Академический Документы
Профессиональный Документы
Культура Документы
1643-1652
Pergamon Press Ltd 19RO. Printed in Great Britain
Abstract - Experiments on free convective heat transfer from electrically heated platinum wires and a
platinum strip to supercritical carbon dioxide were performed for a wide range of bulk conditions. It is shown
that heat transfer can be predicted by a conventional Nusselt-type correlation if the dimensionless numbers
are based on integrated thermophysical properties in order to account for large changes in these properties.
The anomaly of thermal conductivity has to be considered. Agreement between the experimental results and
the correlation is within 10% even for very thin wires when for those a correction factor is introduced.
NOMENCLATURE Subscripts
C, specific heat capacity ; b, bulk condition ;
d, wire diameter ; c, at critical state;
9, gravitational constant ; 1, integrated mean value, equation (4);
h, heat transfer coefficient ; P, at constant pressure ;
h*, specific enthalpy ; PC> pseudocritical (temperature);
1, wire (strip) length; W, heater wall.
P? pressure;
r, co-ordinate normal to surface;
t, temperature [“Cl ; 1. INTRODUCTION
thermodynamic temperature [K] ; THE INCREASINGnumber of applications of heat trans-
2 coefficient of thermal expansion ; fer to supercritical fluids near their thermodynamical
4 thermal conductivity; critical point has revived the interest in investigating
r?, dynamic viscosity ; this particular heat transfer problem in the last two
V, kinematic viscosity; decades. In modern power plants heat is transferred to
Pt density ; supercritical water. Rockets are cooled with supercriti-
Gr, Grashof number ; cal fuel and superconductivity effects are achieved by
gd3 Pb - Pi cooling the conductor with fluids which are close to
__.
Gri,
v? Pi their critical points. Also highly charged machine
Nusselt number; elements like gas turbine blades and computer ele-
Nu,
hd ments are cooled with supercritical fluids.
Nui, > One of the earliest experimental investigations done
on heat transfer to a critical fluid is that of Schmidt et
NC+, modified Nusselt number with 63, [equa- al. [2]. They found that the free convective heat
tion (3)] omitted; transfer coefficient increases considerably as the liquid
J+, Prandtl number; approaches its critical state in a closed loop. This was
2VihF-hb* . attributed to the large changes of the thermophysical
Pri,
li t, - t, ’ properties with temperature in the critical region : the
coefficient of thermal expansion and the specific heat
Ra, Rayleigh number ;
capacity strongly increase, the dynamic viscosity de-
Rai, Gri Pri ;
creases. Meanwhile it is known that also the thermal
Ra*, modified Rayleigh number with A), [equa-
conductivity increases in the vicinity of the critical
tion (3)] omitted.
point. All these changes improve the mechanism of
heat transfer and particularly that of free convection. A
single closed pipe filled with carbon dioxide so that a
near critical state is reached within this ‘heat pipe’
*The material presented here is taken from a Dr.-Ing. thesis exhibits increases in efficient thermal conductivity up
submitted to the Universitlt Stuttgart by R. J. Neumann - to 12000 times the conductivity of copper [3].
bibliography reference rll. It is difficult to solve free convection problems
t Assistant Professor- of Mechanical Engineering, Un- analytically because the temperature and the velocity
iversidad de Concenci6n. Casilla 53-C. Concewi6n. Chile.
1 Professor, Ins&t fii; Thermodynamik und WLimetech-
fields depend upon each other. In the critical region the
nik, Universitst Stuttgart, Pfaflenwaldring 6, D-7000 Stutt- integration of the governing differential equations is
gart 80, Germany. aggravated by the fact that the thermophysical proper-
1643
1644 ROIXUfo J. NI.IIMA?Nand EKIC‘HW. P. HAHYI
ties are extremely temperature dependent. Many au- of the thermal diffusivity and the geometry of the
thors used a modified substitution method developed heating surface.
by Ostrach [4] and Sparrow and Gregg [5] to
integrate the differential equations. 2. THERMOPHYSICAI, PROPERTIES
In most cases the evaluation was for heat transfer For carbon dioxide the following critical proper-
from a flat plate. Reimann and Grigull [6] solved the ties were taken : t, = 31.04 C, pC = 71.81 bar and p,
problem for the free convective heat transfer from thin = 468 kg/m3. Figure 1 shows that density /) and
wires. The agreement of the numerical solutions with enthalpy h* are extremely temperature dependent in
the experimental results obtained by all these authors the critical region, particularly in the vicinity of the
is fairly good only when the fluid is not in a near critical critical point. As a consequence, the coefficient of
state. Deviations in the near critical state may be due to thermal expansion /I’and the specific heat capacity cp
the neglection of the anomaly in thermal conductivity respectively, show a sharp peak at certain tempera-
near the critical point, in the older works. tures called therefore pseudocritical temperatures, t,.
In calculating heat transfer coefficients it is cus- The pseudocritical temperature I,, increases with
tomary to use Nusselt correlations. For free con- pressure (Table 1). The peaks in cp and /1 tend to
vection flow with dynamic forces being negligible infinity at the critical point and become smaller as the
compared to viscous forces pressure increases. All thermodynamic properties used
here were calculated with the canonical equation of
Nu =f(Ra, (1) state for carbon dioxide [ 111.
where The dynamic viscosity q [12] is also shown in Fig. 1.
It decreases with temperature, proportionally to den-
gd3PAt sity. The thermal conductivity i. shows a sharp peak
Ra=Gr,Pr; Gr=------ Pm%. (2)
vz ’ IL at the pseudocritical temperature.
Although no explanation has been found for this
Rigorously, a function of this kind can only be applied strange behaviour its existence is commonly accepted.
to solve physically similar heat transfer problems. This The thermal conductivity can be expressed as follows
means that the velocity distribution is in a fixed
relation to the temperature distribution, corrected i = ;,(T, p = 0) + ;({I) + Ai. (3)
only by the Prandtl number. This holds when the The first two terms represent the well-known de-
physical properties in equation (2) are not temperature pendence of a transport property on temperature and
dependent. It has been tried to determine Nusselt density. The third term Ai. accounts for the abnormal
correlations for free convection heat transfer near the peaks in the vicinity of the pseudocritical tempera-
critical point by evaluating the thermophysical pro-
perties at an arithmetic mean temperature and in-
troducing a suitable correction term. This method T,K
works when the critical or pseudocritical state in the 300 305 310
p,-s- ” ’ ” ” ’ ” ” ”
fluid is not approached too closely (k 2 K, e.g. chapter
5).
In the early fifties Goldmann [7] found that heat
transfer to supercritical water by forced convection is
sometimes accompanied by a noise like a ‘whistle’ and
by an increase in the heat transfer coefficient. He
postulated a special heat transfer mechanism which
was called by him ‘pseudoboiling’. Since then similar
phenomena have been observed at free convection
from electrically heated wires to supercritical carbon
dioxide by Knapp and Sabersky [8], Hahne et al. [9]
and Nishikawa et al. [lo]. The increase in heat transfer
started with oscillations of the wire at temperature
differences not less than 70K. No satisfactory expla-
nation for this unusual phenomenon has been pub-
lished yet.
The experiments described in this paper were de-
signed to find a correlation for heat transfer by free
convection from thin wires and a flat strip to
supercritical carbon dioxide near the critical point.
For this purpose, the extreme temperature dependence
of the thermophysical properties is taken into account.
Experiments on pseudoboiling phenomena were per- FIG. 1. Thermophysical properties of carbon dioxide at
formed and are reported in [l J. They exhibited effects 75 bar.
Free convective heat transfer to supercritical carbon dioxide 1645
PRESSURE VESSEL
SUPPORTING DEVICE
HEATING ELEMENT
COPPER CONNECTION PLATE
TUBE WITH THERMOCOUPLES
GLASS WINDOWS
INSULATED TANK
iHERMOSTAi
COOLER
AUXlLiARY PRESSURE VESSEL
PISION PUMP
THERMOSTAT
PREC!SION BOURWN-GAUGE MANOMETER
DC POWER SUPPLY
PRECISION RESISTANCE
MULTIPLE PRECISION SWITCH
VOLTMETER
LINE VOLTAGE PLOTTER
REFERENCE VOLTAGE AOJUSTER
VOLTAGE AMPLIFIER
THERMOCOUPLE SELECTOR
REFERENCE VOLTAGE ADJUSTER
ZERO VOLTMETER
BOTTLE
VENT VALVE
SAFEIY VALVE
either slightly decreasing or being approximately Generally, the heat transfer coefficients decrease as
constant with wall temperature. It is thus obvious that the wire diameter increases. The lowest values were
for tb > t,, the heat transfer coefficient is primarily a measured for the 5mm high strip. Compared to a
function oft, while for t, < t, it is a function of both t, 0.05mm dia wire the heat transfer coefficients for a
and tb. 0.1 mm wire were about 25 % lower, for the 0.3 mm and
Figure 5 gives the heat transfer coefficient for the for the strip they were roughly about 30% and 60%
same wire at a bulk pressure of 80 bar. The same lower respectively. Nevertheless, for all these heating
qualitative observations as described above can be elements the dependence of the heat transfer coefficient
made under these test conditions. Quantitatively the on the temperatures t,, t, and t, and on pressure was
peaks of the heat transfer coefficient at t, close to t, confirmed.
are smaller than before. At other supercritical pres- Boiling-like phenomena were not observed with any
sures as listed in Table 1 similar results were obtained of the platinum heating elements.
Bulk
pressures, bar Bulk Wall
Heating element (pseudocritical temperatures temperatures
Pt, 1 = 67mm temp, “C) (“C) approx (“C)
10, 25, 31
0.3 mm wire 74, 75, 80, 90 31.5, 32, 35. up to 200°C
50
IO I
I
this state increase much more than Nusselt numbers,
h max calculate d=O.l mm. pt even for peak heat transfer coefficients, so that in the
= 1.07
p=7Lbar, lb i 10°C 0 plot Nu vs Ra within a wide range of scatter a
0,9
ltpc =31,15T) zooc A
25% 0
dependence Nu x constant is falsely pretended. This is
o,fJ
30% 0 also observed in a correlation for horizontal slots filled
31% t
with CO, [14].
3159: A
35oc v
In order to take account of variable properties,
47 + 50-c x
I Nusselt [ 151 suggested to use an integrated mean value
I I ItI l- Calculated CUNPS
for any property P, as calculated from
1 1,
p=- P(t)dt. (4)
TV- tb s lb
n
-10 20 30
II?/ LO 50 60
temperatures (at constant pressure) there are different
properties in a thermal layer adjacent to the heating
surface, maintaining some resistance to the heat
t, PC 1
transport. Consequently, the heat transfer coefficient
remains finite. (In our experiments, at the best, heat
FIG. 4. Heat transfer coefficient from a O.lmm Pt wire at transfer coefficients of h = 9700 W/m’ K were mea-
74 bar vs wall temperature. sured and h = 10700 W/m2 K was calculated.)
The coefficient of thermal expansion and the specific
heat capacity tending towards infinity in the critical
5. CORRELATIONS state have to be especially considered in the aforemen-
Although the thermophysical properties in the tioned respect. They can be approximated as:
critical region are extremely temperature dependent
and heat transfer processes are not physically similar,
it is commonly accepted to use equation (1) for the
correlation of data.
A simple correlation plot, Nu vs Ra,with thermal
properties taken at film temperature (tb + t,)/2 results c~=(~)~=(~),=(~),. (6a)
in anomalies for data obtained for wall temperatures
near the pseudocritical state [9] : Rayleigh numbers in These approximations have the disadvantage that the
mean values only depend on the two boundary
temperatures and do not take into account an extreme
change of properties within the thermal layer. In order
to introduce an integrated mean property into equa-
tion (5a), Kato et al. [16] proposed to approximate the
product /I. At as
arithmetic mean temperature; this gives At = t, - below it: t, < t,, < t,
(tb + t,)/2 and provides a better fit to cp Case III : all temperatures are above the pseudocriti-
= dh*/dr. cal value: t, < t, < t,.
For the thermal conductivity the integrated mean
Figure 6 presents Nui as a function of Rai for a 0.1 mm
value iLiwas taken as a first choice, despite the fact that
dia platinum wire at 75 bar and various bulk tempera-
it exhibits an abnormal increase close to the critical
tures. For comparison a conventional, empirical cor-
point.
relation by van der Hegge Zijnen [18]
With these properties, the following numbers were
obtained : Nui = 0.35 + 0.25Ra:” + 0.45Raf’4 (10)
gd’ [lb - I’i is represented as a line. Deviations of _t 10% to this
Rai = Gri Pr,, Gri = 2 z ~--,
“i Pi line are also plotted in the diagram. This correlation
was selected because it is based on a great number of
Pr, = 2>V - hb* experiments as performed by many investigators on
I ~ (7)
Li t, - t, horizontal cylinders and wires for many fluids, it
The Nusselt number is derived from the boundary covers a range of 10m4 < Ra, < 109.
condition The experimental data for the cases I and III fall well
within the f 10% scatter region of equation (lo), since
the physical properties do not change extremely there.
When, however, t, approaches c,, the properties
This means that all the heat dissipated by the wall is change considerably and the heat transfer coefficients
transferred through a thin layer adherent to the wall, reach some maximum. For this case (II) the diagram
only by conduction. Thus the thermal conductivity in shows that Rai increases more rapidly than Nui and
the Nusselt number should be evaluated at wall loops are formed outside the lower boundary of the
temperature [ 171. However, due to its anomaly in the scatter region. At high temperature differences (e.g.
critical region, the thermal conductivity may increase (t, - tb) > 80 K ; tb = 1O’C) both numbers decrease
considerably while - as experiments show - the heat again and the systematic deviations converge in the
transfer coefficient does not increase correspondingly. & 10 % region of equation (10). The largest deviations
The experimental results proved the following Nusselt result for the lowest t,. As t, increases the deviations
number definition to be the most suitable decrease; for tb also close to t,, the experimental data
fall within the + 10% scatter region.
Nui = hd At higher bulk pressures, similar but smaller syste-
ii matic deviations were obtained. A comparison for
three different wire diameters at the same bulk con-
In the following analysis three different cases shall be
ditions is shown in Fig. 7. The deviations from
discerned :
equation (10) become smaller as the wire diameter
Case I: all temperatures are below the respective increases.
pseudocritical temperature: t, < t, < t, Figure 8 shows the experimental data for a flat strip,
Case II : the wire temperature equals or exceeds the simulating a vertical wall, at 74 bar and various bulk
pseudocritical value while the bulk temperature is still conditions. The characteristic length din equations (7)
7
6
5
Nut
1
15
p=7Lbor, Pt itb < tpcl
12
d = 0,05mm, tb = 1ooc 0
10 25oc x
N”,8
6
and (9) was changed to the height of the strip (5 mm). increase considerably when t, is relatively low. Thus, as
Again for comparison a correlation developed by iLi increases, Nui somewhat stagnates causing the
Krischer [ 191 for a wide range of Grashof and Prandtl aforementioned loops.
numbers is plotted in the diagram. The scatter of This suggests that the thermal conductivity may not
experimental data is small confirming that a correction be evaluated properly from equation (4) or the anom-
function has to be considered only for thin wires and aly is overestimated by equation (3). Since the
case II. experimental data for small temperature differences
In all plots the systematic deviations are largest and t, close to t,, are well represented by equation (lo),
when t, is relatively low and t, exceeds t, only by little however, the second possibility might be excluded.
(case II). Then the peak values of fi and cp are within Some authors argue that the anomaly of thermal
the integration limits and Gr, increases considerably. conductivity is not significant for the heat transfer in
The increase of cp is partially counterbalanced by the this case. The contrary can be proven by calculating
increase of 3,,within a short temperature range, causing Nui as a function of Ra, without considering the
Pri to increase only moderately. On the other hand, as anomaly term A2 in equation (3) and plotting such a
shown in Fig. 4, the heat transfer coefficient does not correlation. This was done in Fig. 9 where now very
lo-- , 1
a-
1X=0,35
+0.25Ro?+ 0.15Ra;"‘
Ll
10 LO IO' 10' IO' 10' 106 10'
Ra',
FIG.9.Correlation ofmeasured data for various wirediameters without considering the thermal conductivity
anomaly term A%in Nui and Rae
distinct deviations from equation (10) can be observed. and T being the thermodynamic temperature T
These start when the wall temperatures approach = t + 273.15.
closely the pseudocritical and they are largest when the This correlation (11) was used to obtain the calcu-
thermal layer spans the pseudocritical state. These lated curves h vs t, in Figs. 4 and 5.
deviations appear for every diameter investigated here. In Fig. 10 the correlation of data according to
The finding of a decrease in deviation of data from equation (11) is shown. The fit of data as also presented
equation (10) with increasing wire diameter (Fig. 7) in Fig. 7 is better here. All data are within a + 10 %
suggests a special consideration for thin wires. For scatter.
these, the thermal boundary layer thickness increases For technical use those temperature ranges were
unproportionally compared to the diameter, and examined for which simple film temperatures cl
conductivity prevails in a larger region around the = (t,,, + Q/2 suffice to determine the thermophysical
wire. As t, exceeds t, at low t, presumably only a very properties in the numbers of equation (10) and this
thin film has a high thermal conductivity, while the equation still yields agreement within + 10%. For the
mean thermal resistance of the boundary layer is still conditions listed in Table 1 these ranges are
high; thus the heat transfer coefficient does not
tb < Tw < $a, -2K and t,>tb>tpc+2K.
increase considerably. For surfaces with a greater
curvature radius, the thermal resistance due to con- Thus free convection heat transfer in the critical region
duction prevails only in a very thin film on the wall and can be predicted from equation (10) without perfor-
convection governs the heat transfer process. The ming integrations as long as bulk and wall tempera-
aforementioned influence of thermal conductivity is tures are either 2K below or above the respective
counterbalanced by the buoyancy effects. pseudocritical temperature.
The deviations principally depend upon c, - t,,,
t, - t, and p. For each of these terms a separate
correction function was determined from the experim-
ental data and the following modified correlation is
proposed for case II and thin wires 6. CONCLUSIONS
Nui = (0.35 + 0.25Raj'8 + 0.45Ra!'"(l +fif&-' Heat transfer coefficients for free convection heat
transfer in the near critical region can be predicted by a
(11)
conventional, and well-established Nusselt corre-
where lation. If a region of f 2 K around the pseudocritical
temperature is spared the thermophysical properties in
fi = xc-x, the NW and Ra numbers may simply be based on film
temperature.
Very close to the pseudocritical state where property
fi = tanh(30y)
changes are large, good agreement of experimental
data and the aforementioned correlation can be ob-
tained with integrated mean property values. The
anomaly of thermal conductivity has to be taken into
Free convective heat transfer to supercritical carbon dioxide 1651
40
p.EObar, 0
I& 3,o
$1
2 2,0
I,6
1.3
‘1,O
FIG. 10. Correlation of measured data according to the corrected equation (11).
account. For thin wires a correction factor introduced transfer to carbon dioxide near the critical point, Inl. J.
Heat Mass Transfer 9, 41-51 (1966).
in the correlation yields & 10% deviation. Boiling-
9. E. W. P. Hahne, G. Feurstein and U. Grigull, Free
like phenomena do not occur on platinum wires up to convective heat transfer in the supercritical region, Proc.
investigated temperatures of 600°C. 5th Int. Heat Transf. Conf., Tokyo, Vol. 3, pp. 5-9 (1974).
10. K. Nishikawa, T. Ito and H. Yamashita, Free convective
heat transfer to a supercritical fluid. Mem. Faculty Engng,
Kyushu Univ. 30, 2 (1970).
11. R. Meyer-Pittroff and U. Grigull, Eine kanonische
REFERENCES Zustandsgleichung fiir Kohlendioxid, Wiirme- und St&l
iibertragung 3, 134- 145 (1973).
1. R. J. Neumann, Zur WIrmeiibertragung durch freie 12. M. P. Vukalovich and V. V. Altunin, Thermophysical
Konvektion im kritischen Gebiet von Kohlendioxid, Dr.- Properties of Carbon Dioxide. Collet, London (1968).
Ing. thesis Universitlt Stuttgart (1979). 13. B. Le Neindre, R. Tufeu, P. Bury and J. V. Sengers,
2. E. Schmidt, E. R. G. Eckert and U. Grigull, WBrmetran- Thermal conductivity ofcarbon dioxide and steam in the
sport durch Fliissigkeiten in der N;ihe Ihres kritischen supercritical region, Ber. Bunsenges. Phys. Chem. 77(4),
Zustands, Jahrbuch der Deutschen Luftfahrtforschung 2, 262-275 (1963).
53-58 (1939). 14. E. Hahne, Natural convection heat transfer through an
3. E. Hahne, WHrmetransport durch natiirliche Konvek- enclosed horizontal layer of supercritical carbon dioxide,
tion in Medien nahe ihrem kritischen Zustand, Int. J. W&me- und Stoffibertraguny 1, 190-196 (1968).
Heat Mass Transfer 8, 481-497 (1965). 15. W. Nusselt, Das Grundgesetz des Wkmelbergangs,
4. S. Ostrach, An analysis of laminar free-convection flow Gesundheits-lngenieur 42, 477-482 (1915).
and heat transfer about a flat plate parallel to the 16. H. Kato, N. Nishiwaki and H. Hirata, Studies on the heat
direction of the generating body force, NASA Report transfer of fluids at a supercritical pressure, Bull. Japan
1111 (1953). Sot. Mech. Engrs 11(46), 654-663 (1968).
5 E. M. Sparrow and J. L. Gregg, The variable fluid- 17. R. Gregorig, Verallgemeinerter Ausdruck fiir den EinfluB
property problem in free convection, J. Heat Transfer temperaturabhgngiger Stoffwerte auf den turbulenten
8OC(3), 879-886 (1958). WLrmetibergang, Wiirme- und Stoffibertragung 3,26-40
6 M. Reimann and U. Grigull, Analytische Untersuchung (1970).
von freier Konvektion und Filmsieden in laminaren 18. B. G. Van der Hegge Zijnen, Modified correlation
Grenzschichten mit temperaturabhgngigen Stoffwerten, formulae for the heat transfer by natural and by forced
W&-me- und Stqffiibertraguny 8, 167-178 (1975). convection from horizontal cylinders, Appl. Scient. Res.
I. K. Goldmann, Heat transfer to supercritical water at 6A, 129-140 (1956).
5000 psi flowing at high mass flow rates through round 19. 0. Krischer, Die wissenschafrlichen Grundlagen der
tubes, Int. Dee. Heat Transfer 3, 561-568 (1961). Trocknungstechnik, S. 139, 2. Auflage. Springer, Berlin
8. K. K. Knapp and R. H. Sabersky, Free convective heat (1963).
R&sum6 ~ Des expiriences sur la convection thermique naturelle autour de fils et d’un ruban en platine
chat&% electriquement, dans du gaz carbonique supercritique, ont it& conduites pour un large domaine de
conditions. On montre que le transfert thermique peut gtre ivalut par une formule du type de Nusselt si les
nombres sans dimension sont basks sur des propri&s thermophysiques intCgrtes de fagon B tenir compte
des grandes variations de ces propri&Cs. L’anomalie de conductivit6 thermique est consid&&e. On obtient un
accord meilleur que 10% entre les resultats expbimentaux et la formule, m&me pour les fils trts fins lorsque,
pour ceux-ci, un facteur correctif est introduit.
1652 RODOLFO J.NEUMANN and ERICH W. P. HAHNF
AHHoTaw- ~poBeneHo3KcnepHMeHTanbHoeHccnenoBaHHecBo60nHoKoHBeKTHe~oronepeHocaTenna
OT HBI-FBBeMbIX 3JleKTPHWCKHM TOKOM n.NTHHOBbIX npOBO,IO’,‘ZK II MaTAHOBOfi JIeHTb, K CBCPX-
KPHTH’ieCKOfi OKHCN yrJIepOna B mHpOKOM BHana3OHe H3MeHCHHR yC,IOBHii B Cone. nOKa3aH0, 'IT0
Tennoo6MeH MOXCHO paccrHTarb c nOMOmbro 06btHHoro ypaBHeHHn nono6aa. ConepBtamero Uicno
HyccenbTa,eCnH KpHTepHH nono6en OCHOBaHbl Ha HHTerpanbHbIX TeMO+H3HWCKHX XapaKTepHCTHKaX
nna yveTa Hx H3MeHeHHs B tmipoKobi nHana30He. Cnenyer TaK*e yHeTbrBaTb aHoManbHoe H3MeHeHHe
TenBOnpOBOnHOCTH3KCnepHMeHTa,IbHbIeLK+HHbte COBnanaIoT C pe3ynbTaTaMH paWeTOB C TO'tHOCTbtO
no 104; naxtenpu HCnO."b30BaHHH npoeonorer O'ieHbManor0 .nriaMeTpa,KornaHeO6XODriMO BBOBHTb
nOnpaBOHHbIH KO+@HnHeHT.