1-s2.0-S1290072918312419-main
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
International Journal of Thermal Sciences 137 (2019) 276–287
ContentslistsavailableatScienceDirect
InternationalJournalofThermalSciences
journal homepage: www.elsevier.com/locate/ijts
Experimentalandnumericalinvestigationofforcedconvectionheattransfer
inporouslatticestructuresproducedbyselectivelasermelting
T
J.Y.Ho a ,K.C.Leong a,∗ ,T.N.Wong b
a SingaporeCentrefor3DPrinting,SchoolofMechanicalandAerospaceEngineering,NanyangTechnologicalUniversity,50NanyangAvenue,Singapore,639798,Republic
ofSingapore
b SchoolofMechanicalandAerospaceEngineering,NanyangTechnologicalUniversity,50NanyangAvenue,Singapore,639798,RepublicofSingapore
ARTICLEINFO
Keywords:
Single-phase
Forcedconvection
Latticestructures
Porousmedia
Selectivelasermelting
ABSTRACT
Forced convection heattransfer in four structured porousmaterials produced by selectivelasermelting (SLM)
was experimentally and numerically studied. The porous materials are lattice structures consisting of periodic
arrangementsofRhombi-Octetunitcells.Thelatticestructureshavesimilarporosity(ε)butareofdifferentunit
cellsizesof5mm,7mm,10mmand12mm.Thisinvestigationaimstocharacterizeandevaluatethethermohydraulicpropertiesofthisnewclassoflatticestructures.Thehydrodynamicandheattransfercharacteristicsof
the lattices structures such as the permeability (K), inertia coefficient (C E ) and Nusselt number (Nu) were determinedexperimentallyinanairflowchannelinwhichtheReynoldsnumber(Re)canbevariedbetween1300
and7000.Usingalinearheatconductionsetup,thestagnanteffectivethermalconductivities(k eff )ofthevarious
lattice structures were determined and a relationship between k eff and the ligament width (d) of the lattice
structure was obtained. Based on the local thermal non-equilibrium model, numerical simulations were performed
to determine the interfacial heat transfer coefficients (h sf ) of the lattice structures. Our results showed
thatthepressuredrop(ΔP)andNusseltnumber(Nu)ofthelatticestructuresincreasewithdecreasingdandthe
highestNuof906wasobtainedwiththeL1latticestructurewhichhasthesmallestligamentwidth.Thelattice
structures alsoexhibited high k eff values which were up to5.5timeshigher thanthatofthe aluminum foams.
Duetotheorderlyarrangementsofthelatticestructures,theirpermeability-basedfrictionfactors(f⋅Da 1/2 )were
foundtobelowerthanthemetallicfoams.However,theirh sf valueswerealsolower.TheColburnj-factorand
thermalefficiencyindex(η)ofthelatticestructuresweredeterminedandfoundtobehigherthanthoseofthe
commercialmetallicfoamsandconventionalpinfinheatsinks.Insummary,thisinvestigationdemonstratesthe
promisinguseofanewclassoflatticestructures(Rhombi-Octet)forenhancingsingle-phaseforcedconvection
cooling.
1. Introduction
Metallic foams have demonstrated high potential in enhancing
forcedconvectionheattransfer[1,2].Theseopencellfoamshavelarge
surfacearea-to-volumeratiosandhighporosities,arelightandconsist
ofintertwiningporousnetworkswhichinducetortuousfluidpathsand
promotefluidmixing.Thesekeyfeatureshaverenderedthemsuitable
for convectivecoolingapplications suchas in electronicthermalmanagement
devices and compact heat transfer devices in the aerospace,
automobileanddefenseindustries[3,4].
Duetothesereasons,theenhancementofsingle-phaseheattransfer
using metallic foams has been extensively investigated. For instance,
Boomsmaetal.[5]demonstratedthatcompressedmetallicfoamscould
reduce the thermal resistance of a heat exchanger by almost half as
comparedtoplateheatexchangers.BhattacharyaandMahajan[6]investigatednaturalconvectioninaluminummetallicfoamheatsinksfor
application in electronic cooling. It was reported that the metallic
foams with finned structure enhanced heat transfer rates more significantlyascomparedtotheblockmetallicfoamsandtheheattransfer
coefficients were found to increase with increasing fin density. Leong
and Jin [7,8] conducted oscillating and steady flow experiments
through a channel filled with aluminum foams of different pore densities.
Under oscillating flow condition, it was determined that the
average Nusselt number increases with increasing pore density and
kineticReynoldsnumber.Inaddition,thealuminumfoamsunderboth
oscillating and steady flow also exhibited higher heat transfer performance
than conventional finned heat sinks. Hatami and Ganji [9]
theoretically investigated the thermal performance of circular porous
∗ Correspondingauthor.
E-mailaddress:mkcleong@ntu.edu.sg(K.C.Leong).
https://doi.org/10.1016/j.ijthermalsci.2018.11.022
Received13July2018;Receivedinrevisedform9November2018;Accepted20November2018
Available online 02 December 2018
1290-0729/ © 2018 Elsevier Masson SAS. All rights reserved.
J.Y.Hoetal. International Journal of Thermal Sciences 137 (2019) 276–287
Nomenclature
A basearea(m 2 )
A s , t totalheattransferarea(m 2 )
C E inertiacoefficient
d ligamentwidthordiameter(m)
d p porediameter(m)
Da Darcynumber
D h hydraulicdiameter(m)
f modifiedfrictionfactor–Eq.(11)
f 1 frictionfactor–Eq.(30)
H heightoflatticestructure(m)
h¯ averageheattransfercoefficient(W/m 2 ⋅K)
h sf interfacialheattransfercoefficient(W/m 2 ⋅K)
j Colburnj-factor
k bulkthermalconductivity(W/m·K)
k eff stagnanteffectivethermalconductivity(W/m·K)
K permeability(m 2 )
l ligamentlength(m)
L length(m)
Nu NusseltnumberbasedonflowchannelD h
Nu d Nusseltnumberbasedond
q'' heatflux(W/m 2 )
Q heatrate(W)
Re ReynoldsnumberbasedonflowchannelD h
Re d Reynoldsnumberbasedond
St Stantonnumber
T temperature(°C)
U averageairvelocity(m/s)
V volume(m 3 )
GreekSymbols
ε porosity
ρ density(kg/m 3 )
μ dynamicviscosity(kg/m⋅s)
η efficiencyindex
Subscripts
c coldside
h hotside
f fluid
s solid
w wall
in inlet
finswithdifferentsectionshapesandmaterialssubjectedtoconvection
and radiation heat transfer. Using the least square method, the fin
equations for the porous fins were solved to obtain the most exact
analyticalsolution.DogonchiandGanji[10]subsequentlyextendedthe
study to a moving fin where the Differential Transformation Method
(DTM) was employed to solve the fin equation analytically. Their results
show that the fin temperature distribution decreases with increasing
Peclet number and radiation-conduction parameter. CylindricalpipesfilledwithmetalfoamswereanalyticallystudiedbyLuetal.
[11].Bysolvingthelocalthermalnon-equilibriumheattransfermodel
and the Brinkman-extended Darcy momentum model for the porous
media, it was concluded that the effects of foam thermal conductivity
werelesssignificantatlowReynoldsnumbers.Inaddition,forfoamsof
lowthermalconductivities,varyingtheporedensityalsohadlittleinfluence
on the heat transfer performance. These findings suggest that
low thermal conductivity and highly porous foams can be used to reduce
cost and pressure drop. Despite demonstrating improvements in
convective cooling, metallic foams typically suffer from low effective
thermalconductivitiesofbetween2and7W/m·K[12].Inaddition,the
poorinterconnectivityoftheirinternalporestructureshasalsoresulted
inhighresistancetofluidflowwhichlimitstheirpracticalapplications
suchasinplate-finheatexchangers[13].
In the recent years, periodic lattice structures have emerged as
multi-functionalmaterialswiththepotentialforbothloadbearingand
heat transfer. In comparison with the stochastic network of metallic
foams, periodic lattice structures consist of orderly unit cell arrangements.
The unit cells are usually an assembly of cylindrical struts. By
varying the strut arrangements, different unit cell configurations such
asKagomé,honeycombandpyramidalstructurescanbeproduced[14].
Owingtotheuniformityofthelatticestructureunitcellarrangements,
high strength and stiffness at a given weight can be achieved. Furthermore,
the strut diameter and length of the unit cell can also be
adjusted to produce a homogenous three-dimensional network with
enhanced thermal and hydrodynamic performances. Investigations on
thelatticestructures’mechanicalpropertiesofvarioustopologieshave
beenreportedbyLiuetal.[15]andHyunetal.[16].Acomprehensive
review of the thermal and hydrodynamic characteristics of periodic
latticestructuresisprovidedbyLuetal.[17].
The convection heat transfer of air in lattice structures of the tetrahedral
unit cell was investigated by Kim et al. [18]. Due to the
anisotropicityoftheunitcell,thepressuredropsvariedunderdifferent
unit cell orientations. However, heat transfer rates were almost independent
of the unit cell orientation. It was determined that the formation
of horseshoe vortices and flow field-to-lattice structure interactions
at the vertices of the unit cell increased the overall
enhancements in Nusselt number. Subsequently, Tian et al. [19] experimentally
characterized square and diamond-shaped periodic cellularstructures
madefrom wire-screen laminatesand determined that
theoverallthermalefficiencyindices(η)ofthecellularstructureswere
approximately 3 times larger than stochastic copper metallic foams.
Thesehigher ηvaluesofthecellularstructureswerelargelyduetotheir
lowerflowresistances.Theuseofthehollowmicro-latticestructureina
cross-flow heat exchanger was demonstrated by Maloney et al. [20].
The micro-lattice was fabricated by electroplating nickel onto a sacrificial
polymer micro-lattice. Experiments were performed by running
hotwater through theinternalflow channels ofthe micro-latticewith
cold water flowing over its external surfaces. The overall thermal
conductances per unit volume of the heat exchanger were determined
to be between 0.84W/cm 3·K and 1.58W/cm 3·K and a correlation was
developedtopredictthethermalperformancesoftheheatexchangerat
differentlengthscales.Morerecently,Sonetal.[21]conductedasystematic
study on the forced convection heat transfer of air in lattice
structuresofdifferentporosities.Thetetrahedralunitcellwasusedand
the porosity was varied by using struts of different diameters. The
Nusseltnumbersofthelatticestructureswereobservedtoincreasewith
decreasingporosityandthefrictionfactorswerefoundtobeanorderof
magnitude smaller than those of the metallic foams of similar porosities.Itwasalsoconcludedthatthelatticestructureshavesimilarheat
transferperformanceasthemetallicfoamsbutwithsignificantlylower
flowresistances.
Fromtheabovereview,itcanbeseenthatperiodiclatticestructures
have superior thermal and hydraulic characteristics and their performances
are mainly dependent on the solid heat conduction paths and
the fluid-solid interactions. These mechanisms are in turn affected by
the lattice structure parameters such as the unit cell topology, strut
diameter and porosity. However, likely due to the limitation of conventional
manufacturing techniques, existing unit cell designs of the
lattice structures are largely limited to simple geometries such as
square, diamond or pyramidal shapes. To further enhance the lattice
structures thermal and hydraulic performances, new unit cell
277
J.Y.Hoetal. International Journal of Thermal Sciences 137 (2019) 276–287
geometries with high effective thermal conductivities and low flow
resistances can be developed. Lattice structures of different strut diameters
and porosities should also be produced so that parametric studiescanbeconductedtounderstandtheireffectsonthefluidflowfield
andtoobtaintheperformancematricesforcoolingapplications.
These challenges can be overcome by the recent advancements in
additive manufacturingtechniques.Selectivelasermeltingis abranch
ofadditivemanufacturing(AM)whichusesahigh-powerlasertomelt
and fuse the metallic base powder based on a predefined computeraideddesign.Complexthree-dimensionaldesignscanbe
fabricatedby
melting consecutive layers of powder over each other. The main advantage
of SLM over other AM technologies is its ability to produce
metallic components that are more than 99% dense for heat transfer
applications. Recently, SLM has been explored to fabricate novel
structures forsingle-phaseforcedconvection[22,23],nucleateboiling
[24,25] and condensation [26] studies where significant heat transfer
enhancementshavebeenreported.Inaddition,anewtypeofperiodic
lattice structure was also investigated by Ho et al. [27] for enhancing
the thermal and hydraulic performances of water-cooled cold plates.
The lattice structure, produced using SLM, consists of orderly arrangements
of the Rhombi-Octet unit cell and was employed as the
enhanced feature in a cylindrical flow channel. Their results showed
that the lattice structure exhibited up to 383% enhancement in the
averageheattransfercoefficient(h¯)ascomparedtoanemptychannel
and up to 51% higher h ave values as compared to the best performing
coldplatewithtwistedtapeinsert.Furthermore,inarecentstudyusing
waterascoolant,itwasalsodeterminedthatthenewlatticestructure
demonstrated higher h ave values than the conventional metallic foam
whilehavingsimilarpressuredrop[28].
AsacontinuationofapreviousstudybyHoetal.[27],thepresent
investigation focuses on characterizing the forced convection heat
transferandpressuredropacrossthelatticestructuresofRhombi-Octet
unitcellsusingairasthecoolingmedium.TheRhombi-Octetstructure
isselectedforthisinvestigationasitcanachievehighpackingdensity
withitsligamentarrangements.Inaddition,theperiodicarrangements
of the unit cells have shown potential in enhancing forced convection
heat transfer [27]. However, in the previous work [27], studies were
only performed to investigate the potential application of lattice
structuresasinsertsinthewater-cooledcoldplates.Inordertoachieve
abetter understandingoftheforcedconvectionheattransfer mechanisms
of this new type of lattice structure, a fundamental study to determine
its thermo-hydraulic properties such as permeability, inertia
coefficient, effective and interfacial thermal conductivities with differentunitcellsizesneedstobecarriedout.Inthisinvestigation,four
latticeheatsinksmadefromrepetitionsoftheRhombi-Octetunitcellof
the same porosity but of different ligament widths and surface-to-volume
ratios were fabricated. Experiments were conducted in a horizontal
flow channel to determine their permeabilities, inertia coefficientsandNusseltnumbers.Inaddition,theeffectivestagnantthermal
conductivitiesoftheselatticestructureswerealsomeasured.Usingthe
experimental data and numerical simulation, the interfacial heat
transfercoefficientsbetweenthefluidandsolidphaseswereevaluated
and the possible fluid-solid interactions were elucidated. Finally, the
thermal and hydraulic performances of these lattice structures were
compared against other enhanced structures and metallic open cell
foams. To the best of the authors’ knowledge, there is no published
work onthe fundamentalstudies offorcedconvectionheattransfer in
lattice structures of the Rhombi-Octet unit cell. Therefore, it is hoped
that from the present investigation, new insights on the transport mechanisms
can be obtained and the data collected can be used for the
developmentofenhancedcoolingdeviceswiththisnewclassoflattice
structures.
2. Fabricationandcharacterizationofspecimens
The SLM 280 (SLM Solutions GmbH) facility at the Future of
Manufacturing Laboratory 1 of the Singapore Centre for 3D Printing
(SC3DP) in Nanyang Technological University (NTU), Singapore was
employedtofabricatethespecimensusedinthepresentinvestigation.
The machine which consists of a Gaussian distributed Yb:YAG laser
Fig.1.(a)IsometricdrawingsofaRhombi-Octetunitcell,(b)comparisonofdifferentunitcellsizesand(c)projectedviewandvariousporetypesofaunitcell.
278
J.Y.Hoetal. International Journal of Thermal Sciences 137 (2019) 276–287
withmaximumpowerof400Wandlaserbeamspotsizeof80μmwas
utilizedtomeltandfusethebasemetallicpowder.Analuminumalloy
powder,AlSi10Mg,of20μm–63μmsizewasusedasthebasemetallic
powder to produce the test specimens for the experimental investigation.Aluminum
alloywasselectedasit hashigh thermalconductivity
of between 103W/m·K and 190W/m·K and is light as compared to
otheralloys.Thelasermeltingprocesswascarriedoutinthemachine's
buildchamberwhereinertargongaswasfirstusedtoflushthechamber
toattainanoxygenleveloflessthan0.2%soastominimizeoxidation
and combustion of powder. Subsequently, the first layer of AlSi10Mg
metallicpowderwasdistributedevenlyonthebase-platebyarecoater
and the laser beam was directed to melt the powder based on a preprogrammed
model. Upon completion of the laser melting process for
thefirstlayer,thebase-platewasthenloweredbyone-layerthicknessof
50μm and the process was repeated until the parts are fully constructed.Inthepresentinvestigation,alaserpowerof350W,scanning
speed of 1150mm/s and hatching spacing, which is the spacing between
adjacent laser scanning tracks, of 0.17mm were used in the
fabricationbasedonpriorsuccessesofbuildingdensebulkobjects.
Four lattice heat sinks (L1 – L4) with the bulk dimensions of
40mm×90mm×15mmweredesignedandfabricatedfortheforced
convection heat transfer experiments. Each lattice structure is integratedontoabaseplateofdimensions40mm×90mm×3.5mmas
a single built piece and consists of periodic arrangements of the
Rhombi-Octetunitcells.ThedrawingofaunitcellisshowninFig.1(a)
andacomparisonofvariousunitcellsizesisshowninFig.1(b).Each
unitcellconsistsof18squareand24triangularfacesandexamplesof
the square and triangular faces are highlighted in blue and red in
Fig. 1(a), respectively. Each square is positioned at 45° to its adjacent
squares to form a Rhombicuboctahedron core whereasat each corner,
the triangles are arranged to form an Octet truss-like structure. A detaileddescriptionoftheunitcellgeometrycanbefoundinRef.[27].In
this study, the lattice heat sinks were made of repetitions of 5mm,
7mm, 10mm and 12mm unit cells. All the unit cells have the same
porosity(ε)of0.85butduetothedifferenceintheirunitcellsize,their
ligament length (l), ligament width (d), average pore diameter (d p,ave )
and surface area-to-volume ratio (A st /V) are different. The ligament
length(l)representsthestrutlengthonthesquarefaceoftheunitcell
andtheligamentwidth(d)isthethicknessofeachstrut.Repetitionsof
unit cells were arranged adjacent to each other to obtain the bulk dimensions
of 40mm×90mm×15mm of the heat sinks. However, in
somecases,theunitcellscouldnotbefittedintothebulkdimensionsof
the heat sinkand this resulted inthe truncation ofthese unit cells.As
theportionofunitcellbeingtruncateddiffersfordifferentunitcellsize,
this resulted the difference in porosity among the different heatsinks,
asshowninTable1.TheprojectedviewoftheRhombi-Octetunitcellis
showninFig.1(c).Itcanbeseenthateachunitcellconsistsofsixtypes
ofpores(d p,1 –d p,6 ).Inordertocomputed p,ave ,thehydraulicdiameter
of each pore was determined and d p,ave was computed by taking the
area-weighted average of all the pores using Eq. (1). The geometrical
parametersofalltheunitcellsaresummarizedinTable1.Imagesofthe
fourlatticeheatsinksareshowninFig.2(a)–(d).Thelatticeheatsinks
are named L1, L2, L3 and L4 and they are made of repetitions of
Rhombi-Octet unit cells of sizes 5mm, 7mm, 10mm and 12mm, respectively.
In addition to the lattice heat sinks, a heat sink with cylindricalpinfinarray(Fig.3(a))andaheatsinkwithparabolicpinfin
array(Fig.3(b))werealsofabricatedforcomparison.Thespecimensof
cylindricalpinfinandparabolicpinfinarraysarenamedasC1andP1,
respectively. The pin fins are in staggered arrangements. The cylindricalpinfinshaveadiameterof3.24mm,heightof15mmandlateral
and longitudinal fin-to-fin spacings of 4mm and 10mm, respectively.
On the other hand, the parabolic pin fins have a base diameter of
5.65mmwhichreduces to0.9mmatthefintip.Similartothecylindrical
pin fin array, the height of the parabolic fins is 15mm while its
lateral and longitudinal fin-to-fin spacings are 4mm and 10mm, respectively.
d
p,
ave
=
n d
n A
A
i p,
i i
i
i
The forced convection heat transfer performances of the lattice
structuresaresignificantlyaffectedbytheirstagnanteffectivethermal
conductivities (k eff ). Therefore, the k eff values of the lattice structures
were determined experimentally. For the thermal conductivity measurements,
another four specimens with the same lattice structures as
L1,L2,L3andL4werefabricated.ThesefourspecimensarenamedLc1,
Lc2, Lc3 and Lc4 and have unit cell sizes of 5mm, 7mm, 10mm and
12mm,respectively.ImagesofthespecimensareshowninFig.4.They
have bulk dimensions of 33.25mm×34.55mm×38mm. Each specimenconsistsofatopfacesheetof9mmheight,abottomfacesheetof
14mm height and a lattice structure of 15mm height. The lattice
structureissandwichedbetweenthetopandbottomfacesheets.Asthe
facesheetsandlatticestructurewerefabricatedasonebuiltpiece,there
is no thermal contact resistance at the face sheet and lattice structure
interface.Finally,asolidAl6061-T6blockandasolidAlSi10Mgblock
of the same bulk dimensions (33.25mm×34.55mm×38mm) were
fabricated and their thermal conductivities were also determined for
comparison.
3. Experimentalsetupandprocedures
3.1. Thermalconductivitytestfacility
Fig. 5 shows a schematic of the experimental facility used to determinek
eff ofthelatticestructures.Themeasurementswereperformed
on specimens Lc1, Lc2, Lc3 and Lc4. A copper block fitted with three
cartridge heaters is located above the top face sheet of the test specimen.
The cartridge heaters are connected to a variable transformer
whichallowstheheatratefromtheheaterstobecontrolled.Usingan
ammeter and a voltmeter, the current (I) and voltage input (V) to the
heatersweredeterminedandtheoutputheatrates(Q)fromtheheaters
were computed. A heat sink is located below the bottom face sheet
wheretheheatisdissipated.Tapwaterwasusedasthecoolingmedium
andwasallowedtorunthroughtheheatsinkflowchannels.Tominimizeheatlosses,Teflonisusedastheinsulatingmaterialanditencloses
thecopper,heatersandthetestspecimen.Alayerofelastomericfoamis
usedasasecondlayerofinsulationandisappliedontheoutersurface
oftheTeflon.AsdepictedinFig.6(a),eightK-typethermocouples(T h,1
–T h,8 )arefittedintothetopfacesheetofthetestspecimenat6.5mm
abovethetopfacesheetandlatticestructureinterface(Δy 1 ).Asshown
inFig.6(b),fourK-typethermocouples(T c,1 –T c,4 )areinstalledatthe
bottom face sheet at 4mm below the bottom face sheet and lattice
structureinterface(Δy 2 ).Inordertodeterminek eff ofthevariouslattice
structures,theone-dimensionalFourier'slawasshowninEq.(2a)was
firstusedtocomputethetemperatureatthetopfacesheetandlattice
structure interface (T top ), where T h is the average of T h,1 – T h,8 and
k AlSi10Mg isthethermalconductivityofbulkAlSi10Mg.Usingthesame
approach,thetemperatureatthebottomfacesheetandlatticestructure
interface(T bottom )wasobtainedusingEq.(2b),whereT c istheaverage
of T c,1 – T c,4 . Finally, with Eq. (2c), k eff of the lattice structure can be
determined. Throughout the experiments, six additional K-type thermocouples
were installed on the surface of the elastomeric foam to
estimate the heat loss to the surroundings. Based on the temperatures
recorded from these thermocouples and the correlations for natural
Table1
Geometricalparametersofthelatticestructureheatsinks.
Unitcellsize(mm) ε d(mm) l(mm) d/l d p,ave (mm) A st /V(m −1 )
5 0.8402 0.42 1.88 0.22 1.00 1355
7 0.8428 0.59 2.62 0.22 1.22 968
10 0.8458 0.84 3.76 0.22 1.54 678
12 0.8538 0.99 4.49 0.22 1.75 565
(1)
279
J.Y.Hoetal. International Journal of Thermal Sciences 137 (2019) 276–287
Fig.2.Imagesoflatticeheatsinks(a)L1,(b)L2,(c)L3and(d)L4usedforforcedconvectionheattransferinvestigation.
Fig.3.Imagesof(a)cylindricalpinfinheatsink–C1and(b)parabolicpinfin
heatsink–P1.
convectionprovidedinRef.[29],therateofheatlosswasdetermined
tobelessthan2.7%.UsingthemethodbyMoffat[30],theuncertainty
of k eff can be computed by Eq. (3). In this experimental setup, the accuracies
of the ammeter and voltmeter are 0.01 A and 1.0V, respectively
whereas the thermocouples were calibrated to an accuracy
of±0.5°C. Using Eq. (3), the maximum measurement uncertainty of
k eff wasfoundtobe±3.5%.
Priortothestartoftheexperimentswiththelatticestructures,the
accuracyof thesetup wasvalidatedusingasolidaluminum (Al6061-
T6)block.Fromtheheatratesoftheheaters(Q),cross-sectionalareaof
the aluminum block (A) and the temperature gradient ( dT
dy ) computed
fromthethermocouplemeasurements,thethermalconductivityofbulk
Al-6061(k Al )wasfoundtobe164W/m⋅Kwhichisclosetothevalueof
167W/m⋅K reported in Ref. [31], indicating good accuracy of the experimental
facility. Using the same procedures, the thermal conductivityofthebulkAlSi10Mg(k
AlSi10Mg )wasdeterminedtobe153W/
m⋅K.
Fig. 5.Schematic of experimental facility for thermal conductivity measurements.
T
top
=
8
T
i=
1 h, i ( Q/ A)
y1
8
k
AlSi10 Mg
(2a)
T
k
bottom
eff
4
T
i=
1 c, i ( Q/ A)
y2
= +
4 kAlSi10 Mg
(2b)
( Q/ A)
=
( T T )/ y
top bottom (2c)
dk
2 2 2 2 2 2
eff dI dV dA d y
dTtop
dTbottom
= + + + + +
keff
I V A y Ttop
Tbottom
Ttop
Tbottom
(3)
Fig.6.(a)Topview oftopfacesheetthermocouple positions,and (b)bottom
viewofbottomfacesheetthermocouplepositions.
3.2. Forcedconvectionheattransfertestfacility
ThetestfacilityshownschematicallyinFig.7wasemployedforthe
forced convection heat transfer investigation. This setup is similar to
thatofLeongetal.[32]andconsistsofaTeflonairflowchannelandan
Fig.4.Specimensfork eff measurements(a)Lc1–5mmunitcell,(b)Lc2–7mmunitcell,(c)Lc3–10mmunitcell,and(d)Lc4–12mmunitcell.
280
J.Y.Hoetal. International Journal of Thermal Sciences 137 (2019) 276–287
acrylic cover. Air supplied by an auto-balance compressor entered the
inlet contraction section of 9:1 contraction ratio. Subsequently, it was
channeled into a straight section where the various lattice structures
and pin fin heat sinks were placed. Two wire meshes of 0.32mm
aperture size which served as the air flowstraighteners were installed
alongthestraightsectionoftheflowchannelandupstreamofthetest
specimens.Ananemometerwaspositionedbetweenthewiremeshesto
measure the average air flow velocity (U) and a K-type thermocouple
was placed after the second wire mesh to measure the inlet air temperature
(T in ). A layer of elastomeric foam, which served as an additionallayerofinsulation,wasadheredontotheexternalsurfacesofthe
Teflonflowchanneltominimizeheatlosstothesurrounding.Thetest
sectionshowninFig.8(a)consistsofacopperblockandfourcartridge
heaters.Thecopperblockhasdimensionsof90mm×40mm×6mm
and each cartridge heater has a maximum heat rate of 60W and a
diameter of 6.35mm. The cartridge heaters were connected to a variable
transformer which enabled the output heat rate (Q) from the
heaters to be varied. Prior to the conduct of the experiments, the test
specimens were adhered onto the copper block using thermally conductiveepoxy.AsshowninFig.8(b),nineK-typethermocouples(T
w,1
–T w,9 )wereinsertedintothebaseplateoftestspecimenstoobtainthe
wall temperatures. Each thermocouple was positioned at 10mm apart
andwaslocatedat1.5mmbelowthebaseplatetopsurface.UsingEqs.
(4)–(7),theaverageheattransfercoefficient(h¯),Nusseltnumber(Nu)
andReynoldsnumber(Re)canbedetermined.Itshouldbenotedthat
inEq.(5),AdenotestheprojectedareaofthebaseplateandinEqs.(6)
and (7), D h is the hydraulic diameter of the flow channel. An inclined
manometer, connected to two pressure taps installed directly before
and after the test specimens, was used to measure the pressure drop
(ΔP) across the test specimens at different air velocities. Finally, eight
K-type thermocouples were also placed on the surface of the elastomericfoam.Usingthetemperaturesrecordedfromthesethermocouples
andthecorrelationsfornaturalconvectionofRef.[29],therateofheat
losstothesurroundingswasestimatedtobelessthan1.3%.Usingthe
methodbyMoffat[30],theuncertaintyofNucanbecomputedbyEq.
(8). The maximum measurement uncertainty of Nu was found to
be±5.6%.
T
w
=
9
T
i=
1 w,
i
9
h¯
Q/
A
=
( T T )
w in (5)
= hD ¯
h
Nu
kf (6)
(4)
f
Re = UD h
µ
f (7)
dNu
dI dV dA dTw
= + + + +
Nu I V A T T
4. Resultsanddiscussions
4.1. Pressuredrop
2 2 2 2 2
w
in
dTin
T T
Fig. 9showsthe measuredpressure dropper unitlength (ΔP/L) of
the lattice structures at different air velocities (U). It can be seen that
ΔP/L decreases with increasing unit cell size or increasing ligament
width(d)ofthelatticestructure.Forexample,atU=3.4m/s,ΔP/Lof
the L1 lattice structure is 29.3kPa/m whereas those of the L2 and L3
lattice structures are 21.8kPa/m and 13.4kPa/m, respectively. This
comparison shows that the increase in d from L1 to L2 resulted in a
reductioninΔP/Lof34.4%whereasincreasingdfromL2toL3resulted
inafurtherreductioninΔP/Lof62.7%.However,furtherincreaseofd
fromL3toL4resultedinalesssignificantreductioninΔP/L.Ascanbe
seenfromTable1,theincreaseindalsoresultedintheincreaseinthe
average pore diameter (d p,ave ). Due to a larger d p,ave , the resistance to
thefluidflowpathreducesandthereforeresultedinlowerΔP/Lvalues.
ThisexplainsthesignificantreductioninΔP/LoftheL2andL3lattice
structuresascomparedtoL1.However,theincreaseofd p,ave fromL3to
L4producedalesssignificantreductioninΔP/L.Thisislikelyduetothe
largerligamentwidth(d)ofL4eventhoughitsd p,ave valueislargerthan
L3. The large d value of L4 may have resulted in a higher form drag
whichproducedasmallerreductioninΔP/L.
The pressure drop across a porous medium under non-Darcy flow
conditionscanbeexpressedasaquadraticfunctionofU.Alsoknownas
theForchheimer-extendedDarcyequation,theexpressionshownasEq.
(9)relatesΔPtothepermeability(K)andinertiacoefficient(C E )ofthe
porous medium. While K depicts the ability of a porous medium to
transmit fluid, C E accounts for the effects of form drag and is strongly
dependent on the unit cell design and arrangement of the porous
medium. By employing the porous medium analogy on the lattice
structures,curvefittingoftheresultsofFig.9intheformofquadratic
functions was performed. By comparing the coefficients of the curvefittedresultswithEq.(9),KandC
E ofthevariouslatticestructureswere
determined and shown in Table 2. Using the Darcy number (Da) expressionofEq.(10),Da
1/2 wasalsocomputedandshowninTable2.As
expected,theKvaluesincreasewithincreasingligamentwidth(d)and
increasing average pore diameter (d p,ave ). For the C E values, the reductioninC
E wasobservedastheunitcellsizeincreases,i.e.,fromL1to
w
in
(8)
Fig.7.Schematicoftestfacilityforforcedconvectionheattransferinvestigation.
281
J.Y.Hoetal. International Journal of Thermal Sciences 137 (2019) 276–287
Fig.8.(a)Schematicoftestsection,and(b)positionsofthermocouples(T w, 1 –T w, 9 )inthebaseplateofaheatsink.
f Da
1/2
=
P
L
f
K
U
2
(12)
f Da
1/2
1
= + CE
ReDa1/2 (13)
4.2. Effectivethermalconductivity
Fig.9.Comparisonofpressuredropperunitlength(ΔP/L)ofvariousRhombi-
OctetlatticestructuresL1–L4.
L3.However,furtherincreaseintheunitcellsizefromL3toL4resulted
in a slight increase in the C E value. As explained above, this larger C E
value of L4 has resulted in a higher form drag as compared to L3 and
couldbeduetothelargerligamentwidthoftheL4structure.
The modified friction factor (f) which relates the pressure drop
across the lattice structures and the flow channel hydraulic diameter
(D h ) can be expressed as Eq. (11). For comparison against different
types of porous media, permeability (K) is commonly used as the
characteristiclengthandthiscanbeachievedbyusingtheproductoff
andDa 1/2 asshowninEq.(12).BymultiplyingD h /ρ f·U 2 onbothsidesof
Eq.(9),anexpressionwhichrelatesf⋅Da 1/2 ,Re⋅Da 1/2 andC E asshown
inEq.(13)canbederived.UsingtheexperimentallyobtainedK,C E and
Da 1/2 valuesofthevariouslatticestructures,asshowninTable2,the
f⋅Da 1/2 values at different Re⋅Da 1/2 were computed and presented in
Fig.10.Fornickel(Ni)foams,BeaversandSparrow[33]showedthata
C E valueof0.74providedagoodapproximationwhereasforaluminum
(Al)foams,Kim etal.[34]proposedaC E valueof0.1.ByusingtheK
values of 0.352×10 −7 m 2 and 0.51×10 −7 m 2 for one of the foam
configurations reported in Refs. [33,34], respectively and their corresponding
C E values, the f⋅Da 1/2 values for the Ni and Al foams are
plottedinFig.10.Likelyduetotheorderlyunitcellarrangementofthe
lattice structures, lower C E values were achieved with the lattice
structures as compared to the Ni and Al foams. This resulted in lower
frictionfactorsasshowninFig.10.
µ f f E
2
P
= +
L K U C
K U
Da = K Dh 2 (10)
f =
P
L
Dh
U
f
2
(9)
(11)
Using the experimental facility depicted in Fig. 5, the effective
thermal conductivities (k eff ) of the lattice structures (Lc1 – Lc4) were
measured where air was the fluid medium. The results are shown in
Fig.11.Amongthefourspecimens,Lc1hasthesmallestRhombi-Octet
unitcell of 5mmandligament width (d) of0.42mmwhereasLc4has
thelargestunitcellof12mmanddof0.99mm.Itcanbeseenthatk eff
decreaseswithincreasingunitcellsizeandincreasingdwithLc1having
thehighestk eff valueof22.8W/m 2·K.FromFig.1(a),itcanbeseenthat
all the unit cells have the same internal structure design. In order to
achieve the required unit cell size, each unit cell is scaled proportionally
in three directions. However, in order to fit the lattice structures
betweenthetopandbottomfacesheetsasshowninFig.4,theunitcell
hastobetruncated.Forinstance,theverticaldistancebetweenthetop
andbottomfacesheetsis15mm.Thisenablesthree5mmunitcellsto
bestackedverticallytoformtheLc1specimen.However,onlyoneand
ahalf10mmunitcellscanbefittedverticallybetweenthefacesheets.
Thetruncationoftheunitcellstructureshasresultedinthedifference
inporositiesofthespecimensandcouldhavecausedthedifferencesin
thek eff values.
A comparison of k eff of the Rhombi-Octet structures (Lc1 – Lc4) in
this study and the aluminum foams reported in Ref. [35] is shown in
Fig. 12, where the ligament width/diameter (d) is selected as the
characteristic length. Calmidi and Mahajan [35] presented a comprehensivestudytodeterminek
eff ofthealuminumfoamswheretheporosities
(ε) of the foams range from 0.9005 to 0.9726. This range of
aluminum foam porosities is commonly reported [36–38] and to the
best of the authors’ knowledge, there is no known reported works on
aluminumfoamswiththesameporosityastheRhombi-Octetstructure
used in forced convection heat transfer investigation. From Fig. 12, it
can be seen that for the same d value, the k eff values of the Rhombi-
Octet structures are significantly higher than those of the aluminum
foams. For instance, for d in the range of 0.55–0.6mm, the Rhombi-
Octet structure (Lc2) has a k eff value of 19W/m·K whereas the aluminum
foamhas ak eff valueof 6.7W/m·K. On the other hand,at dof
0.4–0.42mm,thek eff valueoftheRhombi-Octetstructure(Lc1)is5.5
times that of the aluminum foam. The higher k eff ofthe Rhombi-Octet
structure is likely due to its higher packing density, which resulted in
Table2
K,C E andDa 1/2 ofthevariouslatticestructures.
Specimenname L1 L2 L3 L4
K(10 −10 m 2 ) 34.0 47.6 75.5 82.2
C E 0.0446 0.0381 0.0302 0.0320
Da 1/2 0.00267 0.00316 0.00398 0.00415
282
J.Y.Hoetal. International Journal of Thermal Sciences 137 (2019) 276–287
Fig. 10.Comparison of f ⋅Da 1/2 of various Rhombi-Octet lattice structures
againstthenickelandaluminumfoams.
Fig.11.Effectivethermalconductivity(k eff )ofvariouslatticestructures.
Fig.12.Comparisonofk eff ofRhombi-Octetlatticestructuresagainstaluminum
foams.
lower porosity as compared to the aluminum foams. In addition, the
geometry of the unit cell which forms the porous matrix and their orderlyarrangementmayhavealsoresultedinmoreeffectiveconduction
pathsleadingtothehighk eff valuesofLc1–Lc4.
4.3. Forcedconvectionheattransferperformance
The forced convection heat transfer performance of the various
latticeheatsinks(L1–L4)asshowninFig.2wereinvestigatedinthe
airflowchannelofFig.7.TheexperimentalresultsareshowninFig.13
where Nusselt numbers (Nu) are plotted against the Reynolds number
(Re).AsshowninEqs.(6)and(7),NuandRewerecomputedbasedon
Fig.13.ExperimentalNuoftheRhombi-OctetlatticestructuresatdifferentRe.
the flow channel hydraulic diameter (D h ) and the experiments were
performed for Re values between 1300 and 7000. It can be seen from
Fig.13thatforeachlatticeheatsink,NuincreaseswithincreasingRe.
Inaddition,anincreaseinNuwithreducingligamentwidth(d)canalso
beobservedwherethelatticeheatsinkwiththesmallestd(L1)exhibits
thehighestNu.Forinstance,forRebetween5000and5500,L1hasa
Nuvalueof814andisapproximately38%higherthanthatofL4.Heat
transferfromtheheatedbaseisdominatedbyconductionthroughthe
latticestructurematrixandbyconvectionfromthesolidtothefluidat
theinterfaces.AsshowninFig.11,L1hasthehighestk eff ascompared
to other lattice heat sinks. This significantly improves the removal of
heatfromthebaseoftheheatsinkbyconduction.Inaddition,asshown
inTable1,theL1structureismadeofthe5mmunitcellwhichhasthe
largest surface area-to-volume ratio (A s,t /V), this also increases the
available area in contact with the fluid and enhances heat transfer by
convection.
From the above analysis, it can be observed that the average convectionheattransfercoefficient(h¯)ofthelatticestructuresareaffected
byA s,t /Vandk eff .Inaddition,inthesubsequentsection(Section4.4),it
isalsoshownthattheinterfacialheattransfercoefficient(h sf )between
the fluid and solid phases is a significant factor affecting Nu. For a
porousstructurewherethethermalconductivityofthesolidmaterialis
high,itisexpectedthatthetemperaturesoftheligamentsarecloseto
the base temperature where heat is applied [34]. Under such circumstances,
the thermal boundary layers that formed over the ligaments
and the total heat transfer area have significant influences on the
thermalperformanceoftheporousmatrix.Thetotalheattransferarea
representedbytheA s,t /Vratioisrelatedtothesizeoftheunitcelland
fromTable1,itcanbeobservedthatA s,t /Visinverselyproportionalto
d.Ontheotherhand,thecharacteristicsofthethermalboundarylayer
thatdevelopedovereachligamentdependsonthesizeoftheligament
and the distance between the adjacent ligaments (or the pore size).
Sincethefactorsaffectingtheconvectionheattransferperformancesof
the lattice structures are shown to be directly affected by d, it is
therefore appropriate to obtain a relationship between h¯ and d for all
the lattice structures investigated. On this basis, the Nusselt number
andReynoldsnumberbasedontheligamentwidth,viz.,Nu d andRe d ,
can be defined as Eqs. (14) and (15), respectively. Using the experimental
data obtained, Nu d and Re d for L1 – L4 were computed and
plotted in Fig. 14. From the figure, it can be seen that the datapoints
collapsedontoasinglelinewhichcanbedescribedbyapowerfunction.
Usingthenon-linearregressionmethod,acorrelationbetweenNu d ,Re d
andPrforthelatticestructuresisderived.Thecorrelationisshownin
Eq. (16) and plotted in Fig. 14. It can be seen that the correlation
providedagoodpredictionoftheexperimentalresultswith90%ofthe
data within±3.5% deviation from the predicted values. A maximum
deviationof9%existsatthehighestRe d valueof313.
283
J.Y.Hoetal. International Journal of Thermal Sciences 137 (2019) 276–287
c , u· T = ·[ k , · T ] + A h ( T T ) (19)
f p f f f eff f sf sf s f
0 = ·[ ks, eff · Ts ] + Asf hsf ( Tf Ts
)
(20)
kf , eff = kf
(21)
ks, eff = keff kf , eff
(22)
A uniforminlet air velocity profile, zeropressure atthe outlet and
noslipconditionboundaryconditionsatthewallswereprescribedfor
thefluiddomain.Inaddition,thefluidwasassumedtohaveauniform
inlet temperature and adiabatic conditions were imposed on the solid
and fluid domains at y=H. At y=0, where a constant heat flux was
applied, the solid and fluid phases are coupled through the boundary
conditionsofEqs.(23)and(24)[32].
Fig. 14.Plot of Nu d versus Re d of the various Rhombi-Octet lattice structures
andcomparisonagainstcorrelationofEq.(16).
= hd ¯
Nud
kf (14)
f
Re = Ud
d
µ
f (15)
Nud = 0.895Red 0.65 Pr
0.37 (16)
4.4. Interfacialheattransfercoefficient
The interfacial heat transfer coefficient (h sf ) determines the energy
transfer between the solid and fluid phases at the interface. When the
temperature difference between the solid and fluid phases cannot be
neglected, the solid-to-fluid interface significantly affects h sf which in
turninfluencestheconvectionheattransferperformancesofthelattice
structures. In order to determine h sf , a numerical simulation was performed
and the simulation results were compared against the experimental
results of the various lattice structures. The two-dimensional
computational domain shown in Fig. 15 was used for the numerical
simulation.Ithasaheightof15mmandlengthof90mmandissimilar
to the dimensions of the lattice heat sinks (L1 – L4). With the assumptionoflaminarflowandbyusingthevolumeaveragingmethod,
the steady continuity and momentum equations of the fluid phase are
written as Eqs. (17) and (18). Dukhan et al. [39] showed that, for
metallicfoams,turbulentflowisfullydevelopedforRe⋅Da 1/2 > 50.As
showninFig.10,itisreasonabletoassumethattheflowiswithinthe
laminarregimesinceRe⋅Da 1/2 ofthelatticestructuresarebetween3.5
and 30. The permeability (K) and inertia coefficient (C E ) are obtained
fromTable2andporosity(ε)of0.85whichcorrespondstotheporosity
ofaunitcellwasused.Byemployingthelocalthermalnon-equilibrium
(LTNE)model,theenergyequationsforthefluidandsolidphasescan
be derived as Eqs. (19) and (20), respectively. The LTNE model was
selected duetothe largedifference in thermalconductivityvalues between
the solid and fluid phases. In Eq. (19), k f,eff is the effective
stagnant thermal conductivity of the fluid phase and was estimated
usingEq.(21).Fromthemeasuredk eff ofthelatticestructuresshownin
Fig.11,theeffectivethermalconductivityofthesolidphase(k s,eff )can
be computed using Eq. (22). Finally, A sf denotes the interior surface
area-to-volume ratio and the A s,t/ V values in Table 1 were used. It
shouldbenotedthatinthissimulation,theeffectofthermaldispersion
wasnotconsideredasitwasfoundtobenegligiblewhenthedifference
in thermal conductivities between the solid and fluid phases is large
[35].
· u = 0 (17)
µ µ u
( u· ) u = P +
2u
K
f f f f E
2
C u u
K
(18)
q
Tf
= kf eff + k
y
, s,
eff
Ts
y
(23)
Tf = Ts (24)
The simulation was performed using the “Comsol Multiphysics”
software based on the governing equations and boundary conditions
described above. A convergence criterion of 10 −4 was prescribed and
mesh independence checks were conducted where the results were
obtained with approximately 14,000 mesh elements. For each lattice
structure, simulation was performed for the air velocities (U) ranging
from0.5m/sto5m/satintervalsof0.5m/sandwithh sf rangingfrom0
to 300W/m 2·K at intervals of 10W/m 2·K. Thereafter, the Nu values
obtainedfromthesimulationwerecomparedagainsttheexperimental
Nu values and the h sf values that produced the same numerical and
experimentalNuvaluesweredetermined.Thisprocedurewasrepeated
forthedifferentlatticestructuresandairvelocitiesandtheresultsare
plottedinFig.16,whereNu d,sf isdefinedinEq.(25).Theresultsshow
thattheNu d,sf valuesofthelatticesstructuresincreaseswithincreasing
Re d andthesevaluescollapsedontoasinglelinewhichcanbedescribed
byapowerfunction.This,therefore,suggeststhattheh sf isalsostrongly
influencedbytheligamentwidthofthelatticestructure.Byemploying
the non-linear regression method, a correlation which characterizes
Nu d,sf asshowninEq.(26)isderived.
Nu d , sf
= h sf d
k
Nu = 0.227Re Pr
d,
sf
f
0.608 0.37
d
(25)
(26)
A correlation which relates Nu d,sf and Re d in aluminum foams was
proposed by Calmidi and Mahajan [35] as shown in Eq. (27). This
correlation assumes that the fluid flow through a porous matrix is
analogous to crossflow over cylinders and employs the ligament diameter(d)ofthealuminumfoamsasthecharacteristiclengthforNu
d,sf
and Re d . In addition, as the mean fluid velocity through the porous
regionwasconsidered,the ε −0.5 termwasincludedintheequation.For
the purpose of comparison against the Nu d,sf values of the lattice
structuresobtainedfromthisinvestigation,thepropertiesofthe40PPI
aluminumfoamreportedinRef.[35]wereused,where ε=0.9132and
d=0.00025m. The Nu d,sf values were computed using Eqs. (26) and
(27)forRe d between25and300andtheresultsareshowninFig.17.
As shown in this figure, the Nu d,sf values of the aluminum foams are
consistentlyhigherthanthatoftheRhombi-Octetlatticestructures.At
Fig.15.Schematicofcomputationaldomain.
284
J.Y.Hoetal. International Journal of Thermal Sciences 137 (2019) 276–287
lattice structures have higher heat transfer performances as compared
tothealuminumfoams.Forinstance,atRe·Da 1/2 of16,thecompressed
and non-compressed aluminum foams exhibit j values of 0.125 and
0.101,respectivelywhereasthebestperformancelatticeheatsink(L1)
has a j value of 0.168. The better heat transfer performances of the
latticeheatsinksareduetheirsubstantiallyhigherk eff values(asshown
in Fig. 12) as compared to the aluminum foams. In addition, even
though the interfacial heat transfer coefficients (h sf ) of the lattice
structures are slightly lower than the aluminum foams (as shown in
Fig.17),theadvantageofhigherk eff outweighstheeffectsoflowerh sf
andresultedinhigherjvalues.
Fig.16.PlotofNu d,sf versusRe d ofthevariousRhombi-Octetlatticestructures.
low Re d , the Nu d,sf values of the aluminum foams are approximately
70% higher than those of the lattice structures. However, as Re d increases,
the difference in Nu d,sf between the lattice structures and the
aluminum foams decreases. At the highest Re d of 300, Nu d,sf of the
aluminum foams is approximately 32% higher than the lattice structures.Duetothestochasticnatureoftheporousnetwork,thealuminum
foam could have induced tortuous fluid paths which increases fluid
mixing and enhances the interfacial heat transfer rate. On the other
hand,duetotheorderlyarrangementofthelatticestructures,mixingof
the fluid may not be significant which explains the lower Nu d,sf as
compared to the aluminum foams. However, it can also be observed
fromFig.2thattheSLM-producedlatticestructureshavehighsurface
roughness.InarecentinvestigationbyHoetal.[24],itwasdetermined
thattherootmeansquare(rms)roughnessofaplainsurfacefabricated
by SLMwas approximately 7.32μm.This is significantlyrougher than
thecorrespondingvalueof0.25μmforapolishedplaincommercialAl-
6061 surface.The highlyroughsurfaces ofthelatticestructures could
have induced fluid mixing near the wall region which helped to
maintain moderately high Nu d,sf values of between 1.5 and 6.4 as
showninFig.17.
hsf
d
Nud,
sf = = 0.52 Red
Pr
k
f
4.5. Performanceevaluation
0.5 0.5 0.37
(27)
The Colburn j-factor as defined by Eq. (28) is a dimensionless
parameterofforcedconvectionheattransfer.UsingtheNu d correlation
of Eq. (16) that was derived for the lattice structures, j was computed
forthevariousairvelocitiestested.Fig.18showstheplotofjoverthe
rangeofReforlatticestructuresL1–L4.ItshouldbenotedthatReis
basedonthehydraulicdiameteroftheairflowchannel.Itcanbeseen
that j decreases with increasing Re and the rate of reduction in j also
decreases with increasing Re. In addition, j also increases with decreasing
ligament width (d) and the L1 specimen with the smallest d
exhibits the highest j values. This trend of increasing heat transfer
performance of the lattice structure with decreasing d is similarly observedinFig.13.
Fig. 19 compares the j values of the lattice structures against the j
valuesofthealuminumfoamsreportedbyBoomsmaetal.[5]andKim
et al. [34] for various values of Re⋅Da 1/2 . It should be noted that
Boomsma et al. [5] investigated on compressedaluminum foams with
waterasthecoolingmediumwhereasKimetal.[34]reportedtheheat
transfer performance of aluminum foams of 10, 20 and 40 PPI under
unidirectionalairflowanddevelopedacorrelationtocharacterizeNu.
For comparison against the lattice structures, the better performing
compressedaluminumfoamofRef.[5]wasselectedandthecorrelation
of Ref.[34]forthe40PPIaluminum foamwasconverted toanequationintermsoftheColburnj-factor.TheirresultsareshowninFig.19.
Itcanbeseenthatthejvaluesofthelatticestructuresarehigherthan
those of the aluminum foams of Refs. [5,34]. This indicates that the
j = StPr 2/3 (28)
Even though the lattice structures demonstrated significant enhancementsinheattransfercoefficients,thepressuredropsacrossthese
structures are also high. Therefore, a performance metric which accountsfortheimprovementinheattransferandtheassociatedpenalty
in pressure drop for the various structures is required. The thermal
efficiency index (η) of Eq. (29) developed by Tian et al. [19] from a
dimensionalanalysisofthepumpingpowerandheattransferisusedfor
comparisonofthedifferentenhancedsurfaces.Thisindexevaluatesthe
heat transfer performances of the enhanced surfaces at constant
pumping power and had been used for comparing stochastic metallic
foams,periodicstructures,packedbedsandlouveredfins[19].Fig.20
showstheefficiencyindicesofthevariouslatticeheatsinksfordifferent
values of Re. In addition, experiments were also performed on the cylindrical
(C1) and parabolic (P1) pin fin heat sinks of Fig. 3 to obtain
theirthermalandhydrauliccharacteristics.TheefficiencyindicesofC1
andP1weresubsequentlycomputedandplottedinFig.20.Inaddition,
using the experimental results and the respective hydraulic diameter
(D h )oftheflowchannelsreportedinRefs.[5,34],the ηandRevaluesof
thealuminumfoamswerealsocomputedbyEqs.(6),(7),(29)and(30)
andtheirvaluesarepresentedinFig.20forcomparison.Itcanbeseen
that,exceptforthealuminumfoamofRef.[34],theefficiencyindices
ofallthespecimensincreaselinearlywithincreasingRe.Theefficiency
indicesofthelatticeheatsinksaresignificantlyhigherthanthoseofthe
aluminum foams and pin fin heat sinks (C1 and P1). In addition, it is
also interesting to note that the efficiency indices of the lattice heat
sinkscollapsedontoastraightline.Thisindicatesthat,atthesameRe,
theheattransferperformanceofthelatticeheatsinksincreaseslinearly
withf 1 1/3 .
f
1/3
= Nu/ f 1
=
P Dh
L ( U /2)
1 2
f
(29)
(30)
Finally, Fig. 21 shows a comparison of the heat transfer and hydraulicperformancesoftheRhombi-OctetstructuresandtheDiamond
Fig. 17.Comparison of Nu d,sf of the Rhombi-Octet lattice structures against
metalfoamsatdifferentRe d .
285
J.Y.Hoetal. International Journal of Thermal Sciences 137 (2019) 276–287
Octet structures could likely due to the high k eff and high surface
roughnesswhichimprovesfluidmixing.
Re
Nu
H
H
f
= UH
µ
f (31)
= hH ¯
kf (32)
f
=
P H
L ( U /2)
2 2
f
(33)
5. Conclusions
Fig.18.Colburnj-factorsofRhombi-OctetlatticestructuresatdifferentRe.
Fig.19.ComparisonofColburnj-factorofRhombi-Octetlatticestructuresand
aluminumfoamsatdifferentRe⋅Da 1/2 .
Inthisinvestigation,forcedconvectionheattransferperformanceof
anewclassoflatticestructures(Rhombi-Octet)wasexperimentallyand
numerically investigated. Experiments were performed in an air flow
channelforRevaluesbetween1300and7000.Alinearheatconduction
setupwasusedtomeasuretheeffectivethermalconductivities(k eff )of
the lattice structures. Based on the local thermal non-equilibrium
model, numerical simulations were performed to understand the interfacialheattransferperformanceofthelatticestructures.Thethermohydraulicpropertiesofthelatticestructureswerecharacterizedandthe
keyfindingsofthepresentstudyaresummarizedasfollows:
1. Thepressuredrops(ΔP/L)acrossthelatticestructureswerefoundto
increase with decreasing ligament width (d). Using the
Forchheimer-extended Darcy equation, the permeability (K) and
inertia coefficient (C E ) of the lattice structures were obtained. The
permeability-basedfrictionfactors(f⋅Da 1/2 )ofthelatticestructures
areshowntobelowerthanthoseofthenickelandaluminumfoams.
The better hydraulic performances of the lattice structures are attributedtotheirlowerC
E valuesascomparedtothemetallicfoams.
2. The effective thermal conductivities (k eff ) of the lattice structures
weremeasured.Thek eff valuesofthelatticestructureswerefoundto
Fig.20.Comparisonofefficiencyindices(η)ofRhombi-Octetlatticestructures,
aluminumfoamsandpinfinheatsinksfordifferentRevalues.
and Square shaped lattice structures reported by Tian et al. [19]. It
should be noted that the hydraulic diameter in the definitions of Re H ,
Nu H and f 2 is the height of the lattice structures (H) as shown in Eq.
(31–33). Tian et al. [19] performed a systematic investigation on Diamond
and Square shaped lattices of different wire diameter and aperture
size and the lattice frame of each shape with the best thermal
performance was selected for comparison. As shown in Fig. 21 (a), at
low Re H , the friction factors (f 2 ) of the Rhombi-Octet are higher than
those of Ref. [19]. However, the friction factor of the Diamond and
SquareshapedlatticeincreaseswithincreasingRe H .ForRe H >300,the
f 2 valuesoftheDiamondshapedlatticearehigherthanalltheRhombi-
Octetstructures.Ontheotherhand,asshowninFig.21(b),Nu H ofthe
all the Rhombi-Octet structures are higher than the Diamond and
Squareshapedlattices.ThebetterthermalperformancesoftheRhombi-
Fig.21.Comparison of(a)f 2 and (b)Nu H ofthe Rhombi-Octet,Diamondand
Squareshapedlatticestructures.
286
J.Y.Hoetal. International Journal of Thermal Sciences 137 (2019) 276–287
decrease with increasing d and the highest k eff of 22.8W/m⋅K was
recorded. The k eff values of the lattice structures are significantly
higher (up to 5.5 times) than the aluminum foams. This could be
attributed to their higher packing density and orderly unit cell arrangementswhichresultedinmoreeffectiveheatconductionpaths.
3. Theinterfacialheattransfercoefficients(h sf )ofthelatticestructures
wereobtainedbycomparingthenumericalandexperimentalresults
and a relationship between Nu d,sf and Re d was established. The h sf
values areshown toincreasewith decreasingd,withL1exhibiting
thehighesth sf .However,thelatticestructuresexhibitbetween32%
and 70% poorer Nu d,sf as compared to the results for aluminum
foamsreportedintheliterature.
4. The Colburn j-factors and efficiency indices (η) of the lattice structures
were computed and compared against the results for aluminum
foams reported in the literature and conventional pin fin
heat sinks. The j values for the lattice structures are consistently
higherthanthoseofthecompressedandnon-compressedaluminum
foamsandthe ηvaluesofthelatticestructuresarealsosignificantly
higherthan theconventionalcylindricalandparabolicpinfinheat
sinks.
5. A correlation which relates Nu d and Re d of the lattice structures is
developed where 90% of the experimental results fall within±
3.5%deviationfromthepredictedvalues.Themaximumdeviation
of9%isobtainedatthehighestRe d valueof313.
6. Duetothehighforcedconvectionheattransferperformanceofthe
Rhombi-Octetlatticestructures,theycanbeusedasair-cooledheat
sinks for high heat flux electronic components such as the central
processing units (CPU) and random-access memory (RAM) chips
foundinhighperformanceserverunits.Inthisregard,thedataand
correlationpresentedinthisstudycouldbeusedaspredictivetools
bythermaldesignengineers.
Acknowledgements
The authors would like to acknowledge the assistance of Mr. Noor
MuhammadBinNoorKhalidandMs.LinLiNgforconductingsomeof
the experiments presented in this paper. The SLM equipment used in
thisresearchissupportedbytheNationalResearchFoundation,Prime
Minister's Office, Singapore under its Medium-Sized Centre funding
scheme.
References
[1] C.Y.Zhao,Reviewonthermaltransportinhighporositycellularmetalfoamswith
opencells,Int.J.HeatMassTran.55(2012)3618–3632.
[2] X.Ding,L.Lu,C.Chen,Z.He,D.Ou,Heattransferenhancementbyusingfourkinds
ofporousstructuresinaheatexchanger,Appl.Mech.Mater.52–54(2011)
1632–1637.
[3] G.B.Ribeiro,J.R.BarbosaJr.,A.T.Prata,Performanceofmicrochannelcondensers
withmetalfoamsontheair-side:applicationinsmall-scalerefrigerationsystems,
Appl.Therm.Eng.36(2012)152–160.
[4] K.-C.Chen,C.-C.Wang,Performanceimprovementofhighpoweredliquid-cooled
heatsinkvianon-uniformmetalfoamarrangement,Appl.Therm.Eng.87(2015)
41–46.
[5] K.Boomsma,D.Poulikakos,F.Zwick,Metalfoamsascompacthighperformance
heatexchangers,Mech.Mater.35(2003)1161–1176.
[6] A.Bhattacharya,R.L.Mahajan,Metalfoamandfinnedmetalfoamheatsinksfor
electroniccoolinginbuoyancy-inducedconvection,J.Electron.Packag.128(2006)
259–266.
[7] K.C.Leong,L.W.Jin,Anexperimentalstudyofheattransferinoscillatingflow
throughachannelfilledwithanaluminumfoam,Int.J.HeatMassTran.48(2005)
243–253.
[8] K.C.Leong,L.W.Jin,Effectsofoscillatingfrequencyonheattransferinmetalfoam
heatsinksofvariousporedensities,Int.J.HeatMassTran.49(2006)671–681.
[9] M.Hatami,D.D.Ganji,Thermalperformanceofcircularconvective-radiative
porousfinswithdifferentsectionshapesandmaterials,EnergyConvers.Manag.76
(2013)185–193.
[10] A.S.Dogonchi,D.D.Ganji,Convection-radiationheattransferstudyofmovingfin
withtemperature-dependentthermalconductivity,heattransfercoefficientand
heatgeneration,Appl.Therm.Eng.103(2016)705–712.
[11] W.Lu,C.Y.Zhao,S.A.Tassou,Thermalanalysisonmetal-foamfilledheatexchangers.PartI:metal-foamfilledpipes,Int.J.HeatMassTran.49(2006)
2751–2761.
[12] V.V.Calmidi,R.L.Mahajan,Theeffectivethermalconductivityofhighporosity
fibrousmetalfoams,J.HeatTran.121(1999)466–471.
[13] S.Y.Kim,J.W.Paek,B.H.Kang,Flowandheattransfercorrelationsforporousfinin
aplate-finheatexchanger,J.HeatTran.122(2000)572–578.
[14] K.-J.Kang,Wire-wovencellularmetals:thepresentandfuture,Prog.Mater.Sci.69
(2015)213–307.
[15] J.-S.Liu,T.J.Lu,Multi-objectiveandmulti-loadingoptimizationofultralightweight
trussmaterials,Int.J.SolidStruct.41(2004)619–635.
[16] S.Hyun,A.M.Karlsson,S.Torquato,A.G.Evans,SimulatedpropertiesofKagomé
andtetragonaltrusscorepanels,Int.J.SolidStruct.40(2003)6989–6998.
[17] T.J.Lu,L.Valdevit,A.G.Evans,Activecoolingbymetallicsandwichstructureswith
periodiccores,Prog.Mater.Sci.50(2005)789–815.
[18] T.Kim,H.P.Hodson,T.J.Lu,Fluid-flowandendwallheat-transfercharacteristicsof
anultralightlattice-framematerial,Int.J.HeatMassTran.47(2004)1129–1140.
[19] J.Tian,T.Kim,T.J.Lu,H.P.Hodson,D.T.Queheillalt,D.J.Sypeck,H.N.G.Wadley,
Theeffectsoftopologyuponfluid-flowandheat-transferwithincellularcopper
structures,Int.J.HeatMassTran.47(2004)3171–3186.
[20] K.J.Maloney,K.D.Fink,T.A.Schaedler,J.A.Kolodziejska,A.J.Jacobsen,
C.S.Roper,Multifunctionalheatexchangersderivedfromthree-dimensionalmicrolatticestructures,Int.J.HeatMassTran.55(2012)2486–2493.
[21] K.N.Son,J.A.Weibel,V.Kumaresan,S.V.Garimella,Designofmultifunctional
lattice-framematerialsforcompactheatexchangers,Int.J.HeatMassTran.115
(2017)619–629.
[22] L.Ventola,F.Robotti,M.Dialameh,F.Calignano,D.Manfredi,E.Chiavazo,
P.Asinari,Roughsurfaceswithenhancedheattransferforelectronicscoolingby
directmetallasersintering,Int.J.HeatMassTran.75(2014)58–74.
[23] M.Wong,I.Owen,C.J.Sutcliffe,A.Puri,Convectiveheattransferandpressure
lossesacrossnovelheatsinksfabricatedbyselectivelasermelting,Int.J.HeatMass
Tran.52(2009)281–288.
[24] J.Y.Ho,K.K.Wong,K.C.Leong,SaturatedpoolboilingofFC-72fromenhanced
surfacesproducedbyselectivelasermelting,Int.J.HeatMassTran.99(2016)
107–121.
[25] K.K.Wong,K.C.Leong,Saturatedpoolboilingenhancementusingporouslattice
structuresproducedbySelectiveLaserMelting,Int.J.HeatMassTran.121(2018)
46–63.
[26] J.Y.Ho,K.C.Leong,Filmwisecondensationonmicro-finsurfacesproducedby
SelectiveLaserMelting,Proceedingsofthe13thInternationalConferenceonHeat
Transfer,FluidMechanicsandThermodynamics(HEFAT2017),2017,pp.85–9017
–19July,Portorož,Slovenia.
[27] J.Y.Ho,K.C.Leong,Cylindricalporousinsertsforenhancingthethermalandhydraulicperformanceofwater-cooledcoldplates,Appl.Therm.Eng.121(2017)
863–878.
[28] J.Y.Ho,K.C.Leong,EnhancedThermalPerformanceofaWater-cooledColdPlate
withPorousInsertsFabricatedbySelectiveLaserMelting,Proceedingsofthe13th
InternationalConferenceonHeatTransfer,FluidMechanicsandThermodynamics
(HEFAT2017),2017,pp.860–86517–19July,Portorož,Slovenia.
[29] Y.A.Çengel,A.J.Ghajar,HeatandMassTransfer:FundamentalsandApplications,
fifthed.,McGraw-HillEducation,NewYork,USA,2015.
[30] R.J.Moffat,Usinguncertaintyanalysisintheplanningofanexperiment,J.Fluid
Eng.107(1985)173–178.
[31] ASMInternational,MetalsHandbook,tenthed.,vol.2,TheMaterialsCompany,
1990PropertiesandSelection:NonferrousAlloysandSpecial-PurposeMaterials.
[32] K.C.Leong,H.Y.Li,L.W.Jin,J.C.Chai,Numericalandexperimentalstudyofforced
convectioningraphitefoamsofdifferentconfigurations,Appl.Therm.Eng.30
(2010)520–532.
[33] G.S.Beavers,E.M.Sparrow,Non-Darcyflowthroughfibrousporousmedia,J.Appl.
Mech.36(1969)711–714.
[34] S.Y.Kim,B.H.Kang,J.-H.Kim,Forcedconvectionfromaluminumfoammaterials
inanasymmetricallyheatedchannel,Int.J.HeatMassTran.44(2001)1451–1454.
[35] V.V.Calmidi,R.L.Mahajan,Forcedconvectioninhighporositymetalfoams,J.
HeatTran.122(2000)557–565.
[36] R.Singh,H.S.Kasana,Computationalaspectofeffectivethermalconductivityof
highlyporousmetalfoams,Appl.Therm.Eng.24(2004)1841–1849.
[37] Z.Dai,N.Nawaz,Y.G.Park,J.Bock,A.M.Jacobi,Correctingandextendingthe
Boomsma-Poulikakoseffectivethermalconductivitymodelforthree-dimensional,
fluid-saturatedmetalfoams,Int.Commun.HeatMassTran.37(2010)575–580.
[38] H.Yang,M.Zhao,Z.L.Gu,L.W.Jin,J.C.Chai,Afurtherdiscussionontheeffective
thermalconductivityofmetalfoam:animprovedmodel,Int.J.HeatMassTran.86
(2015)207–211.
[39] N.Dukhan,Ö.Bağcı,M.Özdemir,Metalfoamhydrodynamics:flowregimesfrom
pre-Darcytoturbulent,Int.J.HeatMassTran.77(2014)114–123.
287