Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

International Journal of Heat and Mass Transfer 118 (2018) 1276–1283

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer


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

High power density air-cooled microchannel heat exchanger


Beomjin Kwon 1, Nicholas I. Maniscalco 1, Anthony M. Jacobi, William P. King ⇑
Department of Mechanical Science and Engineering, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA

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

Article history: We present single-phase heat transfer in a compact cross-flow microchannel heat exchanger, with air
Received 8 August 2017 flowing through the heat exchanger to remove heat from a closed-loop flow of refrigerant R245fa. The
Received in revised form 13 November 2017 1 cm3 heat exchanger was monolithically fabricated from a block of copper alloy using micro-
Accepted 13 November 2017
electrical-discharge machining. Air carrying channels of diameter 520 lm were oriented in cross-flow
to the refrigerant-carrying channels of size 2.0  0.5 mm2. High-speed air flowed with Reynolds number
between 1.2  104 and 2.05  104, which corresponded to air speeds between 20 and 100 m/s, while
Keywords:
refrigerant flowed at Reynolds number between 1000 and 2300. Using an equivalent fin model and finite
Compact heat exchanger
Cross-flow
element simulations, we predicted the heat exchanger performance and used the simulations to interpret
Microchannel the measured behavior. Temperature, pressure, and flow rates were measured over a variety of operating
Micro-electro-discharge machining conditions to determine heat transfer rate, j-factor, and friction factor. We observed a maximum power
Single-phase density of 60 W/cm3 when the air inlet temperature was 27 °C and the refrigerant inlet temperature was
80 °C. The high speed of air flow caused large friction on the air side, resulting in goodness factor j/f near
0.5. This work demonstrates that high power density can be achieved in miniature heat exchangers, and
that micromachined metal devices can enable this performance. The results could be broadly applied to
other types of microchannel devices.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction explored microchannels etched into silicon, with feature sizes of


few hundreds of micrometers [2,5]. The patterned silicon wafers
High-flux thermal management is a primary design concern for could also be anodically bonded to assemble perforated-plate heat
advanced computing, communications, and vehicle systems as exchangers [6]. Recent advancements in micromachining enabled
their heat fluxes are expected to exceed 1000 W/cm2 [1]. Although the use of metal for the microchannel heat exchangers. Microscale
dielectric liquid cooling systems have long been considered for flow passages were created by micromachining thin metal foils
such high-flux systems, there remains a need for reduced volume (100 mm thick copper or stainless steel) [7,8] or stainless steel
and weight. As a compact and light-weight cooling solution, plates [9] which were then welded or soldered to form cross-
microchannel coolant flows have been attractive, because decreas- flow heat exchangers. Microfabircation techniques for polymeric
ing the hydraulic diameter can lead to significant increases in the structures such as LIGA [10] or stereolithography [11] could also
convective heat transfer coefficient and area of the cooling surface create microchannel heat exchangers, although the low thermal
[2]. Furthermore, small-volume heat exchangers help to reduce the conductivity of polymer was considered a drawback.
amount of working fluid, and some fluids can cause detrimental Metal microchannel heat exchangers are of interest due to their
effects on environment [3]. Compact heat exchangers are also high thermal conductivity, strength, and fracture toughness.
desired for mobile applications, where reduced size and weight Microfabrication in metals has recently become attractive, due to
can improve overall performance efficiencies of vehicles. improvements in accuracy and resolution that can be achieved in
Over the last few decades, various methods have been metals microfabrication, along with lowered costs [12,13]. Unlike
employed to fabricate microchannel heat exchangers [4]. Moti- conventional silicon microfabrication, which has significant fixed
vated by integration with electronic systems and the opportunity costs, the economics of micro-manufacturing in metal are attrac-
to leverage microelectronics fabrication, many early studies tive, especially for lower volume applications.
Micro-electro-discharge machining (mEDM) is a meso-scale
manufacturing process that selectively removes an electrically
⇑ Corresponding author. conductive material by plasma discharge [12]. Combined with an
E-mail address: wpk@illinois.edu (W.P. King). electrode cutting tool, mEDM provides a high accuracy, relatively
1
B. Kwon and N.I. Maniscalco contributed equally to this work.

https://doi.org/10.1016/j.ijheatmasstransfer.2017.11.068
0017-9310/Ó 2017 Elsevier Ltd. All rights reserved.
B. Kwon et al. / International Journal of Heat and Mass Transfer 118 (2018) 1276–1283 1277

high throughput, low-cost setup, and large design freedom. mEDM


can fabricate features smaller than 10 mm in stainless steel and
titanium [14]. When fabricating multiple rows of perforated chan-
nels, mEDM does not require welding of separate plates, which [7–
9,15] may be expensive and result in partially clogged channels
due to imperfect welding [15]. In this study, we demonstrate the
use of mEDM to fabricate high density and high aspect ratio
microchannels into a 1 cm3 cubic copper alloy.
Here, we design and experimentally assess the performance of a
cross-flow microchannel heat exchanger fabricated by mEDM. A
heat transfer model was developed to recommend the diameter,
length, and the number of fluid channels that minimize the pres-
sure gradient while exploiting the benefits of high heat transfer.
The fabricated heat exchanger was experimentally characterized
to demonstrate high-density power and to determine its ther-
mofluidic properties.

2. Modeling heat exchanger performance

Fig. 1(a) shows a schematic of the cross-flow heat exchanger,


which utilizes two unmixed working fluids in a cross flow config-
Fig. 1. (a) Schematic of the cross-flow heat exchanger and key design parameters.
uration. The heat exchanger has overall dimensions 1 cm  1 cm 
Color represents the temperature distribution predicted by a finite element model
Lch,air, where Lch is the channel length along the flow direction, and when Tair,i = 27 °C, Reair = 1000, Tref,i = 60 °C and Reref = 1000. (b) Side view of the
the subscript air indicates the air channels. Air flows through the heat exchanger showing modeled thermal resistances. The air side thermal
circular microchannels and a dielectric refrigerant, R245fa, flows resistance is modeled as a fin with efficiency gf. (For interpretation of the
through the rectangular minichannels. The air side has an array references to colour in this figure legend, the reader is referred to the web version of
this article.)
of microchannels with a diameter dair, and a center-to-center spac-
ing s. The refrigerant side has parallel minichannels having a cross
section of 2  0.5 mm2, with each row containing four channels. T ref ;o ¼ ð2  KÞT ref ;i =ð2 þ KÞ þ 2KT air;i =ð2 þ KÞ ð3Þ
Color in Fig. 1(a) shows the temperature distribution predicted
from finite element simulations that are discussed later. where thermal effectiveness e = 1  exp(1/Rth,totalCair), K = CairC/
Neglecting the conductive resistance of the thin refrigerant Cref, and C is heat capacity rate of the fluid. The heat transfer
channel wall, the thermal resistance between the refrigerant and between the two fluids is calculated as Q = Cair (Tair,o  Tair,i) = Cref
the air can be modeled as two resistances in series: a convection (Tref,i  Tref,o).
thermal resistance on the hot refrigerant side and a fin thermal The pressure drop (DP) and the total pumping power (Ppump) of
resistance on the cold air side (see Fig. 1b) [16]. The fin resistance the channels was modeled using the Darcy-Weisbach equation
accounts for both conduction in the solid heat exchanger and con- [18]. This equation predicts DP due to core friction as
vection between the solid heat exchanger and the air. In the equiv-
alent fin model, heat flows from the refrigerant channel surface (fin DPfric ¼ 0:5f qu2 Lch =dh ð4Þ
base) to air via the solid heat exchanger (fin). The total thermal
where f is the friction factor, q is the fluid density, u is the axial
resistance (Rth, total) is
velocity, and dh is the hydraulic diameter. The axial velocity is u =
! mRe/dh where m is the fluid kinematic viscosity. With DPfric, for the
1 1 1 air and refrigerant flows, the total pumping power, Ppump, is obtain-
Rth;total ¼ þ ð1Þ
Nf href Aref gf hair Af able as Ppump = (NAcuDPfric)air + (NAcuDPfric)ref.
Several thermofluid correlations are necessary to predict the
where N is the unit number, h is the convective heat transfer coef- friction factors and convective heat transfer coefficients. The model
ficient, A is the area for convective heat transfer, and gf is the fin assumes that the air flow Re is in the range 1000  Reair  2.0 
efficiency. The subscripts ref and f denote the refrigerant channels 104. We estimate the friction factor based on the Petukhov correla-
and fin, respectively, and in this model NfAf is total air-side area. tion, f = (0.790lnRe  1.64)2, and Nu based on the Dittus-Boelter
A fin parameter (m) is helpful for estimating gf, and is defined m correlation, Nu = 0.023Re0.8Pr0.4 for fully developed turbulent flow
= [hairpf/kfAc,f]0.5, where p is the perimeter, k is the thermal conduc- [19]. For the refrigerant flow, the model assumes Re near 1000.
tivity, and Ac is the fin cross-section area. Assuming an adiabatic fin We predict the friction factor and Nu using the Phillips correlations
tip, the fin efficiency can be calculated gf = tanh(mLf)/mLf, where Lf is for laminar developing flow in the entrance region of a rectangular
the fin length, defined as the distance between the refrigerant chan- duct [5].
nel wall and the farthest row of the air channels. We also used a three-dimensional FEM (COMSOL Multiphysics)
Calculating the heat transfer requires the air temperature at to simulate the thermofluidic characteristics and performance of
inlet (Tair,i) and outlet (Tair,o) and refrigerant temperature at inlet the heat exchanger. FEM calculates an energy balance equation
(Tref,i) and outlet (Tref,o). For a simplified model, we assumed that for an internal incompressible fluid flow
the refrigerant temperature linearly varies along the channel. Our
three-dimensional finite element model (FEM), described below, qAc;ch cp u  rT ¼ r  Ac;ch krT þ hpðT w  TÞ ð5Þ
predicts that the temperature profile along the refrigerant channel
where cp is the heat capacity at constant pressure and the subscripts
is nearly linear, which supports this assumption. Then, the outlet
ch and w denote the channel and channel wall, respectively. This
temperatures are estimated as below [17].
equation states that the volumetric energy change is equal to the
sum of the conductive heat transfer within the fluid and the convec-
T air;o ¼ ½K þ 2ð1  eÞT air;i =ð2 þ KÞ þ 2eT ref ;i =ð2 þ KÞ ð2Þ
tive heat transfer through the channel wall. In addition to the
1278 B. Kwon et al. / International Journal of Heat and Mass Transfer 118 (2018) 1276–1283

energy balance equation, the FEM calculates a continuity equation


and momentum ‘including the pressure drop due to viscous shear.
The boundary conditions were Tair,i = 27 °C, Tref,i = 60 °C, and Reref
= 1000. Reair was varied between 1000 and 6000. For simplification,
the Reynolds number evaluated at the channel inlet represents Re
for each flow condition, since Re may not be constant along the
channel. The FEM used the same thermofluid correlations as the
equivalent fin model (EFM).

3. Heat exchanger design

We predicted the heat exchanger performance in order to


select design parameters such as diameter (dair), length (Lch,air)
of the air channels, and the number of refrigerant channel rows
(Nref). The model assumed the spacing between air channels (s)
as 1.25dair to ensure a reasonable fabrication tolerance. To validate
EFM, we compared its prediction to the FEM calculation. Fig. 2
shows the predicted outlet temperature (To), pressure drop (DP),
and total thermal resistance (Rth, total) as a function of Reair. For this
calculation, we assumed dair = 500 mm, Lch,air = 10 mm, and Nref = 2.
As the air flow rate increases, Tair,o and Tref,o decrease slightly due
to the decrease of Rth, total. Overall, the EFM agrees well with FEM,
although it exhibits a deviation of 20–30% in Rth, total. The EFM
may underestimate Rth,total because it models the fin as straight
and rectangular, rather than with the actual geometry that fea-
tures a longer conduction path. Fig. 1(a) shows the temperature Fig. 2. Comparison between the equivalent fin model (line) and finite element
distribution is not one-dimensional, but instead there is some model (symbols) when the air channel diameter is 500 mm, air channel length is 10
temperature variation across the width of the heat exchanger. mm and the number of refrigerant channel rows is two. Temperature at the channel
Nevertheless, the EFM does reasonably well in modeling the heat exits, pressure drop, and total thermal resistance as a function of Reair.

exchanger performance and can be used to aid in choosing design


parameters.
Fig. 3 shows the heat exchanger performance as a function of
the air channel diameter (dair) calculated by EFM, for several values
of Reair. This calculation assumes Lch,air = 10 mm and Nref = 2. As dair
decreases, both the convection coefficient and surface area density
increase, and consequently the heat transfer per volume (Q 000 )
increases. However, when dair decreases below 200 mm, the pres-
sure drop substantially increases for the corresponding Reair,
resulting in a large pumping power (Ppump) and poor heat transfer
efficiency (Q/Ppump).
Fig. 4 shows the effect of the air channel length (Lch,air) on the
heat exchanger performance when dair = 500 mm and Nref = 2. As
Lch,air increases, Q 000 decreases and pressure drop increases, leading
to the decrease of Q/Ppump. In a cross-flow heat exchanger, the
longer channel length results in smaller the mean temperature dif-
ference between the hot and cold fluids which results in reduced
Q 000 .
Fig. 5 shows the heat exchanger performance with varied num-
ber of refrigerant channels when dair = 500 mm and Lch,air = 10 mm.
When Nref decreases, the heat exchanger possess more air channels
but fewer refrigerant channels. Thus, the thermal resistance ratio
of the air side and the refrigerant side is affected by Nref. This pre-
diction shows that the thermal resistance ratio is close to unity
when Nref = 2, and Rth,total is the minimum when Nref = 3. The max-
imum Q 000 occurs when Nref is either 2 or 3 depending on Reair.
The modeling results of Figs. 3–5 provide overall guidance on
the heat exchanger design. To achieve a large Q 000 and small DP,
the air channels must be short and have a reasonably small diam-
eter. Furthermore, the thermal resistance match between the air
and refrigerant channels is important to determine the number
of fluid channels. We therefore selected dair = 500 mm, Lch,air = 10
Fig. 3. Calculated performance of the heat exchanger as a function of the air
mm, and Nref = 2. We chose to operate the air side near Reair = channel diameter when the air channel length is 10 mm and the number of
10000, which supports high heat transfer while allowing operation refrigerant channel rows is two. The plots show power density (Q 000 ), heat rate per
within the air pressure available in our test facility (<1 MPa). pumping power, pumping power, pressure drop, and total thermal resistance.
B. Kwon et al. / International Journal of Heat and Mass Transfer 118 (2018) 1276–1283 1279

4. Experiment

We fabricated the heat exchanger using mEDM. The heat


exchanger material is copper alloy C14500, which is 0.4% tellurium
by weight. As compared to pure copper, C14500 is easy to machine
and still exhibits a high thermal conductivity of 355 W/m K [20].
The fabrication employed a carbon electrode tool having the
inverse shape of the desired channel geometry. Fig. 6(a) and (b)
shows the fabricated heat exchanger which contains 150 air chan-
nels and 8 refrigerant channels. We measured the key dimensions
using a calibrated microscope. The air channel diameter is 520 mm
± 57 mm, center-to-center spacing is 640 mm ± 16 mm, and channel
surface roughness (e) is about 2 mm. Thus, the relative roughness
(e/dair) corresponds to 0.004. For the refrigerant channel, the rel-
ative roughness is 0.0025. To integrate the heat exchanger into
the test facility, a manifold was designed as shown in Fig. 6(c).
The manifold consists of steel cage and interface structures made
of garolite. At each part that connects the heat exchanger and fluid
tubes, a port exists to install a thermocouple.
Fig. 6(d) shows the setup for characterizing the heat exchanger
which includes an air supply and a refrigerant loop. In the air sup-
ply, compressed air at 27 °C enters a regulator that controls the
upstream pressure up to 862 kPa. Then, after flowing through a
Coriolis-effect mass flow meter, with an uncertainty of ±1%, and
the heat exchanger, air is exhausted to the laboratory. The refriger-
ant loop supplies refrigerant to the heat exchanger at the desired
flow rate and temperature. A pump circulates the refrigerant and
its mass flow rate is controlled by changing the pump motor speed.
Fig. 4. Calculated performance matrices of the heat exchanger as a function of the The mass flow rate is measured using a Coriolis-effect mass flow
air channel length when the air channel diameter is 500 mm and the number of
refrigerant channel rows is two. Power density, heat rate per pumping power, and
meter with an uncertainty of ±0.1%. Before entering the heat
pressure drop. exchanger, the refrigerant is heated to either 60 °C or 80 °C by an
evaporator, and after exiting the heat exchanger, the refrigerant
is subcooled using a condenser. T-type thermocouples are used
to measure the temperature and pressure transducers provide
the pressures both upstream and downstream of the test section.
The thermocouples are calibrated using an ice-point bath to
improve the relative uncertainty to ±0.1 °C. The nominal uncer-
tainty of the pressure transducers is ±0.25% of full scale, which is
equivalent to the uncertainty of ±1.7 kPa for air flow measurement
and ±17 Pa for refrigerant flow measurement.
The heat transfer coefficients, hair and href, were calculated using
the Wilson plot method [21]. When evaluating hair, we changed the
refrigerant flow rate while holding the air flow rate fixed and mea-
suring the heat loads for both fluids. Similarly, to estimate href, the
air flow rate was changed and the refrigerant flow rate was con-
stant. The heat transfer coefficients were calculated as

air Pr air ðlair =lair;w Þ


1=3 0:14
hair ¼ c1 ðkair =dair ÞRe0:8 ð6Þ

href ¼ c2 ðkref =dref ÞðReref Prref dref =Lch;ref Þ 3 ðlref =lref ;w Þ0:14
c
ð7Þ

where c1, c2, c3 are unknown coefficients, Pr is the Prandtl number,


and l is dynamic viscosity. dref is the hydraulic diameter. To solve
the unknown coefficients, we combine Eqs. (1), (6), and (7) as
follows.

1 ðhair gf Af =c1 Þ 1
Rth;total ðhair gf Af =c1 Þ ¼ þ ð8Þ
c2 ðhref Aref =c2 Þ c1
As Eq. (8) is in a linear form of Y = X/c2 + 1/c1, the slope and y-axis
intercept of the Wilson plot line correspond to 1/c2 and 1/c1, respec-
tively. An iterative regression of Eq. (8) determines all the
unknowns if Rth,total is available from the experimental data by
Rth,total = FDTlm/Q, where F is a correction factor for cross-flow
Fig. 5. Calculated performance matrices of the heat exchanger as a function of the
number of refrigerant channel rows when the air channel diameter is 500 mm and
heat exchanger with both flows unmixed [22], and DTlm is the
the air channel length is 10 mm. Power density, heat rate per pumping power, and logarithmic mean temperature difference. Q is obtainable from
total thermal resistance. Q = Cair (Tair,o  Tair,i) = Cref (Tref,i  Tref,o). The above procedure
1280 B. Kwon et al. / International Journal of Heat and Mass Transfer 118 (2018) 1276–1283

Fig. 6. (a, b) Images of heat exchanger, showing front and side views, and detail of
the air channels. (c) Assembly of the heat exchanger in the manifold. (d) Schematic
of a test setup for the heat exchanger.

provided c1 = 0.028, c2 = 1.36, and c3 = 0.5. The Wilson plot method


is accurate when employed for fully developed flow. To account for
the developing flow, a correction factor for a laminar flow in a
circular tube [24] can be applied, and this correction requires
knowledge of the channel wall temperature. For thermally develop-
ing laminar flow (100 < Re < 1400), the correction factor modifies
the heat transfer coefficient by 4–20%.
In order to calculate friction factor, we considered pressure drop
for a pipe flow. The total pressure drop (DP) is the sum of the pres-
sure drops by an entrance effect (DPent), an exit effect (DPexi), flow
acceleration (DPacc), and friction (DPfric). Based on the channel
geometry, flow rate, and fluid density, it is possible to estimate
DPent, DPexi, and DPacc [16,23]. Thus, DPfric = DP  DPent  DPexi  Fig. 7. Fluid flow characteristics of the (a) air channels and (b) refrigerant channels
DPacc. Then, the friction factor is determined by [23] as a function of Re when Tair,i = 27 °C. Pressure drop (top) and Fanning friction factor
(bottom).
f ¼ DPfric ðq=G2 ÞðAc;ch =AÞ ð9Þ
where G is the mass flux of the channel.
density variation is much more significant in the air channels. DPent
5. Results and discussion and DPexi are comparatively small in air channels; however, DPent is
comparable to DPfric in the refrigerant flow due to a small ratio of
We evaluated the heat exchanger characteristics and perfor- cross-section area to the frontal area of the refrigerant channels
mance when Reair varied between 12000 and 20500 and Reref var- (0.09). The use of flow distributer would help to reduce DPent in
ied between 1000 and 2300. Fig. 7 shows the pressure drop (DP) the refrigerant flow [25]. The measured f in the air channels is
and friction factor (f) as a function of Re. Fig. 7 also shows separate about an order of magnitude larger than those estimated by Petu-
components of the pressure drop, calculated from models. The khov correlation for fully developed turbulent flow [19] as well as
pressure drop due to core friction (DPfric) is the primary factor in by the Colebrook correlation [19]. However, the measured f com-
both air and refrigerant channels. DPacc is not negligible in air pares well with the f calculated by the Fanno model for an isother-
channels in contrast to refrigerant channels, indicating that the mal compressible flow [26], indicating that the air density change
B. Kwon et al. / International Journal of Heat and Mass Transfer 118 (2018) 1276–1283 1281

Fig. 9. Goodness factor, j/f, of the (a) air channels (Tair,i = 27 °C) and (b) refrigerant
channels (Tref,i = 60 °C) as a function of inlet Reynolds number.

Table 1
Thermofluidic correlations used by the equivalent fin model.

Air flow Refrigerant flow


Re 12,000–20,500 1000–2300
Flow type Turbulent Laminar
Friction coefficient Pertukhov correlation –
(channel design)
Friction coefficient Fanno model
(comparison to
experimental data)
Convection heat transfer Dittus-Boelter Phillips correlation
coefficient (channel correlation
design)
Convection heat transfer Experimentally Experimentally
coefficient reduced values reduced values
(comparison to (Modified Wilson plot (Modified Wilson plot
experimental data) method) method)
Fig. 8. Heat transfer characteristics of the (a) air channels and (b) refrigerant
channels as a function of Reynolds number when Tair,i = 27 °C. Convection heat
transfer coefficient (top) and Colburn j-factor (bottom).

and Gnielinski correlations. The original forms of these correlations


along the channel length is not negligible. In the refrigerant chan- are for fully developed turbulent flow in circular tubes. Since the
nel, the measured f agrees with the Curr correlation for developing air flow is a developing flow, correction factors were applied
laminar flow in rectangular channel (aspect ratio = 0.25) [27] and [30]. Approximately, the thermal entry length in the air channel
Colebrook correlation. The roughness of the channel wall can cause is 10dair = 5 mm; thus, about a half of the channel length is sub-
an increase in f or an early transition to turbulence, and this was ject a convection coefficient larger than that predicted by correla-
not observed in the measurements. The roughness effect mostly tions for fully developed flow. For the refrigerant flow, measured
affects incompressible turbulent flows [28,29] even when the rela- href and j-factor are close to or greater than the Phillips [5] and
tive roughness is smaller than 0.5%. Shah-London [30] correlations for developing laminar flow in rect-
Fig. 8 shows the convection heat transfer coefficient (h) and angular channels. The discrepancy between the correlations is due
j-factor as a function of Re. For the air flow, measured hair and to the different local Nusselt number at the entrance region. The
j-factor compare well with the predictions of the Dittus-Boelter entrance Nusselt number is 26.7 by the Phillips correlation while
1282 B. Kwon et al. / International Journal of Heat and Mass Transfer 118 (2018) 1276–1283

Fig. 10 shows the power density (Q 000 ) as a function of either


Reynolds number or pumping power (Ppump). Here, EFM employs
the experimentally obtained thermofluidic properties such as f
and h. Table 1 lists the empirical correlations used by the EFM.
Over the range of the Reynolds numbers employed, Q 000 is 20–40
W/cm3 when Tref,i = 60 °C and is 40–60 W/cm3 when Tref,i = 80 °C.
The enhancement of Q 000 is more pronounced by the increase of
the refrigerant flow rate than the adjustments of the air flow rate,
indicating that the decrease of the refrigerant-side Rth more helps
to reduce the mismatch of Rth between the working fluids. As com-
pared to our predictions of the heat exchanger performance in
Figs. 3–5, measured Q 000 is larger by more than 20%. The enhance-
ment of Q 000 is primarily due to the large value of hair benefiting
from the thermally developing region existing over the large por-
tion of the air channels. To leverage the large convection efficiency
of the thermally developing flow, it is, therefore, interesting to
design the channel short. Fig. 10(c) shows that the increase of Ppump
results in a slight increase of Q 000 . Here, the change of Ppump is
mainly related to the variation of the air flow rate instead of the
refrigerant flow rate, thus it did not influence Q 000 much similar to
the result of Fig. 10(a).

6. Conclusions

In conclusion, we report a cross-flow microchannel heat


exchanger of size 1 cm3 and capable of heat transfer up to 60 W.
The heat exchanger was monolithically fabricated from a block of
copper alloy using lEDM. Air channels of diameter 520 lm carried
air flows of velocity 20–100 m/s, which corresponded to Reair of
1.2  104 to 2.05  104. Rectangular channels of size 0.5  2 mm
carried refrigerant R245fa, which corresponded to Reref of 1000–
2300. The heat exchanger exhibited a maximum power density
of 60 W/cm3 when the air flow at inlet had temperature of 27 °C
and Reair was 20,500 while the refrigerant flow at inlet had temper-
ature of 80 °C and Reref of 2300. The implementation of a high-
speed compressible air flow in microchannels with a short length
(thermal entry length) permitted the exceptionally high heat
transfer coefficients (>2000 W/m2 K). In addition, the optimized
numbers of air channels (Nair = 150) and refrigerant channels (Nref
= 8) helped to match the thermal resistances between the working
fluids and to reduce the total thermal resistance below 5 K/W. This
work helps to understand the heat transfer in compact microchan-
nel devices, and demonstrate a high-power-density of a microma-
chined metal heat exchanger.

Fig. 10. Heat exchanger power density (Q 000 ) as a function of (a) Air side Reynolds Acknowledgment
number Reair, (b) Refrigerant side Reynolds number Reref, and (c) pumping power
Ppump.
We gratefully acknowledge support from the Defense Advanced
Research Projects Agency. This work was also supported by the
National Science Foundation Engineering Research Center for
Power Optimization of Electro Thermal Systems (POETS) with
it is 20 by the Shah-London correlation. The entrance region
cooperative agreement EEC-1449548.
effects become more important at higher Reynolds numbers, in
part explaining the increasing h value with Reynolds number.
Fig. 9 shows the area goodness factor (j/f) as a function of the
Conflict of interest
Reynolds number. The ratio j/f is about 0.047 for the air flow and
about 0.47 for the refrigerant flow. The ratio j/f based on measure-
The authors declare no conflict of interest.
ments is relatively larger than j/f obtained by correlations mostly
due to the discrepancy in j-factor. In both channels, j/f increases
with the Reynold number, indicating that the flow is not fully Appendix A. Supplementary material
developed. For enhancing j/f particularly in the air channels, oper-
ating in the incompressible regime is a possible method, as this Supplementary data associated with this article can be found, in
would lower f. However, reducing the air flow rate should be well the online version, at https://doi.org/10.1016/j.ijheatmasstransfer.
balanced between j/f gain and the heat transfer rate loss. 2017.11.068.
B. Kwon et al. / International Journal of Heat and Mass Transfer 118 (2018) 1276–1283 1283

References [15] M.V.V. Mortean, L.H.R. Cisterna, K.V. Paiva, M.B.H. Mantelli, Development of
diffusion welded compact heat exchanger technology, Appl. Therm. Eng. 93
(2016) 995–1005.
[1] I. Mudawar, Assessment of high-heat-flux thermal management schemes, IEEE
[16] A.F. Mills, Heat Transfer, Prentice Hall, Upper Saddle River, NJ, 1999.
Trans. Compon. Packag. Technol. 24 (2) (2001) 122–141.
[17] L. Cabezas-Gómez, H.A. Navarro, J.M. Saíz-Jabardo, Thermal Performance
[2] D.B. Tuckerman, R.F.W. Pease, High-performance heat sinking for VLSI, IEEE
Modeling of Cross-Flow Heat Exchangers, Springer, New York, NY.
Electron. Device Lett. 2 (5) (1981) 126–129.
[18] F. Kreith (Ed.), The CRC Handbook of Thermal Engineering, CRC Press LLC, Boca
[3] P.A. Kew, D.A. Reay, Compact/micro-heat exchangers – their role in heat
Raton, FL, 2000.
pumping equipment, Appl. Therm. Eng. 31 (5) (2011) 594–601.
[19] T.L. Bergman, A.S. Lavine, F.P. Incropera, D.P. Dewitt, Fundamentals of Heat and
[4] S. Ashman, S.G. Kandlikar, A review of manufacturing processes for
Mass Transfer, John Wiley & Sons Inc, Hoboken, NJ, 2011.
microchannel heat exchanger fabrication, in: ASME 4th International
[20] J.R. Davis, Copper and Copper Alloys, ASM International, Novelty, OH, 2001.
Conference on Nanochannels, Microchannels, and Minichannels, Parts A and
[21] J. Fernández-Seara, F.J. Uhía, J. Sieres, A. Campo, A general review of the wilson
B, ASME, 2006, pp. 855–860.
plot method and its modifications to determine convection coefficients in heat
[5] R.J. Phillips, Forced-Convection, Liquid-Cooled, Microchannel Heat Sinks M.S.
exchange devices, Appl. Therm. Eng. 27 (17) (2007) 2745–2757.
thesis, Massachusetts Institute of Technology, Cambridge, MA, 1987.
[22] A.S. Tucker, The LMTD correction factor for single-pass crossflow heat
[6] M.J. White, G.F. Nellis, S.A. Klein, W. Zhu, Y. Gianchandani, An experimentally
exchangers with both fluids unmixed, J. Heat Transfer 118 (2) (1996) 488.
validated numerical modeling technique for perforated plate heat exchangers,
[23] W.M. Kays, A.L. London, Compact Heat Exchangers, Krieger Pub, Co, Malabar,
J. Heat Transfer 132 (11) (2010) 111801.
FL, 1998.
[7] C.R. Friedrich, S.D. Kang, Micro heat exchangers fabricated by diamond
[24] M.S. Kim, M.A. Kedzierski, Single-phase heat transfer and pressure drop
machining, Precis. Eng. 16 (1) (1994) 56–59.
characteristics of an integral-spine fin within an annulus, J. Enhanc. Heat
[8] W. Bier, W. Keller, G. Linder, D. Seidel, K. Schubert, H. Martin, Gas to gas heat
Transf. 3 (3) (1996) 201–210.
transfer in micro heat exchangers, Chem. Eng. Process.Process Intensif. 32 (1)
[25] L. Luo, Y. Fan, W. Zhang, X. Yuan, N. Midoux, Integration of constructal
(1993) 33–43.
distributors to a mini crossflow heat exchanger and their assembly
[9] R. Nacke, B. Northcutt, I. Mudawar, Theory and experimental validation of
configuration optimization, Chem. Eng. Sci. 62 (13) (2007) 3605–3619.
cross-flow micro-channel heat exchanger module with reference to high mach
[26] A.H. Shapiro, The Dynamics and Thermodynamics of Compressible Fluid Flow,
aircraft gas turbine engines, Int. J. Heat Mass Transf. 54 (5) (2011) 1224–1235.
John Wiley & Sons Inc, New York, NY, 1953.
[10] C. Harris, M. Despa, K. Kelly, Design and fabrication of a cross flow micro heat
[27] R.M. Curr, D. Sharma, D.G. Tatchell, Numerical predictions of some three-
exchanger, J. Microelectromechanical Syst. 9 (4) (2000) 502–508.
dimensional boundary layers in ducts, Comput. Methods Appl. Mech. Eng. 1 (2)
[11] C.S. Roper, R.C. Schubert, K.J. Maloney, D. Page, C.J. Ro, S.S. Yang, A.J. Jacobsen,
(1972) 143–158.
Scalable 3D bicontinuous fluid networks: polymer heat exchangers toward
[28] Y. Ji, K. Yuan, J.N. Chung, Numerical simulation of wall roughness on gaseous
artificial organs, Adv. Mater. 27 (15) (2015) 2479–2484.
flow and heat transfer in a microchannel, Int. J. Heat Mass Transf. 49 (7–8)
[12] G.L. Benavides, L.F. Bieg, M.P. Saavedra, E.A. Bryce, High aspect ratio meso-
(2006) 1329–1339.
scale parts enabled by wire micro-EDM, Microsyst. Technol. 8 (6) (2002) 395–
[29] S.G. Kandlikar, S. Joshi, S. Tian, Effect of surface roughness on heat transfer and
401.
fluid flow characteristics at low reynolds numbers in small diameter tubes,
[13] K. Ahmmed, C. Grambow, A.-M. Kietzig, Fabrication of micro/nano structures
Heat Transf. Eng. 24 (3) (2003) 4–16.
on metals by femtosecond laser micromachining, Micromachines 5 (4) (2014)
[30] R.K. Shah, D.P. Sekulic, Fundamentals of Heat Exchanger Design, John Wiley &
1219–1253.
Sons Inc, Hoboken, NJ, 2003.
[14] T. Li, Q. Bai, Y.B. Gianchandani, High precision batch mode micro-electro-
discharge machining of metal alloys using DRIE silicon as a cutting tool, J.
Micromech. Microeng. 23 (9) (2013) 95026.

You might also like