocity was 38,000 for the wall bounded tube
bundle. The maximum average helium velocity in the tube bundle
was 26.67 m/s.
Another tube bundle was also simulated to verify the code.
The geometry is the same as that used in the experiments by
Simonin and Barcouda [8] in order to compare with their measurements of the velocity elds. There are 7 rows and 10 or 11
columns of tubes. The tube is in staggered arrangement as shown
in Fig. 3. The tube diameter is 21.7 mm. There are total 498,800
elements in this model. For the verication case, the maximum
average velocity was 1.06 m/s. The working uid was water with
constant properties at 20  C. The operating pressure is atmosphere pressure. The Reynolds number based on the maximum

Table 1
Grid size inuence on Nusselt number and friction factor.
Element number

442, 632

221, 948

115, 163

61, 955

Nu

152.66
0.21785

150.24
0.21637

162.17
0.17922

186.05
0.15619

Fig. 5. Velocity measure position.

130

X. Li et al. / International Journal of Thermal Sciences 75 (2014) 127e139

Fig. 6. Velocity distribution. (a) x 0, u. (b) x 0, v. (c) x 11, u. (d) x 11, v. (e) x 16.5, u. (f) x 16.5, v. (g) y 0, u. (h) y 0, v. (i) y 22.5, u. (j) y 22.5, v.

average velocity was 18,000. Uniform velocity was used for the
inlet. Fully developed condition was used for the outlet.
2.2. Governing equations and numerical scheme
The governing equations were the time-dependant threedimensional incompressible Reynolds averaged NaviereStokes
equations and the energy equation. The pressure variance in the

computational domain is less than 0.01 MPa, it is very little compared


with the operating pressure of 7 MPa, so incompressible assumption
is reasonable. Standard ke3 model and Reynolds stress model proposed in Fluent software were used for the turbulence modeling.
The governing equations together with the turbulence models
were solved using the commercial CFD code, Fluent 6.3, which uses
the nite volume method. Two-layer model was used for the near
wall turbulence modeling, so the near wall turbulence effect can be

X. Li et al. / International Journal of Thermal Sciences 75 (2014) 127e139

131

Fig. 6. (continued).

calculated in the viscous sub-layer. Velocity-pressure coupling was


treated with the SIMPLEC algorithm. The convection terms were
discretized with the QUICK scheme. The time step was 0.0001 s for
all the cases.
2.3. Grid independence analysis
The grid independence study was conducted for the case with 7
rows and 5 columns tube bundle as described in Section 2.1. The
boundary conditions were the same as in Section 2.1. Four sets of
grid were generated, the element numbers are 442, 632, 221, 948,
115, 163 and 61, 955. The calculated average Nusselt numbers and
friction factors are shown in Table 1. The increment of element
number from 221, 948 to 442, 632 only results in the increase of
1.6% for Nu and 0.7% for x. The calculated local time averaged velocity magnitude distribution along the line between the row 4
column 1 and row 4 column 2 tubes with different grid size are
shown in Fig. 4. From Fig. 4(a), it can be seen that the velocity
difference is very little among the four sets of grids. Fig. 4(b) gives a
clearer picture. When the grid size increased from 221, 948 to 442,
632, the velocity magnitude increment is below 0.2%. So grid
number of 222, 000 was chosen to be used for the simulation.
3. Code verication
3.1. Velocity eld and turbulence intensity
Correct prediction of the velocity eld and turbulence intensity
is important for the numerical analysis of the ow and heat transfer

characteristics in the tube bundle. Simonin and Barcouda [8]


detailedly measured the averaged velocity eld and root mean
square velocity uctuations of cross ow over a 2D tube bundle. The
geometry of the tube bundle is shown in Fig. 3. The measurement
positions are shown in Fig. 5. Measurements showed that the ow
was fully periodic after 4 rows of tubes.
Three different turbulent models, standard ke3 model, SST keu
model and Reynolds stress model, are used for the simulation. The
time averaged numerical results are compared with the measurements of Simonin and Barcouda [8]. The u and v velocity at the
positions of x 0, x 11 mm, x 16.5 mm, y 0, and y 22.5 mm
are shown in Fig. 6. It can be seen that the time averaged velocities
predicted by standard ke3 model and Reynolds stress model agree
well with experiments, while the SST keu model is not so good and
is not presented. The turbulence intensity at the positions of x 0,
x 11 mm, x 16.5 mm, y 0, and y 22.5 mm are shown in Fig. 7.
The agreement with the measurements is good. Standard ke3
model and Reynolds stress model will be used for the simulation in
this paper.
3.2. Strouhal number
The Strouhal number of the cross ow over the tube bundle was
analyzed using statistical method. The frequency of the drag coefcient on the tubes can be attained using the Fast Fourier Transform (FFT) analysis. The standard ke3 model and Reynolds stress
model give almost the same results. Fig. 8 shows the results of FFT
analysis. The peak amplitude of FFT of Cd represents the dominant
frequency and the Strouhal number. The Strouhal number value of

132

X. Li et al. / International Journal of Thermal Sciences 75 (2014) 127e139

Fig. 7. Turbulence intensity distribution. (a) x 0, u. (b) x 11, u. (c) x 16.5, u. (d) y 0, u. (e) y 22.5, u.

row 4 column 5 tube from the statistical analysis is shown in Fig. 8.


It can be seen that St 0.35. This agrees well with the measured
results given by Zukauskas et al. [35] with S1/D 1.58, S2/D 1.32.
3.3. Nusselt number and friction factor
A free tube bundle was also simulated to see whether the
method can predict the average heat transfer and pressure drop
well. The geometry of the free tube bundle was the same as that in

Fig. 1. The boundary conditions were almost the same except that
the up and down bounding walls were changed to periodic conditions. The ow and heat transfer were simulated with Reynolds
numbers from 10,000 to 50,000. As shown in Fig. 9, the Nusslet
numbers and friction factors are compared with empirical correlations. The Nusselt number correlation is proposed by Zukauskas
and Ulinskas [4] as in Eq. (1). The friction factor correlation is
proposed by Idelchik [36] as shown in Eq. (2). It can be seen that the
differences between the numerical and empirical correlation for

X. Li et al. / International Journal of Thermal Sciences 75 (2014) 127e139

133

Fig. 10. Drag and lift coefcients of row 4 column 3 tube.

Fig. 8. Strouhal number of the row 4 column 5 tube in the staggered tube bundle of
Simonin and Barcouda [8].

represents the lift coefcient. The drag and lift coefcients are
calculated using Eqs. (3) and (4).

both Nusselt numbers and friction factors are below 25%. The differences maybe induced by both the uncertainties of the correlations and the numerical simulations.

Cd

Fp;x Fv;x
0:5ru2N DH

(3)

Nu 0:27Re0:63 Pr 0:36

Cl

Fp;y Fv;y
0:5ru2N DH

(4)

x 0:38

W D
 0:94
LD

(1)
0:59 
0:5
WD 2
W
Re0:2 LD
1
D

(2)

The velocity eld results showed that this numerical method is


capable of predicting the ow in a tube bundle. The Strouhal
number results showed the transient characteristics of the ow in
the tube bundle can also be simulated. The numerical method can
also predict the average heat transfer and pressure drop.

where Fp,x is the x direction pressure force, Fv,x is the x direction


viscous force, Fp,y is the y direction pressure force, Fv,y is the y direction viscous force. All the forces are calculated by integrating

4. Results and discussion


The transient ow characteristics will be introduced rst. Then
the time and local averaged ow and heat transfer characteristics
will be presented and analyzed detailedly.
4.1. Transient ow characteristics
Fig. 10 is the transient drag and lift coefcients of the row 4 and
column 3 tube. Solid line represents the drag coefcient, dotted line

Fig. 9. Nusselt number and friction factor.

Fig. 11. Strouhal number. (a) Row 4 column 3 tube. (b) Row 4 column 1 tube.

134

X. Li et al. / International Journal of Thermal Sciences 75 (2014) 127e139

along the wall surfaces. D is the tube diameter, H is the tube height
in z direction, it is unit in this case.
From Fig. 10, the drag and lift coefcients vary periodically with
time. There are total 34 periods for the drag coefcient, while there
are 17 periods for the lift coefcient. The oscillation period of the lift
coefcient is 0.32/17 z 0.0188 s. Compared with the transient velocity eld, it can be found that one period represents a variation of
the wake direction. Fig. 10 also shows that the variation of the lift
coefcient is much larger than the drag coefcient. This coincides
with the experimental measurements by Zhan et al. [13]. The experiments of Buresti and Lanciotti [12] also showed that the lift
coefcient of a single cylinder will increase when a wall approximates, while the drag coefcient will decrease. The lift coefcient
can be higher than the drag coefcient when the gap between the
wall and the tube is very small.
Fig. 11(a) shows the FFT analysis of the drag coefcient of row 4
column 3 tube as shown in Fig. 10. Fig. 11(b) shows the FFT analysis

of the drag coefcient of row 4 column 1 tube. It can be seen that


the Strouhal number of row 4 column 3 tube is 0.06, while that of
row 4 column 1 tube is 0.03. This is different from that of a free tube
bundle [35]. The wall effect may be the reason. The measurements
of Bearman and Zdravkovich [10] showed that the wall suppressed
the vortex shedding frequency for a single cylinder when
Re 45,000. Buresti and Lanciotti [12] also found that Strouhal
number decreased a little when decreasing the gap size between
the tube and the wall. Besides the geometry in Fig. 1, a tube bundle
with 33 rows of tubes with inlet and outlet region was also simulated to investigate the inlet condition inuence. The tube bank
arrangement is the same as in Fig. 1. The Strouhal number attained
from the 33 rows tube bundle simulation is the same as in Fig. 11.
This means the periodic condition is applicable.
Fig. 12 shows the transient velocity elds from time t 0.4750 s
to t 0.4938 s. From the velocity elds at different times shown in
Fig. 12, we can see that the uid in the tube bundle sways as it

Fig. 12. Transient velocity eld. (a) t 0.475 s (b) t 0.4797 s (c) t 0.4844 s (d) t 0.4891 s (e) t 0.4938 s.

X. Li et al. / International Journal of Thermal Sciences 75 (2014) 127e139

passes the tubes. This shows the intrinsic characteristics of unsteadiness of the cross ow over tube bundles. From Fig. 12(a), the
wakes of the middle three columns (column No. 2, 3 and 4) of tubes
in the rst row point to upper right, the wakes of the middle three
columns of tubes in the third and fourth rows point to lower right,
the wakes of the middle three columns of tubes in the sixth and
seventh rows point to upper right, while the wakes of the middle
three columns of tubes in the second and fth rows are in transition. In Fig. 12(b), it is the tubes in the third and seventh rows are in
transition. In Fig. 12(c), the directions of the wakes of the middle 3
columns of tubes are opposite to those in Fig. 12(a), and with the
second and fth rows in transition. In Fig. 12(d), the tubes of the
fourth and seventh rows are in transition. In Fig. 12(e), the directions of the wakes of the middle 3 columns of tubes return to the
states in Fig. 12(a). So the time period of one sway of the wakes is
about 0.0188 s. This coincides with the oscillation periods of the lift
coefcient shown in Fig. 10. From the pictures in Fig. 12, we can see
that the wakes of the tubes in the middle 3 columns can sway up
and down freely, but the wakes of the tubes in the side columns
(column No. 1 and 5) are conned by the walls. They can hardly
sway to the walls. This is the difference between the present wall
bounded tube bundle and a free tube bundle.

135

Fig. 14. Flow resistance coefcients of the ow passages.

The time averaged velocity, temperature and turbulence intensity distributions in the ow passages in the tube bundle are
shown in Fig. 13. The horizontal coordinates are the ow passage
numbers, the vertical coordinates are the averaged velocity, temperature and turbulence intensity. The ow passage locations and
its numbering method are illustrated in the small picture at the top
of Fig. 13. The solid lines numbered 1 to 6 represent locations of the
ow passages, where the time and local average was carried out.
The velocity, temperature and turbulence intensity values shown in
Fig. 13(a), (b) and (c) are time and area averaged on the locations at
the solid lines numbered 1 to 6 in several ow periods. The ow
passages numbered 1 and 6 are the side passages which have the
bounding wall as one of its walls. The widths of passage No. 1 and
No. 6 are only half that of the passages numbered 2 to 5. The
averaged velocity, temperature and turbulence intensity in Fig. 13
show symmetric distribution.
From Fig. 13(a), the velocities in the middle passages (No. 2, 3, 4,
and 5) are lower than that in the two side passages (No. 1 and 6).
The side passages numbered 1 and 6 have the highest velocities.
This seems to be incorrect due to that the bounding walls compose

one of the walls of passage No. 1 and 6, which will make the viscous
forces of passage No. 1 and No. 6 larger than that of the passages in
the middle. This can be explained from the ow resistance coefcients shown in Fig. 14. The ow resistance of passage No. 1 is
calculated by the sum of the drag force on its adjacent wall and half
the drag force on tube No. 1, while the ow resistance of passage
No. 2 is calculated by the sum of half the drag forces on both tube
No. 1 and tube No. 2 (the tube number can be seen in the small
picture at the top of Fig. 15). The calculation of the ow resistance
coefcient of passage No. 6 is similar to passage No. 1, while that of
passage No. 3, 4, 5 are similar to passage No. 2. From Fig. 14, the ow
resistances of the six passages have a symmetric distribution. The
ow resistances of passage No. 1 and 6 are lower than that of
passage No. 2, 3, 4 and 5. The higher ow resistances in the middle
of the tube bundle cause the lower velocity.
Why the ow resistance of passage No. 1 and 6 are lower? This
should be explained from the time and area averaged viscous and
pressure force coefcients on the tubes and walls shown in Fig. 15.
From Fig. 15, the pressure forces on the tubes in x direction are
much larger than the viscous forces in the same direction. Also the
pressure forces in y direction on the tubes are much larger than the
viscous forces in the same direction. So the pressure forces dominate the drag and lift forces on the tubes. The rectangular symbols
in Fig. 15 clearly show that the pressure forces on the middle 3
tubes (tube No. 2, 3, 4) in x direction (the drag force) are higher than

Fig. 13. Velocity, temperature and turbulence intensity distributions in the ow passages in the tube bundle.

Fig. 15. Pressure and viscous force coefcients on tubes and walls.

4.2. Time averaged ow characteristics

136

X. Li et al. / International Journal of Thermal Sciences 75 (2014) 127e139

that on the wall adjacent tubes (tube No. 1 and 5). So the ow resistances of the middle four passages (No. 2, 3, 4 and 5) are higher
than that of the two side passage (No. 1 and 6).
Then why the drag forces on the middle 3 tubes are higher than
that on the wall adjacent tubes? This can be explained from the
time and local area averaged pressure distributions on the tubes
shown in Fig. 16. As shown in Fig. 12, the ow in the tube bundle is
unsteady and the wake sways with time. The wakes of the tubes
adjacent to the walls cannot sway towards the walls. The wakes
determine the pressure distributions on the tubes, so the walls will
modify the pressure forces on the wall adjacent tubes. Fig. 16 shows
the pressure distributions on the tubes. Zero degree represents the
incoming ow direction, and the circumferential direction goes
clockwise (from the top view). From Fig. 16(a), we can see that
pressure distribution on tube No. 1 and No. 5 is not symmetric, the
side adjacent to the walls have relatively lower pressure. The
asymmetry of the pressure distributions on tubes numbered 2 and
4 is very small. The pressure distribution on tube No. 3 is totally
symmetric. So the pressure forces in the y direction of tube No. 1
and 5 are higher. The detailed comparison of Fig. 16(a), (b) and (c)
shows that the pressure in the region of 112.5e247.5 (where are
the wake regions) of tube No. 1 and 5 are higher than that of the
tube No. 2 and 4, while the pressure in the regions of 0.0e67.5 and
292.5e360 are almost the same. So the pressure resistances in x
direction on tube No. 1 and 5 are relatively lower. From the above
analysis we can see that it is the wall effect that results in the
different drag and lift forces on the tubes adjacent to the walls. This
nally results in the low ow resistance and higher uid velocity in
the two side ow passages (No. 1 and 6).
From Fig. 13(b), we can see that the temperatures of the uids
are higher in the two side passages, and the temperatures of the
uids in the middle passages are almost the same. The higher

temperatures in the two side passages are due to the higher velocities. The heat transfer coefcients on the tubes cannot vary
signicantly (we can see later in Fig. 19), so the uids in the two side
passages with larger ow rates will have higher temperatures. The
uniformity of the temperatures in passage No. 2, 3, 4 and 5 can be
explained from Fig. 17. Fig. 17 shows the time and local area averaged longitudinal and transverse velocities in the tube bundles
(absolute velocities are used during the average). The locations of
the ow passages are shown in the small picture at the top of Fig. 17.
Flow passage No. 1, 2, 3, 4, 5 and 6 are the same as in Figs. 13 and 14,
the average velocities in these passages are in the longitudinal direction. Flow passage No. 1.5, 2.5, 3.5, 4.5 and 5.5 are the transverse
passages, the average velocities in these passages are in the
transverse direction. From Fig. 17, we can see that the averaged
absolute transverse velocity in passage No. 3.5 can be more than
half the averaged velocity in the streamwise direction. The transverse velocities in passage No. 2.5 and 4.5 are slightly lower than
that in passage No. 3.5, while the transverse velocities in passage
No. 1.5 and 5.5 are only 30% of that in passage No. 3.5. This means
that the uids in passage No. 2, 3, 4 and 5 exchange greatly. So the
uid temperatures in passage No. 2, 3, 4 and 5 are almost the same.
While the two side passages numbered 1 and 6 have relatively less
exchange of uids with the middle passages, so their temperatures
can be more different from that in the middle.
Fig. 13(c) is the turbulence intensity distribution in the tube
bundle. The turbulence intensities in the two side passages
numbered 1 and 6 are relatively lower, while that in the middle
passages numbered 2, 3, 4 and 5 are almost uniform. This is due to
the large exchange of uids in the middle four ow passages
numbered 2, 3, 4 and 5. In the two side passages, due to the viscous
effect induced by the two bounding walls and the suppression of
the tube wakes, the turbulence is suppressed, and the turbulence

Fig. 16. Local pressure distribution on tubes. (a) Pressure on tube No. 1 and 5. (b) Pressure on tube No. 2 and 4. (c) Pressure on tube No. 3.

X. Li et al. / International Journal of Thermal Sciences 75 (2014) 127e139

Fig. 17. Averaged absolute transverse and longitudinal velocities in the tube bundle.

intensity are smaller than that in the middle of the tube bundle.
This will certainly inuences the heat transfer characteristics.
4.3. Time averaged heat transfer characteristics
The time and area averaged local heat ux on the tubes are
shown in Fig. 18. As in Fig. 16, zero degree represents the incoming
ow direction, the circumferential direction goes clockwise (from
the top view). From Fig. 18(a), we can see that the local heat ux of
tube No. 1 and 5 have opposite distributions, and they both have
higher heat ux at the near wall side. This is due to the higher uid

137

temperatures in the near wall ow passages as shown in Fig. 13. The


wall effects are responsible for the asymmetric distribution. The
highest local heat ux appears at 45 and 315 directions but not
the zero direction. This is due to the inline arrangement of the tube
bundle. The local heat ux of tube No. 2 and 4 in Fig. 18(b) shows
that they also have opposite distributions, but the difference are
very small, which means the wall effects are much lower. Fig. 18(c)
shows that the heat transfer of tube No. 3 has a symmetric distribution. From Fig. 18, we can see that the walls inuences on the
near wall column of tubes are obvious, while the inuences on the
tubes in the middle of the bundle are very small.
The time and area averaged heat transfer characteristics in the
tube bundle are shown in Fig. 19. Fig. 19(a) shows the average heat
ux on the ve tubes. Tube No. 1 and 5 have the highest heat uxes
which are 10% higher than that of tube No. 3. This is due to the
higher uid temperatures in the passages numbered 1 and 6 as
shown in Fig. 13(b). The heat uxes of tube No. 2 and 4 are slightly
higher than that of tube No. 3. In order to calculate the heat transfer
coefcients on the tubes, Fig. 19(b) gives the average temperatures
of the uids just before and behind each tube, as shown in the small
picture at the top of Fig. 19. Fig. 19(b) shows that the temperatures
of the uids ow over tube No. 1 and 6 are about 30  C higher than
that ow over tube No. 2, 3 and 4. Fig. 19(c) shows the average heat
transfer coefcients of each tube. The average heat transfer coefcients are calculated using Eq. (5).

q

h 
Tup Tdown 2  Tw

(5)

where q is the local heat ux. Tup is the uid temperature just
before the tube, Tdown is the uid temperature just behind the tube,
the positions are shown by the solid or dashed lines before and

Fig. 18. Local heat ux distribution on tubes. (a) Heat ux on tube No. 1 and 5. (b) Heat ux on tube No. 2 and 4. (c) Heat ux on tube No. 3.

138

X. Li et al. / International Journal of Thermal Sciences 75 (2014) 127e139

3. The turbulence intensity in the near wall ow passages and the


heat transfer coefcient of the near wall tubes are smaller than
that in the middle due to the wall effects. The heat transfer
coefcient of the tubes near the wall is 10% lower than that in
the middle.
Acknowledgments
This work was nancially supported by National Natural Science
Foundation of China (51006061) and National S&T Major Project
(Grant No. ZX06901).
References

Fig. 19. Heat transfer characteristics in the tube bundle.

behind each tube in the small picture at the top of Fig. 19. Tw is the
tube wall temperature.
The local heat ux was calculated using Eq. (6). The near wall
turbulent thermal conductivity was zero.

 
vT
q l
vn wall

(6)

where l is the uid thermal conductivity.


Fig. 19(c) shows that tube No. 1 and 5 have the lowest heat
transfer coefcients while tube No. 2, 3 and 4 have almost the same
heat transfer coefcients. The heat transfer coefcients of No. 1 and
No. 5 tubes are 10% lower than that of No. 2, 3 and 4 tubes. This is
due to the lower turbulence intensities in the near wall regions as
shown in Fig. 13(c). The measurements of Katinas and Tumosa also
showed that the maximum decrease of the heat transfer coefcient
of the tube banks induced by the wall can be 20% [18].
5. Conclusions
Development of nuclear technologies requires high heat transfer rate and more accurate heat transfer calculation. The shell side
heat transfer in the once through steam generators of high temperature gas-cooled reactors are cross ow heat transfer over wall
bounded tube bundles. This paper reported a numerical study
about the cross ow and heat transfer characteristics in a wall
bounded tube bundles. Numerical method was veried using
experimental measurement and empirical correlations before the
simulation. The ow in wall bounded tube bundles has the intrinsic
transient ow characteristics of swaying as in free tube bundles, but
the walls modify the ow and heat transfer signicantly. The ow
and heat transfer characteristics are analyzed detailed in this paper,
the main conclusions are:
1. The walls suppressed the wakes of near wall tubes and prevent
the wakes from swaying towards them. This modies the
pressure distribution on the wall adjacent tubes and make near
wall ow passages have lower ow resistance. The uid velocity and uid temperatures are higher in the near wall ow
passages.
2. The averaged absolute transverse velocity in the middle of the
tube bundle can reach 50% of the averaged streamwise velocity,
while the transverse velocity in the near wall ow passages
reduces to only 15%. There is great mixing in the middle of the
tube bundle. Temperature and turbulence intensity in the
middle of the tube bundle are uniform.

[1] Z.X. Wu, Z.Y. Zhang, World development of nuclear power system and high
temperature gas-cooled reactor, Chinese Journal of Nuclear Science and Engineering 20 (3) (2000) 211e231 (in Chinese).
[2] M.M. Zdravkovich, Flow Around Circular Cylinders, Oxford University Press,
New York, 2003.
[3] A. Zukauskas, J. Ziugzda, Heat Transfer of a Cylinder in Crossow, Hemisphere
Publishing Corporation, New York, 1985.
[4] A. Zukauskas, R. Ulinskas, Heat Transfer in Tube Banks in Crossow, Hemisphere Publishing Corporation, New York, 1988.
[5] M. Coutanceau, J.R. Defaye, Circular cylinder wake congurations: a ow
visualization survey, Applied Mechanics Reviews 44 (5) (1991) 255e305.
[6] H.J. Niemann, N. Holscher, A review of recent experiments on the ow past
circular cylinders, Journal of Wind Engineering and Industrial Aerodynamics
33 (1e2) (1990) 197e209.
[7] S.S. Paul, M.F. Tachie, S.J. Ormiston, Experimental study of turbulent crossow in a staggered tube bundle using particle image velocimetry, International Journal of Heat and Fluid Flow 28 (3) (2007) 441e453.
[8] O. Simonin, M. Barcouda, Measurements and prediction of turbulent ow
entering a staggered tube bundle, in: Proceedings of the Fourth International
Symposium on Applications of Laser Anemometry to Fluid Mechanics, Lisbon,
1988.
[9] D.S. Weavera, M. Elkashlana, On the number of tube rows required to study
cross-ow induced vibrations in tube banks, Journal of Sound and Vibration
75 (2) (1981) 265e273.
[10] P.W. Bearman, M.M. Zdravkovich, Flow around a circular cylinder near a plane
boundary, Journal of Fluid Mechanics 89 (1978) 33e47.
[11] F. Angrilli, S. Bergamaschi, V. Cossalter, Investigation of wall induced modications to vortex shedding from a circular cylinder, ASME Journal of Fluid
Engineering 104 (4) (1982) 518e522.
[12] G. Buresti, A. Lanciotti, Mean and uctuating forces on a circular cylinder in
cross-ow near a plane surface, Journal of Wind Engineering and Industrial
Aerodynamics 41 (1e3) (1992) 639e650.
[13] J.X. Zhan, J.J. Wang, P.F. Zhang, Forces on a near-wall circular cylinder, Journal
of Hydrodynamics Series B 16 (6) (2004) 658e664.
[14] T. Nishino, G.T. Roberts, X. Zhang, Vortex shedding from a circular cylinder
near a moving ground, Physics of Fluids 19 (2) (2007) 025103.
[15] R.Q. Duan, S.Y. Jiang, Numerical investigation of gas ow distribution and
thermal mixing in helically coiled tube bundle, Journal of Nuclear Science and
Technology 45 (7) (2008) 704e711.
[16] S. Aiba, Heat transfer around a circular cylinder near a plane surface, ASME
Journal of Heat Transfer 107 (4) (1985) 916e921.
[17] G. Miskinis, A. Zukauskas, P. Daujotas, Heat transfer from a rough cylinder in
cross ow near a plane wall, Heat Transfer Research 25 (2) (1993) 183e191.
[18] V. Katinas, A. Tumosa, Heat transfer and ow past tube bundles in the wall
region, Heat Transfer Research 25 (2) (1993) 161e164.
[19] C. Moulinec, M.J.B.M. Pourquie, B.J. Boersma, T. Buchal, F.T.M. Nieuwstadt,
Direct numerical simulation on a cartesian mesh of the ow through a tube
bundle, International Journal of Computational Fluid Dynamics 18 (1) (2004)
1e14.
[20] C. Liang, G. Papadakis, Large eddy simulation of cross-ow through a staggered tube bundle at subcritical Reynolds number, Journal of Fluids and
Structures 23 (8) (2007) 1215e1230.
[21] H.R. Barsamian, Y.A. Hassan, Large eddy simulation of turbulent crossow in
tube bundles, Nuclear Engineering and Design 172 (1e2) (1997) 103e122.
[22] D. Bouris, G. Bergles, Two dimensional time dependent simulation of the
subcritical ow in a staggered tube bundle using a subgrid scale model, International Journal of Heat and Fluid Flow 20 (2) (1999) 105e114.
[23] S. Benhamadouche, D. Laurence, LES, coarse LES, and transient RANS comparisons on the ow across a tube bundle, International Journal of Heat and
Fluid Flow 24 (4) (2003) 470e479.
[24] P. Rollet-Miet, D. Laurence, J. Ferziger, LES and RANS of turbulent ow in
tube bundles, International Journal of Heat and Fluid Flow 20 (3) (1999)
241e254.
[25] S.S. Paul, S.J. Ormiston, M.F. Tachie, Experimental and numerical investigation
of turbulent cross-ow in a staggered tube bundle, International Journal of
Heat and Fluid Flow 29 (2) (2008) 387e414.

X. Li et al. / International Journal of Thermal Sciences 75 (2014) 127e139


[26] J.K. Watterson, W.N. Dawes, A.M. Savill, A.J. White, Predicting turbulent ow
in a staggered tube bundle, International Journal of Heat and Fluid Flow 20 (6)
(1999) 581e591.
[27] Y.Q. Wang, P. Jackson, T.J. Phaneuf, Turbulent ow through a staggered tube
bank, Journal of Thermophysics and Heat Transfer 20 (4) (2006) 738e747.
[28] Y.Q. Wang, P.L. Jackson, Turbulent modeling applied to ow through a staggered tube bundle, Journal of Thermophysics and Heat Transfer 24 (3) (2010)
534e543.
[29] A. Ridluan, A. Tokuhiro, Benchmark simulation of turbulent ow through a
staggered tube bundle to support CFD as a reactor design tool. Part I: SRANS
CFD simulation, Journal of Nuclear Science and Technology 45 (12) (2008)
1293e1304.
[30] A. Ridluan, A. Tokuhiro, Benchmark simulation of turbulent ow through a
staggered tube bundle to support CFD as a reactor design tool. Part II: URANS
CFD simulation, Journal of Nuclear Science and Technology 45 (12) (2008)
1305e1315.
[31] S. Bouhairie, V.H. Chu, Two-dimensional simulation of unsteady heat transfer
from a circular cylinder in crossow, Journal of Fluid Mechanics 570 (2007)
177e215.
[32] X.W. Li, J.A. Meng, Z.Y. Guo, Turbulent ow and heat transfer in discrete
double inclined ribs tube, International Journal of Heat and Mass Transfer 52
(3e4) (2009) 962e970.
[33] X.W. Li, J.A. Meng, Z.X. Li, Enhancement mechanisms for single-phase turbulent heat transfer in micro-n tubes, Journal of Enhanced Heat Transfer 15 (3)
(2008) 227e242.
[34] Fluent 6.0 Users Guide, Fluent Inc, Lebanon, NH, 2001.
[35] A. Zukauskas, R. Ulinkas, V. Katinas, J. Karni, Fluid Dynamics and Flow-induced
Vibrations of Tube Banks, Hemisphere Publishing Corporation, New York, 1988.
[36] I.E. Idelchik, Handbook of Hydraulic Resistance, Hemisphere Publishing Corporation, Washington DC, 1986.

139

Cl,v: viscous force coefcient in y direction, Cl;v Fv;y =0:5ru2N DH, dimensionless
D: tube diameter [m]
f: oscillation frequency of forces acting on the tube, [Hz]
Fp: pressure force, [N]
Fv: viscous force, [N]
h: convective heat transfer coefcient, [Wm2 C1]
H: tube length (it is unit in this paper) or the distance between the wall and
the tube, [m]
q
2 k=u , dimensionless
I: turbulence intensity, I
N
3
k: turbulence kinetic energy, k 12 u0i u0i , [m2 s2]
p: pressure [Pa]
Pr: Prandtl number, dimensionless
q: local heat ux, [W m2]
Re: Reynolds number, Re rumaxD/m, dimensionless
S1: transverse pitches, [m]
S2: longitudinal pitches, [m]
St: Strouhal number, St fD/umax, dimensionless
t: time [s]
T: temperature, [ C]
Tdown: uid temperature behind tubes, [ C]
Tup: uid temperature before tubes, [ C]
Tw: tube wall temperature, [ C]
uN: inlet velocity, [m s1]
umax: maximum average velocity in the tube bundle, [m s1]
y: dimensionless distance normal to the wall (yu*/y), dimensionless

Nomenclature

Greek symbols
2 3
3 : turbulence dissipation rate, [m s
]
l: molecular thermal conductivity, [W/m  C]
m: dynamic viscosity [kg m1 s1]
x: friction factor, dimensionless
r: density, [kg m3]

Cd: drag coefcient, Cd Cd,p Cd,v, dimensionless


Cd,p: pressure force coefcient in x direction, Cd;p Fp;x =0:5ru2N DH, dimensionless
Cd,v: viscous force coefcient in x direction, Cd;v Fv;x =0:5ru2N DH, dimensionless
Cl: lift coefcient, dimensionless
Cl,p: pressure force coefcient in y direction, Cl;p Fp;y =0:5ru2N DH, dimensionless

Subscripts
p: pressure
v: viscous
w: wall
x: x direction
y: y direction

Вам также может понравиться