You are on page 1of 10

International Communications in Heat and Mass Transfer 57 (2014) 140149

Contents lists available at ScienceDirect

International Communications in Heat and Mass Transfer


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

Performance effects of heat transfer and geometry on heat pipe thermal


modules under forced convection
Rong-Tsu Wang a, Ya-Wei Lee b, Sih-Li Chen c, Jung-Chang Wang d,
a

Department of Airline and Transport Service Management, Vanung University, No.1, Van-Nung Rd., Chung-Li, Tao-Yuan 32061, Taiwan, ROC
Department of Mechatronic, Energy, and Aerospace Engineering, Chung Cheng Institute of Technology, National Defense University, Tauyuan 33551, Taiwan, ROC
Department of Mechanical Engineering, National Taiwan University, Taipei 10617, Taiwan, ROC
d
Department of Marine Engineering, National Taiwan Ocean University (NTOU), Keelung 20224, Taiwan, ROC
b
c

a r t i c l e

i n f o

Available online 2 August 2014


Keywords:
Heat pipe
Heat sink
Thermal module
Thermal resistance
Least square method

a b s t r a c t
The geometry and heat transfer effects on heat pipes embedded heat sinks-cooling system are investigated in the
present paper. In the forced convection system, two different heat pipe geometrical shapes of L and U congurations are taken into account. This study adopts a versatile superposition method and least-square estimators with
thermal resistance network analysis to design and experiment their geometry and heat transfer effects under
different fan speeds and heat source areas. The results suggest that the characteristics of system performance
under varying speeds and areas are signicantly different from those under altering speeds and areas. When
the thermal performances of these two congurations are 0.04 C/W of U-shaped heat pipes at 78.85 W, and
L-shaped heat pipes are lowest 1.04 C/W at 34 W, respectively, the lowest thermal resistances of the representative L- and U-shaped heat pipeheat sink thermal modules are respectively 0.25 C/W and 0.17 C/W under
twin fans of 3000 RPM and 30 30 mm2 heat sources. The result of this work is a useful thermal management
method to facilitate rapid analysis and has provided a useful insight into the design of heat pipe cooling systems.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
The extended surface n is usually added to enhance the rate of heat
removal for traditional air cooling techniques. The conventional way to
dissipate heat from microprocessors especially in Central Processing
Units (CPUs) and Graphic Processing Units (GPUs) cooling was forced
convection using a fan with a heat sink directly. However, with the
advances in computer and semiconductor manufacturing industry,
integral circuit design is getting more and more complex so that the
electronic components of CPUs, GPUs, and LED lighting lamps (LEDs)
are made toward the small size, increased power and high efciency
development. They generate more and more heat and heat ux is significantly increased. Thus, the cooling problems of electronic device are
daily major [13]. The manufacture technology of embedded heat
pipes into heat sinks is rapid developments to ensure faster dissipation
of the heat which is a quite obvious application in high-performance
cooling devices. Therefore, heat pipe thermal modules that transfer
energy away from the heat source through convection mechanism
possessing simple metal heat sinks and fans are used to solve the hotspot problems [47]. Technical development related with the application of two-phase ow heat transfer assembly to thermal modules has
Communicated by W.J. Minkowycz.
Corresponding author at: No.2, Beining Rd., Keelung 20224, Taiwan, ROC.
E-mail address: jcwang@ntou.edu.tw (J.-C. Wang).

http://dx.doi.org/10.1016/j.icheatmasstransfer.2014.07.023
0735-1933/ 2014 Elsevier Ltd. All rights reserved.

become mature. Heat pipe absorbs large amounts of latent heat from a
heat source through the phase change of working uid inside and transfers rapidly heat ow to the other side by vapor form without any uid
machinery. The computer-aided thermal design of heat sink with heat
pipe/vapor chamber progresses a high-quality thermal condition development [8]. The heat pipe thermal module has better thermal performance and is that most of the heat rst transfers to the evaporator
of a heat pipe, so the evaporated liquid working uid of the heat pipe
produces steam. Steam releases heat through condensation and recondenses into liquid returning to the evaporator, driven by capillary
force, while the rest of the heat capacity is removed from the heat sink
through forced fan convection [9,10]. The thermal dissipation performance of some heat pipes bent into the required geometric shape and
embedded into the metal base plate or n stack compares well with
that of a vapor chamber. Wang et al. [11,12] had experimentally studied
the thermal resistances of an aluminum heat sink with horizontallyembedded two and four U-shaped heat pipes of 6 mm diameter under
xed heat source area and single fan. They showed that two heat
pipes embedded in the base plate carry 36% of the total dissipated
heat capacity from the heat source, while 64% of heat capacity was
delivered from the base plate to the ns. Moreover, if the temperature
of the heat source is not allowed to exceed 70 C, the total heat adsorption power of heat sink with two and four embedded heat pipes will not
respectively exceed 131 W and 164 W. Finally, a program was developed using Visual Basic to rapidly calculate the thermal performance

R.-T. Wang et al. / International Communications in Heat and Mass Transfer 57 (2014) 140149

Nomenclature
k
L
Q
Qb
R
Rt
Rh
T

thermal conductivity, W/mC


distance between evaporation and condensation sections of heat pipe, meter
total heat transfer rate, Watt
heat transfer rate from base plate to ns, Watt
thermal resistance, C/W
total thermal resistance, C/W
heat pipe thermal resistance, C/W
temperature, C

Subscripts
a
air/ambient
b
base plate
f
n
s
heat source/dummy heater
n
position of embedded L-/U-shaped heat pipes

of a heat sink with embedded heat pipes [13]. Mohamed et al. [14,15]
investigated also the thermal performance of a heat sink with nned
U-shaped heat pipes for optimum L-ratio (ratio of the evaporator section length to the condenser section length) of the U-shaped heat
pipe, which was found to be dependent on heat pipe diameter and the
n spacing and was of practical engineering importance in the optimum
design of the heat sink. The performance analysis of a nned U-shaped
heat pipe used for desktop PC-CPU cooling was estimated for both
natural and forced convection modes under steady state condition.
Russel et al. [16] identied the effect of orientation on the performance of the U-shaped heat pipe with grooved and sintered wick structures. Thermal modules with U-shaped heat pipes are currently used for
CPU/GPU cooling. Natural and forced convection can both be used to obtain optimum results for minimum thermal resistance and lead to the
generation of high capillary forces for anti-gravity applications [17].
The thermal performance of a heat sink with nned U-shaped heat
pipes is carried out to compatible research for a wide range of highfrequency microprocessors. Besides, another heat pipe bent into metal
base plate is the L shape. One set of risers of the L-shaped heat pipes
functions as the evaporating section while the other set attached to
ns acts as the condensing section. This shape of heat pipeheat sink
is particularly well-suited for cooling electronic components such as
microprocessors using forced convection. Wang [18] vertically arranged
six L-type heat pipes in such a way that the bottom acts as the evaporating section and the risers act as the condensing section, and derived a
mathematical model including all major components from the thermal
interface through the heat pipes and ns. A Windows-based computer
program also uses an iterative superposition method to predict the
thermal performance. Thermal performance testing shows that a representative heat sink with six L-type heat pipes will carry 160 W and has
reached a minimum thermal resistance of 0.22 C/W. The total thermal
resistance varies according to the functionality of the L-type heat pipes.
In recent years, heat pipe-based two-phase ow heat transfer modules
have emerged to effectively reduce the temperature of small-area LED
lighting lamps with higher degrees of heat ux [19,20]. Wang [21,22]
analyzed and designed the optimum thermal performance of a
at heat pipe-thermal module application in a high-end VGA card
cooling system, which is able to cope with a heat ux GPU of over
62.5 W/cm2. The optimum total thermal resistance of a at heat
pipe-thermal module is 0.232 C/W at a high power GPU of 180 W
and inclination angle of 180. The technical development of two-phase
ow heat transfer assembly to thermal modules has matured and is
one of the best options, especially in LED thermal problems. Lu et al.

141

[23] used the at heat pipe (FHP or vapor chamber) to improve the thermal characteristics of a high power LED (light-emitting diode) package.
The obtained results indicated that the junction temperature of the LED
is about 52 C for an input power of 3 W and, thus, the total thermal
resistance of LED system is 8.8 K/W. Wang et al. [2426] presented a
successful experimental analysis with VCTM V1.0 to develop a 30 W
high-power LEDs vapor chamber-based plate, nding that the thermal
performance of the LED vapor chamber-based plate was an improvement over that of the LED copper-based plate with an input power
above 5 W. Results show that the maximum effective thermal conductivity is 870 W/mC and, compared with the experimental value, the
calculating error is no more than 5%. The LED vapor chamber-based
plate successfully resolved the hot-spot problem of 30 W high-power
LEDs.

(a) U-shaped heat pipes thermal module

(b) L-shaped heat pipes thermal module


Fig. 1. Heat pipe thermal modules (a) U-shaped heat pipe thermal module (b) L-shaped
heat pipe thermal module.

142

R.-T. Wang et al. / International Communications in Heat and Mass Transfer 57 (2014) 140149

The heat pipe thermal module studied in the present study is shown
in Fig. 1. The present study is to experiment two serial fans set in the
opposite sides of the heat sink, providing a double volume ow rate of
a single fan different from the ones before and to investigate their
thermal performances by using the superposition and least square
smoothing methods [27].
The U-shaped heat pipe thermal module is composed of an aluminum base plate with ns, two heat pipes, and an aluminum n stack.
The dimensions of base plate are 52 42 7.5 mm3 and its ns are
10 42 7.8 mm3. The n stack is an array of 35 ns, 0.4 mm thick
with a 2.0 mm gap. Its dimensions are 42 mm 82 mm. This type of
thermal module has two parallel heat pipes. The total length of single
heat pipe is 265 mm with an evaporating length of 30 mm directly
contacting the heat source, and two condensing lengths of 117.5 mm
embedded in the ns. One end of the heat pipes is inserted into the
base plate to form the evaporation section, while the other end is embedded in the ns to form the condensation section. Another L-shaped
heat pipes heat sink consisted of a copper base plate, six heat pipes,
and an aluminum n stack. The overall dimensions of the L-shaped
heat pipes heat sink are 117 85 83 mm3, which consists of six
single heat pipes with aluminum ns 83 mm 70 mm, 0.4 mm of
thickness, a count of 27 and a pitch of 2.8 mm. The copper base plate
measures 66 51 7 mm3. These six L-shaped heat pipes are
156 mm in length, including 50 mm embedded in the base plate to
carry heat from the heat source to the ns for dissipation. The heat
pipe is bent at a right angle to be an L-shaped heat pipe and has a
51 mm evaporator and a 105 mm condenser. However, single U- or
L-shaped heat pipe has 6 mm diameter and is made of copper with a
sintered copper powder metal wick of 10 5 m effective pore radius
and pure water as working uid.
2. Methodology
2.1. Thermal resistance network and superposition method
Fig. 2 exhibits the thermal resistance network of the heat pipe thermal modules acquired by the superposition method [11]. The superposition rule is sufcient to present the given heat transfer geometry
problem geometry in Fig. 2 as the sum or difference of some geometry
with known conguration factors (Fig. 2 in Refs. [11,18]), which can
be expressed as the parameters. Therefore, superposition principle is
also often stated in the summation rule. We may divide the heat transfer
problem into a number of simpler problems each satised the required
homogeneity conditions and all adding up to the problem in the present

study. Homogeneity means that a linear differential equation or a linear


boundary condition is satised by a function with arbitrary constants.
Consequently, the number of the present heat transfer problems is
equal to the number of non-homogeneities involved in the formulation
of the actual problem. The heat diffusion problem is derived and reduced to a number of simpler parameters by employing superposition
method in the present paper. Finally, superposition method affects the
homogeneity but not the type of differential equation or boundary conditions. And each heat transfer problem involved in a superposition rule
has the same geometry. The related derivation processes are similar to
those shown in Ref. [11]. In summary, the superposition method can
be used to reduce a linear heat diffusion problem to a number of simpler
problems, which shows that heat transfer rate Q transfers heat from the
heat source to the ns through the base plate and the embedded heat
pipes.
From Fig. 2 heat transferred from heat source to the lower surface of
the base plate is divided into two parts, which heat transfer rate Q transfers heat from the heat source to the ns through the base plate and the
heat pipes. The rst part,that is the base plate transfers heat to the heat
sink of the second metal block and ns, is the bypass heat ow Qb. And
the second,that is the heat pipes embedded asymmetrically in the base
plate transfer heat from the base plate to ns, is the bypass heat ow Qn,
which transfers heat from the evaporator of the heat pipe to the condenser and ns, and next dissipates in the ambient surroundings due
to the forced convection. The subscript of n denotes the position of the
embedded heat pipes. The bypass heat ow Qb can be evaluated by
comparing the temperature differences corresponding to the input
heating power as claried in Ref. [11] resulting to the thermal performance of the thermal module which is estimated separately by experiments with and without embedded heat pipes of the thermal module.
From the superposition method, the bypass heat ow Q equals the
sum of Qb and Qn as shown in Eq. (1). Symbol n equals 1 to 2 for the
U-shaped heat pipe thermal module; otherwise, n is 1 to 6 for the
L-shaped heat pipe thermal module. Consequently, Qn thus equals Q
minus Qb.
Q Qb

Qn

n1

The measured temperatures of the U-shaped and L-shaped heat pipe


thermal modules as show in Fig. 3. Ts is the temperature of the heat
source, Tb is the temperature of the base plate, Tbu is the temperature
of the upper base plate, Thpe,n is the temperature of the evaporator
of the single embedded heat pipes, Thpc,n is the temperature of the
condenser of the single embedded heat pipes and Ta is the ambient
temperature. The total thermal resistance Rt can be expressed as the
sum of the thermal contact resistance Rif and the thermal resistances
on the pathways of Qb and Qn, which is
Rt Rif

Fig. 2. Thermal resistance network.

2
or6
X

!
1
1
:

Rb Rhs n1 Rbh;n Rh;n R f;n


2
or6
X

In Eq. (2), Rt is dened as the temperature difference (the temperature of heat source Th minus the ambient temperature Ta) divided by the
total heating power Q. The Rif is the interface resistance, dened as the
effective temperature difference at the interface (Th minus the temperature at the center of the lower surface of the base plate Td) divided by
Q. The Rb is the base plate thermal resistance, dened as the temperature difference (Td minus the average temperature at the upper surface
of the base plate Tu) divided by Qb. The Rhs is the base plate convection
resistance, dened as the temperature difference (Tu minus Ta) divided
by Qb. The Rbh,n is the base plate heat pipe thermal resistance, dened as
the temperature difference (Td minus the temperature of the evaporation section of the heat pipes Ten) divided by Qn. The Rh,n is the heat

R.-T. Wang et al. / International Communications in Heat and Mass Transfer 57 (2014) 140149

143

(a) U-shaped heat pipes-heat sink

(b) L-shaped heat pipes-heat sink


Fig. 3. Position of measured temperatures (a) U-shaped heat pipe thermal module (b) L-shaped heat pipe thermal module.

pipe resistance, dened as the temperature difference (Ten minus the


temperature of the condensation section of the heat pipes Tcn) divided
by Qn. The Rf,n is the heat pipe convection resistance, dened as the temperature difference (Tcn minus Ta) divided by Qn. As symbol of n equals
1 to 2, the thermal module is U-shaped heat pipeheat sink including

nine individual thermal resistances. And the thermal module is


L-shaped heat pipeheat sink with twenty-one thermal resistances
when n is 1 to 6. The theoretical heat pipe thermal resistance from
Ref. [18] can be obtained through the temperature difference between
Thpe,n and Thpc,n as shown in Eq. (3). The ratio of bypass heat of the

Fig. 4. Experimental apparatus.

144

R.-T. Wang et al. / International Communications in Heat and Mass Transfer 57 (2014) 140149

heat pipe to total heat is utilized through Newton's method and the
GaussSeidel iteration method.

Rh

Thpe;n Thpc;n
Qn


4

0:5940:032  Q n 4:781  10
7

1:596  10

 Q n

 Q n

2.2. Experimental procedure and least-squares estimation


The experimental apparatus employed to measure the thermal performance of thermal module is shown in Fig. 4. The thermal module is
located on a dummy heater and connected to all the apparatus. The
dummy heater supplies a xed amount of heating power to simulate
the real heat source (CPU, GPU, etc.). The dummy heater consisted of a
copper block of 15 50 100 mm3 with two embedded 35 electrical
resistance rods of 10 mm diameter. And two different contacting areas
of heat source (15 15 mm2 and 30 30 mm2, respectively) are utilized
to attach to the heat pipe evaporator in the present experiment. An adjustable DC power supplier of ADC50-10 possesses a maximum voltage
of 50 V and current of 10 amps to allow calculation of input power
under an error range 0.5%. The heating power can be increased to 300
W in steps of 30 W. Bakelite and adiabatic materials prevent heat loss
around the copper block. Thermal grease with 2.5 W/mk is applied to
the mounting surface of the heater block to reduce the thermal interface
resistance. Two same fans of 80 80 25 mm3 are mounted at the right
and left sides of the cooling module to disperse heat through forced convection to ambient air while maintaining a constant ow direction and
rotational speed. In a forced convection system, the uid movement is
by an external force as fans. If the uid velocity is large, then turbulence
is induced. And the mixing of hot and cold air is more efcient, there is
an increase in heat transfer. The commercial specications of the present fan are that the air volume ows and pressures are respectively
from 10.9 to 40.8 CFM and 41.454 to 2.94 Pa between 800 to
3000 RPM with an operating voltage of 12 V and maximum current of
0.26 A. And the noise values are 13 to 28 dB. Therefore, the maximum
noise maybe 28 dB when a single fan is running at 3000 RPM. Fan
speed can be set at 1000, 2000 and 3000 RPM. The GL-800 data recorder
can measure 40 channels including temperature, voltage, and humidity,
with a sampling time of 104 s. Instrumentation consisted of type T
thermocouples with a diameter of 0.32 mm and a measuring range of

200 C350 C are mounted to all measuring positions involving the


dummy heater, base plate, heat pipes, ns and surroundings. The thermocouples used in the present experiment have a measurement error of
0.5 C. The data recorder manufactured has a measurement error of
1%. The power supply unit has a measurement error of 0.5%. Finally,
the maximum error for the thermal resistance is thus within 5%.
Investigations on the experimental procedure were divided into two
parts. The rst part referring to the superposition method [11] is to
measure the thermal performance of the entire heat sink with all heat
pipes intact. After separately testing the whole thermal performance
of the heat sink with embedded U-shaped and L-shaped heat pipes in
the above experiments, the second part is the heat sink experiments
without heat pipe function. Let all U- (n = 12) and L-shaped heat
pipes (n = 16) in the heat sink fail to function successively. The thermal performances are measured as the U-shaped and L-shaped heat
pipes are successively disabled in positions n = 12 and 16 by cutting
through their adiabatic region, heating them to over 100 C, and
allowing the working uid to vaporize and escape. Because heat ow
could neither travel by two-phase working uid ow or by heat pipe
walls conduction, this cutting process is to reduce the conductance of
the heat pipe to zero. The heat capacity transference from the base
plate and the U-shaped and L-shaped heat pipes to the ns can be obtained by comparing the individual results across the same temperature
differences, respectively. As the geometry is symmetric, the analysis is
carried out for half of the U-shaped heat pipe. The related processes
are presented in Ref [18], and are not presented here. The heat sink
experiments without heat pipes function are then performed, and the
corresponding thermal resistances are determined. The heat pipe is a
two-phase heat transfer component, and its thermal resistance will
change with the heating power. We express and derive a mathematical
model to predict the theoretical thermal resistance of heat pipe utilized
in the present study. Rather than conducting thermal performance
experiments of independent single heat pipes to reverse calculate the
experimental thermal resistance of a single heat pipe through thermal
resistance analysis and bypass heat ow, this study used the leastsquare estimator formula [27] to t the experimental values as the
theoretical thermal resistance of the heat pipe. The present formula is
based on our thermal performance experiments of heat pipe thermal
modules that the rates of change in input heating power and temperature during the execution of an application depends on the temperature
differences between the evaporating and the condensing sections.
Based on the single heat pipe relationship established by Ref. [18] and

Table 1
Relationship between interface thermal resistances Rif and input power.
Heat input (W)

Area of heat source (mm2)

Area of heat source (mm2)

15 15

30 30

Two fans

One fan

Two fans

Fan speeds (RPM)

L-shaped

U-shaped

30
60
90
120
150
180
240
300
30
60
90
120
150
180
240
300

One fan

Fan speeds (RPM)

1000

2000

3000

1000

2000

3000

1000

2000

3000

1000

2000

3000

0.033
0.033
0.032
0.034
0.034

0.033
0.035
0.033
0.034
0.034
0.035

0.031
0.03
0.039
0.038
0.038
0.037

0.037
0.037
0.034
0.036
0.036
0.037

0.033
0.034
0.034
0.034
0.037
0.036

0.037
0.037
0.037
0.037
0.04
0.038

0.033
0.032
0.032
0.034
0.034

0.033
0.035
0.037
0.037
0.037
0.035

0.033
0.032
0.033
0.033
0.035
0.036

0.037
0.033
0.034
0.036
0.037
0.039

0.033
0.032
0.032
0.031
0.033
0.034

0.037
0.035
0.032
0.037
0.037
0.039

0.036

0.037

0.038
0.036
0.037

0.032

0.033

0.034
0.035

0.037

0.036

0.036
0.037
0.037

0.033

0.033

0.033
0.034
0.035

0.036

0.036

0.037
0.037
0.037

0.032

0.033

0.031
0.03
0.033

0.035

0.036

0.035
0.034

0.037

0.035

0.036
0.035

0.035

0.034

0.035
0.035
0.035

0.033

0.034

0.034
0.035
0.035

0.033

0.034

0.033
0.034
0.035

0.033

0.035

0.034
0.035
0.035

R.-T. Wang et al. / International Communications in Heat and Mass Transfer 57 (2014) 140149

145

Table 2
Relationship between total thermal resistances and input power.
Heat input (W)

Area of heat source(mm2)

Area of heat source(mm2)

15 15

30 30

Two fans

One fan

Two fans

Fan speeds (RPM)

L-shaped

U-shaped

30
60
90
120
150
180
240
300
30
60
90
120
150
180
210
240
300

One fan

Fan speeds (RPM)

1000

2000

3000

1000

2000

3000

1000

2000

3000

1000

2000

3000

0.55
0.5117
0.4956
0.49
0.4927
0.5

0.417
0.407
0.404
0.399
0.377

0.4367
0.42
0.4144
0.42
0.4353
0.4483

0.397
0.37
0.366
0.353
0.336
0.318
0.307

0.4133
0.3933
0.3867
0.3983
0.4187
0.4383

0.343
0.297
0.288
0.287
0.282
0.281
0.281

0.6433
0.62
0.6133
0.6117
0.6117
0.6217

0.557
0.507
0.501
0.491
0.49

0.5167
0.4717
0.4589
0.4642
0.47
0.4844

0.417
0.362
0.351
0.343
0.338
0.346

0.4367
0.415
0.4078
0.4192
0.432
0.45

0.363
0.312
0.303
0.301
0.298
0.298

0.3767

0.3733

0.3667
0.3617

0.29

0.276

0.271

0.268
0.274

0.275

0.2708

0.2656
0.2621
0.2613

0.207

0.196

0.19

0.188
0.184

0.245

0.2408

0.2477
0.2479
0.2497

0.175

0.17

0.164

0.161
0.158

0.47

0.4667

0.455
0.455

0.397

0.383

0.354

0.355

0.2983

0.295

0.2878
0.2867
0.286

0.233

0.228

0.228

0.222
0.222

0.26

0.2525

0.2478
0.2467
0.252

0.192

0.184

0.18

0.178
0.177

the general least-squares problem, the output of thermal resistance of a


single heat pipe is assumed to be as follows [26]. And then, they were
acquired by using the least square method to t the discrete points, on
the basis of this, the navigation line detection algorithm was proposed.
Detailed descriptions of Least Square Method can be found in the literature [27].
3. Results and discussions
Table 1 reveals the thermal interface resistance Rif of L- and
U-shaped heat pipeheat sinks to the total heating power. Rif is dependent on mechanical and thermal properties as well as surface roughness
properties and contact pressure between two contacting surfaces.
Moreover, thermal interface materials with high thermal conductivity
can validly reduce the thermal interface resistance. Given a constant
input power, the different between the two thermal modules is not
large. The reason is that the thermal conductivity of thermal grease is
the same when there is not much change in temperature. The Rif is
approximately 0.035 C/W when the heating power is between 30 W
and 300 W. Therefore, the Rif may be treated as a constant in this experiment. The magnitude of Rif is a major part in thermal management and
control of electronic devices and, hence, may signicantly affect the
thermal performance of such thermal module. Table 2 demonstrates
the total thermal resistance Rt of the L- and U-shaped heat pipeheat
sinks to the total heating power under variable fan speeds and heat
source areas. Generally, the total thermal resistances decrease as fan
speeds increase. The thermal module should be improved thermal performance at higher fan speeds. Forced convection is a primary function

of the uid velocity and uid properties, and it is a secondary function


of the temperature because the uid properties are temperature dependent. Empirical dimensionless analysis formulations combined by
Nusselt number (Nu), Reynolds number (Re) and Prandtl number (Pr)
are used to calculate convection heat transfer for forced convection.
The magnitude of Re is used to judge if there is laminar or turbulent
ow in a forced convection. The Nu is almost constant for fully developed laminar channel ow; otherwise, it is no longer as constant for
turbulent ow region. Coolant ows from fans are regulated so the
convection in the channels of heat sink was clearly in either the turbulent or laminar ow region and, as expected, the former gave lower
thermal resistance than the latter. The Rt exhibits small changes for
fan speeds of 2000 RPM and 3000 RPM and 30 30 mm2 heat source.
This is because of larger area of heat source and lower heat ux than
15 15 mm2 heat source at same heating power. The respective average total thermal resistances are 0.17 C/W (3000 RPM), 0.19 C/W
(2000 RPM) and 0.28 C/W (1000 RPM) for L-shaped heat pipe
heat sink, and 0.25 C/W (3000 RPM), 0.27 C/W (2000 RPM)
and 0.37 C/W (1000 RPM) for U-shaped heat pipeheat sink. The differences of Rt are both about 0.1 C/W between 2000 and 1000 RPM.
The total thermal resistances have not large change at the dual fans
speeds of 1000 RPM and above input power of 90 W at 30 30 mm2
area of heat source. The differences of total thermal resistances between
fan speeds of 2000 RPM and 3000 RPM are almost the same; the thermal
module has well thermal performance at 2000 RPM at 30 30 mm2
area of heat source.
From Table 2, the total thermal resistances Rt progressively decrease
as input powers increase, with the lowest values Rt between 90 W to

Table 3
The bypass heat ow ratios.
L-shaped

U-shaped

Area of Heat
Source(mm2)

Ratio

Two fans,
2000 RPM

One fan,
3000 RPM

One fan,
1000 RPM

30 30

I (n = 1,6)
II (n = 2,5)
III (n = 3,4)
Qb/Q
I (n = 1,6)
II (n = 2,5)
III (n = 3,4)
Qb/Q

28%
27%
30%
15%
35%
27%
25%
13%

27%
30%
29%
14%
33%
25%
25%
17%

47%
20%
10%
23%
38%
28%
20%
14%

15 15

Ratio

Two fans,
2000 RPM

Two fans,
3000 RPM

One fan,
2000 RPM

One fan,
3000 RPM

(n = 1 + 2)

68%

68%

69%

67%

Qb/Q

32%

32%

31%

33%

(n = 1 + 2)

67%

65%

72%

63%

Qb/Q

33%

35%

28%

37%

146

R.-T. Wang et al. / International Communications in Heat and Mass Transfer 57 (2014) 140149

150 W for 15 15 mm2 heat source, and the smaller heat source area
results in higher heat ux. After the heating power increases to 150 W
(67 W/cm2), Rt rises and launches because of the operating performance limits of the embedded heat pipes. This is the heating power
increasing to the point that the embedded heat pipes are starting to
function, and the total thermal resistance shows a decreasing trend.
The heat pipes are unable to sustain excessive higher heat capacity at
heating powers of above 150 W for 15 15 mm2 heat source, causing
the evaporation section to produce more amounts of vapor than that
of condensed liquid, resulting in the heat pipes losing performance,
thereby increasing total thermal resistance. Unlike the 30 30 mm2
heat source, Rt presents a mild parabolic curve because of the function
of the embedded heat pipes and higher heat ux. Thus, the Rt values
of 15 15 mm2 heat source are all higher than those of 30 30 mm2
heat source. Raising fan speed improves thermal performance of
present thermal module. Consequently, the fan should be kept above a
gradated speed to acquire better heat diffusion benets. The total thermal resistances of single fan operation are higher than those of dual fan
operation due to heat convection mechanism under identical heat
source areas and the fan speeds. Normally, twin-fan convection can
enhance the thermal performances than the usage of single fan at low
fan speeds. In present experimental results, the values of total thermal
resistance Rt are nearly identical levels between a single fan speed of
3000 RPM and twin fans of 2000 RPM. In addition to thermal interface
resistances Rif and total thermal resistances Rt, the rest of thermal resistances including base plate resistances Rb, n-base convective thermal
resistances Rhs, base to heat pipe resistances Rbh,n, heat pipe resistances
Rh,n, and n-pipe convective resistances Rf,n are able to obtain based on
the individual bypass heat ow. Table 3 shows the bypass heat ow
ratios of L- and U-shaped heat pipeheat sink thermal modules to the
total heat capacity according to superposition method and the same
experimental scheme as Ref. [11]. The single U-shaped heat pipe has a
higher heat dissipation ratio than that of single L-shaped heat pipe at
the identical condition.
From Table 3, three groups of dual L-shaped heat pipes are marked
as I of (Q1 + Q6) / Q, II of (Q2 + Q5) / Q and III of (Q3 + Q4) / Q. The
sum of the ratio of the six embedded L-type heat pipes is between 77%
and 87%. For large volume ow rates (two fan 2000 RPM and one fan
3000 RPM), each bypass heat ow ratio (groups IIII) is almost 30%.
Group I is dissipated heat capacity by the two L-shaped heat pipes closest to the air inlet with higher heat transfer coefcient at single fan condition, and group III is the furthest away from the air inlet as shown in
Fig. 3(b), which is why the ratio of group I heat-transfer path of 47%
and 38% is larger than that of the group III heat-transfer path of 10%
and 20% for single fan of 1000 RPM. At 1000 RPM, the impact of n impedance on the ow eld decreases the volume ow rate from the air
inlet to the heat pipe, while the volume ow rate to the six L-shaped
heat pipes increases from 77% (30 30 mm2 heat source) to 87% (15
15 mm2 heat source). For embedded two U-shaped heat pipe thermal
module, the bypass heat ow ratio of the single U-shaped heat pipe is
found to increase as the heat source area decreases. The reason is
when the heat source area is reduced; the ratio of the total area of the
contact area of the heat pipe to the heat source increases, the bypass
heat ow ratio will raise as the heat source area is decreased. The experimental results are good agreement with theoretical suppose. Table 4
displays the heat pipe thermal resistance regression curves of the Lshaped thermal module which applied the quadratic polynomial equations [27]. The heat pipe is a two-phase heat transfer element, and its
thermal resistance value will change with the heat ow resulting from
the multiphase behavior of the working uid inside heat pipe. Six Lshaped heat pipes have the lowest thermal resistances of the order of
1.04 C/W, 2.07 C/W, 1.92 C/W, 2.76 C/W 2.19 C/W, 1.7 C/W between 34 W and 40 W. And from Ref. [27] the U-shaped heat pipe has
the highest thermal resistance below 10 W, following a gradual decline as heat ow increases with the rate of decline slowing to
25 W, and then gradually increasing after 35 W. Therefore, the

Table 4
Theoretical thermal resistance equations of L-shaped heat pipes [27].
Rh1 (W/mK)
Rh2 (W/mK)
Rh3 (W/mK)
Rh4 (W/mK)
Rh5 (W/mK)
Rh6 (W/mK)

3.66 101
(Q1 = W)
6.31 101
(Q2 = W)
5.42 101
(Q3 = W)
8.46 101
(Q4 = W)
6.93 101
(Q5 = W)
6.42 101
(Q6 = W)

1.31 102 Q1 + 1.92 104 (Q1)2


2.46 102 Q2 + 3.15 104 (Q2)2
2.61 102 Q3 + 3.7 104 (Q3)2
3.3 102 Q4 + 4.26 104 (Q4)2
2.58 102 Q5 + 3.33 104 (Q5)2
1.76 102 Q6 + 2.2 104 (Q6)2

(a) U-shaped thermal module

(b) L-shaped thermal module


Fig. 5. Relationships of the base plate resistances toward the heating power (a) U-shaped
heat pipe thermal module (b) L-shaped heat pipe thermal module.

R.-T. Wang et al. / International Communications in Heat and Mass Transfer 57 (2014) 140149

theoretical thermal resistance of the U-shaped heat pipe was Rh =


1.59 10 1 0.3 102 Q1 + 0.19 104 (Q1)2, where the lowest
thermal resistance is 0.04 C/W at the heat ow Q1 of 78.95 W. And its
highest effective thermal conductivity is 1852 W/(mK).
Fig. 5 indicates the relationships of the base plate resistances Rb toward the heating power Qb. These thermal resistance curves on the
paths Q ib closed to horizontal. Thus, Rb obtained in this experiment
may be considered constant as heating power increases, since the components that transfer heat through this path of heating power Qb, transferred from the base plate without function of the heat pipes
transferring heat capacity to the ns, are all solid. This phenomenon is
anticipated and can be construed as the thermal physical properties
(eg. density, thermal conductivity) of these components which are the
same when there is not much change in temperature. For U-shaped
heat pipes, Rbvalues are approximately between 0.15 C/W and

(a) U-shaped thermal module

(b) L-shaped thermal module


Fig. 6. Relationships of the n-base convective thermal resistances toward the heating
power (a) U-shaped heat pipe thermal module (b) L-shaped heat pipe thermal module.

147

0.24 C/W at 30 30 mm2 heat source, and are approximately between


0.51 C/W and 1.10 C/W at 15 15 mm2 heat source, respectively. For
L-shaped heat pipes, Rb values are approximately between 0.49 C/W
and 0.52 C/W at 30 30 mm2 heat source, and are approximately
between 1.18 C/W and 1.84 C/W at 15 15 mm2 heat source,
respectively. This reason is when the heat source area is reduced, the
spreading of the thermal resistances will be increased and the base
plate resistances will be changed to larger values. Fig. 6 shows the
relationships of the n-base convective thermal resistances Rhs toward
the heating power Qb. These Rhsvalues reduced as volume ow rates increased. The base to heat pipe resistances Rbh will increase when heat
source area decreases as shown in Fig. 7. From the experimental
results, the values of Rbh will increase twice when the heat source size
is reduced from 30 30 mm2to 15 15 mm2. Fig. 8 exhibits the
n-pipe convective resistances Rf with input power. The n-pipe

(a) U-shaped thermal module

(b) L-shaped thermal module


Fig. 7. Relationships of the base to heat pipe resistances toward the heating power.

148

R.-T. Wang et al. / International Communications in Heat and Mass Transfer 57 (2014) 140149

4. Conclusions

(a) U-shaped thermal module

The thermal performances of heat pipe thermal modules have


been analyzed through careful experiments and versatile methods for
electronic cooling in the present study. Particular attention is paid to
the inuences of geometric variations of heat pipes upon the performance of the cooling system. The effects of fan speeds and heat source
areas on the performance have also been examined. For embedded six
L-shaped heat pipe thermal modules, the average heat capacity ratio
of a single heat pipe is 14%. The lowest total thermal resistance of the
L-shaped heat pipeheat sink module is 0.17 C/W under dual fans at
3000 RPM for a 30 mm 30 mm heat source. For embedded two
U-shaped heat pipes module, a single heat pipe can dissipate an average
heat capacity ratio of 33%. The lowest total thermal resistance of the
U-shaped heat pipeheat sink module is 0.25 C/W under dual fans at
3000 RPM for a 30 mm 30 mm heat source. Intentionally, the fan
should be kept above a certain speed over 1000 RPM to obtain better
cooling efciency. The heat ow ratio of total heat ow through the
heat pipe is also different under variable heat source areas. The smaller
they had large variation of the heating power resulting from higher
heat ux causing the heat pipe to reach its operating limits more easily.
Finally, the present study has established a useful method of thermal
performance solution of the heat transfer model which enables us to
apply it in thermal management and control.
Acknowledgments
The author gratefully acknowledges the nancial support from
NSC101-2221-E-019-042- for the present study. Finally, the author
would like to thank all colleagues and students who contributed to
this study.
References

(b) L-shaped thermal module


Fig. 8. Relationships between the n-pipe convective resistances with input power.

convective resistance Rfdoes not change much as the heating power


increases, so they can be considered constant in this experiment. The
volume ow rate and heat sink design affect the heat transfer coefcient
and n efciency. It should be recognized that when ns are in the 90%
efciency range, an increase of several percent efciency could require a
considerable lengthening of the n. The heat transfer coefcients of dual
fans and high fan speed are larger than those of one fan or low fan speed.
The convection heat transfer coefcient is most often calculated using
empirical formulations based on convection correlations. However,
turbulent ow will result in a larger pressure drop such that with a
given fan, the uid ow rate will be reduced. Forced convection heat
transfer from a given surface is a function of the local ow velocity.
Therefore, wind speed caused by the fan should not be affected by the
size of the heat source area, and in addition to the single fan speed,
2000 RPM fans have higher n-pipe thermal resistances.

[1] J.-C. Wang, S.-L. Chen, Heat transfer engineering applications, Air Cooling Module
Applications to Consumer Electronic Products, I,nTech Open Access Publisher,
2011, ISBN 978-953-307-361-3. 339366 (Chapter 14).
[2] Y.-W. Chang, et al., Heat pipe for cooling of electronic equipment, Energy Convers.
Manag. 49 (2008) 33983404, http://dx.doi.org/10.1016/j.enconman.2008.05.002.
[3] T.-E. Tsai, et al., Dynamic test method for determining the thermal performances of
heat pipes, Int. J. Heat Mass Transf. 53 (2010) 45674578.
[4] J.-C. Wang, et al., Experimental investigations of thermal resistance of a heat sink
with horizontal embedded heat pipes, Int. Commun. Heat Mass Transfer 34
(2007) 958970, http://dx.doi.org/10.1016/j.icheatmasstransfer.2007.03.015.
[5] R.-T. Wang, et al., Experimental analysis for thermal performance of a vapor
chamber applied to high-performance servers, J. Mar. Sci. Technol. (Taiwan) 19
(2011) 353360.
[6] S.W. Chi, Heat Pipe Theory and Practice: A Sourcebook, 1st ed Hemisphere Pub.
Corp, 1976. 130 (Chapter 1).
[7] J.-C. Wang, Investigations on non-condensation gas of a heat pipe, Engineering 3
(2011) 376383.
[8] J.-C. Wang, Computer aided design: technology, types and practical applications,
Computer Aided-Thermal Module Design, N
, ova Publisher, 2012, ISBN 978-162257-346-2. 134 (Pub. Date: 2012-December, Chapter 1).
[9] P. Dunn, D.A. Reay, Heat Pipes, 1st ed Pergamon Press, Oxford; New York, 1976.
1887 (Chapter 2).
[10] R.-T. Wang, J.-C. Wang, Analyzing the pressure-difference phenomenon between
the condensing and boiling sections in a heat pipe cooling system, Int. Commun.
Heat Mass Transfer 39 (2012) 390398.
[11] J.-C. Wang, Superposition method to investigate the thermal performance of heat
sink with embedded heat pipes, Int. Commun. Heat Mass Transfer 36 (2009)
686692, http://dx.doi.org/10.1016/j.icheatmasstransfer.2009.04.008.
[12] J.-C. Wang, et al., Program for rapid computation of the thermal performance of a
heat sink with embedded heat pipes, J. Chin. Soc. Mech. Eng. 31 (2010) 2128.
[13] R.-T. Wang, A tting, simple and versatile window program (HSHPTM) design using
lumped parameters and one-dimensional thermal resistance models, Heat Mass
Transf. 49 (2013) 291297.
[14] Mohamed H.A. Elnaggar, et al., Characterization of working uid in vertically
mounted nned U-shape twin heat pipe for electronic cooling, Energy Convers.
Manag. 62 (2012) 3139.
[15] Mohamed H.A. Elnaggar, et al., Experimental analysis and FEM simulation of nned
U-shape multi heat pipe for desktop PC cooling, Energy Convers. Manag. 52 (2011)
29372944, http://dx.doi.org/10.1016/j.enconman.2011.03.001.
[16] M.K. Russel, et al., The effect of orientation on U-shaped grooved and sintered wick
heat pipes, Appl. Therm. Eng. 31 (2011) 6976.

R.-T. Wang et al. / International Communications in Heat and Mass Transfer 57 (2014) 140149
[17] T.-S. Liang, Y.-M. Hung, Experimental investigation on the thermal performance and
optimization of heat sink with U-shape heat pipes, Energy Convers. Manag. 51
(2010) 21092116.
[18] J.-C. Wang, L-type heat pipes application in electronic cooling system, Int. J. Therm.
Sci. 50 (2011) 97105.
[19] J.-C. Wang, Thermal module design and analysis of a 230 Watt LED illumination lamp
under three incline angles, Microelectronics Journal 45 (4) (2014) 416423 (April).
[20] J.-C. Wang, Thermoelectric transformation and illuminative performance analysis of
a novel LED-MGVC device, Int. Commun. Heat Mass Transfer 48 (2013) 8085
(November).
[21] J.-C. Wang, 3-D numerical and experimental models for at and embedded heat
pipes applied in high-end VGA card cooling system, Int. Commun. Heat Mass
Transfer 39 (2012) 13601366.
[22] J.-C. Wang, Development of vapour chamber-based VGA thermal module, Int. J.
Numer. Methods Heat Fluid Flow 20 (2010) 416428, http://dx.doi.org/10.1108/
09615531011035811.

149

[23] X.-Y. Lu, et al., Thermal analysis of high power LED package with heat pipe heat sink,
Microelectron. J. 42 (2011) 12571262.
[24] J.-C. Wang, Thermal investigations on LEDs vapor chamber-based plates, Int.
Commun. Heat Mass Transfer 38 (2011) 12061212.
[25] J.-C. Wang, Analyzing thermal module developments and trends in high-power LED,
Int. J. Photoenergy 2014 (2014), http://dx.doi.org/10.1155/2014/120452 (Article ID
120452, May, 11 pages).
[26] J.-C. Wang, et al., Development of 30 Watt high-power LEDs vapor chamber-based
plate, Int. J. Heat Mass Transf. 53 (2010) 39004001.
[27] J.-C. Wang, U- and L-shaped heat pipes heat sinks for cooling electronic components
employed a least square smoothing method, Microelectron. Reliab. 54 (6/7) (2014)
13441354 (June).