You are on page 1of 21

International Journal of Heat and Mass Transfer 166 (2021) 120809

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer


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

Experimental heat transfer and flow simulations of rectangular


channel with twisted-tape pin-fin array
S.W. Chang a,∗, P.-S. Wu b, W.L. Cai a, C.H. Yu a
a
Department of Systems and Naval Mechatronic Engineering, National Cheng Kung University, No. 1, University Road, Tainan City, 70101, Taiwan R.O.C.
b
Department of Mechanical and Automation Engineering, Da-Yeh University, No. 168, University Road, Dacun Changhua 51591, Taiwan, R.O.C.

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

Article history: In this study, an innovative twisted-tape pin-fin array was proposed for passive heat transfer enhance-
Received 20 July 2020 ment (HTE) to improve the aerothermal performance of channel flows. The full-field Nusselt number
Revised 14 October 2020
distributions, Fanning friction coefficients and thermal performance factors of an improved channel were
Accepted 3 December 2020
measured at Reynolds numbers of 50 0 0, 750 0, 10,0 0 0, 12,50 0, and 15,0 0 0. The turbulent channel flows
Available online 16 December 2020
were numerically explored using the ANSYS-Fluent code to perform correlative analysis between the
Keywords: aerothermal measurements and numerical flow results. A set of aerothermal data and numerical flow
Twisted Tape Array results was selected to illustrate the properties of HTE and the responsive flow mechanisms. Because of
Heat Transfer Enhancement the vortical flows induced by each twisted tape in the array, the average measured Nusselt numbers and
Pin-Fin Channel Fanning friction coefficients of the improved channel were 5.26–4.73 and 39.4–38.5 times higher than
those of a plain tube, respectively; thus, the thermal performance factor (TPF) of the improved channel
was 1.55–1.39 when 50 0 0 < Re < 15,0 0 0. Compared with the pin-fin channels with different geometries
reported in the literature, the twisted-tape pin-fin channel designed in this study provided a higher de-
gree of HTE and thus a higher TPF at the same pressure drop penalty and pumping power. On the basis of
the experimental results, two empirical correlations that were used to evaluate the average endwall Nus-
selt number and Fanning friction coefficient of the designed pin-fin channel were proposed for relevant
applications.
© 2020 Elsevier Ltd. All rights reserved.

1. Introduction amine the effects of streamwise spacing on the turbulent struc-


ture of the pin-fin channel. By decreasing the streamwise spac-
A pin-fin channel has various industrial applications, such as ing between pin-fins, the vortex shedding frequency and thus the
the cooling of electronic chipsets and gas turbine blades and the Strouhal number increased. However, the disturbances generated
heat transfer enhancement for a heat exchanger. When a coolant in the upstream pin row, which caused instability in the shear
flows through a patterned pin fin in a channel, the stagnation flow layers to break the Karman vortices and periodic vortex shedding
at the leading edge of each pin and the separated flows generated behind the pin fins, was Re-dependent [2]. In a later study, Os-
by the adverse pressure gradients behind each pin induce complex tanek and Thole [3] performed correlative flow and heat transfer
vortical flow structures that are dependent on the pin shape, pin- analysis for a pin-fin channel. With small streamwise pin pitches,
array arrangement, and channel geometry. Lawcon et al. [1] con- the quasi-periodic fluid motions were attenuated. The level of ran-
ducted experiments to determine the effects of pin spacing on dom motions was not considerably affected by such pin-fin spac-
the aerothermal performance of the pin-fin channels when 50 0 0 ing. Because the heat transfer rates in a pin-fin channel generally
< Re < 30,0 0 0 (Re is the Reynolds number). The spanwise and increased with a decrease in the pin pitch, the periodic fluid mo-
streamwise pin spacing considerably influenced the pressure drop tions caused by the pin fins did not play a dominant role in heat
and heat transfer performance, respectively. Ostanek and Thole transfer enhancement (HTE) [3].
[2] adopted time-resolved particle image velocimetry (PIV) to ex- At each pin-endwall junction, a pair of horseshoe vortices is
triggered due to localized flow instability and boundary-layer re-
development. The downstream convection of the lifted horseshoe

Corresponding author. vortices in the channel flow stimulates dynamic interactions be-
E-mail addresses: swchang@mail.ncku.edu.tw (S.W. Chang), tween the horseshoe vortices and the separated vortices behind
pswu@mail.dyu.edu.tw (P.-S. Wu), p18071018@mail.ncku.edu.tw (W.L. Cai), each pin to cause vortex breakdown, which affects spatial transport
p16074010@ncku.edu.tw (C.H. Yu).

https://doi.org/10.1016/j.ijheatmasstransfer.2020.120809
0017-9310/© 2020 Elsevier Ltd. All rights reserved.
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

tween pin rows to the unstable shedding vortices agitated by sep-


Nomenclature arated flows. Because of the boundary-layer redevelopment from
the leading edge of each pin, the heat transfer rate increases over
English symbols the surface of each pin [5]. Chyu et al [5] and VanFossen [6] re-
A effective endwall heat transfer surface area (m2 ) spectively reported that the Nusselt numbers (Nu) over the pin
Cp specific heat at constant pressure (Jkg-1 K-1 ) surfaces were 35% and 10-20% higher than those on the channel
d hydraulic diameter of a rectangular channel (m) endwall. However, unlike the boundary-layer-type channel flow, in
dp diameter of the circular cap of a twisted-tape pin which the local Nu decays toward a developed level in the down-
fin (m) stream direction, the HTE mechanisms attributed to a pin-fin ar-
f Fanning friction factor of a pin-fin chan- ray grow along the channel to compete against the increased near-
nel = P/[(0.5ρWm 2 )](d/4L ) wall thermal resistance due to the thickened boundary layers. In
E total fluid energy (J/kg) their experimental study, Axtmann et al. [7] observed that row-
f∞ Fanning friction factor evaluated from the Blasius resolved Nu exhibited an initial increase and reached a peak level
equation at the third to fifth pin row. This streamwise Nu variation was
H channel height (m) Re-independent [7]. The advantages of a pin-fin channel are its
HTE heat transfer enhancement extended cooling surfaces, improved structural integrity, and en-
hloss heat loss coefficient (Wm-2 K-1 ) hanced heat transfer rates. However, the augmented form and fric-
k turbulence kinetic energy (J/kg) tional drags as well as the turbulent wakes considerably increase
keff effective thermal conductivity (Wm-1 K-1 ) the pressure drop penalties for this channel.
kf thermal conductivity of the coolant (Wm-1 K-1 ) For controlling the pressure drop enhancement of a pin-fin
kw thermal conductivity of the scanned channel end- channel, Chang et al. [8], Siw et al. [9], and Moores et al. [10] have
wall (Wm-1 K-1 ) proposed the detached pin-fin concept, which enables the reduc-
L channel length (m) tion of the form and frictional drags by eliminating the horseshoe
m ˙ coolant mass flow rate (kgs-1 ) vortices at each pin–endwall junction. The detached pin fin im-
Nu local Nusselt number of the channel end- proves the aerothermal performance of channel flows mainly by
wall = (q f d )/[(Tw − Tb )k f ] substantially reducing the pressure drops across the pin-fin array,
Nu∞ Dittus-Boelter Nusselt number reference which leads to an increased thermal performance factor (TPF) at
P static pressure of the fluid (Pa) the same pumping power [8–10]. Moreover, increased uniformity is
P channel entry-to-exit average pressure difference observed in the heat transfer distributions of the channel endwalls
(Nm-2 ) because of the elimination of horseshoe vortices at the multiple
qf convective heat flux (Wm-2 ) pin–endwall junctions [8,9]. Moores et al. [10] performed pressure
qgen total heat flux generated by the Joule heating of the drop and heat transfer measurements using liquid crystal thermog-
channel wall (Wm-2 ) raphy at Re values between 200 and 20,0 0 0 to examine the influ-
qloss heat loss flux through the convective, conductive or ence of the pin-tip clearance to diameter ratio on the aerothermal
radiative pathway (Wm-2 ) performance. The pin tip-clearance to diameter ratio affected the
qrad radiative heat flux (Wm-2 ) pressure drops to the greatest extent when Re < 50 0 0, with the
Re Reynolds number = ρ Wm d/μ effect being significantly diminished when Re = 15,0 0 0. In com-
Sh volumetric heat sources (Wm-3 ) parison with the heat transfer performances generated by the at-
Sx,y x,y-wise pitch ratio of the pin-fin array tached pin-fin array, the higher heat transfer rates per unit pump-
TPF thermal performance factor evaluated at a constant ing power were reported when the dimensionless pin-tip clearance
pumping power = (Nu/N u∞ )/[( f / f∞ )1/3 ] was less than 0.2. Jadhav and Balaji [11] numerically examined the
Tamb ambient temperature (K) turbulent flow and heat transfer characteristics of a rectangular
Tb fluid bulk temperature (K) channel with detached circular pin fins extending from two op-
Tf fluid local temperature (K) posite walls. The pin-tip clearance incurred significant thermal im-
Tw wall temperature (K) pacts at low fluid velocities; however, this effect weakened at fluid
t thickness of the scanned channel endwall (m) velocities higher than 2 m/s. The diameter of the pin fin was the
ui , j turbulent fluctuating velocity (ms-1 ) geometric parameter with the greatest effect on the pressure drop
W channel width (m) and thermal resistance.
Wm mean through-flow velocity (ms-1 ) The flow and heat transfer characteristics of pin-fin channels
x, y, z coordinates (m) also vary with the shape and arrangement of pin fins and the type
y pitch of the twisted tape of the pin fin (m) of coolant. The thermal performance of micro pin fins has been
intensively examined for microchannel HTE. Laminar flow with a
Greek symbols low Reynolds number is a common feature of micro pin fins in
ε turbulence dissipation rate (m2 s-3 ) microchannels. A literature review on these works is summarized
εt thermal emissivity in the following subsections.
σ Stefan-Boltzmann constant
μ dynamic viscosity of the coolant (kgm-1 s-1 )
1.1. Shape and arrangement of pin fins
μe f f effective viscosity for turbulent flow (kgm-1 s-1 )
ρ fluid density (kgm-3 ) Goldstein and Chen [12] conducted mass transfer and pres-
(τ ij )eff , stress tensor (Nm-2 ) sure loss measurements at Re values between 30 0 0 and 18,0 0 0
Subscripts for a rectangular channel with three types of pin fins, namely one
channel entrance uniform-diameter circular pin and two stepped-diameter circular
in
channel exit pins. Complex flow structures that consisted of repeated horse-
out
shoe vortices and pin wakes were visualized using oil-lampblack
paint. The downflow from the stepped-diameter pins caused them
phenomena. Ames and Dvorak [4] attributed the endwall HTEs be- to have different flow characteristics to the uniform-diameter pin.

2
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

The flow near the channel endwall and the heat transfer rate var- with rectangular pin columns of variable pin lengths and longitudi-
ied with each pin row. These parameters exhibited stable patterns nal pin pitches. A pin-fin column with the larger pin lengths at the
after the third pin row. The array-averaged Sherwood numbers for channel entry region and with the shorter longitudinal pin pitches
the pin surfaces were higher than their endwall averaged Sher- at the channel exit region was an effective HTE arrangement. The
wood numbers. Among the three types of pin-fins, the stepped- arrangement in which the pin lengths were in high-to-low order
diameter circular pin with the shorter stepwise length exhibited exhibited the optimal thermal performance, with the maximum
the optimal aerothermal performance. Tanda [13] experimentally thermal performance factor reaching 1.65.
studied the aerothermal properties of a rectangular channel with
seven arrangements of diamond-shaped pin-fin arrays using liq- 1.2. Type of coolant
uid crystal thermography. The heat transfer rate of the rectangular
channel was 4.4 and 1.65 times higher than that of a plain chan- Liao et al. [21] numerically examined the heat transfer perfor-
nel at a constant flow rate and pumping power, respectively. Uzol mance of a pin-fin array in a wedge duct using air, steam and
and Camci [14] measured the Nusselt numbers, friction factors, and mist–steam mixture as coolants. The mist–steam mixture exhibited
wake flow fields downstream of two-row staggered elliptical and the lowest friction coefficient and highest thermal performance
circular pin-fin arrays in a channel using liquid crystal thermog- factor and thus considerably improved the HTE properties of the
raphy and total-pressure probe wake surveys at Re values between pin-fin array in the wedge duct. Duangthongsuk and Wongwises
18,0 0 0 and 86,0 0 0. A set of PIV flow measurements at Re = 18,0 0 0 [22] experimentally examined the aerothermal performance of a
were used to correlate the heat transfer measurements. A larger heat sink constructed with miniature pin fins using ZnO-water and
zone of low-momentum wake was observed for the circular pin- SiO2 -water as the working fluids. The heat transfer rates with the
fin array than for the elliptical pin-fin array. The turbulent kinetic ZnO–water coolant were approximately 3%–9% higher than those
energy in the wake zone of the circular pin-fin array was higher with the SiO2 –water nanofluid. The addition of nanoparticles only
than that in the wake zone of the elliptical pin-fin array. On aver- showed a small effect on the pumping power.
age, the friction factors for the circular pin fins were 46.5%–59.5%
higher than those for the elliptical pin fins. However, the average 1.3. Micro pin-fin channels
endwall Nusselt numbers within the wake were approximately 27%
higher for the circular pin fins than for the elliptical pin fins. Wang Shrouded pin-fin arrays with tip clearance have been integrated
et al. [15] numerically and experimentally examined the flow, fric- in silicon microchannels for achieving increased HTE benefits un-
tion and heat transfer characteristics of rectangular channels with der laminar flow conditions for electronic cooling. Tullius et al.
staggered pin-fin arrays of circular, elliptical and drop-shaped cross [23] conducted a numerical study to optimize the laminar convec-
sections at Re values between 4800 and 8200. The streamlined tive heat flux of minichannels enhanced by circular, square, trian-
drop-shaped pin fins suppressed flow separation; thus, the rectan- gular, elliptical, hexagonal, and diamond-shaped pin fins made of
gular channel with drop-shaped pin fins had a higher aerodynamic various materials. The pin-fin height was optimized for the square
efficiency than that with circular pin fins. However, compared with detached pin fins with various tip clearances. Among the various
the circular pin fins, the HTE effect of the teardrop shaped pin fin pin-fin configurations studied, the triangular fins with larger fin
is more moderate than the increase in aerodynamic efficiency. At a height, smaller fin width and spacing exhibited the optimal heat
constant flow rate, the tear-drop shaped pin fins exhibited a higher transfer performance. Roth et al. [24] measured the heat transfer
thermal performance index than did the circular pin fins. Shafeie rates in straight silicon microchannels with shrouded pin-fin arrays
et al. [16] numerically analyzed the laminar flow and heat trasnfer that had staggered and in-line arrangements at tip clearance to pin
of pin-fin microchannel heat sinks and pin-fin heat sinks by ar- diameter ratios between 0.5 and 0.77 and Re values between 9 and
ranging the pin-fin arrays in an oblique or staggered manner. The 246. Because a considerable portion of the coolant stream passed
entropy generation minimization method was adopted to optimize through the clearance bypass region, the aforementioned channels
the cooling performance. At medium or high pumping powers, the exhibited superior HTE properties to those of a plain channel only
heat flux transferred by the pin-fin heat sinks was less than that at small clearance to diameter ratios. Adewumi et al. [25] per-
transferred by the optimum simple pin-fin microchannel heat sink. formed a numerical study with a goal-driven optimization tool to
At low pumping powers, the pin-fin heat sinks marginally outper- optimize the geometry of an integrated microchannel and micro
formed the optimum pin-fin microchannel heat sink. pin fins for different solid volumes to minimize the peak tempera-
For increasing the efficiency of a thermoelectric generator ture. The minimized peak temperature decreased with an increase
(TEG), Pandit et al. [17] used three-dimensional circular, triangu- in the Bejan number.
lar, hexagonal, and diamond-shaped partial pin fins in rectangu- For optimizing the aerothermal performance of microreactors
lar channels to enhance the heat transfer on the hot side of the using micro pin fins, Mei et al. [26] numerically studied the ef-
TEG. Among the various pin-fin shapes tested, the diamond-shaped fect of the tip clearance of the detached pins on the aerothermal
pin fins exhibited the highest HTE effectiveness. Moon and Kim performance at 33 < Re < 350. Unlike the effect of tip clearance
[18] numerically explored the flow and heat transfer characteristics for a pin-fin channel with a turbulent flow, the aerothermal per-
of a rectangular channel with fan-shaped pin fins. Multiobjective formances of micro pin-fin arrays were sensitive to the tip clear-
optimization of the fan-shaped pin fins was conducted to maxi- ance with the promoted HTE effectiveness and the reduced pres-
mize the heat transfer rate and minimize friction loss simultane- sure drops. Abdoli et al. [27] numerically studied the aerothermal
ously. Compared with the circular pin fins, the fan-shaped pin fins characteristics of micro pin-fin arrays using the ANSYS-Fluent code
caused higher HTE. Recently, Caliskan et al. [19] strung five thin to examine the thermal effects of micro pin-fin shapes on the cool-
hexagonal and cylindrical fins around a circular pin to form hexag- ing of high-heat-flux electronic chips with a single hot spot. The
onal pin fins (HPFs) and cylindrical pin fins (CPFs), respectively, cross-sections of the micro pin fins had circular, hydrofoil, mod-
for developing pin-fin arrays for rectangular channels. The HTE of ified hydrofoil and symmetric convex shapes. At a constant flow
the staggered pin-fin array generally exceeded that of the in-line rate, the micro pin fins with the hydrofoil shape exhibited the
array. Although the HPFs and CPFs both exhibited high HTE, the optimal aerothermal performance by moderating the flow separa-
average Nu of the HPFs was higher than that of the CPFs. Rezaee tions. For the same total cross-sectional area, although the channel
et al. [20] experimentally and numerically explored the hydrother- with hydrofoil pin fins had a larger frictional area compared with
mal characteristics of water-cooled, parallel straight channels fitted the channel with the micro circular pin fins, the pumping power

3
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

and overall convective heat flux for the channel with hydrofoil pin into the online condition monitoring program for calculating in-
fins were 30.4% and 3.2% smaller, respectively, than those of the stant Re value at the channel entrance. The airflow rate was fine-
channel with micro circular pin fins. Thus, the form drags associ- tuned during each heat transfer test to compensate for the changes
ated with the flow separation exceeded those associated with the in density and viscosity due to fluid temperature variation. At the
frictional drag, leading to the larger influence on pumping power. channel entrance, the Re value was controlled to within +1% of the
Yang et al. [28] explored the heat transfer and pressure drop char- test value. Because the Prandtl number variations caused by fluid
acteristics of microchannel heat sinks with staggered pin fins of property variations were less than +1% under all test conditions,
rhombus, hydrofoil and sine shapes. Among these symmetrical pin the effect of the Prandtl number on the aerothermal performance
fins, the sine pin fins exhibited the highest heat removal capacity of the proposed twisted-tape pin-fin channel was not investigated.
with the lowest pressure drop penalty.
Wan and Joshi [29] performed a numerical and experimental 2.1. Experimental equipment
study to investigate the single-phase thermal and fluid charac-
teristics of microgaps by fitting staggered pin-fin arrays on one Figure 1(a) displays the experimental equipment, the construc-
wall. The effects of the pin-fin diameter, height, and spacing on tional details of test channel and the geometric specifications of
the aforementioned characteristics were studied at Re values be- the twisted-tape array. As displayed in Fig. 1(a), a rotary-type com-
tween 22 and 357. The effect of the friction factor on Re system- pressor fed pressurized air into a tank. A dehumidifier was in-
atically weakened as Re increased from 22 to 357. At the same Re stalled downstream of the air tank for providing cold dry airflow
value, the pressure drop of the square pin-fin array was consid- to the test channel. The airflow rate was adjusted using a nee-
erably higher than that of the circular pin-fin array; thus, at the dle valve located upstream of the test channel. The flow rate and
same pumping power, the circular micro pin-fin array exhibited pressure of the airflow into the test channel were measured us-
a superior thermal performance to that of the square pin-fin ar- ing a digital mass flow meter and pressure gauge, respectively. The
ray. Ambreen et al. [30] numerically studied the combined effect online condition monitoring program facilitated the evaluations of
of the pin-fin shape and nanofluid coolant on the fluid flow and fluid viscosity, density, and Reynolds number at the channel en-
thermal performance of a pin-fin heat sink. The staggered pin-fin trance using the detected mass flow rate (m˙ ), inlet fluid pressure
arrays comprised micro pin fins with square, circular, and triangu- (Pin ), and temperature (Tf ,in ). Because the fluid properties were
lar cross-sections, and the total pressure drop was limited in the temperature- and pressure-dependent, the instantaneous measure-
range of 570- 2760 Pa. Irrespective of the pin-fin configuration, the ments of m˙ , Pin and Tf ,in were constantly fed into the condition
nanofluid coolant enhanced the heat exchange capacity. However, monitoring program for calculating the Reynolds number at the
the HTE achieved with an appropriate pin-fin configuration ex- channel entrance. The needle valve was accordingly adjusted to
ceeded that achieved with the nanofluid. At the highest flow rate, regulate m˙ for maintaining the inlet Re to within +5% of the target
the average Nusselt numbers obtained with the nanofluid-cooled value.
circular, square, and triangular pin fins were 23.1%, 16.5%, and 8% As depicted in Fig. 1 (a), prior to entering the test channel, the
higher, respectively, than that obtained with the water-cooled tri- airflow was directed into the divergent entry air plenum chamber
angular pin fins. (1), in which the flow was stratified by a honeycomb and steel
With advancements in manufacturing technologies, HTE has wire. The inlet fluid temperature was measured using a thermo-
been achieved for pin-fin channels [1–30] by changing the pin couple that penetrated into the core of the air plenum chamber
shape and array arrangement as well as through array miniatur- (1), which is connected to a rectangular test channel (2). The chan-
ization. Inspired by the early research of Goldstein and Chen [12], nel length (L), width (W), and height (H) were 225, 100 and 25
who developed three-dimensional pin fins for generating down- mm, respectively. The channel width-to-height ratio (aspect ratio)
flow from the steps of pins to promote HTE, the present study pro- and the channel hydraulic diameter (d) were 4 and 40 mm, re-
posed an innovative twisted-tape array as a passive HTE method spectively. The characteristic length selected to define Re, Nu, and f
for improving the aerothermal performance of channel flows. With was d. For presenting the experimental and numerical results, the
the proposed twisted-tape array, the large form drags associated x-y-z coordinate system was adopted. The origin of this coordinate
with typical blunt pins were reduced. Moreover, a vertical swirl system was at the inlet corner of the channel back wall, as indi-
was induced around each twisted-tape pin for acquiring increased cated in Fig. 1(a). The dimensionless x, y and z coordinates were
HTE. The turbulent flow structures in the proposed pin-fin chan- defined as x/L, y/W, and z/H, respectively. The scanned Tw endwall
nel were generated using the ANSYS-Fluent code for interpretating was specified as the x-y plane at z/H = 1 on the front endwall.
the measured full-field Nusselt number distributions, Fanning fric- The test channel comprised a Teflon channel back wall (3), side-
tion coefficients, and thermal performance factors. On the basis of walls (4) and a front frame (5). The twisted-tape pin fins (6) were
the experimental data, two sets of heat transfer and pressure drop arranged in staggered manner, with the axial and spanwise pin
correlations were devised for relevant applications. pitches being 2.5 times the pin diameter (dp ). All the twisted-tape
pin fins had an anticlockwise twist orientation. As indicated by the
2. Experimental details image in by Fig. 1(a), a 2-mm-thick and 15-mm-long twisted plate
was mounted between two circular caps. The height and diameter
The full-field Nusselt number distribution, Fanning friction co- of each circular cap were 5 and 10 mm, respectively. The width
efficient, and thermal performance factor of the proposed twisted- of the twisted tape was 8 mm. The ratio between the twist pitch
tape pin-fin channel were measured at Re values of 50 0 0, 750 0, and tape width was 1.875. By retaining the two circular caps that
10,0 0 0, 12,50 0, and 15,0 0 0. At each tested Re value, the heating were jointed to the channel front and back endwalls, horseshoe
power was adjusted to raise the maximum wall temperature to vortices were triggered at each pin-endwall junction. Although a
373K. Steady-state raw data were acquired for evaluating Re, Nu swirl was induced orthogonally to the bulk stream and centered
and f. The steady-state condition was assumed when the variations on each twisted-tape pin fin for HTE, the large form drag associ-
in several successive Tw scans were less than +0.3K. To satisfy this ated with the blunt shape of a solid pin fin decreased. The twisted
steady-state condition, it took 40-50 minutes of waiting period af- pin fins were adiabatic because they were fabricated through the
ter regulating heating power or, and, airflow rate. To control Re at three-dimensional printing method using polylactide as the raw
the target value, the inlet fluid temperature and pressure, ambi- material. A circular cap that embedded the twisted tape was con-
ent pressure and mass flow rate of the coolant were constantly fed solidated with the Teflon back wall. Each twisted-tape element was

4
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

Fig. 1. (a) Experimental equipment, test channel, and twisted-tape array and (b) energy conservation scenario for a small element from the scanned channel wall.

inserted into a 2.25-mm-deep rectangular groove on the circular The entire assembly of the test channel was tightened using axial
cap at the Teflon back wall. In Fig. 1(a), the x-wise and y-wise pin bolts. To prevent air leakage, sealant was applied to all the clear-
pitches (Sx and Sy , respectively) are 2.5dp . ances between the components.
The two heated channel endwalls (7 and 8) were made from The wall temperature (Tw ) distribution on the heated front end-
a 0.1-mm-thick stainless-steel foil. The stainless-steel heating foil wall was scanned from the backside of the stainless-steel heating
(7), which acted as the front channel wall, was sandwiched be- foil (7) using an infrared radiometer (IR) (11) at a scan rate of 60
tween the Teflon front frame (5) and two channel sidewalls (4). frames per second. The IR camera lens was parallel to the scanned
The back heating foil (8) was inlaid on the Teflon back wall (3). To heating foil (7). The maximum uncertainty for the Tw scan was
secure each twisted-tape pin fin on the channel front wall (heat- 0.3K according to the calibration report of the IR system. To en-
ing foil), the outer circular Teflon lid, which comprised a section hance the infrared emission, the scanned surface of the heating
of thread, was inserted into the in-channel twisted-tape pin fin to foil was painted black with a thermal emissivity of 0.98. The layer
tightly lock the heating foil (7) between the lid and the twisted- of black paint had a thickness of approximately 30 μm, and the
tape pin fin. The two axial ends of each stainless-steel heating foil variation in the local thickness was less than 5%. Because all the
(7 and 8) were sandwiched between the two pairs of copper plates Biot numbers for the scanned heating foil were less than 0.001,
(9 and 10) embedded in the entry and exit Teflon flange of the the scanned Tw distribution on the backside of the heating foil (7)
test channel. Electrical cables were connected with copper plates was representative of that at the fluid-wall interface over the inner
(9 and 10) to feed heating power through the two heating foils. channel wall.

5
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

The thermocouple (12) was located at the center of the entry For polished stainless-steel foils, the emissivity coefficients ε t ,A
plane in the air plenum chamber (1) to measure Tf ,in . Six equally and ε t ,B in Eq. (4) are both 0.09. At the highest wall temperature
spaced thermocouples (13) were installed at the exit plane of the (Tw ) and the lowest fluid bulk temperature (Tb ), the maximum qrad
test section to measure the exit fluid temperatures. The averaged between the two parallel front and back heating foils was 21.4
fluid temperature detected by the six thermocouples (13) was con- Wm-2 , which was less than 0.35% of qgen . Because the qloss (radiation)
sidered the measured exit fluid temperature (Tf ,out ). On the basis value evaluated using Eq. (3) was approximately 11.7% of qgen , most
of the measured local convective heat flux (qf ) and air mass flow of the radiative heat loss flux was transferred into the surround-
rate (m˙ ) at each tested Re value, the axial increase in the fluid bulk ings from the back black surface of the front heating wall (foil).
temperature (Tb ) over an axial segment was evaluated using the Although the test range of Biot number permitted a diminished
enthalpy balance method. An enthalpy accountancy method was wall temperature difference across the 0.1-mm-thick stainless-steel
conducted from the channel entrance to the channel exit to eval- heating foil, the local wall temperature gradients, which were in-
uate the local fluid bulk temperatures. When the discrepancy be- fluenced by the nonuniform Nu distribution, modified the qf dis-
tween the calculated and measured fluid bulk temperatures was tribution over the inner surface of the heated wall (foil). In gen-
less than +10%, the corresponding set of raw data was adopted for eral, conductive heat fluxes are transferred from locations with
evaluating the full-field Nusselt number distribution. higher Tw values to locations with lower Tw values. This wall con-
The Fanning friction factor was measured on the basis of the duction effect, which modifies the local qf distribution, is gov-
pressure drop across the test channel. Two pressure taps that were erned by the component kw t (∂ 2 Tw /∂ x2 + ∂ 2 Tw /∂ y2 ) in Eq. (1).
linked to a digital micro-manometer (14) were located at the inlet However, for turbulent flows, the temporal turbulent fluctuations
and outlet of the test section as indicated in Fig. 1(a). The precision cause the corresponding spatial Tw oscillations to affect the eval-
of the adopted digital micromanometer was 0.1 Pa. uation of the second derivatives of Tw . The numerical filtration of
each Tw scan is a crucial step prior to the calculation of the sec-
2.2. Data processing method ond derivative of Tw . Thus, the numerical average scheme in the
MATLAB code was used to filter each Tw scan before implement-
The most important aspect for generating high-quality Nu data ing the finite-difference scheme that evaluates the local values of
for the present test channel was the evaluation of the local qf dis- kw t (∂ 2 Tw /∂ x2 + ∂ 2 Tw /∂ y2 ) in Eq. (1). Due to the thin heating foil
tribution over the heating foil where the Tw field was scanned. and relatively low kw value for the stainless-steel heating foil, the
When applying the energy conservation principle to a small el- conductive heat flux was between 4.36% and 8.72% of qgen .
ement from 0.1-mm-thick stainless-steel heating foil indicated in To determine the heat loss flux characteristics (qloss
Fig. 1(b), the various heat loss fluxes through the conductive, free-
(free convection) + qloss (conduction) ) of the test channel, heat loss
convective and radiative pathways; and the wall conduction effects calibration tests were conducted after blocking off the test channel
within the heating foil affected the qf distribution. In each steady- by filling a wool layer into the airway. Without flow in the test
state condition, the energy conservation principle for the small el- channel, Eq. (1) is reduced to the following equation:
ement from the heating foil with a heat transfer area of dxdy fol-  
lowed the following equation: qgen − qloss(freeconvection) + qloss(conduction) − qloss(radiation)
   2 2 
qgen − q f − qloss(freeconvection) + qloss(conduction) − qloss(radiation) + kw t ∂ Tw /∂ x +∂ Tw 2 /∂ y2 = 0 (5)
 
+ kw t ∂ Tw 2 /∂ x2 +∂ Tw 2 /∂ y2 = 0 (1) Following the aforementioned procedures for calculating
The thermal conductivity and thickness of the stainless-steel qloss(radiation) and kw t (∂ 2 Tw /∂ x2 + ∂ 2 Tw /∂ y2 ) using the steady-state
heating foil are denoted by kw and t, respectively. In Eq. (1), qgen , is Tw maps obtained at five ascending qgen values, the heat loss
the supplied heat flux generated by the Joule heating power con- term in Eq. (5) (qloss (free convection) + qloss (conduction) ) was measured
verted from the electricity, which is calculated as follows: at the five qgen values. When the free-convective heat transfer
coefficient on the back side of the heating wall (foil) remains
qgen =(EI )/A (2) invariant in a steady state during each heat loss test, Eq. (5) is
where E and I represent the measured voltage and current fed converted into Eq. (6), in which hloss is the heat loss coefficient
into the two opposite channel walls (foils), respectively, and A rep- originated from qloss (free convection) + qloss (conduction) .
resents the effective endwall heat transfer surface area. The heat   
loss fluxes emitted from the back side of the scanned heating wall qgen dxdy − hloss (Tw − Tamb )dxdy − qloss(radiation)
(foil) to the surroundings (qloss (free convection) + qloss (conduction) ) are A A A
functionally related to the local channel wall-to-ambient tempera- 
ture difference. The characteristics of these fluxes were determined
+ kw t ∇ 2 Tw = 0 (6)
A
through a set of heat loss calibration tests which will be described
later. According to the Newton cooling law for free convection and
The Stefan-Boltzmann equation was used to evaluate the local the Fourier conduction law for conductive heat loss, the combined
radiative heat loss flux (qloss (radiation) ). The local radiative heat loss heat loss flux (qloss (free convection) + qloss (conduction) ) is a function of
flux from the back side of the channel wall (foil) to the surround- the wall-to-ambient temperature difference. The parameter hloss is
ings was calculated from the measured Tw map using the following obtained using Eq. (7).
equation:     
hloss Tw − Tamb = qgen − σ εt Tw4 − Tamb
4
+ kw t ∇ 2 Tw (7)
qloss(radiation) = σ εt (Tw − Tamb )
4 4
(3) A A
where σ , ε t and Tamb are the Stefan-Boltzmann constant, thermal
where Tw is the average wall temperature determined from the
emissivity of the black paint of 0.98, and ambient temperature, re-
steady-state Tw map detected in each heat loss test. After evaluat-
spectively. Another stream of radiative heat flux was generated be-
ing the terms on the right-hand side of Eq. (7) for each set of heat
tween the two polished stainless-steel foils of the front and back
loss calibration test results, the hloss value was determined for each
channel walls. This radiative heat flux (qrad ) was determined as the
measured Tw − Tamb value. The variation in hloss in the adopted test
thermal radiation between two parallel gray bodies as follows:
  range of (Tw − Tamb ) was less than 5.6%. After the hloss value was
qrad = σ Tw 4 − T f 4 /[1/εt,A + 1/εt,B + 1] (4) determined for the current testing equipment, the local value of

6
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

qloss (free convection) + qloss (conduction) in Eq. (1) was calculated using precision was specified as ±0.01 mm for a channel hydraulic diam-
Eq. (8), in which Tw is the local wall temperature measured in the eter of 40 mm; thus, the error percentage of the channel hydraulic
heat transfer test conducted at the ambient temperature Tamb . diameter was 0.025%.
  The precision errors of the temperature measurements were
qloss ( f reeconvection ) + qloss (conduction ) = hloss (Tw − Tamb ) (8)
mainly attributed to the adopted IR system, which had an un-
For the generated heat transfer data, the value of qloss certainty of 0.3 K. For the wall-to-fluid temperature difference
+ qloss (conduction) was in the range of 8.20%–16.5% of
(free convection) (T = Tw − Tb ) of 32–54 K, the error percentage of T mea-
qgen . surement was 0.93%. The calibration report of the mass flow me-
After evaluating all the heat flux components in Eq. (1) using ter specified the maximum error percentage of m˙ measurement as
the measured Tw map and Tamb value obtained in each heat trans- 0.2%. Using the error propagation method, the maximum error per-
fer test, the local qf distribution was determined. The local Nu on centage of qf was determined to be 4.3%. The maximum error for
the scanned channel endwall was calculated using the following the measurement of P, which ranged from 21.7 to 153 Pa, was
equation: 0.46%. According to the precision and bias errors for qf , Tw − Tb ,
m˙ , P, μ, and kf , the root-mean-square experimental uncertainties
Nu = q f d/[k f (Tw − Tb )] (9)
at 95% confidence interval for Nu, Re, and f were estimated to be
where kf and d are the thermal conductivity of air at the local fluid +8.21%, +3.62%, and +7.68%, respectively.
bulk temperature (Tb ) and the channel hydraulic diameter, respec-
tively. 3. Numerical method
The Fanning friction factor in each heat transfer test was mea-
sured using the detected pressure drop (P) across the test chan- The turbulent airflows passing through the twisted-tape pin-fin
nel according to the following equation: channel were simulated using the ANSYS-Fluent code (ANSYS Inc.,
  Canonsburg, Pennsylvania, USA) at the same Reynolds numbers se-
f = P/(0.5ρWm 2 ) × (d/4L ) (10)
lected for the heat transfer measurements. In the simulation, the
where ρ is the coolant density determined according to the ideal control-volume method was used to solve the steady-state equa-
gas law at the channel entry pressure and temperature and Wm is tions for the Reynolds-averaged mass, momentum, and energy for
the averaged fluid velocity evaluated from the m˙ , ρ , and channel predicting the time-averaged fluid velocity, pressure, and tempera-
inlet cross-sectional area values measured when no pin fin is used. ture as well as the turbulence properties. The spatial discretization
The Reynolds number at the channel entrance was constantly scheme based on the Green–Gauss node was used to discretize
evaluated by the online condition monitoring program during each the continuity, momentum, energy, and turbulence equations that
heat transfer test using the following equation: were solved iteratively using the SIMPLE algorithm. According to
the Cartesian coordinate system of the ANSYS-Fluent code [33],
Re = ρinWm d/μ (11)
which is different from the coordinate system selected for pre-
The coolant viscosity (μ) was calculated from the measured senting the experimental and numerical results, the time-averaged
fluid inlet temperature (Tf ,in ). continuity, momentum, and energy equations can be expressed us-
The thermal performance factor (TPF) concept of Gee and Webb ing Eqs. (13)-(15), respectively [33].
[31], according to which the HTE ratio of St/St∞ is normalized ∂
by (f/f∞ )1/3 to satisfy the comparative criterion of fixed pumping ( ρ ui ) = 0 (13)
∂ xi
power, was used for assessing the HTE efficiency of the proposed
twisted-tape pin-fin channel. The reference friction factor (f∞ ) and 

∂   ∂p ∂ ∂ ui ∂ u j 2 ∂ uk
Nusselt number (Nu∞ ) for evaluating the TPF are customarily ob- ρ ui u j = − + μ + − δ
tained using the Dittus-Boelter correlation and Blasius equation for ∂xj ∂ xi ∂ x j ∂ x j ∂ xi 3 i j ∂ xk
turbulent flows, respectively. To compare the St/St∞ ratios for var- ∂  
ious HTE methods at constant Re and Pr values, the St/St∞ ratios + −ρ ui  u j  (14)
∂xj
were reduced to Nu/Nu∞ . In each heat transfer test condition, the 
thermal performance factor was determined using the following ∂   ∂ ∂T
u j (ρ E + p) = ke f f + ui (τi j )e f f + Sh (15)
equation: ∂xj ∂xj ∂xj
 1/3

T P F = (Nu/N u∞ )/ ( f / f∞ ) (12) where E and keff are the total energy and the effective thermal con-
The determination of experimental uncertainties of Re, Nu and ductivity, respectively; (τ ij )eff is the deviatoric stress tensor that
f followed the statistical inference of Kline and McClintock [32] for represents the viscous heating; and Sh represents the volumetric
estimating the uncertainty in the single-sample experiments. The heat sources. The deviatoric stress tensor (τ ij )eff is defined as fol-
lows:
sources of bias and precision errors for these dimensionless pa-

rameters originated from the measurements of qf , Tw -Tb , m˙ , P   ∂ ui ∂ u j 2 ∂ uk


and the fluid properties such as μ and kf . The precision errors of τi j e f f = μe f f + − μe f f δ (16)
∂ x j ∂ xi 3 ∂ xk i j
each measurement device were determined from the calibration
report provided by the manufacturer, whereas the bias errors of Where μeff is the effective viscosity and δ ij is the Kronecker
the aforementioned measurements were estimated as the standard delta.
deviation of a set of raw data detected by the measurement de- To evaluate the Reynolds stresses [−ρ ui  u j  in Eq. (14)] for clos-
vice. The percentage errors of the fluid properties were defined as ing the formulation, the fluctuating velocities were solved using a
the maximum discrepancies between the correlation results and turbulence model. In the simulation in this study, the renormaliza-
the table values. Based on the standard deviations for the mea- tion group (RNG) k − ε model comprising the turbulence transport
surements at the same reference conditions, the error percentages equations summarized in Eqs. (17) and (18) was used. The details
of heater power, temperature, airflow rate, and P were 1.2%, 3.1%, regarding the selection of the RNG k − ε turbulence model are pre-
2.3% and 3.6%, respectively. A comparison of the table and corre- sented later in this section.

lated values of kf and μ indicated that the error percentages of kf ∂ ∂ ∂k


and μ were ±0.06% and ±0.05%, respectively. The manufacturing
( ρ k u i) = α μ + Gk + Gb − ρε (k equation ) (17)
∂ xi ∂xj k ef f ∂xj

7
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

Table 1
Boundary and inlet or outlet conditions, thermal-physical properties of the coolant and geometric parameters of the twisted-tape pin-fin channel.

Boundary and inlet/outlet conditions


Uniform heat flux at channel wall, qf = 5000 Wm-2 , Pout = 1.013 bar (uniform)
Reynolds number (Re) 5000 7500 10000 12500 15000
Wm (uniform inlet fluid velocity, ms-1 ) 1.96 2.94 3.92 4.90 5.88
Turbulent intensity at channel inlet (%) 5.51 5.24 5.11 4.92 4.81
Temperature, pressure and thermal-physical properties of fluid at channel inlet
Tf (uniform, K) kf (Wm-1 K-1 ) Cp (Jkg-1 K-1 ) ρ (kgm-3 ) Pr
300 0.02624 1004.9 1.177 0.707
Geometric parameters of twisted-tape pin-fin channel
Channel aspect ratio (W/H) x-wise pin pitch (mm) y-wise pin pitch (mm) dp (mm) d (mm) Twist pitch (mm) (y/dp = 1.875)
100mm/25mm = 4 25 25 10 40 15

∂ ∂ ∂ε ε
(ρε ui ) = αε μe f f + C1ε (Gk + G3ε Gb )
∂ xi ∂xj ∂xj k
ε2
− C2ε ρ (ε equation ) (18)
k
In the k − ε equations, Gk and Gb represent the turbulence ki-
netic energy generated due to the mean fluid velocity gradients
and buoyancy, respectively. In Eqs. (17) and (18); C1ε , C2ε and G3ε
are model constants. Moreover, αk and αε are the inverse effective
Prandtl numbers for k and ε , respectively. The main differences be-
tween the RNG k − ε and standard k − ε models include the eval-
uation of C2ε in the ε Eq. (18), the computation of the inverse ef-
fective Prandtl numbers (αk and αε ), and the evaluation of the tur-
bulent viscosity, μe f f . These modifications improve the accuracy of
the RNG k − ε model for rapidly strained and swirling flows [33].
The numerical procedures for the specification of boundary
conditions, testing of grid independency, selection of the turbu-
lence model, and validation of the simulation results by comparing
the calculated f and endwall-averaged Nu against the experimental
data are presented in the following text.
The uniform fluid velocities at Re values of 50 0 0, 750 0, 10,0 0 0,
12,50 0, and 15,0 0 0 and a uniform fluid temperature of 300 K were
specified at the channel entry plane. The uniform turbulence in-
tensity at the channel entry plane is recommended to be 0.16Re-1/8
[33]. The complex vortical flows induced by the present twisted-
tape pin-fin array caused high fluid velocity gradients and hence
nonuniform velocities over the channel exit plane; thus, the fluid
pressure was specified as the atmospheric condition at the chan-
nel exit for acquiring superior convergence and accuracy to those
achieved with the outflow boundary condition. Through the selec-
tion of appropriate functional options in ANSYS-Fluent code, the
fluid properties were set as temperature-dependent and the fluid
density was determined using the ideal gas law. The convergence
criterion was set with the residuals of iterations within 1E-6 for
Fig. 2. (a) Variation in the endwall-averaged Nu against the element number and
energy equation and 1E-4 for the other equations. Table 1 sum-
(b) mesh structure adopted in the numerical simulation.
marizes the boundary and inlet or outlet conditions, the thermal-
physical properties of the coolant at the channel entrance, and the
geometries of the twisted-tape pin-fin channel.
On the basis of the calculated Tw and pressure distribu- for various grid structures. Fig. 2 displays the variation in the
tions over the channel inlet plane, the endwall-averaged Nus- endwall-averaged Nusselt number with the number of elements
selt numbers and Fanning friction facrors were determined us- [Fig. 2(a)] and the mesh structures adopted in the numerical sim-
ing Eqs. (9) and (10), respectively. The fluid bulk temperatures for ulation [Fig. 2(b)]. Fig. 2(a) illustrates the gradual decline in the
Nu calculations were determined through the aforementioned en- endwall-averaged Nusselt number toward a stable value as the
thalpy balance method using the airflow rate and heat flux speci- number of elements increases. The endwall-averaged Nusselt num-
fied in the numerical method. For the predictions of f, the pressure bers for two mesh structures with 35.3 million and 58.3 million
drops across the twisted-tape pin-fin channel (P) at all the sim- elements exhibited a difference of less than 1%, which was defined
ulated Re values were calculated from the sectional average static as the grid independency requirement. For economic computation,
pressures over the channel inlet and outlet planes. the mesh structure with 35.3 million elements was selected for the
The grid independency test for the twisted-tape pin-fin channel numerical study.
was conducted using the RNG k-ε turbulence model at Re = 15,0 0 0 The grid used in the simulation as well as the three mesh struc-
by comparing the endwall-averaged Nusselt numbers evaluated tures across the AA´, BB´, and CC´channel cross-sections are illus-

8
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

Fig. 3. Comparisons of the numerically predicted and experimentally determined (a) endwall-averaged Nusselt numbers and (b) f-factors for the twisted-tape pin-fin channel
when 50 0 0 < Re < 15,0 0 0.

trated in Fig. 2(b). The regional zoom-in plots exhibiting the grids The selection of the turbulence model and validation of the nu-
in the flow field adjacent to these solid surfaces for the AA´, BB´, and merical results were performed by comparing the average Nusselt
CCśections are also displayed in Fig. 2(b). A multi-grid mesh struc- numbers and Fanning friction factors predicted using various tur-
ture was constructed using tetrahedral core elements and 20-layer bulence models with the experimental measurements. Fig. 3 de-
refined near-wall hexagonal meshes. The highly dense grids adja- picts the variations obtained in the endwall-averaged Nusselt num-
cent to the channel walls and the curly surfaces of all the twisted- ber [Fig. 3(a)] and Fanning friction factor [Fig. 3(b)] with the Re
tape pin fins are depicted in Fig. 2(b). In the aforementioned mesh value when using the standard k-ε , RNG k-ε and shear-stress
structure, the length ratios between the channel-core tetragonal transport (SST) k-
models. The experimental results displayed in
meshes and the joined boundary layer fine meshes were main- Figs. 3(a) and 3(b) served as the criteria for turbulence model se-
tained to be less than 1.5 at their interface. Enhanced wall treat- lection and result validation. As displayed in Figs. 3(a) and 3(b),
ment [33] was applied to determine the near-wall fluid velocities the endwall-averaged Nusselt numbers and Fanning friction factors
and temperatures, both of which were sensitive to the fluid vis- predicted by the RNG k–ε model exhibited the lowest discrepan-
cosity. The turbulence properties were calculated from the chan- cies with the experimental results (12.0%–0.40% and 7.03%–2.94%,
nel core to viscous sublayer. The body size control parameter was respectively) when 50 0 0 < Re < 15,0 0 0. The superior and satis-
selected to control the element size. A total of 20 prism layers factory validation results obtained with the RNG k–ε model are
with a wall-to-core growing rate of 1.2 were used to construct mainly attributed to its improved ε equation. The RNG k–ε model
meshes normal to the heated channel wall to resolve the large is more suitable than the standard k–ε model for analyzing flows
near-wall fluid velocity and temperature gradients. Through using a with high strain rates as well as mixed high- and low-Reynolds-
grid structure that satisfied the defined grid independency require- number regions. Therefore, the RNG k-ε turbulence model was
ment, the y+ values for the near-wall elements were controlled to adopted in the numerical investigation.
be less than unity (0.01–0.07).

9
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

Fig. 4. Contours and vectors of the fluid velocity projected onto the xz planes at y/W values of 0.5 (centerline), 0.575, 0.625, 0.675, and 0.75 when Re = 10,0 0 0.

Fig. 5. Velocity contours and vectors projected onto the yz planes adjacent to the channel entrance at x/L = 0.056, 0.111, 0.167, and 0.222 as well as yz planes near the
channel exit at x/L = 0.778, 0.833, 0.889, and 0.944 when Re = 10,0 0 0.

10
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

Fig. 6. Distributions of the z-wise velocity components on the yz planes at (a) x/L = 0.056, 0.167, 0.833, and 0.944 across the pin rows and (b) x/L = 0.111, 0.222, 0.778, and
0.889 between the pin rows when Re = 10,0 0 0.

4. Results and discussion 1. Such upward and downward flows are unique phenomena for
the twisted-tape pin-fin channel and affect the endwall heat trans-
4.1. Numerical flow characteristics fer properties. Additional details regarding the distributions of the
z-wise velocity components on the yz planes are described later in
The turbulent flow structures of the twisted-tape pin-fin chan- this section. Although the characteristic patterns of the flow dis-
nel are typified by the simulation results acquired at Re = 10,0 0 0. tributions on the xz planes at y/W = 0.5, 0.625 and 0.75 are sim-
Fig. 4 depicts the contours and vectors of the fluid velocity pro- ilar, the sidewall effect weakens the axial momentum, leading to
jected onto the xz planes at y/W values of 0.5 (centerline), 0.575, systematic reductions in the axial fluid velocity as the xz plane
0.625, 0.675, and 0.75 when Re = 10,0 0 0. moves from y/W = 0.5 (centerline plane) toward the channel side-
As displayed in Fig. 4, the uniformly distributed entry flow has wall at y/W = 0.75 (Fig. 4). The xz planes at y/W = 0.575 and
periodic velocity variations along the test channel after the airflow 0.675 have mirror-image velocity distributions. The flow structures
traverses the first pin row. On the xz planes sectioned through the on the xz planes at y/W = 0.575 and 0.675 are affected by the
column of the twisted-tape pin fins, the low-velocity zone behind swirling streams induced along the two twisted-tape pin columns
each twisted-tape pin fin that corresponds to the low-momentum adjacent to these planes. Half the height of each twisted surface of
wake zone, is nonuniformly distributed along the channel height the pin fin is in the same direction as the main stream, whereas
but symmetrical about the middle plane of z/H = 0.5. The triple the remaining height of the twisted surface is in the opposite di-
low-momentum wake zones behind the two circular caps and the rection to the main stream. Consequently, the low-velocity zones
inlaid twisted tape are clearly visible after the second pin row. The developed on the xz planes at y/W = 0.575 and 0.675 are alterna-
x-wise span of such triple low-velocity zones is approximately 1.5 tively shifted about the z/H = 0.5 centerline, as displayed in Fig. 4.
dp . Local flow accelerations are directed into the twisted section Due to the absence of pin-associated form drags on the xz planes
of the downstream pin fins. Heat transfer impacts associated with at y/W = 0.575 and 0.675 as well as the blockage effect caused by
the wake zones behind each pin and the localized flow acceleration the two adjacent pin columns, the flow velocities on the xz planes
downstream the low velocity zones will be later examined in the at y/W = 0.575 and 0.675 are generally higher than those on the
next section. xz planes at y/W = 0.5, 0.625, and 0.75 (Fig. 4).
Unlike the situation for a solid pin fin, the typical stagnation Fig. 5 displays the contours and vectors of the fluid velocity
flow attached to the entire leading edge of the twist section of projected onto the yz planes adjacent to the channel entrance at
each pin fin is weakened. Instead, the air stream is guided to swirl x/L = 0.056, 0.111, 0.167, and 0.222 as well as the yz planes near
along the curly surface of the twist section to distort the x-wise the channel exit at x/L = 0.778, 0.833, 0.889, and 0.944. In con-
mainstream flow toward the two channel endwalls at z/H = 0 and trast to the weakened axial momentum of the fluid flow near the

11
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

Fig. 7. Distributions of the y-wise velocity components on the yz planes at (a) x/L = 0.056, 0.167, 0.833, and 0.944 across the pin rows and (b) x/L = 0.111, 0.222, 0.778, and
0.889 between the pin rows when Re = 10,0 0 0.

Fig. 8. Distributions of the vorticity on the yz planes at x/L = 0.056, 0.167, 0.833, and 0.944 across the pin rows and at x/L = 0.111, 0.222, 0.778, and 0.889 over the
mid-sections between the pin rows when Re = 10,0 0 0.

12
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

Fig. 9. Distributions of the turbulence kinetic energy on the yz planes at (a) x/L = 0.056, 0.167, 0.833, and 0.944 across the pin rows and (b) x/L = 0.111, 0.222, 0.778, and
0.889 at the mid-sections between pin rows when Re = 10,0 0 0.

two channel sidewalls, the cross-flow velocities near the two chan- circular pin-fin arrays [2,3], Ostanek and Thole observed that the
nel sidewalls at x/L> 0.167 are locally amplified. In [14], the flow fluid motions were dominated by periodic or quasi-periodic vor-
separation on the pin surface was reported to push the upstream tex shedding and affected by the disturbances from the upstream
wake toward the channel sidewall. This phenomenon can lead to wakes that broke up to form Karman vortices. Such flow structures
momentum enrichments such as those near the two channel side- are considerably different from those depicted in Figs. 4 and 5 and
walls illustrated in Fig. 5. A cross examination of the compatible are characterized by strong channel cross flows comprising coher-
velocity projections on the yz planes near the entry and exit of ent multiple vortices that are distributed symmetrically about the
the pin-fin channel indicates that the strengths of the secondary diagonal axis across each yz plane. The thermal properties of the
flows on the yz planes are enhanced in the downstream direction. twisted-tape pin-fin array are closely related to the z-wise fluid ve-
In general, the sectional flow structures in each yz plane are diag- locity components, which enhance the core–wall momentum and
onally symmetric. On the yz planes across the first, second, eighth, energy exchanges, and y-wise velocity components, which influ-
and ninth pin rows at x/L = 0.056, 0.167, 0.833, and 0.944, the ence the spanwise heat transfer variations over each channel end-
cores of the low-velocity zones are trapped aside the two diag- wall. The distributions of the z-wise and y-wise fluid velocity com-
onal lee sides of each twisted tape. Fluids coming from another ponents on the sectioned yz planes are illustrated in Figs. 6 and 7,
pair of windward sides of each twisted tape are tripped over the respectively.
curly surface to swirl toward the channel front (z/H = 1) and back The magnitudes of the z- and y-wise velocity components on
(z/H = 0) walls. Large-velocity components are emitted from the the yz planes across the pin rows, which act as vortex generators
two diagonal windward surfaces of each twisted tape toward the for each twisted-tape pin fin, are approximately 1.5–2 times higher
front (z/H = 1) and back (z/H = 0) walls, as illustrated in each yz than those of the z- and y-wise velocity components on the mid-
plot across the pin row displayed in Fig. 5. The flows emitted from pin yz planes. Thus, different ranges of the z- and y-wise velocity
the curly surface of each twisted tape induce vortical flow cells be- components are selected for constructing Figs. 6 and 7 for clearer
tween two adjacent twisted tapes in each pin row on each yz plane illustration of the distribution patterns. Diagonally symmetric dis-
sectioned through the pin row. The flows originating from the two tributions observed for the z- and y-wise velocity components on
twisted tapes along the two channel sidewalls evolve into multi- each yz plane. As depicted in Fig. 6(a), for the yz planes across
ple corner vortices that swirl into downstream locations. On the the pin rows, alternate positive and negative velocity zones are ob-
yz planes sectioned through the mid-pin locations at x/L = 0.111, served along the front (z/H = 1) and back (z/H = 0) endwalls from
0.222, 0.778, and 0.889 (Fig. 5), an S-shaped low-velocity zone is y/W = 0 to y/W = 1. Positive and negative z-wise velocity compo-
developed behind the curly tail of each twisted tape. On these yz nents are consistently observed around the upper and lower halves
planes, the multiple vortical flow cells located downstream of each of each twisted tape in each plot in Fig. 6(a). Multiple z-wise veloc-
pin row are diagonally symmetric with the high flow momentums ity cells with positive and negative magnitudes along the channel
originating along the two channel sidewalls. With regard to the front (z/H = 1) and back (z/H = 0) endwalls are more evident for
typical turbulent flow structures in channels with the staggered the mid-pin-row yz planes displayed in Fig. 6(b) than for the yz

13
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

Fig. 10. Fluid temperature distributions on the (a) xz and (b) yz planes when Re = 10,0 0 0.

14
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

Fig. 12. (a) Full-field and sectioned, (b) x-wise, and (c) y-wise Nu distributions
when Re = 15,0 0 0.
Fig. 11. (a) Full-field and sectioned, (b) x-wise, and (c) y-wise Tw distributions when
Re = 15,0 0 0.

each twisted tape include the skewed anticlockwise vertical swirl


planes across the pin rows depicted in Fig. 6(a). The upward and around each twisted-tape pin fin. Near-wall cross flow is gener-
downward velocity components displayed in Fig. 6 are attributed ated in the y-direction at the side edges of the two circular caps
to the vertical swirl tripped around each twisted-tape pin fin. Be- attached on the front and back endwalls. On the yz plane at the
cause alternate upward and downward velocity components are mid-section between two adjacent pin rows, the skewed clockwise
observed over the mid-pin-row yz planes, swirls are expected to vertical swirl is induced behind each twisted-tape pin fin by the
form downstream of each twisted-tape pin fin. Thus, fluid momen- upstream vortex around the twisted tape [Fig. 7(b)]. In particular,
tum exchanges in upward and downward directions are caused by the negative and positive y-wise velocity cells along the two op-
the multiple vertical swirls that enhance the core-endwall momen- posite channel front and back endwalls at z/H = 1 and z/H = 0 on
tum and energy transmissions. These enhancements lead to overall the yz planes depicted in Fig. 7(b) are larger than those on the yz
HTEs relative to the heat transfer of the solid pin-fin channels (de- planes across the pin rows displayed in Fig. 7(a). Due to the nu-
scribed in the next section). merous generated near-wall y-wise velocity components, spanwise
The distributions of y-wise velocity components on the yz heat transfer variations are expected over the channel front and
planes (Fig. 7) also indicate the effects of the z-wise swirls gen- back endwalls (described later in the next section by presenting
erated around the twisted tapes on the yz planes. As indicated the measured endwall Nusselt number distributions). Although the
in each plot of Fig. 7(a), the positive and negative y-wise veloc- pressure drags formed due to the conventional blunt solid pin fins
ity components around the upper-right and lower-left portions of decrease when using the twisted-tape pin fins, shear layers are in-

15
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

Fig. 13. Measured endwall Nu distributions at Re = (a) 50 0 0, (b) 7500, (c) 10,000, (d) 12,500, and (e) 15,000.

duced between the boundaries of multiple flow cells. These layers vorticity stripes at x/L = 0.944. Because of the entry and exit ef-
increase the local vorticity and turbulence intensity, which results fects on the vorticity distribution, the endwall heat transfer prop-
in enhancements in the frictional drags and energy transmissions erties are influenced by the channel entry effect and the upstream
in the pin-fin channel. effect caused by the prescribed flow exit condition.
Figures 8 and 9 depict the distributions of the vorticity and tur- In general, the distribution patterns of turbulence kinetic en-
bulent kinetic energy, respectively, on the yz planes at x/L = 0.056, ergy on the yz planes displayed in Fig. 9 are similar to the pat-
0.167, 0.833, and 0.944 across the pin rows and at x/L = 0.111, terns of the vorticity maps depicted in Fig. 8. However, the high
0.222, 0.778, and 0.889 between the pin rows when Re = 10,0 0 0. vorticity within the boundary layer displayed in Fig. 8 is converted
The high vorticity and turbulent kinetic energy developed on the into low turbulence kinetic energy in Fig. 9 because the near-wall
mid-row yz plane behind each pin row in Figs. 8 and 9 indi- low-momentum fluids are subjected to large viscous drags Because
cate that the turbulence and vorticity are mainly generated within the distribution patterns of the vorticity and turbulence kinetic en-
the wake zones behind the twisted-tape pin fins. As displayed in ergy are highly similar in the turbulent core, the presence of shear
Fig. 8, a high vorticity is observed within the thin boundary layers layers between the multiple vortical flow cells acts as the main
along the peripheries of the channel walls and twisted-tape pin flow physic for turbulence production in the pin-fin channel. Axial
fins. On the yz planes across the pin rows, the flow regions with augmentation of the overall turbulent kinetic energy is observed
low velocities but high velocity gradients at the upper-left and along the pin-fin channel, especially for the flow regions along the
lower-right portions of each twisted tape depicted in Fig. 5 cor- two channel sidewalls. The axially enhanced turbulent kinetic en-
respond to the high-vorticity regions in Fig. 8. On the yz planes at ergy and secondary flows, which are displayed in Figs. 9 and 5, re-
the mid-pin-row sections, diffused twin-“ϩ”-shaped high-vorticity spectively, counteract the increased thermal barrier caused by the
stripes are developed behind each twisted-tape pin fin, as illus- thickened boundary layers along the two opposite channel end-
trated in Fig. 8. The vorticity stripes on each yz plane at the mid- walls to affect the Nusselt number distributions (will be shown in
pin-row section are associated with the upstream vertical swirl the next section).
around each twisted-tape pin fin. Axial (x-wise) variations in the The combined effects of the momentum transport properties,
vorticity distribution are mainly observed in the regions adjacent namely the fluid velocity field and spatial distributions of the vor-
to the channel entrance and exit. The “ϩ”-shaped high-vorticity ticity and turbulence kinetic energy (Figs. 4–9), cause the fluid
stripes between two adjacent pin fins and along the two channel temperature distributions (Tf ) to affect the heat transfer perfor-
sidewalls on the yz planes across the first pin row at x/L = 0.056 mance on the endwalls with uniform heat fluxes. Because the
are transformed into three straight y-wise stripes at x/L = 0.167 change in the constant-pressure specific heat of the coolant with Tf
through the second pin-row. Due to the upstream effect caused by variation is less than 0.73%, the simulated fluid bulk temperature
the uniform atmospheric pressure distribution at the channel exit, increases linearly in x-wise direction at each Re value. Fig. 10 de-
the straight y-wise high-vorticity stripes along the two channel picts the Tf distributions over the xz and yz planes [Figs. 10(a)
sidewalls at x/L = 0.833 transformed again into “ϩ”-shaped high- and 10(b), respectively] when Re = 10,0 0 0. Fig. 10(a) illustrates

16
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

stream developments of vertical swirls tripped by the twisted-tape


pin fins, a difference gradually emerges in the thicknesses of the
high-temperature layers on the endwall between the flow regions
near the sidewall edges at y/W = 0 and 1. As depicted in Fig. 10(b)
, for the sectional plots near the flow exit at x/L> 0.778, the wavy
high-temperature layers attached along the upper endwall near the
sidewall edge at y/W = 1 are generally thicker than those adjacent
to the sidewall edge at y/W = 0. Because the thermal boundary
layers create heat transfer barriers, the increase in the thickness
of the high-temperature layer from y/W = 0 to y/W = 1 at x/L>
0.778 along the edge at z/H =1, where the wall temperatures are
scanned, reflects the y-wise Nu decrease from y/W = 0 to y/W = 1.
Due to the symmetric fluid temperature distributions about the
diagonal of each yz plane [Fig. 10(b)], the reversed y-wise Nu in-
crease from y/W = 0 to y/W = 1 due to the thinner high-Tf thick-
ness along the bottom channel endwall at z/H = 0 is expected in
the flow region located away from the channel entrance. The y-
wise Tw and Nu variations over the scanned channel endwall are
described in the following section, and the measured Tw and Nu
maps are used for the detailed examination of the endwall heat
transfer properties.

4.2. Experimental results

4.2.1. Heat transfer performances


Because the two channel endwalls have uniform heat fluxes,
their heat transfer properties are directly reflected by the wall
temperature (Tw ) distributions. Fig. 11 depicts the full-field and
sectioned, x-wise, and y-wise Tw distributions [Figs. 11(a)–11(c),
respectively] at Re = 15,0 0 0. On the scanned heated endwall,
which corresponds to the upper endwall at z/H = 1 in the cur-
rent computational fluid dynamics simulation, y-wise Tw variation
emerges at x/L = 0.056 [Fig. 11(a)]. At further downstream loca-
tions, the vertical swirls tripped by the twisted-tape pin-fin array
cause the bulk shift of the high-Tw zone toward the upper-right
corner of Fig. 11 (a) from the low-Tw zone near the channel en-
Fig. 14. Variations in (a) Nu and (b) Nu/Nu∞ with Re for the pin-fin channels pre-
sented in [7,14,34–37] and the test channel proposed in this study. trance. Consequently, the values of the axial Tw profiles across all
the pin columns displayed in Fig. 11(b) systematically increased
from y/W = 0.25 toward y/W = 0.75 and from y/W = 0.125 to-
ward y/W = 0.875. However, this y-wise Tw increase is less signif-
the manner in which the axial Tf increases with the z-wise pen- icant at the flow regions adjacent to the first pin row as depicted
etrations of hot fluids from the two endwalls toward the chan- in Fig. 11(b). The variation of axial fluid bulk temperature (Tb ) is
nel core. Along the pin-fin channel, the uniform Tf distribution at also included in Fig. 11(b). The typical heat transfer effect associ-
the entry yz plane gradually changes into a symmetric distribution ated with the boundary layer development is represented by the
about the central plane at z/H = 0.5. The thickness of each high- reduced Tw − Tb value as the flow region approaches the channel
temperature layer adjacent to the channel front or back endwall in- entrance. Behind each circular cap of the twisted-tape pin fin, the
creases along the pin-fin channel, which reflects the axial growth axial Tw increases toward a local peak at approximately 2/3 of the
of thermal boundary layers on the two opposite heated endwalls pin-pitch location. Subsequently, the axial Tw decreases until the
and leads to the axial development of the z-wise Tf gradient in the flow impinges the front edge of the downstream circular pin cap.
downstream direction. In the sectional plots depicted in Fig. 10(a), Because of the upstream effects on the distributions of the vortic-
the high temperature regions behind the two circular caps of each ity and turbulence kinetic energy (Figs. 8 and 9, respectively), Tw
twisted-tape pin fin are slightly extended from their leading-edge decreases near the channel exit [Fig. 11(b)].
counterparts. In particular, on the two xz planes at y/W = 0.575 Fig. 11(c) displays the y-wise Tw variations along the pin-fin
and 0.675 along the mid-sections of two adjacent row columns, channel, which mainly respond to the velocity and Tf distributions
the high-Tf layers adjacent to the two channel endwalls are not on the yz planes [Figs. 5–7 and Fig. 10(b), respectively]. In each plot
uniformly distributed but have wavy forms. Considering the axial of Fig. 11(c), two pairs of compatible y-wise Tw profiles near the
Tf increase illustrated in Fig. 10(a), the temperature ranges of the channel entrance and exit are compared. Because of the increase in
Tf maps for the yz planes sectioned along the pin-fin channel dis- the axial Tf along the pin-fin channel, the two pairs of y-wise Tw
played in Fig. 10(b) are selected. profiles near the channel entrance and exit are clearly distinguish-
As illustrated in Fig. 10(b), the high-temperature layers along able. The Tw drops toward the two adiabatic channel sidewalls are
the two channel endwalls at z/H = 0 and z/H = 1 have wavy forms. detected for all the y-wise Tw profiles displayed in Fig. 11(c). Away
The multiple vortical flows adjacent to the two channel endwalls from the channel entrance at y/W = 0.778 and 0.833 as well as
disturb the development of boundary layers. At the entry flow re- along the sections at y/W = 0.889 and 0.944, the overall spanwise
gion at x/L < 0.222, the distributions of the high-temperature lay- Tw increases from the sidewall edge at y/W = 0 to the opposite
ers on the two endwalls near the sidewall edges at y/W = 0 and channel sidewall at y/W = 1 [Fig. 11(c)]. Such an increase is not
y/W = 1 are similar as shown by Fig. 10(b). Along with the down- observed in the y-wise Tw in the flow region near the channel en-

17
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

Fig. 15. Variations in (a) f (b) f/f∞ with Re for the pin-fin channels presented in [7,14,34–37] and the test channel proposed in this study.

trance because the vortical flows immersing in the bulk coolant try effect associated with the development of boundary layers is
stream are still developing. Due to the presence of the near-wall observed at the entry region on the endwall, the axial growth of
vortical flows in the yz planes (Fig. 5), the y-wise Tw profiles ex- the vortical flows tripped by the twisted-tape pin fin considerably
hibit wavy patterns in Fig. 11(c). increases the local Nu level adjacent to the upwind face of each
Fig. 12 depicts the Nu distributions corresponding to the Tw re- pin fin located downstream of the fifth pin row [Fig. 12(b)].
sults displayed in Fig. 11. As displayed in Fig. 12(a), high Nusselt As displayed in Fig. 12(b), the local Nusselt numbers systemati-
numbers develop along the sidewall at y/W = 0 and in the re- cally decrease as the y/W section moves toward the sidewall edge
gion adjacent to the channel entrance. After the cooling airflow at y/W = 1 from y/W = 0. The y-wise Nu variation along the pin-
traverses three pin rows, the y-wise Tw increases from the channel fin channel is depicted in Fig. 12(c). Because the secondary flow
sidewall at y/W = 0 to the upper-right endwall corner depicted in fields are developing in the yz planes near the flow entrance at
Fig. 11(a), which causes the particular endwall Nu distribution il- x/L = 0.167 and 0.222 as well as at x/L = 0.056 and 0.111, no ev-
lustrated in Fig. 12(a). The low-Nu zone along the sidewall edge at ident y-wise Nu decays are observed in the wavy Nu profiles. For
y/W = 1 after the fourth pin row shrinks and becomes skewed to- the further downstream locations at x/L = 0.778, 0.833, 0.889, and
ward the upper-right endwall corner displayed in Fig. 12(a). In the 0.944, Nu decays can be observed in the corresponding y-wise Nu
whole flow region, the HTE caused due to flow impingement at profiles. Along the two adiabatic channel sidewalls, local Nu eleva-
the upwind face of each pin fin induces a semiring-shaped high- tions are consistently observed [Fig. 12(c)].
Nu zone in front of each pin fin. Behind each pin fin, the local The measured endwall Nu distributions for all the tested
Nusselt numbers decay axially and then recover, as indicated by Reynolds numbers have similar patterns as those displayed in
the x-wise Nu profiles in Fig. 12(b). Although the typical flow en- Fig. 12. Fig. 13 depicts the full-field endwall Nu distributions at

18
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

Fig. 16. Variations in (a) Nu/Nu∞ with f/f∞ and (b) the TPF with Re for the compared pin-fin channels.

Re = (a) 50 0 0, (b) 750 0, (c) 10,0 0 0, (d) 12,50 0, and (e) 15,0 0 0. As compared with a smooth-tube channel. The Nu levels of the ellip-
illustrated in each plot of Fig. 13, the local Nu and area-averaged tical pin-fin array with a pin-pitch ratio of 2 [34] are slightly lower
Nusselt number (Nu) increase as Re increases from 50 0 0 to 15,0 0 0. than those measured for the present pin-fin channel. By normaliz-
In addition to such HTE, an increase in the Re value enhances the ing the Nu data presented in Fig. 14(a) with Nu∞ , the variations
degree of shrinkage and skewness of the low-Nu zone toward the in the HTE effectiveness can be determined by plotting Nu/Nu∞
upper-right corner of the channel endwall. against Re for the various pin-fin channels [7,14,34–37]. A compar-
The variation in the endwall average Nusselt number (Nu) with ison of the HTE effectiveness of the various pin-fin channels is pre-
the Reynolds number is displayed in Fig. 14(a). The aforemen- sented in Fig. 14(b). All the Nu/Nu∞ ratios reported in [7, 14, 34-
tioned figure also presents a comparison of the Nu values mea- 37] and the present study decrease as Re increases. As displayed in
sured for the pin-fin channels with various geometric configura- Fig. 14(b), the Nu/Nu∞ ratios for the channels with circular, cubic
tions [14,34–37]. For each type of pin-fin array, the Nu values in- and diamond-shaped pin fins are in the range of 3.2–1.3 when the
crease with an increase in the Reynolds number in accordance Re value is between 50 0 0 and 20,0 0 0 [7, 34-37]. While the Nu/Nu∞
with the power law. Relative to the Dittus-Boelter correlation level ratios for the pin-fin channel with the elliptical pin-fins [14] are
indicated in Fig. 14(a), the channel exhibits a considerable HTE

19
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

approximately 4.5–4.0 for Re values between 50 0 0 and 15,0 0 0. For presented in Fig. 16(b). In general, the thermal performance fac-
the same Re range, the Nu/Nu∞ ratios of the present channel are tors of the pin-fin channels with dense arrays [34-37] are less than
5.26–4.73. The Re-driven decay rate of the aforementioned ratio is unity due to the large pressure drops associated with them. For
lower for the present channel than it is for the channel with ellip- channels with HPFs and CPFs [19], the TPF exceeds unity when Re
tical pin films. The results displayed in Fig. 14 verify that a favor- < 6400. Above Re values of 6400, the TPF becomes less than unity
able HTE property is achieved with the twisted-tape pin-fin array. due to the high f/f∞ ratios. Although the Sx /dp and Sy /dp values
Because forced heat convection vanishes when Re = 0, the Nu re- (both values are 2.5) of the twisted-tape pin-fin channel are similar
sults displayed in Fig. 14(a) can be correlated with the power law to those of the pin-fin arrays reported in [36,37], the present chan-
equation for relevant applications as follows: nel has higher Nu/Nu∞ ratios; thus, the TPF of the present channel
for Re values between 50 0 0 and 15,0 0 0 is 1.55–1.39. The lower f/f∞
Nu = 0.233Re0.707 (19)
values with the corresponding Nu/Nu∞ ratios between 3.13 and 2.4
for the pin-fin channels presented by Chyu et al. [36] result in the
4.3. Pressure drop and thermal performance higher TPF values in the range of 1.54-1.77 at Re values between
12,0 0 0 and 19,0 0 0 [Fig. 16(b)].
The increases in f associated with the HTE displayed in
Fig. 14 are comparatively examined for the pin-fin channels pre-
sented in [7,14,34–37] and the test channel proposed in this study. 5. Conclusions
Fig. 15 depicts the variations in f and f/f∞ with Re [Fig. 15(a) and
15(b), respectively] for these channels. The friction factors eval- In this study, a twisted-tape pin-fin array was proposed for the
uated from the Blasius equation are also presented in Fig. 15(a) HTE of a channel flow. The turbulent flow field of the present
to highlight that the presence of various types of pin-fin ar- channel was numerically analyzed to determine the flow mech-
rays caused a considerable pressure drop in the aforementioned anisms influencing the measured endwall heat transfer property,
channels compared with the pressure in the smooth reference Fanning friction factor, and thermal performance factor at Re val-
tube. ues between 50 0 0 and 15,0 0 0. On the basis of the flow simula-
As displayed in Figs. 15(a) and 15(b), the f data and the f/f∞ tions and experimental heat transfer results, the following conclu-
ratios reported in [7, 14, 34-37] have three ranges due to the dif- sion was drawn:
ferent pin-fin pitches and shapes. All the f values presented in
Fig. 15(a) decrease with an increase in the Reynolds number. Dense 1 The air stream tripped along the curly surface of each twisted-
pin fins with a pin pitch to pin diameter ratio of 2 [14, 35] gen- tape pin fin induced multiple vortices in the pin-fin array to
erally have higher f values and f/f∞ ratios than do those with a generate upward and downward flows normal to the channel
ratio of 2.5 [36]. Elliptical pin fins with a pin pitch to pin diam- endwall for promoting the core-to-wall fluid mixing, which led
eter ratio of 2.17-2.15 [34] have significantly higher f values and to higher HTE than that achieved with conventional pin fins.
f/f∞ ratios compared with the other two data clusters, which en- Because of the identical twist orientation for all the twisted-
ables the high HTE effectiveness of the elliptical fins to be offset tape pin fins, the vortical interactions at downstream locations
(Fig. 14). The f values and f/f∞ ratios of the twisted-tape pin-fin resulted in the generation of a strong near-wall cross flow along
channel with Sx /dp = Sy /dp = 2.5 are similar to those of the dense the front and back endwalls. This cross flow increased the heat
pin-fin arrays [14,35,37], as displayed in Fig. 15. The f/f∞ ratios for transfer rates but caused spanwise Nu variations. Shear layers
the present pin-fin channel vary from 39.4 to 38.5 when the Re were formed between the boundaries of multiple flow cells and
value changes from 50 0 0 to 15,0 0 0, as illustrated in Fig. 15(b). The amplified the vorticity and turbulent intensity in the wake re-
format of the Blasius equation is used to formulate Eq. (20) to cal- gion behind each pin fin. Thus, in the present channel, HTE was
culate the f factor for correlating the f with the Reynolds number accompanied with increased frictional and turbulent drags.
data of the twisted-tape pin-fin channel. 2 On the scanned channel endwall, high-Nu zones emerged along
the y/W = 0 sidewall and channel entry region. Downstream
f = 2.498Re−0.227 (20)
of the third pin row, the increase in the y-wise Tw from chan-
The relationships among Nu/Nu∞ , f/f∞ , and Re for the compared nel sidewall at y/W = 0 to the upper-right endwall corner gen-
pin-fin channels are displayed in Fig. 16(a). The thermal perfor- erated a skewed low-Nu zone along the y/W =1 sidewall. Lo-
mance improvement associated with HTE is indicated by the shift cal HTE through impinging flow resulted in the formation of a
of the data cluster toward the upper-left corner in a Nu/Nu∞ versus semiring-shaped high-Nu zone at the upwind face of each pin-
f/f∞ plot. In a Nu/Nu∞ versus f/f∞ plot, the steeper the data string, fin cap on the channel endwall. Local Nusselt numbers in the
the more sensitive the Nu/Nu∞ ratio is to the f/f∞ variation. The wake region behind each pin decayed axially and then recov-
Nu/Nu∞ ratios decrease as Re increases for all the compared pin-fin ered to reach the local Nu peak at the stagnation point on the
channels [Fig. 16(a)]. However, f/f∞ can increase or decrease with leading edge of the successive pin. Considerable local Nu eleva-
an increase in Re; thus, two downward orientations are observed tions were observed along the two channel sidewalls because
for the data spreads in Fig. 16(a). Due to the steeper Nu/Nu∞ ver- of the strong near-wall flows formed.
sus f/f∞ increase for the present twisted-tape pin-fin channel with 3 In addition to an increase in Nu, an increase in Re caused the
the Nu/Nu∞ variation from 5.6 to 3.85 in the f/f∞ range of 39.4 to low-Nu zone to shrink and skew toward the upper-right corner
38.5, and for the pin-fin channels reported in [36], the HTE ben- on the channel endwall. As Re increased from 50 0 0 to 15,0 0 0,
efits for these two types of pin-fin channels are more sensitive the Nu/Nu∞ ratio systematically decreased from 5.26 to 4.73,
to the f augmentations. Regardless of the geometric characteristics which was the highest Nu/Nu∞ range among the compared pin-
of a pin-fin array, the data spread for each pin-fin channel moves fin channels (Fig. 14).
away from the upper-left corner of Fig. 16(a) with an increase in 4 For the twisted-tape pin-fin channel with Sx /dp = Sy /dp = 2.5,
the Re value. This result indicates that the thermal performance the f value decreased when the Re increased. The f/f∞ ratio of
of all the compared channels deteriorates with an increase in the the present channel decreased from 39.4 to 38.5 when Re in-
higher Reynolds number. creased from 50 0 0 to 15,0 0 0.
The thermal performance factors evaluated at constant pump- 5 Because of its high HTE effectiveness, the TPF of the present
ing powers according to the suggestions of Gee and Webb [31] are channel at Re values between 50 0 0 and 15,0 0 0 was 1.55–1.39.

20
S.W. Chang, P.-S. Wu, W.L. Cai et al. International Journal of Heat and Mass Transfer 166 (2021) 120809

6 Two sets of empirical correlations that permit the evaluations [14] O. Uzol, C. Camci, Heat transfer, pressure loss and flow field measurements
of Nu and f using Re as the determining parameter were de- downstream of staggered two-row circular and elliptical pin-fin arrays, ASME
J. Heat Transfer 127 (2005) 458–471.
vised for relevant applications. [15] F. Wang, J. Zhang, S. Wang, Investigation on flow and heat transfer character-
istics in rectangular channel with drop-shaped pin fins, Propulsion and Power
Disclosure Statement Research 1 (2012) 64–70.
[16] H. Shafeie, O. Abouali, G.Ahmadi K.Jafarpur, Numerical study of heat transfer
performance of single-phase heat sinks with micro pin-fin structures, Applied
A manuscript entitled “Experimental heat transfer and flow Thermal Engineering (2013) 68–76.
simulations of rectangular channel with twisted-tape pin-fin array” [17] J. Pandit, M. Thompson, S.V. Ekkad, S.T. Huxtable, Effect of pin fin to channel
height ratio and pin fin geometry on heat transfer performance for flow in
by S.W. Chang, P.-S. Wu, W.L. Cai and C.H. Yu is submitted to Int. J.
rectangular channels, Int. J. Heat Mass Transfer 77 (2014) 359–368.
Heat Mass Transfer for consideration. [18] M.-A. Moon, K.-Y. Kim, Analysis and optimization of fan-shaped pin–fin in a
There is no actual or potential conflict of interest including any rectangular cooling channel, Int. J. Heat Mass Transfer 72 (2014) 148–162.
[19] S. Caliskan, A. Dogan, I. Kotcioglu, Experimental investigation of heat transfer
financial, personal or other relationships with other people or or-
from different pin fin in a rectangular channel, Experimental Heat Transfer 32
ganizations within three (3) years of beginning the above refer- (2019) 376–392.
enced work submitted that could inappropriately influence (bias) [20] M. Rezaee, M. Khoshvaght-Aliabadi, A.A. AbbasianArani, S.H. Ma-
their work. zloumic, Heat transfer intensification in pin-fin heat sink by changing
pin-length/longitudinal-pitch, Chemical Engineering & Processing Intensifica-
tion 141 (2019) 107544 1-13.
Declaration of Competing Interest [21] G. Liao, X. Wang, J. Li, F. Zhang, A numerical comparison of thermal perfor-
mance of in-line pin–fins in a wedge duct with three kinds of coolant, Int. J.
Heat and Mass Transfer 77 (2014) 1033–1042.
None. [22] W. Duangthongsuk, S. Wongwises, An experimental study on the thermal and
hydraulic performances of nanofluids flow in a miniature circular pin fin heat
Acknowledgement sink, Experimental Thermal and Fluid Science 66 (2015) 28–35.
[23] J.F. Tullius, T.K. Tullius, Y. Bayazitoglu, Optimization of short micro pin fins in
minichannels, Int. J. Heat Mass Transfer 55 (2012) 3921–3932.
The research work was supported by Ministry of Science and [24] R. Roth, G. Lenk, K. Cobry, P. Woias, Heat transfer in freestanding microchan-
Technology, Taiwan, under MOST 108-2221-E-006-090-MY3 and nels with in-line and staggered pin fin structures with clearance, Int. J. Heat
Mass Transfer 67 (2013) 1–15.
MOST 108-2622-8-006-014 projects.
[25] O.O. Adewumi, T. Bello-Ochende, J.P. Meyer, Constructal design of combined
microchannel and micro pin fins for electronic cooling, Int. J. Heat Mass Trans-
References fer 66 (2013) 315–323.
[26] D. Mei, X. Lou, M. Qian, Z. Yao, L. Liang, Z. Chen, Effect of tip clearance on
[1] S.A. Lawson, A. A.Thrift, K. A.Thole, A. Kohlib, Heat transfer from multiple the heat transfer and pressure drop performance in the micro-reactor with
row arrays of low aspect ratio pin fins, Int. J. Heat Mass Transfer 54 (2011) micro-pin–fin arrays at low Reynolds number, Int. J. Heat Mass Transfer 70
4099–4109. (2014) 709–718.
[2] J.K. Ostanek, K.A. Thole, Wake development in staggered short cylinder arrays [27] A. Abdoli, G. Jimenez, G.S. Dulikravich, Thermo-fluid analysis of micro pin-fin
within a channel, Exp. Fluids 53 (2012) 673–697. array cooling configurations for high heat fluxes with a hot spot, Int. J. Thermal
[3] J.K. Ostanek, K.A. Thole, Effect of streamwise spacing on periodic and random Sciences 90 (2015) 290–297.
unsteadiness in a bundle of short cylinders confined in a channel, Exp. Fluids [28] D. Yang, Z. Jin, Y. Wang, G. Ding, G. Wang, Heat removal capacity of laminar
53 (2012) 1779–1796. coolant flow in a micro channel heat sink with different pin fins, Int. J. Heat
[4] F.E. Ames, L.A. Dvorak, Turbulent transport in pin fin arrays: experimental data Mass Transfer 113 (2017) 366–372.
and predictions, ASME Journal of turbomachinery 128 (2006) 71–81. [29] Z. Wan, Y. Joshi, Pressure drop and heat transfer characteristics of pin fin en-
[5] M.K. Chyu, Y.C. Hsing, T.P. Shih, V. Natarajan, Heat transfer contributions of hanced microgaps in single phase microfluidic cooling, Int. J. Heat Mass Trans-
pins and endwall in pin-fin arrays: effects of thermal boundary condition mod- fer 115 (2017) 115–126.
eling, ASME Journal of Turbomachinery 121 (1999) 257–263. [30] T. Ambreen, A. Saleem, C.W. Park, Pin-fin shape-dependent heat transfer and
[6] G.J. VanFossen, Heat-transfer coefficients for staggered arrays of short pin fins, fluid flow characteristics of water- and nanofluid-cooled micropin-fin heat
ASME Journal of Engineering for Power 104 (1982) 268–276. sinks: Square, circular and triangular fin cross-sections, Applied Thermal En-
[7] M. Axtmann, R. Poser, J. von Wolfersdorf, M. Bouchez, Endwall heat transfer gineering 158 (2019) 113781 1-15.
and pressure loss measurements in staggered arrays of adiabatic pin fins, Ap- [31] D.L. Gee, R.L. Webb, Forced convection heat transfer in helically rib-roughened
plied Thermal Engineering 103 (2016) 1048–1056. tubes, Int. J. Heat Mass Transfer 23 (1980) 1127–1135.
[8] S.W. Chang, T.L. Yang, C.C. Huang, K.F. Chiang, Endwall heat transfer and pres- [32] S.J. Kline, F.A. McClintock, Describing uncertainties in single sample experi-
sure drop in rectangular channels with attached and detached circular pin-fin ments, Mechanical Engineering 75 (1953) 3–8.
array, Int. Journal of Heat and Mass Transfer 51 (2008) 5247–5259. [33] Fluent Inc. FLUENT 19.2 Ansys Help – Fluent Theory Guide, 2019.
[9] S.C. Siw, M.K. Chyu, T.I.-P. Shih, M.A. Alvin, Effects of pin detached space on [34] L. Tarchi, B. Facchini, S. Zecchi, Experimental investigation of innovative inter-
heat transfer and pin-fin arrays, ASME J. Heat Transfer 134 (2012) 081902 1-9. nal trailing edge cooling configurations with pentagonal arrangement and el-
[10] K.A. Moores, J. Kim, Y.K. Joshi, Heat transfer and fluid flow in shrouded pin fin liptic pin fin, International Journal of Rotating Machinery 109120 (2008) 1–10.
arrays with and without tip clearance, Int. J. Heat and Mass Transfer 52 (2009) [35] W. Bai, D. Liang, W. Chen, M.K. Chyu, Investigation of ribs disturbed entrance
5978–5989. effect of heat transfer and pressure drop in pin-fin array, Applied Thermal En-
[11] R.S. Jadhav, C. Balaji, Fluid flow and heat transfer characteristics of a vertical gineering 162 (2019) 114214 1-8.
channel with detached pin-fin arrays arranged in staggered manner on two [36] M. K. Chyu, C. H. Yen and S. Siw, Comparison of heat transfer from staggered
opposite endwalls, Int. J. Thermal Sciences 105 (2016) 57–74. pin fin arrays with circular, cubic and diamond shaped elements, GT2007-
[12] R.J. Goldstein, S.B. Chen, Flow and mass transfer performance in short pin-fin 28306, 991-999, ASME Turbo Expo 2007: Power for Land, Sea, and Air, Mon-
channels with different fin shapes, Int. J. Rotating Machinery 4 (1998) 113–128. treal, Canada May 14–17, 2007.
[13] G. Tanda, Heat transfer and pressure drop in a rectangular channel with dia- [37] S.C. Lau, J.C. Han, T. Batten, Heat transfer, pressure drop and mass-flow rate in
mond-shaped elements, Int. J. Heat and Mass Transfer 44 (2001) 3529–3541. pin fin channels with long and short trailing edge ejection holes, ASME Journal
of Turbomachinery 111 (1989) 116–123.

21

You might also like