Applied Energy 195 (2017) 426438

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Numerical analysis of a vertical double-pipe single-flow heat exchanger

applied in an active cooling system for high-power LED street lights
Gerd Schmid a, Zun-Long Huang a, Tai-Her Yang b, Sih-Li Chen a,
a
Department of Mechanical Engineering, National Taiwan University, No. 1, Sec. 4, Roosevelt Road, Taipei 10617, Taiwan, ROC
b
Giant Lion Know-How Co., Ltd., P.O. Box: 108-276, Taipei, Taiwan, ROC

h i g h l i g h t s g r a p h i c a l a b s t r a c t

Active air cooling system for high-

power LED street lights is analyzed.

Heat is dissipated from the lamp post
by natural convection.

Axisymmetric numerical simulation
using ANSYS Fluent has been
performed.

Center-In-Side-Exit (CISE) flow
configuration has shown higher heat
transfer rates.

Effect of thermal conductivity of
center pipe material is less than 3%.

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

Article history: The present study examines the use of a vertical double-pipe single-flow heat exchanger as part of an
Received 28 October 2016 active air cooling system for a 150 W LED street light. The air is circulated inside the lamppost by an inter-
Received in revised form 19 January 2017 nal fan to form a closed-loop system. The heat is dissipated to the surrounding air by natural convection,
Accepted 10 March 2017
reaching Rayleigh numbers up to Ra = 6.5  1010. Experiments with a 5 m high prototype were conducted,
Available online xxxx
and the data were used to validate the numerical model. The experimental results show that the LED
excess temperature can be lowered to about 42 C. A two-dimensional axisymmetric numerical simula-
Keywords:
tion was performed to study the influence of various parameters, including pipe length, material conduc-
Forced-air cooling system
Light emitting diode
tivity, flow direction, pipe diameter ratio, and mass flow rate, on the heat transfer rate. The findings show
Double-pipe heat exchanger that the additional heat loss created by extending the lamppost largely depends on the flow rate. When
Street light extending the lamppost from 3 to 5 m at a high mass flow rate of 0.014 kg/s, the heat loss increases by
Axisymmetric simulation 34.1% to 120.2 W. The numerical study was also used to visualize the hydrodynamic boundary layers
on the surface of the lamppost and the temperature contours in and outside of the heat exchanger.
2017 Elsevier Ltd. All rights reserved.

1. Introduction tised up to 50,000 h, which is about 50 times longer than incandes-

cent and 45 times longer than fluorescent light sources [1]. One of
Light emitting diodes (LED) are gradually replacing traditional the main challenges in designing LED applications is ensuring ade-
light sources in commercial and residential sectors. As well as quate thermal management to remove and dissipate the heat pro-
higher energy efficiency, LEDs have a longer lifetime, often adver- duced by the LED and to guarantee reliable and safe operation.
High temperatures can damage the p-n junction, lower the lumi-
Corresponding author. nous efficiency, shift the wavelength, and reduce the quantum effi-
E-mail address: slchen01@ntu.edu.tw (S.-L. Chen). ciency of phosphor, which all negatively affect the lifetime of the

http://dx.doi.org/10.1016/j.apenergy.2017.03.054
0306-2619/ 2017 Elsevier Ltd. All rights reserved.
 G. Schmid et al. / Applied Energy 195 (2017) 426438 427

Nomenclature

A flow cross section (m2) op outer pipe

cp thermal heat capacity (J/kg K) out outlet
D diameter of lamppost (m) s surface
f friction factor
GrL Grashof number Greek symbols
g gravitational acceleration (m/s2) b volume expansion coefficient (1/K)
k turbulent kinetic energy (m2/s2) C generalized diffusion coefficient
L length of inner pipe (m) e dissipation rate of turbulence (m2/s3)
m_ mass flow rate (kg/s) l dynamic viscosity (kg/m s)
NuL Nusselt number lt turbulent dynamic viscosity (kg/m s)
Q_ heat loss (W) m kinematic viscosity (m2/s)
RaL Rayleigh number mt turbulent kinematic viscosity (m2/s)
T temperature (K) q density (kg/m3)
u, v, w velocity in different directions (m/s)
Acronyms
Subscripts Al aluminum
a ambient CISE Center-In-Side-Exit flow
c center Fg fiberglass
exp experimental GHE ground heat exchanger
FC fan chamber LED light emitting diode
ip inner pipe PVC polyvinyl chloride
j junction SICE Side-In-Center-Exit flow
num numerical

LED [2]. The maximum allowable junction temperature (Tj) for safe axial fans, which were arranged on top of a heat sink inside an alu-
operation varies among manufacturers, but is in the range of 120 minum housing, from which the heat is dissipated by natural con-
140 C. However, lower temperatures are desirable because they vection to the surroundings. Their results showed that the cooling
severely influence the lifespan of LEDs, as an increase in Tj of 10 system can reduce the temperature of a 68 W LED by up to 27.9 C,
15 C can reduce the lifetime by 50% [3]. compared to a case with only the heat sink. The main drawback of
Most studies on thermal management of LEDs have focused on this design is that the hot internal air is contained inside the hous-
the design and optimization of passive heat sinks [48], which are ing and only the lamp head case acts as an external cooling body.
often combined with heat pipes [911] or vapor chambers [12,13] Liquid cooling systems are more complex than forced air cooling
to improve the thermal performance by reducing the spreading systems, but they can provide higher cooling capacities. They usu-
resistance. Two recent studies found that perforations [8] or open- ally consist of a cold plate attached to the LED, a reservoir to store
ing slits [5] on the base of free-hanging oversized heat sinks can the liquid, a pump to drive it, and a liquid-to-air heat exchanger to
increase the air flow through the fin array and reduce the thermal dissipate the heat to the surroundings. Ramos-Alvarado et al.
resistance by up to 30.5% and 36.7%, respectively. Ye et al. [10] numerically compared multiple cold plates with different micro-
designed a heat sink with parallel vertical fins and three embedded channel designs [23] and, in a subsequent study, applied the best
heat pipes for high-power LED applications. Experimental and cold plate design to an array of 64 high-power LEDs with various
numerical analyses were carried out to investigate the optimal power inputs and flow rates [20]. Their simulations showed a Tj
fin spacing to achieve the smallest thermal resistance. The experi- of about 80 C at a total input power of 192 W and a pumping
ments with an 80 W LED showed that the optimized design can power of 1 W. However, the study considered an array of individ-
limit the temperature of the heat sink base to about 70 C. While ual LEDs instead of single high-power LED chip, and the pumping
passive cooling systems are extremely reliable and do not require power of 1 W is only derived from the numerical pressure drop
additional power consumption, they are usually limited to power across the cold plate and does not represent the complete cooling
inputs of less than 100 W. Active cooling solutions, such as forced system.
air [1419] or liquid cooling systems [2022], are more capable of The present study analyzes a novel forced-air cooling system for
maintaining the junction temperature of high-power LEDs under a LED street lights, which uses the lamppost as an air-to-air heat
critical limit. Deng et al. [14] conducted an experimental evalua- exchanger to dissipate heat to the surrounding air. The air-flow
tion of single-and dual-synthetic jet actuators combined with a inside the lamppost is driven by an internal fan, and the whole sys-
heat sink. The system consumed about 3.2 W of power and was tem is sealed to prevent dust and debris from entering the system;
used to cool a 30 W LED chip. The results showed that the this reduces maintenance cycles, lowers the risk of system failure,
single-synthetic jet actuator could maintain the base temperature and thus makes it especially suitable for extreme climatic condi-
at 46.5 C and the double jet could reduce it by 4 C. Synthetic jets tions. This type of vertical single-flow double-pipe heat exchanger
can be integrated in very compact systems, but for higher power has a similar working principle to coaxial ground heat exchangers
ranges, fans have a better overall heat dissipation capability. Geis- (GHEs) [2429]. However, the aim is not to transfer the heat from
ler [15] developed an active thermal solution consisting of multiple one pipe to the other, as in conventional double-pipe heat
fans combined with a densely folded copper heat sink. His design exchangers with two inlets and outlets [30], but rather to transfer
could maintain a dummy heater with a 600 W input at a base it through the outer pipe wall to the surrounding air. According to
excess temperature over ambient of 85 C. Iaronka et al. [16] the experimental findings of Jalaluddin et al. [24], coaxial GHEs
designed and investigated a closed-loop forced air cooling system have a 61% higher heat exchange rate than U-tube GHEs at a depth
for high-power LED street lamps. The system was driven by five of 20 m and a flow rate of 4 l/min. In fact, Acuna and Palm [27]
428 G. Schmid et al. / Applied Energy 195 (2017) 426438

found that the thermal performance of coaxial GHEs can be about

100% higher than that of U-pipes, based on their experimental find-
ings with a 190 m deep borehole. Yekoladio et al. [26] performed
an analytical analysis to determine the optimum diameter ratio
of a coaxial GHE to reduce the total pressure drop and found that
the optimal ratio of the inner to outer diameter is 0.653 for turbu-
lent and 0.683 for laminar pipe flow. In a recent study, Holmberg
et al. [28] numerically analyzed the influence of the flow direction
and the depth on the performance of a coaxial GHE in the range of
200500 m. They concluded that for depths of up to 200 m, the
flow direction has little or no effect on the performance; however,
for deeper boreholes, they recommended using the center pipe as
an inlet for heat dissipation and the annular space as an inlet for
heat extraction. They also suggested using high mass flow rates
or a center pipe with low thermal conductivity to reduce the heat
transfer between the center pipe and the annular space. Jalaluddin
and Miyara [29] also investigated the influence of the center pipe
material on the thermal performance of a coaxial GHE. Their
numerical simulations indicated that replacing a polyethylene cen-
ter pipe (k = 0.35 W/m K) with one made of PVC (k = 0.15 W/m K)
can increase the heat exchange rate by about 12%.
While similarities exist between coaxial GHEs and the cooling
system of this study, the results of the former studies cannot be
directly applied, because the heat is transferred to the surrounding
air by natural convection instead of being conducted to the soil. In
addition, the pipe lengths of the coaxial GHEs in the aforemen-
tioned studies range from 20 to 500 m, while those of the present
study are between 2 and 5 m. Accordingly, the aim of this study is
Fig. 1. (a) Photographic view of the experimental study, (b) photographic view of
to present the first comprehensive analysis of a vertical single-flow the internal pipe, (c) schematic of lamppost with thermocouple locations.
double-pipe air-to-air heat exchanger by investigating the influ-
ence of various design parameters and boundary conditions on
the thermal performance. Apart from being used for LED street
lights, this type of cooling system can also be integrated with
active cooling systems of other pole-mounted electronic equip-
ment, such as small power transformers [31] or solar panels
[32,33].

2. Experimental method

2.1. Configuration of experimental system

Fig. 1 shows the experimental apparatus and the working prin-

ciple of the present active cooling system. The waste heat from the
LED is conducted to the lamp head, from where it is convected to
the internal air flow via a two-way heat exchanger. The air is circu-
lated in a closed-loop by an integrated fan and guided by an inter-
nal pipe along the inner surface of the lamppost. The heat is then
dissipated by natural convection from the lamppost to the sur-
rounding air. The fan is located at the lower end of the lamppost
inside a fan chamber. The entire system is sealed to prevent dust
and debris from entering.
Owing to its overall height of more than 5 m, the complete test
rig was placed inside a two floor research building, as displayed in
Fig. 1(a). This way, the air surrounding the test equipment was free
from air currents that might otherwise have influenced the mea-
surements. Fig. 1(c) shows the main components of the cooling
system including the lamp head, lamppost, internal pipe, and cen-
trifugal fan inside the fan chamber. All measurement instruments
were located on the second floor, as illustrated in Fig. 2. Fig. 2. Photographic view of the upper part of the test rig with measurement
The fan is the central component of any forced air cooling sys- equipment.
tem. The present study used a centrifugal AC fan, manufactured by
Oriental Motor Co. Ltd., with a confirmed input power of 20 W to
circulate the air. The AC fan, which can generate a maximum vol- lamppost. The fan chamber had a diameter of 32 cm to provide suf-
umetric flow rate of 1.6 m3/min or a maximum static pressure of ficient space for recirculating the air and the fan was connected to
160 Pa, was located in a circular chamber at the lower end of the a variac transformer to vary the fan speed and control the flow
 G. Schmid et al. / Applied Energy 195 (2017) 426438 429

utes to obtain the average maximum velocity in m/s, the standard

deviation in m/s, and the turbulence intensity in%. The average
velocity was then determined using Eq. (1), which is valid for tur-
bulent pipe flow:
umax
uav g p 1
1 1:3 f
The friction factor f was determined from the Moody diagram
[35], where the pipe surface was considered smooth. In the next
step the mass flow rates were calculated using the average veloc-
ities. These mass flow rates were then used in the numerical anal-
ysis as the inlet boundary condition. Five different flow velocities
Fig. 3. Photographic view of the lamp head.
were set for each setup by manually adjusting the variac. Table 2
provides an overview of the maximum velocities measured during
velocity of the system. The air flow was measured using a TSI the experiments and the calculated mass flow rates.
TA465 air velocity meter, in combination with a straight thermal Fig. 1(c) presents locations and names of the temperatures that
hot-film anemometer (model number: TSI960). At the top, the were monitored using T-Type thermocouples connected to a data
fan was connected to a 3 m long internal pipe, which could be recorder model MV200 by YOKOGAWA. The outer surface temper-
removed and changed by disassembling the outer steel pipe of ature (Ts) of the lamppost was taken as the average of four mea-
the lamppost. Two different internal pipes were compared, one surement points:
made of aluminum (Al) and the other of PVC. These pipes were T fan T s1 T s2 T s3
connected to the lamp head with a flexible hose to account for Ts 2
4
the bend of the lamppost (Fig. 1(b)).
Tipc2 and Topc2 measured the inlet and outlet temperatures for
As shown in Fig. 3, the lamp head, where the LED is located,
the numerical lamppost study. Another thermocouple was con-
compromises a heat sink, a flow distributor, a flow diverter, a
nected to the temperature measurement point of the LED chip to
lamppost connector, and a lid. The flow distributor connects the
monitor the temperature of the LED (TLED). Because of the large size
central pipe with the heat sink and regulates the ratio of center
of the cooling system, it usually took up to four hours to reach ther-
to side fin channels. The flow diverter, as the name implies, diverts
mal stability, and the system reacted very slowly to changes in
the flow from one heat sink section to the other. The heat sink is
ambient temperature. Therefore, after reaching thermal stability,
divided into three different flow sections: one in the middle in
the temperature measurements were recorded for at least two
one direction, and two on the sides in the opposite direction. Nota-
more hours to accurately capture the steady-state temperatures.
bly, this study only focuses on analyzing the heat dissipation from
There are two possible flow directions with different heat transfer
the lamppost and not on analyzing and optimizing the lamp head
paths. In the Center-In-Side-Exit (CISE) flow configuration, the fluid
design. A COB LED produced by Lighten Corp. with an input power
flows down the center pipe and back up along the outer pipe (Fig. 4
of 150 W and a color temperature of 3000 K was used as a power
(a)). In the Side-In-Center-Exit configuration (SICE), the hot fluid
source and was mounted to the center of the heat sink base. Table 1
enters the heat exchanger from the outer pipe and exits via the
presents the properties of the materials used in the experimental
internal pipe (Fig. 4(b)). The overall heat loss from the lamppost
analysis.
to the surrounding air can be expressed using the following conser-
vation of energy equation [35]:
2.2. Data reduction
Q_ mc
_ p T in  T out 3
Fig. 1(c) indicates the measurement point of the flow velocity.
The velocity is measured at the center of the pipe, where the flow where Tin and Tout are the mean fluid temperatures at the inlet and
reaches its maximum. Using the built-in turbulence function of the _ is the mass flow rate, and cp is the thermal heat capacity.
outlet, m
flow meter, the air flow was measured for three consecutive min-  Q_ =L and is used
The heat loss rate per unit length is defined as Q

Table 1
Material properties from the lamppost analysis (25 C) [34].

Material q (kg/m3) cp (J/kg K) k (W/m K) m (kg/m s) b (1/K)

Air 1.127 1005.0 0.0271 1.697e05 0.0032
Aluminum 2719.0 871.0 202.4
Steel 7817.0 461.0 14.3
PVC 50.0 840.0 0.2

Table 2
Flow conditions for internal pipe with smooth surface.

umax (m/s) Re number Friction factor uavg (m/s) _ (kg/s)

m
11.0 22,727 0.02509 9.12 0.0099
9.5 19,628 0.0260 7.85 0.0085
8.0 16,529 0.02713 6.59 0.0071
6.5 13,430 0.0286 5.33 0.0058
5.0 10,331 0.03062 4.07 0.0044
3.5 7231 0.03371 2.83 0.0031
2.0 4132 0.03953 1.59 0.0017
430 G. Schmid et al. / Applied Energy 195 (2017) 426438

3. Numerical method

3.1. Numerical domain

This numerical study focuses on the heat loss from the lamp-
post to the surrounding air (dotted area in Fig. 1(c)). The lamppost
can be described as a vertical double-pipe single-flow heat exchan-
ger with natural convection to the surrounding air. The cross-
section of the lamppost is circular, and the heat loss to the sur-
rounding air is assumed uniform; hence, the flow is modeled as
axisymmetric with a rectangular two-dimensional domain, as
shown in Fig. 5(a). This simplification reduces the time taken for
the simulations by several orders of magnitude. According to
[37], turbulent flows are intrinsically three-dimensional because
certain mechanisms, such as vortex stretching or tilting, cannot
occur in two dimensions; nevertheless, they can still be two-
dimensional in a statistical sense. The geometry was thus drawn
as a two-dimensional line model consisting of four separate zones:
two solid regions for the internal pipe and the lamppost, and two
fluid regions, one inside the lamppost and the other representing
the surrounding air, as indicated in Fig. 5(b) and (c). The interfaces
between these fluid and solid zones are modeled as coupled walls,
which enable an accurate simulation of the real heat transfer
through those boundaries [38]. The grid is structured, built up from
quadrilateral elements, and aligned with the flow direction. The
grid cells are very dense close to the pipe walls to accurately model
the boundary layers.

3.2. Flow model selection

Fig. 4. Heat transfer paths of double pipe single flow heat exchanger for different
flow configurations: (a) Center-Inlet-Side-Exit (CISE), (b) Side-Inlet-Center-Exit The natural convection from vertical cylinders can be treated as
(SICE). a vertical plate if the following criterion is true [35]:
35L
DP 1=4
6
GrL
to express the performance of the heat exchanger. If the heat loss is
known, the average heat transfer coefficient where GrL is the Grashof number based on the height of the lamp-
post, defined as
Q_ gbT s  T a L3
hav 4 GrL 7
As T s  T a m2
In the present study, the diameter of the lamppost is fixed at
and the Nusselt number
D = 0.101 m and the right-hand-side of Eq. (6) is estimated to be
larger than 0.16 m. Consequently, the curvature effects are not
h av  L
NuL 5 negligible and the lamppost cannot be treated as a vertical plate.
k
Boetcher [39] gives a comprehensive overview of studies about
natural convection from circular cylinders. Because the length of
can be calculated, where the lamppost height L is used as the char-
the lamppost is more than 3 m, the Rayleigh numbers in the pre-
acteristic length.
sent analysis reach values of more than 1010. Hence, the flow on
the outside of the lamppost has to be considered as turbulent free
2.3. Experimental uncertainty convection. Among the various Nusselt number correlations, Kreith
et al. [40] proposed the most suitable for the present case, which in
The error range of T-type thermocouples is (0.15% + 0.5 C) for the turbulent region, 109  RaL  1012, is reported as
a temperature range of 0 C  T  350 C. The anemometer used to
NuL 0:021  Ra2=3
L 8
measure the air flow velocity has an operating range from 0 to
50 m/s, and the accuracy is either 3% or 0.015 m/s, whichever This study focuses on a mass flow ranging from 0.002 to
is greater. The uncertainties of the experimental measurements 0.014 kg/s. The corresponding Reynolds numbers for the flow in
are estimated following the recommendations by Holman [36]. the internal and outer pipes are ranging from 876 to 25,000. The
The measured quantities are the temperatures, the flow velocity flow in the internal pipe is entirely in the transitional and fully tur-
and the power input of the centrifugal fan. Uncertainties in tem- bulent region, while that in the outer pipe reaches from the lami-
perature differences were within 0.6 C and in power measure- nar to the turbulent region. Currently, no practical turbulence
ments within 2.0% of the readings. The uncertainties of the model can reliably predict all types of turbulent flows with suffi-
volumetric air flow rates were within 3% of full scale readings, cient accuracy and reasonable computational cost [38]. This study
and the standard deviation provided by the instrument ranged compared four different turbulence models with Eq. (8) to deter-
from 0.03 m/s at the highest velocities to 0.07 m/s at the lower mine which model is closest to the data. The four models are the
range of velocities. standard k-e, realizable k-e, the standard k-x, and the k-x SST
 G. Schmid et al. / Applied Energy 195 (2017) 426438 431

Fig. 5. (a) Axisymmetric section of lamppost, (b) detail view of computational model with different domains, (c) full 2D-model showing mesh and boundary conditions.

model. These are all two-equation models based on the Reynolds- 3.3. Governing equations and boundary conditions
averaged Navier-Stokes (RANS) method, where the turbulent fluc-
tuations are averaged over time [37]. The theoretical background The governing equations for continuity, momentum, and energy
of the turbulence models can be found in [38]. Fig. 6 presents the conservation together with the realizable k-e turbulence model can
comparison results of the turbulence models. All design points be expressed as follows [37,42]:
reach Rayleigh numbers of over 1010, which are considered to be Continuity equation
fully turbulent [39]. The results of both k-e models are almost iden-
@ui
tical, showing very good agreement with the turbulent correlation 0 9
@xi
by Kreith et al. (Eq. (8)). These two turbulence models characterize
the turbulent field in terms of turbulent kinetic energy (k) and the Momentum equation
viscous dissipation rate (e). However, the standard k-e model is   
@ui uj 1 @p @ @uj @ui
only valid for fully turbulent flows, which is not the case in the pre-  m mt 10
@xi q @xi @xj @xi @xj
sent study [41]. The realizable k-e model was thus chosen for the
present simulations. Energy equation
  
@ui T @ m mt @T
11
@xi @xi Pr Prt @xi
Turbulent kinetic energy k equation
  
@ui k @ mt @k
m Ce 12
@xi @xi rk @xi
Turbulent energy dissipation e equation
  
@ui e @ mt @ e e2
m c1 Ce  c2 13
@xi @xi re @xi k
with C being the generation of k given by
 
@ui @uj @ui
C mt 14
@xj @xi @xi
and mt is calculated as
2
k
mt 15
Fig. 6. Comparison of turbulence models applied to the lamppost simulation. e
432 G. Schmid et al. / Applied Energy 195 (2017) 426438

The empirical constants in Eqs. (12) and (13) were assigned as the pipe surface. Following the validation, the numerical models
[38]: c1 = max[0.43, l/(lt + 5)], c2 = 1.9, rk = 1.0, re = 1.2. The were used to study how various design parameters affect the ther-
enhanced wall treatment is used at the near-wall. Fig. 5 mal performance.
(b) and (c) shows the boundary conditions including the axis, a
mass flow inlet, pressure outlets (at the exit of the lamppost and 4.1. Experimental LED temperatures
at the boundaries of the ambient region), and wall boundaries with
zero heat flux at the top and bottom of the lamppost. A uniform Figs. 7 and 8 show the experimental results of the active cooling
temperature and a mass flow normal to the boundary with a turbu- system for different flow directions and pipe materials. In Fig. 7,
lent intensity of 1% were defined at the inlet. The gauge pressure at the excess temperatures (Tex = TLED  Ta) of the LED are compared
the pressure outlets were set to the static pressure of the environ- for different volume flows ranging from 1.5 to 8.5 m3/s. Overall,
ment. The back flow conditions at the boundaries of the ambient it can be seen that the CISE flow leads to a slightly lower excess
region were assumed as 0.1% turbulence with a length scale of temperature than the SICE flow for both pipe materials. A compar-
0.1 m. The backflow at the outlet of the pipe was set to 323 K, ison of the results for the PVC and Al pipes shows that the pipe
and 1% turbulence with the corresponding hydraulic diameter of material has only a very small influence on the excess temperature.
0.0275 m. However, the PVC seems to have a marginally better thermal per-
formance. The lowest excess temperature of Tex = 44.2 C could be
3.4. Numerical setup achieved at V_ = 0.0072 m3/s with CISE and the PVC pipe. At low vol-
ume flows, under 0.002 m3/s, the excess temperature rose to
The simulation was carried out using the software ANSYS Fluent almost 60 C, and when the active cooling system was turned off
14.5. The solver was set to pressure-based, steady-state flow, and completely, an excess temperature of Tex = 72.3 C was measured.
absolute velocity formulation. The material properties were taken In Fig. 8, the experimental results are compared for different fan
as equal to those in the experimental analysis, listed in Table 1. power inputs ranging from 8 W to 20 W. When the fan power was
The Boussinesq model, which assumes that the density is constant reduced by 25% from 20 W to 15 W, the excess temperature
in all solved equations except for the body force term in the increased on average by only 1.5 C. Further reductions to 12, 10,
momentum equation, was chosen for the air density. The SIMPLEC and 8 W led to average increases of 2.3, 4.4, and 4.3 C, respec-
pressure-velocity coupling scheme was used and the gradient tively. These findings show that it is necessary to find a balance
option was set to least square cell based. For better accuracy, the between the thermal performance and power consumption, which
spatial discretization for momentum, the turbulent kinetic energy, depends on the design parameters and the equipment used.
and the turbulent dissipation rate and energy were all set to sec-
ond order upwind. The convergence criterion for the residuals of 4.2. Validation of numerical model
the governing equations was set to 105.
Figs. 9 and 10 show the direct comparisons of the experimental
and numerical results. The experiments were conducted for both
3.5. Grid independence study
flow directions and for two different internal pipe materials (Al
and PVC). Each numerical simulation used the same mass flow rate
A grid independence study was conducted to confirm that the
mesh density is high enough to capture all of the relevant features.
Table 3 summarizes the results of the grid independence study.
First, an initial grid with 28,390 elements was defined, and the
influence of grid refinements at various boundaries was then
checked. The surface-averaged temperature at the outlet (Tout)
and the heat loss to the surrounding air (Q_ ) were the monitored
parameters. The grid was finer adjacent to the walls to resolve
the boundary layer flow. The number of nodes in y-direction of
both solid regions (internal and outer post) was kept constant at
four. The analysis showed that the initial grid was sufficiently
accurate, because further refinements brought no significant
changes; the initial grid was therefore chosen for the numerical
analysis.

4. Results and discussion

This section first presents the experimental results for the com-
plete lamppost, which were then used to validate the numerical
model by comparing the temperatures at the flow outlet and on Fig. 7. Experimental excess temperatures at different volume flows.

Table 3
_ = 0.014 kg/s).
Grid independence for lamppost simulation (Tin = 333 K, Ta = 303 K, m

# Mesh elements y-nodes inner pipe y-nodes outer pipe y-nodes ambient x-nodes total Tout (K) Q_ (W)
1 28,390 40 50 70 171 326.44 90.60
2 29,410 50 50 70 171 326.45 90.50
3 29,410 40 60 70 171 326.43 90.68
4 32,810 40 50 100 171 326.45 90.47
5 39,120 40 50 70 241 326.43 90.67
6 50,530 40 50 70 311 326.43 90.72
 G. Schmid et al. / Applied Energy 195 (2017) 426438 433

Fig. 8. Experimental excess temperatures for different fan power inputs. Fig. 11. Heat loss over mass flow rate for different flow directions and pipe
materials (Ta = 303 K, Tin = 333 K).

and the same inlet and ambient temperatures as the corresponding

experimental case. The inlet temperature was either the value
measured at Tipc2 for CISE or at Topc2 for SICE, as indicated in
Fig. 1(c). Fig. 9 compares the temperatures at the pipe outlet (Topc2
for CISE and Tipc2 for SICE), which show a good agreement within
5%. Fig. 10 presents the comparison results of the experimental
and numerical average surface temperatures of the outer pipe.
The experimental surface temperature was calculated using Eq.
(2), and the area-weighted average over the entire lamppost was
considered for the simulation. As can be seen, all numerical values
are about 510% higher than the experimental ones, which might
be because four measuring points are not enough to accurately
capture the average surface temperature or because the numerical
model does not include the lamp stand under the fan chamber,
which also dissipates heat. Nonetheless, these errors are consid-
ered sufficient verification of the numerical model and setup.

4.3. Influence of varying boundary conditions

Fig. 9. Comparison of outlet temperatures between the experiment and the Fig. 11 shows the influence of the flow direction (CISE vs. SICE)
simulation. and the material of the center pipe on the heat loss for mass flow
rates ranging from 0.002 to 0.014 kg/s. Materials with lower ther-
mal conductivity can reduce the thermal contact between the
down and up streaming air flow. Besides PVC and Al, which were
also tested experimentally, the simulation included a pipe made
of fiberglass (Fg) with an extremely small conductivity of
k = 0.04 W/m2 K. The fiberglass pipe displays a very small improve-
ment compared to the PVC pipe, of 2.1% and 1.5% over the mea-
sured temperature range for CISE and SICE, respectively, and the
difference between the Al and PVC pipes is less than 0.5% for both
flow directions. Hence, it can be concluded that the thermal con-
ductivity has only a small influence on the overall heat loss of
the present configuration. However, the flow direction clearly
affects the rate of heat loss. On average, the CISE flow direction
has an 11.4% higher heat loss than SICE for all three considered
materials.
Fig. 12 illustrates the influence of the inlet temperature on the
rate of heat loss for different mass flow rates. The inlet tempera-
ture was varied in steps of 10 K ranging from 313 to 353 K, while
the ambient temperature was kept constant at 303 K. For higher
inlet temperatures, the rate of heat loss increases linearly, with
higher slopes for higher velocities. The increase in the heat loss rate
is significant when the flow rate is changed from 0.002 to 0.004 kg/
Fig. 10. Comparison of the average pipe surface temperature between the s, but is much smaller when the flow rate is changed from 0.01 to
experiment and the simulation. 0.014 kg/s. Changing the inlet temperature increases the heat loss
434 G. Schmid et al. / Applied Energy 195 (2017) 426438

Table 4
Heat loss rates per unit length.

_ (kg/s)
m 0.002 0.006 0.01 0.014
 (W/m) (CISE)
Q 1.81 8.15 12.58 15.73
 (W/m) (SICE)
Q 2.65 9.95 14.27 17.14

small area ratios lead to high pressure drops and should be avoided
[26]. Fig. 14 shows the influence of the pipe area ratio on the heat
loss for different mass flow rates. Increasing the area ratio from
0.16 to 0.54 can slightly increase the overall heat transfer for both
flow directions. A further increase from 0.54 to 1.0 can enhance the
SICE flow but has no effect on the CISE flow. For the SICE configu-
ration, the maximum increase is up to 12.4% at the highest mass
flow rate of 0.014 kg/s, while the CISE flow only increases by
4.8% at the same mass flow rate. The influence of the pipe area ratio
is negligible for low mass flow rates of m_ = 0.002 kg/s.
Fig. 12. Influence of inlet temperature on heat transfer (Ta = 303 K, PVC pipe, CISE).

4.5. Temperature profiles on centerlines

on average by 1.1 W/K at m_ = 0.002 kg/s, 2.4 W/K at m_ = 0.006 kg/s,
_ = 0.01 kg/s, and 3.6 W/K at m
3.1 W/K at m _ = 0.014 kg/s. Fig. 15 illustrates the temperature profiles on the centerlines of
the inner and outer pipes (annulus) for different flow configura-
4.4. Influence of geometrical parameters tions. For the CISE flow direction (Fig. 15(a) and (c)), the tempera-
ture decreases steadily from inlet to outlet. For the lower mass flow
All the experiments were conducted with a pipe length of 3 m. rate of m _ = 0.002 kg/s (Fig. 15(a)), the temperature drop over both
For the numerical analysis the length of the center pipe was varied pipes is almost equal, while at a higher flow rate of m _ = 0.014 kg/
from 2 to 5 m, which covers the practical range of the present heat s (Fig. 15(c)), most of the temperature drop occurs during the flow
exchanger. Fig. 13 depicts the heat transfer from the lamppost to through the outer pipe. Conversely, for the SICE flow direction
the surrounding air for different pipe lengths and flow directions. (Fig. 15(b) and (d)), which has completely different temperature
As expected, longer pipes lead to larger heat transfer rates, and profiles, the flow has a steep temperature drop along the outer cen-
the increase per unit length becomes more significant at higher terline, but increases again during the flow back up the center pipe.
mass flow rates. When extending the lamppost from 3 to 5 m, Fig. 15 also compares the influence of the center pipe material. The
the heat loss increases by 34.1% to 120.2 W at a high mass flow rate PVC and Al pipes have marginally different temperature profiles,
of 0.014 kg/s; however, at a low flow rate of 0.002 kg/s, it only but they result in almost the same exit temperature in all four
increases by 8.4% to 34.8 W. Table 4 lists the individual heat loss cases. This finding confirms the previous observation that an inner
rates per unit length in the range of 25 m. The SICE flow has pipe with a low conductivity cannot significantly prevent the heat
higher heat loss rates per unit length, which shows that the ther- transfer from the inner to the outer pipe. In addition, a significant
mal advantage of the CISE flow slightly decreases for longer pipes. temperature drop occurs for the flow through the fan chamber,
This study examined the influence of the ratio of the inner and especially for the CISE flow at low flow rates, which can be attrib-
outer pipe cross sections (Aip/Aop) on heat transfer using a PVC uted to a highly turbulent separated flow with swirl inside the fan
inner pipe, with a wall thickness of 3.5 mm. The diameter of the chamber and to the high temperature gradient between ambient
center pipe (Dip) was gradually increased, while the diameter of and pipe flow. The fan chamber is responsible for 42% of the total
the outer pipe (Dop) was kept constant. Table 5 shows the individ- temperature drop at m _ = 0.002 kg/s and for 38% at m_ = 0.014 kg/s.
ual parameters, of which the inner diameter is taken as the pipe
diameter. This analysis does not consider the influence of the area 4.6. Influence of the fan chamber diameter
ratio on the pressure drop; however, in general, too large and too
In the fan chamber, the air flow is diverted from the center pipe
to the outer pipe or vice versa. It was shown that the heat loss dur-
ing the flow through the fan chamber can account for more than
40% of the total heat loss. This part analyzes the influence of the
diameter of the fan chamber (dFC) in relation to the outer pipe
diameter (dop) on the thermal performance. In real applications,
the minimum value of dFC /dop is limited by the size of the fan used
to drive the system. Fig. 16 shows that at low mass flow rates, the
fan shape has a negligible influence on the heat loss rate. At higher
flow rates, the influence of the shape becomes more apparent,
especially for the CISE flow direction. At m_ = 0.014 kg/s, increasing
dFC/dop from 1 to 3.5 increases the heat loss rate by 29.1% and
17.6% for CISE and SICE, respectively.

4.7. Temperature and flow visualization

In this section, the results of the lamppost simulation are visu-

alized by plotting flow and temperature contours. Fig. 17 displays
Fig. 13. Heat loss over lamppost height for different flow conditions. the velocity vectors and the hydrodynamic boundary layers along
 G. Schmid et al. / Applied Energy 195 (2017) 426438 435

Table 5
Pipe ratios of lamppost and internal pipe.

Di (cm) Do (cm) Di/Do Inner pipe cross section Aip (cm2) Outer pipe cross section Aop (cm2) Aip/Aop
3.5 9.7 0.36 9.62 60.04 0.16
4.5 9.7 0.46 15.90 52.66 0.30
5.5 9.7 0.57 23.76 43.71 0.54
6.5 9.7 0.67 33.18 33.18 1.00

Fig. 14. Influence of the pipe ratio on the rate of heat loss (PVC pipe). Fig. 16. Heat loss for different ratios of dFC/dop.

_ = 0.002 kg/s, (b) SICE at m

Fig. 15. Temperature profiles on centerlines (a) CISE at m _ = 0.002 kg/s, (c) CISE at m
_ = 0.014 kg/s, (d) SICE at m
_ = 0.014 kg/s.
436 G. Schmid et al. / Applied Energy 195 (2017) 426438

_ = 0.002 kg/s, (b) CISE at m

Fig. 17. Comparison of boundary layer on surface of lamppost: (a) SICE at m _ = 0.002 kg/s, (c) SICE at m
_ = 0.014 kg/s, (d) CISE at m
_ = 0.014 kg/s.

the outer pipe surface of the lamppost. The thickness of the bound- The flow velocities and the shape of the boundary layers are very
ary layer is shown via isolines for velocities of 0.010.2 m/s, and similar for both flow directions on the upper part of the lamppost.
the flow on the pipe surface reaches velocities of over 0.4 m/s. However, on the lower part, the fan chamber clearly influences the
shape of the boundary layer depending on the flow direction
Fig. 18 shows the temperature contours inside the lamppost
and the distinguished thermal boundary layer for different config-
urations (scaled by 200% in radial direction). For the SICE flow
(Fig. 18(a) and (c)), a considerable amount of air is cornered at
the bottom of the fan chamber and does not take part in the flow
circulation; this becomes especially apparent at small mass flow
rates and is not desired, because it reduces the effective heat trans-
fer surface, which lowers the thermal performance. This phe-
nomenon also explains the sharp temperature drops at the
centerlines previously observed in Fig. 15(b) and (d). A possible
solution to this issue is to reduce the distance between the lower
exit of the internal pipe and the bottom of the fan chamber.

5. Conclusion

This paper analyzed a vertical double-pipe single flow heat

exchanger cooled by natural convection as part of an active cooling
system for LED street lamps, where the heat is dissipated via the
lamppost. This closed-loop cooling system is a viable solution
when passive cooling is insufficient. Possible applications include
large illuminations at industrial docks, stadiums, parking lots,
and storage or factory facilities, which often require large LED
arrays with combined power inputs of up to 500 W. The following
conclusions can be used as a guide when designing such an
installation:

Fig. 18. Temperature contours in and around the lamppost for different flow
1. Experiments and simulations have shown that more heat can be
conditions (scaled by 200% in radial direction): (a) SICE at m_ = 0.002 kg/s, (b) CISE at dissipated from the lamppost when the air flows down the cen-
m_ = 0.002 kg/s, (c) SICE at m
_ = 0.014 kg/s, (d) CISE at m
_ = 0.014 kg/s. ter pipe and back up via the outer pipe (CISE). The increase in
 G. Schmid et al. / Applied Energy 195 (2017) 426438 437

heat loss at a pipe length of 3.0 m is 8.2% at a mass flow of Therm Eng 2011;31:22219. http://dx.doi.org/10.1016/j.
applthermaleng.2011.03.003.
0.002 kg/s and up to 12.4% at a mass flow of 0.014 kg/s. How-
[10] Ye H, Li B, Tang H, Zhao J, Yuan C, Zhang G. Design of vertical fin arrays with
ever, this advantage decreases for longer pipes. heat pipes used for high-power light-emitting diodes. Microelectron Reliab
2. The heat dissipation can be slightly improved by using a center 2014;54:244855. http://dx.doi.org/10.1016/j.microrel.2014.05.004.
pipe with lower thermal conductivity. Pipes made of fiberglass [11] Moon S-H, Park Y-W, Yang H-M. A single unit cooling fins aluminum flat heat
pipe for 100 W socket type COB LED lamp. Appl Therm Eng 2016. http://dx.doi.
and PVC with thermal conductivities of 0.04 and 0.5 W/m2 K org/10.1016/j.applthermaleng.2016.11.077.
can increase the rate of heat loss by about 2.8% and 0.7% com- [12] Wang J-C. Thermal investigations on LED vapor chamber-based plates. Int
pared to Al with k = 204 W/m2 K. Commun Heat Mass Transfer 2011;38:120612. http://dx.doi.org/10.1016/j.
icheatmasstransfer.2011.07.002.
3. The additional heat loss by extending the lamppost largely [13] Ma J, Fu X, Hu R, Luo X. Effect of inclination angle on the performance of a kind
depends on the flow rate. When extending the lamppost from of vapor chamber. J Solid State Light 2014;1. http://dx.doi.org/10.1186/
3 to 5 m at a high mass flow rate of 0.014 kg/s, the heat loss s40539-014-0012-7.
[14] Deng X, Luo Z, Xia Z, Gong W, Wang L. Active-passive combined and closed-
is increased by 34.1% to 120.2 W; however, at a low flow rate loop control for the thermal management of high-power LED based on a dual
of 0.002 kg/s, it is only increased by 8.4% to 34.8 W. synthetic jet actuator. Energy Convers Manage 2017;132:20712. http://dx.
4. In general, mass flow rates smaller than 0.006 kg/s should be doi.org/10.1016/j.enconman.2016.11.034.
[15] Geisler K. Parametric design of a low-profile, forced convection heat sink for
avoided due to poor thermal performance, especially for longer high-power, high-density LED arrays. In: Intersociety conference on thermal
pipe lengths. The experiments also highlighted the importance and thermomechanical phenomena in electronic systems (ITHERM). Orlando,
of optimizing the flow rate to find the right balance between the FL; 2014. p. 4758. http://dx.doi.org/10.1109/ITHERM.2014.6892263.
[16] Iaronka O, Bender VC, Marchesan TB. Finite element analysis of a closed
power requirement for the fan and the reduction in LED junc-
cooling system applied to thermal management of LED luminaires. In:
tion temperature. Brazilian power electronics conference. Gramado, Brazil; 2013. p. 11539.
5. The shape and dimensions of the fan chamber influence the http://dx.doi.org/10.1109/COBEP.2013.6785260.
thermal performance of the cooling system. Hence, the design [17] Ma L, Yang Y, Liu J. Cooling of high power LEDs through ventilating ambient air
to front surface of chip. Heat Mass Transfer 2013;49:8594. http://dx.doi.org/
of the fan chamber needs to be carefully adjusted to the target 10.1007/s00231-012-1067-x.
flow rate to avoid airflow dead zones at the bottom. [18] Tsai H-L, Le PT. Self-sufficient energy recycling of light emitter diode/
thermoelectric generator module for its active-cooling application. Energy
Convers Manage 2016;118:1708. http://dx.doi.org/10.1016/j.
Conflict of interest enconman.2016.03.077.
[19] Jeng T-M. Combined convection and radiation heat transfer of the radially
finned heat sink with a built-in motor fan and multiple vertical passages. Int J
Author C has applied for a patent titled Heat-dissipating struc- Heat Mass Transfer 2015;80:41123. http://dx.doi.org/10.1016/j.
ture having embedded support tube to form internally recycling ijheatmasstransfer.2014.09.043.
[20] Ramos-Alvarado B, Feng B, Peterson GP. Comparison and optimization of
heat transfer fluid and application apparatus (EP 2818818 A3,
single-phase liquid cooling devices for the heat dissipation of high-power LED
US 2015/0000875 A1), which is related to the active cooling system arrays. Appl Therm Eng 2013;59:64859. http://dx.doi.org/10.1016/j.
of the present study. applthermaleng.2013.06.036.
[21] Yang C-Y, Liu W-C, Yeh C-T, Tsai S-N. A novel liquid cooling system for high
power LED street lamp semiconductor thermal measurement and
Acknowledgement management symposium (SEMI-THERM). San Jose, CA; 2013. p. 1513.
http://dx.doi.org/10.1109/SEMI-THERM.2013.6526820.
[22] Hsieh S-S, Hsu Y-F, Wang M-L. A microspray-based cooling system for high
The authors gratefully thank Giant Lion Know-How Co., Ltd. for powered LEDs. Energy Convers Manage 2014;78:33846. http://dx.doi.org/
supporting this project by manufacturing the lamppost and the 10.1016/j.enconman.2013.10.066.
[23] Ramos-Alvarado B, Li P, Liu H, Hernandez-Guerrero A. CFD study of liquid-
cooling system for the experimental analysis. This research did
cooled heat sinks with microchannel flow field configurations for electronics,
not receive any specific grant from funding agencies in the public, fuel cells, and concentrated solar cells. Appl Therm Eng 2011;31:2494507.
commercial, or not-for-profit sectors. http://dx.doi.org/10.1016/j.applthermaleng.2011.04.015.
[24] Jalaluddin, Miyara A, Tsubaki K, Inoue S, Yoshida K. Experimental study of
several types of ground heat exchanger using a steel pile foundation. Renew
References Energy 2011;36:76471. http://dx.doi.org/10.1016/j.renene.2010.08.01.
[25] Beier RA, Acua J, Mogensen P, Palm B. Borehole resistance and vertical
temperature profiles in coaxial borehole heat exchangers. Appl Energy
[1] Thorseth A. Characterization, modeling, and optimization of light-emitting
2013;102:66575. http://dx.doi.org/10.1016/j.apenergy.2012.08.007.
diode systems [PhD Thesis]. Kgs. Lyngby: Technical University of Denmark;
[26] Yekoladio PJ, Bello-Ochende T, Meyer JP. Design and optimization of a
2011.
downhole coaxial heat exchanger for an enhanced geothermal system (EGS).
[2] Liu S, Luo X. LED packaging for lighting applications: design manufacturing and
Renew Energy 2013;55:12837. http://dx.doi.org/10.1016/j.
testing. Singapore: John Wiley and Sons (Asia) Pte Ltd.; 2011. http://dx.doi.org/
renene.2012.11.035.
10.1002/9780470827857.
[27] Acua J, Palm B. Distributed thermal response tests on pipe-in-pipe borehole
[3] Mahalingam R. Air cooling for LED lighting. In: Lasance CJM, Poppe A, editors.
heat exchangers. Appl Energy 2013;109:31220. http://dx.doi.org/10.1016/j.
Thermal management for LED applications. New York: Springer Science and
apenergy.2013.01.024.
Business Media; 2014. p. 26798. http://dx.doi.org/10.1007/978-1-4614-
[28] Holmberg H, Acuna J, Nss E, Snju OK. Thermal evaluation of coaxial deep
5091-7.
borehole heat exchangers. Renew Energy 2016;97:6576. http://dx.doi.org/
[4] Jang D, Yook S-J, Lee K-S. Optimum design of a radial heat sink with a fin-
10.1016/j.renene.2016.05.048.
height profile for high-power LED lighting applications. Appl Energy
[29] Jalaluddin Miyara A. Thermal performance investigation of several types of
2014;116:2608. http://dx.doi.org/10.1016/j.apenergy.2013.11.063.
vertical ground heat exchangers with different operation mode. Appl Therm
[5] Schmid G, Yang L-H, Yang T-H, Chen S-L. Influence of inter-fin base length on
Eng 2012;3334:16774. http://dx.doi.org/10.1016/j.
thermal performance of free hanging horizontal base heat sinks. Appl Therm
applthermaleng.2011.09.030.
Eng 2016;108:122636. http://dx.doi.org/10.1016/j.
[30] Omidi M, Farhadi M, Jafari M. A comprehensive review on double pipe heat
applthermaleng.2016.08.034.
exchangers. Appl Therm Eng 2017;110:107590. http://dx.doi.org/10.1016/j.
[6] Wang J-C. Thermal module design and analysis of a 230W LED illumination
applthermaleng.2016.09.027.
lamp under three incline angles. Microelectron J 2014;45:41623. http://dx.
[31] Schmid G, Hsu C-Y, Chen Y-T, Yang T-H, Chen S-L. Performance study of an oil-
doi.org/10.1016/j.mejo.2014.01.016.
immersed power transformer with shallow geothermal cooling. J Therm Sci
[7] Shen Q, Sun D, Xu Y, Jin T, Zhao X. Orientation effects on natural convection
Eng Appl 2016;8. http://dx.doi.org/10.1115/1.4031889.
heat dissipation of rectangular fin heat sinks mounted on LEDs. Int J Heat Mass
[32] Kaiser AS, Zamora B, Mazn R, Garca JR, Vera F. Experimental study of cooling
Transfer 2014;75:4629. http://dx.doi.org/10.1016/j.
BIPV modules by forced convection in the air channel. Appl Energy
ijheatmasstransfer.2014.03.085.
2014;135:8897. http://dx.doi.org/10.1016/j.apenergy.2014.08.079.
[8] Jeong MW, Jeon SW, Kim Y. Optimal thermal design of a horizontal fin heat
[33] Teo HG, Lee PS, Hawlader MNA. An active cooling system for photovoltaic
sink with a modified-opening model mounted on an LED module. Appl Therm
modules. Appl Energy 2012;90:30915. http://dx.doi.org/10.1016/j.
Eng 2015;91:10515. http://dx.doi.org/10.1016/j.
apenergy.2011.01.017.
applthermaleng.2015.08.001.
[34] Lienhard IV JH, Lienhard VJH. A heat transfer textbook. Cambridge,
[9] Lin Z, Wang S, Huo J, Hu Y, Chen J, Zhang W, et al. Heat transfer characteristics
Massachusetts: Phlogiston Press; 2010.
and LED heat sink application of aluminum plate oscillating heat pipes. Appl
438 G. Schmid et al. / Applied Energy 195 (2017) 426438

[35] Cengel YA, Cimbala JM. Fluid mechanics: fundamentals and applications. 1st [40] Kreith F, Manglik RM, Bohn MS. Principles of heat transfer. 7th ed. Stamford,
ed. New York: McGraw-Hill; 2006. CT: Cengage Learning; 2011. p. 3056.
[36] Holman JP. Experimental methods for engineers. 7th ed. McGraw-Hill [41] Bakker A. Lecture 10 turbulence models. Applied computational fluid
Mechanical Engineering; 2007. dynamics. Dartmouth College, Hanover, NH; 2006. http://www.bakker.
[37] Andersson B, Andersson R, Hakansson L, Mortensen M, Sudiyo R, Wachem BV, org/dartmouth06/engs150/10-rans.pdf [accessed 13.03.03].
et al. Computational fluid dynamics for engineers. 1st ed. Edinburgh Building, [42] Yang J-F, Zeng M, Wang Q-W. Numerical investigation on shell-side
Cambridge, UK: Cambridge University Press; 2012. performances of combined parallel and serial two shell-pass shell-and-tube
[38] ANSYS, Inc., Ansys Fluent theory guide Release 14.5. Canonsburg, PA; 2013. heat exchangers with continuous helical baffle. Appl Energy
[39] Boetcher SKS. Natural convection from circular 2015;139:16374. http://dx.doi.org/10.1016/j.apenergy.2014.11.029.
cylinders. Heidelberg: Springer; 2014. http://dx.doi.org/10.1007/978-3-319-
08132-8.