J Heat Transfer 2006 Vol 128 N05

Journal of
Heat Transfer
Published Monthly by ASME
VOLUME 128 • NUMBER 5 • MAY 2006

Editor, YOGESH JALURIA „2010…
Associate Editors
S. ACHARYA „2006…
N. K. ANAND „2006… RESEARCH PAPERS
L. C. BURMEISTER „2008…
B. FAROUK „2006… Thermodynamic Properties
S. V. GARIMELLA „2007…
C. P. GRIGOROPOULOS „2006… 421 Analysis of Enthalpy Approximation for Compressed Liquid Water
A. HAJI-SHEIKH „2008…
A. M. JACOBI „2008… Milivoje M. Kostic
Y. JOSHI „2008…
S. G. KANDLIKAR „2007…
J. M. KHODADADI „2007… Heat Exchangers
J. LAGE „2008…
J. H. LIENHARD V „2006… 427 A Model for Condensate Retention on Plain-Fin Heat Exchangers
P. M. LIGRANI „2006…
R. M. MANGLIK „2008… A. I. ElSherbini and A. M. Jacobi
C. H. OH „2007…
R. PITCHUMANI „2007… Porous Media
R. P. ROY „2007…
B. SUNDEN „2008…
K. A. THOLE „2007… 434 Heat Transfer in a Porous Electrode of Fuel Cells
W. W. YUEN „2008… J. J. Hwang
Past Editors
V. DHIR
J. R. HOWELL 444 A Correlation for Interfacial Heat Transfer Coefficient for Turbulent Flow
R. VISKANTA Over an Array of Square Rods
G. M. FAETH Marcelo B. Saito and Marcelo J. S. de Lemos
K. T. YANG
E. M. SPARROW
453 Convective Heat Transfer in a Rectangular Channel Filled With Sintered
HEAT TRANSFER DIVISION Bronze Beads and Periodically Spaced Heated Blocks
Chair, MICHAEL K. JENSEN
Vice Chair, RODNEY W. DOUGLASS Sheng Chung Tzeng
Past Chair, R. D. SKOCYPEC
PUBLICATIONS COMMITTEE
Micro/Nanoscale Heat Transfer
Chair, ARTHUR G. ERDMAN
465 Heat Conduction in Nanofluid Suspensions
OFFICERS OF THE ASME
President, RICHARD E. FEIGEL Peter Vadasz
Executive Director,
VIRGIL R. CARTER Combustion and Reactive Flows
Treasurer,
THOMAS D. PESTORIUS
478 Study of Infrared CO2 Radiation From Liquid-Fueled Combustor
PUBLISHING STAFF
Y. Levy, M. Lev, and V. Ovcharenko
Managing Director, Publishing
PHILIP DI VIETRO
Manager, Journals
Conduction
COLIN McATEER
Production Assistant 484 Numerical and Experimental Investigation of Thermal Signatures
MARISOL ANDINO of Buried Landmines in Dry Soil
F. Moukalled, N. Ghaddar, H. Kabbani, N. Khalid, and Z. Fawaz
TECHNICAL BRIEFS
495 Natural Convection From a Square Disk in the Intermediate
Rayleigh Range
Arnout Willockx, Gilbert De Mey, Christophe T’Joen,
Hendrik-Jan Steeman, and Michel De Paepe
Transactions of the ASME, Journal of Heat Transfer
共ISSN 0022-1481兲 is published monthly by The American
Society of Mechanical Engineers, Three Park Avenue,
New York, NY 10016. Periodicals postage paid at
New York, NY and additional mailing offices.
POSTMASTER: Send address changes to Transactions
of the ASME, Journal of Heat Transfer, c/o THE
AMERICAN SOCIETY OF MECHANICAL ENGINEERS,
22 Law Drive, Box 2300, Fairfield, NJ 07007-2300.
CHANGES OF ADDRESS must be received at Society
headquarters seven weeks before they are to be effective.
Please send old label and new address.
STATEMENT from By-Laws. The Society shall not be
responsible for statements or opinions advanced in papers
or ... printed in its publications 共B7.1, Para. 3兲.
COPYRIGHT © 2006 by The American Society of Mechanical „Contents continued on inside back cover…
Engineers. For authorization to photocopy material for
internal or personal use under those circumstances not falling
within the fair use provisions of the Copyright Act, contact the
Copyright Clearance Center 共CCC兲, 222 Rosewood Drive,
Danvers, MA 01923, tel: 978-750-8400, www.copyright.com.
Request for special permission or bulk copying should
be addressed to Reprints/Permission Department.
Canadian Goods & Services Tax Registration #126148048
Downloaded 04 Feb 2011 to 194.85.80.107. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
„Contents continued…
Journal of Heat Transfer Volume 128, Number 5 MAY 2006
499 Least-Squares Finite Element Analysis for Transient Radiative Transfer in Absorbing and Scattering Media
W. An, L. M. Ruan, H. P. Tan, and H. Qi
504 On the Entropy Generation Formula of Radiation Heat Transfer Processes

L. H. Liu and S. X. Chu
The ASME Journal of Heat Transfer is abstracted and indexed in the

following:
Applied Science and Technology Index, Chemical Abstracts, Chemical Engineering and
Biotechnology Abstracts (Electronic equivalent of Process and Chemical Engineering),
Civil Engineering Abstracts, Compendex (The electronic equivalent of Engineering
Index), Corrosion Abstracts, Current Contents, E & P Health, Safety, and Environment,
Ei EncompassLit, Engineered Materials Abstracts, Engineering Index, Enviroline (The
electronic equivalent of Environment Abstracts), Environment Abstracts, Environmental
Engineering Abstracts, Environmental Science and Pollution Management, Fluidex,
Fuel and Energy Abstracts, Index to Scientific Reviews, INSPEC, International Building
Services Abstracts, Mechanical & Transportation Engineering Abstracts, Mechanical
Engineering Abstracts, METADEX (The electronic equivalent of Metals Abstracts and
Alloys Index), Petroleum Abstracts, Process and Chemical Engineering, Referativnyi
Zhurnal, Science Citation Index, SciSearch (The electronic equivalent of Science
Citation Index), Theoretical Chemical Engineering
Downloaded 04 Feb 2011 to 194.85.80.107. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Analysis of Enthalpy
Approximation for Compressed
Liquid Water
It is custom to approximate solid and liquid thermodynamic properties as being a func-
tion of temperature only, since they are virtually incompressible, and Pdv boundary work
may be neglected. Furthermore, in classical literature, for isothermal compression pro-
cesses, a general “improvement” and correction for liquid enthalpy approximation is
Milivoje M. Kostic given by adding the “pressure correction,” vdP, to the corresponding saturation value. It
Northern Illinois University, is shown that such correction given for isothermal processes is generally valid for isen-
DeKalb, IL 60115-2854 tropic processes only. Analysis of water real properties, over the saturation temperature
e-mail: kostic@niu.edu range and a wide pressure range up to 100 MPa, shows that the recommended correc-
tions are only beneficial for higher pressures at smaller temperatures (below 200° C),
insignificant for smaller pressures at most of the temperatures, about the same but op-
posite sign (thus unnecessary) for intermediate temperatures and pressures, and more
erroneous (thus counterproductive and misleading) for higher temperatures and pres-
sures, than the corresponding saturation values without any correction. The misconcep-
tion in the literature is a result of the erroneous assumption, that due to incompressibility
for liquids in general, the internal energy is less dependent on pressure than
enthalpy. 关DOI: 10.1115/1.2175090兴
Keywords: enthalpy, water, thermodynamic properties, thermodynamic analysis,

isothermal, isentropic
1 Introduction
Since solids and liquids are virtually 共but not exactly兲 incom-
pressible, then the compression work, Pdv, could be neglected and
thus many properties virtually will not be a function of pressure 共4兲
but temperature only, such as specific internal energy, u, etc. Fur-
thermore, any process is also at the same time an isochoric,
constant-volume process. Namely, the isobaric, constant-pressure Furthermore, for isothermal processes 共dT = 0 and du ⬇ 0, then
process will be a simultaneously constant-volume process for an dh ⬇ vdP兲, and finally, for finite pressure difference change from
incompressible substance, so that specific heat at constant pres- saturated pressure, Psat, corresponding to the given temperature,
sure, c p, and constant volume, cv, are the same, or approximately T, the specific enthalpy with correction, hcorr共T , P兲 at that tem-
the same for virtually incompressible real solids and liquids, par- perature, T, and any pressure, P, will be 关1,2兴:
ticularly when compared to vapors and gases, i.e.:
u ⬇ u共T兲 ⬇ usat共T兲 and c p ⬇ cv ⬇ c共T兲共1兲
共5a兲
Even the specific enthalpy for a liquid 共from here on word “spe-
cific” will be assumed and omitted for brevity兲, can be approxi-
mated to be independent from pressure and conveniently taken to
be equal to the corresponding saturated liquid value at the given 共5b兲
temperature:
h共P,T兲 ⬇ h共T兲 ⬇ hsat共T兲共2兲 where, hsat共T兲 and vsat共T兲 are liquid saturation enthalpy and liquid
saturation specific volume at given temperature, T, respectively. It
However, enthalpy is unique, since it is explicitly defined as a
is stated in many references, including 关1,2兴, that the above equa-
function of pressure, namely:
tions 共5a兲 and 共5b兲 are recommended as the correction for isother-
h⬅u+P·v thus, h共T, P兲 = u共T, P兲 + P · v ⬇ usat共T兲 + P · v mal, liquid enthalpy dependence on pressure, and that it is more
accurate than a simple, approximation without correction, hsat 共Eq.
共3兲
共2兲兲.
Therefore, it is common in most engineering references, including It is the objective of this paper to point out the erroneous gen-
classical and widely used thermodynamics textbooks 关1,2兴, to eral recommendations in the literature. The correction 共Eq. 共5兲兲, as
evaluate the change of enthalpy, assuming incompressibility 共dv recommended in many references, is only useful for higher pres-
= 0兲, but taking correction for pressure increase as: sures at smaller temperatures, but is actually more erroneous 共thus
counterproductive and misleading兲 for higher temperatures and
pressures, and is about the same 共but opposite sign, thus not nec-
Contributed by the Heat Transfer Division of ASME for publication in the JOUR- essary兲 for intermediate temperatures, than the simple approxima-
HEAT TRANSFER. Manuscript received December 13, 2004; final manuscript
NAL OF
received November 1, 2005. Review conducted by John H. Lienhard V. Paper pre-
tion 共Eq. 共2兲兲 without any correction. Corresponding analysis us-
sented at the 2004 ASME International Mechanical Engineering Congress 共IM- ing real water data 关3兴 and physical justification are presented
ECE2004兲, November 13–19, 2004, Anaheim, California, USA. below.
Journal of Heat Transfer Copyright © 2006 by ASME MAY 2006, Vol. 128 / 421
Downloaded 06 Dec 2010 to 194.27.225.72. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
422 / Vol. 128, MAY 2006
Table 1 Compressed liquid water property data at 260° C †3‡ and different enthalpy corrections
⌬h or
P v u h s Cv Cp csound ␮JT ⌬h⌬u ⌬h⌬v ⌬h⌬P ⌬hcorr −兰␮CpdP hsat-h hcorr-h
共MPa兲共m3 / kg兲共kJ/kg兲共kJ/kg兲共kJ/ kg K兲共kJ/ kg K兲共kJ/ kg K兲共m/s兲共K/MPa兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲
4.69sat 0.001276 1129.0 1135.0 2.8849 3.1301 4.9856 1105.3 0.03472 0.00 0.00 0.00 0.00 0.00 0.00 0.00
5 0.001276 1128.5 1134.9 2.8841 3.1299 4.9804 1107.2 0.03373 −0.50 0.00 0.39 0.39 −0.05 0.10 0.49
10 0.001265 1121.6 1134.3 2.8710 3.1257 4.9019 1135.8 0.01884 −7.40 −0.08 6.74 6.77 −0.70 0.70 7.47
15 0.001256 1115.1 1134.0 2.8586 3.1220 4.8336 1162.6 0.00582 −13.90 −0.20 13.05 13.15 −1.00 1.00 14.15
20 0.001247 1109.0 1134.0 2.8469 3.1187 4.7733 1187.8 −0.00568 −20.00 −0.35 19.31 19.53 −1.00 1.00 20.53
25 0.001239 1103.2 1134.2 2.8357 3.1158 4.7195 1211.7 −0.01595 −25.80 −0.53 25.52 25.91 −0.75 0.80 26.71
30 0.001231 1097.8 1134.7 2.8250 3.1131 4.6711 1234.3 −0.02519 −31.20 −0.74 31.70 32.30 −0.26 0.30 32.60
35 0.001224 1092.5 1135.4 2.8148 3.1107 4.6272 1256.0 −0.03355 −36.50 −0.98 37.84 38.68 0.42 −0.40 38.28
40 0.001217 1087.6 1136.3 2.8050 3.1084 4.5871 1276.7 −0.04117 −41.40 −1.24 43.94 45.06 1.28 −1.30 43.76
45 0.001211 1082.8 1137.3 2.7955 3.1063 4.5502 1296.6 −0.04816 −46.20 −1.52 50.01 51.44 2.30 −2.30 49.14
50 0.001204 1078.2 1138.4 2.7864 3.1043 4.5162 1315.8 −0.05458 −50.80 −1.82 56.05 57.82 3.46 −3.40 54.42
55 0.001199 1073.8 1139.7 2.7775 3.1024 4.4847 1334.3 −0.06051 −55.20 −2.12 62.05 64.20 4.76 −4.70 59.50
60 0.001193 1069.6 1141.1 2.7690 3.1006 4.4553 1352.2 −0.06601 −59.40 −2.45 68.03 70.58 6.17 −6.10 64.48
65 0.001187 1065.5 1142.7 2.7607 3.0989 4.4279 1369.5 −0.07113 −63.50 −2.80 73.98 76.96 7.70 −7.70 69.26
70 0.001182 1061.6 1144.3 2.7526 3.0973 4.4022 1386.3 −0.07590 −67.40 −3.15 79.91 83.34 9.32 −9.30 74.04
75 0.001177 1057.7 1146.0 2.7447 3.0957 4.3781 1402.6 −0.08037 −71.30 −3.52 85.80 89.72 11.03 −11.00 78.72
80 0.001172 1054.0 1147.8 2.7371 3.0943 4.3553 1418.5 −0.08456 −75.00 −3.90 91.68 96.10 12.83 −12.80 83.30
85 0.001167 1050.5 1149.7 2.7297 3.0928 4.3338 1433.9 −0.08850 −78.50 −4.29 97.53 102.48 14.71 −14.70 87.78
90 0.001163 1047.0 1151.6 2.7224 3.0914 4.3134 1449.0 −0.09220 −82.00 −4.69 103.35 108.86 16.67 −16.60 92.26
95 0.001158 1043.6 1153.7 2.7153 3.0901 4.2940 1463.7 −0.09570 −85.40 −5.09 109.15 115.24 18.69 −18.70 96.54
100 0.001154 1040.3 1155.8 2.7084 3.0888 4.2756 1478.1 −0.09901 −88.70 −5.51 114.93 121.62 20.77 −20.80 100.82
Transactions of the ASME
Table 2 Compressed liquid water property data at different temperatures and pressures †3‡
Journal of Heat Transfer
T=4°C T = 20° C T = 50° C T = 100° C T = 150° C
P v u h v u h v u h v u h v u h
共MPa兲共m3 / kg兲共kJ/kg兲共kJ/kg兲共m3 / kg兲共kJ/ kg兲共kJ/ kg兲共m3 / kg兲共kJ/ kg兲共kJ/ kg兲共m3 / kg兲共kJ/kg兲共kJ/kg兲共m3 / kg兲共kJ/kg兲共kJ/kg兲
Sat 0.001000 16.81 16.81 0.001002 83.912 83.914 0.001012 209.33 209.34 0.001044 419.06 419.17 0.001091 631.66 632.18
5 0.000998 16.81 21.80 0.001000 83.609 88.607 0.001010 208.59 213.64 0.001041 417.64 422.85 0.001088 629.55 634.98
10 0.000995 16.79 26.75 0.000997 83.308 93.281 0.001008 207.86 217.94 0.001039 416.23 426.62 0.001084 627.27 638.11
15 0.000993 16.77 31.66 0.000995 83.007 97.934 0.001006 207.15 222.23 0.001036 414.85 430.39 0.001081 625.05 641.27
20 0.000990 16.74 36.55 0.000993 82.708 102.57 0.001004 206.44 226.51 0.001034 413.50 434.17 0.001078 622.89 644.45
25 0.000988 16.70 41.40 0.000991 82.409 107.18 0.001001 205.75 230.79 0.001031 412.17 437.95 0.001075 620.78 647.66
30 0.000986 16.65 46.23 0.000989 82.112 111.77 0.000999 205.07 235.05 0.001029 410.87 441.74 0.001072 618.73 650.89
35 0.000984 16.59 51.02 0.000987 81.815 116.34 0.000997 204.40 239.31 0.001027 409.60 445.54 0.001069 616.72 654.14
40 0.000981 16.53 55.79 0.000985 81.520 120.90 0.000995 203.75 243.56 0.001025 408.35 449.33 0.001066 614.77 657.42
45 0.000979 16.46 60.53 0.000982 81.225 125.44 0.000993 203.10 247.80 0.001022 407.13 453.14 0.001064 612.85 660.71
50 0.000977 16.38 65.24 0.000980 80.931 129.95 0.000991 202.46 252.03 0.001020 405.93 456.94 0.001061 610.98 664.02
55 0.000975 16.30 69.93 0.000978 80.637 134.45 0.000989 201.84 256.26 0.001018 404.76 460.75 0.001058 609.15 667.35
60 0.000973 16.21 74.59 0.000977 80.345 138.94 0.000988 201.22 260.47 0.001016 403.61 464.56 0.001056 607.36 670.69
65 0.000971 16.12 79.23 0.000975 80.053 143.40 0.000986 200.61 264.68 0.001014 402.47 468.37 0.001053 605.61 674.05
70 0.000969 16.01 83.84 0.000973 79.762 147.85 0.000984 200.01 268.88 0.001012 401.36 472.19 0.001051 603.89 677.43
75 0.000967 15.91 88.43 0.000971 79.471 152.29 0.000982 199.42 273.07 0.001010 400.27 476.00 0.001048 602.21 680.81
80 0.000965 15.80 93.00 0.000969 79.182 156.70 0.000980 198.84 277.26 0.001008 399.20 479.82 0.001046 600.56 684.21
85 0.000963 15.68 97.55 0.000967 78.893 161.10 0.000978 198.26 281.44 0.001006 398.14 483.64 0.001043 598.95 687.63
90 0.000961 15.56 102.07 0.000965 78.604 165.49 0.000977 197.70 285.61 0.001004 397.11 487.46 0.001041 597.36 691.05
95 0.000959 15.44 106.58 0.000964 78.317 169.86 0.000975 197.14 289.77 0.001002 396.09 491.29 0.001039 595.81 694.48
100 0.000958 15.31 111.06 0.000962 78.031 174.22 0.000973 196.59 293.92 0.001000 395.09 495.11 0.001036 594.29 697.93
T = 200° C T = 250° C T = 300° C T = 350° C
P v u h v u h v u h v u h
共MPa兲共m3 / kg兲共kJ/kg兲共kJ/kg兲共m3 / kg兲共kJ/ kg兲共kJ/ kg兲共m3 / kg兲共kJ/ kg兲共kJ/ kg兲共m3 / kg兲共kJ/kg兲共kJ/kg兲
Sat 0.001157 850.47 852.27 0.001252 1080.8 1085.8 0.001404 1332.9 1345.0 0.00174 1642.1 1670.9
5 0.001153 847.91 853.68 0.00125 1079.5 1085.7 Psat = 8.5879 MPa Psat = 16.529 MPa
10 0.001148 844.31 855.80 0.001241 1073.4 1085.8 0.001398 1329.4 1343.3
15 0.001144 840.84 857.99 0.001233 1067.6 1086.1 0.001378 1317.6 1338.3
20 0.001139 837.49 860.27 0.001225 1062.2 1086.7 0.001361 1307.1 1334.4 0.001665 1612.7 1646.0
25 0.001135 834.24 862.61 0.001218 1057.0 1087.4 0.001346 1297.6 1331.3 0.001599 1583.9 1623.9
30 0.00113 831.10 865.02 0.001211 1052.0 1088.4 0.001332 1288.9 1328.9 0.001553 1562.2 1608.8
35 0.001126 828.06 867.48 0.001205 1047.3 1089.4 0.001320 1280.8 1327.0 0.001517 1544.5 1597.6
40 0.001122 825.10 870.00 0.001199 1042.7 1090.7 0.001308 1273.3 1325.6 0.001488 1529.3 1588.8
45 0.001119 822.23 872.57 0.001193 1038.4 1092.0 0.001298 1266.2 1324.6 0.001464 1515.9 1581.8
50 0.001115 819.45 875.19 0.001187 1034.2 1093.5 0.001288 1259.6 1324.0 0.001443 1503.9 1576.1
55 0.001111 816.73 877.85 0.001182 1030.1 1095.1 0.001279 1253.3 1323.6 0.001424 1493.1 1571.4
60 0.001108 814.09 880.55 0.001176 1026.2 1096.8 0.001270 1247.3 1323.5 0.001407 1483.1 1567.5
MAY 2006, Vol. 128 / 423
65 0.001104 811.51 883.29 0.001171 1022.5 1098.6 0.001262 1241.6 1323.6 0.001391 1473.8 1564.3
70 0.001101 809.00 886.07 0.001167 1018.8 1100.5 0.001254 1236.1 1323.9 0.001377 1465.2 1561.6
75 0.001098 806.56 888.89 0.001162 1015.3 1102.4 0.001247 1230.9 1324.4 0.001364 1457.1 1559.5
80 0.001095 804.17 891.73 0.001157 1011.8 1104.4 0.001240 1225.9 1325.1 0.001353 1449.5 1557.7
85 0.001091 801.83 894.61 0.001153 1008.5 1106.5 0.001233 1221.1 1325.9 0.001341 1442.3 1556.3
90 0.001088 799.55 897.51 0.001149 1005.3 1108.7 0.001227 1216.4 1326.8 0.001331 1435.5 1555.3
95 0.001086 797.32 900.44 0.001145 1002.1 1110.9 0.001221 1212.0 1327.9 0.001321 1429.0 1554.5
100 0.001083 795.14 903.40 0.001141 999.06 1113.1 0.001215 1207.6 1329.1 0.001312 1422.8 1554.0
424 / Vol. 128, MAY 2006
Table 3 Compressed liquid water enthalpies, h, their approximation differences, hsat − h „Eq. „2……, and hcorr − h „Eq. „5……, and related percentages „Note: More erroneous approxima-
tions are indicated in bold.…
T °C → 4°C 20° C 50° C 100° C 150° C 200° C 250° C 300° C 350° C

h 共kJ/kg兲 h 共kJ/kg兲 h 共kJ/kg兲 h 共kJ/kg兲 h 共kJ/kg兲 h 共kJ/kg兲 h 共kJ/kg兲 h 共kJ/kg兲 h 共kJ/kg兲
P hsat − h hcorr − h hsat − h hcorr − h hsat − h hcorr − h hsat − h hcorr − h hsat − h hcorr − h hsat − h hcorr − h hsat − h hcorr − h hsat − h hcorr − h hsat − h hcorr − h
共MPa兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲共kJ/kg兲
% % % % % % % % % % % % % % % % % %
Sat. 16.81 83.91 209.34 419.17 632.18 852.27 1085.80 1345.00 1670.90
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0%
5 21.80 88.61 213.64 422.85 634.98 853.68 1085.70 Vapor
−4.98 0.02 −4.69 0.31 −4.30 0.75 −3.68 1.43 −2.80 2.13 −1.41 2.57 0.10 1.38 Psat=
−22.9% 0.1% −5.3% 0.4% −2.0% 0.4% −0.9% 0.3% −0.4% 0.3% −0.2% 0.3% 0.0% 0.1% 8.5879 MPa
10 26.75 93.28 217.94 426.62 638.11 855.80 1085.80 1343.30 Vapor
−9.93 0.07 −9.37 0.65 −8.60 1.51 −7.45 2.88 −5.93 4.46 −3.53 6.24 0.00 7.54 1.70 3.68 Psat=
−37.1% 0.3% −10.0% 0.7% −3.9% 0.7% −1.7% 0.7% −0.9% 0.7% −0.4% 0.7% 0.0% 0.7% 0.1% 0.3% 6.529 MPa
15 31.66 97.93 222.23 430.39 641.27 857.99 1086.10 1338.30
−14.85 0.15 −14.02 1.00 −12.89 2.28 −11.22 4.33 −9.09 6.75 −5.72 9.83 −0.30 13.50 6.70 15.70
−46.9% 0.5% −14.3% 1.0% −5.8% 1.0% −2.6% 1.0% −1.4% 1.1% −0.7% 1.1% 0.0% 1.2% 0.5% 1.2%
20 36.55 102.57 226.51 434.17 644.45 860.27 1086.70 1334.40 1646.00

−19.74 0.27 −18.66 1.38 −17.17 3.06 −15.00 5.76 −12.27 9.02 −8.00 13.33 −0.90 19.16 10.60 26.62 24.90 30.94
−54.0% 0.7% −18.2% 1.3% −7.6% 1.4% −3.5% 1.3% −1.9% 1.4% −0.9% 1.5% −0.1% 1.8% 0.8% 2.0% 1.5% 1.9%
30 46.23 111.77 235.05 441.74 650.89 865.02 1088.40 1328.90 1608.80

−29.41 0.59 −27.86 2.20 −25.71 4.64 −22.57 8.63 −18.71 13.49 −12.75 20.15 −2.60 29.97 16.10 46.17 62.10 85.54
−63.6% 1.3% −24.9% 2.0% −10.9% 2.0% −5.1% 2.0% −2.9% 2.1% −1.5% 2.3% −0.2% 2.8% 1.2% 3.5% 3.9% 5.3%
40 55.79 120.90 243.56 449.33 657.42 870.00 1090.70 1325.60 1588.80

−38.98 1.03 −36.99 3.08 −34.22 6.25 −30.16 11.47 −25.24 17.86 −17.73 26.73 −4.90 40.19 19.40 63.51 82.10 122.94
−69.9% 1.8% −30.6% 2.6% −14.0% 2.6% −6.7% 2.6% −3.8% 2.7% −2.0% 3.1% −0.4% 3.7% 1.5% 4.8% 5.2% 7.7%
50 65.24 129.95 252.03 456.94 664.02 875.19 1093.50 1324.00 1576.10

−48.43 1.58 −46.04 4.05 −42.69 7.90 −37.77 14.30 −31.84 22.17 −22.92 33.11 −7.70 49.91 21.00 79.15 94.80 153.04
−74.2% 2.4% −35.4% 3.1% −16.9% 3.1% −8.3% 3.1% −4.8% 3.3% −2.6% 3.8% −0.7% 4.6% 1.6% 6.0% 6.0% 9.7%
60 74.59 138.94 260.47 464.56 670.69 880.55 1096.80 1323.50 1567.50

−57.78 2.23 −55.03 5.08 −51.13 9.58 −45.39 17.11 −38.51 26.40 −28.28 39.31 −11.00 59.12 21.50 93.69 103.40 179.04
−77.5% 3.0% −39.6% 3.7% −19.6% 3.7% −9.8% 3.7% −5.7% 3.9% −3.2% 4.5% −1.0% 5.4% 1.6% 7.1% 6.6% 11.4%
Transactions of the ASME
80 93.00 156.70 277.26 479.82 684.21 891.73 1104.40 1325.10 1557.70

−76.19 3.82 −72.79 7.36 −67.92 13.04 −60.65 22.72 −52.03 34.69 −39.46 51.26 −18.60 76.56 19.90 120.18 113.20 223.64
−81.9% 4.1% −46.4% 4.7% −24.5% 4.7% −12.6% 4.7% −7.6% 5.1% −4.4% 5.7% −1.7% 6.9% 1.5% 9.1% 7.3% 14.4%
100 111.06 174.22 293.92 495.11 697.93 903.40 1113.10 1329.10 1554.00
−94.25 5.76 −90.31 9.87 −84.58 16.62 −75.94 28.30 −65.75 42.78 −51.13 62.72 −27.30 92.89 15.90 144.26 116.90 262.14
−84.9% 5.2% −51.8% 5.7% −28.8% 5.7% −15.3% 5.7% −9.4% 6.1% −5.7% 6.9% −2.5% 8.3% 1.2% 10.9% 7.5% 16.9%
2 Analysis
Compressed liquid water properties 关3兴, for different pressure at
260° C, are presented in Table 1 and selected properties for dif-
ferent temperatures in Table 2. In addition, the corresponding cor-
rections 共⌬h⌬u, ⌬h⌬v, and ⌬h⌬P, see Eqs. 共4兲–共8兲兲 are tabulated
along with differences of the approximated enthalpies, without
共hsat − h兲 and with correction 共hcorr − h兲 from the real enthalpy val-
ues, h, in the last two columns in Table 1, respectively, for satu-
rated liquid water and compressed liquid water up to 100 MPa.
Cumulative water enthalpy data with the corresponding differ-
ences and related percentages are presented in Table 3 for a wide
range of temperatures between triple and critical points, and, a
wide range of pressures from saturation up to 100 MPa. Note that
the tabulated enthalpy differences with correction is based on Eq.
共5兲, i.e., using correction ⌬hcorr ⬇ ⌬h⌬P only, as generally recom-
mended in the literature, while corrections ⌬h⌬u, and ⌬h⌬v are Fig. 1 Isothermal and isentropic compression of saturated liq-
neglected 关1,2兴. uid water
The corrections in Tables 1 and 3 are calculated using the fol-
lowing equations:
pressure increase due to a strong decrease of internal energy, mak-
⌬h⌬u = ⌬u = u共P,T兲 − usat共T兲共6兲
ing the recommended positive correction, ⌬hcorr, to be erroneous
冕冋册
N−1 and thus counterproductive and misleading, see Tables 2 and 3.
The ⌬h⌬u correction is not included in Eq. 共5兲 even though its
兺
1
⌬h⌬v = P · dv ⬵ 共Pi + Pi+1兲 · 共vi+1 − vi兲共7兲
2 magnitude may be, and sometimes is, larger than the included
i=1sat
⌬hcorr ⬇ ⌬h⌬P correction. Therefore, the recommended enthalpy
冕兺冋册
N−1 correction for isothermal compression in the literature is appropri-
1 ate for the isentropic processes 共see additional justification be-
⌬h⌬P = v · dP ⬵ 共vi + vi+1兲 · 共Pi+1 − Pi兲 ⬇ ⌬hcorr 共8兲
2 low兲, but not appropriate for isothermal processes, although it
i=1sat
may sometimes be beneficial, due to an erroneous assumption that
Note that in Eq. 共5兲 the fluid is assumed to be incompressible, internal energy is not, and enthalpy is, dependent on pressure. It is
while in Eqs. 共6兲–共8兲 variability in v, although small, is taken into quite the opposite in Table 1, see how the corresponding values
account, thus the differences in ⌬hcorr and ⌬h⌬P values in Table 1. 共u and h兲 change with pressure at constant temperature of 260° C.
Furthermore, for the isothermal compression, the correction ⌬h⌬v The above physical justifications could be confirmed using the
is much smaller than the corrections ⌬h⌬u and ⌬h⌬P, and thus it corresponding differential property correlations obtained using the
may be neglected. However, corrections ⌬h⌬u 共negative兲 and cor- Maxwell’s relations 关1兴:
rections ⌬h⌬P 共positive兲 are comparable in magnitude but oppo-
site in sign, so it is better not to take the corrections, as in Eq. 共2兲,
than to take only the correction ⌬hcorr ⬇ ⌬h⌬P, as in Eq. 共5兲. It will
be shown below that the recommended enthalpy correction for the
isothermal compression 关1,2兴, is actually valid for isentropic
compression.
It appears from real data values in Tables 1 and 2 that the
magnitude of the negative correction ⌬h⌬u 共i.e., ⌬u兲 increases
with both pressure and temperature, while positive correction
⌬h⌬P 共i.e., 兰vdP兲 depends mostly on pressure since the specific
volume, v, does not change significantly. As is evident from data
in Table 3, the correction, Eq. 共5兲关1,2兴, as a general improvement 共9兲
for compressed liquid enthalpy calculation, is not justified in gen-
eral, but for smaller temperatures only 共less than 200° C兲. How-
ever, it is more erroneous 共and thus counterproductive and mis-
leading兲 for higher temperatures 共above 200° C兲 and particularly
at higher pressures, where a simple approximation 共Eq. 共2兲兲, with-
out any correction, is more accurate.
In Fig. 1, the compression of saturated liquid water, state 共f sat兲
at 260° C, is presented for isothermal compression to state 共T50兲
at 50 MPa, and for isentropic compression to state 共s50兲 to the
same pressure of 50 MPa 共see Table 1 for property data兲.
Since du = ␦q − ␦w 共where ␦w = Pdv兲, and real liquids 共in this
case water兲 are not exactly incompressible, then during isentropic
共10兲
compression 共␦q = Tds = 0兲 there will be some negative work 共Pdv兲
and an increase of internal energy and temperature 共in this case for From Eq. 共9兲 it is evident that change in internal energy for the
11.7° C, from 260° C to 271.7° C兲关3兴. However, during the iso- isothermal process is zero 共du = 0兲 only for ideal incompressible
thermal compression, to cool and maintain constant water tem- fluids 共dv = 0兲, but for real liquids 共dv ⫽ 0兲 the bracketed term
perature, there must be some heat transfer out, q, and in the pro- with dv in Eq. 共9兲 is not zero since real liquids are not exactly
cess the internal energy, u, will be decreased with pressure incompressible.
increase at constant temperature 共see the corresponding data in Equation 共10兲 confirms that pressure correction 共vdP兲 is always
Tables 1 and 2; note ⌬h⌬u = ⌬u ⬍ 0 if ⌬P ⬎ 0兲. At high tempera- appropriate for isentropic processes 共dh = Tds + vdP = vdP for s
tures, 300° C and above, even specific enthalpy is decreasing with = const兲, but not for isothermal processes as given in many refer-
Journal of Heat Transfer MAY 2006, Vol. 128 / 425
ences 关1,2兴. For example, for an isentropic compression of uids in general, the internal energy is less dependent on pressure
liquid water from saturation at 260° C to 50 MPa, see Fig. 1 than enthalpy. The literature recommendations may be erroneous
and data in Table 1 共Psat = 4.69 MPa, hsat = 1135.0 kJ/ kg, ssat and thus counterproductive and misleading, as is the case for liq-
= 2.8849 kJ/ kg K兲, the temperature and enthalpy will increase to uid water at higher temperatures and pressures. For intermediate
271.7° C and hs50 = 1191.7 kJ/ kg, respectively 关3兴. The isentropic pressures and temperatures, the enthalpy corrections recom-
enthalpy increase, hs50 − hsat = 共1191.7− 1135.0兲 kJ/ kg= 56.7 kJ/ mended in the literature are unnecessary, since the errors are about
kg is virtually identical to the corresponding enthalpy correction the same in magnitude 共but opposite in sign兲 as if the correspond-
共⌬h⌬P = 兰vdP = 56.05 KJ/ kg兲 ⬇ 共⌬hcorr = 57.82 kJ/ kg兲, where the ing saturated enthalpy values without any corrections were used.
small differences in Table 1 are due to an integral approximation The isothermal corrections are only beneficial for very high pres-
sures at smaller temperatures 共below 200° C for liquid water兲.
with discrete summation 共Eq.共8兲兲 or assuming v = vsat = const in
In summary, the recommended pressure corrections in the clas-
Eq. 共5兲 during the process.
sical reference textbooks for isothermal, liquid enthalpy approxi-
For the isothermal process the enthalpy pressure-correction
mation are not appropriate and often insignificant, unnecessary or
could be evaluated using the three corrections, Eqs. 共6兲–共8兲, or
more erroneous than the simple approximation using the corre-
integrating Eq. 共9兲 or 共10兲, or combined Eqs. 共9兲 and 共10兲, i.e.:
sponding saturated values. Furthermore, it is shown here that the
recommended pressure correction for isothermal processes are
actually valid in general for enthalpy correction for isentropic
processes.
⫽
Nomenclature
c , c p , cv ⫽ specific heat, at constant pressure, or constant
volume
csound ⫽ speed of sound
⌬h ⫽ enthalpy correction including all relevant cor-
共11兲 rections, Eq. 共12兲
Since the Joule-Thomson coefficient is defined as ␮JT = 共⳵T / ⳵ P兲h, ⌬hcorr ⫽ enthalpy correction due to the change of ⌬Psat
then Eq. 共11兲 could be expressed as: = P − Psat, Eq. 共5a兲
⌬h⌬u ⫽ enthalpy correction due to the change of ⌬u,
关⌬h兴T = 关h共P,T兲 − hsat共T兲兴T = ⌬h⌬u + ⌬h⌬v + ⌬h⌬P
Eq. 共6兲
冕 ⌬h⌬v ⫽ enthalpy correction due to the change of ⌬v,

P
= 关− C P␮JT兴dP Eq. 共7兲
Psat
⌬h⌬P ⫽ enthalpy correction due to the change of ⌬P,
冋册 Eq. 共8兲
N−1
兺
1
⬵ 共C P␮JT,i + C P␮JT,i+1兲共Pi+1 − Pi兲共12兲 h ⫽ liquid enthalpy at any P and T
i=1sat 2 hcorr ⫽ liquid enthalpy approximation with vsat⌬Psat
Note that real data values for both, the ␮JT and CP are available correction as recommended in the literature
关3兴, and for the isothermal enthalpy correction, ⌬h, Eq. 共12兲 was hsat ⫽ saturation liquid enthalpy
used instead of Eqs. 共6兲–共8兲, see Table 1. P ⫽ pressure
Compressed liquid water enthalpies 共h兲, different corrections, q ⫽ heat transfer per unit of mass
their approximation differences 共hsat − h, and hcorr − h兲 and the cor- s ⫽ entropy
responding error percentages, with regard to the saturation value, T ⫽ temperature
are presented in Table 3, without any correction 共Eq. 共2兲兲 and the u ⫽ internal thermal energy
correction recommended in the literature 共⌬hcorr ⬇ ⌬h⌬P, Eq. 共5兲兲, v ⫽ specific volume
for a wide range of temperatures and pressures. The corrections, w ⫽ boundary work per unit of mass
as recommended in 关1,2兴共Eq. 共5兲兲, are only beneficial for higher ␮JT ⫽ The Joule-Thomson coefficient
pressures at smaller temperatures, insignificant for smaller pres- Subscripts
sures at most of the temperatures, about the same but opposite s50 ⫽ isentropic compression from saturated to
sign 共thus unnecessary兲 for intermediate temperatures and pres- 50 MPa
sures, and more erroneous 共thus counterproductive and mislead- sat ⫽ for saturated liquid
ing兲 for higher temperatures and pressures, than the simple ap- T50 ⫽ isothermal compression from saturated to
proximation without any correction 共Eq. 共2兲兲, as seen from Tables
50 MPa
1 and 3.
3 Conclusion References
An analysis with physical justification, supported by water real 关1兴 Cengel, Y. A., and Boles, M. A., 2002, Thermodynamics, An Engineering
Approach, 4th ed., McGraw-Hill, New York, Sec. 2.5.3, p. 84 and Chap. 11,
enthalpy data, regarding liquid enthalpy approximation, is pre- pp. 603–626.
sented here. The analysis shows, and a conclusion is drawn, that 关2兴 Moran, M. J., and Shapiro, H. N., 2000, Fundamentals of Engineering Ther-
the recommendation in the classical reference textbooks for im- modynamics, 4th ed., Wiley, New York, Sec. 3.3.6, p. 109.
provement of enthalpy calculation of compressed liquids, by ac- 关3兴 Lemmon, E. W., McLinden, M. O., and Friend, D. G., 2005, “Thermophysical
Properties of Fluid Systems,” NIST Chemistry WebBook, NIST Standard Ref-
counting for pressure dependence, is not generally valid for iso- erence Database No. 69, P. J. Linstrom, and W. G. Mallard, eds., National
thermal processes. The misconception in the literature is a result Institute of Standards and Technology, Gaithersburg, MD 共http://
of the erroneous assumption that, due to incompressibility for liq- webbook.nist.gov/chemistry/fluid兲.
426 / Vol. 128, MAY 2006 Transactions of the ASME
A Model for Condensate
Retention on Plain-Fin Heat
A. I. ElSherbini2
e-mail: asherbini@kisr.edu.kw
Exchangers1
Building and Energy Technologies Department, A model has been developed for predicting the amount of condensate retained as drops
Kuwait Institute for Scientific Research, on the air-side of heat exchangers operating under dehumidifying conditions. For a coil
P.O. Box 24885 Safat, 13109, Kuwait with a given surface wettability, characterized by the advancing contact angle, the maxi-
mum diameter for a retained drop is obtained from a balance between gravitational and
A. M. Jacobi surface tension forces. A logarithmic function is used to describe the size-distribution of
e-mail: a-jacobi@uiuc.edu drops on fins, based on the fraction of fin-area covered by liquid. The volumes of indi-
Department of Mechanical and Industrial vidual drops are calculated by a geometric method for approximating the three-
Engineering, dimensional shapes of drops on vertical and inclined surfaces. The total volume of con-
University of Illinois at Urbana-Champaign, densate accumulated on a coil is then found by multiplying the size-distribution and
1206 W. Green St., Urbana, IL 61801 volume functions and integrating over all drop diameters. The model is successful in
predicting measurements by other researchers of the mass of condensate retained on
plain-fin heat exchangers. The critical fin spacing to avoid the formation of condensate
bridges is also predicted. 关DOI: 10.1115/1.2175091兴
Keywords: condensation, retention, heat exchangers
Introduction ditions 共see for example 关7–9兴兲. The data in the literature suggest
that the quantity of retained condensate and its mode play signifi-
In air conditioning and refrigeration systems, evaporators usu-
cant roles in characterizing the thermal and hydraulic performance
ally operate below the dewpoint of conditioned air. Thus, under
of heat exchangers under wet conditions. Therefore, developing a
normal air-cooling and dehumidification conditions, water con-
model that can predict condensate retention on heat exchangers
denses and accumulates on the air-side surfaces of heat exchang-
will provide a useful tool for designing heat exchangers for wet
ers. Condensate retained on an evaporator affects the indoor air
operation.
quality by providing a medium for biological activity. In addition,
Some researchers 关10,11兴 presented models to calculate the
condensate retention can have a significant impact on the perfor-
amount of condensate retained on finned-tubes, in order to deter-
mance of the heat exchanger.
mine the condensation heat transfer coefficient for pure vapors.
The effects of condensate retention on the thermal and hydrau-
Jacobi and Goldschmidt 关7兴 studied moisture condensation on
lic performance of heat exchangers have been studied by many
finned-tubes and proposed a model for condensate bridging that
researchers. Retained condensate generally causes the friction fac-
helped explain coil performance at different airflow rates. These
tor to increase. For example, Yin and Jacobi 关1兴 found the friction
simple models, however, were restricted to annular geometry and
factor for a plain-fin coil under wet conditions to be 70% higher
the filmwise mode of the condensation.
than that for a dry coil. The degradation in hydraulic performance
Korte and Jacobi 关6兴 developed a model for predicting the
is less pronounced at a large fin spacing and high Reynolds num-
amount of condensate retained on plain-fin heat exchangers with
bers, Re 关1–6兴. Water drops are believed to roughen the heat ex-
wide fin spacings. The volume of retained condensate was ob-
changer surfaces and increase friction factors. Also, condensate
tained by integrating a size distribution function, multiplied by the
removal from the surface is harder at smaller fin spacing and
volumes of individual drops, over all drop sizes, as will be ex-
lower Re, especially when condensate “bridges” form. The in-
plained later. A balance of forces was used to find the maximum
creased retention results in more blockage of the air flow path and
drop diameter, Dmax, to set the integration limit. The model
higher friction. Condensation may increase or decrease heat trans-
showed reasonable success in predicting condensate retention on a
fer, depending on the heat exchanger geometry and operating con-
heat exchanger tested by Korte and Jacobi. However, simplifica-
ditions. For plain-fin coils, condensate decreases sensible heat
tions in calculating the shapes and size distributions of drops have
transfer at low Re, but it increases heat transfer at high Re. Wang
limited the accuracy and applicability of the model.
et al. 关5兴 reported such a behavior with the transition occurring at
The objective of this article is to develop a model for predicting
Re= 2000 based on tube diameter. Wet conditions enhance heat
the steady-state mass of condensate retained as drops on heat ex-
transfer at larger fin spacing, f s, and degrade it at smaller f s. Yin
changers. To achieve this goal, the volumes of drops held on the
and Jacobi 关1兴 observed a 55% decrease in heat transfer under wet
fins of a heat exchanger need to be reliably determined under
conditions for a plain-fin coil with a fin spacing of 1.3 mm. The various surface conditions. Then, the steady-state distribution of
roughness caused by droplets is thought to result in the heat trans- drop sizes on a fin surface is considered. Finally, the total mass of
fer enhancement, while bridging at low Re and narrow fin spacing condensate is found to close the model.
can be responsible for the degradation. Heat exchangers with
other geometries have also been tested under dehumidifying con-
Retention Model
1
Presented at the 2004 ASME Heat Transfer/Fluids Engineering Summer Confer- Consider a plain-fin heat exchanger operating under dehumidi-
ence 共HT-FED2004兲, July 11, 2004–July 15, 2004, Charlotte, NC. fication conditions. At steady state, the mass of condensate re-
2
Corresponding author.
Contributed by the Heat Transfer Division of ASME for publication in the JOUR-
moved from the coil by drainage or fly-out is equal to the mass
NAL OF HEAT TRANSFER. Manuscript received January 23, 2005; final manuscript re- added by condensation. Thus, the amount of water retained on the
ceived October 19, 2005. Review conducted by A. Haji-Sheikh. heat exchanger is constant. Measurements of retained condensate
关1,6兴 show that the mass reaches steady-state values. For wide fin
spacing, where inter-fin bridges cannot form, the total mass of
retained condensate can be expressed as
M = ␳w 冕冕
AT D
V共D兲n共D兲dDdA 共1兲
where ␳w is the density of water, V共D兲 is the volume of a droplet

with diameter D, and n共D兲 is the number of droplets with diam-
eter D per unit area and per droplet diameter. The integration is
performed for all drop sizes over the entire surface area of the heat
exchanger, AT. This formulation of the condensate retention
model, reduces the problem into finding the volume function, V,
and the size-distribution function, n, for drops. These functions
are written in terms of D to emphasize the dependence on diam-
eter; however, they also depend on other variables, such as the
wettability 共contact angles兲 and inclination angle of the heat trans-
fer surfaces. The maximum diameter, Dmax, is also needed as a
limit for the integration in Eq. 共1兲, and will prove useful when
calculating n共D兲. In the next sections, we discuss the volume and
size-distribution functions.
Volume Function
Finding the volume of a drop requires knowledge of its shape.
For a small drop on a horizontal fin, the shape can be defined by
a spherical cap. However, a large drop on a vertical surface cannot
be described by a simple function. The shapes of liquid drops on
vertical and inclined surfaces have been studied by many investi- Fig. 1 „a… Profile of a liquid drop fit by parts of two circles. „b…
gators, as in 关12–15兴. These studies disagreed on key parameters Contact line at the base of the drop. The two-circle method
affecting drop shapes, such as contact angles and contact lines, as approximates the profiles and volumes of drops on vertical or
explained in 关16兴. Moreover, they did not provide a reliable func- inclined fins.
tion relating the volume of a drop to its diameter and to surface
conditions. In a related paper 关17兴, we developed a method for
approximating the shapes of liquid drops on vertical and inclined
surfaces. The method is briefly explained below. and ␪2. Thus, the two-circle method provides an approximation
for the volume of a drop in the form
Two-Circle Model. Consider a liquid drop on a fin surface that
冕
is inclined an angle ␣ from the horizontal. For an arbitrary cross
section taken at an azimuthal angle, ␾, the profile displays two V= F共␨共␾兲, ␪共␾兲兲d␾ 共4兲
contact angles, ␪1 and ␪2, as in Fig. 1. The two-circle method
approximates the profile at the given cross section by parts of two
circles sharing a common tangent, as shown in the figure. The where ␨共␾兲 is a function describing the contact line 共contour兲 of
drop volume is found by integrating all profiles along the circum- the drop, and ␪共␾兲 describes the contact-angle variation along the
ference of the drop base. The total volume is divided into two drop circumference. The method was validated against experi-
parts, mental measurements.
V = VI + VII 共2兲 Contact Angle and Contour Functions. Investigators have
The integrals can be expressed as predicted different forms for the function ␪共␾兲, which describes
冕冋
␲/2 the contact angle, ␪, as it varies with the azimuthal angle, ␾, for a
共y b − Y 0兲3 drop on an inclined surface 共see for example 关11,12,15兴兲. Also,
VI = 共2 − 3 cos ␪1 + cos3 ␪1兲 + X0共y b − Y 0兲2␪1
0
3 several shapes have been reported for the contour, ␨共␾兲, at the
base of the drop 关16兴. We have conducted experiments, reported in
2 2
− X0共y a − Y 0兲共y b − Y 0兲sin ␪1 + X20共y b − y a兲 − y 3b + y 2by c 关18兴, to investigate the general shapes of these two functions.
3 3 Based on observations and analysis, it was concluded that the
冉冉冊冊册
contact-angle variation is best fit by a third-degree function that
X20 ␲ yc takes the form
− X20y b + y c + X0y 2b − sin−1 d␾ 共3a兲
3 2 yb
and cos ␪max − cos ␪min 3 cos ␪max − cos ␪min 2
cos ␪共␾兲 = 2 ␾ −3 ␾
冕冋冉冊 ␲3 ␲2
3␲/2
2 2 X20 y 3a
VII = yc yb + − y 2by a + − X20y a + X0y ay b sin ␪2 + cos ␪max 共5兲
␲
3 3 3
冉冉冊冊册
− X0y 2b sin−1
yc
yb
␲
− + ␪2
2
d␾ 共3b兲
where ␪max and ␪min are the maximum and minimum contact
angles in a drop, respectively. Such angles are observed in the
symmetry plane 共␾ = 0 in Fig. 1兲 of the drop.
where the variables y a, y b, y c, X0, and Y 0 are as shown in Fig. 1. The drop contour was found to be best fit by an ellipse. Let L
Expressions for these variables have been developed in 关17兴 in and w denote the major and minor axes of the ellipse, respectively.
terms of the radius of the drop base, ␨, and the contact angles, ␪1 The radius of the ellipse is described as
Fig. 2 Minimum contact angle, normalized by the maximum
angle, as it varies with the Bond number for different liquids
and surfaces. A single relation fits the data with a coefficient of
determination r2 = 0.90.
L
␨共␾兲 = 共6兲
冑cos2 ␾ + ␤2 sin2 ␾
where ␤ is the aspect ratio 共␤ = L / w兲. The ellipse can also be Fig. 3 A force balance on a drop on an inclined fin, used to
find the maximum drop diameter
described by an equivalent diameter, D, which yields the same
contour area, and its aspect ratio. Using the equivalent diameter,
the contour equation becomes
␨共␾兲 =
D
2
冑 ␤
cos ␾ + ␤2 sin2 ␾
2 共7兲
␪min
␪A
= 0.01 Bo2 − 0.155 Bo + 0.97 共11兲
Equations 共5兲 and 共7兲 define the inputs needed for the volume Similarly, an empirical relation was obtained for the aspect ratio
function in Eq. 共4兲. Thus, the volume of a drop can be found, of the contour,
given its diameter, contour aspect ratio, and maximum and mini-
mum contact angles,
␤ = 1 + 0.096 Bo 共12兲
V= 冕 H共D, ␤, ␪max, ␪min兲d␾ 共8兲

This relation compares well to measurements from the litera-
ture. More details can be found in 关18兴. Equations 共10兲–共12兲 can
now be used to find the volume of any droplet on the heat ex-
Contact Angle and Contour Values. In order to predict the changer. Consider a plain-fin evaporator with surface wettability
volumes of all drops, the variables ␤, ␪max, and ␪min need to be characterized by the advancing contact angle ␪A. For any droplet
determined at different conditions. These variables were measured of diameter D, the Bond number can be calculated from Eq. 共9兲,
and reported 关18兴 for a range of drop sizes on surfaces with vari- and the values of ␪max, ␪min, and ␤ can be found from Eqs.
ous inclination angles and contact angles. Two liquids were tested 共10兲–共12兲. The volume of the drop is obtained by a numerical
on eight surfaces to cover a wide range of advancing and receding integration of Eq. 共8兲, which is based on the two-circle approxi-
contact angles. Notice that ␪max and ␪min refer to the contact mation. Thus, the V共D兲 function in Eq. 共1兲 is determined. The next
angles in a drop, while the advancing and receding contact angles step in the retention model is to find the maximum drop diameter
refer to those angles characteristic of the liquid-surface combina- that can be held on the heat-exchanger surface.
tion. The results were presented in terms of the Bond number, Bo,
which is a measure of gravitational to surface tension forces,
␳wgD2 sin ␣ Maximum Drop Diameter
Bo = 共9兲 In order to set the limits of the integration in Eq. 共1兲, the mini-
␥
mum and maximum drop diameters, Dmin and Dmax, are required.
where ␳w is the liquid density, g is the acceleration of gravity, D is Typical values measured for Dmax are in the order of millimeters.
the equivalent drop diameter, ␣ is the surface inclination angle, The lower limit, Dmin, is several orders of magnitude smaller, and
and ␥ is surface tension of the liquid-vapor interface. For a given is conveniently taken as 10 ␮m. Smaller sizes do not add signifi-
liquid drop at constant temperature, an increase in Bo indicates an cant value to the area covered by drops or their total volume. The
increase in drop diameter and/or surface inclination. The maxi- maximum diameter is obtained from a balance of forces acting on
mum contact angle in a drop, ␪max, was found to be almost equal a drop on a fin surface, as in Fig. 3. Earlier experiments show
to the advancing contact angle, ␪A, for all tested conditions air-flow forces to have small effects on condensate retention over
flow rates typical to the current conditions 关6兴. Therefore, drag
␪max = ␪A 共10兲 forces will be neglected and the balance is performed between
The minimum contact angle of a drop was found to decrease as gravitational and surface-tension forces acting in the x-z plane.
the Bond number is increased for various drop sizes and surface- The surface tension force acting on the drop of Fig. 3 is found
inclination angles. In addition, the ratio 共␪min / ␪A兲 followed a by integrating the differential forces over the contour of the drop.
single curve for all tested surfaces, as shown in Fig. 2. This be- Due to symmetry, the components of the surface-tension force in
havior allowed describing the minimum contact angle by the gen- the z direction will cancel; the resulting surface-tension force acts
eral relation in the x direction and can be calculated from
冕
␲
Fsx = − 2␥ ␨ cos ␪ cos ␾d␾ 共13兲
0
where ␥ is the liquid surface tension, and ␪ and ␾ are as defined

earlier. The surface tension is assumed to be constant, and changes
associated with the temperature distribution of the fin are ne-
glected. By substituting for cos ␪ from Eq. 共5兲 and using the
equivalent diameter of the drop, D, instead of 2␨, the integration
leads to 共see 关19兴兲
24
Fsx = ␥D共cos ␪min − cos ␪A兲共14兲
␲3
where ␪max = ␪A.
The gravitational force acting on the drop of Fig. 3 in the Fig. 4 Size-distribution functions for drops condensing on
x-direction can be written as plane surfaces, obtained by different researchers
Fgx = − ␳wgV sin ␣ 共15兲

which can be rearranged, using an equivalent diameter, to be
n b = Q bD Bb for 0.2Dmax 艋 D 艋 Dmax 共20b兲
V
Fgx = − ␥D Bo 3 共16兲 where ns and nb denote distributions for small and big drops,
D respectively, and Qs, Bs, Qb, and Bb are constants. Previous inves-
A force balance on the drop in the x direction is obtained by tigators have associated different values to these constants. We
setting will present a simple approach for finding the best values for such
constants. Our model will follow prior work and neglect edge
Fgx + Fsx = 0 共17兲 effects in describing droplet distributions. Although the fin edges
ElSherbini 关16兴 showed that the force balance leads to a relation can be expected to influence droplet sizes, the number of droplets
that takes the form near the edges is small in comparison to the total number of drop-
lets on the fin.
␪min = G共␪A,Bo兲. 共18兲 Figure 4 compares the size-distribution functions obtained by
If the volume of the drop is calculated based on approximating its different researchers. The figure shows that earlier studies of ns
shape as a spherical cap, Eq. 共18兲 can be expressed as agreed on the slope, Bs, of the distribution function for small
drops. In fact, the distribution of Graham 关20兴 and one of the
cos ␪min = cos ␪A +
576
冉
␲4 2 − 3 cos ␪av + cos3 ␪av
sin3 ␪av
冊Bo 共19兲
distributions of Tanasawa and Ochiai 关22兴 resulted in the same Bs
value of −2.73. Adopting this value, three other constants 共Qs, Qb,
Bb兲 still need to be determined. In the present analysis, these con-
where ␪av = 共␪min + ␪A兲 / 2. A more reliable version of Eq. 共18兲 uses stants will be considered as parameters depending on the maxi-
the two-circle method to find the drop volume. Equations 共18兲 and mum diameter and the surface area covered by drops. The fraction
共11兲 can be solved simultaneously to obtain the critical values for of area covered by small drops can be found from
Bo and ␪min. The maximum Bond number, Bomax, is then used to
冕冉冊
0.2Dmax
find Dmax from Eq. 共9兲. The critical drop sizes found by this analy- ␲D2
Acov,s = ns dD 共21a兲
sis agreed well with measurements reported in the literature, as 4
Dmin
shown in 关19兴.
and the fraction of area covered by large drops is
冕冉冊
Dmax
␲D2
Size-Distribution for Drops Acov,b = nb dD 共21b兲
0.2Dmax
4
Drops condensing on a vertical or inclined surface grow by
It is important to use consistent units in Eq. 共21兲. The fraction
condensation and coalescence until they reach the maximum di-
of area covered by drops is the total area of drops per unit area,
ameter, if they are not swept by other sliding drops. At maximum
which is dimensionless. The size distribution function is a number
size, the force due to gravity overcomes surface-tension forces
and the drop starts sliding. The work of early researchers 关20–23兴 per unit area per diameter, with dimensions of 共length兲−3. It is
showed that the distribution of drop sizes can be described by a common to describe n as the number of drops per cm2 per ␮m and
steady-state function which has a logarithmic nature. The logarith- D in ␮m. Using these units, Eq. 共21兲 can be integrated to become
mic function changes its slope when drop sizes become large and ␲ Qs
the drop growth mechanism changes. The growth of small drops is Acov,s = 关共0.2Dmax兲Bs+3 − Dmin
Bs+3
兴共22a兲
4共10兲8 Bs + 3
dominated by condensation, while the growth of large drops is
more influenced by coalescence. Other distribution functions were and
also proposed 关24–26兴. They agreed with measurements for small
␲ Qb
drops, but did not capture the change in distribution for large Acov,b = 关DBb+3 − 共0.2Dmax兲Bb+3兴共22b兲
drops. Graham 关20兴 noticed that the change in slope for the size- 4共10兲 Bb + 3 max
8
distribution function occurred at 共D / Dmax兲 of 0.2. Yin and Jacobi The continuity of the size-distribution function requires that ns
关1兴 confirmed that finding. The measurements of Tanasawa and = nb at D = 0.2Dmax. Therefore,
Ochiai 关22兴 indicate that the change occurs when 共D / Dmax兲 is in
the vicinity of that number 共0.18–0.26兲. Therefore, the size distri- Qb = Qs共0.2Dmax兲共Bs−Bb兲共22c兲
bution function can be written as Using the maximum drop diameter, Dmax, and the fractions of area
covered by small and large drops, Acov,s and Acov,b, as inputs, Eq.
n s = Q sD Bs for Dmin 艋 D 艋 0.2Dmax 共20a兲
共22兲 can be solved for the three unknowns: Qs, Qb, and Bb. The
and maximum diameter is determined from a force balance as de-
scribed in the previous section. The area covered by drops is an
important parameter for n that needs to be found.
The fraction of area covered by small and large drops calcu-
lated from 关20–23兴 ranged between 0.75 and 0.85, with maximum
drop diameters of 2–2.9 mm. However, the test surfaces used
were circular with diameters of only 2–4 cm. Fin-and-tube heat
exchangers typically have much larger lengths, where falling
drops sweep greater areas of the fin surfaces. Therefore, the frac-
tional area covered by drops, Acov, should be smaller for such heat
exchangers. Yin and Jacobi 关1兴 measured the area covered by
drops at three locations on a heat-exchanger fin with a length of
20 cm. The average fractional area covered by drops was 0.34,
with a maximum diameter of 4.2 mm. The average Acov for the
small and large portions of the size distribution function were 0.16
and 0.18, respectively. These values are considered representative
of the geometry and conditions of fin-and-tube heat exchangers,
and will be used to compare the predicted condensate retention to
measurements for coils with fin lengths of 15 and 27 cm.
The analysis presented in this section generalizes the size- Fig. 5 Heat-exchanger geometry used for comparing the
distribution function by allowing the constants of its two portions model to measurements of retained condensate
to vary for different conditions. The fractional areas covered by
small and large drops, along with the calculated value of Dmax,
can be used in Eq. 共22兲 for obtaining the constants defining ns and per unit area should remain constant as long as bridging does not
nb. Defining these size-distribution functions completes the model take place. Korte and Jacobi 关6兴 presented condensate retention
for predicting the mass of retained condensate. measurements for a fin-and-tube heat exchanger with a fin spacing
of 6.4 mm. The coil had vertical aluminum fins, as shown in Fig.
Mass of Retained Condensate 5, with an advancing angle, ␪A, of about 90 deg. At very low air
The mass of condensate retained as drops on a heat exchanger velocity, the condensate per unit area becomes 129 g / m2. Shin
is calculated from Eq. 共1兲, which can be divided into two parts in and Ha 关27兴 also tested an aluminum heat exchanger with the
order to represent small and large drops, same advancing contact angle and a fin spacing of 5.4 mm at a
冕冕
0.2Dmax
low face velocity of 0.6 m / s. Figure 6 compares the measure-
ments of Korte and Jacobi 关6兴 and Shin and Ha 关27兴 to the pre-
M s = ␳w V共D兲ns共D兲dDdA 共23a兲 dictions of the present retention model and the model of Korte and
AT Dmin
Jacobi. The new model succeeds in predicting the two indepen-
and dent measurements of condensate retention. The simplified
冕冕
Dmax model of Korte and Jacobi overpredicts the mass of the retained
M b = ␳w V共D兲nb共D兲dDdA 共23b兲 condensate.
AT 0.2Dmax
The success of the current model confirms the validity of its
components. The predictions of drop volumes, contact angles,
where maximum diameters, and retentive forces had been validated
M = Ms + Mb 共23c兲 separately in earlier works. When considered together, the model
predicts the mass of condensate with excellent accuracy.
The procedure for calculating the mass of condensate retained
on a heat exchanger can now be summarized. The inputs to the Predicting Condensate Bridges
model are the inclination angle of the coil, ␣, measured from the
horizontal, and the advancing contact angle, ␪A.3 The fractional The proposed retention model is applicable for heat exchangers
areas covered by small and large drops, Acov,s, and Acov,b, are with small fin densities. As the fin density increases, drops from
considered known for the given wettability and dimensions of the adjacent fins coalesce together to form bridges and significantly
coil surfaces. The maximum Bond number is obtained by numeri- change the distribution of condensate on the coil. Drop-shape
cally searching for the solution to Eqs. 共11兲 and 共18兲. The maxi- analysis can be used to develop an equation for the critical fin
mum diameter is then calculated from Eq. 共9兲. The minimum di- spacing for condensate bridging to occur. The height-to-diameter
ameter, Dmin, is taken as 10 ␮m, and Bs = −2.73. The values of ratio for a drop is expressed as 共see 关16兴兲,
Dmax, Acov,s and Acov,b are used in Eq. 共22兲 to obtain the constants
defining the size-distribution functions. The integrations in Eqs.
共23a兲 and 共23b兲 are performed for diameters between the lower
and upper limits. For each increment in diameter, Eqs. 共9兲 and
共12兲 are used to find the contact angles and aspect ratio, which are
substituted into Eq.共8兲 and integrated to calculate the volume cor-
responding to that diameter. The volume is multiplied by the size
distribution for that diameter and the integration continues. The
mass of condensate retained per unit area is finally found from Eq.
共23c兲.
The model developed in this work is used to compare predic-
tions of condensate retention to experimental data. Notice that
changing the fin spacing may change the total area of the heat
exchanger and total mass of retained condensate, but the retention
Fig. 6 Predicted mass of condensate retained on heat ex-
3
The receding contact angle, ␪R, is not required, but can be incorporated—if changers compared to measurements reported by Shin and Ha
know—into calculating Dmax. †27‡ and Korte and Jacobi †6‡, and to an earlier model
volumes of individual drops are found using the two-circle ap-
proximation. The maximum diameter and fractional areas covered
by drops are used to define the logarithmic functions describing
the distribution sizes for drops. The total amount of retained con-
densate is found by integrating the volume and size-distribution
functions over drop diameters. The model successfully predicts
the condensate retention measurements by other researchers. The
critical fin spacing to avoid bridges is also calculated from drop
geometry.
Nomenclature
Acov ⫽ fraction of area covered by drops
AT ⫽ total surface area of a heat exchanger
B ⫽ constant in Eq. 共20兲
Fig. 7 Critical fin spacing, beyond which condensate bridges
cannot occur, as a function of the advancing contact angle Bo ⫽ Bond number
D ⫽ equivalent diameter of drop
F,G,H ⫽ functions
h 冑␤ fs ⫽ fin spacing
冉冊冉冊
= 共24兲 Fs ⫽ surface tension force
D sin ␪max sin ␪min Fg ⫽ gravitational force
+
1 − cos ␪max 1 − cos ␪min g ⫽ acceleration of gravity
h ⫽ drop height
where ␤ is the aspect ratio of the drop contour, and ␪max and ␪min
are the maximum and minimum contact angles, respectively. For L ⫽ major axis of ellipse
drops on two fins to coalesce into a bridge, the fin spacing cannot M ⫽ total mass
exceed the value of 2hmax, where hmax is the height of the largest n ⫽ size distribution of drops 共number per unit
drop that can be retained on a fin. Thus, the critical fin spacing is area, per diameter兲
Q ⫽ constant in Eq. 共20兲
2Dmax冑␤ r2 ⫽ coefficient of determination
冉冊冉冊
f s,cr = 共25兲 Re ⫽ Reynolds number
sin ␪A sin ␪R
+ V ⫽ volume
1 − cos ␪A 1 − cos ␪R w ⫽ minor axis of ellipse
where ␪A and ␪R are the advancing and receding contact angles, X 0, Y 0 ⫽ variables in Fig. 1
respectively, that are characteristic of the fin surface. Heat ex- y a, y b, y c ⫽ variables in Fig. 1
changers with fin spacings greater than f s,cr will not have conden-
sate bridges. Greek Symbols
The critical fin spacing, f s,cr, sets the limit for the applicability ␾ ⫽ azimuthal angle
of the current condensate retention model. It also provides a tool ␣ ⫽ inclination angle of a surface
for designing heat exchangers for wet conditions of operation. In ␤ ⫽ aspect ratio of contour
order to avoid condensate bridging, which degrades coil perfor- ␥ ⫽ surface tension of the liquid-vapor interface
mance, a fin spacing larger than f s,cr should be used. Figure 7 ␨ ⫽ radius of contour
shows the variation of f s,cr with advancing contact angle, where ␪ ⫽ contact-angle
the variables ␪R, Dmax, and ␤ are calculated based on analysis of ␪ 1, ␪ 2 ⫽ contact angles, see Fig. 1
critical-size drops. The critical fin spacing increases with the ad- ␳w ⫽ mass density of water
vancing angle to a maximum, and then it decreases. This behavior Subscripts
of f s,cr follows the pattern of the maximum drop size, since f s,cr is A ⫽ advancing
proportional to Dmax 共Eq. 共25兲兲. Critical drop sizes are discussed b ⫽ for big drops
in more detail in another work 关19兴. It is interesting to note that cr ⫽ critical
both hydrophilic and hydrophobic coatings can be used to reduce max ⫽ maximum
f s,cr and prevent the formation of bridges. Retention measure- min ⫽ minimum
ments by Korte and Jacobi 关28兴 support the predicted critical fin R ⫽ receding
spacing. A coil with a fin spacing of 3.2 mm showed bridging s ⫽ for small drops
behavior. After extended operation, the advancing contact angle x ⫽ in the x direction
decreased, causing f s,cr to fall below 3.2 mm and the bridging
behavior disappeared. Acknowledgment
Conclusions The authors would like to acknowledge the support of the Air
Conditioning and Refrigeration Center 共ACRC兲 of the University
In this article, we developed a model for predicting the steady- of Illinois at Urbana-Champaign.
state mass of the condensate retained on plain-fin heat exchangers.
Coils are assumed to have wide fin spacing so that condensate
bridges cannot form. The model takes the inclination angle and References
the advancing contact angle of the coil surfaces as inputs. The 关1兴 Yin, J., and Jacobi, A. M., 2000, “Condensate Retention Effects on the Air-
Side Heat Transfer Performance of Plain and Wavy-Louvered Heat Exchang-
fractional areas covered by small and large drops are also known ers,” Report No. TR-158, ACRC, University of Illinois at Urbana-Champaign,
from observations of coils with similar wettability and surface IL.
dimensions. Application of the model should be constrained to 关2兴 McQuiston, F. C., 1978, “Heat, Mass, and Momentum Transfer Data for Five
surfaces with roughness and materials that manifest advancing Plate-Fin-Tube Heat Transfer Surfaces,” ASHRAE Trans., 84, pp. 266–293.
关3兴 McQuiston, F. C., 1978, “Correlation of Heat, Mass, and Momentum Transport
contact angles from 45 deg to 120 deg, because the empirical Coefficients for Plate-Fin-Tube Heat Transfer Surfaces with Staggered Tubes,”
basis for the model is restricted to that range. The maximum di- ASHRAE Trans., 84, pp. 294–308.
ameter is obtained from a balance of forces acting on a drop. The 关4兴 Eckels, P. W., and Rabas, T. J., 1987, “Dehumidification: On the Correlation of
Wet and Dry Transport Processes in Plate Finned-Tube Heat Exchangers,” 关16兴 ElSherbini, A. I., 2003, “Modeling Condensate Drops Retained on the Air-Side
ASME J. Heat Transfer, 109, pp. 575–582. of Heat Exchangers,” Ph.D. thesis, University of Illinois at Urbana-
关5兴 Wang, C. C., Hsieh, Y. C., and Lin, Y. T., 1997, “Performance of Plate Finned Champaign, IL.
Tube Heat Exchangers Under Dehumidifying Conditions,” ASME J. Heat 关17兴 ElSherbini, A. I., and Jacobi, A. M., 2004, “Liquid Drops on Vertical and
Transfer, 119, pp. 109–117. Inclined Surfaces: Part II—A Method for Approximating Drop Shapes,” J.
关6兴 Korte, C. M., and Jacobi, A. M., 2001, “Condensate Retention Effects on the Colloid Interface Sci., 273, pp. 566–575.
Performance of Plain-Fin-and-Tube Heat Exchangers: Retention Data and 关18兴 ElSherbini, A. I., and Jacobi, A. M., 2004, “Liquid Drops on Vertical and
Modeling,” ASME J. Heat Transfer, 123, pp. 926–936. Inclined Surfaces: Part I—An Experimental Study of Drop Geometry,” J. Col-
关7兴 Jacobi, A. M., and Goldschmidt, V. W., 1990, “Low Reynolds Number Heat loid Interface Sci., 273, pp. 556–565.
and Mass Transfer Measurements of an Overall Counterflow, Baffled, Finned- 关19兴 ElSherbini, A. I., and Jacobi, A. M., 2005, “Retention Forces and Contact
Tube, Condensing Heat Exchanger,” Int. J. Heat Mass Transfer, 33, pp. 755– Angles for Critical Liquid Drops on Non-Horizontal Surfaces,” J. Colloid In-
765. terface Sci. 共in press兲.
关8兴 Wang, C. C., Du, Y. J., Chang, Y. J., and Tao, W. H., 1999, “Airside Perfor-
关20兴 Graham, C., 1969, “The Limiting Heat Transfer Mechanisms of Dropwise
mance of Herringbone Fin-and-Tube Heat Exchangers in Wet Conditions,”
Condensation,” Ph.D. thesis, Massachusetts Institute of Technology, MA.
Can. J. Chem. Eng., 77, pp. 1225–1230.
关21兴 Graham, C., and Griffith, P., 1973, “Drop Size Distributions and Heat Transfer
关9兴 Kim, G. J., and Jacobi, A. M., 2000, “Condensate Accumulation Effects on the
in Dropwise Condensation,” Int. J. Heat Mass Transfer, 16, pp. 337–346.
Air-Side Thermal Performance of Slit-Fin Surfaces,” Report No. CR-26,
ACRC, University of Illinois, Urbana. 关22兴 Tanasawa, I., and Ochiai, J., 1973, “Experimental Study on Dropwise Conden-
关10兴 Beatty, K. O., Jr., and Katz, D. L., 1948, “Condensation of Vapors on Outside sation,” Bull. JSME, 16, pp. 1184–1197.
of Finned Tubes,” Chem. Eng. Prog., 44, pp. 55–70. 关23兴 Tanaka, H., 1975, “Measurements of Drop-Size Distributions During Transient
关11兴 Rudy, T. M., and Webb, R. L., 1985, “An Analytical Model to Predict Con- Dropwise Condensation,” ASME J. Heat Transfer, 97, pp. 341–346.
densate Retention on Horizontal Integral-Fin Tubes,” ASME J. Heat Transfer, 关24兴 Sugawara, S., and Michiyoshi, I., 1956, “Dropwise Condensation,” Mem. Fac.
107, pp. 361–368. Eng., Kyoto Univ., Kyoto University, 18, pp. 84–111.
关12兴 Brown, R. A., Orr, F. M., Jr., and Scriven, L. E., 1980, “Static Drop on an 关25兴 Wu, Y. T., Yang, C. X., and Yuan, X. G., 2001, “Drop Distributions and
Inclined Plate: Analysis by the Finite Element Method,” J. Colloid Interface Numerical Simulation of Dropwise Condensation Heat Transfer,” Int. J. Heat
Sci., 73, pp. 76–87. Mass Transfer, 44, pp. 4455–4464.
关13兴 Milinazzo, F., and Shinbrot, M., 1988, “A Numerical Study of a Drop on a 关26兴 Rose, J. W., and Glicksman, L. R., 1973, “Dropwise Condensation–The Dis-
Vertical Wall,” J. Colloid Interface Sci., 121, pp. 254–264. tribution of Drop Sizes,” Int. J. Heat Mass Transfer, 16, pp. 411–425.
关14兴 Rotenberg, Y., Boruvka, L., and Neumann, A. W., 1984, “The Shape of Non- 关27兴 Shin, J., and Ha, S., 2002, “The Effect of Hydrophilicity on Condensation
axisymmetric Drops on Inclined Planar Surfaces,” J. Colloid Interface Sci., Over Various Types of Fin-And-Tube Heat Exchangers,” Int. J. Refrig., 25,
102, pp. 424–434. pp. 688–694.
关15兴 Dimitrakopoulos, P., and Higdon, J. J. L., 1999, “On the Gravitational Dis- 关28兴 Korte, C. M., and Jacobi, A. M., 1997, “Condensate Retention and Shedding
placement of Three-Dimensional Fluid Droplets from Inclined Solid Sur- Effects on Air-Side Heat Exchanger Performance,” Report No. TR-132,
faces,” J. Fluid Mech., 395, pp. 181–209. ACRC, University of Illinois, Urbana.
Heat Transfer in a Porous
Electrode of Fuel Cells
The thermal-fluid behaviors in a porous electrode of a proton exchange membrane fuel
cell (PEMFC) in contact with an interdigitated gas distributor are investigated numeri-
cally. The porous electrode consists of a catalyst layer and a diffusion layer. The heat
J. J. Hwang transfer in the catalyst layer is coupled with species transports via a macroscopic elec-
Professor, Dean of Graduate School trochemical model. In the diffusion layer, the energy equations based on the local thermal
e-mail: azaijj@mdu.edu.tw nonequilibrium (LTNE) are derived to resolve the temperature difference between the
Research Center for Advanced Science and solid phase and the fluid phase. Parametric studies include the Reynolds number and the
Technology, Stanton number (St). Results show that the wall temperature decreases with increasing
Mingdao University, Stanton number. The maximum wall temperatures occur at the downstream end of the
Changhua, 52345 Taiwan module, while the locations of local minimum wall temperature depend on the Stanton
numbers. Moreover, the solid phase and the fluid phase in the diffusion layer are ther-
mally insulated as StⰆ 1. The diffusion layer becomes local thermal nonequilibrium as
the Stanton number around unity. The porous electrode is local thermal equilibrium for
StⰇ 1. Finally, the species concentrations inside the catalyst and diffusion layers are also
provided. 关DOI: 10.1115/1.2175092兴
Keywords: porous electrode, proton exchange membrane fuel cell (PEMFC), catalyst
layer, diffusion layer, local thermal nonequilibrium (LTNE)
1 Introduction been devoted to simultaneously predict the fluid and solid tem-
peratures inside the porous electrode of a fuel cell in the open
Study of fluid flow and heat transfer in a porous medium is a
literature 关17兴.
subject of continuous interest for decades because of the wide In the present study, an LTNE model is proposed to simulate the
range of engineering applications such as solar receivers, building thermal-fluid transports in the porous electrode of a fuel cell. A
thermal insulation materials, packed bed heat exchangers, and en- multiphysics model coupling heat/mass transfer with electro-
ergy storage units. Recently, the emerging field of fuel cell ther- chemical kinetics is developed. This model is capable of predict-
mal management has received particular attention due to the rise ing the thermal-fluid and electrochemical behaviors, such as flow
of the importance of new energy technology. The gas diffusion velocities, fluid/solid phase temperatures, reactant concentrations,
electrode of a fuel cell is a typical porous medium. The heat and current densities in the porous electrode of fuel cells. More-
dissipation by the electrochemical reaction in the active layer of a over, the heat/mass transfer mechanisms together with the thermal
fuel cell should be removed thoroughly to keep proper operation. pathways in a fuel cell can be clearly identified. It would be ben-
It not only optimizes an operating temperature range but also eficial for further accurate analyses of the fuel-cell thermal perfor-
keeps a high uniformity of temperature distribution in the porous mance by considering the temperature-dependent physical proper-
electrode. Therefore, a superior thermal management becomes ties inside a fuel cell. The present research along with 关17兴 is the
critical for safely operating and prolonging a fuel cell. first attempt to develop an LTNE model to simultaneously predict
A variety of analytical and/or numerical studies based on the the fluid and solid temperatures inside a porous electrode of a fuel
assumption of local thermal equilibrium 共LTE兲 have been con- cell.
ducted 关1,2兴 to investigate the transport phenomena in the porous
medium. Under the LTE assumption, the so-called one-equation
model is employed to obtain the phase-mixing temperatures in the 2 Mathematical Numerical Model
porous medium. It is simple and straightforward, but is applicable The sectional view of a typical porous electrode in contact with
only if the temperature difference between the solid and the fluid an interdigitated flow field of a PEMFC is shown schematically in
phases is very small. When the temperature difference is signifi- Fig. 1. Figure 2 shows the dimensions of the present computa-
cant, it becomes a local thermal nonequilibrium 共LTNE兲. The one- tional module. The anodic reaction is neglected due to its fast
equation model is not more valid, and should be modified with the kinetics 关18兴. Therefore, only the oxygen reduction reaction
two-equation model, which treats the solid-phase temperature and 共ORR兲 on the cathode is considered 关19兴 in the present model.
the fluid-phase temperature separately. An analysis of heat transfer
in a porous medium based on the two-equation model is more O2 + 4H+ + 4e− ↔ 2H2O 共1兲
involved because it requires interstitial information inside the po- The oxygen-water vapor mixture enters the porous cathode form
rous medium such as heat transfer coefficient between the fluid the module inlet 共inlet channel兲, then transverses the diffusion
and solid phases and the interfacial surface area 关3–7兴. Due to this layer to the catalyst layer. The electrochemical reaction occurring
difficulty, most investigations about the porous electrode of a fuel in the catalyst layer consumes the oxygen 共CO2兲 and, meanwhile,
cell either used the one-equation model to obtain the phase- produces the water vapor 共CH2O兲. In addition, the heat generation
mixing temperatures 关8–11兴 or did not consider thermal transports 共Q兲 due to the overpotential and irreversibility heating during the
at all 关12–16兴. Up to the present time, quite a few efforts have ever electrochemical reaction is removed by either the fluid or the
solid.
The assumptions in this model include 共i兲 mixtures are ideal
NAL OF HEAT TRANSFER. Manuscript received February 6, 2005; final manuscript re- gas, 共ii兲 flow is steady, laminar, and incompressible 共iii兲 thermal
ceived October 21, 2005. Review conducted by N. K. Anand. physical properties are constant, 共iv兲 porous electrode is homoge-
434 / Vol. 128, MAY 2006 Copyright © 2006 by ASME Transactions of the ASME
␳fu
⳵v
⳵x
⳵v
+ ␳fv = −
⳵y
⳵p
⳵y
+␮
⳵ 2v ⳵ 2v
+
⳵x2 ⳵ y 2
␮
− ␧v
␬
冉冊共4兲
The last terms on the RHS of Eqs. 共3兲 and 共4兲 are the traditional
Darcy dissipation which is used to account for the pressure drop
due to viscous friction at the surfaces of carbon fibers.
As for the energy equations, the LTNE model is used to de-
scribe the thermal behavior in the diffusion layer. The energy
equations for solid and fluid phases, respectively, are 关5兴
0 = ks,eff 冉 ⳵ 2T s ⳵ 2T s
⳵x2 ⳵y
冊
+ 2 − hsSV共Ts − T f 兲共5兲
共 ␳ c p兲 f u
⳵T f
⳵x
+ 共 ␳ c p兲 f v
⳵T f
⳵y
= k f,eff
⳵x2
冉
⳵ 2T f ⳵ 2T f
+ 2 + hsSV共Ts − T f 兲
⳵y
冊
共6兲
The source term on the RHS of Eq. 共6兲, hsSV共Ts − T f 兲, means the
convective heat transfer from the carbon fiber to the fluid, and
vice versa for the source term in Eq. 共5兲. The effective thermal
Fig. 1 Schematic drawing of porous electrode of the interdigi- conductivities of both phases are respectively defined as
tated flow field
ks,eff = 共1 − ␧兲ks 共7兲
k f,eff = ␧k f 共8兲
neous and isotropic, 共v兲 water exits as vapor only, and 共vi兲 both
inlet-fluid temperature and rib-surface temperature are uniform. In the catalyst layer, physically, the electrochemical reaction
occurs at the interface of reactant 共fluid兲 and catalyst 共solid兲 under
2.1 Governing Equations. In both diffusion and catalyst lay- a fixed temperature. Therefore, the fluid and solid phases in the
ers, the steady volume-averaged continuity and momentum equa- catalyst layer have the same local temperatures, i.e.,
tions that govern fluid flow are given by
⳵共␳ f u兲 ⳵共␳ f v兲 T f = Ts 共9兲
+ = S␾ 共2兲
⳵x ⳵y
␳fu
⳵u ⳵u
+ ␳fv = −
⳵p
+␮ 冉
⳵u ⳵u
+
2
␮ 2
− ␧u 冊共3兲
共 ␳ c p兲 f u
⳵T f
⳵x
+ 共 ␳ c p兲 f v
⳵T f
⳵y
= kc,eff
⳵x2
冉
⳵ 2T f ⳵ 2T f
+ 2 + j T␩
⳵y
冊共10兲
⳵x ⳵y ⳵x ⳵x2 ⳵ y 2 ␬
The source term j T␩ represents the overpotential heating by the
electrochemical reaction. According to the Butler-Volmer correla-
tion 关20兴, the relationship between the local transfer current den-
sity and the reactant concentrations can be described as the fol-
lowing equation;
j T = ␣1 冉冊冉冊
c O2
cO2,ref
− ␣2
c H2O
cH2O,ref
2
共11兲
␣1 and ␣2 are electrochemical coefficients depending on the ex-

change current density and the overpotential on the electrode sur-
faces. They are regarded as constants in the present model. The
effective thermal conductivity of the catalyst layer can be deter-
mined by the following equation 关21兴
1
kc,eff = − 2kc + 共12兲
␧C 1 − ␧C
+
2kc + k f 3kc
where kc is the weight-averaged conductivity between the ionic
conductor 共such as Nafion™兲 and the electric conductor 共such as
Pt/ C兲. ␧C is the porosity of the catalyst layer.
In the catalyst layer, the energy equation is coupled with the
species concentration 共Eqs. 共10兲 and 共11兲兲. Therefore, prior to
solving the temperature distribution it is necessary to determine
the species concentration distribution in the porous electrode. The
species transports of the oxygen and water vapor in the porous
cathode can be governed by the following equations
Fig. 2 Configuration of the computational domain

u
⳵ c O2
⳵x
+v
⳵ c O2
⳵y
= DO2,eff 冉 ⳵ 2c O2
⳵x2
+
⳵ 2c O2
⳵y2
冊 + S O2 共13兲
Table 1 Governing equations of the present model
S⌽
Equations Diffusion layer Catalyst layer Eq.
⳵U ⳵V
Mass + = S⌽ 0 S⌽ 共17兲
⳵X ⳵Y
Momentum U
⳵U
⳵X
+V
⳵U
⳵Y
=−
⳵ P 1 ⳵ 2U ⳵ 2U
+ +
⳵X Re ⳵X2 ⳵Y 2
+ S⌽ 冉冊 −
1
Re Da
␧U −
1
Re Da
␧ CU 共18兲
U
⳵V
⳵X
⳵V
+V =−
⳵Y
⳵ P 1 ⳵ 2V ⳵ 2V
+ +
⳵Y Re ⳵X2 ⳵Y 2
+ S⌽ 冉冊 −
1
Re Da
␧V −
1
Re Da
␧ CV 共19兲
Energy, ⳵ 2␪ s ⳵ 2␪ s
diffusion 0= + + S⌽ Bi共␪ f − ␪s兲 ¯ 共20兲
⳵X2 ⳵Y 2
layer
U
⳵␪ f
⳵X
+V
⳵␪ f
=
1 ⳵ 2␪ f ⳵ 2␪ f
+
⳵Y Re Preff ⳵X2 ⳵Y 2
冉
+ S⌽ 冊 St共␪s − ␪ f 兲 ¯ 共21兲
Energy, ␪s = ␪ f ¯ ¯ 共22兲
catalyst
layer
U
⳵␪ f
⳵X
+V
⳵␪ f
=
1 ⳵ 2␪ f ⳵ 2␪ f
+
⳵Y Re Preff ⳵X2 ⳵Y 2
冉
+ S⌽ 冊 ¯
1
Re Preff
关Ec1CO2 − Ec2共CH2O兲2兴共23兲
Species U
⳵ C O2
⳵X
+V
⳵ C O2
⳵Y
=
1
Re ScO2 ⳵X2
+ 冉
⳵ 2C O2 ⳵ 2C O2
⳵Y 2
+ S⌽ 冊 0
1
Re ScO2
关RM1CO2 − RM2共CH2O兲2兴
共24兲
U
⳵ C H2O
⳵X
+V
⳵ C H2O
⳵Y
=
1
Re ScH2O ⳵X2
+ 冉
⳵ 2C H2O ⳵ 2C H2O
⳵Y 2
+ S⌽ 冊 0
1
Re ScH2O
关RN1CO2 − RN2共CH2O兲2兴
共25兲
u
⳵ c H2O
⳵x
+v
⳵ c H2O
⳵y
= DH2O,eff 冉 ⳵ 2c H2O
⳵x 2 +
⳵ 2c H2O
⳵y 2 冊 + SH2O 共14兲 St =
h sS V␦
共 ␳ c p兲 f u d
, ScO2 =
␯
DO2,eff
In the catalyst layer, the source terms for oxygen transport and
water-vapor transport are SO2 = j T / 4F and SH2O = −共jT / 2F兲, respec- ␯ ␣ 1␦ 2
tively. In the diffusion layer they are nothing. The effective diffu- ScH2O = , RM1 =
DH2O,eff 4FctotDO2,eff
sivities of the oxygen and water vapor in the porous electrode,
DO2,eff and DH2,eff, follows the Bruggemann model 关18兴, i.e.,
DO2,eff = ␧␶DO2 共15兲 ␣ 2␦ 2 ␣ 1␦ 2
R M2 = , RN1 =
4FctotDO2,eff 2FctotDH2O,eff
DH2O,eff = ␧␶DH2O 共16兲
Using the nondimensional parameters of ␣ 2␦ 2 k f,eff ␣1␩␦2
RN2 = , Rk = , Ec1 =
x y ṁ u v 2FctotDH2O,eff ks,eff ks,eff共T f,in − Tr兲CO2,ref
X= , Y= , ud = , U= , V=
␦ ␦ ␳ f A␧ ud ud
and
p c O2 cO2,ref
P= , C O2 = , CO2,ref = ␣2␩␦2
␳ f u2d ctot ctot Ec2 = ,
ks,eff共T f,in − Tr兲CH
2
2O,ref
c H2O cH2O,ref T − Tr
C H2O = , CH2O,ref = , ␪= the above governing equations can be deduced to dimensionless
ctot ctot T f,in − Tr
forms of Table 1.
u d␦ ␬ ␮共c p兲 f h sS V␦ 2 2.2 Boundary Conditions. Conditions for the boundaries of
Re = , Da = , Preff = , Bi =
␯ ␦2 k f,eff ks,eff the present computational domain can be summarized as follows.
factory. A typical simulation requires about 200 min of central
processing unit time on a Pentium IV 2.8 GHz PC.
3 Results and Discussion

3.1 Prior to Predictions. Two important issues should be
clarified before the subsequent presentation and discussion of the
predicted results, i.e., variation 共or typical values兲 of the param-
eters and validation of the model.
To perform a practical simulation of thermal behaviors in a fuel
cell, a realistic operation condition of fuel cells should be known.
Table 2 summarizes the physical, geometric, and flow prosperities
of a typical porous electrode of a PEMFC. Most data in the table
are from the manufacturer and the textbook. Some of them are
assumed, such as the flow properties of interstitial heat transfer
coefficient 共hs兲 and surface-to-volume ratio 共SV兲. In addition, from
the data listed in Table 2, typical values or ranges of the nondi-
mensional parameters can be calculated. For example, the Rey-
nolds number based on the pore velocity and the diffusion layer
thickness becomes
Fig. 3 Grid test by comparing the axial velocity at Y = 2.0

ud␦ 0.239 · 400 ⫻ 10−6
Re = = =6
␯ 1.586 ⫻ 10−5
which is typical in the following predictions.
1. At the module inlet, Up to the present time, it is hard to measure the local tempera-
⳵␪s ture inside the porous electrode of a fuel cell. Strictly speaking, no
P = Pin, ␪ f = ␪ f,in, = Bi共␪s − ␪ f 兲 data about the local thermal-fluid characteristics such as local
⳵X temperatures and heat transfer coefficients inside the porous elec-
共26兲 trode of a fuel cell are available in the open literature. Therefore,
CO2 = CO2,in, CH2O = CH2O,in a macroscopic comparison of polarization curve is made between
2. At the rib 共current collector兲 surfaces, the present predictions and previous experiments 关23兴, which have
been widely used to verify a fuel cell simulation model. Figure 4
⳵P compared the predicted polarization curves with the measured
= 0, U = V = 0, ␪s = 1
⳵X ones. The cell potential in the experiment is measured directly,
共27兲 while in the present prediction the cell potential, neglecting the
⳵ C O2 ⳵ C H2O overpotential on the anode side, is calculated as
⳵␪ f
= St共␪s − ␪ f 兲, = =0 Vcell = VOC − ␩ − I · R⍀ 共31兲
⳵X ⳵X ⳵X
3. At the interface of the membrane and the catalyst layer where the Ohmic resistance, R⍀, is set to a constant value of
共X = −0.1兲, 0.2 ⍀ · cm2 through the electrode, which is a typical value mea-
sured in the experiment 关24兴. Generally speaking, the agreement
⳵ P ⳵ ␪ s ⳵ ␪ f ⳵ C O2 ⳵ C H2O for the comparison above is satisfactory, indicating that the results
U = V = 0, = = = = = 0 共28兲 obtained by the present computation are reliable.
⳵X ⳵X ⳵X ⳵X ⳵X
4. At the symmetric planes 3.2 Velocity Distributions. Figure 5 shows the distribution of
flow velocity magnitude, 共U2 + V2兲1/2, at several Y stations in the
⳵ P ⳵ U ⳵ V ⳵ ␪ s ⳵ ␪ f ⳵ C O2 ⳵ C H2O porous electrode. The Reynolds number and Darcy number are
= = = = = = = 0 共29兲
⳵Y ⳵Y ⳵Y ⳵Y ⳵Y ⳵Y ⳵Y fixed at Re= 6 and Da= 9.83⫻ 10−6, respectively. It is seen that the
velocities are higher in the diffusion layer than in the catalyst
5. At the module outlet,
layer. This is because a higher porosity and a higher permeability
⳵␪ f ⳵␪s ⳵ C O2 ⳵ C H2O are accompanied by the diffusion layer 共Table 2兲. In addition, the
P = Pout, = 0, = Bi共␪s − ␪ f 兲, = =0 velocities near the shoulder surface 共X = 0.4兲 are higher than those
⳵X ⳵X ⳵X ⳵X
adjacent to the catalyst layer. It is reasonable because the flow
共30兲 paths from the inlet channel to the outlet channel are shorter for
the flow closer to the shoulder surface. A shorter flow path has a
2.3 Numerical Methods. The governing equations are nu- lower pressure-drop penalty, which allows more fluids to pass
merically solved by the control-volume-based finite element through. The effect of Reynolds number on the velocity distribu-
method. The discretization procedure as well as the numerical tions at module middle 共Y = 2.0兲 is shown in Fig. 6. It is seen that
scheme has been described in detailed elsewhere 关22兴, and is not the velocity increases with increasing Reynolds number in both
elaborated on here. In the present work, all computations are per- catalyst and diffusion layers.
formed on 503,160 共X by Y兲 structured, orthogonal meshes. Ad-
ditional runs for the coarser meshes, 403,120, and the finer 3.3 Porous Matrix Temperatures. Figures 7–9 compare the
meshes, 603,180, are taken for a check of grid independence. As distributions of the fluid-phase temperature 共␪ f 兲 and the solid-
shown in Fig. 3, the maximum discrepancy in the axial velocity at phase temperature 共␪s兲 along the X direction at several Y stations
Y = 2.0 between the grid sizes of 503,160 and 603,180 is only 共Y = 0.5, 2.0, 2.8, and 3.6兲 in the porous electrode under various
0.6%. In addition, results indicate a maximum change of 0.6% in Stanton numbers. The Reynolds number and Darcy number are
wall temperature distribution between the solutions of 503,160 fixed at Re= 6, and Da= 9.83⫻ 10−6, respectively.
and 603,180 grids. All of the above discrepancies are so small that In Fig. 7, clearly, both solid and fluid phases have the same
the accuracy of the solutions on a 503,160 grid is deemed satis- temperatures in the catalyst layer 共−0.25⬍ X ⬍ 0兲. It starts with a
Table 2 Porous electrode data sheet
Property name Solid phase Fluid phase
Physical Material Carbon fiber Oxygen

propertiesa Bulk density, ␳ 1100 kg/ m3 1.13 kg/ m3
Specific resistance, ␴−1 115 ␮␵ m ¯
Thermal conductivity, k 1.71 W / m Kc 0.0268 W / m K
Coeff. of thermal expansion, ␤ 9.03⫻ 10−7 / K 3.333⫻ 10−3 / K
Viscosity, ␯ ¯ 1.5863⫻ 10−5 m2 / s
Thermal diffusivity, ␣ f ¯ 2.2353⫻ 10−5 m2 / s
Geometric Diffusion layer porosity, ␧ 48%

properties Catalyst layer porosity, ␧C 42%
Diffusion layer thickness, ␦ 400 ␮m
Catalyst layer thickness, ␦C 100 ␮m
Diffusion layer tortuosity, ␶ 1.5
Catalyst layer tortuosity, ␶ 1.5
Average pore diameter, d 33 ␮m
Diffusion layer permeability, ␬b 1.573⫻ 10−12 m2
Catalyst layer permeability, ␬Cb 1.023⫻ 10−12 m2
Flow properties Interstitial heat transfer coefficient, hs 100– 1000 W m−2 共assumed兲
Surface-to-volume ratio, SV 1.03⫻ 103 – 1.03⫻ 106 m−1 共assumed兲
Pore velocity, ud for I = 1.0– 3.0 A at 0.239– 0.717 m / s
stoichiometric flow ratio 5
Nondimensional Conductivity ratio, Rk 0.01447

parameters Prandtl number, Pr 0.709
Reynolds number, Red 6
Stanton number, St 1.39
共hs = 100 W m−2 and SV = 10,000 m−1兲
a
Data from manufacture 共Toray兲.
b
ASTM E1530.
c
Oxygen at 25° C with 15% RH, p = 0.15 atm, plate thickness: 2.5 cm, 51 l / m2 s.
local maximum value at the interface of the catalyst layer and the hot solid-matrix transfers a part of heat to the cold fluid via con-
electrolyte membrane 共X = −0.25兲, decreases slightly and then vection 共hs兲. At the middle of the module, i.e., Y = 2.0, the solid-
drops sharply as it approaches the diffusion layer 共X = 0兲. In the phase temperatures decrease sharply and linearly from X = 0 to
diffusion layer, the fluid-phase and solid-phase temperatures sepa- X = 1.0, meaning that a large amount of heat passes through the
rate from X = 0 due to the LTNE effect. At the station of Y = 0.5, a solid matrix and then is removed out from the rib surface by
low fluid-phase temperature from the module inlet 共X = 1.0兲 in- conduction. It is interesting to note that this is the shortest con-
creases linearly as the flow approaches the catalyst layer 共X = 0兲. ductive 共solid-phase兲 pathway form the catalyst layer to the rib
Meanwhile, the solid-phase temperature decreases along the X shoulder. At this station, the fluid-phase temperatures are higher
direction from the interface of the catalyst layer and the diffusion
layer 共X = 0兲 to the module inlet 共X = 1.0兲. During this stage, the
Fig. 4 Comparison of the present predictions with the experi- Fig. 5 Velocity magnitudes in the porous electrode at several
mental results Y stations, Da= 9.83Ã 10−6, Re= 10
fixed temperature ␪s = 1. As for the cross-section cutting across the
module outlet, i.e., Y = 3.6, both the fluid-phase and solid-phase
temperatures are significantly lifted up.
When the Stanton number increases to St= 14.73, as shown in
Fig. 8, the temperature difference between the solid and fluid
phases has been narrowed down. This trend is more significant as
the flow moves downstream. It means that the porous electrode
approaches the local thermal equilibrium. As the Stanton number
further increases to St= 1.473⫻ 103, as shown in Fig. 9, the curves
of the fluid-phase and solid-phase temperatures merge together,
except for the region near the module inlet 共Y = 0.5, 0.8⬍ X
⬍ 1.0兲. A large similarity in the temperature distribution between
the solid phase and the fluid phase gives an indication that the
thermal-fluid field in the porous electrode has become local ther-
mal equilibrium.
3.4 Wall Temperature Distributions. Figures 10–12 show
the temperature distributions along the interface of the electrolyte
membrane and catalyst layer 共X = −0.25兲, which is referred to as
the wall temperature ␪w hereafter. Note that the wall temperature
is always a local maximum temperature with respect to the X
Fig. 6 Effect of Reynolds number on the velocity magnitude in
direction 共see Figs. 7–9兲, As far as the fuel cell thermal manage-
the porous electrode, Da= 9.83Ã 10−6 ment concern, to keep the maximum wall temperature within a
reasonable range has a benefit of prolonging fuel cells, which
requires a cooling scheme to ensure the removal of the heat dis-
than the solid-phase ones. This indicates that the heat possessed sipation by electrochemical reaction.
with the high-temperature fluid not only is carried downstream by Figure 10 shows the effect of Stanton number on the wall tem-
bulk motion but also is convected to the solid matrix, which is perature distribution along the Y direction. The Darcy number and
then conducted out through the rib surfaces. At the station of Y Reynolds number are fixed at Da= 9.83⫻ 10−6 and Re= 6, respec-
= 2.8, the fluid-phase temperature is further elevated, while the tively. It is clearly observed from this figure that the wall tempera-
solid-phase temperature distribution is puffed up a little in the ture decreases with increasing Stanton number. In addition, the
middle region since the RHS boundary 共X = 1.0兲 is confided by a maximum wall temperatures occur at the downstream end of the
Fig. 7 Comparison of fluid-phase and solid-phase temperature distributions inside the porous
electrode for Re= 6, Da= 9.83Ã 10−6, and St= 0.74
electrode for Re= 6, Da= 9.83Ã 10−6, and St= 14.73
module 共Y = 1.6兲. However, the locations of local minimum wall wall temperature distribution separates from Y = 0.2 as the Rey-
temperature depend on the Stanton numbers, which will be dis- nolds number varies. The wall temperature decreases with increas-
cussed later. In general, the wall temperature distributions can be ing Reynolds number. It is clear from this figure that the wall
sorted by three groups, depending on the magnetite of the Stanton temperature distribution neither has a large valley in the module
number. The first group is collected under small Stanton numbers, middle nor increases downstream monotonously. This trend shows
StⰆ 1 共St= 0.0074, 0.0118, and 0.0147, open symbol lines兲. It is a mixed thermal-transport mode in combination of convection and
seen that the wall temperatures increase downstream monoto- conduction in the porous electrodes of fuel cells.
nously and fall into a single curve almost. This trend is largely
3.5 Species Concentration Distributions. The oxygen and
similar to a developing flow over a thermal flat plate. At a small
water-vapor concentration distributions along the X direction at
value of Stanton number, the source terms of Eqs. 共20兲 and 共21兲
are neglected. Therefore, the thermal pathways of convection and several Y stations in the porous electrode are shown in Figs. 13
conduction in the diffusion layer are independent. In other words, and 14, respectively. The composition of the oxygen-water vapor
there is no heat transfer between the solid phase and the fluid mixture at the module inlet is CO2 = 0.9 and CH2O = 0.1. The Rey-
phase due to drastically small interfacial heat transfer coefficients nolds number and Stanton number are fixed Re= 6 and St= 0.74,
and/or surface area. The second group goes with the Stanton num- respectively. It is seen from Fig. 13 that at the station cutting
ber around unity 共0.88艋 St艋 1.47, broken lines兲. It starts with a across the module inlet, i.e., Y = 0.4, high-concentrated oxygen
little drop of wall temperature and then sharply increases down- from the module entrance 共X = 1.0兲 decreases as the flow moves
stream. As for the third group of curves 共1.473⫻ 103 艋 St closely to the catalyst layer. The inclination of CO2 curves de-
艋 1.473⫻ 104, solid symbol lines兲, the wall temperature decreases creases as the flow moves downstream. It is further seen from
first and then increases to a local maximum at the downstream Figs. 13 and 14 that the oxygen/water vapor concentration
end. A significant local minimum occurs at the module middle decreases/increases along the Y direction due to the ORR on the
共Y = 0.8兲. The effect of Reynolds number on the local wall tem- reaction surface. The lowest values of CO2 are found at the inter-
perature distributions at the Stanton number of St= 1.473⫻ 103 is face between the catalyst layer and the electrolyte membrane 共X
further shown in Fig. 11. It is seen from this figure that the wall = −0.25兲. The CH2O distributions shown in Fig. 14 just compensate
temperature at the upstream half module decreases with increasing for those of CO2. It increases along the flow direction due to the
Reynolds number, but is not affected by the Reynolds number at convective accumulation downstream of the water vapor gener-
the downstream half module. The wall temperature distributions ated in the catalyst layer.
are roughly symmetrical about the local minimum at the module
middle 共Y = 0.8兲 where it faces the cold shoulder 共rib兲 surface. A
valley distribution of wall temperature gives an indication that the 4 Conclusions
conduction dominates the thermal transport in the porous medium. Detailed heat transfer characteristics inside a fuel cell porous
Under LTNE conditions 共St= 1.03兲, as shown in Fig. 12, the electrode in contact with a typical interdigitated flow field have
electrode for Re= 6, Da= 9.83Ã 10−6, and St= 1.473Ã 103
been studied numerically. A two-dimensional multiphysics model scribe the effective diffusivities of the oxygen and water vapor in
coupling heat/mass transfer with electrochemical kinetics has the porous electrode of a fuel cell. Results show that the wall
been performed to simulate thermal-fluid/species transports inside temperature decreases with increasing Stanton number. The local
a porous electrode of a fuel cell. In thermal analysis, the local maximum wall temperature always occurs at the downstream end
thermal nonequilibrium 共LTNE兲 model is first implemented to in- of the module, while the location of the local minimum depends
vestigate the heat transfer in the fuel cell porous electrode. In on the Stanton number. Three categories of thermal-fluid behav-
species transports, the Bruggemann model is employed to de- iors in the porous electrode have been characterized according to
Fig. 10 Effect of Stanton number on the wall temperature dis- Fig. 11 Effect of Reynolds number on the wall temperature
tributions for Da= 9.83Ã 10−6, and Re= 6 distributions for Da= 9.83Ã 10−6 and St= 1.473Ã 103
Fig. 12 Effect of Reynolds number on the wall temperature
distributions for Da= 9.83Ã 10−6 and St= 1.03
Fig. 14 Water vapor concentration distributions of along the
reaction surfaces, Re= 6, Da= 9.83Ã 10−6, and St= 0.74
the numerical results. The solid phase and fluid phase in the dif-
fusion layer are thermally independent for a small value of Stan-
ton number 共StⰆ 1兲. As the Stanton number approaches unity, the ⫽ water vapor mole concentration at inlet,
porous electrode becomes local thermal nonequilibrium 共LTNE兲 mol m−3
with a strong thermal interaction 共heat transfer兲 between the solid C H2O ⫽ normalized water vapor concentration,
and fluid phases. Under the conditions of StⰇ 1, the porous elec- cH2O / ctot
trode is local thermal equilibrium. CH2O,ref ⫽ normalized water vapor mole concentration at
inlet, cH2O,ref / ctot
Acknowledgment c O2 ⫽ oxygen concentration, mol m−3
This work was partly sponsored by the National Science Coun- cO2,ref ⫽ oxygen mole concentration at inlet, mol m−3
cil of the Taiwan, ROC under Contract No. NSC 92-2212-E-451- C O2 ⫽ normalized oxygen concentration, cO2 / ctot
002.
CO2,ref ⫽ normalized oxygen mole concentration at inlet,
cO2,ref / ctot
Nomenclature
cp ⫽ specific heat at constant temperature,
Bi ⫽ Biot number J kg−1 K−1
d ⫽ pore diameter of the porous medium, m ctot ⫽ total mole concentration of the reacting fluid,
cH2O ⫽ water vapor mole concentration, mol m−3 mol m−3
cH2O,ref D H2O ⫽ binary diffusivity of water vapor in the oxy-
gen, m2 s−1
DH2O,eff ⫽ effective diffusivity of water vapor in the gas
diffusion layer, m2 s−1
D O2 ⫽ binary diffusivity of oxygen in the water va-
por, m2 s−1
DO2,eff ⫽ effective diffusivity of oxygen in the gas diffu-
sion layer, m2 s−1
Da ⫽ Darcy number
Ec1 , Ec2 ⫽ coefficients in Eq. 共23兲
F ⫽ Faraday’s constant, 96,487 C mol−1
I ⫽ current, A
hs ⫽ interstitial heat transfer coefficient in the po-
rous medium, W m−2 K−1
k ⫽ thermal conductivity, W m−1 K−1
L ⫽ length of the computational domain, m
ṁ ⫽ mass flux, kg m−2 s−1
p ⫽ pressure, Pa
P ⫽ nondimensional pressure
Pr ⫽ Prandtl number
Q ⫽ heat dissipation by the electrochemical reac-
tion, W
Re ⫽ Reynolds number
Fig. 13 Oxygen concentration distributions of along the reac- RM1 , RM2 ⫽ coefficients in Eq. 共24兲
tion surfaces, Re= 6, Da= 9.83Ã 10−6, and St= 0.74 RN1 , RN2 ⫽ coefficients in Eq. 共25兲
R␵ ⫽ Ohmic resistance, ␵ Porous Media,” Handbook of Heat Transfer in Porous Media, 1st ed., K. Vafai,
ed., Decker, New York.
SV ⫽ surface area-to-volume ratio, m−1 关4兴 Kaviany, M., 1995, Principles of Heat Transfer in Porous Media, 2nd ed.,
Sc ⫽ Schmidt number Springer, Berlin.
St ⫽ Stanton number 关5兴 Hwang, J. J., Hwang, G. J., Yeh, R. H., and Chao, C. H., 2002, “Measurement
S⌽ ⫽ nondimensional source term of Interstitial Convective Heat Transfer and Fictional Drag for Flow Across
Metal Foams,” ASME J. Heat Transfer, 124, pp. 120–129.
S␾ ⫽ source term 关6兴 Alazmi, B., and Vafai, K., 2002, “Constant Wall Heat Flux Boundary Condi-
T ⫽ temperature, K tions in Porous Media Under Local Thermal Non-Equilibrium Conditions,”
u,v ⫽ velocity components in the x and y directions, Int. J. Heat Mass Transfer, 45, pp. 3071–3087.
respectively, m s−1 关7兴 Alazmi, B., and Vafai, K., 2004, “Analysis of Variable Porosity, Thermal Dis-
persion, and Local Thermal Non-equilibrium on Free Surface Flows Through
U,V ⫽ nondimensional velocity components in the x Porous Media,” ASME J. Heat Transfer, 126, pp. 389–399.
and y direction 关8兴 Bernardi, D. M., and Verbrugge, M. W., 1991, “Mathematical Model of a Gas
Vcell ⫽ cell voltage, V Diffusion Electrode Bonded to a Polymer Electrolyte,” Am. Inst. Chem. Eng.
Voc ⫽ open circuit voltage, V Symp. Ser., 37, pp. 1151–1163.
关9兴 Springer, T. E., Zawodzinski, T. A., and Gottesfeld, S., 1991, “Polymer Elec-
x,y ⫽ coordinate system, m trolyte Fuel Cell Model,” J. Electrochem. Soc., 138共8兲, pp. 2334–2342.
X,Y ⫽ nondimensional x and y coordinate 关10兴 Nguyen, T. V., and White, R. E., 1993, “A Water and Heat Management Model
for Proton-Exchange-Membrane Fuel Cells,” J. Electrochem. Soc., 140共8兲, pp.
Greek Symbols 2178–2186.
␣1 , ␣2 ⫽ coefficients in Eq. 共11兲关11兴 Ackmann, T., de Haart, L. G. J., Lehnert, W., and Stolten, D., 2003, “Modeling
␦ ⫽ thickness of the diffusion layer, m of Mass and Heat Transport in Planar Substrate Type SOFCs,” J. Electrochem.
Soc., 150, pp. A783–A789.
␦C ⫽ thickness of the catalyst layer, m 关12兴 Yuan, J., Rokni, M., and Sunden, B., 2003, “Three-Dimensional Computa-
␧ ⫽ porosity of the diffusion layer tional Analysis of Gas and Heat Transport Phenomena in Ducts Relevant for
␧C ⫽ porosity of the catalyst layer Anode-Supported Solid Oxide Fuel Cells,” Int. J. Heat Mass Transfer, 46, pp.
␬ ⫽ permeability, m2 809–821.
关13兴 Yi, J. S., and Nguyen, T. V., 1998, “An Along the Channel Model for Proton
␳ ⫽ density, kg m3 Exchange Membrane Fuel Cells,” J. Electrochem. Soc., 145, pp. 1149–1159.
␩ ⫽ cathodic overpotential, V 关14兴 Dutta, S., Shimpalee, S., and Van Zee, J. W., 2001, “Numerical Prediction of
␪ ⫽ nondimensional temperature Mass-Exchange Between Cathode and Anode Channels in a PEM Fuel Cell,”
␶ ⫽ tortuosity Int. J. Heat Mass Transfer, 44, pp. 2029–2042.
关15兴 Ju, H., and Wang, C. Y., 2004, “Experimental Validation of a PEM Fuel Cell
Subscript Model by Current Distribution Data,” J. Electrochem. Soc., 151, pp. A1954–
A1960.
C ⫽ catalyst layer 关16兴 Um, S., and Wang, C. Y., 2004, “Three-Dimensional Analysis of Transport and
D ⫽ diffusion layer Electrochemical Reactions in Polymer Electrolyte Fuel Cells,” J. Power
eff ⫽ effective Sources, 125, pp. 40–51.
f ⫽ fluid phase 关17兴 Hwang, J. J., 2006, “Thermal-Electrochemical Modeling of a PEM Fuel Cell,”
J. Electrochem. Soc., 153, pp. A216–A224.
in ⫽ inlet 关18兴 Meredith, R. E., and Tobias, C. W., 1962, in Advances in Electrochemistry and
loc ⫽ local Electrochemical Engineering, 2, C. W. Tobias, ed., Interscience, New York.
o ⫽ oxygen 关19兴 Hwang, J. J., Chen, C. K., Savinell, R. F., Liu, C. C., and Wainright, J., 2004,
out ⫽ outlet “A Three-Dimensional Numerical Simulation of the Transport Phenomena in
the Cathodic Side of a PEMFC,” J. Appl. Electrochem., 34, pp. 217–224.
r or ref ⫽ reference 关20兴 Oldham, H., and Myland, J., 1994, Fundamentals of Electrochemical Science,
s ⫽ solid phase Academic, New York.
tot ⫽ total 关21兴 Hwang, J. J., Chen, C. K., and Lai, D. Y., 2005, “Detailed Characteristic
Comparison Between Planar and MOLB-Type Sofcs,” J. Power Sources, 143,
pp. 75–83.
References 关22兴 Hwang, J. J., Chao, C. H., Ho, W. Y., Chang, C. L., and Wang, D. Y., 2005,
关1兴 Quintard, M., and Whitaker, S., 1995, “Local Thermal Equilibrium for Tran- “Effect of Flow Orientation on the Thermal-Electrochemical Transports in a
sient Heat Conduction: Theory and Comparison With Numerical Experi- PEM Fuel Cell,” J. Power Sources 共in press兲.
ments,” Int. J. Heat Mass Transfer, 38, pp. 2779–2796. 关23兴 Hwang, J. J., Lo, K. H., Wang, S. H., and Tsay, K. C., 2001, The 25th Con-
关2兴 Sahraoui, M., and Kaviany, M., 1994, “Slip and No-Slip Temperature Bound- ference on Theoretical and Applied Mechanics, Taichung, Taiwan, ROC.
ary Condition at the Interface of Porous, Plain Media: Convection,” Int. J. Heat 关24兴 Hwang, J. J., and Hwang, H. S., 2002, “Parametric Studies of a Double-Cell
Mass Transfer, 37, pp. 1029–1044. Stack of PEMFC Using Grafoil Flow-Field Plates,” J. Power Sources, 104,
关3兴 Quintard, M., and Whitaker, S., 2000, “Theoretical Modeling of Transport in pp. 24–32.
A Correlation for Interfacial Heat
Transfer Coefficient for Turbulent
Flow Over an Array of Square
Rods
Marcelo B. Saito
Interfacial heat transfer coefficients in a porous medium modeled as a staggered array of
Marcelo J. S. de Lemos1 square rods are numerically determined. High and low Reynolds k-⑀ turbulence models
Mem. ASME are used in conjunction of a two-energy equation model, which includes distinct transport
e-mail: delemos@ita.br equations for the fluid and the solid phases. The literature has documented proposals for
macroscopic energy equation modeling for porous media considering the local thermal
Departamento de Energia - IEME, equilibrium hypothesis and laminar flow. In addition, two-energy equation models have
Instituto Tecnológico de Aeronáutica - ITA, been proposed for conduction and laminar convection in packed beds. With the aim of
12228-900 - São José dos Campos - SP, Brazil contributing to new developments, this work treats turbulent heat transport modeling in
porous media under the local thermal nonequilibrium assumption. Macroscopic time-
average equations for continuity, momentum, and energy are presented based on the
recently established double decomposition concept (spatial deviations and temporal fluc-
tuations of flow properties). The numerical technique employed for discretizing the gov-
erning equations is the control volume method. Turbulent flow results for the macroscopic
heat transfer coefficient, between the fluid and solid phase in a periodic cell, are
presented. 关DOI: 10.1115/1.2175150兴
Keywords: turbulence modeling, porous media, heat transfer coefficient
1 Introduction medium. Sahraouri and Kaviany 关12兴 performed direct numerical

simulations of premixed combustion in a two-dimensional porous
Convection heat transfer in porous media has been extensively
medium made of in-line and staggered arrangements of square
investigated due to its many important engineering applications.
cylinders. Results therein were limited to the laminar flow regime.
The wide applications available have led to numerous investiga-
Saito and de Lemos 关13兴 considered local thermal nonequilib-
tions in this area. Such applications can be found in solar receiver
rium and obtained the interfacial heat transfer coefficient for lami-
devices, building thermal insulation, heat exchangers, energy stor-
age units, etc. From the point of view of the energy equation there nar flow using a single unit cell with local instantaneous transport
are two different models, local thermal equilibrium model and the equations.
two energy approach. The first model assumes that the solid tem- In all of the above, only laminar flow has been considered.
perature is equal to the fluid temperature, thus local thermal equi- When treating turbulent flow in porous media, however, difficul-
librium between the fluid and the solid-phases is achieved at any ties arise due to the fact that the flow fluctuates with time and a
location in the porous media. This model simplifies theoretical volumetric average is applied 关14兴. For handling such situations, a
and numerical research, but the assumption of local thermal equi- new concept called double decomposition has been proposed for
librium between the fluid and the solid is inadequate for a number developing macroscopic models for turbulent transport in porous
of problems 关1–4兴. In recent years more attention has been paid to media 关15–19兴. This methodology has been extended to nonbuoy-
the local thermal nonequilibrium model and its use has increased ant heat transfer 关20兴, buoyant flows 关21–24兴, mass transfer 关25兴,
in theoretical and numerical research for convection heat transfer and double diffusion 关26兴. In addition, a general classification of
processes in porous media 关5,6兴. models has been published 关27兴. Recently, the problem of treating
Kuwahara et al. 关7兴 proposed a numerical procedure to deter- interfaces between a porous medium and a clear region, consider-
mine macroscopic transport coefficients from a theoretical basis ing a diffusion-jump condition for laminar 关28兴 and turbulence
without any empiricism. They used a single unit cell and deter- fields 关29–31兴, have also been investigated under the concept first
mined the interfacial heat transfer coefficient for the asymptotic proposed by 关15–19兴. Following this same concept, de Lemos and
case of infinite conductivity of the solid phase. Nakayama et al. Rocamora 关32兴 have developed a macroscopic turbulent energy
关8兴 extended the conduction model of Hsu 关9兴 for treating also equation for a homogeneous, rigid, and saturated porous medium,
convection in porous media and the monographs of 关3,10,11兴 fully considering local thermal equilibrium between the fluid and solid
document forced convection in porous media. Having established matrix.
the macroscopic energy equations for both phases, useful exact Motivated by the foregoing, this work focuses on turbulent flow
solutions were obtained for two fundamental heat transfer pro- through a packed bed, which represents an important configura-
cesses associated with porous media, namely, steady conduction tion for efficient heat and mass transfer and suggests the use of
in a porous slab with internal heat generation within the solid, and equations governing thermal nonequilibrium involving distinct en-
also, thermally developing flow through a semi-infinite porous ergy balances for both the solid and fluid phases. Accordingly, the
use of such two-energy equation model requires an extra param-
1
eter to be determined, namely, the heat transfer coefficient be-
Corresponding author.
tween the fluid and the solid. The contribution herein consists in
NAL OF HEAT TRANSFER. Manuscript received April 15, 2005; final manuscript re- proposing a new correlation for obtaining the interfacial heat
ceived October 28, 2005. Review conducted by N.K. Anand. transfer coefficient for turbulent flow in a packed bed. The bed is
modeled as an infinite staggered array of square rods and the k2
range of Reynolds number, based on the size of the rod, is ex- ␮t = ␳c␮ f ␮ 共10兲
⑀
tended up to 107. In-line rod arrangement is not considered here as
the objective of this work to first consolidate results for staggered where, ␳ is the fluid density, p is the pressure, ␮ represents the
arrays. Future investigations shall consider different array ar- fluid viscosity.
rangements as well as distinct rod shapes, such as elliptical and In the above equation set ␴k, ␴⑀, c1, c2, and c␮ are dimension-
circular rods. less constants, whereas f 2 and f ␮ are damping functions. The tur-
The next sections detail the basic mathematical model, includ- bulence model constants are
ing the mean and turbulent fields for turbulent flows. Although the
discussion of turbulent motion in porous media is not presented in c␮ = 0.09, c1 = 1.5, c2 = 1.9, ␴k = 1.4, ␴⑀ = 1.3
this work, the definitions and concepts to calculate the interfacial
For the high Re model the standard constants of Launder and
heat transfer coefficient for macroscopic flows are presented.
Spalding 关33兴 were employed.
Also, the time averaged energy equations become:
Energy-fluid phase:
2 Governing Equations
共␳c p兲 f 关ⵜ · 共ūT̄ f 兲兴 = ⵜ · 共k f ⵜ T̄ f 兲 − 共␳c p兲 f ⵜ · 共u⬘T⬘f 兲共11兲
2.1 Microscopic Transport Equations. Microscopic trans-
port equations or local time-averaged transport equations for in- Energy-solid phase 共porous matrix兲:
compressible fluid flow in a rigid homogeneous porous medium
have already been presented in the literature and for that they are ⵜ · 共ks ⵜ T̄s兲 + Ss = 0 共12兲
just presented here 关32兴. The governing equations for the flow and
energy for an incompressible fluid are given by: 2.2 Decomposition of Flow Variables in Space and Time.
Continuity: Macroscopic transport modeling of incompressible flows in po-
rous media has been based on the volume-average methodology
ⵜ·u=0 共1兲 for either heat 关34兴 or mass transfer 关35,36兴. If time fluctuations of
Momentum: the flow properties are also considered, in addition to spatial de-
冋册
viations, there are two possible methodologies to follow in order
⳵u to obtain macroscopic equations: 共a兲 application of time-average
␳ + ⵜ · 共uu兲 = − ⵜp + ␮ⵜ2u 共2兲
⳵t operator followed by volume-averaging 关37–42兴, or 共b兲 use of
volume-averaging before time-averaging is applied 关43–45兴.
Energy-fluid phase:
再冎
However, both sets of macroscopic mass transport equations are
⳵Tf equivalent when examined under the recently established double
共 ␳ c p兲 f + ⵜ · 共uT f 兲 = ⵜ · 共k f ⵜ T f 兲 + S f 共3兲 decomposition concept 关15–19兴. As mentioned, the double decom-
⳵t position concept has been published in a number of worldwide
Energy-solid phase 共porous matrix兲: available journal articles 关15–32兴 and does not need to be repeated
here.
⳵ Ts
共 ␳ c p兲 s = ⵜ · 共ks ⵜ Ts兲 + Ss 共4兲 2.3 Macroscopic Flow and Energy Equations. When the
⳵t
average operators are simultaneously applied over Eqs. 共1兲 and
where the subscripts f and s refer to fluid and solid phases, re- 共2兲, macroscopic equations for turbulent flow are obtained. Vol-
spectively. Here, T is the temperature, k f is the fluid thermal con- ume integration is performed over a REV Refs. 关14,46兴, resulting
ductivity, ks is the solid thermal conductivity, c p is the specific in
heat, and S is the heat generation term. If there is no heat genera- Continuity:
tion either in the solid or in the fluid, one has further S f = Ss = 0.
For turbulent flows the time averaged transport equations can ⵜ · ūD = 0 共13兲
be written as:
Continuity: where, ūD = ␾具ū典 and 具ū典 identifies the intrinsic 共liquid兲 average
i i
of the time-averaged velocity vector ū.

ⵜ · ū = 0 共5兲 Momentum:
冋冉冊册
Momentum:
⳵ ūD ūDūD
␳ f 关ⵜ · 共uu兲兴 = − ⵜp̄ + ⵜ · 兵␮关ⵜū + 共ⵜū兲T兴 − ␳u⬘u⬘其共6兲 ␳ +ⵜ·
⳵t ␾
where the low and high Reynolds k-⑀ model is used to obtain the = − ⵜ共␾具p̄典i兲 + ␮ⵜ2ūD − ⵜ · 共␳␾具u⬘u⬘典i兲
冋册
eddy viscosity, ␮t, whose equations for the turbulent kinetic en-
ergy per unit mass and for its dissipation rate read: ␮␾ cF␾␳兩ūD兩ūD
− ūD + 共14兲
Turbulent kinetic energy per unit mass: K 冑K
␳ f 关ⵜ · 共ūk兲兴 = ⵜ · 冋冉冊册
␮+
␮t
␴k
ⵜ k − ␳u⬘u⬘:ⵜū − ␳⑀ 共7兲 where the last two terms in Eq. 共14兲 represent the Darcy and
Forchheimer contributions by 关47兴. The symbol K is the porous
Turbulent kinetic energy per unit mass dissipation rate: medium permeability, cF is the form drag or Forchheimer coeffi-
cient, 具p̄典i is the intrinsic average pressure of the fluid, and ␾ is the
␳ f 关ⵜ · 共ū⑀兲兴 = ⵜ · 冋冉冊册␮+
␮t
␴⑀
ⵜ ⑀ + 关c1共− ␳u⬘u⬘:ⵜū兲 − c2 f 2␳⑀兴
⑀
k
porosity of the porous medium.
The macroscopic Reynolds stress, −␳␾具u⬘u⬘典i, appearing in Eq.
共8兲共14兲 is given as
Reynolds stresses and the eddy viscosity is given by, respectively, 2

− ␳␾具u⬘u⬘典i = ␮t␾2具D̄典␯ − ␾␳具k典iI 共15兲
2 3
− ␳u⬘u⬘ = ␮t关ⵜū + 共ⵜū兲兴 − ␳kI T
共9兲
3 where
具D̄典␯ = 2 关ⵜ共␾具ū典i兲 + 关ⵜ共␾具ū典i兲兴T兴
1
共16兲 i.e., thermal dispersion due to spatial deviations, turbulent heat
flux due to temporal fluctuations, turbulent thermal dispersion due
is the macroscopic deformation tensor, 具k典i = 具u⬘u⬘典i / 2 is the in- to temporal fluctuations, and spatial deviations and local
trinsic turbulent kinetic energy, and ␮t␾, is the turbulent viscosity, conduction.
which is modeled in 关27兴 similarly to the case of clear flow, in the Using these gradient type diffusion models, we can write:
form, Turbulent heat flux:
2 − 共␳c p兲 f 共␾具u⬘典i具T⬘f 典i兲 = Kt · ⵜ具T̄ f 典i 共22兲
具k典i
␮ t␾ = ␳ c ␮ i 共17兲 Thermal dispersion:
具⑀典
The intrinsic turbulent kinetic energy per unit mass and its dis- − 共␳c p兲 f 共␾具 iūiT f 典i兲 = Kdisp · ⵜ具T̄ f 典i 共23兲
sipation rate are governed by the following equations:
册冋冉册
Turbulent thermal dispersion:
␳ 冋 ⳵
⳵t
共␾具k典i兲 + ⵜ · 共ūD具k典i兲 = ⵜ · ␮+
␮ t␾
␴k
冊 ⵜ 共␾具k典i兲 − 共␳c p兲 f 共␾具 iu⬘ iT⬘ f 典i兲 = Kdisp,t · ⵜ具T̄ f 典i
Local conduction:
共24兲
− ␳具u⬘u⬘典i:ⵜūD + ck␳
␾具k典i兩ūD兩
冑K ⵜ· 冋冕 1
⌬V Ai
册
nik f T f dA = K f,s · ⵜ具T̄s典i
冋冕册
− ␳␾具⑀典i 共18兲
冋册冋冉冊册
1
⳵ ␮ t␾ ⵜ· nik f Ts dA = Ks,f · ⵜ具T̄ f 典i 共25兲
␳ 共␾具⑀典i兲 + ⵜ · 共ūD具⑀典i兲 = ⵜ · ␮+ ⵜ 共␾具⑀典兲
i ⌬V Ai
⳵t ␴⑀
For the above shown expressions, Eqs. 共20兲 and 共21兲 can be writ-
具⑀典i ten as:
+ c1共− ␳具u⬘u⬘典 :ⵜūD兲 i i
具k典
⳵ 具T̄典i
␾具⑀典i兩ūD兩具⑀典i
2 兵共␳c p兲 f ␾其 + 共␳c p兲 f ⵜ · 共uD具T f 典i兲
+ c 2c k␳ − c 2␳ ␾ i ⳵t
冑K 具k典
= ⵜ · 兵Keff,f · ⵜ具T̄ f 典i其 + hiai共具Ts典i − 具T f 典i兲共26兲
共19兲
where, ck, c1, c2, and c␮ are nondimensional constants. The sec- ⳵ 具T̄典i
ond terms on the left-hand side of Eqs. 共18兲 and 共19兲 represent the 兵共1 − ␾兲共␳c p兲s其 = ⵜ · 兵Keff,s · ⵜ具Ts典i其 + hiai共具Ts典i − 具T f 典i兲
⳵t
generation rate of 具k典i and 具⑀典i, respectively, due to the mean gra-
共27兲
dient of ūD. The third terms in the same equations are the genera-
tion rates due to the action of the porous matrix 共see 关16兴兲. where, Keff,f and Keff,s are the effective conductivity tensor for
Similarly, macroscopic energy equations are obtained for both fluid and solid, respectively, given by:
fluid and solid phases by applying time and volume average op-
Keff,f = ␾k f I + K f,s + Kt + Kdisp + Kdisp,t 共28兲
erators to Eqs. 共3兲 and 共4兲. As in the flow case, volume integration
is performed over a REV, resulting in Keff,s = 关共1 − ␾兲ks兴I + Ks,f 共29兲
共 ␳ c p兲 f 冋
⳵ ␾具T f 典i
⳵t
+ ⵜ · 兵␾共具ū典i具T f 典i + 具 iuiT f 典i
and I is the unit tensor. Details of interfacial convective heat trans-
fer coefficient are presented next section.
册
In order to be able to apply Eq. 共26兲, it is necessary to deter-
mine the dispersion and conductivity tensors in Eq. 共28兲, i.e., K f,s,
+ 具u⬘典i具T⬘f 典i + 具 iu⬘ iT⬘ f 典i兲其
Kt, Kdisp, and Kdisp,t. Following Kuwahara and Nakayama 关38兴
冋册
and Quintard et al. 关5兴, this can be accomplished for the thermal
= ⵜ · k f ⵜ 共␾具T f 典i兲 +
1
⌬V
冕Ai
nik f T f dA
dispersion and conductivity tensors, K f,s and Kdisp, by making use
of a unit cell subjected to periodic boundary conditions for the
flow and a linear temperature gradient, to represent the porous
+ hiai共具Ts典i − 具T f 典i兲共20兲 medium. The dispersion and conductivity tensors are then ob-
再冎
tained directly from the microscopic results for the unit cell, using
⳵ 共1 − ␾兲具Ts典i Eqs. 共23兲 and 共25兲. Besides, it can be used for the following cor-
共 ␳ c p兲 s relations for the transverse and longitudinal components of the
再冎
⳵t thermal dispersion tensor, which are valid for PeD 艌 10:
=ⵜ· ks ⵜ 关共1 − ␾兲具Ts典i兴 −

1
⌬V
冕 Ai
niksTs dA
共kdis兲xx
kf
= 2.1
PeD
共1 − ␾兲0.1
, for longitudinal dispersion 共30兲
− hiai共具Ts典i − 具T f 典i兲共21兲共kdis兲yy

= 0.052共1 − ␾兲0.5 PeD, for transverse dispersion
where, hi and ai are the interfacial convective heat transfer coef- kf
ficient and surface area per unit volume, respectively. 共31兲
2.4 Macroscopic Two-Energy Equation Modeling. In order The turbulent heat flux and turbulent thermal dispersion terms,
to apply Eqs. 共20兲 and 共21兲 to obtain the temperature field for Kt and Kdisp,t, which can not be determined from such a micro-
turbulent flow in porous media, the underscored terms have to be scopic calculation, are modeled through the eddy diffusivity con-
modeled in some way as a function of the intrinsically averaged cept, similarly to Nakayama and Kuwahara 关42兴. It should be
temperature of solid phase and fluid, 具Ts典i and 具T f 典i. To accom- noticed that these terms arise only if the flow in the porous me-
plish this, a gradient-type diffusion model is used for all the terms, dium is turbulent, whereas the thermal dispersion terms exist for
Fig. 2 Nonuniform computational grid
Fig. 1 Physical model and coordinate system metrically ordered medium a repetitive cell can be identified. Pe-
riodic boundary conditions were then applied for obtaining the
temperature distribution under fully developed flow conditions. A
both laminar and turbulent flow regimes. Starting out from the numerical correlation for the interfacial convective heat transfer
time averaged energy equation coupled with the microscopic coefficient was proposed by Kuwahara et al. 关7兴 for laminar flow
modeling for the “turbulent thermal stress tensor” through the as,
microscopic eddy diffusivity, ⌫T = ␮t / ␴T, one can write:
␯t
h iD
kf
冉
= 1+
4共1 − ␾兲
␾
1
2
冊
+ 共1 − ␾兲1/2 Re0.6 1/3
D Pr , valid for
− 共␳c p兲 f u⬘T⬘f = 共␳c p兲 f ⵜ T̄ f 共32兲
␴T 0.2 ⬍ ␾ ⬍ 0.9 共37兲
where ␴T is the turbulent Prandtl number which is taken here as a Equation 共37兲 is based on porosity dependency and is valid for
constant. packed beds of particle diameter D.
Applying the volume average to the resulting equation, one Saito and de Lemos 关13兴 obtained the interfacial heat transfer
obtains the macroscopic version of the “turbulent thermal stress coefficient for laminar flows though an infinite square rod; this
tensor,” given by: same physical model will be used here for obtaining the interfacial
␯ t␾ heat transfer coefficient hi for turbulent flows.
− 共␳c p兲 f 具u⬘T⬘f 典i = 共␳c p兲 f ⵜ 具T̄ f 典i 共33兲 The flow through an infinite square rod can be associated with
␴T flow across a bundle of tubes. Furthermore the heat transfer coef-
where we have adopted the symbol ␯t␾ to express the macroscopic ficient related to a tube is determined by its position in the pack-
age. The tube rows of a bundle are either aligned or staggered in
version of the eddy viscosity, ␮t␾ = ␳ f ␯t␾.
the direction of the fluid velocity. In this work the geometric ar-
Equation 共33兲 is the sum of the turbulent heat flux and the rangement is staggered 共see Fig. 1兲. For the staggered configura-
turbulent thermal dispersion found by Rocamora and de Lemos tion Zhukauskas 关27兴 has proposed a correlation of the form,
关20兴. In view of the arguments given above, the turbulent heat flux
and turbulent thermal dispersion components of the conductivity h iD
= 0.022 Re0.84
D Pr
0.36
共38兲
tensor, Kt and Kdisp,t, respectively, will be expressed as: kf
␯ t␾ where the values 0.022 and 0.84 are constants for tube bank in
Kt + Kdisp,t = ␾共␳c p兲 f I 共34兲 cross flow and for this particular case 2 ⫻ 105 ⬍ ReD ⬍ 2 ⫻ 106.
␴T
2.5 Interfacial Heat Transfer Coefficient. In Eqs. 共20兲 and

共21兲 the heat transferred between the two phases can be modeled 3 Periodic Cell and Boundary Conditions
by means of a film coefficient hi such that, The macroscopic hydrodynamic and thermodynamic behavior
hiai共具Ts典i − 具T f 典i兲 =
1
⌬V
冕
⌳i
ni · k f ⵜ T f dA =
1
⌬V
冕 ⌳i
ni · ks ⵜ Ts dA
of practical interest can be obtained from the direct application of
the first principles to viscous flow and heat transfer at a pore scale.
In reality, however, it is impossible to resolve the details of the
共35兲 flow and heat transfer fields within a real porous medium. Na-
kayama et al. 关8兴 and Kuwahara et al. 关7兴 modeled a porous me-
where, ai = Ai / ⌬V. dium in terms of obstacles arranged in a regular pattern, and
Wakao et al. 关48兴 obtained a heuristic correlation for closely solved the set of the microscopic governing equations, exploiting
packed bed, of particle diameter D and compared their with ex- periodic boundary conditions.
perimental data. This correlation for the interfacial heat transfer In order to evaluate the numerical tool to be used in the deter-
coefficient is given by, mination of the film coefficient given by Eq. 共35兲, a test case was
h iD run for obtaining the flow field in a periodic cell, which is here
= 2 + 1.1 Re0.6
D Pr
1/3
共36兲 assumed to represent the porous medium. Consider a macroscopi-
kf cally uniform flow through an infinite number of square rods of
For numerically determining hi, Kuwahara et al. 关7兴 modeled a lateral size D, placed in a staggered arrangement and maintained
porous medium by considering an infinite number of solid square at constant temperature Tw. The periodic cell or representative
rods of size D, arranged in a regular triangular pattern 共see Fig. 1兲. elementary volume, ⌬V is schematically showed in Fig. 1 and has
They numerically solved the governing equations in the void re- dimensions 2H ⫻ H. Computations within this cell were carried
gion, exploiting to advantage the fact that for an infinite and geo- out using a nonuniform grid, as shown in Fig. 2, to ensure that the
results were grid independent. The Reynolds number ReD
= ␳ūDD / ␮ was varied from 104 to 107 and the porosity, ␾ = 1
− 共D / H兲2.
The numerical method utilized to discretize the flow and energy
equations in the unit cell is the finite control volume approach.
The SIMPLE method of Patankar 关49兴 was used for handling Eqs.
共1兲–共4兲 the velocity-pressure coupling. Convergence was moni-
tored in terms of the normalized residue for each variable. The
maximum residue allowed for convergence check was set to 10−9,
being the variables normalized by appropriate reference values.
For fully developed flow in the cell of Fig. 1, the velocity at
exit 共x / H = 2兲 must be identical to that at the inlet 共x / H = 0兲. Tem-
perature profiles, however, are only identical at both the cell exit
and inlet if presented in terms of an appropriate nondimensional
variable. The situation is analogous to the case of forced convec-
tion in a channel with isothermal walls. Due to the periodicity of Fig. 3 Velocity profile in fully developed pipe flow
the model and a single structural unit as indicated in Fig. 1 may be
taken as a calculation domain. The equations used for turbulent
flow in the unit cell are Eqs. 共5兲, 共6兲, and 共11兲. Computations are based on the Darcy velocity, the length of
Thus, boundary conditions and periodic constraints are given
by: structural unit H, and the temperature difference 共T̄B共x兲 − T̄w兲, as
On the solid walls 共Low Re Model兲: references scales.
⳵ 2k 3.1 Film Coefficient hi. Determination of hi is here obtained

ū = 0, k = 0, ⑀=␯ , T̄ = T̄w 共39兲 by calculating, for the unit cell of Fig. 1, an expression given as,
⳵ y2
Qtotal
On the solid walls 共High Re Model兲: hi = 共45兲
Ai⌬Tml
ū 1 u2␶ c␮3/4k3/2 where Ai = 8Dx1. The overall heat transferred in the cell, Qtotal, is
= ln共y +E兲, k= , ⑀= w
,
u␶ ␬ c␮1/2 ␬yw giving by,
共␳c p兲 f c␮1/4k1/2
w 共T̄ − T̄w兲
Qtotal = 共H − D兲␳ūBc p共T̄B兩outlet − T̄B兩inlet兲共46兲
冉冊
qw = 共40兲
Prt The bulk mean velocity of the fluid is given by:
ln共y +w兲 + cQ共Pr兲
where,
␬
冕 ū dy
冕
ūB共x兲共47兲
u␶ =
␶w
␳
冉冊 1/2
, y +w =
y wu ␶
␯
, dy
cQ = 12.5 Pr2/3 + 2.12 ln共Pr兲 − 5.3 for Pr ⬎ 0.5 and the logarithm mean temperature difference, ⌬Tml is,
where, Pr and Prt are Prandtl and turbulent Prandtl number, re- 共T̄w − T̄B兩outlet兲 − 共T̄w − T̄B兩inlet兲
spectively, qw is wall heat flux, u␶ is wall-friction velocity, y w is ⌬Tml = 共48兲
the coordinate normal to wall, ␬ is a constant for turbulent flow ln关共T̄w − T̄B兩outlet兲共w̄ − T̄B兩inlet兲兴
past smooth impermeable walls or von Kármán’s constant, and E Equation 共46兲 represents an overall heat balance on the entire
is an integration constant that depends on the roughness of the cell and associates the heat transferred to the fluid to a suitable
wall. For smooth walls with constant shear stress E = 9. temperature difference ⌬Tml. As mentioned earlier, Eqs. 共1兲–共4兲
On the symmetry planes: were numerically solved in the unit cell until conditions 共42兲 and
共43兲 were satisfied.
¯ ⳵k ⳵⑀
⳵ ū ⳵␯
= = = =0 共41兲
⳵y ⳵y ⳵y ⳵y
where ū and ¯␯ are components of u.
On the periodic boundaries: 4.1 Periodic Flow. Results for velocity and temperature fields
were obtained for different Reynolds numbers. In order to assure
ū兩inlet = ū兩outlet, ¯␯兩inlet = ¯␯兩outlet, k兩inlet = k兩outlet, ⑀兩inlet = ⑀兩outlet that the flow was hydrodynamically and thermally developed in
the periodic cell of Fig. 1, the governing equations were solved
共42兲
repetitively in the cell, taking the outlet profiles for ū and ␪ at the
冏冏冏冏
exit and plugging them back at the inlet. In the first run, uniform
T̄ − T̄w T̄ − T̄w velocity and temperature profiles were set at the cell entrance for
␪兩inlet = ␪兩outlet Û = 共43兲 Pr= 1 giving ␪ = 1 at x / H = 0. Then, after convergence of the flow
T̄B共x兲 − T̄w inlet T̄B共x兲 − T̄w outlet and temperature fields, ū and ␪ at x / H = 2 were used as inlet
The bulk mean temperature of the fluid is given by: profiles for a second run, corresponding to solving again the flow
for a similar cell beginning in x / H = 2. Similarly, a third run was
冕 ūT̄ dy
carried out and again outlet results, this time corresponding to an
axial position x / H = 4, were recorded. This procedure was re-
peated several times until ū and ␪ did not differ substantially at
冕
T̄B共x兲 = 共44兲
both inlet and outlet positions. Figure 3 further shows that the
ū dy velocity profile here obtained in with a low Re model has a good
agreement within the laminar and the wall log layers.
Grid independence studies are summarized in Fig. 4, which
presents results for Nu as a function of the number of grid points.
The subscript conv refers to the asymptotic value as the grid in-
creases. The figure indicates that for grids greater than 12,000,
errors in Nu are less that 1%. For that, all results presented below
considered this grid size.
Nondimensional velocity and temperature profiles are shown in
Figs. 5 and 6, respectively, showing that the periodicity con-
straints imposed by Eqs. 共42兲 and 共43兲 was satisfied for x / H ⬎ 4.
For the entrance region 共0 ⬍ x / H ⬍ 4兲, ␪ profiles change with
length x / H being essentially invariable after this distance. Under
this condition of constant ␪ profile, the flow was considered to be
macroscopically developed for ReD up to 107.
For the low Re model, the first node adjacent to the wall re-
Fig. 4 Grid independence study quires that the nondimensional wall distance be such that y +
= u␶y ␳ / ␮ 艋 1. To accomplish this requirement, the grid needs a
greater number of points close to the wall leading to computa-
tional meshes of large sizes. As a further code validation for tur-
bulent flow calculation, which uses the k − ⑀ model, a developing
turbulent channel flow has been solved for Re= 5 ⫻ 104, where Re
is based on the duct hydraulic diameter.
4.2 Developed Flow and Temperature Fields. Macroscopi-
cally developed flow field for Pr= 1 and ReD = 5 ⫻ 104 is presented
in Fig. 5, corresponding to x / D = 6 at the cell inlet. The expression
“macroscopically developed” is used herein to account for the fact
that periodic flow has been achieved at that axial position. Figures
7–9 show distributions of pressure, isotherms, and turbulence ki-
netic energy in a microscopic porous structure, obtained at ReD
= 105 for cases of ␾ = 0.65. Pressure increases at the front face of
the square rod and drastically decreases around the corner, as can
be seen from the pressure contours shown in Fig. 7.
Fig. 5 Dimensionless velocity profile for Pr= 1 and ReD = 5 Temperature distribution is shown in Fig. 8. Colder fluid im-
Ã 104 pinges on the left-hand side of the rod yielding strong temperature
gradients on that face. Downstream the obstacles, fluid recircula-
tion smooths temperature gradients and deforms isotherms within
the mixing region. When the Reynolds number is sufficiently high
共not shown here兲, thermal boundary layers cover the rod surfaces
indicating that convective heat transfer overwhelms thermal dif-
fusion. Figure 9 presents levels of turbulence kinetic energy,
which are higher around the rod corners where a strong shear
layer is formed. Further downstream the rods in the weak region,
steep velocity gradients appear due to flow deceleration, also in-
creasing the local level of k.
Once fully developed flow and temperature fields are achieved,
for the fully developed condition 共x ⬎ 6H兲, bulk temperatures
were calculated according to Eq. 共44兲, at both inlet and outlet
positions. They were then used to calculate hi using Eqs.
共45兲–共48兲. Results for hi are plotted in Fig. 10 for ReD up to 107.
Also plotted in this figure are results computed with correlation
Fig. 6 Dimensionless temperature profile for Pr= 1 and ReD 共37兲 by Kuwahara et al. 关7兴 using ␾ = 0.65. The figure seems to
= 5 Ã 104
Fig. 7 Nondimensional pressure field for ReD = 105 and ␾ = 0.65
Fig. 8 Isotherms for Pr= 1, ReD = 105, and ␾ = 0.65
Fig. 9 Turbulence kinetic energy for ReD = 105 and ␾ = 0.65
indicate that both computations show a reasonable agreement for

laminar results. In addition, numerical results for turbulent flow
using low and High Re models are also presented in this figure.
h iD
kf
= 0.08冉冊
ReD
␾
0.8
Pr1/3 for
Figure 11 shows numerical results for the interfacial convective ReD

heat transfer coefficient for various porosities 共␾ = 0.44, ␾ = 0.65, 1.0 ⫻ 104 ⬍ ⬍ 2.0 ⫻ 107 valid for
␾
and ␾ = 0.90兲. Results for hi are plotted for ReD up to 107. In order
to obtain a correlation for hi in the turbulent regime, all curves 0.2 ⬍ ␾ ⬍ 0.9 共49兲
were first collapsed after plotting them in terms of ReD / ␾, as Equation 共49兲, which gives the heat transfer coefficient for tur-
shown in Fig. 12. Furthermore, the least squares technique was bulent flow, is compared with numerical results obtained with low
applied in order to determine the best correlation, which lead to a and high Re models. Such comparison is presented in Fig. 13,
minimum overall error. Thus, the following expression is here which also shows computations using correlations given by Eqs.
proposed: 共36兲 and 共37兲 by Wakao et al. 关48兴 and Zhukauskas 关50兴, respec-
tively. The agreement between the present correlation, other cor-
relations in the literature, and the numerical simulations stimulates
further investigation on this subject, contributing towards the
building of a more general expression for the interfacial heat
transfer coefficient for porous media.
5 Concluding Remarks
A computational procedure for determining the convective co-
efficient of heat exchange between the porous substrate and the
working fluid for a porous medium was detailed. As a preliminary
result, macroscopically uniform laminar and turbulent flow
through a periodic cell of isothermal square rods was computed,
considering periodical velocity and temperature fields. Quantita-
tive agreement was obtained when comparing laminar results
herein with simulations by Kuwahara et al. 关7兴. For turbulent
flows, low and high Reynolds turbulence models were employed
in order to obtain the interfacial heat transfer coefficient. A corre-
lation was then proposed for such coefficients. Further work will
be carried out in order to simulate fully turbulent flow and heat
Fig. 10 Effect of ReD on hi for Pr= 1 and ␾ = 0.65 transfer in a porous medium formed by arrays of elliptic, cylin-
Fig. 11 Effect of porosity on hi for Pr= 1 Fig. 13 Comparison of the numerical results and various cor-
relations for ␾ = 0.65
ReD ⫽ Reynolds number based on D and the macro-

scopically uniform velocity
T ⫽ temperature
T̄ ⫽ time averaged temperature
u ⫽ microscopic velocity
uD ⫽ Darcy or superficial velocity 共volume average
of u兲
Greek Symbols
␣ ⫽ fluid thermal diffusivity
⌬V ⫽ representative elementary volume
⌬V f ⫽ fluid volume inside ⌬V
␮ ⫽ fluid dynamic viscosity
␮t ⫽ eddy viscosity
Fig. 12 Comparison of the numerical results and proposed ␮ t␾ ⫽ macroscopic eddy viscosity
correlation ␯ ⫽ fluid kinematic viscosity
␳ ⫽ fluid density
␪ ⫽ dimensionless temperature
drical and transverse elliptic rods, displaced in in-line as well as ␾ ⫽ ␾ = ⌬V f / ⌬V, porosity
staggered arrangements. Ultimately, it is expected that a more ␴T ⫽ turbulent Prandtl number
general correlation for hi be obtained to be used in conjunction ␸ ⫽ general variable
with macroscopic two-energy equation models. 具␸典i ⫽ intrinsic average
具␸典␯ ⫽ volume average
Nomenclature
i
␸ ⫽ spatial deviation
Ai ⫽ interface total area between the fluid and solid
Acknowledgment
cF ⫽ Forchheimer coefficient
cp ⫽ fluid specific heat The authors are thankful to CNPq and FAPESP, Brazil, for their
D ⫽ square rods of lateral size financial support during the course of this research.
hi ⫽ interfacial convective heat transfer coefficient
H ⫽ periodic cell height References
I ⫽ unit tensor 关1兴 Schumann, T. E. W., 1929, “Heat Transfer: Liquid Flowing Through a Porous
K ⫽ permeability Prism,” J. Franklin Inst., 208, pp. 405–416.
k ⫽ turbulence kinetic energy per unit mass 关2兴 Quintard, M., 1998, “Modeling Local Non-Equilibrium Heat Transfer in Po-
rous Media,” in Proc. 11th Int. Heat Transfer Conf., Kyongyu, Korea, Vol. 1,
kf ⫽ fluid thermal conductivity pp. 279–285.
ks ⫽ solid thermal conductivity 关3兴 Kuznetsov, A. V., 1998, “Thermal Nonequilibrium Forced Convection in Po-
Kdisp ⫽ dispersion conductivity tensor rous Media,” Chap. 5 in Transport Phenomena in Porous Media, D. B. Ingham
and I. Pop, eds., Elsevier, Oxford, pp. 103–129.
K f,s ⫽ two-equation model effective thermal conduc- 关4兴 Kaviany, M., 1995, Principles of Heat Transfer in Porous Media, 2nd ed.,
tivity tensor in fluid phase Springer, New York.
Ks,f ⫽ two-equation model effective thermal conduc- 关5兴 Quintard, M., Kaviany, M., and Whitaker, S., 1997, “Two-Medium Treatment
tivity tensor in solid phase of Heat Transfer in Porous Media: Numerical Results for Effective Properties,”
Adv. Water Resour., 20, pp. 77–94.
Kt ⫽ turbulence conductivity tensor 关6兴 Ochoa-Tapia, J. A., and Whitaker, S., 1997, “Heat Transfer at the Boundary
Kdisp,t ⫽ turbulent dispersion tensor Between a Porous Medium and a Homogeneous Fluid,” Int. J. Heat Mass
Nu ⫽ Interfacial Nusselt number; Nu= hiD / k f Transfer, 40, pp. 2691–2707.
关7兴 Kuwahara, F., Shirota, M., and Nakayama, A., 2001, “A Numerical Study of
P ⫽ pressure Interfacial Convective Heat Transfer Coefficient in Two-Energy Equation
P* ⫽ P*共P − Pmin兲 / 共Pmax − Pmin兲, nondimensional Model for Convection in Porous Media,” Int. J. Heat Mass Transfer, 44, pp.
Pressure 1153–1159.
关8兴 Nakayama, A., Kuwahara, F., Sugiyama, M., and Xu, G., 2001, “A Two-
Pr ⫽ Pr= ␯ / ␣, Prandtl number Energy Equation Model for Conduction and Convection in Porous Media,” Int.
PeD ⫽ Peclet number based on D and the macroscopi- J. Heat Mass Transfer, 44, pp. 4375–4379.
cally uniform velocity 关9兴 Hsu, C. T., 1999, “A Closure Model for Transient Heat Conduction in Porous
Media,” ASME J. Heat Transfer, 121, pp. 733–739. Occupied by a Porous Layer Considering the Stress Jump at the Interface,” Int.
关10兴 Nield, D. A., and Bejan, A., 1992, Convection in Porous Media, Springer, New J. Heat Mass Transfer, 46共26兲, pp. 5113–5121.
York. 关30兴 de Lemos, M. J. S., 2005, “Turbulent Kinetic Energy Distribution Across the
关11兴 Bear, J., 1972, Dynamics of Fluids in Porous Media, American Elsevier, New Interface Between a Porous Medium and a Clear Region,” Int. Commun. Heat
York. Mass Transfer, 32共1-2兲, pp. 107–115.
关12兴 Sahraouri, M., and Kaviany, M., 1994, “Direct Simulation Versus Volume- 关31兴 de Lemos, M. J. S., and Silva, R. A., 2006, “Turbulent Flow Over A Layer Of
Averaged Treatment of Adiabatic, Premixed Flame in a Porous Medium,” Int. A Highly Permeable Medium Simulated With A Diffusion-Jump Model For
J. Heat Mass Transfer, 37共18兲, pp. 2817–2834. The Interface,” Int. J. Heat Mass Transfer, 49共3-4兲, pp. 546–556.
关13兴 Saito, M. B., and de Lemos, M. J. S., 2005, “Interfacial Heat Transfer Coef- 关32兴 de Lemos, M. J. S., and Rocamora, F. D., 2002, “Turbulent Transport Model-
ficient for NonEquilibrium Convective Transport in Porous Media,” Int. Com- ing for Heated Flow in Rigid Porous Media, in Proceedings of the Twelfth
mun. Heat Mass Transfer, 32共5兲, pp. 667–677. International Heat Transfer Conference, Grenoble, France, August 18–23, pp.
关14兴 Gray, W. G., and Lee, P. C. Y., 1977, “On the Theorems for Local Volume 791–795.
Averaging of Multiphase System,” Int. J. Multiphase Flow, 3, pp. 333–340. 关33兴 Launder, B. E., and Spalding, D. B., 1974, “The Numerical Computation of
关15兴 Pedras, M. H. J., and de Lemos, M. J. S., 2000, “On the Definition of Turbu- Turbulent Flows,” Comput. Methods Appl. Mech. Eng., 3, pp. 269–289.
lent Kinetic Energy for Flow in Porous Media,” Int. Commun. Heat Mass 关34兴 Hsu, C. T., and Cheng, P., 1990, “Thermal Dispersion in a Porous Medium,”
Transfer, 27共2兲, pp. 211–220. Int. J. Heat Mass Transfer, 33, pp. 1587–1597.
关16兴 Pedras, M. H. J., and de Lemos, M. J. S., 2001, “Macroscopic Turbulence 关35兴 Whitaker, S., 1966, “Equations of Motion in Porous Media,” Chem. Eng. Sci.,
Modeling for Incompressible Flow Through Undeformable Porous Media,” 21, pp. 291–300.
Int. J. Heat Mass Transfer, 44共6兲, pp. 1081–1093.
关36兴 Whitaker, S., 1967, “Diffusion and Dispersion in Porous Media,” AIChE J.,
关17兴 Pedras, M. H. J., and de Lemos, M. J. S., 2001, “Simulation of Turbulent Flow
13共3兲, pp. 420–427.
in Porous Media Using a Spatially Periodic Array and a Low-Re Two-Equation
关37兴 Masuoka, T., and Takatsu, Y., 1996, “Turbulence Model for Flow Through
Closure,” Numer. Heat Transfer, Part A, 39共1兲, pp. 35–59.
Porous Media,” Int. J. Heat Mass Transfer, 39共13兲, pp. 2803–2809.
关18兴 Pedras, M. H. J., and de Lemos, M. J. S., 2001, “On the Mathematical De-
关38兴 Kuwahara, F., Nakayama, A., and Koyama, H., 1996, “A Numerical Study of
scription and Simulation of Turbulent Flow in a Porous Medium Formed by an
Array of Elliptic Rods,” J. Fluids Eng., 123共4兲, pp. 941–947. Thermal Dispersion in Porous Media,” ASME J. Heat Transfer, 118, pp. 756–
关19兴 Pedras, M. H. J., and de Lemos, M. J. S., 2003, “Computation of Turbulent 761.
关39兴 Kuwahara, F., and Nakayama, A., 1998, “Numerical Modeling of Non-Darcy
Flow in Porous Media Using a Low Reynolds k-⑀ Model and an Infinite Array
of Transversally-Displaced Elliptic Rods,” Numer. Heat Transfer, Part A, Convective Flow in a Porous Medium, Heat Transfer 1998, Proceedings of the
43共6兲, pp. 585–602. 11th Int. Heat Transf. Conf., Kyongyu, Korea, Taylor and Francis, Washington,
关20兴 Rocamora, F. D. Jr., and de Lemos, M. J. S., 2000, “Analysis of Convective D.C., Vol. 4, pp. 411–416.
Heat Transfer of Turbulent Flow in Saturated Porous Media,” Int. Commun. 关40兴 Kuwahara, F., Kameyama, Y. Yamashita, S. and Nakayama, A., 1998, “Nu-
Heat Mass Transfer, 27共6兲, pp. 825–834. merical Modeling of Turbulent Flow in Porous Media Using a Spatially Peri-
关21兴 de Lemos, M. J. S., and Braga, E. J., 2003, “Modeling of Turbulent Natural odic Array,” J. Porous Media, 1共1兲, pp. 47–55.
Convection in Saturated Rigid Porous Media,” Int. Commun. Heat Mass 关41兴 Ergun, S., 1952, “Fluid Flow Through Packed Columns,” Chem. Eng. Prog.,
Transfer, 30共5兲, pp. 615–624. 48, pp. 89–94.
关22兴 Braga, E. J., and de Lemos, M. J. S., 2004, “Turbulent Natural Convection in 关42兴 Nakayama, A., and Kuwahara, F., 1999, “A Macroscopic Turbulence Model
a Porous Square Cavity Computed with a Macroscopic k-⑀ Model,” Int. J. Heat for Flow in a Porous Medium,” J. Fluids Eng., 121, pp. 427–433.
Mass Transfer, 47共26兲, pp. 5639–5650. 关43兴 Lee, K., and Howell, J. R., 1987, “Forced Convective and Radiative Transfer
关23兴 Braga, E. J., and de Lemos, M. J. S., 2005, “Heat Transfer in Enclosures Within a Highly Porous Layer Exposed to a Turbulent External Flow Field,” in
Having a Fixed Amount of Solid Material Simulated with Heterogeneous and Proceedings of the 1987 ASME-JSME Thermal Engineering Joint Conf., Ho-
Homogeneous Models,” Int. J. Heat Mass Transfer, 48共23-24兲, pp. 4748– nolulu, Hawaii, ASME, New York, Vol. 2, pp. 377–386.
4765. 关44兴 Antohe, B. V., and Lage, J. L., 1997, “A General Two-Equation Macroscopic
关24兴 Braga, E. J., and de Lemos, M. J. S., 2000, “Laminar Natural Convection in Turbulence Model for Incompressible Flow in Porous Media,” Int. J. Heat
Cavities Filled with Circular and Square Rods,” Int. Commun. Heat Mass Mass Transfer, 40共13兲, pp. 3013–3024.
Transfer, 32共10兲, pp. 1289–1297. 关45兴 Getachewa, D., Minkowycz, W. J., and Lage, J. L., 2000, “A Modified Form of
关25兴 de Lemos, M. J. S., and Mesquita, M. S., 2003, “Turbulent Mass Transport in the k-⑀ Model for Turbulent Flow of a Incompressible Fluid in Porous Media,”
Saturated Rigid Porous Media,” Int. Commun. Heat Mass Transfer, 30共1兲, pp. Int. J. Heat Mass Transfer, 43, pp. 2909–2915.
105–113. 关46兴 Slattery, J. C., 1967, “Flow of Viscoelastic Fluids Through Porous Media,”
关26兴 de Lemos, M. J. S., and Tofaneli, L. A., 2004, “Modeling of Double-Diffusive AIChE J., 13, pp. 1066–1071.
Turbulent Natural Convection in Porous Media,” Int. J. Heat Mass Transfer, 关47兴 Forchheimer, P., 1901, “Wasserbewegung durch Boden,” Z. Ver. Deutsch. Ing.,
47共19-20兲, pp. 4221–4231. 45, pp. 1782–1788.
关27兴 de Lemos, M. J. S., and Pedras, M. H. J., 2001, “Recent Mathematical Models 关48兴 Wakao, N., Kaguei, S., and Funazkri, T., 1979, “Effect of Fluid Dispersion
for Turbulent Flow for Saturated Rigid Porous Media,” J. Fluids Eng., 123共4兲, Coefficients on Particle-to-Fluid Heat Transfer Coefficients in Packed Bed,”
pp. 935–940. Chem. Eng. Sci., 34, pp. 325–336.
关28兴 Silva, R. A., and de Lemos, M. J. S., 2003, “Numerical Analysis of the Stress 关49兴 Patankar, S. V., 1980, Numerical Heat Transfer and Fluid Flow, Hemisphere,
Jump Interface Condition for Laminar Flow over a Porous Layer,” Numer. Washington, D.C.
Heat Transfer, Part A, 43共6兲, pp. 603–617. 关50兴 Zhukauskas, A., 1972, “Heat Transfer from Tubes in Cross Flow,” Adv. Heat
关29兴 Silva, R. A., and de Lemos, M. J. S., 2003, “Turbulent Flow in a Channel Transfer, 8, pp. 93–160.
Convective Heat Transfer in a
Rectangular Channel Filled With
Sintered Bronze Beads and
Periodically Spaced Heated
Blocks
This work numerically investigated the steady state fluid flow and heat transfer behaviors
associated with a sintered porous channel that contains periodically spaced heated
Sheng Chung Tzeng blocks. Some typical cases are experimentally examined in this study. The relevant varied
Associate Professor parameters were the average bead diameter 共d兲, the relative block height 共h / H兲, the
Department of Mechanical Engineering, relative block width 共w / H兲, the relative block spacing 共s / H兲, and the Reynolds number
Chienkuo Technology University,
(Re). The numerical results revealed a lack of global recirculation in regions between the
Changhua 500, Taiwan, R.O.C.
blocks, where the forced convective heat transfer was low, but the heat in those regions
e-mail: tsc@ctu.edu.tw; tsc33@ms32.hinet.net
was transferred to the metallic block by conduction through porous media, before being
dissipated into the fluid that passed over the zone above the heated block. Additionally,
the relevant parameters considerably affect the local Nusselt number distribution along
the periphery of the block surface. The average Nusselt number for each block decreased
along the direction of the flow until it reached its fully developed value. The Nusselt
number increased with h / H or Re in the fully developed region. The effect of h / H on the
fully developed Nusselt number became stronger as Re increased and w / H decreased.
The effects of s / H and d on the fully developed Nusselt number were insignificant over
the ranges of parameters considered herein (d = 0.7 and 1.16 mm, h / H = 0.12– 0.59,
w / H = 0.24– 0.47, s / H = 0.24– 0.7, and Re= 1019– 5059). Finally, this study summarized
the average Nusselt number for different configurations of the heated blocks with various
d, h / H, w / H, s / H, and Re. 关DOI: 10.1115/1.2175151兴
Keywords: sintered porous channel, periodically spaced heated blocks, convective heat
transfer
1 Introduction nels. Details of the flow and thermal fields for various Reynolds
numbers and Grashof numbers were provided. They reported the
The electronics industry is manufacturing denser and more
largest heat transfer in the first block, declining progressively
powerful products, and so requires better cooling technology.
downstream. They also demonstrated that the cooling efficiency of
Typically, the power density values at the chip level of plastic
the vertical channel was slightly higher than that of the horizontal
DIPs and PGA ceramic packages are around 10 and 25 W / cm2,
channel because of the buoyancy effect. Young and Vafai 关5,6兴
respectively 关1,2兴. The cooling technology by conventional air-
numerically studied in a comprehensive manner the fluid and ther-
handling methods is insufficient in most cases. Therefore, more mal transport in a two-dimensional channel that contained a
effective cooling is required. Investigating the convective cooling heated block or multiple heated blocks. The effects of the number
of heated blocks mounted on a channel wall is valuable because of blocks and their width, height, and spacing as well as the heat-
such blocks are used to simulate heated electronic components. ing method, solid thermal conductivity, and flow rate were con-
Some investigations of channels with heated block or filled with sidered. Their results indicated that smaller, more widely spaced
various porous materials were reported. blocks were associated with better heat transfer performance. Nar-
Kang et al. 关3兴 experimentally studied mixed convective heat row gaps between blocks were found to allow upstream thermal
transfer from a heated module on a horizontal plate. Their data transport, which, in some cases, actually heated the upstream
revealed that the heat transfer and fluid flow characteristics varied blocks.
strongly with the mixed convection parameter, Gr/ Re5/2. A tran- Hunt and Tien 关7兴 experimentally measured forced convection
sition range for their system in the external flow was observed to in channels filled with various porous foam materials. Forced con-
be Gr/ Re5/2 = 0.9– 1.1. The average Nusselt number was found to vection in horizontal square channels through packed spheres was
drop as Gr/ Re5/2 increased. Kang et al. also determined that the investigated by Chou et al. 关8兴. Hwang and Chao 关9兴 and Hwang
Nusselt number at the left face of the heated module was the et al. 关10兴 measured heat transfer through channels filled with a
largest of any of the three faces because of the impact of the porous media made of sintered beads, and numerically solved the
external flow. Kim et al. 关4兴 numerically examined mixed convec- two-equation model to make accurate predictions. Peterson and
tive flow around three blocks in both horizontal and vertical chan- Chang 关11兴 experimentally investigated two-phase dissipation in
porous channels of various sizes filled with sintered copper beads.
Calmidi and Mahajan 关12兴 experimentally and numerically exam-
NAL OFHEAT TRANSFER. Manuscript received May 16, 2005; final manuscript received ined the forced convection in highly porous aluminum foams,
October 31, 2005. Review conducted by Jose L. Lage. using air as the cooling fluid. Angirasa 关13,14兴 studied the forced
convective heat transfer in metallic fibrous materials with various
porosities and fiber thicknesses. Jiang et al. 关15,16兴 investigated
heat transfer by forced convection in sintered porous channels.
The studies cited herein provide insights into the transfer of heat
by porous metallic materials. The results of these investigations
reveal that a porous material can be used as an extremely effective
heat sink, possibly providing an effective means of cooling circuit
boards on which heat-generating electronic components are
mounted.
Recently, porous materials are employed to promote heat trans- Fig. 1 Schematic diagram of the porous channel with heated
fer in a channel with discrete heated sources. Rizk and Klein- electronic components
streuer 关17兴 analyzed laminar forced convection on discrete heated
blocks in a porous channel. They presented empirical correlations
between Reynolds number and heat transfer coefficient associated channel can be specified using the following volume-averaged
with the porous material and the solid blocks. By comparing the conservation equations of continuity, momentum and energy
traditional open channel with their results indicated that the heat 关22–24兴.
transfer enhancement of the porous channel was about 50%.
Hadim 关18兴 numerically studied forced convection in a partially ⳵u ⳵v
+ =0 共1兲
porous channel that contained discrete heat sources on the bottom ⳵x ⳵ y
冉冊冉冊
wall. He reported that heat transfer increased as the Darcy number
declined, especially at the leading edge of each heat source. ␳ f ⳵u ⳵u ⳵p ␮ ␳ f CF 冑 2 2 ␮ ⳵ 2u ⳵ 2u
2 u +v =− − u− 共 u + v 兲u + +
Hadim and Bethancourt 关19兴 extended the work of Hadim 关18兴. ␧ ⳵x ⳵y ⳵x K 冑K ␧ ⳵x2 ⳵ y 2
They numerically examined the effects of the spacing between
heat sources and the width of the heat source on heat transfer. 共2兲
冉冊冉冊
Their results indicated that increasing the spacing between heat
sources insignificantly increased the heat transfer. As the width of ␳ f ⳵v ⳵v ⳵p ␮ ␳ f CF 冑 2 2 ␮ ⳵ 2v ⳵ 2v
2 u +v =− − v− 共 u + v 兲v + +
the heat source decreased, the heat transfer enhancement was only ␧ ⳵x ⳵y ⳵y K 冑K ␧ ⳵x2 ⳵ y 2
moderate. Oild-Amer et al. 关20兴 analyzed the effect of porous
inserts between the heated blocks. They found that porous inserts 共3兲
increased the heat transfer rate on the vertical sides of the blocks.
The heat transfer from multiple heated blocks with porous covers
in a channel was studied by Huang et al. 关21兴. The global recir-
冉
共 ␳ C p兲 f u
⳵T f
⳵x
+v
⳵T f
⳵y
冊= h fsa fs共Ts − T f 兲 + 共k*f + kd兲
⳵x2
冉
⳵ 2T f ⳵ 2T f
+ 2
⳵y
冊
culation caused by the blocks substantially increased the heat 共4兲
transfer rate on both the top and the downstream faces of the
second and the subsequent blocks.
This work was motivated by the fact that few numerical studies
have addressed the effects of the size and spacing of the heated
0 = h fsa fs共T f − Ts兲 + k*s 冉 ⳵ 2T s ⳵ 2T s
⳵x2
+ 2
⳵y
冊共5兲
blocks on the fluid flow and the convective heat transfer. The lack where K is the permeability, CF is the inertial coefficient, ␧ is the
of experimental data in this area has required the completion of porosity, k*s is the effective conductivity of the solid and depends
experimental measurements for comparison with the numerical on both the geometry of the porous medium and the conductivity
results. of the solid, k*f is the effective conductivity of the fluid and is
This work numerically investigated the fluid flow behavior and given simply as ␧k f , kd is the thermal dispersion conductivity, a fs
the heat transfer mechanisms associated with a porous channel. is the surface area of the fluid-solid interface per unit bulk volume
The channel filled with a porous media made of sintered bronze of the packed bed, and h fs is the heat transfer coefficient between
beads was discretely heated through periodically spaced metallic the fluid stream and the solid matrix. Notably, the relevant empiri-
blocks and cooled by air flow. Some typical cases were experi- cal coefficients in the present numerical model, such as K, CF, ks*,
mentally examined. The local temperatures and the heat transfer kd, and h fs, generally do not have universal values. This is because
characteristics of each heated block were studied. The varied pa- that these empirical coefficients mainly depend on the geometry
rameters were the average bead diameter 共d兲, the relative block of the porous medium, the thermal properties of the solid and the
height 共h / H兲, the relative block width 共w / H兲, the relative block fluid, and the flow rate. Additionally, the sintered process changes
spacing 共s / H兲, and the Reynolds number 共Re兲. The findings can the inner structure of the sintered porous material, resulting in
be used to improve the cooling performance of heated blocks, different values of K, CF, k*s , kd, and h fs from those for the non-
where the blocks are used to simulate heated electronic compo- sintered porous medium. Moreover, all these empirical coeffi-
nents, by sintered metallic porous media. cients are difficultly measured. This work contributes to build a
numerical model for sintered porous channels by measuring the K,
CF, and k*s values and finding the suitable empirical formula of kd
and h fs in the open literatures. The section of the test apparatus
2 Numerical Analysis
and instrumentation addresses the values of K, CF, and k*s consid-
2.1 Basic Assumptions and Governing Equations. Figure 1 ered herein. The thermal properties, kd, a fs, and h fs, of the sintered
presents the configuration of the problem studied herein. It in- bronze beads are taken from Hwang and Chao 关9兴, Hwang et al.
volves air flow through a parallel plate channel with periodically 关10兴, Chen and Hsu 关25兴, Chen et al. 关26兴, Kar and Dybbs 关27兴,
spaced blocks on the bottom plate. The blocks are heated under and Gamson et al. 关28兴
isoflux conditions and the upper and bottom plates are assumed to
be adiabatic. The air enters the channel at uniform temperature Ti kd = lDt Pr Red k f 共6兲
and uniform velocity ui. The other assumptions and simplifica-
l = 兵1 + 共3d p/H*兲关exp共− H*/3d p兲 − 1兴其共7兲
tions are as follows: 共1兲 the porous medium is homogenous and
isotropic; 共2兲 the fluid flow is steady state, laminar and incom- H* = H for the regions between blocks
pressible; 共3兲 the thermophysical properties of the fluid and po-
rous media do not depend on temperature. Hence, the flow in the = 共H − h兲 for the regions above blocks 共8兲
h fs = 0.004 冉冊冉冊
dv
dp
0.35
kf
dp
Pr0.33 Re1.35
d for Red 艋 75
= 1.064 冉冊
kf
dp
Pr0.33 Re0.59
d for Red
艌 350 for the porous medium

= 0 for the blocks 共9兲
a fs = 20.346共1 − ␧兲␧2/d p for the porous medium

= 0 for the blocks 共10兲
where l represents the wall function for transverse thermal disper- Fig. 2 Schematic diagram of numerical domain with boundary
conditions
sion, Dt is the thermal dispersion constant that equals 0.375, d p is
the average bead diameter; dv 共=4␧ / a fs兲 is the average void diam-
eter, the bead Reynolds number Red is based on the bead diameter
and the local velocity, and h fs is evaluated by interpolation at
Red = 75– 350. Additionally, according to the results of Fu and very large 共kb is the thermal conductivity of the block and k f is the
Huang 关29兴, the local Nusselt number distribution of the random thermal conductivity of the fluid兲. The value of kb / k f in the
porosity model are more similar to those of the constant porosity present work is very high because the copper block is used as a
model than those of the variable porosity model. Therefore, a heat spreader. Therefore, the isoflux condition is applied to the
constant porosity value was used in this numerical simulation. bottom surfaces of the blocks in this work. The temperature gra-
Equations 共1兲–共5兲 can be nondimensionalized by introducing dients of the solid phase and the fluid phase are zero on the adia-
the following variables: batic wall. The fluid is assumed to enter the inlet at a uniform
temperature, Ti. The fluid phase at the exit is assumed to meet the
x y u v ␮/␳ f K zero diffusion condition. The following characteristics of the solid
X= , Y= , U= , V= , Pr = , Da = , phase, reported by Hwang et al. 关10兴, are assumed. At the inter-
H H ui ui k f /共␳C p兲 f H2
faces between the blocks and the porous media, the nonslip con-
␳ f u iH 冑u2 + v2 T − Ti h fsa fsH2 dition, and the continuity of temperature and heat flux are as-
Re = , UM = , ␪= , Nu fs = sumed to apply
␮ ui qwH/k f kf
The vorticity-transport equation is determined by replacing the

共11兲 ⳵␪s
⳵X
= 1.29lDt 冉冊冉冊冉冊
dv
dp
0.5
kf
k*s
H
dv d ␪s
Pr1.4 Re1.5 at X = 0 共17兲
velocity components with the vorticity 共␻兲 and the stream func-
tion 共⌿兲
⳵U ⳵V ⳵⌿ ⳵⌿
⳵␪s
⳵X
= − 0.2776lDt *冉冊冉冊
kf
ks
H
dp d 共␪s − ␪ f 兲
Pr1.4 Re1.5 at X = L/H
␻=− + , U= , V=− 共12兲共18兲
⳵Y ⳵X ⳵Y ⳵X
The dimensionless governing equations as follows:
⳵␪s ⳵␪s ⳵␪ f
⳵ 2⌿ ⳵ 2⌿ U = V = 0, ␪ f = ␪ s, kb = k*s + k*f
−␻= + 共13兲 ⳵N ⳵N ⳵N
⳵X2 ⳵Y 2
at block/porous interfaces 共19兲
U
⳵␻
⳵X
+V
⳵␻
⳵Y
=−
␧2
Re · Da
␻−
␧ 2C FU M
冑Da
␻+
␧ 2C F ⳵ U M
冑 Da
U
⳵Y
−V
⳵U M
⳵X
冉冊 where l, Dt, d p, dv, and Red were defined according to Eqs.
共6兲–共10兲; kb is the thermal conductivity of the block, and N is the
冉冊
dimensionless coordinate normal to the block surface. The validity
␧ ⳵ 2␻ ⳵ 2␻ and accuracy of the exit boundary conditions used herein was
+ + 共14兲
Re ⳵X2 ⳵Y 2 verified by comparing velocity and temperature profiles obtained
冉冊
for Lo / H = 0.93 with those obtained for an extended computational
⳵␪ f ⳵␪ f Nu fs 1 k*f + kd ⳵2␪ f ⳵2␪ f domain Lo / H = 6. The results had a good agreement between the
U +V = 共␪s − ␪ f 兲 + + two cases.
⳵X ⳵Y Re Pr Re Pr k f ⳵X2 ⳵Y 2
共15兲 2.3 Numerical Procedure. The governing equations were
discretized using the power-law scheme proposed by Patankar
0 = Nu fs共␪ f − ␪s兲 + 冉
k*s ⳵2␪s ⳵2␪s
+
k f ⳵X2 ⳵Y 2
冊共16兲
关31兴. The stream function equation and the vorticity transport
equation were solved simultaneously and iteratively 关32兴. The ve-
locity components U and V were obtained from the stream func-
In the blocks, Da= ⌿ = 0, yielding ␻ = 0 and then U = V = 0. Also, tion. When the velocity components had been determined, the
Nu fs = k*f = kd = 0. energy equation was solved to yield the temperature distribution.
All discretized algebraic equations were solved using the strong
2.2 Boundary Conditions. Figure 2 schematically depicts the implicit solver 共SIS兲 algorithm developed by Lee 关33兴. The con-
numerical domain and the boundary conditions. Roache 关30兴 elu-
cidated the boundary conditions of the vorticity-transport equa- vergence criterion was Max兩共F共n兲 − F共n−1兲兲 / 共F共n兲共n兲
max − Fmin兲兩艋 1
tion. Additionally, when the blocks are employed to simulate the ⫻ 10−5, where F was ⌿, ␻, ␪ f or ␪s. The local Nusselt number
electronic components, two heat input approaches are typically 共Nu兲 on the channel wall and the local Nusselt number 共Nup兲
utilized. One is isoflux at the block base and the other is uniform along the periphery of the block surface are defined as
冉冊冒
volumetric energy generation in the block. However, Young and
Vafai 关6兴 demonstrated that differences between the isotherms in q wH k*s ⳵␪s k*f ⳵␪ f
Nu = =− + ␪w 共20兲
the two heating cases are very small when the value of kb / k f is k f 共Tw − Ti兲 k f ⳵Y k f ⳵Y
Table 1 Results from grid dependency simulations Table 2 Physical properties of sintered bronze beads
Nu for the Nu for the Nu for the Samples 1 2

Grid size first block third block fifth block
0.70 1.16
average bead diameter, d 共mm兲共0.62–0.80兲a 共1.00–1.42兲a
201⫻ 81 1771.3 889.6 681.5
151⫻ 61 1771.2 889.6 681.5 porosity, ␧ 0.37 0.39
151⫻ 41 1764.6 887.1 679.3 thermal conductivity, ks* 共W/m °C兲 10.73 10.45
101⫻ 21 1752.3 880.7 675.4 permeability, K ⫻ 1010 共m2兲 2.83 6.70
inertial coefficient, CF 0.239 0.171
a
The diameter ranges of beads
Nup =
qw,pH
k f 共T p − Ti兲
冉
=−
k*s ⳵␪s,p
k f ⳵N
+
k*f ⳵␪ f,p
k f ⳵N
冊冒 ␪p 共21兲
test section was made of 40-mm-thick Bakelite. The dimensions
where qw is the heat flux of the heated wall, qw,p is the heat flux of the channel section were 52 mm⫻ 8.5 mm. The channel was
along the periphery of the block surface, ␪w is the dimensionless filled with test specimens of size 52 mm⫻ 8.5 mm⫻ 52 mm. The
temperature on the channel wall conjugated with the blocks or the test specimens were made of sintered bronze beads with periodi-
porous medium, ␪ p is the dimensionless block surface tempera- cally inserted copper blocks. The separation between each pair of
ture, and N is the dimensionless coordinate normal to the block copper blocks of size 52 mm⫻ 3 mm⫻ 4 mm was 4 mm. A film
surface. The average Nusselt number 共Nu兲 for each block is the heater was fixed on the bottom surface of the copper blocks.
integral average value of Nu 共defined in Eq. 共20兲兲. It is obtained Available with the thermal grease as its adhesion agent, the top
from the block length. plate of the film heater was provided with 52 mm⫻ 8.5 mm
2.4 Grid Dependency and Code Validation. A nonuniform ⫻ 52 mm Teflon material for insulation so as to prevent the stain-
rectangular grid system was employed. The typical case with less film heater from contacting the sintered porous channel,
sample-1, Li / H = 0.93, Lo / H = 0.93, n = 5, h / H = 0.36, s / H = 0.47, which would affect the thermocouple measurement signal. The
w / H = 0.47, Re= 3057 was used for grid tests. Table 1 lists the wall heat fluxes supplied to the film heater from the electronic
results from grid dependency simulations. Based on the grid de- power were 1.5– 4.3 W / cm2 for 50– 250 l / min flow rate.
pendence study, the grid parameters 共151⫻ 61兲 listed in the sec- The average diameter of the sintered bronze beads in the chan-
ond row were used for all computations. nel was 0.70 mm and 1.16 mm. Table 2 provides the relevant
In order to validate the code, numerical results were compared parameters associated with the porous properties, including the
with the relevant cases reported by Hadim 关18兴. Hadim 关18兴 stud- permeability 共K兲, the inertial coefficient 共CF兲, the effective solid
ied the problem that in a porous channel with four strip heat conductivity 共k*s 兲, and others. The K and CF values were deter-
sources for ␧ = 0.97, CF = 0.1, Li / H = 1, Lo / H = 7, n = 4, h / H = 0, mined by the method reported by Hunt and Tien 关7兴. The value of
s / H = 1, w / H = 1, Pr= 10, and 共k*s + k*f 兲 / k f = 1 at Da= 1 ⫻ 10−6 and k*s was measured by performing several experiments of one-
1 ⫻ 10−3, as well as Re= 20 and 1000. Since the numerical study dimensional heat conduction 共the test method can be found in Ref.
performed by Hadim 关18兴 was achieved by the one-equation 关34兴兲. The total error in the measurement of the effective solid
model, a sufficiently large value was employed for Nu fs in the thermal conductivity 共k*s 兲 was ±3.7%. Notably, the effective solid
present model. The results, as shown in Fig. 5, agree to better than
3% with data reported by Hadim 关18兴.
3 Experiments
3.1 Test Apparatus and Instrumentation. The experimental
setup displayed in Fig. 3 was divided into four main parts: 共1兲 air
supply system; 共2兲 test section; 共3兲 test specimens; and 共4兲 data
acquisition system. The air compressor 共MITSUISEIKI 10 HP兲
blew air into the air tank. Then, the air flowed through a filter to
remove the oil, water, and particles. Finally, the air entered the test
section after it passed through the straightener made of an un-
heated porous specimen. The air flow rate was controlled by an
electric digital flowmeter with the range of 0 – 1000 l / min. The
Fig. 4 Physical model of the test section and positions of ther-

Fig. 3 Experimental apparatus mocouples from bottom view
tures were greater than 36° C at the fifth heated block. The system
was assumed to be in a steady state when the temperature varied
by no more than 0.2° C in 25 min.
3.2 Uncertainty Analysis. The maximum error in the flow
rate was about ±5.0% at 50 l / min. The real temperature on the
bottom surface of the test specimen should be estimated by Tw
− qw␦ / kg, where ␦ was the distance between the temperature mea-
surement point to the bottom surface of the test specimen, and kg
was the thermal conductivity of the combination of the thermal
grease and the Teflon material. Since the ␦ was hardly estimated,
the Tw was used to replace the real temperature on the bottom
surface of the test specimen. Therefore, the errors in the tempera-
ture measurements were resulted from the inaccuracies in the re-
corder readings 共about ±0.2° C兲, and the contact resistance be-
tween the film heater and the bottom surface of the test specimen
共about ±2 ° C兲. The maximum errors in determining the heat flux
共qw兲, the channel height 共H兲, and fluid conductivity 共k f 兲 were
estimated within ±3, ±1, and ±1%, respectively. Uncertainties in
parameters were estimated by using the root-sum-square method
of Kline and McClintock 关35兴 and Moffat 关36兴. The experimental
data obtained herein revealed that the uncertainties in the Rey-
nolds number and the Nusselt number 共as shown in Eq. 共20兲兲 were
Fig. 5 Comparison of Nu on each heat source with that found ±5.6 and ±22.0%, respectively.
in Ref. †18‡

conductivity determined using ks共1 − ␧兲, which is commonly used The study was conducted to discuss the convective heat transfer
in a numerical study of heat transfer in porous media, greatly in a rectangular channel filled with sintered bronze beads and
exceeds the experimental value. Jiang et al. 关15兴 found the same periodically spaced heated blocks. The relevant varied parameters
result. were h / H, s / H, w / H, Re, as listed in Table 3. The range of Rey-
Eleven OMEGA T-TT-30 thermocouples for measuring the wall nolds numbers was Re= 1019– 5095.
temperature were embedded on the interior wall surface of the
flow passage, as presented in Fig. 4. Another three thermocouples 4.1 Experimental Validation. Figures 6共a兲 and 6共b兲 compare
were employed to monitor the ambient temperature, the air tem- the numerical predictions with the experimental result. Figure 6共a兲
perature at the channel inlet, and the air temperature at the channel shows the nondimensional wall temperature in the flow direction
outlet. A total of 14 thermocouples were connected to the data at various Reynolds numbers. The data reveal that the dimension-
recorder 共YOKOGAWA DA100兲, which will transform the poten- less wall temperature fell as the Reynolds number increased. This
tial difference signal into a temperature value. All the temperature phenomenon is attributed to the effect of forced convection. The
data printed out by the data recorder are then inputted into the measured temperature difference 共Tw − Ti兲, heat flux 共qw兲, and the
computer and converted into data for further analysis. The inlet air uncertainties in the nondimensional temperature are listed in Table
temperature was about 23° C. The temperature difference of air 4. The comparison results indicated that most of the numerical
between inlet and outlet was about 10– 16° C. The wall tempera- predictions were consistent with the measured temperature of the
Table 3 Relevant varied parameters for sample, h / H, s / H, w / H, Re
Li / H = 0.93, Lo / H = 0.93, n = 5, Re= 1019– 5095

h / H = 0.12 h / H = 0.36 h / H = 0.59
Sample-1, w / H = 0.24,
共L / H = 3.962兲
s / H = 0.235 Sample-1, w / H = 0.47, Sample-1, w / H = 0.47, Sample-1, w / H = 0.47,
共L / H = 5.136兲共L / H = 5.136兲共L / H = 5.136兲
Sample-2, w / H = 0.47,
共L / H = 5.136兲
Sample-1, w / H = 0.24, Sample-1, w / H = 0.24, Sample-1, w / H = 0.24,

共L / H = 4.901兲共L / H = 4.901兲共L / H = 4.901兲
共L / H = 6.075兲共L / H = 6.075兲共L / H = 6.075兲
Sample-2, w / H = 0.47, Sample-2, w / H = 0.47, Sample-2, w / H = 0.47,
共L / H = 6.075兲共L / H = 6.075兲共L / H = 6.075兲
Sample-1, w / H = 0.24,
共L / H = 5.840兲
共L / H = 7.014兲共L / H = 7.014兲共L / H = 7.014兲
Sample-2, w / H = 0.47,
共L / H = 7.014兲
Fig. 7 Streamlines for d = 0.7 mm, h / H = 0.36, w / H = 0.47, s / H
= 0.47, and Re= 3057
at the walls were also consistent with the experimental data 共the
average deviation between numerical predictions and experimen-
tal data are less than 16%兲, indicating that the numerical model
used herein accurately simulated the flow and thermal behavior in
sintered porous channels that contained multiple heated blocks.
4.2 Flow Behavior and Heat Transfer Mechanism. Insert-
ing metallic blocks into the porous media increased the complex-
ity of the overall fluid flow behavior and the heat transfer mecha-
nism. This section presents numerical results including
streamlines, temperature contours, and velocity vectors for typical
cases, to elucidate fluid flow behavior and the heat transfer
mechanism.
Figure 7 depicts the streamlines in a typical case, with d
= 0.70 mm, h / H = 0.36, w / H = 0.47, s / H = 0.47, and Re= 3057.
Unlike when many obstacles are present in an empty channel 关6兴
or in a channel partially filled with porous material 关18,19兴, no
global recirculation occurred between blocks in a sintered porous
channel. All numerical tests revealed this phenomenon, because
the sintered bronze beads with low permeability prevented global
Fig. 6 Comparison of numerical predictions with experimental

values „a… wall temperatures; „b… Nusselt numbers
wall. Figure 6共b兲 displays the Nusselt numbers for each heated
block. Experimental results reveal that the Nusselt numbers at the
walls of the upstream blocks exceeded those of the downstream
blocks. Furthermore, the Nusselt numbers at the walls of the
downstream blocks approached constants at a specified Reynolds Fig. 8 Velocity vector field for d = 0.7 mm, h / H = 0.36, w / H
number. The numerically predicted Nusselt number distributions = 0.47, s / H = 0.47, Re= 3057, and around the second block
Table 4 Measured temperature difference „Tw − Ti… „°C… for d = 0.70 mm
Average
uncertainty
qw共⫻10−4兲 in ␪w
Re 共W / m2兲 x / H= 0.47 1.17 1.64 2.11 2.58 3.05 3.52 3.99 4.46 4.93 5.63 ⫾%a
1019 1.5 5.6 8.3 11.0 12.1 14.4 15.6 16.6 17.0 17.5 17.9 16.4 16.2
2038 3.2 3.6 8.2 12.5 13.4 16.9 18.5 20.0 20.6 21.4 21.8 18.6 14.4
3057 3.8 1.5 6.8 10.5 10.4 14.1 15.0 16.6 16.5 17.6 17.5 13.9 17.6
4076 4.3 0.6 6.3 10.0 9.4 13.3 13.8 15.4 14.9 16.1 15.6 11.3 19.3
5095 4.3 0.1 5.6 9.0 8.4 11.9 12.3 13.5 13.0 13.8 13.4 8.7 22.0
a
Except the data at x / H = 0.47
Fig. 10 Fluid isotherms for d = 0.7 mm, h / H = 0.36, w / H = 0.47,
Fig. 9 Solid isotherms for d = 0.7 mm, h / H = 0.36, w / H = 0.47, s / H = 0.47, and Re= 3057 „Note: the isothermal in the block re-
s / H = 0.47, and Re= 3057. „a… Solid temperature for all computa- gion is the block temperature…. „a… Fluid temperature for all
tional domain. „b… Solid temperature around the second block. computational domain. „b… Fluid temperature around the sec-
„c… Solid temperature around the fifth block. ond block. „c… Fluid temperature around the fifth block.
recirculation. It matches the results of earlier works 关18,19兴, and Re= 3057. The temperature distributions around the second
which indicated that the global recirculation did not extend to the block and the fifth block were plotted in detail to elucidate the
flow in porous material but was confined to the open region of the heat transfer mechanism. Figure 9共a兲 demonstrates that the solid
channel. Figure 8 plots the velocity vectors for a selected param- temperature distributions around the first four blocks were similar
eter. The velocity vector between the second block and the third to each other and that around the fifth block differed from the
block was examined. This map also shows an absence of global others. The contours of solid temperature in Fig. 9共b兲 indicate that
recirculation, which greatly improves cooling. According to Mar- the metallic block absorbed heat from the heated bottom surface
tin et al. 关21兴, who studied the application of porous inserts to via two paths. The first path is direct conduction from the heated
enhance heat transfer in a recirculating flow over a backward- bottom surface; the other is indirect conduction through the po-
facing step, porous inserts not only prevent the formation of a rous media in the bottom portion of the right face of the block.
global recirculation zone but also eliminate the locally minimal Therefore, although the forced convective heat transfer between
Nusselt number near the backward-face. This characteristic helps the blocks was low, the heat in those regions was transferred to the
to reduce the locally maximal wall temperature. Additionally, the metallic block and then dissipated into the fluid that flowed over
quantity of air that entered the bottom portions between the blocks the zone above the block. However, in Figs. 9共c兲 and 10共c兲, the
was lower than the average flow rate and the flow velocity in the temperature contours reveal that the final block did not absorb
zone above the block was higher than the average velocity. These heat from the porous media and none of the blocks took heat from
conditions implied that more heat was dissipated from the top face the fluid. It needs to note that the isothermal in the block region
of the block than from other faces of the block. Figures 9 and 10 shown in Fig. 10 is the block temperature. The purpose of plotting
present the contours of the solid and fluid temperatures in a typi- block isotherms in Fig. 10 is to present the temperature gradient
cal case, with d = 0.70 mm, h / H = 0.36, w / H = 0.47, s / H = 0.47, and the heat transfer direction.
Fig. 11 Effects of h / H and Re on local Nusselt number distribution along the periphery of the block surface
4.3 Effects of Bead Diameter, Block Height, Length, Spac- = 0.24– 0.7兲. Nup at the left and top faces were lower in the down-
ing, and Flow Rate on Thermal Characteristics. This section stream blocks than in the upstream blocks. On the left faces of the
systematically addresses the effects of the parameters of interest, blocks, the Nup has maximal value at the upper corners because of
including block height, length, spacing, and flow rate, on the heat fluid acceleration. Nup was largest at the left corner of the top
convection characteristics of the system, which include the Nus- faces declining rapidly to a local minimum value. In cases with
selt number distribution along the periphery of the heated block small h / H, large Re and large s / H, Nup increased slightly near the
surface and the average Nusselt numbers for heated blocks. right corner, because the blowing effect causes more fluid to turn
4.3.1 Nusselt Number Distribution Along the Periphery of the around the right corner of the top face. The Nup at the right faces
Heated Block Surface. This section presents the local Nusselt of the blocks varied differently. At the right faces, Nup was typi-
number 共Nup兲 distributions along the periphery of the block sur- cally higher at the downstream blocks than at the upstream blocks.
face to clarify the differences among the heat transfers at all of the However, the Nup at the right faces of the upstream blocks in-
blocks in the porous channel. Figure 11 plots the effects of h / H creased more as h / H fell; Re increased or s / H increased, because
and Re on Nup. The typical cases involved d = 0.7 mm, w / H as h / H falls, Re increases or s / H increases, and more cold flow
= 0.47, s / H = 0.47, various h / H values 共h / H = 0.12– 0.59兲, and enters the regions among the heated blocks. The first block had
various Re values 共Re= 1019– 5095兲. Figure 12 shows the effects much larger Nup values along its left face than did the other
of d, w / H, and s / H on Nup. The typical cases involved h / H blocks, because of the uniform entry flow. Therefore, the large
= 0.36, Re= 5095, various d values 共d = 0.7 and 1.16 mm兲, various Nup along the left face of the first block caused more heat to be
w / H values 共w / H = 0.24 and 0.47兲, and various s / H values 共s / H transferred across its right face. Notably, the negative Nup along
Fig. 12 Effects of d, w / H, s / H on local Nusselt number distribution along the periphery of the block surface
the right faces of the blocks gradually became positive as h / H the fully developed region. The effect of h / H on the fully devel-
fell; Re increased or s / H increased, especially in the upper por- oped Nusselt number was stronger as Re increased and w / H de-
tions of the right faces of the blocks. creased. Notably, a comparison of the Nu* values at low w / H
4.3.2 Average Nusselt Numbers of Heated Blocks. Nu* is 共w / H = 0.24兲 and high w / H 共w / H = 0.47兲 reveals that in the
taken as the average Nu for the bottom of the heated blocks to former case, Nu* is larger when h / H is higher 共h / H = 0.59兲, and in
determine the average heat dissipation of the heated blocks for the latter case, Nu* is larger when h / H is lower 共h / H = 0.36 and
various d, h / H, w / H, s / H, and Re 0.12兲. In the fully developed region, the effects of s / H and d on
the average Nusselt number were not so evident over the ranges of
w
Nu* = Nu 共22兲 parameters considered herein. Some articles 关4,37–40兴 have indi-
H cated that the fully developed Nu* in a nonporous channel with
where Nu is the integral average value of the Nu 共defined in Eq. periodically spaced heated blocks happened from the third row to
共20兲兲 determined from the block length. Figure 13 plots the aver- the tenth row of the block array. This work arranged five heated
age Nusselt number 共Nu*兲 of heated blocks with various d, h / H, blocks in the sintered porous channel. Using the deviation be-
w / H, s / H, and Re. The data reveals that for a specific h / H and tween two neighbor blocks as the reference of the fully developed
Re, Nu* is close to a particular value downstream. Restated, the Nusselt number, the difference in average Nusselt number de-
array of heated blocks approaches the expected fully developed creased downstream and generally less than 10% was found be-
values. The average Nusselt number increased with h / H or Re in tween the fourth and fifth blocks for all numerical cases. There-
Fig. 13 Average Nusselt numbers on different blocks with various d, h / H, w / H, s / H, and Re
fore, Fig. 13 demonstrates that the Nu* of the fifth block gradually blocks. For the top faces, Nup was highest at the left corner,
reached the fully developed value. For ranges of parameters con- decreasing rapidly toward a local minimum value. In cases
sidered herein 共d = 0.7 and 1.16 mm, h / H = 0.12– 0.59, w / H with low h / H, high Re, and high s / H values, Nup increased
= 0.24– 0.47, s / H = 0.24– 0.7, and Re= 1019– 5059兲, the Nu* of slightly near the right corner. Nup on the right faces was
the fifth block is between 133 and 573. typically higher at the downstream blocks. However, Nup on
the right faces of the upstream blocks increased more
5 Conclusions quickly as h / H fell, Re increased or s / H increased, than did
the corresponding value of other blocks.
This work numerically investigated the fluid flow behavior and 5. The average Nusselt number values 共Nu*兲 of upstream
the heat transfer mechanism in a porous channel that contains
blocks exceeded those of downstream blocks and ap-
periodically spaced heated metallic blocks. The results were ex-
proached a constant value downstream for specific h / H and
perimentally validated for some typical cases. The porous medium
was sintered bronze material. The varied parameters were the av- Re. The Nu* increased with h / H or Re in the fully devel-
erage bead diameter 共d兲, the relative block height 共h / H兲, the rela- oped region. The effect of h / H on the fully developed Nus-
selt number became stronger as Re increased and w / H de-
tive block width 共w / H兲, the relative block spacing 共s / H兲, and the
creased. However, the effects of s / H and d on the fully
Reynolds number 共Re兲. The following conclusions were drawn: developed Nusselt number were not very obvious over the
1. The numerical predictions were consistent with the experi- ranges of parameters considered herein. The Nu* of the fifth
mental results, indicating that the numerical model em- block gradually reached the fully developed value, being
ployed herein can accurately simulate the flow and thermal between 133 and 573.
behavior in a sintered porous channel that contains multiple
heated blocks. Acknowledgment
2. No global recirculation occurred among the blocks in the The authors would like to thank the National Science Council
sintered porous channel. The amount of air that entered the of the Republic of China for financially supporting this research
bottom regions between the blocks was lower than the aver- under Contract No. NSC 93-2622-E-270-003-CC3.
age flow rate and the flow velocity in the zone above the
block was higher than the average velocity. These conditions
reveal that more heat was dissipated from the top face of the Nomenclature
block than from other faces of the block. a fs ⫽ surface area of the fluid-solid interface per unit
3. Although less heat was transferred by forced convection in volume, m−1
the regions among the blocks than elsewhere, the heat in CF ⫽ inertial coefficient
those regions was transferred to the metallic block by con- C p ⫽ specific heat of fluid at constant pressure,
duction through the porous media, before being dissipated J/kg/°C
into the fluid that passed over the zone above the block. d ⫽ average bead diameter, m
However, the final block did not absorb heat from the porous dv ⫽ average void diameter= 4␧ / a fs, m
media and no block absorbed heat from the fluid. Da ⫽ Darcy number, K / H2
4. On the left and top faces, Nup was lower at the downstream Dt ⫽ thermal dispersion constant
H ⫽ height of the channel, m tions,” Int. J. Heat Mass Transfer, 35, pp. 2941–2952.
关5兴 Young, T. J., and Vafai, K., 1998, “Convective Cooling of a Heated Obstacle in
h ⫽ height of the block, m
a Channel,” Int. J. Heat Mass Transfer, 41, pp. 3131–3148.
h fs ⫽ heat transfer coefficient between the fluid 关6兴 Young, T. J., and Vafai, K., 1998, “Convective Flow and Heat Transfer in a
stream and the solid matrix, W / m2 / ° C Channel Containing Multiple Heated Obstacles,” Int. J. Heat Mass Transfer,
K ⫽ Permeability, m2 41, pp. 3279–3298.
k ⫽ thermal conductivity, W/m/°C 关7兴 Hunt, M. L., and Tien, C. L., 1988, “Effects of Thermal Dispersion on Forced
l ⫽ wall function for transverse thermal dispersion Convection in Fibrous Media,” Int. J. Heat Mass Transfer, 31, pp. 301–309.
L ⫽ length of the channel, m 关8兴 Chou, F. C., Lien, W. Y., and Lin, S. H., 1992, “Analysis and Experiment of
Li ⫽ length of channel upstream from the blocks, m Non-Darcian Convection in Horizontal Square Packed-Sphere Channels—1.
Forced Convection,” Int. J. Heat Mass Transfer, 35, pp. 195–205.
Lo ⫽ length of channel downstream from the blocks, 关9兴 Hwang, G. J., and Chao, C. H., 1994, “Heat Transfer Measurement and Analy-
m sis for Sintered Porous Channels,” ASME J. Heat Transfer, 116, pp. 456–464.
n ⫽ number of blocks 关10兴 Hwang, G. J., Wu, C. C., and Chao, C. H., 1995, “Investigation of Non-
Nu ⫽ Nusselt number Darcian Forced Convection in an Asymmetrically Heated Sintered Porous
p ⫽ pressure, Pa Channel,” ASME J. Heat Transfer, 117, pp. 725–732.
Pr ⫽ Prandtl number 关11兴 Peterson, G. P., and Chang, C. S., 1998, “Two-phase Heat Dissipation Utilizing
qw ⫽ heat flux of the heated wall, W / m2 Porous-Channels of High-Conductivity Material,” ASME J. Heat Transfer,
120, pp. 243–252.
qw,p ⫽ heat flux along the periphery of the block sur-
关12兴 Calmidi, V. V., and Mahajan, R. L., 2000, “Forced Convection in High Poros-
face, W / m2 ity Metal Foams,” ASME J. Heat Transfer, 122, pp. 557–565.
Re ⫽ Reynolds number based on the height of chan- 关13兴 Angirasa, D., 2002, “Forced Convective Heat Transfer in Metallic Fibrous
nel, ␳ f uiH / ␮ Materials,” ASME J. Heat Transfer, 124, pp. 739–745.
Red ⫽ Reynolds number based on the average bead 关14兴 Angirasa, D., 2002, “Experimental Investigation of Forced Convection Heat
diameter, ␳ f 冑u2 + v2d / ␮ Transfer Augmentation With Metallic Fibrous Materials,” Int. J. Heat Mass
Transfer, 45, pp. 919–922.
s ⫽ spacing of the block, m
关15兴 Jiang, P. X., Li, M., Lu, T. J., Yu, L., and Ren, Z. P., 2004, “Experimental
T ⫽ temperature, °C Research on Convection Heat Transfer in Sintered Porous Plate Channels,” Int.
u ⫽ fluid flow velocity in the streamwise direction, J. Heat Mass Transfer, 47, pp. 2085–2096.
m/s 关16兴 Jiang, P. X., Li, M., Ma, Y. C., and Ren, Z. P., 2004, “Boundary Conditions
U ⫽ dimensionless fluid flow velocity in the stream- and Wall Effect for Forced Convection Heat Transfer in Sintered Porous Plate
wise direction Channels,” Int. J. Heat Mass Transfer, 47, pp. 2073–2083.
U M ⫽ dimensionless magnitude of the velocity, 关17兴 Rizk, T. A., and Kleinstreuer, C., 1991, “Forced Convective Cooling of a
冑u2 + v2 / ui Linear Array of Blocks in Open and Porous Matrix Channels,” Heat Transfer
Eng., 12, pp. 40–47.
v ⫽ fluid flow velocity in the crosswise direction, 关18兴 Hadim, A., 1994, “Forced Convection in Porous Channel With Localized Heat
m/s Sources,” ASME J. Heat Transfer, 116, pp. 465–472.
V ⫽ dimensionless fluid flow velocity in the cross- 关19兴 Hadim, H. A., and Bethancourt, A., 1995, “Numerical Study of Forced Con-
wise direction vection in a Partially Porous Channel With Discrete Heat Sources,” ASME J.
w ⫽ length of the block, m Electron. Packag., 117, pp. 46–51.
x , y ⫽ Cartesian coordinates, m 关20兴 Ould-Amer, Y., Chikh, S., Bouhadef, K., and Lauriat, G., 1998, “Forced Con-
vection Cooling Enhancement by Use of Porous Material,” Int. J. Heat Fluid
X , Y ⫽ dimensionless Cartesian coordinates
Flow, 19, pp. 251–258.
Greek Symbols 关21兴 Huang, P. C., Yang, C. F., Hwang, J. J., and Chiu, M. T., 2005, “Enhancement
␧ ⫽ porosity of Forced-Convection Cooling of Multiple Heated Blocks in a Channel Using
Porous Covers,” Int. J. Heat Mass Transfer, 48, pp. 647–664.
␮ ⫽ viscosity of fluid, kg/m/s 关22兴 Slattery, J. C., 1972, Momentum, Energy and Mass Transfer in Continua,
␪ ⫽ dimensionless temperature, 共T − Ti兲 / 共qwH / k f 兲 McGraw-Hill, New York, pp. 57–65.
␳ ⫽ density, kg/ m3 关23兴 Martin, A. R., Saltiel, C., and Shyy, W., 1998, “Heat Transfer Enhancement
␻ ⫽ vorticity With Porous Inserts in Recirculating Flows,” ASME J. Heat Transfer, 120, pp.
⌿ ⫽ stream function 458–467.
关24兴 Angirasa, D., and Peterson, G. P., 1999, “Forced Convection Heat Transfer
Superscripts Augmentation in a Channel With a Localized Heat Source Using Fibrous Ma-
*
⫽ effective terials,” ASME J. Electron. Packag., 121, pp. 1–7.
- ⫽ average 关25兴 Cheng, P., and Hsu, C. T., 1986, “Applications of Van Driest’s Mixing Length
Theory to Transverse Thermal Dispersion in Forced Convective Flow through
Subscripts a Packed Bed,” Int. Commun. Heat Mass Transfer, 13, pp. 613–625.
b ⫽ block 关26兴 Cheng, P., Hsu, C. T., and Chowdhury, A., 1988, “Forced Convection in the
d ⫽ thermal dispersion Entrance Region of a Packed Channel With Asymmetric Heating,” ASME J.
Heat Transfer, 110, pp. 946–954.
f ⫽ fluid 关27兴 Kar, K. K., and Dybbs, A., 1982, “Internal Heat Transfer Coefficients of Po-
fs ⫽ fluid-solid interface rous Metals,” Proc. Winter Annual Meeting of ASME, Phoenix, AZ, pp. 81–91.
i ⫽ at the channel inlet 关28兴 Gamson, B. W., Thodos, G., and Hougen, O. A., 1943, “Heat, Mass and
p ⫽ along the periphery of the block surface Momentum Transfer in Flow of Gases,” Trans. AIChE, 39, pp. 1–35.
s ⫽ solid matrix 关29兴 Fu, W. S., and Huang, H. C., 1999, “Effects of a Random Porosity Model on
w ⫽ channel wall Heat Transfer Performance of Porous Media,” Int. J. Heat Mass Transfer, 42,
pp. 13–25.
关30兴 Roache, P. J., 1972, Computational Fluid Dynamics, Hermosa, Albuquerque,
New Mexico, pp. 139–173.
References 关31兴 Patankar, S. V., 1980, Numerical Heat Transfer and Fluid Flow, McGraw-Hill,
关1兴 Mahalingam, M., and Berg, H. M., 1984, “Thermal Trend in Component Level Hemisphere, New York, pp. 113–137.
Packaging,” Int. J. Hybrid Microelectron., 7, pp. 1–9. 关32兴 Hwang, G. J., and Cheng, K. C., 1970, “Boundary Vorticity Method for Con-
关2兴 Mahalingam, M., 1985, “Thermal Management in Semiconductor Device
vective Heat Transfer With Secondary Flow Application to the Combined Free
Packaging,” Proc. IEEE, 73, pp. 1396–1404.
and Forced Laminar Convection in Horizontal Tubes,” Proc. 4th Int. Heat
关3兴 Kang, B. H., Jaluria, Y., and Tewari, S. S., 1990, “Mixed Convection Transport
From an Isolated Heat Source Module on a Horizontal Plate,” ASME J. Heat Transfer Conf., Versailles, Paris, Vol. 4, Paper No. NC3.5.
Transfer, 112, pp. 653–661. 关33兴 Lee, S. L., 1989, “A Strong Implicit Solver for Two-Dimensional Elliptic
关4兴 Kim, S. Y., Sung, H. J., and Hyun, J. M., 1992, “Mixed Convection from Differential Equations,” Numer. Heat Transfer, Part B, 16, pp. 161–178.
Multiple-Layered Boards With Cross-Streamwise Periodic Boundary Condi- 关34兴 Calmidi, V. V., and Mahajan, R. L., 1999, “The Effective Thermal Conductiv-
ity of High Porosity Fibrous Metal Foams,” ASME J. Heat Transfer, 121, pp. Array of Protruding Elements in Single-Phase Forced Convection,” Int. J. Heat
466–471. Mass Transfer, 33, pp. 2659–2669.
关35兴 Kline, S. J., and Mcclintock, F. A., 1953, “Describing Uncertainties in Single- 关39兴 Lehmann, G. L., and Pembroke, J., 1991, “Forced Convection Air Cooling of
Sample Experiments,” Mech. Eng. 共Am. Soc. Mech. Eng.兲, 75, Jan., pp. 3–8. Simulated Low Profile Electronic Components: Part 1-Base Case,” ASME J.
关36兴 Moffat, R. J., 1986, “Contributions to the Theory of Single-Sample Uncer-
Electron. Packag., 113, pp. 21–26.
tainty Analysis,” ASME J. Fluids Eng., 104, pp. 250–260.
关37兴 Souza Mendes, P. R., and Santos, W. F. N., 1987, “Heat Transfer and Pressure 关40兴 Jubran, B. A., Swiety, S. A., and Hamdan, M. A., 1996, “Convective Heat
Drop Experiments in Air-Cooled in Electronic-Component Arrays,” J. Ther- Transfer and Pressure Drop Characteristics of Various Array Configurations to
mophys. Heat Transfer, 1, pp. 373–378. Simulate the Cooling of Electronic Modules,” Int. J. Heat Mass Transfer, 39,
关38兴 Garimella, S. V., and Eibeck, P. A., 1990, “Heat Transfer Characteristics of an pp. 3519–3529.
Heat Conduction in Nanofluid
Suspensions
The heat conduction mechanism in nanofluid suspensions is derived for transient pro-
Peter Vadasz cesses attempting to explain experimental results, which reveal an impressive heat trans-
Professor and Chair fer enhancement. In particular, the effect of the surface-area-to-volume ratio (specific
Fellow ASME
area) of the suspended nanoparticles on the heat transfer mechanism is explicitly ac-
e-mail: peter.vadasz@nau.edu
counted for, and reveals its contribution to the specific solution and results. The present
Department of Mechanical Engineering,
analysis might provide an explanation that settles an apparent conflict between the recent
Northern Arizona University,
experimental results in nanofluid suspensions and classical theories for estimating the
P.O. Box 15600,
effective thermal conductivity of suspensions that go back more than one century (Max-
Flagstaff, AZ 86001
well, J.C., 1891, Treatise on Electricity and Magnetism). Nevertheless, other possible
explanations have to be accounted for and investigated in more detail prior to reaching
a final conclusion on the former explanation. 关DOI: 10.1115/1.2175149兴
Keywords: nanofluids, nanoparticles suspension, heat transfer enhancement, effective

thermal conductivity, effective heat capacity, dual-phase-lagging
1 Introduction While these investigations were not done in detail but mainly at
the very basic level, Keblinski et al. 关14兴 show that the “key fac-
The reported breakthrough in substantially increasing the ther-
tors in understanding thermal properties of nanofluids are the
mal conductivity of fluids by adding very small amounts of sus-
ballistic, rather than diffusive, nature of heat transport in the
pended metallic or metallic oxide nanoparticles 共Cu, CuO, Al2O3兲
to the fluid 共Eastman et al. 关1兴, Lee et al. 关2兴兲, or alternatively nanoparticles, combined with direct or fluid mediated clustering
using nanotube suspensions 共Choi et al. 关3兴, Xuan and Li 关4兴兲 effects that provide paths for rapid heat transport.” Recent re-
conflicts with the classical theories 共Maxwell 关5兴, Batchelor 关6兴, search results presented by Xue et al. 关15兴 eliminate the
Batchelor and Green 关7兴, Hamilton and Crosser 关8兴, Jeffrey 关9兴, molecular-level layering of the liquid at the liquid/particle inter-
Davis 关10兴, Lu and Lin 关11兴, Bonnecaze and Brady 关12,13兴兲 of face as a possible heat transfer enhancement mechanism. Xue et
estimating the effective thermal conductivity of suspensions. A al. 关15兴 conclude that “the experimentally observed large en-
very small amount 共less than 1% in terms of volume fraction兲 of hancement of thermal conductivity in suspensions of solid nano-
copper nanoparticles was reported to improve the measured ther- size particles (nanofluids) can not be explained by altered thermal
mal conductivity of the suspension by 40% 共Eastman et al. 关1兴, transport properties of the layered liquid.” Furthermore, while
Lee et al. 关2兴兲, while over a 150% improvement of the effective originally Keblinski et al. 关14兴 estimated that “¼Brownian motion
thermal conductivity at a volume fraction of 1% was reported by is too slow to transport significant amounts of heat through a
Choi et al. 关3兴 for multiwalled carbon nanotubes suspended in oil. nanofluid, a conclusion supported by the results of Molecular Dy-
The procedure used by Eastman et al. 关1兴 for producing nanofluids namics simulations¼” recent research efforts such as Prasher et
containing metallic particles and dispersing them into ethylene al. 关16兴, Das et al. 关17兴, Kumar et al. 关18兴, and Jang and Choi 关19兴
glycol resulted in very little agglomeration. In the case of nano- show although in a very preliminary fashion, practically via an
tube suspensions Choi et al. 关3兴 indicate that well dispersed and order of magnitude analysis, that Brownian motion may explain
stable suspensions were produced and tested. The reported experi- the substantial enhancement of the effective thermal conductivity.
mental results for the effective thermal conductivity undertaken Indeed, Keblinski et al. 关14兴 suggested 共although did not investi-
by using the transient hot wire 共THW兲 method are substantially gate兲 a possibility where Brownian motion combined with ballis-
greater than the predictions based on existing theories and models. tic phonon effects may provide a mechanism for such enhance-
The argument laid forward by Eastman et al. 关1兴 for explaining ment. On the other hand the ballistic phonon transport alone is
such an impressive enhancement of the effective thermal conduc- known to be a substantially less effective mechanism of heat
tivity of the suspension is the fact that “¼the surface-area-to transfer than diffusion as presented in detailed studies at the
volume ratio is 1000 times larger for particles with a 10 nm di- nanoscale level by Chen 关20–22兴. Additional possible mechanisms
ameter than for particles with a 10 ␮m diameter¼” and that “an for enhancement of heat transfer in nanofluid suspensions may
obvious shortcoming” in the classical theory of estimating the also include thermo/electrophoresis and natural convection driven
effective thermal conductivity of suspensions “is the lack of any by the suspended nanoelements 共nanoparticles or nanotubes兲.
predicted dependence of conductivity on particle size.¼it does not However, while the reported results are a direct consequence of
account for the orders-of-magnitude increase in surface-area-to- the presence of nanoelements 共nanoparticles or nanotubes兲 in the
volume ratio that accompany decreasing particle size into the suspension, the measurements were not performed at the nanos-
nanocrystalline regime.” Possible explanations for the divergence cale, but rather at the macro/mesoscale. Therefore, the interest
between theory and experiments were suggested and explored should not be focused only on what occurs at the nanoscale but
very basically by Keblinski et al. 关14兴. Brownian motion of the rather on how the heat transfer at the macro/mesoscale is substan-
particles, molecular-level layering of the liquid at the liquid/ tially affected by a very small presence 共less than 1% in volume兲
particle interface, the nature of heat transport within the nanopar- of an extremely small concentration of suspended nanoelements
ticles and effects of nanoparticle clustering were investigated. 共nanoparticles or nanotubes兲.
The present investigation demonstrates that the transient heat
conduction process in suspensions is fully consistent with the
NAL OF HEAT TRANSFER. Manuscript received March 2, 2005; final manuscript re- classical method of estimating the effective thermal conductivity
ceived October 7, 2005. Review conducted by Jay M. Khodadadi. of suspensions and may provide a valid explanation for the appar-
ent heat transfer enhancement revealed experimentally. Note that
Maxwell’s 关5兴 results as well as their extensions 共Batchelor 关6兴,
Batchelor and Green 关7兴, Hamilton and Crosser 关8兴, Jeffrey 关9兴,
Davis 关10兴, Lu and Lin 关11兴, Bonnecaze and Brady 关12,13兴兲 dem-
onstrating the existence of a constant effective thermal conductiv-
ity were derived at steady state and such an effective thermal
conductivity might not be existent during transients. Furthermore,
if it does exist, the results presented here suggest that a correction
to the analytical formulation of the THW method is required when
the latter is applied to the evaluation of the effective thermal con-
ductivity of suspensions.
The method adopted in this analysis is basically a variation of
the classical method applied to heat conduction in porous media
共Nield and Bejan 关23兴兲 adapted to the problem of heat conduction
in suspensions. In the respective formulation it is already evident
that the transient heat conduction in suspensions is controlled not
only by the effective thermal conductivity but also by a fluid-to-
solid heat transfer coefficient that accounts for the heat flux at the
fluid-solid interface. The latter includes the effect of the surface-
area-to-volume ratio 共specific area兲 as well as the impact of the
heat capacity of the solid particles, which introduce a delayed
response to any temperature variation in the neighboring fluid.
While the results presented may provide a possible explanation
for the apparent heat transfer enhancement revealed experimen-
tally, there are other possible macroscale explanations that might
contribute to the latter, such as hyperbolic heat conduction 共Va-
dasz, Govender, and Vadasz, 关24兴兲, particle driven natural convec-
tion, convection induced by electrophoresis, thermophoresis, etc.
Throughout the paper all variables carrying the subscript 共·兲ⴱ
represent dimensional values while variables lacking the subscript
are dimensionless. Specific dimensional values of variables also
carry the subscript 共·兲ⴱ in order to distinguish them from their
corresponding dimensionless ones. Constant property symbols, al-
though dimensional do not carry the 共·兲ⴱ subscript, for maintaining
simplicity of notation.
2 Transient Hot Wire Method for Estimating the

Thermal Conductivity
An essential component in the following analysis is the appli-
cation of the 共THW兲 method for estimating experimentally the
thermal conductivity of the nanofluid suspension. The THW
method is well established as the most accurate, reliable and ro-
bust technique 共Hammerschmidt and Sabuga 关25兴兲 for evaluating
the thermal conductivity of fluids 共De Groot, Kestin, and Sookia-
zian 关26兴, Healy, De Groot, and Kestin 关27兴, Kestin and Wakeham,
关28兴兲 and solids 共Assael et al. 关29兴兲. It replaced the steady state Fig. 1 „a… An embedded platinum „tantalum… hot wire within a
nanofluid suspension in a cylindrical container. „b… Graphical
methods primarily because of the difficulty in determining that
representation of the averaging method for suspended solid
steady state conditions have indeed been established and for fluids particles in a fluid using a REV.
the difficulty in preventing the occurrence of natural convection
and consequently the difficulty in eliminating the effects of natural
convection on the heat flux. The THW method consists in prin-
ciple of determining the thermal conductivity of a selected straightforward its corresponding application to electrically con-
material/fluid by observing the rate at which the temperature of a ducting liquids needs further attention. The experiments con-
very thin platinum 共or alternatively tantalum兲 wire 共5 – 80 ␮m in ducted in nanofluid suspensions listed above 关1–4兴 used a thin
diameter兲 increases with time after a step change in voltage has electrical insulation coating layer to cover the platinum wire in-
been applied to it. The platinum 共tantalum兲 wire is embedded stead of using the bare metallic wire, a technique developed by
vertically in the selected material/fluid 共see Fig. 1共a兲兲 and serves Nagasaka and Nagashima 关30兴. The latter is aimed at preventing
as a heat source as well as a thermometer. The temperature of the problems such as electrical current flow through the liquid causing
wire is established by measuring its electrical resistance, the latter ambiguity of the heat generation in the wire. Alternatively, Assael
being related to the temperature via a well-known relationship. A et al. 关31兴 used tantalum wires, which were anodized in situ to
Wheatstone bridge is used to measure the electrical resistance Rw form a coating layer of tantalum pentoxide 共Ta2O5兲 which is an
of the wire. The electrical resistance of a potentiometer R3 is electrical insulator.
adjusted until the reading of the galvanometer G shows zero cur- Because of the very small diameter 共micrometer size兲 and high
rent. When the bridge is balanced as indicated by a zero current thermal conductivity of the platinum wire the latter can be re-
reading on the galvanometer G, the value of Rw can be established garded as a line source in an otherwise infinite cylindrical me-
from the known electrical resistances R1, R2, and R3 by using the dium. The rate of heat generated per unit length 共l*兲 of platinum
balanced Wheatstone bridge relationship Rw = R1R2 / R3. wire is therefore q̇l* = iV / l* W m−1, where i is the electric current
While the application of the method to solids and gases is flowing through the wire and V is the voltage drop across the wire.
Solving for the radial heat conduction due to this line heat source for oil, leading to the validity condition t* Ⰷ 13.7 ms for ethylene
leads to a temperature solution in the following closed form that glycol and t* ⬎ 7.2 ms for oil. The long times beyond which the
can be expanded in an infinite series as follows solution 共5兲 can be used reliably are therefore of the order of
冉冊冋冉冊
hundreds of milliseconds, not so long in the actual practical sense.
q̇l* r2* q̇l* 4␣t* r2* r4* On the other hand the experimental time range is limited from
T* = Ei = − ␥ + ln + −
4␲k 4␣t* 4␲k 2
r* 4␣t* 64␣2t2* above as well in order to ensure the lack of natural convection that
册
develops at longer time scales. Xuan and Li 关6兴 estimate this
r6* upper limit for the time that an experiment may last before natural
+ −¯ 共1兲
1152␣3t3* convection develops as about 5 s. They indicate that “An experi-
ment lasts about 5 s. If the time is longer, the temperature differ-
where Ei共·兲 represents the exponential integral function, and ␥
ence between the hot-wire and the sample fluid increases and free
= ln共␴兲 = 0.5772156649 is Euler’s constant. For a line heat source convection takes place, which may result in errors.” Lee et al. 关2兴
embedded in a cylindrical cell of infinite radial extent and filled while using the THW method and providing experimental data in
with the test fluid one can use the approximation r2* / 4␣t* Ⰶ 1 in the time range of 1 – 10 s, indicate in their Fig. 3 the “valid range
Eq. 共1兲 to truncate the infinite series and yield of data reduction” to be between 3 – 6 s. Our estimations evalu-
T* ⬇
q̇l*
4␲k
冋冉冊冉冊册
− ␥ + ln
4␣t*
r2*
+O
r2*
4␣t*
共2兲
ated above confirm these lower limits as a very safe constraint and
we assume that the upper limits listed by Xuan and Li 关6兴 and Lee
et al. 关2兴 are also good estimates. Within this time range the ex-
Equation 共2兲 reveals a linear relationship, on a logarithmic time perimental results should produce a linear relationship, on a loga-
scale, between the temperature and time. Therefore, one way of rithmic time scale, between the temperature and time.
evaluating the thermal conductivity is from the slope of this rela-
tionship evaluated at r* = rw*, for example, rw* being the radius of 3 Problem Formulation of Heat Conduction in Sus-
the platinum wire. For r* = rw* the condition for the series trunca- pensions and Dual-Phase-Lagging
2
tion rw* / 4␣t* Ⰶ 1 can be expressed in the following equivalent To investigate the transient heat conduction in a fluid containing
form that provides the validity condition of the approximation in suspended solid particles 共Figs. 1共a兲 and 1共b兲兲 one considers the
the form phase-averaged equations. The averaging process over a represen-
2 tative elementary volume 共REV兲 VREV 共Fig. 1共b兲兲 can be repre-
r w* sented by introducing the following definitions: VREV = Vs,REV
t* Ⰷ 共3兲
4␣ + V f,REV, where Vs,REV and V f,REV, are the volumes occupied by
2 the solid and fluid phases within the REV, respectively, and ␧
The value of to* = rw* / 4␣ provides a validity condition of the ex- = Vs,REV / VREV is the volume fraction of the solid phase. Then, for
perimental readings, i.e., t* Ⰷ to*. For any two readings of tem- the local temperature variable of any solid particle j, Tsj*, and for
perature T1* and T2* recorded at times t1* and t2* respectively, the the corresponding local temperature variable of the fluid, Tfl*, one
temperature difference 共T2* − T1*兲 can be approximated by using can evaluate the average temperatures over the REV in the form
Eq. 共2兲, as long as the latter validity condition is fulfilled, in the Ts* = 共兺np
i=1Tsj*兲 / n p, where n p = nVREV is the number of solid par-
form ticles within the REV while n is the number of solid particles per
共T2* − T1*兲 ⬇
iV
4␲kl*
ln 冋冉冊册
t 2*
t 1*
共4兲
unit volume, and T f * = 共兰V f Tfl*dV兲 / V f . As a result, one obtains the
average temperatures of the solid and fluid phases, respectively,
associated with the REV located at the coordinates of its centroid,
where we replaced the heat source with its explicit dependence on r*. As the REV can be moved within the domain its centroid
i, V and l*, i.e., q̇l* = iV / l*. From Eq. 共4兲 one can express the moves to different spatial coordinates and the resulting average
thermal conductivity k explicitly in the form phase temperatures vary. The results are the average temperature
冋冉冊册
functions Ts*共r* , t*兲 and T f *共r* , t*兲, where r* are the coordinates of
iV t 2* the centroid at any given position. The resulting phase-averaged
k⬇ ln 共5兲 equations have the form
4␲共T2* − T1*兲l* t 1*
Equation 共5兲 is a very accurate way of estimating the thermal ⳵ T s*
␥s = h共T f * − Ts*兲共6兲
conductivity as long as the validity conditions for appropriateness ⳵t*
of the problem derivations used above are fulfilled. A finite length
of the platinum 共tantalum兲 wire, the finite size of the cylindrical ⳵T f *
container, the heat capacity of the platinum 共tantalum兲 wire, and ␥f = keⵜ2*T f * − h共T f * − Ts*兲共7兲
⳵t*
possibly natural convection effects are examples of possible de-
viations of any realistic system from the one used in deriving Eq. where T f *共r* , t*兲, Ts*共r* , t*兲 are average temperature values for the
共5兲. De Groot et al. 关26兴, Healy et al. 关27兴, and Kestin and Wake- fluid and solid phases respectively taken over a REV that is large
ham 关28兴 introduce an assessment of these deviations and possible enough to be statistically valid but sufficiently small compared to
corrections to the THW readings to improve the accuracy of the the size of the domain. In Eqs. 共6兲 and 共7兲 ␥s = ␳scs and ␥ f = ␳ f c p
results. In general all the deviations indicated above could be represent the heat capacity of the solid and fluid phases, respec-
eliminated via the proposed corrections provided the validity con- tively, with ␳s and ␳ f being the density of the solid and fluid
dition listed in Eq. 共3兲 is enforced as well as an additional condi- phases, respectively, and cs and c p being the specific heat of the
tion that ensures that natural convection is absent. The validity solid and fluid phases, respectively. Similarly, ke is the effective
condition 共3兲 implies the application of Eq. 共5兲 for long times thermal conductivity of the fluid as established by Maxwell 关5兴
only. However, when evaluating this condition 共3兲 to data used in and developed further including higher order corrections and other
the nanofluid suspensions experiments considered in this paper than spherical particle geometries by Batchelor 关6兴, Batchelor and
one obtains explicitly the following values. For a 76.2 ␮m diam- Green 关7兴, Hamilton and Crosser 关8兴, Jeffrey 关9兴, Davis 关10兴, Lu
eter of platinum wire used by Eastman et al. 关1兴, Lee et al. 关2兴, and Lin 关11兴, Bonnecaze and Brady 关12,13兴. Note that Ts*共r* , t*兲 is
Choi et al. 关3兴, and an electrical insulation coating thickness of a function of the space variables represented by the position vec-
10 ␮m, the wire radius is rw* = 4.81⫻ 10−5 m leading to to* tor r* = r*êr + ␾*ê␾ + z*êz, in addition to its dependence on time,
2 2
= rw* / 4␣ = 13.7 ms for ethylene glycol and to* = rw* / 4␣ = 7.2 ms because Ts*共r* , t*兲 depends on T f *共r* , t*兲 as explicitly stated in Eq.
共6兲, although no spatial derivatives appear in Eq. 共6兲. The reason ␥ s␥ f ke ␥ sk e ␥s
for the lack of macroscopic level conduction mechanism in Eq. ␶q = ; ␣e = ; ␶T = =
h共␥s + ␥ f 兲共␥s + ␥ f 兲 h共␥s + ␥ f 兲␣e h
共6兲 representing the heat transfer within the solid phase is the fact
that the solid particles represent the dispersed phase in the fluid 共12兲
suspension and therefore the solid particles can conduct heat be- In Eq. 共12兲 ␶q and ␶T are the heat flux and temperature related time
tween themselves only via the neighboring fluid. When steady lags linked to the dual-phase-lagging 共DuPhlag兲 to be discussed
state is accomplished ⳵Ts* / ⳵t* = ⳵T f * / ⳵t* = 0 leading to local ther- below, while ␣e is the effective thermal diffusivity of the suspen-
mal equilibrium between the solid and fluid phases, i.e., Ts*共r兲 sion. It may be observed from Eq. 共12兲 that there is a dual effect
= T f *共r兲. At steady state, therefore, Maxwell’s 关5兴 method and its of the heat capacities on the effective parameters of the uncoupled
further derivations 共Batchelor 关6兴, Batchelor and Green 关7兴, system in the sense that the heat flux time lag ␶q is affected by the
Hamilton and Crosser 关8兴, Jeffrey 关9兴, Davis 关10兴, Lu and Lin heat capacities of the solid and fluid phases as thermal capacitors
关11兴, Bonnecaze and Brady 关12,13兴兲 can be applied to evaluate the connected in series following the relationship 1 / ␥es = 1 / ␥s + 1 / ␥ f
effective thermal conductivity ke of the suspension. The parameter = 共␥s + ␥ f 兲 / ␥s␥ f , while the effective thermal diffusivity is affected
h, carrying units of W m−3 K−1, represents an integral heat trans- by the heat capacities of the solid and fluid phases as thermal
fer coefficient for the heat conduction at the solid-fluid interface capacitors connected in parallel following the relationship ␥ep
within an REV. It is assumed to be independent of time and its
= 共␥s + ␥ f 兲. The boundary and initial conditions applicable to the
general relationship to the surface-area-to-volume ratio 共specific
uncoupled equation 共11兲 are obtained from the original boundary
area兲 is derived in Appendix A.
and initial conditions 共9兲 in combination with Eq. 共7兲 and include
For the particular case pertaining to this investigation that con-
two initial conditions because Eq. 共11兲 is second order in time
sists of a thin hot wire embedded in a cylindrical container insu-
lated on its top and bottom sides as presented in Fig. 1 one can r* = ro*:T f * = TC*
assume that the heat is transferred in the radial direction only
leading to the sensible assumption that ⳵Ts* / ⳵␾* = ⳵Ts* / ⳵z*
= ⳵T f * / ⳵␾* = ⳵T f * / ⳵z* = 0, hence rendering Eqs. 共6兲 and 共7兲 to the
following operator form
r* = rw*:q fr* = − ke 冉冊
⳵T f *
⳵r* rw*
= qo* = const Þ 冉冊
⳵T f *
⳵r* rw*
=−
q o*
ke
冉冊冉冊
冤冥
⳵
冋册
␥s +h −h ⳵T f *
⳵t* T s* t* = 0:T f * = TC* and =0 共13兲
⳵t*
冉冉冊冊
=0 t*=0
⳵ 1 ⳵ ⳵ Tf*
−h ␥f − ke r* +h Equation 共11兲 is identical to the one obtained by replacing the
⳵t* r* ⳵r* ⳵r* Fourier Law and the fluid-solid interface heat transfer with a
共8兲 DuPhlag constitutive relationship between the heat flux and tem-
perature gradient 共Tzou 关32,33兴兲 in the form
subject to the following boundary and initial conditions
q*共r*,t* + ␶q兲 = − keⵜ*T*共r*,t* + ␶T兲共14兲
r* = ro*:T f * = TC* = const
where the relationship between the heat flux and temperature gra-
冉冊冉冊
dient is not instantaneous but rather affected by two time lags, a
⳵T f * ⳵T f * q o* heat flux time lag ␶q, and a temperature gradient time lag, ␶T. By
r* = rw*:q fr* = − ke = qo* = const Þ =−
⳵r* rw* ⳵r* rw* ke expanding Eq. 共14兲 in a Taylor series in time one obtains
⳵q*
q*共r*,t*兲 + ␶q 共r*,t*兲 + O共␶2q兲
t* = 0:T f * = Ts* = TC* = const 共9兲 ⳵t*
where TC* is the ambient temperature and qo* is the heat flux over
the fluid-platinum-wire interface that is related to the Joule heat-
ing of the platinum wire in the form qo* = iV / 共␲dw*l*兲, where dw*
冋
= − ke ⵜT*共r*,t*兲 + ␶T
⳵共ⵜ*T*兲
⳵t*
共r*,t*兲 + O共␶T2 兲册共15兲
and l* are the diameter and the length of the platinum wire, re- Truncating the Taylor series by including terms up to order O共␶q兲
spectively. and O共␶T兲 yields
The coupling between the two equations via the term h共T f *
− Ts*兲 can be removed by evaluating the operator determinant
leading to
q* + ␶q
⳵q*
⳵t*
冋
= − k e ⵜ *T * + ␶ T
⳵共ⵜ*T*兲
⳵t*
册共16兲
冋冉冊冉冉冊冊册
This DuPhlag formulation is applied to the thermal conduction
⳵ ⳵ 1 ⳵ ⳵ energy equation for the effective medium
␥s +h ␥f − ke r* + h − h 2 T i* = 0
⳵t* ⳵t* r* ⳵r* ⳵r*
⳵T*
" i = s, f 共10兲 ␥e + ⵜ* · q* = 0 共17兲
⳵t*
where the index i can take the values s representing the solid by replacing the fluid-solid interface heat transfer term with the
phase or f standing for the fluid phase. Assuming that the value of DuPhlag formulation. Applying now the 共ⵜ* · 兲 operator in Eq.
the heat transfer coefficient h is independent of time one can 共16兲 produces
present Eq. 共10兲 explicitly after dividing it by h共␥s + ␥ f 兲 in the
form
冋 1 + ␶q
⳵
册ⵜ* · q* = − ke 1 + ␶T 冋⳵
ⵜ 2T * 册共18兲
冋冉冊冉冊册
⳵t* ⳵t* *
⳵ 2T * ⳵ T * 1 ⳵ ⳵T* ␶T ⳵ ⳵ 2T *
␶q 2 + = ␣e r* + r* 共11兲 Substituting ⵜ* · q* = −␥e⳵T* / ⳵t* from Eq. 共17兲 into Eq. 共18兲 and
⳵t* ⳵t* r* ⳵r* ⳵r* r* ⳵r* ⳵ r *⳵ t *
dividing the resulting equation by ␥e yields the equation for the
where the index i was removed as the intention is to use this temperature of the effective medium due to DuPhlag, in a form
equation from this point onwards to represent the effective fluid which is identical to Eq. 共11兲 if the ⵜ2* operator is replaced with its
phase, and the following notation was used one-dimensional version 共1 / r*兲⳵ / ⳵r*关r*⳵共·兲 / ⳵r*兴, as follows
␶q
⳵ 2T * ⳵ T *
⳵t2*
+
⳵t*
冋
= ␣e ⵜ2*T* + ␶Tⵜ2*
⳵T*
⳵t*
冉冊册共19兲
cases the Fourier thermal diffusion solution applies and can be
presented in the classical form 共e.g., Özisik 关37兴兲
⬁
where ␥e, ␣e, ␶q, and ␶T are effective parameters which the
DuPhlag formulation cannot relate to the specific properties of the T = − rw ln共r兲 + 兺A e ␬
n=1
n
− 2nt
关Y o共␬n兲Jo共␬nr兲 − Jo共␬n兲Y o共␬nr兲兴
individual phases. However, the original derivation presented
above resulting in Eq. 共11兲 clearly identifies the explicit form of 共27兲
these effective coefficients and their accurate relationship with the
properties of the individual phases in the suspension. where Jo共␬nr兲 and Y o共␬nr兲 are the order 0 Bessel functions of the
A direct and very important property of these parameters is first and second kind, respectively. The eigenvalues ␬n are the
obtained in evaluating the ratio ␶T / ␶q by using Eq. 共12兲 leading to roots of the equation
the following result Jo共␬n兲Y 1共␬nrw兲 − Y o共␬n兲J1共␬nrw兲 = 0,
␶T ␥s and the coefficient An is established from the initial condition
␤= =1+ ⬎1 共20兲
␶q ␥f 共T兲t=0 = 0 in the form
Since the ratio of positive valued properties in the second term of ␲2rwJ21共␬nrw兲关Y o共␬n兲Jo共␬nrw兲 − Jo共␬n兲Y o共␬nrw兲兴
Eq. 共20兲 is always positive, the time lags ratio is always greater An =
than 1, i.e., ␶T / ␶q ⬎ 1. The latter conclusion, which is identical to 2关J2o共␬nrw兲 − J21共␬nrw兲兴
DuPhlag conduction in porous media 共Vadasz 关34–36兴兲 and is The analytical solution to the DuPhlag equation 共22兲 subject to
based on a physical argument, has a profound impact on the fol- the boundary conditions 共24兲 and 共25兲 and initial conditions 共26兲
lowing results. It applies generally to Fourier heat conduction in is obtained by superposition of the steady state, Tst共r兲, and tran-
suspensions subject to the interface heat transfer formulation pre- sient Ttr共r , t兲, solutions in the form
sented in Eqs. 共6兲 and 共7兲, and is not restricted to any specific
geometry or any specific boundary conditions. T = Tst共r兲 + Ttr共r,t兲共28兲
The steady state part of the solution is governed by the equation
4 Analytical Solution
In order to produce an analytical solution to the problem it is
冉冊
d dTst
r
dr dr
=0 共29兲
convenient to render Eq. 共11兲 into a dimensionless form by intro- which is simply integrated subject to the boundary conditions
冉冊
ducing the following dimensionless variables
dTst
q* 共T* − TC*兲ke r* ␣ et * r = 1:Tst = 0; r = r w: =−1 共30兲
q= ; T= ; r= ; t= 共21兲 dr rw
2
q o* q o*r o* r o* r o*
to yield the steady state solution in the form
leading to
共31兲
冉冊冉冊
Tst = − rw ln共r兲
⳵ 2T ⳵ T 1 ⳵ ⳵ T 1 ⳵ ⳵ 2T
Foq 2 + = r + FoT r 共22兲 The transient solution is governed by the equation
⳵t ⳵t r ⳵r ⳵r r ⳵r ⳵r⳵t
where the following two dimensionless groups representing a heat
flux Fourier number, Foq, and a temperature Fourier number, FoT,
Foq
⳵2Ttr ⳵Ttr 1 ⳵ ⳵Ttr
⳵t2
+
⳵t
= r
r ⳵r ⳵r
+ FoT 冉冊
1 ⳵ ⳵2Ttr
r
r ⳵r ⳵r⳵t
冉冊共32兲
emerged. These Fourier numbers are defined in the form which is subject to the homogeneous boundary conditions
Foq =
␣ e␶ q
2
r o*
; FoT =
␣ e␶ T
2
r o*
共23兲
r = 1:Ttr = 0; r = r w: 冉冊
⳵Ttr
⳵r rw
=0 共33兲
and the ratio between them is identical to the ratio between the
and the initial conditions
time lags, i.e., FoT / Foq = ␶T / ␶q. The applicable dimensionless
boundary and initial conditions are obtained by rendering Eq. 共13兲
dimensionless using the scales introduced by Eq. 共21兲 in the form t = 0:Ttr = − Tst共r兲 = rw ln共r兲 and 冉冊 ⳵Ttr
⳵t t=0
=0 共34兲
r = 1:T = 0 共24兲 The solutions to Eq. 共32兲 subject to the boundary conditions 共33兲
冉冊冉冊
and initial conditions 共34兲 is obtained by separation of variables
⳵T ⳵T
r = rw:qr = − =1Þ =−1 共25兲 and expansion in terms of orthogonal eigenfunctions in the form
⳵r rw ⳵r rw Ttr = S共t兲R共r兲, which upon substitution into Eq. 共32兲 produces the
冉冊
following eigenvalue-eigenfunction problems
冉冊
⳵T
t = 0:T = 0 and =0 共26兲 1 d dRon
⳵t t=0 r + ␬nRon = 0 共35兲
r dr dr
Prior to presenting the analytical solution to the DuPhlag prob-
lem represented by Eqs. 共22兲 and 共24兲–共26兲 it is of interest to subject to the homogeneous boundary conditions
冉冊
present the case corresponding to the Fourier solution because this
dRon
will be our reference case, the one being used in the experimental r = 1:Ron = 0; r = r w: =0 共36兲
evaluation of the effective thermal conductivity when using the dr rw
THW method. It can be observed that Eq. 共22兲 degenerates to the
Fourier thermal diffusion equation when Foq = FoT = 0, corre- and
sponding to ␶q = ␶T = 0. The latter may occur either when the heat d 2S n dSn
capacity of the solid is excessively low, i.e., when ␥s → 0, or when + c fn + ␻2nSn = 0 共37兲
dt2 dt
the fluid-solid interface heat transfer coefficient is excessively
large, i.e., when h → ⬁ as can be observed from Eq. 共12兲. In such where
共1 + ␬2nFoT兲 ␬2n FoT ␶T
c fn = ; ␻2n = 共38兲 ␤= = 艋1 共47兲
Foq Foq Foq ␶q
Equation 共35兲 is the Bessel equation of order 0, producing solu- However Eq. 共20兲 shows that based on physical arguments the
tions in the form of Bessel functions time lag ratio ␶T / ␶q is always greater than 1, i.e., ␶T / ␶q ⬎ 1. There-
fore, underdamped 共oscillatory兲 solutions, which require accord-
Ron = A1Jo共␬nr兲 + A2Y o共␬nr兲共39兲 ing to Eq. 共47兲 that ␶T / ␶q ⬍ 1, are being ruled out. Similarly, since
where Jo共␬nr兲 and Y o共␬nr兲 are the order 0 Bessel functions of the the condition for critically-damped solutions is FoT / Foq = ␶T / ␶q
first and second kind, respectively. Applying the boundary condi- = 1, but in reality this ratio is greater than 1, i.e., ␶T / ␶q ⬎ 1,
tions 共36兲 one obtains the equation for the eigenvalues in the form critically-damped solutions are ruled out as well, meaning that the
eignevalues ␭n1 and ␭n2, are real, distinct, and negative. There-
Jo共␬n兲Y 1共␬nrw兲 − Y o共␬n兲J1共␬nrw兲 = 0 共40兲 fore, solution 共45兲 represents the complete form of the solution.
where J1共␬nrw兲 and Y 1共␬nrw兲 are the order 1 Bessel functions of The values of the coefficients B1n and B2n are being evaluated
by using the two initial conditions presented in Eq. 共26兲 or 共34兲.
the first and second kind, respectively. This equation for the ei-
genvalues 共40兲 as well as the eignefunctions 共39兲 are identical to By applying the initial condition 共⳵T / ⳵t兲t=0 = 0 with the solution
the ones obtained for Fourier thermal diffusion as observed in Eq. 共45兲 yields a relationship between the coefficients B1n and B2n in
共27兲 and the text following it. The boundary conditions also yield the form B2n = −共␭n1 / ␭n2兲B1n transforming Eq. 共45兲 to its final
the following form for the eignefunctions form
⬁
Ron = An关Y o共␬n兲Jo共␬nr兲 − Jo共␬n兲Y o共␬nr兲兴共41兲
On the other hand, Eq. 共37兲 is the equation of a linear damped
T = − rw ln共r兲 + 兺 B 关␭
n=1
n n2e
␭n1t
− ␭n1e␭n2t兴关Y o共␬n兲Jo共␬nr兲
oscillator that allows in principle, overdamped, critically-damped

− Jo共␬n兲Y o共␬nr兲兴共48兲
and underdamped solutions depending on the values of c fn and ␻2n.
The solution to Eq. 共37兲 is obtained by solving the auxiliary equa- By applying now the other initial condition 共T兲t=0 = 0 to Eq. 共48兲,
tion for the eigenvalues ␭n multiplying Eq. 共48兲 by Rom = Y o共␬m兲Jo共␬mr兲 − Jo共␬m兲Y o共␬mr兲, in-
tegrating the result over the domain r 苸关rw , 1兴 while using the
␭2n + c fn␭n + ␻2n = 0 共42兲
orthogonal properties of Bessel functions
where c fn and ␻2n
are defined by Eq. 共38兲. Hence, the solutions for
the eigenvalues ␭n are 冕
1
Ron共␬n,r兲Rom共␬m,r兲r dr = 再 0 for m ⫽ n
冎共49兲
冋冑册 rw
N共␬n兲 for m = n
c fn ␻2n
␭n1 = − 1+ 1− 4 2 共43兲 where N共␬n兲 = 兰r1 rR2on共␬n , r兲dr is the norm, produces the follow-
2 c fn w
ing result for Bn
␭n2 = −
c fn
2
冋冑册
1− 1−4
␻2n
c2fn
共44兲 Bn =
− rw关Y o共␬n兲Jo共␬nrw兲 − Jo共␬n兲Y o共␬nrw兲兴
共␭2n − ␭1n兲␬2n
1
N共␬n兲
共50兲
producing the complete solution in the form The norm evaluated for this problem is presented in the form
⬁ 1 ␲2 ␬2nJ21共␬nrw兲
共51兲
T = − rw ln共r兲 + 兺关B
n=1
1ne
␭n1t
+ B2ne ␭n2t
兴关Y o共␬n兲Jo共␬nr兲
=
N共␬n兲 2 关J21共␬nrw兲 − J2o共␬nrw兲兴
and therefore the explicit form of the coefficient Bn in the final
− Jo共␬n兲Y o共␬nr兲兴共45兲 form of the solution 共48兲 is
The solution Sn共t兲 represented by the terms in the first brackets in ␲2rwJ21共␬nrw兲关Y o共␬n兲Jo共␬nrw兲 − Jo共␬n兲Y o共␬nrw兲兴
Eq. 共45兲 is overdamped if for some values of n 艌 1 the condition Bn = 共52兲
2共␭n2 − ␭n1兲关J2o共␬nrw兲 − J21共␬nrw兲兴
4␻2n ⬍ c2fn is satisfied, leading to real values of ␭n1 and ␭n2, and to
the solution 共45兲 above, it is critically damped if for some value of where ␭n1 and ␭n2 are defined by Eqs. 共43兲 and 共44兲, respectively,
n = ncr 艌 1 the condition 4␻2n,cr = c2fn,cr is satisfied, corresponding to and ␬n " n = 1 , 2 , 3 , ¼ are the solutions of the eignevalues equa-
␭n1 = ␭n2 = ␭n,cr = −c fn,cr / 2, leading to a solution of the form tion 共40兲.
关B1n,cre␭n,crt + B2n,crte␭n,crt兴关Y o共␬n,cr兲Jo共␬n,crr兲 − Jo共␬n,cr兲Y o共␬n,crr兲兴
and it is underdamped if for some values of n 艌 1 the condition
4␻2n ⬎ c2fn is satisfied producing complex values of ␭n1 and ␭n2 in 5 Corrections to Experimental Data
the form ␭n1 = ␭rn + i␭in, ␭n2 = ␭rn − i␭in, leading to the following
In order to derive the deviations from the Fourier to the
form of the solution e␭rnt关Bnei␭int + B*ne−i␭int兴关Y o共␬n兲Jo共␬nr兲 DuPhlag thermal conduction solutions and evaluate the required
− Jo共␬n兲Y o共␬nr兲兴 where B*n stands for the complex conjugate of Bn. corrections to the THW experimental method to render it appli-
The condition for underdamped solutions and its associated os- cable to suspensions we use the corresponding Fourier solution to
cillations is further explored to obtain explicit criteria in terms of the same problem by using Eq. 共27兲 leading to the following di-
the primitive parameters of the original system. By using the defi- mensional solution for the wire temperature Tw*共t兲 at r = rw
nitions from Eq. 共38兲 it produces the condition for underdamped
and critically-damped solutions in the form q o*r o*
关Tw*共t兲 − TC*兴 = 关− rw ln共rw兲 + f共t兲兴共53兲
kapp
c2fn 关1 + FoT ␬2n兴2
= 艋1 共46兲 where kapp is the apparent effective thermal conductivity and
4␻2n 4 Foq ␬2n
⬁
兺C
where the equality part applies to critically-damped solutions. An
analysis of inequality 共46兲 presented in Appendix B yields the f共t兲 = n exp共− ␬2nt兲共54兲
n=1
following necessary and sufficient condition for underdamped and
critically-damped solutions to materialize The coefficient Cn is defined in the form
␲2rwJ21共␬nrw兲关Y o共␬n兲Jo共␬nrw兲 − Jo共␬n兲Y o共␬nrw兲兴2 Lee et al. 关2兴, and Choi et al. 关3兴, and two different values of
Cn = 共55兲 effective thermal diffusivity ␣e corresponding to copper in ethyl-
2关J2o共␬nrw兲 − J21共␬nrw兲兴
ene glycol and carbon nanotubes in oil, respectively. The effective
When evaluating the thermal conductivity by applying the THW thermal conductivity used to evaluate the effective thermal diffu-
method and using the Fourier Law one obtains from Eq. 共53兲 sivity was evaluated by using the Hamilton and Crosser 关8兴 equa-
tions that represent an application of Maxwell’s equation 关5兴 to
q o*r o* other than the spherical geometry. As a result, for copper nano-
kapp = 关− rw ln共rw兲 + f共t兲兴共56兲
关Tw*共t兲 − TC*兴 particles in ethylene glycol at a volumetric solid fraction of ␧
= 0.01 one obtains the values of ␣e = 4.227⫻ 10−8 m2 s−1, ke
where the temperature difference 关Tw*共t兲 − TC*兴 is represented by = 0.257 W m−1 K−1 and a time lags ratio of ␤ = ␶T / ␶q = FoT / Foq
the recorded experimental data and the value of the heat flux at the = 2.2857. The corresponding values for carbon nanotubes sus-
fluid-platinum-wire interface qo* is evaluated from the Joule heat- pended in oil are ␣e = 8.045⫻ 10−8 m2 s−1, ke = 0.154 W m−1 K−1
ing of the hot wire in the form qo* = iV / 2␲rw*l*, where 2␲rw*l* is and a time lag ratio of ␤ = ␶T / ␶q = FoT / Foq = 1.113. Consequently
the heat transfer area of the hot wire, with l* being the length of the time conversion from dimensionless to dimensional follows
the platinum wire, i is the electric current flowing through the 2
the respective factoring t* = 共ro* / ␣e兲t = 1.479⫻ 104t s for copper
wire, and V is the voltage drop across the wire.
nanoparticles suspended in ethylene glycol, and t* = 7.769
A method of synthetic experimental emulation data 共SEED兲 is
applied now to evaluate the deviation between Fourier and Dual- ⫻ 103t s for carbon nanotubes suspended in oil. The same values
Phase-Lagging thermal conduction. According to the SEED produce the following validity time range for the application of
the THW method as indicated by Eq. 共3兲 and the text following
method one assumes that the data expressed by 关Tw*共t兲 − TC*兴 rep-
Eq. 共5兲 in the form 13.7 msⰆ t* ⬍ 5 s for copper nanoparticles
resent a different than Fourier conduction solution, in this case a
suspended in ethylene glycol and 7.2 msⰆ t* ⬍ 5 s for carbon
DuPhlag thermal conduction solution. Then we substitute in Eq.
nanotubes suspended in oil. A value of rw = 2 ⫻ 10−3 was used in
共56兲 the values of 关Tw*共t兲 − TC*兴 obtained from the DuPhlag solu-
all computations. If the relationships for h that are applicable to
tion expressed by Eq. 共48兲 in the form
porous media are being 共without justification兲 extrapolated to in-
q o*r o* clude nanofluid suspensions one evaluates very large values of h
关Tw* − TC*兴 = 关− rw ln共rw兲 + g共t兲兴共57兲 leading to very small values for the time lags ␶q and ␶T, and
kact
consequently very small values of the Fourier numbers, Foq and
where kact is the actual effective thermal conductivity and FoT. However, further and a different dependence of h on particle
⬁ size is anticipated when the particle size is reduced to the nano-
g共t兲 = 兺 C̃ 关␭
n=1
n n2 exp共␭n1t兲 − ␭n1 exp共␭n2t兲兴共58兲
scale level predominantly when the ballistic rather than diffusive
nature of heat transfer becomes dominant 共Chen 关20–22兴兲 and
hence reducing by a few orders of magnitude the rate of heat
The coefficient C̃n is related to Cn defined in Eq. 共54兲 by the transfer, implying a consequential reduction of the value of h to
following relationship compensate for the otherwise substantial increase of h as the par-
ticle size decreases. The latter opposing trends in the variation of
Cn the value of h as the particle size is reduced to the nanoscale level
C̃n = 共59兲
共␭n2 − ␭n1兲 suggest much larger values for the time lags ␶q and ␶T, and con-
sequently larger values of the Fourier numbers, Foq and FoT.
where ␭n1 and ␭n2 are defined by Eqs. 共43兲 and 共44兲, respectively. Therefore, a wide range of Fourier numbers Foq and FoT spanning
Following the substitution of the values of 关Tw*共t兲 − TC*兴 obtained over a few orders of magnitude 共but still quite small Fourier num-
from the DuPhlag solution expressed by Eq. 共57兲 into Eq. 共56兲 ber values兲 was used in order to compensate for the lack of accu-
leads to the following ratio between the apparent and actual effec- rate knowledge of the value of the solid-fluid interface heat trans-
tive thermal conductivities fer coefficient h, especially for nanoscale suspensions, as
kapp 关− rw ln共rw兲 + f共t兲兴 discussed in Appendix A. The Fourier number ratio ␤ = FoT / Foq
␴= = 共60兲 used was as indicated above for copper nanoparticles suspended
kact 关− rw ln共rw兲 + g共t兲兴 in ethylene glycol and carbon nanotubes suspended in oil, respec-
where kapp is the apparent effective thermal conductivity obtained tively. Therefore, for each one of these types of suspensions, a
from the Fourier conduction solution while kact is the actual effec- variation of Foq was applied while the corresponding value of FoT
tive thermal conductivity that corresponds to data that follow the can be evaluated by FoT = ␤Foq with the values of ␤ consistent
DuPhlag conduction, and where f共t兲 can be evaluated from Eq. with the ones listed above.
共54兲 while g共t兲 is evaluated from Eq. 共58兲. The ratio between the To conclude so far, a correction factor ␴, as presented in Eq.
two will provide the deviation of the apparent effective thermal 共60兲, can be used in connection with the THW method in order to
conductivity from the actual one. apply the latter method to suspended particles in fluids.
The dimensional values used in the computational results are
identical to the ones used in the experimental setup of Eastman et 6 Results and Discussion
al. 关1兴, Lee et al. 关2兴 for metals and metal oxides suspended in The evaluated dimensionless wire temperature at r* = rw* as a
ethylene glycol, or alternatively for nanotube suspensions in oil
function of the dimensional time corresponding to either nanopar-
used by Choi et al. 关3兴. Equation 共60兲 can be evaluated as a func-
ticles suspended in ethylene glycol and carbon nanotubes sus-
tion of time for any given values of Fourier numbers, Foq , FoT.
pended in oil and a range of values of Foq varying between Foq
The dependence of the results on the Fourier numbers is estab-
= 10−4 and Foq = 10−6 is presented in Figs. 2–7 on a logarithmic
lished due to the dependence of the DuPhlag eigenvalues on Foq
time scale. The objective of the wire temperature results is to
and FoT, which are needed in the evaluation of g共t兲. Other than observe whether the DuPhlag solution produces results that are
that the thermal conductivity ratio depends only on the dimension- approximately linear in time on a logarithmic time scale, as the
less time and on the value of rw = rw* / ro*. To convert the thermal Fourier solution is anticipated to produce, especially for the time
conductivity ratio and express it in terms of the dimensional val- ranges indicated above. The latter is particularly important be-
ues of time we use the scaling definition introduced by Eq. 共21兲 in cause the experimental results presented by Eastman et al. 关1兴, Lee
2
the form t* = 共ro* / ␣e兲t. We use a value of ro* = 0.025 m which is et al. 关2兴, and Choi et al. 关3兴 do not include the wire temperature
the radius of the cylindrical container used by Eastman et al. 关1兴, as a function of time used to evaluate the effective thermal con-
Fig. 3 The dimensionless wire temperature corresponding to
the Fourier as well as the DuPhlag solutions as a function of
time, on a logarithmic time scale, for a Fourier number of Foq
= 10−5 and for copper nanoparticles suspended in ethylene gly-
Fig. 2 The dimensionless wire temperature corresponding to col. „a… Transient solution for the time range 0.5 s < t* < 100 s; „b…
the Fourier as well as the DuPhlag solutions as a function of detail within the time range 0.5 s < t* < 5 s.
col. „a… Transient solution for the time range 0.5 s < t* < 120 s; „b…
detail within the time range 1 s < t* < 10 s. can be seen as approximately linear in time on a logarithmic time
scale, while the Fourier solution is almost accurately linear in time
on a logarithmic time scale, however their slopes are substantially
different. Similar results that apply to copper nanoparticles in eth-
ductivity, with the only exception of a calibration result for the ylene glycol and a Fourier number of Foq = 10−5 are presented in
fluid without the suspended nanoparticles which was presented by Fig. 3共a兲 where the overlapping between the DuPhlag and the
Lee et al. 关2兴. Initial evaluation of the Fourier dimensionless so- Fourier solutions at long times is evident. The corresponding re-
lution 共53兲 was compared first to the “THW” wire temperature sults zoomed within the short time range 0.5 s ⬍ t* ⬍ 5 s are pre-
solution evaluated based on a dimensionless form of Eq. 共2兲 sented in Fig. 3共b兲. They show that within this time range the
clearly identifying a straight line on a logarithmic time scale and DuPhlag solution produces results that are approximately linear in
very small deviations between the two. Actually for dimensional time on a logarithmic time scale, while the Fourier solution is
times above 1 s the relative deviations between the two are less accurately linear in time on a logarithmic time scale, however
than 5% with less than 1% deviations for times larger than 3 s. different slopes are generally associated with each solution. The
The dimensionless wire temperature corresponding to the Fou- results applicable to copper nanoparticles in ethylene glycol and a
rier as well as the DuPhlag solutions as a function of time is Fourier number of Foq = 10−6 are presented in Fig. 4 within the
presented on a logarithmic time scale in Fig. 2共a兲 for a Fourier short time range 0.5 s ⬍ t* ⬍ 5 s, where the overlapping between
number of Foq = 10−4 and for copper nanoparticles in ethylene the DuPhlag and the Fourier solutions at almost all times is re-
glycol. The solution presented in the figure represents the tran- vealed. A slight variation between the DuPhlag and Fourier solu-
sient, and the time range does not include the steady state. The tions within the initial short time range is noticed. Both the
DuPhlag solution overlaps with the Fourier solution for long DuPhlag and Fourier results are almost accurately linear in time
times. The corresponding results zoomed within the short time on a logarithmic time scale, their slopes however varying at short
range 1 s ⬍ t* ⬍ 10 s are presented in Fig. 2共b兲 showing that times.
within this time range the DuPhlag solution produces results that The dimensionless wire temperature corresponding to the Fou-
col. Detail of the transient solution for the time range 0.5 s < t*
< 5 s.
rier as well as the DuPhlag solutions as a function of time is

presented on a logarithmic time scale in Fig. 5共a兲 for a Fourier
number of Foq = 10−4 and for carbon nanotubes in oil. The
DuPhlag solution overlaps with the Fourier solution at long times.
The corresponding results zoomed within the short time range
1 s ⬍ t* ⬍ 10 s are presented in Fig. 5共b兲. The results show that
within this time range the DuPhlag solution does not produce
results that are approximately linear in time on a logarithmic time
scale unless a smaller time range window is used, while the Fou-
rier solution is almost accurately linear in time on a logarithmic
time scale. Their corresponding slopes are substantially different.
Similar results that apply to carbon nanotubes in oil and a Fourier Fig. 5 The dimensionless wire temperature corresponding to
number of Foq = 10−5 are presented in Fig. 6 within the short time the Fourier as well as the DuPhlag solutions as a function of
range 0.5 s ⬍ t* ⬍ 5 s, where the overlapping between the DuPhlag
= 10−4 and for for carbon nanotubes suspended in oil. „a… tran-
and the Fourier solutions at longer times is evident. The figure sient solution for the time range 1 s < t* < 100 s; „b… detail within
shows that within this short time range both the DuPhlag as well the time range 1 s < t* < 10 s.
as the Fourier solutions produce results that are linear in time on a
logarithmic time scale, however different slopes are generally as-
sociated with each solution. The results applicable to carbon nano-
tubes in oil and a Fourier number of Foq = 10−6 zoomed within the beyond kapp / kact = 5 was evaluated for carbon nanotubes in oil,
short time range 0.5 s ⬍ t* ⬍ 5 s are presented in Fig. 7 where the while this ratio varies between kapp / kact = 1.05 and beyond
overlapping between the DuPhlag and the Fourier solutions at kapp / kact = 5 for copper nanoparticles suspended in ethylene glycol.
almost all times is revealed. A slight variation between the The corresponding effective thermal conductivity ratio results for
DuPhlag and Fourier solutions within the initial short time range a Fourier number of Foq = 10−5 is presented in Fig. 9. The figure
is noticed. Both the DuPhlag and Fourier results are accurately reveals a variation of the effective thermal conductivity ratio that
linear in time on a logarithmic time scale. The linearity of the wire may go beyond kapp / kact = 1.5 for times below 1 s for both copper
temperature with time on a logarithmic time scale that was at least nanoparticles suspended in ethylene glycol as well as for carbon
approximately revealed in most of the results presented above nanotubes in oil. Results for a very small Fourier number value of
cannot be used to distinguish a priori between experimental data Foq = 10−6 are presented in Fig. 10 where values of up to and
whether they belong to Fourier or DuPhlag unless moderate or beyond kapp / kact = 1.3 can be reached for times below 0.5 s.
large values of the Fourier number, i.e., Foq 艌 10−5, are consid- The fact that the effective thermal conductivity ratio varies sub-
ered. The latter may therefore confirm the validity of one of the stantially with time within the time range considered and used in
models as the correct one if the previously mentioned range of the experiments makes a definite resolution to the experimental
Fourier numbers applies. Otherwise there is no easy way to dis- puzzle inconclusive, unless the experimental wire temperature
tinguish between the two. versus time data become available for analysis. Nevertheless the
The results of the ratio between the “apparent” and “actual” results presented in this paper may provide a valid explanation.
effective thermal conductivities following Eq. 共60兲 and corre- The computed analytical results show that the apparent thermal
sponding to data of copper nanoparticles suspended in ethylene conductivity evaluated via the Fourier conduction constitutive re-
glycol compared with carbon nanotubes suspended in oil are pre- lationship in conjunction with the THW experimental method
sented in Fig. 8 for a Fourier number of Foq = 10−4. An effective could indeed produce results that show substantial apparent en-
thermal conductivity ratio that varies between kapp / kact = 1.4 and hancement of the effective thermal conductivity of the nanofluid
Fig. 8 The ratio between the “apparent” and “actual” effective
thermal conductivities following Eq. „60… and corresponding to
= 10−5 and for carbon nanotubes suspended in oil. Detail within
copper nanoparticles suspended in ethylene glycol compared
the time range 0.5 s < t* < 5 s.
with carbon nanotubes suspended in oil for a Fourier number
of Foq = 10−4
suspension if the actual conduction process is governed by a ther-

mal conduction process as derived in the present paper. A correc- 7 Conclusions
tion factor ␴ is to be used in the latter case to convert the THW
An attempt in explaining experimental results, which reveal an
data to values appropriate for suspensions. 共Note: the thermal con-
impressive heat transfer enhancement in nanofluid suspensions
ductivity obtained from the Fourier solution is constant, i.e., inde-
was derived based on the heat conduction mechanism for transient
pendent of time. However, the ratio between the derivations of the
processes in suspensions. The effect of the surface-area-to-volume
thermal conductivity via Fourier and DuPhlag methods varies in
ratio 共specific area兲 of the suspended nanoparticles on the heat
time.兲
transfer mechanism was shown to impact substantially on the so-
Future research should focus on the development of an accurate
lution and results. The results show that an apparent effective
estimate for the solid-fluid interface heat transfer coefficient, h, as
thermal conductivity enhancement of the same quantitative values
the particle size is reduced. The latter will allow an accurate and
as the experiments suggest can be obtained if the solid-fluid inter-
directly applicable correction to the THW data, which will make
face heat transfer mechanism is accounted for, leading to DuPhlag
the THW experimental method useful in its application to suspen-
conduction. The analysis provides a possible explanation that
sions.
settles the apparent conflict between the recent experimental re-
Fig. 7 The dimensionless wire temperature corresponding to Fig. 9 The ratio between the “apparent” and “actual” effective
the Fourier as well as the DuPhlag solutions as a function of thermal conductivities following Eq. „60… and corresponding to
time, on a logarithmic time scale, for a Fourier number of Foq copper nanoparticles suspended in ethylene glycol compared
= 10−6 and for for carbon nanotubes suspended in oil. Detail to carbon nanotubes suspended in oil for a Fourier number of
within the time range 0.5 s < t* < 5 s. Foq = 10−5
NAif Asf Asf Vtot 1 Asf asf
aif = = = = = 共A1兲
NV p Vs Vtot Vs ␧ Vtot ␧
where the definition of the solid fraction as ␧ = Vs / Vtot was intro-
duced leading to
asf = ␧aif 共A2兲
For spherical particles suspended in the fluid asf = 6␧ / d p and for
long cylindrical particles 共neglecting the effect of the top and
bottom surfaces on heat transfer兲 asf = 4␧ / d p. We can therefore
separate the direct effect of the specific heat transfer area on the
heat transfer coefficient h in fluid suspensions in the form
h = hsf asf = ␧hsf aif 共A3兲
where hsf = h / asf is the specific heat transfer coefficient carrying
units of W m−2 K−1. Now let us consider the heat flux, qif * in
W m−2, at the interface of one particle leading to the rate of heat-
transferred to/from that particle 共or the rate of internal energy
gained/lost per unit time兲 as Q̇if = qif Aif in W. For the whole sus-
pension the total rate of heat-transferred to/from the solid phase is
Fig. 10 The ratio between the “apparent” and “actual” effec-
Q̇tot = Nqif Aif and within an REV containing n solid particles per
tive thermal conductivities following Eq. „60… and correspond-
ing to copper nanoparticles suspended in ethylene glycol com- unit volume 共n = N / Vtot兲 the rate of heat-transferred to/from the
pared with carbon nanotubes suspended in oil for a Fourier solid phase per unit volume, Q̇sf = Q̇tot / Vtot is
number of Foq = 10−6
Q̇sf = nqif Aif = nV paif qif = ␧aif qif = asf qif 共A4兲
The representation used in Eqs. 共6兲 and 共7兲 for this rate of heat
transferred between the phases per unit volume of suspension was
sults in nanofluid suspensions and classical theories for estimating
the effective thermal conductivity of suspensions. Nevertheless, Q̇sf = h共T f * − Ts*兲 = hsf asf 共T f * − Ts*兲共A5兲
other possible mechanisms have to be accounted for and investi-
gated in more detail prior to reaching a final conclusion. In this Combining Eqs. 共A4兲 and 共A5兲 yields the definition of the heat
respect the present results do not provide a definite conclusion to transfer coefficient in the form
settle the puzzle.
Q̇sf qif
h= = asf 共A6兲
共T f * − Ts*兲共T f * − Ts*兲
Appendix A and the specific heat transfer coefficient
−3 −1
The parameter h, carrying units of W m K , represents an
Q̇sf qif
integral heat transfer coefficient for the heat conduction at the hsf = = 共A7兲
solid-fluid interface within an REV. Its general relationship to the 共T f * − Ts*兲asf 共T f * − Ts*兲
surface-area-to-volume ratio 共specific area兲 is derived in this ap- The general form of the relationships derived theoretically or em-
pendix. The author is not aware of any investigations attempting pirically for the evaluation of the heat transfer coefficient in po-
to evaluate a relationship for the heat transfer coefficient h in fluid rous media has the form
suspensions, however some relationships are available for the re-
spective coefficient in fluid saturated porous media 共Quintard and hd2p
Whitaker 关38兴, Alazmi and Vafai 关39兴, Amiri and Vafai 关40兴, Ni = = s共␧,k f /ks兲共A8兲
kf
Wakao, Kaguei, and Funazkri 关41兴, Wakao and Kaguei 关42兴, Ku-
wahara, Shirota, and Nakayama 关43兴兲. Most of the reported evalu- where Ni is Nield number 共Nield 关44兴兲 d p is a characteristic length
ations of h in porous media, such as Alazmi and Vafai 关39兴, Amiri of the solid phase 共particle size兲, and s共␧ , k f / ks兲 is a function of
and Vafai 关40兴, Wakao et al. 关41兴, Wakao and Kaguei 关42兴, Kuwa- the solid fraction ␧ and the fluid-to-solid thermal conductivity
hara, Shirota, and Nakayama 关43兴, were derived with a particular ratio k f / ks. The implication of Eq. 共A8兲 is that the heat transfer
focus on convection rather than conduction heat transfer. Their coefficient is related to the particle size by the inversely quadratic
applicability and accuracy for conduction are therefore question- relationship
able. Nevertheless, the common aspects of all these derivations or
kf
empirical relationships and their conversion to fluid suspensions h= s共␧,k f /ks兲共A9兲
can be summarized as follows. Consider Aif as the interface heat d2p
transfer area and V p the volume of one suspended particle. Then, While the particular form of the function s共␧ , k f / ks兲 and its pos-
one can define the specific area, aif in m−1, of one particle as sible further dependence on the particle size as well, especially as
the interface area per unit volume of solid particle in the form the particle size is reduced to nanoscale levels, is not established
aif = Aif / V p. Based on this definition the specific area of a sphere is for the case of suspensions, the general dependence of the heat
共aif 兲sphere = 6 / d p, where d p is the spherical particle’s diameter, and transfer coefficient on the particle size is evident. This dependence
the specific area of a cylinder excluding the top and bottom sur- of the heat transfer coefficient on the particle size introduces the
faces is 共aif 兲cylinder = 4 / d p, where d p is the cylindrical particle’s effect of the surface-area-to-volume ratio 共specific area兲 that was
diameter. For N particles in a suspension that has a total volume claimed by Eastman et al. 关1兴 to be missing in the classical models
Vtot the total interface heat transfer area is Asf = NAif , the total of evaluating the effective thermal conductivity of the suspension.
volume of the solid phase is Vs = NV p, and the specific area of all One should however bear in mind that further dependence on
particles per unit volume of suspension, asf = Asf / Vtot, can be ob- particle size is anticipated as the particle size is reduced to the
tained from the following derivation nanoscale level predominantly when the ballistic rather than dif-
fusive nature of heat transfer becomes dominant 共Chen 关20–22兴兲 Two Small Freely-Moving Spheres in a Linear Flow Field,” J. Fluid Mech.,
56, pp. 375–400.
and hence reducing the rate of heat transfer, implying a conse- 关8兴 Hamilton, R. L., and Crosser, O. K., 1962, “Thermal Conductivity of Hetero-
quential reduction of the value of h to somewhat compensate for geneous Two-Component Systems,” Ind. Eng. Chem. Fundam., 1, pp. 187–
the otherwise substantial increase of h as the particle size is re- 191.
duced. In addition one may anticipate an increase of h due to 关9兴 Jeffrey, D. J., 1973, “Conduction Through a Random Suspension of Spheres,”
Proc. R. Soc. London, Ser. A, 335, pp. 355–367.
Brownian motion induced nanoconvection 关19兴 or a decrease due 关10兴 Davis, R. H., 1986, “The Effective Thermal Conductivity of a Composite
to the exceptionally small interface thermal conductance 关45兴. The Material With Spherical Inclusions,” Int. J. Thermophys., 7, pp. 609–620.
latter opposing trends in the variation of the value of h as the 关11兴 Lu, S., and Lin, H., 1996, “Effective Conductivity of Composites Containing
particle size is reduced to the nanoscale level suggest much larger Aligned Spheroidal Inclusions of Finite Conductivity,” J. Appl. Phys., 79, pp.
6761–6769.
values for the time lags ␶q and ␶T, and consequently larger values 关12兴 Bonnecaze, R. T., and Brady, J. F., 1990, “A Method for Determining the
of the Fourier numbers, Foq and FoT. Effective Conductivity of Dispersions of Particles,” Proc. R. Soc. London, Ser.
A, 430, pp. 285–313.
关13兴 Bonnecaze, R. T., and Brady, J. F., 1991, “The Effective Conductivity of
Appendix B Random Suspensions of Spherical Particles,” Proc. R. Soc. London, Ser. A,
The condition for underdamped and critically-damped solutions 432, pp. 445–465.
关14兴 Keblinski, P., Phillpot, S. R., Choi, S. U. S., and Eastman, J. A., 2002,
to Eq. 共37兲 was identified in Eq. 共46兲 and is expressed in the form “Mechanisms of Heat Flow in Suspensions of Nano-Sized Particles 共Nanoflu-
ids兲,” Int. J. Heat Mass Transfer, 45, pp. 855–863.
c2fn 共1 + ␬2n FoT兲2 关15兴
= 艋1 共B1兲 Xue, L., Keblinski, P., Phillpot, S. R., Choi, S. U.-S., and Eastman, J. A., 2004,
4␻2n 4␬2n Foq “Effect of Fluid Layering at the Liquid-Solid Interface on Thermal Transport,”
Int. J. Heat Mass Transfer, 47, pp. 4277–4284.
where the equality part applies to the critically-damped solution, 关16兴 Prasher, R., Bhattacharya, P., and Phelan, P. E., 2005, “Thermal Conductivity
and the inequality to underdamped solutions. Inequality 共B1兲 can of Nanoscale Colloidal Solutions 共Nanofluids兲,” Phys. Rev. Lett., 94, p.
025901.
be expanded to yield 关17兴 Das, S. K., Putra, N., Thiesen, P., and Roetzel, W., 2003, “Temperature De-
pendence of Thermal Conductivity Enhancement for Nanofluids,” ASME J.
共FoT − 2 Foq兲 1
␬4n + 2 ␬2n + 艋0 共B2兲 Heat Transfer, 125, pp. 567–574.
FoT2 FoT2 关18兴 Kumar, D. H., Patel, H. E., Rajeev Kumar, V. R., Sundararajan, T., Pradeep, T.,
and Das, S. K., 2004, “Model for Heat Conduction in Nanofluids,” Phys. Rev.
which has the quadratic form Lett., 93, p. 144301.
关19兴 Jang, S. P., and Choi, S. U.-S., 2004, “Role of Brownian Motion in the En-
y ⬅ ␬4n + b␬2n + c 艋 0 共B3兲 hanced Thermal Conductivity of Nanofluids,” Appl. Phys. Lett., 84共21兲, pp.
4316–4318.
where b = 2共FoT − 2 Foq兲2 / FoT2 and c = 1 / FoT2 ⬎ 0. By treating ␬2n as 关20兴 Chen, G., 1996, “Nonlocal and Nonequilibrium Heat Conduction in the Vicin-
a continuous variable, the function y ⬅ ␬4n + b␬2n + c represents a ity of Nanoparticles,” ASME J. Heat Transfer, 118, pp. 539–545.
关21兴 Chen, G., 2000, “Particularities of Heat Conduction in Nanostructures,” J.
parabola, which has a minimum at ␬2n = −b / 2. Given the estab- Nanopart. Res., 2, pp. 199–204.
lished fact that c = 1 / FoT2 ⬎ 0, the relevant case consistent with real 关22兴 Chen, G., 2001, “Balistic-Diffusive Heat-Conduction Equations,” Phys. Rev.
␬n values is linked to b ⬍ 0 or explicitly to 共FoT − 2 Foq兲 ⬍ 0. For Lett., 86共11兲, pp. 2297–2300.
关23兴 Nield, D., and Bejan, A., 1999, Convection in Porous Media, 2nd ed.,
b ⬍ 0 an underdamped solution is in principle possible provided Springer, New York.
the roots of the quadratic equation y ⬅ ␬4n + b␬2n + c = 0 are real. The 关24兴 Vadasz, J. J., Govender, S., and Vadasz, P., 2005, “Heat Transfer Enhancement
latter implies that the following roots ␬2n1,2 = −b关1 ± 冑1 − 4c / b2兴 / 2 in Nanofluids Suspensions: Possible Mechanisms and Explanations,” Int. J.
Heat Mass Transfer, 48, pp. 2673–2683.
have to be real in order to obtain two real and positive, distinct or 关25兴 Hammerschmidt, U., and Sabuga, W., 2000, “Transient Hot Wire 共THW兲
identical values of ␬2n. For the latter to occur, the discriminant Method: Uncertainty Assessment,” Int. J. Thermophys., 21, pp. 1255–1278.
关26兴 De Groot, J. J., Kestin, J., and Sookiazian, H., 1974, “Instrument to Measure
共b2 − 4c兲 must be nonnegative leading to the Thermal Conductivity of Gases,” Physica 共Utrecht兲, 75, pp. 454–482.
4
FoT2
冋 1−
2 Foq
FoT
册 2
−
4
FoT2
艌0 共B4兲
关27兴
关28兴
Healy, J. J., de Groot, J. J., and Kestin, J., 1976, “The Theory of the Transient
Hot-Wire Method for Measuring Thermal Conductivity,” Physica, 82C, pp.
392–408.
Kestin, J., and Wakeham, W. A., 1978, “A Contribution to the Theory of the
Further expansion of the left-hand side of inequality 共B4兲 pro- Transient Hot-Wire Technique for Thermal Conductivity Measurements,”
duces the final condition for underdamped and critically-damped Physica A, 92, pp. 102–116.
关29兴 Assael, M. J., Dix, M., Gialou, K., Vozar, L., and Wakeham, W. A., 2002,
solutions in the form “Application of the Transient Hot-Wire Technique to the Measurement of the
FoT ␶T Thermal Conductivity of Solids,” Int. J. Thermophys., 23, 615–633.
= 艋1 共B5兲关30兴 Nagasaka, Y., and Nagashima, A., 1981, “Absolute Measurement of the Ther-
Foq ␶q mal Conductivity of Electrically Conducting Liquids by the Transient Hot-
Wire Method,” J. Phys. E, 14, pp. 1435–1440.
where the equality part is applicable to the critically-damped so- 关31兴 Assael, M. J., Chen, C.-F., Metaxa, I., and Wakeham, W. A., 2004, “Thermal
lution, while the inequality is the condition for underdamped so- Conductivity of Suspensions of Carbon Nanotubes in Water,” Int. J. Thermo-
lutions. phys., 25, 971–985.
关32兴 Tzou, D. Y., 1997, Macro-to-Microscale Heat Transfer The Lagging Behavior,
Taylor and Francis, Washington, D.C.
关33兴 Tzou, D. Y., 2001, “Temperature-Dependent Thermal Lagging in Ultrafast
References Laser Heating,” Int. J. Heat Mass Transfer, 44, pp. 1725–1734.
关1兴 Eastman, J. A., Choi, S. U. S., Li, S., Yu, W., and Thompson, L. J., 2001, 关34兴 Vadasz, P., 2005, “Explicit Conditions for Local Thermal Equilibrium in Po-
“Anomalously Increased Effective Thermal Conductivities of Ethylene Glycol- rous Media Conduction,” Transp. Porous Media, 59, pp. 341–355.
Based Nanofluids Containing Copper Nanoparticles” Appl. Phys. Lett., 78, 关35兴 Vadasz, P., 2005, “Absence of Oscillations and Resonance in Porous Media
pp. 718–720. Dual-Phase-Lagging Fourier Heat Conduction,” ASME J. Heat Transfer, 127,
关2兴 Lee, S., Choi, S. U.-S., Li, S., and Eastman, J. A., 1999, “Measuring Thermal pp. 307–314.
Conductivity of Fluids Containing Oxide Nanoparticles,” ASME J. Heat 关36兴 Vadasz, P., 2005, “Lack of Oscillations in Dual-Phase-Lagging Heat Conduc-
Transfer, 121, pp. 280–289. tion for a Porous Slab Subject to Imposed Heat Flux and Temperature,” Int. J.
关3兴 Choi, S. U. S., Zhang, Z. G., Yu, W., Lockwood, F. E., and Grulke, E. A., Heat Mass Transfer, 48共14兲, pp. 2822–2828.
2001, “Anomalous Thermal Conductivity Enhancement in Nanotube Suspen- 关37兴 Özisik, M. N., 1993, Heat Conduction, 2nd ed., Wiley, New York.
sions,” Appl. Phys. Lett., 79, pp. 2252–2254. 关38兴 Quintard, M., and Whitaker, S., 1995, “Local Thermal Equilibrium for Tran-
关4兴 Xuan, Y., and Li, Q., 2000, “Heat Transfer Enhancement of Nanofluids,” Int. J. sient Heat Conduction: Theory and Comparison With Numerical Experi-
Heat Mass Transfer, 21, pp. 58–64. ments,” Int. J. Heat Mass Transfer, 38, pp. 2779–2796.
关5兴 Maxwell, J. C., 1891, A Treatise on Electricity and Magnetism, 3rd ed., Clar- 关39兴 Alazmi, B., and Vafai, K., 2002, “Constant Wall Heat Flux Boundary Condi-
endon Press, 1954 reprint, Dover, New York, pp. 435–441. tions in Porous Media Under Local Thermal Nonequilibrium Conditions,” Int.
关6兴 Batchelor, G. K., 1972, “Sedimentation in a Dilute Dispersion Of Spheres,” J. J. Heat Mass Transfer, 45, pp. 3071–3087.
Fluid Mech., 52, pp. 45–268. 关40兴 Amiri, A., and Vafai, K., 1994, “Analysis of Dispersion Effects and Non-
关7兴 Batchelor, G. K., and Green, J. T., 1972, “The Hydrodynamic Interaction of Thermal Equilibrium, Non-Darcian, Variable Porosity Incompressible Flow
Through Porous Media,” Int. J. Heat Mass Transfer, 37, pp. 934–954. Model for Convection in Porous Media,” Int. J. Heat Mass Transfer, 44, pp.
关41兴 Wakao, N., Kaguei, S., and Funazkri, T., 1979, “Effect of Fluid Dispersion 1153–1159.
Coefficients on Particle-to-Fluid Heat Transfer Coefficients in Packed Beds,” 关44兴 Nield, D. A., 1998, “Effects of Local Thermal Nonequilibrium in Steady Con-
Chem. Eng. Sci., 34, pp. 325–336. vective Processes in a Saturated Porous Medium: Forced Convection in a
关42兴 Wakao, N., and Kaguei, S., 1982, “Effect of Fluid Dispersion Coefficients on Channel,” J. Porous Media, 1, pp. 181–186.
Particle-to-Fluid Heat Transfer Coefficients in Packed Beds,” Heat and Mass 关45兴 Huxtable, S. T., Cahill, D. G., Shenogin, S., Xue, L., Ozisik, R., Barone, P.,
Transfer in Packed Beds, Gordon and Breach, New York. Usrey, M., Strano, M. S., Siddons, G., Shim, M., and Keblinski, P., 2003,
关43兴 Kuwahara, F., Shirota, M., and Nakayama, A., 2001, “A Numerical Study of “Interfacial Heat Flow in Carbon Nanotube Suspensions,” Nat. Mater., 2, pp.
Interfacial Convective Heat Transfer Coefficient in Two-Energy Equation 731–734.
Study of Infrared CO2 Radiation
From Liquid-Fueled Combustor
Y. Levy The combustor that is a 30% flat (rectangular) model of an annular turbojet-engine
combustion chamber was studied. It gave a turbulent nonpremixed kerosene-air flame
M. Lev with equivalence ratios within the range of 0.15–0.75. Measurements of CO2 infrared
e-mail: aerysml@aerodyne.technion.ac.il radiation were performed on exhaust gases from a combustor. The infrared radiation
images were obtained by an IR camera equipped with a narrow bandpass filter that falls
V. Ovcharenko on the CO2 fundamental band. Temperature profiles were measured by the thermocouple
at the combustor outlet. The in-house program INFRAD was used to calculate infrared
Israel Institute of Technology, radiation from the CO2 component of the efflux gases from the combustor. The calcula-
Faculty of Aerospace Engineering, tion was performed, taking into account the effects of radiation emission and absorption
Haifa, 32000, Israel along an optical path. The calculations that allow prediction of “apparent” gas tempera-
tures are those expected to be read by the IR camera. The calculated results are com-
pared with experimental measurements. They are found to be in close agreement.
关DOI: 10.1115/1.2175153兴
Keywords: combustion, IR optical diagnostics, infrared radiation calculation
1 Introduction In the present work, the CO2 infrared emission from efflux
gases of a practical combustor, with a nonpremixed liquid-fueled
For a long time laboratory flames have been successfully inves-
flame, was investigated. The IR images of the efflux were ob-
tigated by studying their radiation emission 关1–5兴. The emission
tained by an IR camera. The camera was equipped with a narrow
diagnostics can reveal the flame structure, temperature, data on
bandpass filter, with transmittance band that lies within the fun-
the chemical composition, and the combustion processes 关5,6兴.
damental CO2 radiation band. No other molecular species radiate
Less has been published about the applications of these methods
within the filter transmittance band. Simulation of the IR radiation
to practical nonpremixed combustions employing liquid fuels.
was carried out, based on thermocouple measurements of tem-
During the last decades, commercial optical instruments were de-
peratures in the efflux gases. The radiation transfer equation was
veloped allowing radiation mapping and spectral imaging of intri-
solved numerically taking into account the emission and absorp-
cate combustion environments such as engine chambers and
tion of radiation along the lines-of-sight. Comparison of the simu-
rocket exhaust plumes 关5,7,8兴. On the other hand, comprehensive lated and measured results is needed to improve the reliability of
computer codes, simulating combustion in complicated nonuni- the diagnostic method. The objective of the present study is to
form nonisothermal conditions became available 关5,6兴. Based on develop a radiation diagnostic method, to enable fast evaluation of
CFD simulation, molecular radiation emission can be calculated the exhaust gases characteristics in full-scale turbojet combustors.
for comparison with experimental data.
The maximum intensity of combustion radiation falls on the
near-infrared 共NIR兲 spectral range. The main IR radiators of com-
bustion gases are CO2, H2O, CO, and carbon soot particles. Typi-
2 Combustor Design and Performance
cal combustion zones are transparent for IR radiation to various
extents, depending on the radiation wavelength. In many cases The combustor, used in the present study, is shown in Fig. 1. It
they are transparent enough to collect radiation not only from is a 30% flat sector model of reversed flow annular combustor of
layers close to the observed surface but also from the whole op- a small turbojet engine. This model was designed to study its
tical path. Difficulty arises from the interpretation of the radiation, internal flow and ignition processes. Three L-shaped vaporizers
accumulated from the whole optical path, along which the gas is 共out of 10 in the complete engine兲 are incorporated in the model,
usually nonuniform and nonisothermal. The problem of obtaining but in the present experiment only two, adjacent to the igniter,
the temperatures and species concentrations along an optical path were used. The air feed duct is connected to the combustor in a
from the radiation data is known as an “inverse problem” and it is reverse-flow arrangement. The liner contains three series of air
far from being solved in general. It seems that a solution of the inlets. One set of holes of 0.8 mm diameter in the liner front 共Fig.
problem in the infrared range depends on: 共a兲 developing multi- 1兲; two others with larger different diameters are drilled in the
wave imaging, that allows different transparency for different outer and inner liner walls to form the optimized graduated air
wavelengths, and 共b兲 developing an infrared emission tomography admission. The combustor has a rectangular exit of 210
technique, in which images at different viewing angles are ob- ⫻ 30 mm2 cross section. We define a Cartesian coordinate system
tained. However, the radiation data measured from outside the gas with its origin in exit plane, see Fig. 1 共A-A兲. The combustor and
flow reflects changes that can occur inside the flow. It can serve liner profiles, as well as the system of air inlets in the outer and
for monitoring the gas flow. The solution of a “direct problem” inner liner walls, shown in Fig. 1, are equivalent to those in a real
共that is finding the radiation output from gas thermodynamic data兲 engine.
in a specific case helps in understanding the results of monitoring. Nonpremixed kerosene/air combustion process takes place, and
can be observed, inside the model combustor through quartz or
sapphire windows. The fuel mass flow-rate was varied from 1 g / s
Contributed by the Heat Transfer Division of ASME for publication in the JOUR- to 3 g / s and the air mass airflow rate was fixed at a value of
NAL OFHEAT TRANSFER. Manuscript received July 19, 2005; final manuscript received
November 6, 2005. Review conducted by Walter W. Yuen. Paper presented at the
100 g / s. This corresponds to global equivalence ratios of 0.15,
ASME Turbo Expo 2005: Land, Sea and Air 共GT2005兲, June 6–9, 2005, Reno, 0.3, and 0.45. The Reynolds number Re⬃ 18, 000 for the model is
Nevada, USA. the same as for the real engine combustor.
Fig. 2 Measured temperature profiles along the Y-axis for dif-
ferent equivalence ratios, X = 10 mm, Z = 0 mm
namic flow pattern in the combustor and the heat release are also
nonuniform. Radiation calculation of nonisothermal gas requires
knowledge of gaseous temperature distribution in the combustion
chamber. To simulate the exhaust temperature field, temperature
profiles were measured by a thermocouple 共Pt/ Pt13% Rh thermo-
couple, with a junction of 1 mm in diameter兲. The most significant
error in temperature measurement by an unshielded thermocouple
originates from the heat loss through radiation from the thermo-
couple junction. The heat balance for the junction is described by
the equation:
Fig. 1 Combustor „all units in mm… h共Tg − Ttc兲 = ␧␴共T4tc − T4amb兲共3兲
where h is the convective heat transfer coefficient, ⑀ is the ther-
mocouple emissivity, and ␴ is the Stefan-Boltzmann constant, Tg
3 Chemical Composition of the Combustion Products
is the gas temperature, Ttc is the thermocouple temperature, and
Normally, the kerosene combustion products are carbon diox- Tamb is the ambient temperature. An emissivity value of 0.2 was
ide, water vapor, nitrogen, and oxygen. The reaction between hy- accepted for the platinum alloy thermocouple bead 关9兴. An esti-
drocarbon fuel and “theoretical air” 共molar concentration of 21% mation of the Nusselt number is required in order to obtain the
O2 and 79% N2兲 at stoichiometric condition can be written as convective heat transfer coefficient. It was estimated as Nu= 10,
CxHy + 共x + y/4兲O2 + 3.76共x + y/4兲N2 using an empirical dependence of Nu as a function of Re0.54
d Pr0.33.
The Reynolds number Red = 300 was estimated based on the ther-
= xCO2 + 共y/2兲H2O + 3.76共x + y/4兲N2 共1兲 mocouple junction diameter d and the flow velocity 共about
For combustion with excess air 共Ø ⬍ 1兲, evaluation of the ex- 70 m / s兲 at the combustor exit. The Prandtl number was assumed
haust gases’ composition should also consider the specific equiva- to be Pr= 0.7, typical for combustion gases. The discrepancy be-
lence ratio, Ø. Assuming the equivalent kerosene molecular for- tween a measured temperature and real gas temperature increases
mula to be C12H24 共x = 12, y = 24兲: when the gas temperature increases. It is negligible up to a tem-
perature of 1000 K and is about 180 K at a temperature of
ØC12H24 + 18O2 + 67.68N2 = 12 Ø CO2 + 12 Ø H2O 2000 K. These discrepancies were calculated for each temperature
+ 18共1 − Ø 兲O2 + 67.68N2 共2兲 and the experimentally obtained temperature values were cor-
rected.
the mole fraction of combustion products can be found 共Table 1兲. The time constant of the thermocouple is estimated by the equa-
tion:
4 Temperature Measurements in the Combustor Ex-
haust Gases ␶ = mc/hA 共4兲
The exhaust gases are not isothermal, mainly due to nonuni- where m is the mass of the thermocouple junction, c is the ther-
form fuel injection and discrete air inlets. The operational regime mocouple specific heat capacity, h is the convective heat transfer
used, in which two vaporizers operated within the combustor, coefficient, and A is the surface area of the thermocouple junction.
causes temperature asymmetry across the exhaust. The aerody- This gives ␶ = 0.5 s. This value is larger than the turbulence time
scale 共about 30 ms for flow velocities of about 70 m / s兲. Conse-
quently, only average values are detected. It can be noted here that
Table 1 Mole fraction of the combustion products in different the oscillation of the measured apparent temperature, obtained by
regimes
the thermocamera has a wavelength of 20 mm 共along the X-axis兲.
Equivalence ratios The temperature measurements were performed along two
lines: transverse 共Y-axis兲 and longitudinal 共X-axis兲 to the flow.
Combustion The transverse line was selected at the exit plane at the midheight
products 0.15 0.3 0.45
of the combustor exit. The measured transverse temperatures are
CO2 0.02 0.04 0.06 shown in Fig. 2 for the regimes with equivalence ratios of 0.15,
H 2O 0.02 0.04 0.06 0.3, and 0.45. The transverse profiles, normalized to the maximum
O2 0.18 0.14 0.11 temperature, differ in the various regimes by no more than 4%.
N2 0.78 0.78 0.77 The combustion is not symmetrical in the combustion chamber.
This is due to several reasons. As can be seen from Fig. 1, the fuel
Fig. 3 Measured temperature profiles along the X-axis for dif-
ferent equivalence ratios, Y = 65 mm, Z = 0 mm Fig. 4 Comparison of experimental and calculated apparent
blackbody temperature read by a PM390 camera with the
4.6 ␮m filter
is supplied by two neighboring injectors, which are located closer

blackbody continuum. The temperature read by the camera is an
to one side of the combustor than to the other side. Because of
this, the combustion zone is moved to one side of the combustor. apparent, or effective, temperature, Teff, which represents a total
The air is supplied through the symmetrical sequences of holes radiation over the range of spectral sensitivity. It is assumed that
that produce nonsymmetrical aerodynamics with respect to the the camera operates in a linear range of radiation intensity, and a
combustion zone. It is obvious that the energy losses are also temperature read by the camera depends only on the number of
nonsymmetrical in this combustion geometry. All the above fac- watts acting on the sensor, and not on the energy spectral distri-
tors lead to the appearance of a peak in the transverse temperature bution. These assumptions were checked by measurement of
profile, which is closer to one side of the combustor. blackbody temperatures, read by the camera, equipped with the
The measured longitudinal profiles for the same regimes are 4.6 ␮m filter, and by corresponding numerical simulation, using
shown in Fig. 3. The location of maximum temperature on the data of the filter transparency and the camera total response. The
Y-axis is the same for all regimes. The maximum gas temperature measured and calculated results are in a good agreement, as can
at the combustor exit plane can serve as a characteristic of a given be seen in Fig. 4.
regime. Characteristic temperatures of 910° C, 1500° C, and Carbon dioxide CO2 is the main radiating species in the com-
bustion products. In spite of significant amounts of water vapor,
1720° C were obtained for the studied regimes. To simulate the
its emittance in the considered range is much lower. However, the
complete temperature field in the X-Y plane, at the midheight of
radiation of H2O is present and should be taken into account.
the combustor exit 共Z = 0兲, it was accepted that any normalized
Because the equivalence ratio is small, it was assumed that the
transverse temperature profile is preserved in each X-coordinate. quantity of carbon monoxide 共CO兲 in the combustion products is
The value of the maximum temperature in a transverse profile is
negligible 关3,4兴. Also the emittance of carbon monoxide in the
assumed to be the measured value as presented in Fig. 3.
considered spectral range is significantly lower than the emittance
of CO2. The possible presence of carbon monoxide 共CO兲 in the
exhaust gases cannot essentially change the measured or calcu-
5 Experimental IR Radiation Study lated results.
Imaging of the combustor exhaust gases was performed by an Figure 5 shows examples of exhaust gas images performed by
infrared camera ThermaCAM® PM 390 关8兴. The PM 390 camera the IR camera equipped with the 4.6 ␮m filter. The temperature
is intended for general infrared inspections in a spectral range of values along the X-axis at the exhaust centerline 共at midheight of
3.4–5.0 ␮m and can be used for planar and local temperature the combustor exit兲, read by the IR camera, are presented in Fig.
measurements. Image acquisition rate of the thermocamera is 50– 6. The thermocamera image of IR radiation shows a two-
60 Hz. The filter used with the camera was a near-infrared inter- dimensional temperature map of the object according to its cali-
ference filter 共Oriel Instruments兲 with peak transmittance of about bration. Temperatures Teff along the centerline of the flow were
80% at 4.6 ␮m and bandwidth 0.2 ␮m. This filter was chosen read from images recorded during the experiment. There is a cer-
because its transmittance band is located within the spectral inter- tain dispersion of the temperatures at similar points of several
val of the fundamental vibrational-rotational emission band of the images, especially in the regimes with high equivalence ratios.
carbon dioxide molecule. IR spectral transmittance of the 4.6 ␮m The dispersion of the apparent temperature is believed to be asso-
filter was measured in the working range. The transmittance of the ciated with the flow turbulence, causing oscillations of the tem-
filter was exploited in calculations of IR radiation. The imaging perature and radiation. Therefore, the temperatures in each point
was performed from a distance of 2.0 m. The spatial resolution of were averaged over five images. These averaged temperatures
the images is about 2.2 mm/ pixel for a distance of 2 m.
The IR camera is calibrated by blackbody radiation and its
readings indicate temperature units. Hence a temperature indica-
tion corresponds to the blackbody radiation integrated over the
range of the camera sensitivity according to its spectral response.
The temperature read by the camera is the real temperature only if
the monitored surface radiates like a blackbody, or like a gray
body, of which the emissivity is constant, at least, within the range
of sensitivity. The emissivity value is installed through the camera
operational menu.
The gas, in principle, is not a blackbody-like radiator. Besides,
the transmittance band of the applied 4.6 ␮m filter is narrower
than the range of camera sensitivity. Therefore, the camera opera- Fig. 5 Planar CO2 radiation images of exhaust gases obtained
tion while using that filter is inconsistent with the calibration by a with a 4.6 ␮m filter for two equivalence ratios
Fig. 7 Experimental and calculated temperature profile along
Fig. 6 Measured Teff, averaged over five images „4.6 ␮m filter… the X-axis for different equivalence ratio, Z = 0 mm, 4.6 ␮m filter
The mole fractions are required for IR radiation calculation.

along the centerline of the flow images are shown in Fig. 6 with
Because they are unknown during the experiment, an assumption
standard deviations 共bars兲. It can be seen that the temperatures
is made that the measured temperatures are as of adiabatic flame
read by the thermocamera, along the line with maximum tempera-
temperatures. Then the equivalence ratios can be calculated, as-
tures 共Y = 65 mm兲, are much lower than those measured by the
suming the kerosene formula as C12H24. The pressure was taken at
thermocouple. This result was expected because the temperature P = 1 bar everywhere in the efflux flow. The mole fractions of the
read by the thermocamera does not correspond to any real tem- equilibrium combustion products were calculated according to the
perature in the flow. It is an apparent temperature corresponding to local temperature and equivalence ratio 关3兴. The temperature field
the IR radiation received by the thermocamera. The apparent tem-
in the plane X-Y was obtained from the thermocouple measuring
peratures are determined by the emitted radiative energy, collected
in compliance with the regime. The radiation at the centerline,
along optical paths from the hot gas volume, by transmittances of
observed from a combustor narrow side, was computed with re-
the gas, by camera optics and filter, and by camera sensitivity.
spect to the X-coordinate.
To compare the experimental apparent temperature 共obtained by
the thermocamera兲 with predictions from the INFRAD program,
6 Results and Discussion integration of the predicted spectral radiations was conducted over
a spectral interval of camera sensitivity. The transmission of the
The in-house INFRAD program was developed for the numeri- filter and the camera response were taken into account. These
cal study of the spectral and radiative properties of hot gases in integrals were compared with similar integrals of the blackbody at
the IR spectral range 关5兴. The program calculates the radiation different temperatures. We assume that the apparent temperature
transfer for nonuniform and nonisothermal molecular gases. It of the gas is that of a blackbody, which emits equal total radiation
uses a technique, known as a statistical narrow-band model. The 共taking into account the camera and filter characteristics兲.
radiative transfer in the narrow band model is formulated for an The calculation of IR radiation, observed through the 4.6 ␮m
array of many lines with centers confined to a narrow spectral filter, including the effect of a certain amount of carbon soot par-
interval ⌬␻ 共␻ is a wave number兲. The average radiance N̄ in a ticles, shows that the occurrence of the soot adds no more than 4%
band ⌬␻ for a nonuniform optical path segment from s = 0 to s to the integral radiation energy. The calculations show that radia-
= S is tion intensity of water vapor in this spectral range is about one
冕
S order of magnitude less than the radiation of CO2. Thus CO2
d¯␶共s兲 molecules contribute the dominant radiation passing through the
N̄ = − N␻0 共s兲 ds 共5兲
0
ds filter band.
Figure 7 shows averaging over five measurement temperature
The band ⌬␻ should be small enough that the Planck function data and simulated apparent temperatures along X, for the camera
N␻0 共s兲 can be considered as approximately constant across ⌬␻. equipped with the 4.6 ␮m filter. Excellent agreement can be seen
¯␶共s兲 is the average spectral transmittance for the ⌬␻: in the regime with the equivalence ratio 0.15 and some discrep-
冕
ancy is observed in the regimes with the equivalence ratio 0.3 and
1 0.45. The camera indicates temperature values higher than the
¯␶共s兲 = ␶␻共␻,s兲d␻ , 共6兲
⌬␻ ⌬␻
predicted temperatures based on thermocouple measurements.
Two sources of this discrepancy can be surmised. One is a wrong
where ␶␻共␻ , s兲 is the spectral transmittance within ⌬␻. The num- accounting for temperature losses by the thermocouple. This can
ber of lines in ⌬␻ is large, therefore ␶␻共␻ , s兲 is a rapidly fluctu- lead to the higher measured temperature and to the higher result in
ating function of ␻. the calculated radiation. Another obvious error is the assumption
An introduction and details of the problem can be found in the that the measured temperatures are the adiabatic flame tempera-
Handbook of Infrared Radiation 关2兴. Detailed analysis of the dif- tures. The adiabatic temperature based on the flow rate measure-
ferent band models and all mathematical details are given in 关10兴. ments, or global adiabatic temperature, can serve only as a calcu-
The approach was tested in 关11兴. The code INFRAD, is based on lated characteristic parameter, which reflects the overall
the mathematical analyses conducted in 关10兴 and was directly equivalence ratio and shows a certain mean temperature in the
tested in 关5兴. The spectral resolution, given by the code, is deter- flow. Due to nonperfect mixing and intense turbulence, special
mined by the available band model parameters data, and is 5 cm−1 variations exist in the equivalence ratio and therefore also in the
in wave numbers for the present work. The input to the program local temperatures. The thermocouple measurements at the exit
includes temperature, pressure, and species mole fraction. The plane reflect more or less the local adiabatic temperature in the
program can simulate the radiation emitted from a local surface combustor. This assumption is more accurate with the heat losses
area. while the flow passes from the combustion zone to the exit plane
Fig. 8 Calculated CO2 radiance spectra for different equiva- Fig. 9 Calculated radiation intensity at 4.6 ␮m depending on
lence ratios, X = 0 mm, Z = 0 mm the X-coordinate for different equivalence ratio, Z = 0 mm
are minimal. However, these losses are not negligible. By consid- duced by water vapor. The range of 3.5– 4.2 ␮m represents a
ering the thermocouple-measured temperatures as adiabatic tem- spectral window where no radiation of molecular species occurs.
peratures, an error is introduced in finding CO2 mole fractions in The radiation within the range of 4.2– 4.8 ␮m is due to the CO2
the exit plane. This error contributes to the discrepancy, which can fundamental band. From the wavelength of 4.7 ␮m the radiation
be seen between the predicted apparent temperatures and mea- of the water vapor spectral band begins. The vertical lines show
sured ones. If the real adiabatic temperatures are higher, the cal- the filter transmittance band, used for the IR measurements. This
culated mole fractions of CO2 are higher. This leads to an increase band is within the strong CO2 radiation band. However, its loca-
of the calculated radiation and, correspondingly, to higher pre- tion on the edge of the band ensures a weak absorption and allows
dicted apparent temperatures. This discrepancy increases in the us to obtain radiation from the whole optical path. The calculated
regimes with higher equivalence ratios because the heat losses are total transparency of the gas in the transverse direction is about
more significant in these regimes and the difference between the 0.8–0.9. Figure 9 shows the calculated radiance intensity at
local adiabatic temperature and the value measured by the ther- 4.6 ␮m depending on the X-coordinate for different equivalence
mocouple is larger. To estimate the effect of variation of CO2 ratio at Z = 0 mm. Unlike the effective temperature, the spectral
molar fraction on the evaluated temperature, it was assumed that radiance intensity is a characteristic independent of the ther-
perfect mixing takes place in the combustor, and species mole mocamera and characterizes a specific optical path. It can also be
fractions are uniform at the combustor exit plane and equal to the expressed in terms of temperatures, by recalculating it to the
average value 共obtained from the chemical balanced equation兲. brightness temperature 共temperature of a blackbody with equal
Due to this approach, in regions of high temperatures the evalu- radiation intensity in a specific wavelength兲.
ated CO2 mole fraction will be lower than the real values. The
discrepancy in the predicted apparent temperature is within the
range of 20– 40 deg. 7 Conclusions
The characteristics wavelength of temperature oscillations, as A diagnostic method for efflux gas monitoring was proposed
observed by the thermocamera, along the X-axis were about and tested in a liquid-fueled model of a jet engine’s combustor.
20 mm. The image acquisition rate of the thermocamera is 50 or The method is based on measuring infrared emission from com-
60 Hz. Thus the time response is less than 20 ms, and therefore bustor efflux flow.
the oscillations of the effective temperature are observed in the Measurements of temperature profiles at the combustor exit
images. In order to compare results of temperature values based plane revealed similar profiles even with different equivalence
on the thermocouple measurements with temperatures measured ratios and the consequent different maximum temperature values.
by the thermocamera, several recorded images were used to This fact allows that an apparent temperature can serve as a char-
smooth the oscillations by averaging the temperature at each acteristic temperature of the combustion gases. The method can be
point. It was found that there are, practically, no oscillations in the used for fast retrieval of the apparent temperature field in different
regimes with an equivalence ratio of 0.15; in terms of effective combustion systems. Upon calibration, variation from nominal
temperatures they are less than 2% of the apparent temperature values in the apparent temperature fields can be easily detected.
共90° C兲, i.e., about 2 ° C. The oscillations in the regimes with It should be noted that the use of several filters with an IR
equivalence ratios of 0.3–0.45 are about 7% that allows smooth- camera will increase the power of the gas flow monitoring by the
ing them using only five images. It is believed that the oscillations IR camera. Particularly, obtaining mole fractions of different spe-
are small enough and turbulence/radiation interactions do not af- cies in combustion products is possible.
fect the radiation values. In the following a possible explanation
of the oscillations in the efflux flow is given. In the regime with
the global equivalence ratio of 0.15, the whole fuel is burned out References
at the combustion zones, where only part of the air is supplied. In 关1兴 Edwards, D. K., 1960, “Absorption by Infrared Bands of Carbon Dioxide Gas
regimes with global equivalence ratios of 0.3–0.45, the air sup- at Elevated Pressures and Temperatures,” J. Opt. Soc. Am., 50, pp. 617–626.
plied to the combustion zones is not sufficient for complete com- 关2兴 Ludwig, C. B., Malkmus, W., Reardon, J. E., and Thomson, J. A. L., 1973,
bustion of all the fuel. An amount of the fuel passes to the dilution Handbook of Infrared Radiation from Combustion Gases, NASA SP-3080.
关3兴 Turns, S. R., 2000, An Introduction to Combustion, Concept and Applications,
zone. It is possible that a small amount of the unburned fuel exits 2nd ed., McGraw-Hill Higher Education, Boston.
the combustor and postcombustion occurs downstream of the exit 关4兴 Baukal, C. E., 2001, The John Zink Combustion Handbook, R. Schwartz, ed.,
plane; hence oscillations are observed in the IR radiation. CRC Press, London.
Figure 8 shows the calculated spectral radiation intensity emit- 关5兴 Avital, G., Cohen, Y., Gamss, L., Kanelbaum, Y., Macales, J., Trieman, B.,
Yaniv, S., Lev, M., Stricker, Y., Sternlieb, A., 2001, “Experimental and Com-
ted from the combustor efflux at X = 0 mm for different equiva- putational Study of Infrared Emission from Underexpanded Rocket Exhaust
lence ratios. The radiation within the range 3.0– 3.5 ␮m is pro- Plumes,” J. Thermophys. Heat Transfer, 15, pp. 377–383.
关6兴 Jun, J. I., Sivathanu, Y. R., and Gore, J. P., 2000, “Thermal Radiation Proper- 关9兴 Bradley, D., and Entwistle, A. G., 1961, “Determination of the Emissivity, for
ties of Turbulent Lean Premixed Methane Air Flames,” in Proceedings of the Total Radiation, of Small Diameter Platinum-10% Rhodium Wires in the Tem-
Combustion Institute, Vol. 28, pp. 391–398. perature Range 600– 1450° C,” Br. J. Appl. Phys., 12, pp. 708–711.
关7兴 Kirby, B. J., and Hanson, R. K., 2000, “Imaging of CO and CO2 Using Infra- 关10兴 Young, S. J., 1977, “Nonisothermal Band Model Theory,” J. Quant. Spectrosc.
red Planar Laser-Induced Fluorescence,” in Proceedings of the Combustion Radiat. Transf., 18, pp. 1–28.
Institute, Vol. 28, pp. 253–259. 关11兴 Young, S. J., 1977, “Evaluation of Nonisothermal Band Model for H2O,” J.
关8兴 http://www.x20.org/FLIR.htm, Infrared PM390 radiometer. Quant. Spectrosc. Radiat. Transf., 18, pp. 29–35.
Numerical and Experimental
Investigation of Thermal
Signatures of Buried Landmines
F. Moukalled
N. Ghaddar1
in Dry Soil
ASME Fellow
e-mail: farah@aub.edu.lb This paper reports a numerical and experimental investigation conducted to study the
thermal signature of buried landmines on soil surface. A finite-volume-based numerical
model was developed to solve the unsteady three-dimensional heat transport equation in
H. Kabbani dry homogeneous soil with a buried mine. Numerical predictions of soil thermal response
were validated by comparison with published analytical and numerical values in addition
N. Khalid to data obtained experimentally. Experiments were performed inside an environmental
chamber and soil temperatures were measured during cooling, using two measurement
Faculty of Engineering and Architecture,
techniques, after exposing the soil surface to a radiant heat flux for a specified period. In
American University of Beirut,
the first technique, the temporal variation of the surface and internal soil temperatures
Beirut, Lebanon
were recorded using thermocouples. In the second technique, the soil surface temperature
was measured using an infrared camera that revealed the thermal signature of the mine.
The transient temperature profiles generated numerically agreed with measurements, and
Z. Fawaz the difference between predicted and measured values was less than 0.3°C at both the soil
Ryerson University,
surface and in depth. The accurate matching of numerical and IR images at the surfaces
School Aerospace Engineering,
was found to strongly depend on the use of a smaller soil thermal conductivity at the
Toronto, Canada, M5B 2K3
surface than at greater depths. The numerical model was used to predict the dependence
of the peak thermal contrast on time, depth, and heating period. The thermographic
analysis, when combined with numerical predictions, holds promise as a method for
detecting shallowly buried land mines. 关DOI: 10.1115/1.2176681兴
Keywords: infrared imaging, detection of buried objects, soil thermal response to un-
steady surface heating, finite-volume method
1 Introduction ertheless, a few projects using IR technology aimed at searching

for individual mines. One such project involved developing a
Objects of dissimilar thermal capacitance absorb and release
short range IR system based on an 8 – 12 ␮m IR sensor and using
heat at different rates and as such have different infrared 共IR兲
neural networks for pattern recognition after segmentation of the
emission characteristics. Therefore, a buried mine heats up and
image. Reported results indicate that the system is capable of
cools down at a rate different from the surrounding soil. This
achieving a successful mine-detection rate of 90% 关6兴.
unequal heat transfer rate manifests itself on the surface above the
Besides accuracy, it is highly desirable for any potential land-
mine through either a higher or a lower soil surface temperature
mine detection technology to be quick and minimally intrusive. IR
than the surface temperature at a location away from the mine.
technology, as a means for remotely detecting buried landmines,
This variation in temperature is denoted in the literature by ther-
has the potential of combining these desirable features. Since the
mal signature 关1–3兴 and appears as a thermal contrast in an image
detection of mines using IR technology is via an image for the soil
taken by an IR camera. Simard 关2,3兴 thoroughly explained the
surface, its degree of success depends on a number of factors
various thermal mechanisms affecting the temperature contrast
including burial depth, soil attributes, moisture content, mine
and related them to the “volume” and “surface” effects. The vol-
properties, as well as the time of day during when the scanning is
ume effect results in an alteration of the heat flow due to the
performed. Bruschini and Gros 关7兴 assessed the potential of using
presence of the buried mine. The surface effect, which enhances
IR imaging as a landmine detection technology and described its
the mine’s signature, is present whenever the soil layer above and
advantages and limitations. They estimated that IR images could
around the mine is disturbed 共e.g., during the first few weeks after
spot the presence of an antitank mine at a maximum burial depth
burial兲. According to numerous studies, IR systems hold promise
of 10– 15 cm. Their work also indicated that the results of IR
as a support technology for specific mine-detection situations 关4兴.
imaging depend heavily on environmental conditions. In a recent
The IR technology, which has the key advantage of being passive,
paper, Deans et al. 关8兴 presented experimental results of a thermal
could be remotely utilized by aerial search, and could cover a
imaging method using microwave heating to enhance the visibility
large area in a relatively short time 关4兴. Infrared thermography is
of buried landmines. They concluded that discovery of mines bur-
best suited for identifying minefields 共global area search兲, rather
ied in dry sand is possible up to a burial depth of 2 cm, which
than searching for individual mines 共local area search兲关5兴. Nev-
differs significantly from the 10– 15 cm suggested by Bruschini
and Gros 关7兴. Consequently, the depth at which mines are detect-
1
Corresponding author. able by an IR sensor strongly depends on the type of mines and
Contributed by the Heat Transfer Division of ASME for publication in the JOUR- the type of soil in which they are buried.
NAL OF HEAT TRANSFER. Manuscript received April 27, 2005; final manuscript re-
Hermann and Ian 关9兴 investigated experimentally the enhance-
ceived October 25, 2005. Review conducted by Ramendra P. Roy. Paper presented at
the 2005 ASME Heat Transfer Summer Conference 共HT2005兲, July 15–22, 2005, ment of thermal signature of buried landmines using IR sensors
San Francisco, California, USA. through volumetric heating via microwave energy sources. In ad-
dition, the authors presented a one-dimensional model for micro-
wave absorption and heat dissipation by moisture-laden soils with
a surrogate buried landmine to model the enhancement of the
signature. Mitchell et al. 关10兴 studied the detection of antiperson-
nel landmines by taking thermal images of the soil surface after
spraying it by hot water with an array of jets. The variation in
conduction and radiation heat transfer due to the water blocked by
the mine and the heating of the mine itself, affected the transient
temperature profiles at the soil surface and reflected the presence
of the buried mine. Moreover, to reduce the false alarm mitigation
共FAM兲 three quantitative numbers based on gray contrast, tem-
perature gradient, and apparent thermal width were used as indi-
cators. Deepak et al. 关11兴 applied other FAM reduction techniques
using three indicators based on circularity, gray scale moment,
and reflection symmetry. The study reported reduction in false Fig. 1 The soil bed and the embedded mine
alarms by factors ranging from four to twelve depending on the
mine size.
Recently, Khanafer and Vafai 关12兴 analyzed numerically using
the finite-element method effect of surface roughness on thermal
conduction equation governs the transport of heat in both soil and
signature of buried landmines over a diurnal cycle. In their work,
mine. Moreover, the target objects buried in it are modeled as
an antitank mine was simulated taking into consideration the ef-
isotropic solids. Assuming negligible temporal variation of the
fect of its striker mechanism. Khanafer et al. also studied the
moisture content, the resulting governing equations for the soil
effect on the soil surface temperature of the mine’s outer metal
and the mine can be written as
case and its top air gap 关13兴. Their results show that the thin metal
outer case and the top air space of a buried antipersonnel mine ⳵
have a noticeable effect on the intensity of the landmine signature Soil: 共␳cT兲s = ⵜ共k ⵜ T兲s 共1兲
⳵t
over a diurnal cycle.
Lopez et al. 关1,14兴 investigated landmine detection numerically ⳵
and experimentally using, respectively, the finite difference Mine: 共␳cT兲m = ⵜ共k ⵜ T兲m 共2兲
method and infrared technology. Their work 关14兴 consisted of ⳵t
comparing thermal signatures in bare soil generated using a three- Where ␳, c, k, and T are the density, specific heat, thermal con-
dimensional thermal model against signatures obtained from in- ductivity, and temperature, respectively. The subscripts s and m
frared images in order to infer the presence of buried objects. refer to soil and mine, respectively. The boundary conditions ap-
Then, Lopez et al. 关14兴 classified these detected objects into either plicable at the boundaries of the physical domain are given by
mine or no-mine categories by solving an inverse heat transfer
problem. The analysis of the time evolution of perturbations cre- At the soil surface: n · ks ⵜ Ts = qnet 共3a兲
ated by such objects constituted the basis of the classification
procedure. Moreover, Sendur and Baertlein 关15兴 conducted a simi- At the mine surface: n · ks ⵜ Ts = n · km ⵜ Tm 共3b兲
lar study to illustrate the effect of the buried landmines on the
surface temperature distribution. Results reported 关14,15兴 good At vertical boundaries: n · ks ⵜ Ts = 0 共3c兲
agreement in the measured and predicted thermal contrast at the
surface of the soil. However, the difference between the predicted At the deep soil below the mine: Ts → T⬁ 共3d兲
and measured absolute temperatures at the surface 关14,15兴 and in where n represents the normal unit vector, T⬁ the deep soil tem-
the soil 关14兴 were higher 共of ±1 ° C to ±2 ° C兲 showing lower perature below the buried mine, and qnet the net heat flux into the
matching levels. top surface of the soil given by
From the above, it seems that combining numerical techniques
with infrared technology holds a promise with regard to develop- qnet = qconv + qrhs − qemis 共4兲
ing a powerful tool for the detection of buried landmines. To this where qconv is the convective heat transfer between the surface of
end, the aim of this investigation is to develop a three-dimensional
the soil and the ambient air, qrhs is the incident radiation heat flux
thermal model using a control volume approach to better under-
on soil surface, qemisis the gray body emission from the soil’s
stand the unsteady heat transport in soil with buried landmines
surface. In this work, the soil is assumed dry and therefore the
and to predict accurately the soil surface temperature and surface
latent cooling of the ground caused by evapotranspiration and
thermal contrast. The model is validated by comparing generated
condensation is neglected. Convective heat transfer between the
results against available theoretical/numerical values and against
soil and the surrounding air is expressed as
measurements taken at controlled ambient conditions in an envi-
ronmental chamber using both an IR video camera of high thermal qconv = Ah共T⬁ − Ts兲共5兲
resolution and thermocouples embedded in the soil. The temporal
In Eq. 共5兲, h is the convective heat transfer coefficient at the soil
variation of the soil surface temperature is generated by subjecting
the soil surface to radiant heating for a specified time interval surface and A the surface area. Initially, the soil, ambient air, and
followed by convection cooling. The formulation and validation mine are considered to be at the same uniform temperature T⬁.
of the numerical model represent the first step toward the future Moreover, assuming the ground surface to behave as a grey body,
development of an IR-based tool for detecting buried landmines the long-wave radiation between the soil surface and the room
using both the expected absolute surface temperature and the surfaces can be expressed as
strength of the thermal contrast as parameters in the mine identi- qemis = ␴␧共T4s − T4⬁兲共6兲
fication technique.
where ␴ = 5.67* 10 W / m K is the Stefan-Boltzmann constant,
−8 2 4
␧ is the mean emissivity of the room surfaces, and Ts is the soil’s

2 Soil Thermal Model Formulation surface temperature. The temperatures of the surfaces surrounding
A schematic of the physical situation and heat exchange pro- the soil-mine configuration in the environmental chamber are as-
cesses is depicted in Fig. 1. The three-dimensional transient heat sumed the same as that of the room air temperature.
共− k ⵜ T兲 f · S f = − k f 冉 TF − T P
d PF
E f + 共ⵜT兲 f · T f 冊共11兲
where E f is the magnitude of E f , and the cross-diffusion term is

treated as a source term in a deferred correction manner, with its
value computed explicitly using the current T field. By substitut-
ing Eqs. 共9兲 and 共11兲 in Eq. 共8兲, the following algebraic equation
is obtained:
Fig. 2 A schematic showing the treatment of the diffusion
term
共atP + aDP 兲T P + 兺
F=NB共 P兲
aD t 0
F T F = a PT P 共12兲
where the superscripts t and D indicate coefficients obtained from

3 Numerical Analysis the discretization of the transient term and of the diffusion fluxes,
respectively, and the superscript 0 designates a value from the
3.1 The Discretization Process. The discretization of the previous time step.
heat conduction equation is performed following the finite-volume To improve the robustness of the solution procedure, which is
method. In this approach, the solution domain is divided into a set iterative in nature, the equations are usually under relaxed. Denot-
of discrete and nonoverlapping elements or control volumes, each ing the under relaxation factor by urf, Eq. 共12兲 becomes
冉冊冉冊
associated with a grid point located at its geometric center. The
discretization process is a two-step procedure. In step 1, the equa- atP + aDP
兺
1 − urf
tions are integrated over a control volume to obtain a discretized TP + aD t 0
F T F = a PT P + 共atP + aDP 兲T共Pn兲
urf F=NB共 P兲 urf
description of the conservation laws. In step 2, an interpolation
profile is used to relate some of the discretized terms from step 1 共13兲
to the discrete nodes in the solution domain. where the superscript 共n兲 refers to values taken from the previous
To perform step 1, Eq. 共1兲 or 共2兲共with the subscript s or m
iteration. This equation can be rewritten as
dropped兲 is integrated over a control volume with the flux com-
ponents transformed into surface integrals using Green’s theorem
关16兴. This procedure yields
a PT P + 兺
F=NB共 P兲
共aFTF兲 = b P 共14兲
冕 ⳵ 共␳cT兲
⳵t
dV = 冖共k ⵜ T兲dS 共7兲
with the coefficients given by
兺
V ⳵S Ef
aF = − k f a P = atP − aF
The discrete form is obtained by replacing the surface integrals d PF F=NB共 P兲
冉冊
of the flux terms by discrete summations over the faces of the
兺
control volume, and the volume integral by the product of the 1 − urf
integrand at the cell center and the cell volume. Upon substitution b P = atPT0P + 共atP + aDP 兲T共Pn兲 + 共k ⵜ T兲 f · T f
urf f=nb共 P兲
into Eq. 共7兲, the resulting discrete equation is written as
共15兲
⳵ 共␳cT兲
⳵t
VP − 兺 ⌫ ⵜ Tf · Sf = 0 共8兲 The system of algebraic equations is solved iteratively using a
f=nb共 P兲 line-by-line tri-diagonal matrix algorithm 关16兴. Moreover, grid
where S f represents the surface area of the control volume face f networks are generated using the transfinite interpolation tech-
and nb共P兲 the faces of cell P. nique 关17兴. Furthermore, since a conservative scheme is used,
In step 2, Eq. 共8兲 is transformed into an algebraic equation with arranging the control volume face to coincide with the mine in-
the time derivative approximated using an Euler-implicit formula- terface ensures energy balance at the mine-soil surface of contact
tion to yield 关16兴 and forces Eq. 共3b兲 to be satisfied identically.
⳵ 共␳cT兲共␳cT兲 P − 共␳cT兲oP 3.2 Validation of the Thermal Model. The above-described
VP = VP 共9兲 numerical procedure was implemented in a finite volume code and
⳵t ⌬t validated by solving several problems of which two tests are pre-
Moreover, the diffusion flux of T through the control volume face sented next.
f can be written as
3.2.1 Test 1: Temperature Distribution in a Semi-Infinite Wall.
共− k ⵜ T兲 f · S f = 共− k ⵜ T兲 f · E f + 共− k ⵜ T兲 f · T f 共10兲 A soil column of depth 2 m represents the semi-infinite wall, with
where the two vectors E and T satisfy the relation S = E + T, with the soil having a density of 2000 kg/ m3, a thermal conductivity of
E being collinear with d PF, the vector joining the grid points P 2.511 W / m K, and a heat capacity of 837.2 J / kg K as used by
Khanafer and Vafai 关12兴. Initially, the soil is maintained at a uni-
and F straddling the control volume face 共Fig. 2兲. The first term
on the right-hand side of Eq. 共10兲 represents a contribution similar form temperature of value 293 K. At time t ⬎ 0, the surface tem-
perature of the soil is increased to 310 K and the temporal evolu-
to the one obtained on orthogonal grids 共i.e., involving TF and
tion of temperature profiles in the soil are predicted using the
T P兲, while the second term on the right-hand side is called cross
developed numerical code. Predicted profiles after 1, 4, 9, 16, and
diffusion or nonorthogonal diffusion and is due to the nonorthogo-
25 h are compared in Fig. 3共a兲 against the exact analytical pro-
nality of the grid system. Several options are available for the
files, assuming that heat flows in the vertical direction only. As
decomposition of S. The over-relaxed approach, in which T is
shown, the numerical code reproduces accurately the analytical
selected to be perpendicular to S 共i.e., E = S / cos ␪, ␪ being defined
profiles, with numerical predictions falling right on top of analyti-
in Fig. 2 and E and S are the magnitudes of E and S, respectively兲,
cal values. This is an indication of the correct implementation of
is found to be the most stable even when using highly distorted
the unsteady and conduction terms in the governing equation.
grids. In this approach, the importance of the term involving T P
and TF increases with increasing grid nonorthogonality. For an 3.2.2 Test 2: Reproduction of Khanafer and Vafai [16]
orthogonal grid, all variations yield the same numerical discreti- predictions. As a further check for accuracy, soil temperature pro-
zation. Expanding Eq. 共10兲 using a fully implicit scheme in time, files in the presence of a buried landmine reported by Khanafer
the expression for the diffusion flux becomes and Vafai 关12兴 following a finite-element approach were repro-
time step of 30 s is used in this work, while Khanafer and Vafai
关12兴 used a variable time step with a minimum value of 0.36 s.
4 Experimental Methodology
As mentioned earlier, the current work is the first foundation in
a large-scale project aiming at developing an IR-based system for
the detection of buried landmines. The objective of this phase is to
develop the numerical model that will be used in generating a
database against which infrared images will be compared for in-
ferring the presence or non-presence of landmines. Therefore, it is
essential for this tool to be capable of generating accurate predic-
tions. Furthermore, to fine tune the model, it is necessary to com-
pare numerical predictions against experimental data. Recording
these measurements in outdoor conditions requires accurate
weather and solar radiation data. To reduce the complexity of the
problem and validate the numerical tool, indoor experiments were
conducted at controlled temperature and humidity conditions in-
side an environmental chamber. The soil surface was exposed to
radiant heat flux for a short period of time and the thermal re-
sponse of the soil surface as well as the inner shallow layers of the
soil were monitored while cooling.
The environmental chamber has the dimensions of 3.2 m
⫻ 4.2 m ⫻ 2.8 m and is equipped with two independent systems
for controlling the air temperature and relative humidity. The pre-
cision in the set conditions of the environmental chamber is
±0.5° C for temperature and ±2% for relative humidity. The test
bed 共width: 1 m, length: 1 m, and height: 0.65 m, see Fig. 4兲
placed inside the chamber is made of a 2 cm thick layer of ply-
wood and is filled to the top with a homogeneous soil that is
typically found in Lebanon and is composed of clay silty sand:
80% sand 共fine兲, 12% silt, and 8% clay. The bed size was chosen
based on preliminary numerical calculations to make sure that the
boundary conditions along the sides of the bed do not affect the
solution around the mine. The soil was sifted and cleaned from
clutter and pebbles and its surface was carefully leveled. The mea-
sured soil density and specific heat were found to be ␳s
= 1430 kg/ m3, and Cs = 750 J / kg K, respectively. The soil thermal
conductivity, measured using a Hilton Thermal Conductivity Unit
TCU 100 of Heat Technology that does the analysis according to
ISO 8301 关18兴, was found to be ks = 0.30± 0.003 W / m K. On the
other hand, at the porous surface layer, the soil thermal conduc-
tivity was measured to be ks = 0.25± 0.003 W / m K. These mea-
surements were performed on soil samples that were initially
dried in a furnace.
Figure 4 shows the physical dimensions of the soil bed, the
embedded mine, and the radiation source. The soil surface is ex-
posed to radiant heat flux for a short period and then thermal
Fig. 3 A plot showing „a… comparison of soil temperature pro- signatures are observed while cooling takes place. The intention
files generated numerically and analytically for the semi-infinite of the experiment is to understand the mechanism by which the
heat conduction problem, „b… comparison of predicted soil tem- thermal signature is induced and to develop and validate a numeri-
perature profiles in the presence of a buried mine against those
cal model capable of accurately predicting the mine’s signature
reported by Khanafer and Vafai †12‡
rather than simulate the actual diurnal heating and cooling cycle.
The radiant heating source is composed of a set of six halogen
lamps rated at 1000 W each 共1 m ⫻ 1 m兲 placed horizontally
above the soil bed at a height of 1 m and backed with parabolic
duced. The simulated mine radius and height in the study were reflectors. The power input to the lamp is monitored for stable
0.125 m and 0.0833 m, respectively. The properties used in their nonfluctuating operation. The arrangement of the radiant lamps
work were: Trinitrotoluene 共TNT兲 thermal conductivity resulted in a nonuniform incident radiation on the soil surface. A
= 0.22344 W / m K, soil thermal conductivity= 2.5 W / m K, soil special CM3 Campbell Scientific Pyranometer of spectral range
density= 2000 kg/ m3, and soil heat capacity= 837 J / kg K. Com- 305– 2800 nm is used to measure the radiative heat flux from the
putations were repeatedly performed while refining the mesh and Halogen lamps at different locations on the soil surface using a
varying the time step until a grid and time step independent solu- grid of 0.05 m ⫻ 0.05 m resulting in 400 readings. The spectral
tion was reached. Predicted temperature profiles at two depths of selectivity of the pyranometer is ±5% 共350– 1500 nm兲, with a
0.076 m and 0.1193 m are compared against similar results re- sensitivity of 10 ␮V / Wm−2 to 35 ␮V / Wm−2 and impedance of
ported in 关12兴. As depicted, predictions are in good agreement 60 ⍀ to 200 ⍀.The measured reflected flux from the soil surface
with each other. The slight difference in values is attributed to amounted to about 20% of the incident flux. The radiant heat flux
some error introduced while extracting data from small figures for at the surface ranged from 2052 W / m2 to 2187 W / m2 in the
comparison and to the different mesh and time step used. A fixed middle of the test bed region of area 0.3 m ⫻ 0.3 m. The unifor-
Two measurement methods are used in this work. The main
measurement instrument is the IR ThermaCAM S60 camera
共7.5– 13 ␮m spectral range兲, which has a temperature sensitivity
of 0.08°C at 30°C and accuracy of ±2% of reading. The camera is
air-cooled and tower mounted horizontally at a height of 0.6– 1 m
from the soil surface. A tripod mounted IR camera, at a lens-to-
soil distance of 1 m angled at 45° from the horizontal, is utilized
to capture images of the soil surface at the peak, when a signifi-
cant variation in temperature is noted. The capture card employing
the ThermaCam Professional 2.8 关19兴 software stores images
共joint photographic experts group 共JPEG兲 and Microsoft bitmap
format 共bmp兲兲 and videos taken by the camera. The surrounding
air temperature 共level temperature兲 and the distance between the
lens and the object are adjusted in the camera and autozoom is
used. A major challenge in the work is processing the acquired
images with a high degree of accuracy. Because IR sensors are
passive sensors that quantify emitted infrared radiation from a
body surface, the radiation measured by the camera is not only
dependent on the temperature of the body, but it is also a function
of the emissivity of the surface. This is why the emissivity setting
of the camera significantly affects the accuracy of the measured
temperature differences. To minimize error, the soil emissivity
was carefully measured following two different procedures and
was found to be equal to 0.92. Another input parameter was the
reflected temperature, which was equal to the ambient temperature
in an outdoor environment and takes into consideration the radia-
tion from the surroundings reflected via the object surface to the
camera. When reflected temperature input to the camera in-
creased, the absolute temperatures decreased, but the difference in
temperature between any points on the surface remained un-
changed. The surrounding objects in the climatic chamber were all
set at the ambient temperature with the exception of the halogen
lamps panel. Since the solar panel is highly reflective and has a
very low emissivity, the camera reflected temperature was put
equal to the chamber air temperature. The camera resolution is
320⫻ 240 pixels capable of registering data at 76 800 pixel loca-
tions on the surface.
The second technique involves using K-type thermocouples to
measure soil temperature with an accuracy of ±0.5° C. Thermo-
Fig. 4 A schematic of „a… the side view of the experimental couples, mounted on a glass rod at the reference location 共x
setup and „b… the top view of the soil bed = 0.675 m, y = 0.505 m兲 were positioned at depths of 0 mm,
4.5 mm, 10.5 mm, 18 mm, and 45 mm. The thermocouples tips
were not bonded to the rod, but were stuck out at a distance of at
mity of the radiation intensity at the soil surface is dependent on least 10 mm from the rod surface into the desired soil locations to
the height of the source lamps above the bed. A height of 1 m was eliminate the effect of the glass rod conductivity on the readings.
found to give reasonable spatial uniformity on the surface above Additional thermocouples were located at the center of the mine’s
the buried mine and sufficient intensity of the radiant flux to gen- top surface and the soil’s surface above the mine center. The soil
erate realistic temperature changes on the surface and in the soil surface temperature was monitored at the sensors reference loca-
layer above the mine. Surface drying of the soil was not a problem tion using a special temperature probe Model 108 Campbell Sci-
since the thermal signature was observed during the cooling pe- entific with a resolution of ±0.3° C over the range −3 ° C to 90° C.
riod when surface temperatures of the soil was less than 35°C. In In addition, a precision infrared temperature sensor 共IRTS-P兲 is
addition, the variation of the relative humidity in the chamber was positioned 0.5 m above the bed corner directed toward the soil
monitored to remain at 50% ± 2%. The continuous renewal of the surfaces above the center of the buried mine. The IRTS-P sensor
air during cooling in the chamber maintained steady temperature has a silicon lens whose field of view is a circle of 1 m diameter
above the soil bed. when the sensor is 3 m away from the target. To accurately pro-
The mine is embedded in the soil at a distance x = 0.5 m and cess infrared images, the air temperature and relative humidity
y = 0.505 m from the origin of the bed, located on the lower left- above the soil bed were monitored. All thermocouples were cali-
hand corner. This location is selected so that the surface of soil brated before the start of the tests and were connected to a Camp-
above the mine is well within the view of the IR camera and is bell Scientific CR23X logger capable of recording one set of read-
exposed to spatially uniform radiative flux during the heating pe- ings per second. Sampling of the temperature readings was done
riod. A common type of antipersonnel mine of rectangular shape every 10 s. A small fan was placed parallel to the heater and away
共13.9 cm⫻ 5.4 cm⫻ 6.7 cm兲 is used. The mine is filled with from the soil bed, at the heating panel level of 1 m, to accelerate
共RTV兲 material that has properties similar to TNT 共specific heat cooling by increasing convection heat transfer. The air speed at a
cm = 1500 J / kg K, thermal conductivity km = 0.2 W / m K, and den- height of 1 m was measured using a handheld airflow meter and
sity ␳m = 1170 kg/ m3兲 and is buried in the soil. The plastic cover was found to be equal to 4.5 m / s ± 0.1 m / s.
of the mine has a thickness of 0.3 cm, specific heat c p The soil bed was conditioned inside the environmental chamber
= 1260 J / kg K, thermal conductivity k p = 0.5 W / m K, and density at an air temperature of 25° C and relative humidity of 50% for at
␳ p = 1760 kg/ m3. The burial depth is an experimental parameter least 48 h for steady state to be reached. Following this condition-
and is varied from 0.5 cm to 2.5 cm. ing period, the soil surface was heated for five minutes, after
Fig. 5 The computational domain
which the heat source was stopped and images of the soil surface
were taken at one-minute intervals to monitor the surface tem-
perature changes during the cooling period. The tests were per-
formed with the fan either on or off to study both forced and
natural convection cooling, respectively. The experiments were
carried out with a mine buried at predetermined depths to examine
the surface signature changes during cooling using the IR camera
and while recording temperature readings at the surface above the
mine and at the reference point location and in depth. The air
temperature in the environmental chamber was monitored at four
different locations.
For natural cooling, the air temperature above the soil increases
by about 5 ° C from its initial value 共the heating period is 5 min兲,
but cools quickly to the chamber temperature within the first five
minutes after shutting off the heat source. When repeating the test
Fig. 6 The variation in time of the predicted and measured
with the fan turned on, the temperature of the air above the soil temperatures at the center of the mine surface at z = 0.01 m, and
increases by 2.5° C only for the same heating period. The interval the soil surface at the reference point at depths of 0.45 cm,
between any two experiments is at least 48 hours to allow the soil 1.05 cm, 1.8 cm, and 4.5 cm for a test conducted in the environ-
and the environmental chamber air to reach thermal equilibrium mental chamber while heating the soil bed for a period of 5 min
as this condition is important for the numerical simulation. and then cooling by „a… radiation and forced convection with
the fan turned on and „b… radiation and natural convection with
5 Result and Discussions the fan turned off
5.1 Experimental Results and Validation of the Numerical

Thermal Model. The use of a three-dimensional conduction
model was dictated by the non-uniform heating of the soil surface terial buried at depths of z = 0.5 cm, 0.75 cm, and 1 cm from the
in the experiments, the geometry of the mine, and the willingness surface at x = 0.5 m and y = 0.505 m. The reference point at which
to validate such as model as it will be used for developing an temperature values arerecorded corresponds to the location x
IR-based tool for detecting buried landmines. = 0.675 m and y = 0.505 m. The environmental chamber is set to
Before comparing numerical results to measured values, the 25° C and 50% relative humidity.
grid and time step were adjusted until a grid and time step inde- Figure 6 shows the temporal variation of the predicted and mea-
pendent solution was obtained. This was accomplished with an sured temperatures at the center of the top surface of the mine
80⫻ 80⫻ 54 grid network with the grid points nonuniformly dis- 共z = 1 cm兲, the soil surface above the center of the mine, and the
tributed as shown in Fig. 5 using a time step of 10 s. All results reference point at depths of 0.45 cm, 1.05 cm, 1.8 cm, and 4.5 cm
presented here were generated using the aforementioned grid sys- for two test cases. Heating of the soil surface in both configura-
tem and time step for the physical domain described in the experi- tions took place over a period of 5 m. The difference between the
mental section and using the measured thermal properties of the two situations is in the cooling method, which occurred either
soil, mine and RTV filling material 共␳s = 1430 kg/ m3, cs with the fan turned on 共Fig. 6共a兲兲 or off 共Fig. 6共b兲兲. As shown in
= 770 J / kg K, cm = 1500 J / kg K, km = 0.2 W / m K, density ␳m Figs. 6共a兲 and 6共b兲, the soil surface temperature is less sensitive to
= 1170 kg/ m3兲. The spatially varying heat flux on the soil surface, changes in the convection heat transfer coefficient during the heat-
needed as a boundary condition in the numerical simulations, was ing period since the heat gain by radiation in that time interval is
calculated at the center of each control volume by interpolating at least one order of magnitude higher than the convection heat
between the nearest two or four experimentally measured incident loss. After turning off the lamps, the surface cooling rate shows a
heat flux values. These interpolated estimates were assumed to steep gradient indicating immediate response with its temperature
prevail over the entire control volume face. The initial tempera- dropping sharply. Comparing profiles in Figs. 6共a兲 and 6共b兲 it is
ture of the soil was set equal to the temperature of the ambient air noticed that the value of the convection heat transfer coefficient
in the environmental chamber. Results are presented for the two affects the cooling rate, but does not affect the time of peak at any
methods of cooling by forced and natural convection. Tests were depth z in the soil. Due to storage effects, the time at which the
carried out using a rectangular shaped mine filled with RTV ma- temperature peaks, for both methods of cooling, increases with
Fig. 7 The test results of the soil surface temperature above the mine center predicted by the
numerical model and measured by the infrared camera at 1 min intervals for a mine burial
depths of 0.005 m, and 0.01 m with a heating period of 5 min and a burial depth of 0.0075 m
with a heating period of 10 min
depth. The numerical model is able to reproduce the actual ther- and the plastic rigid surface and is different than the top air gap
mal processes that take place in the soil since temporal tempera- region within the mine cover in the work of Khanafer and Vafai
ture profiles obtained near the surface and inside the soil are in 关12兴 of 10 mm thickness compared to the interface layer outside
excellent agreement with measurements. The maximum error is the plastic cover of 0.3 mm used in this work. The predicted soil
less than 4% between the measured and numerically predicted surface temperature above the mine and on the mine center agrees
values. This agreement was achieved after rigorous experimenta- well with experimentally measured values using thermocouples.
tion and sensitivity analysis of the effect of the various physical Figure 7 presents a comparison between numerical predictions
parameters on the predicted response. Initially, even though solu- and measurements taken by the infrared camera at 1 min intervals
tions were close to measurements below the surface, it was not for the soil surface temperature above the mine center for burial
possible to match it at the surface. The extensive analysis per- depths of 0.005 m and 0.01 m when the heating period is 5 min
formed identified the soil thermal conductivity in the layer close and of 0.0075 m when the heating period is 10 min. Good agree-
to the surface to be the cause. Due to continuously disturbing the ment is observed where differences between readings and predic-
soil while setting up the experiment, the soil in the layer close to tions is less than 0.3° C. The longer heating period results in
the surface has more air void than deeper soil. Therefore, the higher surface temperatures and a faster cooling rate of the soil.
thermal conductivity of the surface layer is different from the The experiments have been limited to shallow burial depths due to
measured soil value and has to be calculated as a weighted aver- the limitations imposed by the heating source strength and the
age of the thermal conductivities of the soil and air. This conclu- type of soil which has a close thermal conductivity
sion was verified experimentally using a portable Quicktime30 共0.25– 0.3 W / m K兲to the buried RTV material in the mine
Thermal Properties Analyzer 共the device has an accuracy of 5% or 共0.2 W / m K兲. Typical soil conductivities that appeared in other
±0.001 W / m K for thermal conductivity in the range of studies ranged from 0.5– 0.75 W / m K and have shown ability to
0.05– 0.7 W / m K兲. The same situation would be faced with a detect signatures up to 2 – 5 cm depth in outdoor tests 关12,14兴.
newly buried landmine. For the conditions of the experiment, the Further processing of surface images taken by the infrared camera
value measured by the device helped developing the following will assess the thermal contrast between the surface temperature
equation for the soil thermal conductivity, which was used within measured above the mine and at a point away from the mine.
the top 1 cm of the soil bed:
冉
ks,surface = ks 1 −
e−3z
6
冊共16兲
6 Analysis of Thermographic Results and Image Pro-
cessing
where ks,surface is the soil thermal conductivity at depth z 共ex- The output of the IR camera is a JPEG image with its tempera-
pressed in centimeters兲 and ks is the soil thermal conductivity. As ture palette. An algorithm was developed to map the captured
depicted in Figs. 6共a兲 and 6共b兲, the use of Eq. 共16兲 allowed accu- pixels’ attributes to absolute temperatures. The input for this algo-
rate numerical predictions of the soil surface response while also rithm is the IR image and its color temperature palette, while the
predicting accurately the soil thermal response in depth. More- output is a set of temperature values. Figure 8共a兲 shows the ref-
over, the interface of the plastic layer surface of the mine box is erence spatial extent of the camera image, an actual image show-
not ideal due to the porous characteristics of the soil. A thin inter- ing the mine signature, and the temperature distribution along a
face air layer of 0.03 mm thickness was used between the plastic line on the surface of the soil for the experiment whose data are
cover of the mine, which has a thickness of 0.1 mm, and the soil. presented in Fig. 6共a兲. Figure 8共b兲 shows the three-dimensional
The thin interface layer introduced in the solution represents a temperature distribution on the surface as generated by the map-
physical effect resembling the contact resistance between the sand ping algorithm from the IR images at 5 min and 10 min from the
Fig. 8 „a… The soil bed reference image at steady conditions of the environmental chamber, and the IR camera
image during the cooling period, and the temperature distribution plot along a line on the surface of the soil for the
experiment whose data were discussed and shown in Fig. 6„a…; „b… the IR camera image, the 3-D temperature
distribution of the surface as generated by the mapping algorithm from the IR images at 5 min and 10 min from the
onset of cooling, and the corresponding surface images generated by the numerical model. „Heating period
= 5 min, mine depth= 0.01 m, Tⴥ = 25° C, RH= 50%.…
onset of cooling and the corresponding surface images generated of 0.005 m 共5 min heating兲, and 0.0075 m 共10 min heating兲 and
by the numerical model. The mean difference in temperature val- 共b兲 mine burial depth of 0.01 m 共5 min heating兲. The first is for a
ues between the camera images and those produced by the nu- mine buried at 0.005 m under the surface and exposed to 5 min of
merical code is less than 0.2° C while the maximum difference is heating and the other is for a mine buried at 0.0075 m and ex-
0.3° C. The spatial nonuniformity in the camera image is due to posed to 10 min of heating. The shallower mine has given a peak
the incident nonuniform flux from the heating lamps during heat-
contrast of about 4 ° C after one minute from the start of the cool-
ing.
ing process, while the deeper mine has given a maximum contrast
The thermal signature or contrast is defined in this work as the
temperature difference between the point on the soil surface above of 3.5° C directly at the end of the heating period. The time of
the mine center and the reference point. For the results discussed occurrence of the peak contrast depends on both the mine depth
earlier, the numerical and experimental thermal contrasts were and the heating period. The difference in temperature between the
calculated and their values, denoted by TC⬘max, were found to rep- numerical and experimental readings is less than ±0.3° C during
resent the maximum possible temperature difference on the soil cooling, while during heating the error is slightly higher 共±0.5° C兲
surface. Figure 9 shows the predicted thermal contrast, and the due to the sensitivity of the camera to the surface temperature and
measured thermal contrast using the IR images and the thermo- the need to adjust its range for temperatures above 60° C. The
couples’ readings as a function of time for 共a兲 mine burial depth camera images produce more accurate readings of the surface
Fig. 9 A plot of the predicted thermal contrast, and the mea-
sured thermal contrast using the IR images and the thermo-
couples’ readings as a function of time for „a… burial depth of
0.005 m „5 min heating…, and 0.0075 m „10 min heating… and „b…
burial depth of 0.01 m „5 min heating…
temperature than the thermocouple particularly in the period when

heating source is turned off. The surface of the soil is rough and
the tip of the thermocouple is in touch with the air and the soil
grain. At the end of the heating period, the soil temperature is not
in equilibrium with the air temperature in the void space of the
soil. This causes a drop in the measured temperature at that instant
given an error of about 0.7° C due to the sudden change from
heating to cooling at the surface.
For fixed thermal properties of soil and mine material, the de-
pendence of the peak thermal contrast on time, depth, and heating
period was investigated numerically. Figure 10 shows plots of the
thermal contrast as a function of time for 共a兲 a heating period of
Fig. 10 A plot of the thermal contrast as a function of time for
5 min. 共b兲 a heating period of 20 min at different mine depths,
„a… a heating period of 5 min, „b… a heating period of 20 min at
and 共c兲 different heating periods at fixed mine depth. Plots in Figs. different mine depths, and „c… different heating periods at fixed
10共a兲 and 10共b兲 indicate that the thermal contrast decreases with mine depth for the same climatic conditions of the indoor ex-
increasing mine’s burial depth. On the other hand, Fig. 10共c兲 re- periments considered in this work
veals that the peak thermal contrast does not change when the
heating period increases beyond 10 min. Figure 11 presents the
variation with depth of 共a兲 the peak thermal contrast and 共b兲 the
time of the peak occurrence for heating periods of 5 and 20 min. tifying potential target objects. In the real situation, the mine lo-
As shown in Fig. 11共a兲, the value of the peak thermal contrast cation is not known a priori. Moreover, relying on the experimen-
increases with the heating period due to the larger amount of tal data is not enough neither to identify the targets nor to
energy stored in the soil during heating. However, Fig. 11共b兲 in- determine their depths if classified as mines. However, using the
dicates that the time of occurrence of the peak thermal contrast numerical model it is possible to generate a database of solutions
increases with depth and approaches the end of the heating period. against which real images can be compared. The thermal signature
The numerical method has predicted well the transient thermal maps of the database at the different periods can then be used to
response of the soil surface and the resulting thermal signature. aid classifying the target. Among other objectives, future work
Integrating the model with an IR imaging system will help iden- will be directed toward developing a criterion that transforms the
brating the sensors, and conducting the experiments is highly ap-
preciated.
Nomenclature
aP , . . ⫽ coefficients in the discretized equation
bP ⫽ source term in the discretized equation
k ⫽ thermal conductivity 共W / m K兲
urf ⫽ under-relaxation factor
Sf ⫽ surface vector
t ⫽ time 共s兲
T ⫽ temperature 共°C兲
n ⫽ outward unit vector normal to cell face
d PF ⫽ vector joining the grid point P and F
E ⫽ vector collinear with d PF
T ⫽ vector equals to S − E
C ⫽ specific heat 共J / kg K兲
V ⫽ volume of control cell
h ⫽ heat transfer coefficient 共W / m2 K兲
A ⫽ surface area of soil 共m2兲
q ⫽ heat flux 共W / m2兲
x,y,z ⫽ rectangular coordinate system
Greek Symbols
␳ ⫽ density 共kg/ m3兲
⌬t ⫽ time step 共s兲.
␴ ⫽ Stefan-Boltzmann constant 共W / m2 K4兲.
␧ ⫽ emissivity.
Subscripts
f ⫽
refers to control volume face f
P ⫽
refers to the P grid point
F ⫽
refers to the neighbor of the P grid point
⬁ ⫽
refers to ambient conditions
Fig. 11 The variation with depth of „a… the peak thermal con- s ⫽
refers to soil
trast and „b… the time of the peak occurrence m ⫽
refers to mine
net ⫽
refers to net heat flux at soil surface
conv ⫽
refers to convection heat flux at soil surface
information contained in the thermal contrast maps into decision rhs ⫽
refers to incident radiation heat flux on soil
indices that help identifying targets with precision and reduce surface
false alarms. emis ⫽ refers to emitted heat flux from soil surface
NB ⫽ refers to the neighbors of the P grid point
nb ⫽ refers to faces surrounding the P grid point
7 Conclusion
Superscripts
A three-dimensional finite volume simulation tool for the pre-
o ⫽ refers to values from previous time step
diction of buried landmines in dry soil has been developed. The
n ⫽ refers to values from previous iteration
numerical model was validated by comparison with published
t ⫽ refers to transient contribution
data and by experimentation. The model predicts the expected soil
thermal signatures evolving due to the presence of buried objects. D ⫽ refers to diffusion contribution
The generated numerical results match with high accuracy the
experimentally generated IR images of the soil surface when sub- References
jected to heating and cooling. The key element in matching ex- 关1兴 Lopez, P., Sahli, H., Vilarino, D. L., and Cabello, D., 2003, “A Target Identi-
perimental and numerical images stems from the consideration of fication Procedure for the Detection and Classification of Landmines from IR
the change in thermal conductivity of the soil from surface and images,” Proceedings of EUDEM2-SCOT-2003: International Conference on
Requirements and Technologies for the Detection, Removal, and Neutraliza-
shallow surface values to higher value at more depth. The ther- tion of Landmines and UXO, September 15–18, Brussels, Belgium, 5046, pp.
mographic analysis of the model and IR images will further be 242–252.
developed to produce additional information that can help in clas- 关2兴 Simard, J. R., 1996, “Improved Landmine Detection Capability 共ILDC兲: Sys-
sifying the detected objects. tematic Approach to the Detection of Buried Mines Using IR Imaging,” Proc.
SPIE, 2765, pp. 489–500.
关3兴 Simard, J. R., 1997, “Theoretical and Experimental Characterizations of the IR
Technology for the Detection of Low-Metal and Nonmetallic Buried Land-
Acknowledgment mines,” Centre de Recherches pour la Defense Valcartier, Quebec.
The authors acknowledge the support of the Lebanese National 关4兴 Russell, K. L., McFee, J. E., and Sirovyak, W., 1996, “Remote Performance
Prediction for Infra-Red Imaging of Buried Mines,” Proc. SPIE, 2933, pp.
Council for Scientific Research Grant No. LCR-11304-002208. In 762–769.
addition, the authors would like to acknowledge the support of 关5兴 Agrawal, S., Sriram, P., and Mitchell, R., 2001, “Algorithms for IR Imagery
Youssef Jameel Grant for acquisition of the IR camera, sensors, Based Airborne Landmine and Minefield Detection,” Proc. SPIE, Remediation
and support to graduate students. The environmental chamber lab Technologies for Mines and Minelike Targets VI, 4394, pp. 284–295.
关6兴 Ngan, P., 1995, “Development of Automatic Target Recognition for Infra-Red
facility is supported by the American School and Hospital Aid Sensor-Based Close Range Land Mine Detector,” Proc. SPIE, 2496, pp. 881–
共ASHA兲 Grant No. AUB-FEA-793. The help of the research en- 889.
gineer, Samir Berjaoui in performing the sensitivity analysis, cali- 关7兴 Bruschini, C., and Gros, B., 1998, “A Survey of Current Sensor Technology
Research for the Detection of Landmines,” Sustainable Humanitarian Demin- 关14兴 Lopez, P., Sahli, H., Vilarino, D. L., and Cabello, D., 2003, “Detection of
ing: Trends, Mid Valley Press, Verona, VA, pp. 314–325. Perturbations in Thermal IR Signatures: An Inverse Problem for Buried Land
关8兴 Deans, J., Schmithalis, G., and Carter, L. J., 2001, “An Analysis of a Thermal Mine Detection,” Proceedings of EUDEM2-SCOT-2003: International Confer-
Imaging Method for Landmine Detection Using Microwave Heating,” J. Appl. ence on Requirements and Technologies for the Detection, Removal, and Neu-
Geophys., 47, pp. 334–352. tralization of Landmines and UXO, September 15–18, Brussels, Belgium, 1,
关9兴 Hermann, J., and Ian, C., 1999, “Microwave Enhancement of Thermal Land- pp. 385–392.
mine Signatures,” Proc. SPIE, 3710, Part I, pp. 154–166. 关15兴 Sendur, I. K., and Baertlein, B. A., 2000, “Numerical Simulation of Thermal
关10兴 Mitchell, R., Srinivasa, S., and Agarwal, S., 1999, “Detection of Antipersonnel Signatures of Buried Mines Over a Diurnal Cycle,” Proc. SPIE, Detection and
Landmines Based on Waterjet Induced Thermal Images,” Proc. SPIE, 3710, Remediation Technologies for Mines and Minelike Targets V, 4038, pp. 156–
Part I, pp. 180–188.
167.
关11兴 Depak, M., Agarwal, S., Ritech, G., and Swonger, C. W., 2004, “False-Alarm
关16兴 Patankar, S. V., 1980, Numerical Heat Transfer and Fluid Flow, Hemisphere
Mitigation and Feature-Based Discrimination for Airborne Mine Detection,”
Publishing, New York.
Proc. SPIE, 5415, pp. 1163–1173.
关12兴 Khanafer, K., and Vafai, K., 2002, “Thermal Analysis of Buried Land Mines 关17兴 Gordon, W. J., and Theil, L. C., 1982, “Transfinite Mappings and Their Ap-
over a Diurnal Cycle,” IEEE Trans. Geosci. Remote Sens., 40共2兲, pp. 461– plications to Grid Generation,” J. F. Thopmson ed., Numerical Grid Genera-
473. tion, North Holland, New York, pp. 171–192.
关13兴 Khanafer, K., Vafai, K., and Baertlein, B. A., 2003, “Effects of Thin Metal 关18兴 ISO 8301, 1991, Thermal Insulation, Determination of Steady-State Thermal
Outer Case and Top Air Gap on Thermal IR Images of Buried Antitank and Resistance and Related Properties,” Heat Flow Meter Apparatus.
Antipersonnel Land Mines,” IEEE Trans. Geosci. Remote Sens., 41共1兲, pp. 关19兴 Software 2004 “ThermaCamTM Researches Pro 2.8, Copyright©1997-2004,
123–135. FLIR Systems AB.
Journal of
Heat Transfer
Technical Briefs
Natural Convection From a Square

1 Introduction
Disk in the Intermediate Natural convection is found in many engineering applications.
Rayleigh Range Natural convection from a heated object is particularly of interest
for the cooling of electronic components. There are a lot of pub-
lications on natural convection from isothermal bodies 关1–7兴.
Arnout Willockx Mainly, the goal in these studies is to develop correlations be-
tween the dimensionless Rayleigh number 共Ra兲 and Nusselt num-
Department of Fluid, Heat and Combustion Mechanics,
ber 共Nu兲. Most of these studies propose a general correlation for
Ghent University-UGent, natural convection from a body. These correlations vary from each
Sint-Pietersnieuwstraat 41, 9000, other in different ways for a specific body: choice of the charac-
Ghent, Belgium teristic length Lref 关2,4,8兴, and the range of Ra for which they are
valid 共low Ra, laminar Ra, turbulent Ra or multiple ranges兲
e-mail: arnout.willockx@ugent.be 关1,3–5兴. The correlations are obtained from experiments. A com-
mon technique to obtain a wide range of Ra is varying the pres-
sure and the temperature difference between body and ambient air
Gilbert De Mey 关1,2兴. In other experiments only the temperature difference is
Department of Electronic and Information Systems, changed, but then only a small range of Ra is attained 关3兴.
Ghent University-UGent, The correlation for laminar Ra can be derived analytically from
the boundary layer equations for free convection 关9兴 and is of the
Sint-Pietersnieuwstraat 41, 9000, general form
Ghent, Belgium
Nu = C1共Pr兲 · Ra1/4 for 106 ⬍ Ra ⬍ 108 共1兲
Christophe T’Joen Equation 共1兲 is invalid for high Ra 共Ra⬎ 108兲 due to turbulent
Hendrik-Jan Steeman effects and the exponent becomes 1 / 3 for fully turbulent regime.
In between the turbulent and the laminar regime there is a transi-
Michel De Paepe tion zone. In this transition zone, the exponent changes from 1 / 4
to 1 / 3.
The exponent of 1 / 4 also becomes incorrect for Ra⬍ 106. For
Department of Fluid, Heat and Combustion Mechanics, Ra→ 0, Nu is not zero but has a finite value because the boundary
Ghent University-UGent, layer becomes too thick and conductive effects occur. The influ-
Sint-Pietersnieuwstraat 41, 9000, ence of this conductive effect becomes significant for Ra⬍ 106.
Therefore, correlations that are valid for a wide range of Ra
Ghent, Belgium have to be of another form than Eq. 共1兲. For both small and lami-
nar Ra numbers 共Ra⬍ 108兲, two types of correlations exist
This technical brief reports on an experiment in which Ra is var- Nu = C2 + C1共Pr兲 · Ra1/4 共2兲
ied in an intermediate range. It is an interesting range because
the values lie in a transition zone between laminar thin boundary
layer dominated and conduction dominated natural convection. Nu = 关Cn2 + 共C1共Pr兲 · Ra1/4兲n兴1/n 共3兲
The experimental data are compared with existing correlations to
check the accuracy and validity of these correlations for this tran- C2 is the pure conduction limit at Ra= 0 and the second term is the
sitional range of Ra. A new correlation that is only valid in this laminar boundary layer validity limit 共cf. Eq. 共1兲兲. Both equations
are certainly valid in the extreme cases of pure conduction and
transitional zone is also proposed and compared with correlations
pure laminar natural convection. However, the question arises
of the same kind. 关DOI: 10.1115/1.2190693兴 whether they also accurately describe natural convection in the
transition zone. Equation 共3兲 is an asymptotical correlation: it is
Keywords: natural convection, square disk, transitional Rayleigh an asymptotic approach of both modes and normally is more ac-
number curate in the transition zone. It is the objective of this technical
brief to check the accuracy in the transitional zone of correlations
developed for a wide range of Ra of the form of Eqs. 共2兲 and 共3兲.
NAL OF HEAT TRANSFER. Manuscript received February 9, 2005; final manuscript re- Also the value of the exponent of Ra is analyzed for equations of
ceived September 8, 2005. Review conducted by: Jay M. Khodadadi. the form of Eq. 共1兲 for the transitional zone.
Fig. 1 Experimental setup
2 Experiment
The originality of the experiment lies in the fact that Ra is
changed in another way than varying the pressure or the tempera- Fig. 2 Comparison between correlation of Yovanovich „Eq.
ture difference. Since Ra is also determined by the gravitational „6……, asymptotic correlation „Eq. „7…… and the experimental data
acceleration g, it is also possible to vary Ra by varying the accel- with error bars
eration in a test section. A variable acceleration was realized with
a centrifuge. A cylindrical enclosure hangs from an arm of the
centrifuge. The acceleration inside the cylinder changes by rotat- Nu冑A = 1.314Ra冑A
0.202
105 ⬍ Ra冑A ⬍ 106 共4兲
ing the centrifuge and the acceleration is directly determined by
the rotational speed of the centrifuge. A resistor in the shape of a
vertical square disk was put inside the cylinder. The properties of NuL = 1.134RaL0.202 6 ⫻ 104 ⬍ RaL ⬍ 3 ⫻ 105 共5兲
the square disk are an aspect ratio L / D = 0.03 and a length of L For both cases, the exponent n = 0.202 is found, but another
= 0.0255 m. The resistor dissipates power from batteries which are coefficient C was found. The exponent n = 0.202 is smaller than
also placed inside the cylinder. The test section was separated 0.25. An error calculation gave a maximum error of 0.029 on the
from the batteries with an insulating layer. A sketch of the experi- exponent n, so 0.25 is not in the error margin. This confirms that
mental setup is given in Fig. 1. the experimental data lie in the transitional zone, which is the
The power dissipated in the disk is transferred to the environ- range of interest for this brief. So it is not necessary to have an
ment by natural convection, radiation, and conduction. The sur- experiment that attains higher Ra.
face temperature of the disk is measured with a thermographic
camera and the ambient temperature inside the cylinder with a
thermocouple. Radiation and conduction are calculated from the 4 Discussion
measurements. Radiation was calculated with an accurate method 4.1 Correlations for a Wide Range of Ra. The experimental
keeping into account the transparency of the CaF2 window 共95%兲 data are compared with existing correlations for natural convec-
and the emissivities of the disk, container, and window. The cal- tion from a vertical square disk. This comparison gives informa-
culation was based on the method described in 关10兴. The conduction about the extent of accuracy of these correlations for the
tion through the wires connecting the batteries with the disk is transitional zone. Two correlations are discussed. These correla-
calculated with the one-dimensional conduction equation. The tions are developed for isothermal bodies. The thermographic im-
heat transferred by natural convection is then the dissipated power ages of the disk in the experiment show that it is not completely
in the disk minus the heat transferred by radiation and conduction. isothermal. The maximum difference between the highest tem-
This way the relation between Nu and Ra can be calculated 关11兴. perature at the center of the disk and the lowest temperatures at
The amount of radiative heat transfer and natural convection are the edges of the disk is 12% 共if the difference is divided by the
of the same order of magnitude. A detailed description of the mean temperature兲. This can probably cause a small aberration,
experiment and the measurement and calculation method is given but generally a good estimation is expected for a near isothermal
in 关11兴. body by using correlations for isothermal bodies.
4.1.1 Correlation of Yovanovich. Yovanovich 关4,7兴 uses 冑A as
3 Results characteristic length. He proposes a correlation of the form of Eq.
The acceleration inside the cylinder can be varied from 共2兲. For a square disk it is
9.81 to 53 m / s2. The mean surface temperature of the disk varies 1/4
from 84 to 72° C. The experimental data are plotted in a Nu-Ra Nu冑A = 3.19 + 0.561 · Ra冑A 共6兲
graph for two characteristic lengths Lref: the length of the disk L Both constants in Eq. 共6兲 depend on the shape and orientation of
关8兴 and the square root of the surface area of the disk 冑A 关4兴共Fig. the disk and 0.561 depends also on the Prandtl number. This cor-
2兲. The values of Ra lie within a range of 1 decade 共Fig. 2 and relation is valid for all Ra⬍ 108, thus not for turbulent Ra.
Eqs. 共4兲 and 共5兲兲 To make a comparison, both experimental data and Eq. 共6兲 are
A Nu-Ra correlation can be proposed from these experimental plotted in Fig. 2. It seems that the equation of Yovanovich under-
data for both characteristic lengths. Due to the small range of estimates Nu and thus natural convection a bit. However, the larg-
available data, only a correlation of the form Nu= CRan is signifi- est deviation for Nu is 5.2%. The average deviation is 2.2%. If the
cant. This means that the correlation is only with certainty valid in error on the measurement data is taken into account, this error
the range of Ra of the experimental data. This gives the following margin is always larger than the largest deviation 共Fig. 2兲. The
correlations for both characteristic lengths correlation of Yovanovich gives a very accurate prediction for the
experiment and as a result it is valid for the case of a small vection in a Nu-Ra1/4 graph. In the range 106 ⬍ Ra⬍ 108 the curve
rectangular square disk in the transitional range of 105 ⬍ Ra冑A is linear. Once Ra⬍ 106, the curve begins to bend upwards to the
⬍ 106. value of pure conduction 共C2 in Eqs. 共2兲 and 共3兲兲. Because the
exponent of Ra will decrease with decreasing Ra, correlations of
4.1.2 Asymptotic Correlation. Most correlations for natural the form Nu= CRan will only be valid in a narrow range of Ra in
convection from bodies are of an asymptotical form like Eq. 共3兲. the transitional zone. Correlations valid for a wider range of Ra
The length L of the disk is used as characteristic length. The have to be of the form of Eq. 共2兲 or Eq. 共3兲. So correlation 共8兲 of
available values of thickness-to-length ratio in literature for which Kobus and Wedekind can only be valid with certainty in the range
the constants are defined are 0 and 0.1 and higher. The aspect ratio of the experimental Ra 共2 ⫻ 102 ⬍ Ra⬍ 4 ⫻ 104兲. To extend the
of the disk of 0.03 is closer to zero than to 0.1. The following
validity of the correlation to lower and higher Ra, extra experi-
equation is given in 关8兴 for an aspect ratio of 0
mental data are needed. However, Kobus and Wedekind 关3兴 as-
NuL = 关共2.26兲1.07 + 共0.515 · RaL1/4兲1.07兴1/1.07 共7兲 sume that their correlation is also valid for higher Ra. Therefore
they compare it with data from Hassani and Hollands 关2兴 for
The correlation for an aspect ratio of 0 is based on the approxi- higher Ra. They state that their correlation 共Eq. 共8兲兲 under predicts
mate method of Raithby-Hollands 关12兴. The value 1.07 for the Nu a bit and that more experiments need to be done to detect the
exponent was recommended by Hassani and Hollands 关2兴. Both precise limits of the correlation. The authors state that their cor-
experimental data and Eq. 共7兲 are plotted in Fig. 2. The largest relation could be valid up to Ra= 105 but not further, because then
deviation is 7.2% and the average 4.4%. The error margin on the the influence of the laminar boundary layer becomes bigger. This
experimental data is also calculated and shown in Fig. 2. The asks for a larger exponent n. This is shown in Eqs. 共4兲, 共5兲, 共9兲,
predicted values of Nu with Eq. 共7兲 lie within this error margin. and 共10兲. These correlations are for a square disk and a vertical
Actually this is for a disk with a thickness to length ratio of 0, so cylinder and not a circular disk, nevertheless this has probably not
very thin disks like a piece of paper. The disk used in the experi- too much influence on the exponent n. So their assumption that
ment has a ratio of 0.03. Interpolation between the correlations for the correlation can be extrapolated to other Ra needs to be nu-
a ratio of 0 and 0.1 increases the largest deviation with 1% for the anced, because it can be only a very small extrapolation or maybe
ratio of 0.03, but it still lies within the error margins. So this it cannot even be extrapolated.
correlation seems less accurate than that of Yovanovich, but nev- This exponent is only valid till a certain Ra where turbulent
ertheless also predicts the experiment well. Thus the asymptotic effects start to occur 共Ra⬇ 108兲. A same transitional zone is ex-
correlation is also valid in the transitional range of 6 · 104 ⬍ RaL pected between the laminar boundary layer assumption zone and
⬍ 3 · 105. the fully turbulent zone. Then the exponent changes from 0.25 to
4.2 Discussion on Exponent n of Nu= C . Ran. Kobus and 0.33.
Wedekind 关3兴 proposed a correlation for natural convection from a
vertical circular disk. Nevertheless, their research is interesting, 5 Conclusion
because they also performed experiments in the transitional zone An experiment was executed to obtain values of Ra for free
共2 ⫻ 102 ⬍ Ra⬍ 4 ⫻ 104兲. They attained smaller Ra and a larger convection from a square disk that lie in a transition zone between
range of Ra 共2 decades兲 than our experiment. Their correlation has laminar boundary layer dominated and conduction dominated free
the form of Eq. 共1兲 but the exponent for Ra is not 1 / 4 but had to convection.
be determined from the data The measurements confirm that the general correlation of Yo-
vanovich 关4,7兴 and the asymptotic correlation are accurate for
Nu = 1.759共Ra兲0.15 共8兲
these transitional values of Ra. Thus correlations of the simple
The exponent for Ra is smaller than 1 / 4, namely 0.15. In our form: Nu= C2 + C1共Pr兲 · Ra1/4 are as accurate as asymptotic corre-
correlations 共Eqs. 共4兲 and 共5兲兲 the exponent was 0.202. Yovanov- lations: Nu= 关Cn2 + 共C1共Pr兲 · Ra1/4兲n兴1/n for the whole range of lami-
ich 关4兴 also reported a correlation of Sparrow and Ansari 关6兴 for a nar Ra, even in the transition area as shown by the experiment.
vertical cylinder. They used the diameter D of the cylinder as It is explained why extrapolation of a correlation of the form
characteristic length. Yovanovich recalculated their calculation for Nu= CRan is not allowed in the transitional zone and how the
冑A as characteristic length. The two correlations are: exponent n varies in the transitional zone.
0.208
Nu冑A = 1.037Ra冑A 1.4 ⫻ 105 ⬍ Ra冑A ⬍ 1.3 ⫻ 106 共9兲
Acknowledgment
NuD = 0.775Ra0.208 1.4 ⫻ 104 ⬍ RaD ⬍ 1.3 ⫻ 105 共10兲 This research was funded by Ghent University-Ugent BOF/
D
GOA 12051299.
These correlations are similar to our correlations 共4兲 and 共5兲 and
are valid for almost the same range of Ra. The exponent of Ra is
as good as the same for both correlations 共0.208 lies within error Nomenclature
margin of 0.202± 0.029兲. So an exponent of 0.2 seems acceptable a ⫽
thermal diffusivity 共m2 / s兲
for 105 ⬍ Ra冑A ⬍ 106. The coefficients of Ra depend on shape, C1共Pr兲 ⫽
correlation coefficient dependent of Pr
dimensions and orientation of the body, so they are different for C2 ⫽
correlation coefficient
Eqs. 共9兲, 共10兲, 共4兲, and 共5兲. Nevertheless, Sparrow and Ansari’s g gravity acceleration 共=9.81 m / s2兲
⫽
correlation confirms our correlations 共4兲 and 共5兲 and an exponent Lref characteristic length: either 冑A or L 共m兲
⫽
smaller than 0.25. Kobus and Wedekind 关3兴 found an exponent of n ⫽
correlation coefficient
0.15, because their experiments were executed at even lower Ra. Nu average Nusselt number: Nu= h · Lref / ␭
⫽
This confirms the decrease of the exponent n for decreasing Ra in Pr ⫽
Prandtl number
the zone 0 ⬍ Ra⬍ 106. Ra ⫽
average Rayleigh number:
This can be explained as follows. Once Ra⬍ 106, the conduc- Ra= g · ␤ · 共Tw − Ta兲 · Lref
3
/␯·a
tive effect has more influence on the heat transfer, while the in- Tw ⫽ temperature of the resistor 共K兲
fluence of pure natural convection diminishes for decreasing Ra. Ta ⫽ ambient temperature in the cylinder 共K兲
This causes the exponent of Ra to decrease gradually with Ra for
equations of the form Nu= CRan. When Ra→ 0 there is only con- Greek Symbols
duction and n = 0, thus Nu= Cte. This gradual decrease of the ex- ␤ ⫽ thermal expansion coefficient 共1/K兲
ponent n causes a bend in the curve that represents natural con- v ⫽ kinematical viscosity 共m2 / s兲
Subscripts 关5兴 Hassani, A. V., and Hollands, K. G. T., 1989, “Prandtl Number Effect on
冑A ⫽ characteristic length= square root of the surface
External Natural Convection Heat Transfer From Irregular Three-dimensional
Bodies,” Int. J. Heat Mass Transfer, 32共11兲, pp. 2075–2080.
area of the square disk 共m兲关6兴 Sparrow, E. M., and Ansari, M. A., 1983, “A Refutation of King’s Rule for
L ⫽ characteristic length= length of the square disk Multidimensional External Natural Convection,” Int. J. Heat Mass Transfer,
共m兲 26共9兲, pp. 1357–1364.
关7兴 Yovanovich, M. M., and Jafarpur, K., 1993, “Bounds on Laminar Natural
Convection From Isothermal Disks and Infinite Plates of Arbitrary Shape for
References All Orientations and Prandtl Numbers,” HTD 共Am. Soc. Mech. Eng.兲, 264,
pp. 93–110.
关1兴 Chamberlain, M. J., Hollands, K. G. T., and Raithby, G. D., 1985, “Experi-
ments and Theory on Natural Convection Heat Transfer From Bodies of Com- 关8兴 Rohsenow, M. W., Hartnett, J. P., and Cho, Y. I., 1998, Handbook of Heat
plex Shape,” ASME J. Heat Transfer, 107, pp. 624–629. Transfer, McGraw–Hill, New York, Chap. 4, pp. 4.25–4.32.
关2兴 Hassani, A. V., and Hollands, K. G. T., 1989, “On Natural Convection Heat 关9兴 Kays, W. M., and Crawford, M. E., 1993, Convective Heat and Mass Transfer,
Transfer From Three-Dimensional Bodies of Arbitrary Shape,” ASME J. Heat McGraw-Hill, New York, Chap. 17.
Transfer, 111, pp. 363–371 关10兴 Siegel, R., and Howell, J. R., 1992, Thermal Radiation Heat Transfer, 3rd ed.,
关3兴 Kobus, C. J., and Wedekind, G. L., 1995, “An Experimental Investigation Into Hemisphere, Bristol, UK, Chap. 7, pp. 263–272 and 936–939.
Forced, Natural and Combined Forced and Natural Convective Heat Transfer 关11兴 Willockx, A., Steeman, H., De Mey, G., and De Paepe, M., 2005, “Analysis of
From Stationary Isothermal Circular Disks,” Int. J. Heat Mass Transfer, Natural Convection and Radiation by Using Thermography,” 14th Interna-
38共18兲, pp. 3329–3339. tional Conference on Thermal Engineering and Thermogrammetry, June 22–
关4兴 Yovanovich, M. M., 1987, “On the Effect of Shape, Aspect Ratio and Orien- 24, Budapest, Hungary.
tation Upon Natural Convection From Isothermal Bodies of Complex Shape,” 关12兴 Rohsenow, M. W., Hartnett, J. P., and Ganiç, E. N., 1985, Handbook of Heat
HTD 共Am. Soc. Mech. Eng.兲, 82, pp. 121–129. Transfer, McGraw–Hill, New York, Chap. 6, pp. 6.9–6.16.
Least-Squares Finite Element Analysis Monte Carlo method 关8兴. However, little information about finite
element method 共FEM兲 for transient radiative transfer was re-
for Transient Radiative Transfer ported.
in Absorbing and Scattering Media As a numerical method, finite element method has become a
sophisticated technology in mechanics and thermal analyses. It
has many advantages over other methods. For complex geometric
W. An shape, it shows considerable flexibility. Its complication does not
increase with the complexity of geometric shape and boundary
L. M. Ruan conditions. Besides, the approximation for the field variables in a
e-mail: ruanlm@hit.edu.on volume or surface element can vary across the element, and the
variation of field variables in the FEM can be specified to increase
H. P. Tan degrees of approximation. Therefore, the FEM of radiation heat
transfer has been investigated by many researchers.
H. Qi Early FEM for radiative transfer is based on integral form of
RTE, as in the case of Burns 关9兴, who developed a FEM to simu-
School of Energy Science and Engineering, late radiative heat transfer in multidimensional media with non-
Harbin Institute of Technology, gray, non-homogeneous radiative properties. Furmanski and Ban-
150001 Harbin, People’s Republic of China naszek 关10兴 studied the coupled conduction and radiation heat
transfer in a semitransparent medium by FEM. Their finite ele-
ment models need the complex integration and geometrical com-
In some radiative transfer processes, the time scales are usually putation which may require too much calculation in multidimen-
on the order of 10−9 – 10−15 s, so the transient effect of radiation sional problems. Furthermore, it is difficult to deal with
should be considered. In present research, a finite element model, anisotropic scattering and obstacles in the media by their methods.
which is based on the discrete ordinates method and least-squares Very recently, standard Galerkin finite element method 共GFEM兲
variational principle, is developed to simulate the transient radia- has been developed and integrated with DOM to simulate the
tive transfer in absorbing and scattering media. The numerical radiation transfer by Liu 关11兴. This model can solve anisotropic
formulations and detailed steps are given. Moreover, two transient scattering conveniently and needs no complex integration and vis-
radiative transfer problems are investigated and the results are ibility judgment. The accuracy of this model in absorbing and
compared with those by integral method and finite volume anisotropic scattering media is investigated by An et al. 关12兴. The
method. It indicates that the present model can simulate the tran- least-squares finite element method 共LSFEM兲关13兴 and discon-
sient radiative transfer effectively and accurately. tinue finite element method 关14兴 also are presented to simulate the
关DOI: 10.1115/1.2190694兴 radiative transfer. However, all of these finite element methods are
used to simulate the steady radiative heat transfer. Further re-
Keywords: transient radiative transfer, finite element, absorbing
search should be devoted in the transient radiative transfer due to
and scattering media, microscale, non-intrusive diagnostics
the hyperbolic wave equation coupled with the in-scattering inte-
gral term. As Tan 关1兴 suggested, the simulation of transient radia-
tive transfer is an important approach to verify the reliability of
the model from the perspective of numerical simulation. For the
1 Introduction model with considerable accuracy in steady simulation, some dif-
In most of the radiative transfer studies, the temporal variation ficulties such as stability and unrealistic oscillation may be iden-
of radiative signal is not considered since the time scale of radia- tified in transient simulation. It follows that, the objective of the
tive transport is far smaller than the time scales of the observable present study is to verify the accuracy and stability of LSFEM for
properties. For these applications, radiative transfer can be taken transient radiative transfer.
as steady-state process. The transient term of radiative transfer Due to hyperbolic wave characteristic of the transient radiative
equation 共RTE兲 can be neglected. However, recent research on the transfer equation 共TRTE兲, the numerical computation based on
propagation of ultrashort light pulse inside the absorbing and scat- GFEM has an unstable behavior leading to unrealistic oscillations
tering media has led to some interesting applications such as ma- for advection-dominated problems. To eliminate the unrealistic
terial properties diagnostics, optical imaging, laser therapy, par- oscillations, some numerical techniques have been developed in
ticle detection and sizing, remote sensing, etc. For these emerging the past decades. The successful representations in these numeri-
applications, the transient effect of radiation should be considered, cal techniques are the Petrov-Galerkin method 关15兴, Taylor-
because the time scales of such processes are usually on the order
Galerkin method 关16兴, and least-squares method 关17兴. The Taylor-
of 10−9 – 10−15 s. In these applications, the variation of radiative
Galerkin and Petrov-Galerkin method provide much stabilized
signals with time always needs be precisely computed. As pointed
out by Tan 关1兴, the accurate forward calculation is essentially im- numerical solutions by generating numerical diffusion, which is
portant for inverse analysis and estimates. Therefore, the consid- an artificial diffusion as an inherent upwind damping. But for the
eration of the transient term in the radiation transport equation is Petrov-Galerkin method, particularly the Streamline Upwind
necessary. Besides these emerging technologies, in astrophysical Petrov-Galerkin method 关15兴, there is at least one free parameter
radiation hydrodynamics, a better model of the transient radiative to be tuned. The Taylor-Galerkin method is implemented by using
transport is also needed 关2兴. Over the past five years, transient a Taylor series for time discretization before applying space dis-
radiative transfer in participating media has received considerable cretization. The high order derivatives of temporal variable need
attention. Several numerical strategies have been developed, to be deduced and introduced into the discrete equation. Accord-
which include discrete ordinate method 共DOM兲关3,4兴, finite vol- ingly, the Taylor-Galerkin method is relatively complicated.
ume method 共FVM兲关5,6兴, integral equation 共IE兲 models 关1,7兴, and The reason for the success of the LSFEM for fluid flows and
heat transfer problems can be justified as follows: in the first
place, unlike the standard Galerkin and Petrov-Galerkin method,
NAL OFHEAT TRANSFER. Manuscript received June 19, 2005; final manuscript received LSFEM leads to symmetric, positive-definite matrix systems for
October 7, 2005. Review conducted by Walter W. Yuen. advection-dominating transfer equations. Only half of a sparse
matrix needs to be stored. Some highly-efficient matrix solvers solve the TRTE numerically, the angular dependence has to be
such as the preconditioned conjugate gradient method, and the removed first. Similar to the conventional DOM, the angular in-
Cholesky decomposition method can be successfully used. tegration over 4␲ solid angles is replaced by summing up the
Whereas the Galerkin or Petrov-Galerkin method results in a function on the discrete, a finite number of ordinates directions.
large, sparse and non-symmetric matrix system which requires a For specific ordinate direction m, its direction cosines along the
robust iterative solver and a prohibitive amount of memory stor- coordinates x, y, and z can be denoted as ␮m, ␩m and ␰m. Then, the
age and computing time for multidimensional problems. Further, TRTE of m direction can be written as
there is no free parameter to be tuned in the LSFEM. The LSFEM
can be applied in a straightforward manner, much in the same way ⳵Im ⳵Im ⳵Im ⳵Im
+ ␮m + ␩m + ␰m = − 共␬a + ␴s兲Im + ␬aIb
as the Galerkin finite element method. The program for LSFEM C⳵t ⳵x ⳵y ⳵z
can be realized by very little modification in the program of stan- M
dard Galerkin finite element method. Compared with the Petrov- ␴s
Galerkin method in which the tuning parameter may be different +
4␲ 兺I m⬘
⌽m⬘mwm⬘ 共3兲
for high order interpolation element, in the LSFEM, higher order m⬘=1
interpolation element without any modification for the weight where M is the number of discrete directions; the boundary con-
function is used, which is more efficient and accurate for transient ditions can be written as
radiative simulation. Donea and Quartapelle 关18兴 studied the con-
vergence of LSFEM in the convection dominated transport prob- 1 − ␧w
lems compared with that of other FEMs. w = ␧wIbw +
Im
␲ 兩n 兺 Im ⬘
w 兩nw · sm⬘兩wm⬘ 共4兲
As a pioneering and creative research, LSFEM is applied to w·sm⬘兩⬍0
analyze the one-dimensional 共1D兲 steady radiative transfer prob-
lem by Pontaza and Reddy 关13兴. The space-angle coupled and where sm⬘, wm⬘ are the unit vector and the solid angle associated
decoupled least-squares finite element models were proposed to with the direction m⬘; nw is the unit normal vector on the wall. If
simulate the radiative transfer in one-dimensional absorbing and the following simplified coefficients are defined
scattering media in the literature. However, the space-angle ␴s mm
coupled model is difficult to extend to a multidimensional prob- B = ␬a + ␴s − ⌽ wm 共5a兲
lem. Therefore, the present paper employs a space-angle decou- 4␲
pled model and extends this model to transient radiative transfer
simulation in one-dimensional and two-dimensional 共2D兲 partici- ␴s
M
pating media. In the following section of the present paper, the S = ␬ aI b +

4␲ 兺 I m⬘⌽ m⬘mw m⬘ 共5b兲
mathematic formulation and numerical process of LSFEM are m⬘=1,m⬘⫽m
proposed and two transient radiative transfer cases are simulated.
The results produced by the present model are compared with Then Eq. 共3兲 can be simplified as the form
those by IE and FVM, and the efficiency and accuracy of our ⳵Im ⳵Im ⳵Im ⳵Im
model is verified. + ␮m + ␩m + ␰m + BIm = S 共6兲
C⳵t ⳵x ⳵y ⳵z
2 Mathematic Formulation Equation 共6兲 is a hyperbolic partial difference equation about Im
which can be solved numerically by using finite difference method
2.1 Radiative Transfer Equation. For the participating me- or finite volume method, certainly by using finite element method
dia, the transient radiative transfer equation 共TRTE兲 in the Carte- too. The weighted residual method applied to Eq. 共6兲 provides
sian coordinate system can be written as
⳵I共r,⍀,t兲 ⳵I共r,⍀,t兲
C⳵t
+
⳵s
= − 共␬a + ␴s兲I共r,⍀,t兲 + ␬aIb共r,t兲冕冉V
⳵Im
C⳵t
+ ␮m
⳵Im
⳵x
+ ␩m
⳵Im
⳵y
+ ␰m
⳵Im
⳵z
+ BIm − S WldV = 0 共7兲冊
+
␴s
4␲
冕⍀⬘=4␲
I共r,⍀⬘,t兲⌽共⍀⬘,⍀兲d⍀⬘
where V is the domain of solution. For FEM, different weighting
function Wl will lead to different finite element methods. The
unknown variable Im and its derivative can be written as
共1兲
N
where I共r , ⍀ , t兲 is radiative intensity, which is a function of posi-
tion r, direction ⍀ and time t; Ib共r , t兲 is the blackbody radiative
Im = 兺I l=1
l ␸l
m
共8兲
intensity of media, C is propagation speed of radiation transport in

the medium, ␬a and ␴s are the absorption and scattering coeffi- N
⳵Im ⳵␸l
cient, respectively. The ⌽共⍀⬘ , ⍀兲 is the scattering phase function
between incoming direction ⍀⬘ and scattering direction ⍀. The ⳵xi
= 兺 ⳵x I
l=1 i
m
l 共xi = x,y,z兲共9兲
term on left side of the equation represents the gradient of radia-

tive intensity in the direction ⍀. The three terms on right side of where Iml is the radiative intensity at the node l and ␸l is the shape
equation represent the change in intensity due to absorption and function. For standard Galerkin finite element approach, the
outgoing scattering, emission and incoming scattering, respec- weight function Wl is equal to the shape function ␸l, whereas in
tively. When considered the gray and diffuse reflection boundary, the least-squares finite element approach, the weight function can
the radiative boundary condition can be written as following form be expressed as
I共rw,⍀,t兲 = ␧wIb共rw,t兲 +
1 − ␧w
␲
冕
n·⍀⬘⬍0
I共rw,⍀⬘,t兲冉
Wl = ␸l + ␪ · ⌬t* ␮m
⳵␸l
⳵x
+ ␩m
⳵␸l
⳵y
+ ␰m
⳵␸l
⳵z
+ B␸l 冊共10兲
⫻兩n · ⍀⬘兩d⍀⬘ 共n · ⍀ ⬎ 0兲共2兲 where ␪ is the factor of time difference; its value can be 1, 0.5,
2 / 3, or 0, which represent implicit 共Backward-Euler兲, Crank-
where I共rw , ⍀ , t兲 is the leaving intensity on the boundary, Nicolson, Galerkin, and explicit 共Forward-Euler兲, respectively.
I共rw , ⍀⬘ , t兲 is the arriving intensity on the boundary. The term ␧w The time step ⌬t* = C · ⌬t. For the triangle isoparametric element,
is the wall emissivity; n is the unit normal vector on the wall. To the shape function ␸l 共l = i , j , k兲 can be written as a function of x , y
1 values of xl and y l 共l = i , j , k兲 in ␸i are variable by turning i, j, k.
␸i = 关共x j y k − xky j兲 + 共y j − y k兲x + 共xk − x j兲y兴共11兲 The Eq. 共7兲 can be rewritten in matrix form as
2A
n+1兴 = 关K⬘兴关In 兴 + 关R兴
关K兴关Im 共12兲
m
where A is the area of triangle element, xl and y l 共l = i , j , k兲 are the
values of coordinates of three nodes in the triangle element. The with
冋冉
⳵␸i ⳵␸i ⳵␸i
冊册
冕冋
␸i + ␪ · ⌬t* ␮m+ ␩m + ␰m + B␸i
⳵x ⳵y ⳵z
兺
冉冊册
关K兴 = dxdydz 共13兲
⳵␸ j ⳵␸ j ⳵␸ j
e e ⫻ ␸ + ␪ · ⌬t* ␮
j m + ␩m + ␰m + B␸ j
⳵x ⳵y ⳵z
冋冉
⳵␸i ⳵␸i ⳵␸i
冊册
兺冕
␸i − 共1 − ␪兲 · ⌬t* ␮m + ␩m + ␰m + B␸i
⳵x ⳵y ⳵z
冋冉冊册
关K⬘兴 = dxdydz 共14兲
⳵␸ j ⳵␸ j ⳵␸ j
e e ⫻ ␸ + ␪ · ⌬t* ␮
j m + ␩m + ␰m + B␸ j
⳵x ⳵y ⳵z
关R兴 = 兺
e
冕 e
冋
␪ · ⌬t*共Sn+1 + Sn兲␸i ⫻ ␸ j + ␪ · ⌬t* ␮m 冉 ⳵␸ j
⳵x
+ ␩m
⳵␸ j
⳵y
+ ␰m
⳵␸ j
⳵z
+ B␸ j 冊册 dxdydz 共15兲
where 关K兴 is the stiffness matrix, 关R兴 is the right side term. Im
n+1 is
boundary condition expressed in Eq. 共4兲 should be imposed upon
the radiative intensity in m direction at the n + 1 moment. It can be Eq. 共12兲 as the inflow boundary condition. Due to the dependence
solved through the value of Im of both source terms and boundary conditions on the intensities,
n . The stiffness matrix generated
from the LSFEM is symmetric and positive definite 共SPD兲, global iterations are necessary at each time step. At the outset, the
whereas that generated from the Galerkin finite element method is source terms S in Eq. 共5b兲 and irradiation should be computed by
unsymmetric. Consequently, storing of the stiffness matrix for the the initial intensity. Then the stiffness matrix K, K⬘ and right hand
LSFEM is very inexpensive and some fast, robust iterative solu- term R in Eqs. 共13兲–共15兲 are computed. The matrix expressed in
tions for SPD matrix systems can be employed. By solving the Eq. 共12兲 is solved for each direction. At each time step, the com-
Eq. 共12兲, the radiative intensity in m direction can be acquired. putational results replace the previous ones and the iterative pro-
Similarly, we can write out the equations and obtain the radiative cedure will continue until the convergence is met. The processes
intensity in each discrete direction. above are repeated until the end time. The solver of the discretized
equations is based on preconditioned conjugate gradients arith-
2.2 Discretization. The present method involves not only metic.
spatial discretization but also angular discretization. For spatial
discretization, FEM is highly flexible. Spatial grid can be polygo-
nal unstructured element, which is more convenient and easier to 3 Results and Discussion
use than structured grid in the application of complex geometry.
The 4␲ angular domain at any spatial location is divided into a Based on the above theories, a computer code is implemented.
finite number of discrete, non-overlapping solid angle. Many an- Two transient cases are used to evaluate the accuracy of our
gular discretization strategies have been developed in the last model. In these cases, the numerical solution was considered to be
three decades. The Sn-type discretization and azimuthal discreti- convergent when the relative error of radiative intensity is less
zation are two of the most widely used strategies. The former is than 0.01%. All computation is realized in a PC with AMD2100
always employed in the DOM and the latter is associated with the + CPU. The grid-independent checks for all results have already
FVM. Theoretically, the Sn-type discretization is more rigorous been verified in our research.
than azimuthal discretization. The DOM adopts a numerical In the first case of a one-dimensional problem, we compared
quadrature method to evaluate the solid angle integrals. These the results of FEM with those of IE 关1兴, which is considered as
quadrature sets are constructed to be invariant under any 90 deg precise one. The walls are black and at a cold temperature 共0 K兲.
rotation and to satisfy a number of key moments of the radiative At initial time t = 0, the temperature of the left boundary is sud-
intensity. The FVM treats directional discretization by analytical denly raised to provide an unit radiative intensity I0 for all subse-
integration. In our previous study, in spite of the numerical results quent time. Considering the absorbing and isotropic scattering
predicted by these two directional discretization strategies, the media, the optical thickness is 1.0 and the albedo is 0.5. The solid
results of Sn-type discretization are more accurate than that of angular discretization adopts S20 and the spatial domain is dis-
azimuthal discretization. Hence, the Sn-type discretization is em- cretized into 100 uniform linear elements. With Crank-Nicolson
ployed in the following numerical test. However, as pointed out scheme as the time difference scheme, the time step is taken as
by Raithby 关19兴, the azimuthal discretization could be necessary ⌬t * = 0.02. The time t* is defined as t * = Ct. The dimensionless
in some radiative problem such as collimated radiation problem, radiative heat flux Q and incident radiation G are, respectively,
because it allows the angular grid to be adapted to resolve sharp defined as
changes in the intensity with direction.
2.3 Solution Method. The spatial discretization and compu-
Q共x,t兲 =
1
I0
冕4␲
I共x,⍀,t兲cos共s,n兲d⍀
tational processes in each discrete direction are the same, and the
radiative intensity in each direction is solved independently. The 共16兲
Fig. 2 Schematic diagram and quadrangle-element grid „10
Ã 10… in 2D irregular quadrilateral enclosure
the bottom wall at the different time. This case has been investi-
gated by using the FVM 关6兴. As shown in Fig. 3, the results of
FEM and FVM are very close. The maximum relative error is less
6.5%.
4 Conclusion
In the present study, a finite element model has been developed
and applied to simulate transient radiative transfer in one-
dimensional and two-dimensional enclosure. The simulation is
based on least-squares variational principles and DOM. It leads to
a symmetric positive definite coefficient matrix which can be
stored inexpensively and solved efficiently. This model could pro-
duce smooth result even for a longer time step. Compared with the
other oscillation-eliminate numerical techniques, the LSFEM con-
tains the implicit upwind component and does not need to tune
any parameter. The discretization strategy for angular domain
could use either the Sn-type discretization or the azimuthal dis-
cretization. The results of LSFEM show a good agreement with
that of IE and FVM. Therefore, the present model is a promising
approach for the transient radiative transfer, and can easily be
expanded to a three-dimensional problem. Just as the present
model has been developed in a 1D and 2D transient radiative
problem, it is would seem especially fruitful to promote the study
of transient radiative transfer with collimated light irradiation by
LSFEM. In summary, the present model might realize its potential
and make a difference in the transient radiative research.
Acknowledgment
The support of this work by the National Natural Science Foun-
dation of China 共No. 50276014兲 is gratefully acknowledged.
Fig. 1 The heat flux and incident radiation in one-dimensional

slab at different time: „a… the heat flux in the media at different
time; „b… the incident radiation in the media at different time
G共x,t兲 =
1
I0
冕
4␲
I共x,⍀,t兲d⍀
where s present the direction of the radiative intensity I; n is the

direction of coordinate axis x.
As shown in Fig. 1, the results of FEM and IE have a desired
agreement. The maximum relative error is less 4.0%. With the
time elapsing, the incident radiation increases, however the radia-
tive flux near the hot surface decreases due to the backscattering
of radiant energy.
In the second case, a two-dimensional transient radiative trans-
fer problem is studied. As shown in Fig. 2, an irregular quadrilat-
eral enclosure is filled with an absorbing-emitting medium main-
tained at a constant temperature of Tg. The wall is black and cold
Tw = 0 K. The absorption coefficient is 1.0 m−1. The solid angular
discretization adopts S8. The spatial domain is discretized into
40⫻ 40 quadrangular bilinear elements. The Crank-Nicolson
scheme is also employed in the time difference scheme. The time Fig. 3 The heat flux on the bottom wall in 2D irregular quadri-
step is taken as ⌬t * = 0.025. Figure 3 shows the radiative flux on lateral enclosure at different time
Nomenclature diative Transfer,” ASME J. Heat Transfer, 123, pp. 466–475.
关2兴 Balsara, D. W., 1999, “An Analysis of the Hyperbolic Nature of the Equations
A ⫽ area of element of Radiation Hydrodynamics,” J. Quant. Spectrosc. Radiat. Transf., 61共5兲, pp.
B ⫽ modified extinction coefficient, Eq. 共5a兲 617–627.
C ⫽ propagation speed of radiation transport in the 关3兴 Guo, Z., and Kumar, S., 2000, “Discrete-Ordinates Solution of Short-Pulsed
medium Laser Transport in Two-Dimensional Turbid Media,” Appl. Opt., 39共24兲, pp.
G ⫽ dimensionless incident radiation, Eq. 共16兲 4411–4417.
关4兴 Sakami, M., Mitra, K., and Vo-Dinh, T., 2002, “Analysis of Short-Pulse Laser
I ⫽ radiative intensity, W / 共m2 sr兲 Photon Transport Through Tissues for Optical Tomography,” Opt. Lett., 27共5兲,
K, K⬘ ⫽ stiffness matrix of finite element pp. 336–338.
M ⫽ number of discrete directions 关5兴 Chai, J. C., 2003, “One-Dimensional Transient Radiation Heat Transfer Mod-
Q ⫽ dimensionless radiative heat flux, Eq. 共16兲 eling Using a Finite-Volume Method,” Nat. Biotechnol., 44, pp. 187–208.
S ⫽ modified source function, Eq. 共5b兲关6兴 Chai, J. C., 2004, “Transient Radiative Transfer in Irregular Two-Dimensional
Geometries,” J. Quant. Spectrosc. Radiat. Transf., 84, pp. 281–294.
R ⫽ right side term of finite element 关7兴 Wu, C. Y., 2000, “Propagation of Scattered Radiation in a Participating Planar
V ⫽ the domain of solution Medium With Pulse Irradiation,” J. Quant. Spectrosc. Radiat. Transf., 64, pp.
Wl ⫽ the weight function 537–548.
nw ⫽ unit normal vector of boundary surface 关8兴 Hsu, P. F., 2001, “Effects of Multiple Scattering and Reflective Boundary on
q ⫽ radiative heat fluxes, W / m2 the Transient Radiative Transfer Process,” Int. J. Therm. Sci., 40, pp. 539–
549.
r ⫽ the location
关9兴 Burns, S. P., Howell, J. R., and Klein, D. R., 1995, “Finite Element Solution
sm⬘ ⫽ unit vector in the direction m⬘ for Radiative Heat Transfer With Nongray, Nonhomogeneous Radiative Prop-
t ⫽ the time erties,” ASME National Heat Transfer Conference, Vol. 13, pp. 3–10.
wm⬘ ⫽ angular weight 关10兴 Furmanski, P., and Bannaszek, J., 2004, “Finite Element Analysis of Concur-
x, y, z ⫽ coordinate directions rent Radiation and Conduction in Participating Media,” J. Quant. Spectrosc.
Radiat. Transf., 84, pp. 563–573.
⍀ ⫽ solid angle, sr 关11兴 Liu, L. H., 2004, “Finite Element Simulation of Radiative Heat Transfer in
⌽ ⫽ scattering phase function Absorbing and Scattering Media,” J. Thermophys. Heat Transfer, 18共4兲, pp.
␧ ⫽ emissivity 555–557.
␬a ⫽ absorption coefficient, m−1 关12兴 An, W., Ruan, L. M., Qi, H., and Liu, L. H., 2005, “Finite Element Method for
␴s ⫽ scattering coefficient, m−1 Radiative Heat Transfer in Absorbing and Anisotropic Scattering Media,” J.
Quant. Spectrosc. Radiat. Transf., 96共3–4兲, pp. 409–422.
␻ ⫽ scattering albedo 关13兴 Pontaza, J. P., and Reddy, J. N., 2005, “Least-Squares Finite Element Formu-
␸l ⫽ shape function lations for One-Dimensional Radiative Transfer,” J. Quant. Spectrosc. Radiat.
␶ ⫽ optical thickness Transf., 95共3兲, pp. 387–406.
␮, ␩, ␰ ⫽ direction cosines 关14兴 Cui, X., and Li, B. Q., 2005, “Discontinuous Finite Element Solution of 2D
Radiative Transfer With and Without Axisymmetry,” J. Quant. Spectrosc. Ra-
Subscripts diat. Transf., 96共3–4兲, pp. 383–407.
b ⫽ blackbody 关15兴 Hughes, T. J. R., and Brooks, A., 1982, “Streamline Upwind/Petrov-Galerkin
e ⫽ element Formulation for Convection Dominated Flows With Particular Emphasis on
the Incompressible Navier-Stokes Equations,” Comput. Methods Appl. Mech.
i, j, k, l ⫽ node Eng., 32, pp. 199–259.
n ⫽ time step 关16兴 Donea, J., 1984, “A Taylor-Galerkin Method for Convective Transport Prob-
w ⫽ wall lems,” Int. J. Numer. Methods Eng., 20, pp. 101–119.
␪ ⫽ zenith angle 关17兴 Jiang, B. N., and Povinelli, L. A., 1990, “Least-Squares Finite Element
␸ ⫽ round angle Method for Fluid Dynamics,” Comput. Methods Appl. Mech. Eng., 81, pp.
13–37.
Superscripts 关18兴 Donea, J., and Quartapelle, L., 1992, “An Introduction to Finite Element Meth-
m, m⬘ ⫽ indices for ordinate directions ods for Transient Advection Problems,” Comput. Methods Appl. Mech. Eng.,
95, pp. 169–203.
关19兴 Raithby, G. D., 1999, “Discussion of the Finite Volume Method for Radiation
References and its Application Using 3D Unstructured Meshes,” Numer. Heat Transfer,
关1兴 Tan, Z. M., and Hsu, P. F., 2001, “An Integral Formulation of Transient Ra- Part B, 35, pp. 389–405.
On the Entropy Generation Formula of the entropy production and discussed the sign of its time de-
rivative. It is shown that the entropy production due to emission,
of Radiation Heat Transfer absorption, and scattering of radiation is a form bilinear in gener-
alized thermodynamic fluxes and forces. Arpaci 关7–10兴, and Ar-
Processes paci and Esmaeeli 关11兴 studied the radiative deformation and ra-
diation entropy generation. Recently, Caldas and Semiao 关12兴
presented a numerical simulation method of radiative entropy gen-
L. H. Liu1 eration in semitransparent participating media. This method is
e-mail: lhliu@hit.edu.cn completely compatible with standard radiative transfer calculation
like the discrete ordinate ordinates method.
S. X. Chu In the community of heat transfer, traditionally, the formula of
entropy generation rate per unit volume for heat transfer, includ-
School of Energy Science and Engineering, ing conduction or radiation, is written as
Harbin Institute of Technology,
ⵜTM 共r兲
92 West Dazhi Street, Ṡ⵮
gen共r兲 = − q共r兲 · 共1兲
T2M 共r兲
Harbin 150001,
People’s Republic of China where TM is the medium temperature, and q is the total heat flux
vector. According to Eq. 共1兲, the entropy generation rate per unit
volume for radiation heat transfer is often written as 共see, for
example, Arpaci 关7–10兴, and Arpaci and Esmaeeli 关11兴兲
Because thermal radiation is a long-range phenomenon, the local
radiative heat flux is dependent on the temperature distribution of ⵜT M 共r兲
⵮ 共r兲 = − qR共r兲 ·
ṠR,V 共2兲
the entire enclosure under consideration and is not determined by T2M 共r兲
the local temperature gradient. In the community of heat transfer,
traditionally, the conduction-type formula of entropy generation where qR is the radiative heat flux vector. Equation 共2兲 is similar
rate is used to calculate the entropy generation rate of radiation to the formula of entropy generation rate of heat conduction.
heat transfer. In the present study, three counterexamples are con- However, as shown in the next section, careful examination of Eq.
sidered. The discrete ordinates method is employed to solve the 共2兲 reveals that this equation is incorrect in a thermodynamics
radiative transfer equation and then solve the radiative entropy sense. The processes of radiation heat transfer differ from those of
generation rate. The results show that the traditional formulas of conduction heat transfer. One distinguishing feature between con-
entropy generation rate for heat transfer generally cannot be used duction and radiation is the difference in their temperature depen-
to calculate the local entropy generation rate of radiation heat dencies. Conduction heat transfer is well described by Fourier’s
transfer. Only in optically extremely thick situations, the tradi- law, and the local heat flux of conduction is linearly proportional
tional formula of entropy generation rate for heat transfer can be to the local temperature gradient. In the view of irreversible ther-
approximately used to calculate the local entropy generation rate modynamics, the heat flux of conduction is driven by the local
of radiation heat transfer. 关DOI: 10.1115/1.2190695兴 temperature gradient. Therefore, the conductive entropy genera-
tion rate per unit volume can be calculated using Eq. 共1兲. How-
Keywords: entropy generation rate, radiative heat transfer, semi- ever, thermal radiation is generally a long-range phenomenon,
transparent medium, calculation formula therefore the local radiative heat flux is dependent on the tempera-
ture distribution of the entire enclosure under consideration and is
not determined by the local temperature gradient. Only in opti-
cally extremely thick situations, radiation heat transfer processes
1 Introduction can be described by the diffusion approximation and the local heat
Entropy generation is associated with thermodynamic irrevers- flux of radiation is approximately proportional to the local tem-
ibility, which is present in all heat transfer processes and results in perature gradient.
the loss of available work. The contemporary trend in the field of In this paper, we take three counterexamples to show that the
heat transfer and thermal design is to apply the second law analy- traditional formula 共Eq. 共2兲兲 of entropy generation rate for heat
sis of thermodynamics and its design-related concept of entropy transfer cannot be used to calculate the local entropy generation
generation minimization. The entropy generation and its minimi- rate of radiation heat transfer.
zation were investigated extensively by Bejan 关1–3兴, who pre-
sented systematically the concept and the optimization method of
entropy generation minimization. 2 Counterexamples
Thermal radiation is an important factor in the thermodynamic To show the incorrectness of a traditional formula 共Eq. 共2兲兲 of
analysis of many high-temperature systems, such as solar collec- radiation entropy generation, as shown in Fig. 1, we consider a
tors, boilers, and furnaces. The correct evaluation of radiation en- one-dimensional absorbing, emitting, and isotropic scattering gray
tropy generation is important when determining the second-law medium bounded by two blackbody walls. The thickness of the
performance of these energy conversion devices. Planck 关4兴 was semitransparent slab is L = 1.0 m. The optical thickness based on
the first to investigate the interaction between light and matter the slab thickness is denoted as ␶L. The temperatures of the left
with respect to its irreversibility. Thermodynamic aspects of radia- and the right walls are imposed as TW1 and TW2, respectively. The
tive transfer processes were later studied by Wildt 关5兴 who was distribution of medium temperature T M is prescribed. In this
mainly interested in atmospheres in radiative equilibrium. Wildt study, we mainly check the correctness of the traditional formula
derived a number of inequalities for the radiative entropy that 共Eq. 共2兲兲 for radiative entropy generation rate, therefore no con-
follow from macroscopic considerations alone, and was the first to duction entropy generation is considered. For the following nu-
formulate the transfer equation for the radiative entropy. Starting merical analysis, the semitransparent slab is uniformly divided
from basic statistical relations, Kroll 关6兴 analyzed the general form into 1000 parts, and the discrete ordinates method 共S8兲关13兴 is
employed to solve the radiative transfer equation and then solve
Contributed by the Heat Transfer Division of ASME for publication in the JOUR- the radiative entropy generation rate.
NAL OFHEAT TRANSFER. Manuscript received July 30, 2005; final manuscript received
October 21, 2005. Review conducted by Walter W. Yuen. 2.1 Example 1. We first consider the radiative heat transfer
1
Corresponding author. and the radiative entropy generation in the case of ␻ = 0.0,␶L
Fig. 1 One-dimensional semitransparent slab system
= 1.0, TW1 = 1000 K, and T M = TW2 = 500 K. In the view of irrevers-

ible thermodynamics, for the example under consideration, the
one-dimensional system shown in Fig. 1 is equivalent to a system
in which heat is transported from a high-temperature heat source Fig. 3 Radiative heat flux for Example 2
共1000 K兲 to a low-temperature heat source 共500 K兲 by radiation,
but from which no work has been obtained. Therefore, theoreti-
cally, the total radiative entropy generation of the system is is less than zero within the region under consideration. Similar to
冉冊
Examples 1 and 2, this also conflicts with the second law of ther-
1 1
R
SRgen = qW − ⬎0 共3兲 modynamics. Therefore, Eq. 共2兲 is incorrect, and the traditional
1 T M T W1 formula 共Eq. 共2兲兲 of entropy generation rate for heat transfer can-
not be used to calculate the local entropy generation rate of radia-
In this example, the temperature of semitransparent medium is
tion heat transfer.
uniform, therefore the temperature gradient is ⳵T M / ⳵x = 0. Be-
The local radiative entropy generation rate calculated using the
cause TW1 is greater than T M and TW2, as shown in Fig. 2, the axial
formulas of Caldas and Semiao in Ref. 关12兴 is also shown in Fig.
radiative heat flux is greater than zero. The radiative heat transfer 6 for Example 3. By comparison with the results calculated using
process is irreversible, but the local radiative entropy generation Eq. 共2兲, it is shown that, even if both the temperature gradient and
rate calculated using Eq. 共2兲 is Ṡ⵮R,V共r兲 = 0. This conflicts with the the axial radiative heat flux within the region of x / L
thermodynamics analysis of the system. 苸关0.75, 0.95兴 are greater than zero for the case of ␶L = 2, the local
radiative entropy generation rate calculated using the formulas of
2.2 Example 2. In this example, we consider the radiative
Caldas and Semiao 关12兴 is greater than zero. This is consistent
heat transfer and the radiative entropy generation in the case of
with the Second Law of Thermodynamics. For the case of ␶L
␻ = 0.0,␶L = 1.0, TW1 = 1000 K, TW2 = 500 K, and TM = 300
= 40, the local radiative entropy generation rate calculated using
+ 200x / L , K. The temperature gradient is ⳵T M / ⳵x = 200/ L ⬎ 0. As Eq. 共2兲 is close to that calculated using the formulas of Caldas and
shown in Fig. 3, the axial radiative heat flux is greater than zero. Semiao 关12兴. In optically extremely thick situations, the traditional
The local radiative entropy generation rate calculated using Eq. formulas 共Eq. 共2兲兲 of entropy generation rate for heat transfer can
共2兲 is shown in Fig. 4. The local radiative entropy generation rate be approximately used to calculate the local entropy generation
calculated using Eq. 共1兲 is Ṡ⵮R,V共r兲 ⬍ 0. This conflicts with the rate of radiation heat transfer.
second law of thermodynamics.
3 Conclusions
2.3 Example 3. In this example, we consider the radiative
entropy generation in the case of ␻ = 0.5, TW1 = TW2 = 1000 K, and The processes of radiation heat transfer differ from those of
T M = 1000+ 500 sin共2␲x / L兲 , K. The axial radiative heat flux and conduction heat transfer. One distinguishing feature between con-
the local radiative entropy generation rate calculated using Eq. 共2兲 duction and radiation is the difference in their temperature depen-
are shown in Figs. 5 and 6 for the cases of ␶L = 2 and ␶L = 40, dencies. Thermal radiation is a long-range phenomenon. There-
respectively. In the case of ␶L = 2, both the temperature gradient fore, the local heat flux is dependent on the temperature
distribution of the entire enclosure under consideration and is not
and the axial radiative heat flux within the region of x / L
determined by the local temperature gradient. In the present study,
苸关0.75, 0.95兴 are greater than zero. This leads to the result that
three counterexamples are considered. The results show that the
the local radiative entropy generation rate calculated using Eq. 共2兲
Fig. 4 Radiative entropy generation rate computed using Eq.

Fig. 2 Radiative heat flux for Example 1 „2… for Example 2
Fig. 5 Radiative heat flux for Example 3 Fig. 6 Comparison of radiative entropy generation rates for
Example 3
traditional formulas of entropy generation rate for heat transfer

generally cannot be used to calculate the local entropy generation TW1, TW2 ⫽ temperature at walls, K
rate of radiation heat transfer. Only in optically extremely thick ␶L ⫽ optical thickness
situations, can the traditional formula of entropy generation rate ␻ ⫽ single scattering albedo
for heat transfer be approximately used to calculate the local en-
tropy generation rate of radiation heat transfer. References
关1兴 Bejan, A., 1982, Entropy Generation Through Heat and Fluid Flow, Wiley,
Acknowledgment New York.
The supports of this work by the National Natural Science 关2兴 Bejan, A., 1996, Entropy Generation Minimization, CRC Press, New York.
关3兴 Bejan, A., 1997, Advanced Engineering Thermodynamics, 2nd ed., Wiley,
Foundation of China 共50425619兲 and the Science Fund of New York.
Heilongjiang Province for Distinguished Young Scholars 共No. 关4兴 Planck, M., 1959, The Theory of Heat Radiation, Dover, New York.
JC04-03兲 are gratefully acknowledged. 关5兴 Wildt, R., 1956, “Radiative Transfer and Thermodynamics,” Astrophys. J.,
123, pp. 107–116.
关6兴 Kroll, W., 1967, “Properties of the Entropy Production due to Radiative Trans-
Nomenclature fer,” J. Quant. Spectrosc. Radiat. Transf., 7, pp. 715–723.
L ⫽ thickness of semitransparent slab, m 关7兴 Arpaci, V. S., 1986, “Radiative Entropy Production,” AIAA J., 24, pp. 1859–
1860.
q ⫽ heat flux vector 关8兴 Arpaci, V. S., 1987, “Radiative Entropy Production—Lost Heat into Entropy,”
qR ⫽ radiative heat flux vector Int. J. Heat Mass Transfer, 30, pp. 2115–2123.
r ⫽ position vector 关9兴 Arpaci, V. S., 1991, “Radiative Entropy Production—Heat Loss into Entropy,”
in Advances in Heat Transfer, J. P. Hartnett and T. F. Irvine, eds., 21, Aca-
Ṡ⵮
gen ⫽ entropy generation rate of heat transfer per unit demic, New York, pp. 239–276.
volume, W / K m3 关10兴 Arpaci, V. S., 2001, “Thermal Deformation: From Thermodynamics to Heat
Transfer,” ASME J. Heat Transfer, 123, pp. 821–826.
SRgen ⫽ total radiative entropy generation of slab sys- 关11兴 Arpaci, V. S., Esmaeeli, A., 2000, “Radiative Deformation,” J. Appl. Phys.,
tem, W / K m2 87, pp. 3093–3100.
关12兴 Caldas, F., Semiao, V., 2005, “Entropy Generation Through Radiative Transfer
Ṡ⵮
R,V ⫽ radiative entropy generation rate per unit vol- in Participating Media: Analysis and Numerical Computation,” J. Quant. Spec-
ume, W / K m3 trosc. Radiat. Transf., 96, pp. 423–437.
TM ⫽ medium temperature, K 关13兴 Modest, M. F., 2003, Radiative Heat Transfer, 2nd ed., Academic, New York.

J Heat Transfer 2006 Vol 128 N05

Uploaded by

Copyright:

Available Formats

J Heat Transfer 2006 Vol 128 N05

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

J Heat Transfer 2006 Vol 128 N05

Uploaded by

Copyright:

Available Formats

Journal of

VOLUME 128 • NUMBER 5 • MAY 2006

Journal of Heat Transfer Volume 128, Number 5 MAY 2006

504 On the Entropy Generation Formula of Radiation Heat Transfer Processes

The ASME Journal of Heat Transfer is abstracted and indexed in the

Keywords: enthalpy, water, thermodynamic properties, thermodynamic analysis,

T=4°C T = 20° C T = 50° C T = 100° C T = 150° C

T = 200° C T = 250° C T = 300° C T = 350° C

T °C → 4°C 20° C 50° C 100° C 150° C 200° C 250° C 300° C 350° C

20 36.55 102.57 226.51 434.17 644.45 860.27 1086.70 1334.40 1646.00

30 46.23 111.77 235.05 441.74 650.89 865.02 1088.40 1328.90 1608.80

40 55.79 120.90 243.56 449.33 657.42 870.00 1090.70 1325.60 1588.80

50 65.24 129.95 252.03 456.94 664.02 875.19 1093.50 1324.00 1576.10

60 74.59 138.94 260.47 464.56 670.69 880.55 1096.80 1323.50 1567.50

80 93.00 156.70 277.26 479.82 684.21 891.73 1104.40 1325.10 1557.70

Journal of Heat Transfer MAY 2006, Vol. 128 / 425

冕 ⌬h⌬v ⫽ enthalpy correction due to the change of ⌬v,

426 / Vol. 128, MAY 2006 Transactions of the ASME

Keywords: condensation, retention, heat exchangers

where ␳w is the density of water, V共D兲 is the volume of a droplet

428 / Vol. 128, MAY 2006 Transactions of the ASME

V= 冕 H共D, ␤, ␪max, ␪min兲d␾ 共8兲

Journal of Heat Transfer MAY 2006, Vol. 128 / 429

where ␥ is the liquid surface tension, and ␪ and ␾ are as defined

Fgx = − ␳wgV sin ␣ 共15兲

430 / Vol. 128, MAY 2006 Transactions of the ASME

Journal of Heat Transfer MAY 2006, Vol. 128 / 431

432 / Vol. 128, MAY 2006 Transactions of the ASME

Journal of Heat Transfer MAY 2006, Vol. 128 / 433

␣1 and ␣2 are electrochemical coefficients depending on the ex-

Fig. 2 Configuration of the computational domain

Journal of Heat Transfer MAY 2006, Vol. 128 / 435

Equations Diffusion layer Catalyst layer Eq.

436 / Vol. 128, MAY 2006 Transactions of the ASME

3 Results and Discussion

Fig. 3 Grid test by comparing the axial velocity at Y = 2.0

Journal of Heat Transfer MAY 2006, Vol. 128 / 437

Property name Solid phase Fluid phase

Physical Material Carbon fiber Oxygen

Geometric Diffusion layer porosity, ␧ 48%

Nondimensional Conductivity ratio, Rk 0.01447

438 / Vol. 128, MAY 2006 Transactions of the ASME

Journal of Heat Transfer MAY 2006, Vol. 128 / 439

440 / Vol. 128, MAY 2006 Transactions of the ASME

Journal of Heat Transfer MAY 2006, Vol. 128 / 441

442 / Vol. 128, MAY 2006 Transactions of the ASME

Journal of Heat Transfer MAY 2006, Vol. 128 / 443

Keywords: turbulence modeling, porous media, heat transfer coefficient

1 Introduction medium. Sahraouri and Kaviany 关12兴 performed direct numerical

of the time-averaged velocity vector ū.

Reynolds stresses and the eddy viscosity is given by, respectively, 2

Journal of Heat Transfer MAY 2006, Vol. 128 / 445

=ⵜ· ks ⵜ 关共1 − ␾兲具Ts典i兴 −

− hiai共具Ts典i − 具T f 典i兲 共21兲 共kdis兲yy

446 / Vol. 128, MAY 2006 Transactions of the ASME

2.5 Interfacial Heat Transfer Coefficient. In Eqs. 共20兲 and

Journal of Heat Transfer MAY 2006, Vol. 128 / 447

⳵ 2k 3.1 Film Coefficient hi. Determination of hi is here obtained

448 / Vol. 128, MAY 2006 Transactions of the ASME

− hiai共具Ts典i − 具T f 典i兲共21兲共kdis兲yy