Sie sind auf Seite 1von 20

Physica XII, no.

2-3 Juni 1946

HEAT TRANSFER FROM SPHERES


TO FLOWING MEDIA
by H. KRAMERS

Summary
The overall heat transfer coefficient of three small spheres in a forced
flow of air, water and oil was measured; in the latter two media at two
different temperatures. For this a stationnary method was used.
From these experiments the dependence of the heat transfer on the
velocity of flow and on the properties of the flowing substance is derived.
By using the dimensionless numbers characteristic of these quantities,
viz. Nu, ‘Re and Pr, the following generalized expression is obtained:
Nu = 2.0 + 1.3 (Pv)0.15 + 0.66 (PY)O.~I (Rc)O.~O.

For very small Re numbers this equation will give too high va lues of Nz5.
since for Re = 0 NU will be equal to 2.
Further it is shown that the known data on heat transfer fro m cylinders
and wires of circular cross section in a forced flow perpendicular to the
axis can be described very well by:
Nu = 0.42 (PY)O.~’ + 0.57 (PY)‘.~~ (Re)0.50

As to the dependence on the Re and Pr number the character of both


relations given here is the same.
Finally some theoretical results concerning these problems are dis-
cussed, from which it appears that theory has but partly succeeded in
explaining the experimental facts.

5 1. Introduction. Many technical processes involve the exchange


of heat or matter between small solid or liquid particles and a con-
tinuous fluid phase. If the phases are at rest - each in itself and
with respect to one another -, the exchange is governed by such
properties as the specific heat, the heat conductivity and the dif-
fusion coefficient. Most of these cases can be approached by cal-
culation. In general, however, the phases are not at rest and conse-
quently the rate of transfer will be increased. In those cases where
the resistance between the particle surface and the flowing medium

- 61 -
Physica XII 5*
62 H. KRAMERS

is the determining factor for the transfer process, it is important to


know how this resistance depends on the size of the particle, its
velocity relative to the flowing medium and the properties of the
latter.’
With regard to the heat transfer from spheres and small particles
to flowing media a few authors ‘) *) have recorded results which
were obtained by non-stationary measuring methods. As such me-
thods cannot give very accurate results and these experiments were
restricted to some special cases, no general description of this heat
transfer process can be based on their data. Nor has any quantitative
information of this kind been derived from purely theoretical con-
siderations.
Therefore measurements have been carried out to obtain the
heat transfer coefficient u for a sphere as a function of the velocity
z, of the flowing medium and for different values of the sphere radius
Y and the characteristic properties of the flowing medium: the heat
conductivity A, the specific heat cp, the density p and the kinematic
viscosity p/p = v.
These results can be generalized to other cases, as according to
the known similarity considerations 3) in the stationary state the di-
mensionless N u s s e 1 t number, containing the heat transfer coef-
ficient,

has to be a function only of the two dimensionless quantities the


R e y n o 1 d s number, characterizing the conditions of flow, and
the P r a n d t 1 number, combining the properties of the flowing
substance :
Re = ?!
V

This applies only as long as ,,forced flow” exists and the hydro-
dynamical conditions are not affected by temperature differences
(,,free convection”).
When the expression
Nu = f(Re, Pr)
has once been obtained from measurements of a as a function of v
for a set of values of the parameters Y, A, v, p and cp, this same rela-
tion can then be used to calculate the heat transfer coefficient for
HEAT T.RAh’SFER FROM SPHERES TO FLOWIKG MEDIA .63

other situations provided the condition of geometrical similarity


is satisfied.

5 2. Experimental. M e t h o d. If in a sphere of radius r a con-


stant quantity of heat q is generated per unit of time and volume,
this will result in the stationary state in a constant mean tempera-
ture difference At between its furface and the undisturbed flow. The
overall heat transfer coefficient u is then defined by:

Q = 5 w3q = r~ . 4nr2At.
As At is an average value, a should also be considered as such.
Ry measuring Q and At under certain conditions of flow and for
certain sphere sizes we determined the corresponding values of CC
As we were particularly interested in u values at low Re numbers, it
was desirable to keep the size of the spheres to be used as small as
possible. This was effected by making use of high frequency heating,
thus avoiding the rather large dimensions and other difficulties con-
nected with direct electric heating.

Fig. 1. Arrangement for the measurements in liquid flow.

Apparatus. The measuring apparatus is represented dia-


grammatically in fig. 1. A small metal sphere is suspended by thin
thermocouple wires inside the cylindrical coil A. When a high fre-
64 H. KRAMERS

quency current is passed through the coil, the sphere will be heated
as a result of eddy current losses and, if the sphere material is ferro-
magnetic, of hysteresis losses.
For measurements in liquids a liquid flow round the sphere
is obtained in the glass tube B. In this tube, which is 3.4 cm wide,
the liquid flows upward. After leaving the tube it passes through a
measuring cylinder C and a thermostat and is introduced again at
the bottom of the tube. The temperature and the velocity of the
liquid flow can be regulated.
For measurements in air a fan is placed
to compensator under coil A, giving an upward air flow
(fig. 2). The air velocity can be varied by
means of the adjustable diaphragmI). In
order to suppress the larger whirls caused
by the fan, a grid E is placed above it.
The determination of the various quan-
. tities necessary for the evaluation of a will
:A
:
: be discussed below in detail.
:
I
I .
.. :
The temperature difference.
We used three annealed steel spheres (ball
bearings) I, II and III, with a diameter of
1.26, 0.787 and 0.709 cm respectively. A
hole was drilled in the spheres and the
junction of a copper - constantan ther-
mocouple was fastened with solder in the
centre. The thermocouple wires had a
thickness of 0.1 mm; it had been carefully
Fig. 2. Arrangement for calibrated at different temperatures. The
the measurements in .air other (;,cold”) thermocouple junction was
flow. placed in the undisturbed flow. The e.m.f.
resulting from the temperature difference
At between the two junctions was measured by means of a Dies-
selhorst potentiometric compensator.
It should be observed that it is only permissible to use the value
of At, measured in this way, for the calculation of a, if the tempera-
ture differences in the sphere are small compared with At. It can
be shown that the temperature difference At’ between the centre
HEAT TRANSFER FROM SPHERES TO FLOWING MEDIA 65
.
and the surface of a uniformly heated sphere is represented by:

At’ - At -& N_ZI,


s
in which h, and A, stand for the heat conductivities of the surround-
ing substance and of the sphere material respectively. As will be
seen from the results, this correction should be taken into account
only for the experiments in water in consequence of its rather high
thermal conductivity. In that case, however, the correction was
only applied to the largest sphere. Here the thermocouple leads to
the ,,hot” junction were thermally isolated from the sphere, so it
could be assumed that this junction really indicated the temperature
in the centre of the sphere. In the two smaller spheres the place
where the temperature was measured was not so well defined.

T h e h e a t i n g e n e r g y. The magnetic field coil A consisted


of 66 turns of rectangular copper wire (2.5 x 3.0 mm) wound on a
pertinax cylinder with a diameter of 11 cm and a length of 30 cm.
Several additional turns were laid at the ends in order to obtain
a constant field strength within 1 per cent over a distance of 5 cm
in the centre of the coil. The high frequency tension was taken from
several turns of the generator coil of a 2 kW ultrasonic transmitting
installation. It had.a frequency of about 2.6 x lo5 cycles per second
corresponding to a wavelength of 1150 m. The tension could be
varied by regulating the anode tension of the transmitting valve.
The strength of the current passing through the coil was measured
by means of a vacuum thermocouple, which was shunted by a low
resistance.
The maximum effective field strength attainable inside the coil
was about 20 Gauss at a current of 10 A and a tension of 3000 V.
The sphere material being ferromagnetic, a sufficient amount of
heat was generated in the spheres at these rather low field strengths.
The quantity of heat Q developed per unit of time had to be
determined for each sphere separately as a function of the current
through the coil. For this purpose we used a small water calorimeter
consisting of a thin-walled test tube surrounded by a vacuun jacket
for thermal insulation; this was not, however, provided with a re-
flecting metal layer, as this would give rise to additional heating
resulting from eddy currents in it. The water in the calorimeter was
stirred by moving the sphere up and down by mechanical means.
Physica SII 5
66 H. KRAMERS
.-

Although these calorimetric measurements were carried out on a


rather small scale, it was possible to determine the relation between
Q and the current through the coil to within a few per cent.
For a sphere with a magnetic permeability which is independent
of the field strength H, Q may be expected to be proportional to
H2 4). For the spheres used here the proportionality factor (= Q/H2)
was found to diminish about 20 per cent between 5 and 20 Gauss.
Presumably this was caused by the decrease of the effective magne-
tic permeability in this region.

T h e f 1 o w v e 1 o c i t y. For the determination of the velocity


v in air flow a P i t o t tube was used, which was placed at the place
occupied by the sphere before and after each experiment. The pres-
sure difference was measured with a F u e s s micromanometer for
v > 150 cm/set; for velocities between 40 and 150 cm/set this was
done by an electrical apparatus specially designed for measuring
small pressure differences and consisting of a membrane condenser
with an amplifier.
Direct measurement of the velocity of the liquid flow at the
spot occupied by the sphere was hardly possible. Therefore v had
to be derived from the quantity of liquid flowing through the tube
A per unit of rime which gives the mean velocity c in the tube. The
value of v to be used for the computation of the Reynolds number
is proportional to 5, the ratio v/C depending on the velocity distri-
bution in the tube and, to a minor extent, on the dimensions of
the sphere and the tube.
To establish a well defined condition of flow the tube A had to be
rather long and comparatively narrow; at the bottom of the tube a
glass wool filter F was placed (see fig. 1). The velocity of the liquid,
just after leaving F, is constant throughout the cross section of the
tube and has the value 6. As the liquid rises this uniform distribu-
tion gradually becomes parabolic, if the flow is laminar. If L re-
presents the distance from F in the direction of the flow, and D the
tube diameter (here 3.4 cm), then according to several authors “)
a parabolic distribution is approximated within a few per cent, when
5D2
Lv < about 16.

In that case v along the axis of the tube is equal to 26. In our
experiments in oil this condition was satisfied, as the distance from
HEAT TRANSFER FROM SPHERES TO FLOWING MEDIA . 67

the filter to the sphere was about 110 cm. With water parabolic
distribution had not been reached at this height for 5 > 2.5 cm/set,
but in all cases the flow was laminar. For the calculation of u in the
non-parabolic cases we used data furnished by N i k u r a d s e “)-
concerning the transition range from uniform to parabolic velocity
distribution in cylindrical pipes.
The fact that the sphere size relative to the tube diameter is not
negligible calls for two corrections to be made in ZI. First, it is more
correct to use in the calculations the average velocity over the cross
section of the sphere than the velocity along the axis of the tube.
For a parabolic distribution this gives the correction factor
[l - 3 (2r,/D)2] in 21.In the second place the presence of the sphere
itself in the relatively narrow tube causes a disturbance in the velo-
city field, resulting in an increase of the velocity near the sphere.
We tried to account for this by introducing the correction factor
[l + (2v/D)*], as the increase in 5 will be proportional to the ratio
of the areas of the sphere and tube cross sections. Thus the value of
u in the liquid flow was obtained with the following formula:
v=+ +p(g)2).5.
Here ‘p = 2 for parabolic velocity distribution and y < 2 in case
this has not yet been reached.
4 3. The mensurements. The hear rransfer of rhe three spheres was
investigated in five different media, viz. air at 1 atm. and about 20”
C, water at 6.5 and 19.O”C and a spindle oil at 19.0 and 34°C. For the
calculation of the characteristic numbers Nat, Re and Pv the values
of A, cp, p and v of these substances under the circumstances men-
tioned above had to be used. In the case of air and water these
values are sufficiently known. For the oil used they were determined
at different temperatures, the results of which are shown in table I.
TABLE I
Properties of spindle oil used

t 1 20 / 30 1 40 j "C

A.108 0.367 0.364 0.361 calpl SW "C


Y 0.348 0.224 0.153 cm ]sec
CfJ 0.456 0.474 0.516 ml/g "C
P 0.842 0.836 0.826 g/cm~
-364 Pr 242--- 177 -
68 H. KRAMERS

For the evaluation of the heat transfer coefficient at each condi-


tion of flow at least four determinations of At were carried out at
different energies of heating, At varying from 4 to 20°C in air, 1 to
10°C in oil and 1 to 5°C in water. The value of a0 at At = 0 was ob-
tained by lineair extrapolation. In this way, all quantities which
were necessary for further calculation could be taken at the tem-
perature of the undisturbed flow.
This method was also adopted to eliminate a possible influence
on the forced flow round the sphere resulting from free convection,
which may cause the value of u to rise with increasing At. Theory
shows that in cases where free convection must be taken into account
the Nu number will also depend on the value of the G r a s h o f
number

where p is the coefficient of thermal expansion and g the acceleration


of gravity. According to E 1 e n b a a s “) heat transfer from a
sphere to a surrounding medium by free convection only may be
described by the formula
Nat3
(1 - 2/N+’ - GrJ

which states that Nu increases with increasing value of Gr. Thus it


may be expected that a dependence of u on At, caused by free
convection, will exist only in cases where the forced velocity ZI
and the dynamic viscosity *I are small and I and p are large.
In our measurements at relatively large Re numbers x was
found to be independent of At, so in these cases no influence of free
convection OI of the variation of the properties of the flowing sub-
stances with temperature could be perceived. At lower velocities,
however, an increase of u with At was found originating from free
convection. This effect appeared to be the grea.test in the experi-
ments with the largest sphere in the medium with the lowest vis-
cosity, viz. water. This agrees with the above mentioned consider-
ations as to the appearance of this effect. The go values in which the
dependence of M on At was clearly observed have been starred (*)
in the tables II to V which contain the experimental results. In these
tables a0 is expressed in Cal/cm2 set “C and v in cm/set.
HEAT TRANSFER FROM SPHERES TO FLOWING MEDIA 69
.
TABLE 11 TABLE III
-
Air 20°C PI = 0.714
-
Oil 19.0°C
- - -
l’r = 380
-
Sphere Sphere
10B.U” NIL z Rt!
IX.
-
nr.
- - 1
0.8 I 16.9 65 540 3.42; 11.7 0.34 2.65
1.17 24.4 155 1290 4.42* 15.2 0.72 5.53
I I .32 27.6 199 lb60 5.348 18.4 1.12. 8.55
1.40 30.4 252 2100 I 6.40 22.0 1.93 14.7
.- _ 7.00 24.0 2.52 19.2
0.89 11.8 40 210 7.50 25.8 3.12 23.7

II
1.20
1.53
15.6
20.2
85
156
450
840 --
8.00
_
27.5 3.47
.- 26.5

1 .b9 22.3 204 1080 4.61* 9.9 0.27 1.22


1.83 24.2 252 1340 6.00 12.9 0.66 2.90
.- - 6.70 14.6 1.02 4.59
0.97 11.5 40 190 II 7.64 lb.4 1.58 7.10
1.17 13.8 65 300 a.48 18.2 2.07 9.32
1.43 lb.9 110 510 9.45
-
20.3 2.85
.- 12.8

5
III 1.56 18.6 135 640 -
1.72 20.4 197 920 3.86* 7.5 0.10: 0.42
I .99 23.6 270 1280 5.22’ 10.1 0.30 1.23
2.19 25.9 310 1460 5.40 10.4 0.36 1.45
- - III 6.90 13.3 0.90 3.64
8.28 lb.0 1.61 6.50
9.90 19.1 2.83 11.4

-
I 0.05
-
19.5 2.03
- 11.4

With reference to the results shown in these tables, the following


should be remarked.
The measurements in air are the most accurate and have a pos-
sible error of -c 29/, in Nu and in v/Re. The results obtained in oil
flow are estimated to have a probable error of about f 3%. In oil
flow-at 34°C the measurements could not be extended to such low
velocities as at 19”C, because in these cases the oil cooled too much
near the tube wall. The’resulting increase in velocity changed the
velocity distribution to an extent which could not be neglected.
In the experiments in water flow difficulties were encountered
owing to the fact that the spheres were almost immediately covered
with small air bubbles. As a result, the a values obtained were ge-
nerally too low. Moreover, at large Re numbers the velbcity was
evaluated with less certainty than in case of parabolic distribution.
For both reasons in the results obtained at 19°C an error of up to
70 H. KRAMERS

TABLE IV TABLE V

Oil 34’C P’r = 213 Water 6.5’C Pr = 10.7


T

Sphere
103.Q NU ii Re
nr.
-
4.47* 15.7 0.68 IO.1 l5.6* 14.5 0.47 90
5.70 19.8 1.33 19.4 24.6’ 22.8 1.29 246
I 5.92 20.7 1.58 23.2 I 35.0 32.4 2.79 520
7.42 25.9 2.71 39.6 38.6 35.7 3.20 590
-- -- 49 45 5.35 920
6.55 14.3 0.80 6.95
7.04 15.3 1.13 9.75 28.6 16.5 1.06 118
II 7.64 16.6 1.53 13.2 30.6 17.7 1.26 141
9.34 20.4 2.72 23.6 39.3 22.7 2.20 246
-- -- II 49.8 28.8 3.44 370
7.61 14.9 1.14 8.79 57.2 33.1 5.10 520
III 8.61 16.9 1.67 12.8 73.4 42.2 8.40 800
9.95 19.5 2.72 21.0 --
24.8 12.9 0.65 65
34.8 18.1 1.13 113
TABLE VI 35.3 18:4 1.27 127

Water 19.O”C
- -
Pr = 7.3
- 1 III
47.2
51.1
59.7
24.6
26.6
31.1
2.35
3.33
4.30
234
320
400
Sphere
nr.
*o”,a,

29.2*
I
2-
NU

25.9
-
1.88
-
RC

505
74.0
80.6
38.5
42.0
7.8
10.6
670
870

35.6 31.4 2.65 680


45.0 39.9 4.1 1’040
-- -- _-
20.0* 11.1 0.36 57
27.8* 15.6 0.86 136
33.0 18.3 I.15 182
50.7 28.0 3.90 580
.-
23.6* 11.8 0.37 52.2
27.8* 13.8 0.57 79
33.62 16.8 0.90 127
35.0 17.4 1.15 162
III
47.5 23.6 2.65 370
54.0 26.4 3.62 485
55.8 27.8 3.88 520
-

10% is thought possible. At 6°C air bubbles could be largely avoided


by using water direct from the mains instead of from the thermostat.
These results are estimated to be accurate to within f 504.
HEAT TRANSFER FROM SPHERES TO FLOWING MEDIA .71

5 4. Discussion of results. The results are shown in figs. 3 ‘to 6,


in which NW has been plotted against 4Re. As can be seen from these
graphs, for each medium the relation between Nu and Re is indeed
independent of the sphere diameter. The linear relations which are
found to exist between Na and 1/Re can be represented by the fol-
lowing formulae :
Air 20°C Pr = 0.714 Nu = 3.2 + 0.59 (Re)’
Water 19”C, Pr = 7.3 Nu = 4.0 + 1.03 (Re))
Water 6”C, Pr = 10.7 Nu = 3.4 + 1.31 (Re)*
Oil 34°C Pr = 213 Nu = 4.9 + 3.32 (Re)*
Oil 19°C P7 = 380 Nu = 5.0 + 4.32 (Re))

30

NU

10
,
,’
,’
8’
0 -..-. ..
I 200 2 8 I I
0 10 30 40 50
VKG

Fig. 3. Measurements in air.

These formulae only describe the experimental results covering


the range from Nu = 8 up to Nu = 40. For Re = 0 they give values
for Nu varying from 3.2 to 5.0 instead of the theoretical value
-Nu = 2 for this problem of a heated sphere in an infinite medium at
rest. To account for this difference we will change the experimental
formulae into
Nu = 2 + f + s(Re)f.
The constants f and g will be a function of the P r a n d t 1
number. It appears to be possible to represent g to within 5% by
the relation
g = 0.66 (Pry.
72 H. KRAMERS

Only the value for water at 19°C appears to be too low by 1So/b.
This will probably be caused by the air bubbles on the sphere surface.
Although with less accuracy, f can be represented by a similar
expression :
f = 1.3 (P,)O.?

“0 2 4 6
-fiG
Fig. 4. Measurements in spindle oil.

Generally the contribution of f will be small compared to the


last term g (Re)*.

30
Nu

f 20

Fig. 5. Measurements in water, 6°C. Fig. 6. Measurements in water, 19°C.


HEAT TRANSFER FROM SPHERES TO FLOWING MEDIA 73

In this way the following empirical equation between Nu, Re and


Pr is obtained from our experiments for the heat transfer from sphe-
res to flowing media:
Nu = 2.0 + 1.3 (Pr)0.15 + 0.66 (PY)O.~’ (Re)0.50.

This relation is verified to within f 10% in the range 10 < NU < 40


and 0.7 < Pr < 400. For Nzl < 10 it gives too high values for
Nzc. The formula can probably be used up to Re numbers of
\ about 105. As has been shown by the measurement of the hydro-
dynamic resistance of spheres, above this Re value the type of flow
changes considerably, so that a further extrapolation of this formula
will not be permissible.

9 5. Comparison with other investigations on spheres. L j a-


ch o w s k i “) and V y r o u b o w 7) have observed the heating of
small metal spheres in a flow of hot air and have calculated from
these non-stationary measurements the heat transfer coefficient u.
Their results are respectively :

a = 3.4 -g t) (G)‘, 200 < Re < 30,000,


and
0.092
a = 3.02-- 200 < Re -C 3000,
100

all quantities being expressed in technical metric units. Introducing


Nu and Re in these equations we get for air at room temperature:
Nu = 0.61 (Re)) (Ljachowski)
and
Nu = 0.58 (Re)* (Vyroubow).

The dependence on temperature found by V y r o u b o w cor-


responds approximatively to a proportionality of Nu with (Pr)“.
L j a c h o w s k i’s and our own results in air are compared in
fig. 7, where they are represented by the curves a and b respectively.
From Re = 1000 onwards there appears to be good agreement.
But for smaller Re values L j a c h o w s k i’s and especially V y-
r o u b o w’s results are lower than ours.
Data on heat transfer in air flow for very small Re numbers
have been obtained by J o h n s t o n e et al. ‘) and also by L j a-
c h o w s k i 2) from experiments where clouds of small particles
74 H. KR4MERS

were dropped through a heated air zone, after which the rise in
temperature was measured in a calorimeter. The results of -both
authors show a considerable spread. A few of the data in J o h n-
s t o n e’s paper point to Nu values which are less than 2. A syste-
matic effect may have played a part here, this probably being a
circulation of air inside the vertical heating space resulting from
cooling and entrainment by the falling particles. When the velocities
of this ,,reinforced convection” become comparable to the falling
speeds of the particles, this effect may lead to too low Re numbers,
if it is not accounted for. In particular for the heating process as a
whole, investigated by these authors, their conclusions are very

b,._.__._
-.Alr ;

c.,._._......
water 100-c ;ownr*sulh
-crpcr1msnta1
sl+-H-t ------E.L cul.tcd I
OI
0 100 200 500 1000 2wo 5oal lo.ooo 2c.000

-Re
Fig. 7. Comparison with L j a c h o w s k i’s results for spheres.

important. They give, however, no reliable information on the heat


transfer of spheres or small particles individually at very low Re
numbers.
The heat transfer from spheres to flowing water has also been
studied by L j a c h o w s k i, who measured the temperature rise
of small steel spheres and mercury drops, after these had dropped
through a column of hot water. On an average his results can be
described by
Nu = 0.085 (R'e)O.'$ 3500 <Re < 15,000,
the mean deviation from this relation being about f 40%. For
purposes of comparison with this formula (curve c in fig. 7) the lines
HEAT TRAKSFER FROM SPHERES TO FLOWIXG MEDIA 75
.

d and e are drawn representing the values of Nu to be expected from


our own results for water at 6°C (Pr = 10.7) and 100°C (Pr = 1.75)
respectively. There is agreement between c, d and e in so far as c
lies in between d and e. That L j a c h o w s k i’s results for re-
latively small Re numbers, i.e. for small spheres and low rates of
fall, are nearer to the curve e for 100°C might be explained in the
following way; the smaller the spheres are, the quicker they will
assume the temperature of the surrounding liquid. Accordingly,
the results for high Re values will be nearer to curve d, on the low
temperature side.
Finally, we may mention the interesting experiments carried
out by F r B s z 1 i n g B), who measured the rate of evaporation
from small drops of nitrobenzene, aniline and water and from very
small naphthalene spheres in air flow. Since the laws governing
this diffusion process are essentially the same as in case of heat
transfer, we may draw conclusions from F r ij s z 1 i n g’s results
with respect to the heat transfer from spheres by replacing concen-
trations by temperatures and the diffusion coefficient by the quan-
tity ?,/c,,p. In this way the, following relation for Nu is found:

Nu = 2.0 + 0.55 (Pr)* (Re)* , o z 2 Fr 2 y!


. .

Here too the part of Nu which depends on the flow velocity is pro-
portional to (Pr)*. The numerical coefficient of this part, however,
is about 20% lower then was found by us, whereas a small term
dependent on Pr only is not present. In our opinion the absence of
this term is surprising, the more so, since F r ij s z 1 i n g’s experi-
ments extend to very low Re numbers. An explanation for this
discrepancy cannot be offered at the present, but it should be point-
ed out that F r ii s z 1 i n g measured the ratio between the rate of
evaporation in a forced air flow and in air at rest, assuming in the
latter case Nu to be equal to 2. Thus, a slight deviation from this
condition might affect the interpretation of the results considerably.

5 6. Heat transfer from cylinders to flowing media. In 1932 U l-


s a m e r “) reviewed all available experimental data on heat transfer
in the case of infinitely long cylinders and wires of circular cross
section in a forced flow perpendicular to the axis. As results in air,
water and three different oils were present, the influence of the value
76 H. KRAMERS

of Pr could be investigated. U 1 s a m e r showed that these data


could be represented by the following general expressions:
Nu = 0.91 (PY)O.~’ (R~z)O.~~~, 0.1 < Rc < 50
and
Nu = 0.60 (PY)O.~’ (RET)O.~O, 50 < Re < 10,000.
When plotted on a double logarithmic scale these formulae show two
straight lines which intersect at Re = 50. Consequently this des-
cription involves a discontinuity in the value of d(Nw)/d(Re), which
does not exist in reality.
Now we found that the data from U 1 s a m e r’s paper could be
written with sufficient accuracy in the same form as we used for
spheres, viz.
Nu = 1’ + g’ (Re)‘,
f’ and g’ being functions of Pr. Thus, the following relations were
obtained, which cover the experimental facts very well in the region
2 <Nu <20:
Air, Pr = 0.71 Nu = 0.4 + 0.52 (Re)*
Water, Pr = 7.1 Nu = 0.7 + 1.05 (Re)t
White oil, Pr = 29.6 Nu = 0.6 + 1.75 (Re)*
Transformer oil I, Pr = 135 Nu = 1.1 + 2.9 (Re)*
Transformer oil II, Pr = 525 Nu = 1.6 + 4.6 (Re)*
Further, the values of f’ and g’ fitted very well in the expressions:
f’ = 0.42 (Pr)“.20 and g’ = 0.57 (Pr)“.33.
So a general formula for the heat transfer from circular cylinders
in a forced flow, which is based on the experimental data referred to
in U 1 s a m e r’s papen, is given by:
Nu = 0.42 (Pr)“.20 + 0.57 (Pr)“.33 (Re)0.50.
Since for Re = 0 the value of Nu must be zero, this quantity will
deviate from the value given by the above formula for Nu < about 2
In fig. 8 the five linear relations given above for different Pr
numbers have been plotted as Nu against (Re . Pr)S. The slope of
the straight lines (= g’ (PY)-*) becomes smaller with increasing
Pr. At the same time the value of Nu extrapolated to RE = 0 (= f’)
becomes greater. Fig. 8 contains also the theoretical results by
B o u s s i n e s q and K i n g (to be discussed in the next section)
which may be considered to apply to the case where Pr = 0.
HEAT TRANSFER FROM SPHERES TO FLOWING MEDIA 72

Fj7. Theoretical From the above it appears that the ex-


results.
perimental data on the heat transfer from spheres and cylinders in
a forced flow can be represented by similar expressions. Both con-
tain a part proportional to (Pv)” (Re)), which ‘prevails at large Re
numbers, and a small term, which depends but slightly on Pr. The
latter term vanishes at very low Re numbers, since in this region Nu
apparently increases quicker with increasing Re numbers than at
high Re values.

l;ig. 8. Theoretical and experimental results for the heat transfer of


cylinders in clifferent media.

A theoretical treatment of problems of this kind has proved to be


extremely difficult. For the calculation of the amount of heat trans-
ferred from the body to the fluid flow the temperature distribution
in the latter has to be known. For this a partial differential equation,
containing the velocity distribution around the body, with given
boundary conditions has to be solved. In the exceptional case, where
the velocity field around a sphere is known. i.e. the distribution
according to S t o k e s for very low Re numbers, this differential
equation of the temperature field is hardly soluble *).
Nevertheless, several authors have tried to obtain theoretical

*) F u c h s 10) has shown that for Re < I in first approximation Nlc will not differ
appreciably from the value at Rr = 0, if S t o k e s’ distribution law is t&en to be valid.
This is not supported by the experimrntai facts mentioned in this paper, which on the
contrary suggest a quick rise in Nu at very low Re numbers. Even F r 6 s z 1 i n g’s
results do not point in that direction.
78 H. KRAMERS

expressions for the heat transfer from spheres and cylinders starting
from certain simplifying assumptions regarding the velocity field
around the body. These assumptions may be divided into two groups
viz. potential flow and a velocity field according to P r a n d t l’s
boundary layer theory. From both theories the heat transfer appears
to be proportional to (Re)*.
P o t e n t i a 1 f 1 o w can only be realized with incompressible
fluids having no viscosity. Thus, the heat transfer number Nu must
be a function of Re x Pv = Pe' (P C c 1 e t number), since this
dimensionless group does not contain the viscosity.
B o u s s i n e s q’l) has calculated the heat transfer from bodies
of different shape in a potential flow. After an elegant transforma-
tion and assuming the temperature gradient in the direction of the
flow to be negligible as compared with the gradient perpendicular
to it, i.e. for low heat conductivities and high velocities, he obtained
the following results, which are expressed here in dimensionless
form :
for spheres: Nu = 1.13 (Pk)*
for cylinders: NZL = 1.62 (Pe')'.

King 12) has solved B o u s s i n es q’s problem for cylinders


without neglecting the variation of temperature in the direction of
the flow, but using a different boundary condition on the surface of
the body. Whereas B o u s s i n e s q supposes the streamlines
touching the surface to assume immediately its temperature,
K i n g’s calculations are based on the assumption that the heat
flow per unit area is constant. His results can be expressed by
2
Nu = Pt! < 1
1.12-ln Pe' '
and
Nu = 0.32 + 0.80 (PA)', Pe'> 1.
These formulae represent a type of relation between Nu and Re,
which agrees rather well with the experimental data partly from
K i n g himself and referred to by U 1 s a m e r in his paper “).
It should be pointed out, however, that the boundary condition
assumed by K i n g is quite contrary to facts, for it has been shown
by experiments that the heat loss of a cylindrical surface in a forced
flow is much larger on the side facing the flow than on the opposite
side (see e.g. E c k e r t 12)).
HEAT TRANSFER FROM SPHERES TO FLOWING MEDIA 79

For completeness’ sake it should be mentioned here that J o h n-


s t o n e, P i g f o r d and C h a p i n 1) derived a formula for the
heat transfer from spheres in a forced flow. By making the hydro-
dynamically impossible assumption that the fluid flows every-
where parallel to the surface with a velocity equal to the undisturbed
velocity they arrived at an expression, which for high Rti numbers
amounts to
Nzr = 0.714 (Pe’)‘.
For low values of Re, however, the character of their theoretical
curve deviates largely from the experimental data.
The theory of the laminary b o u n d a r y 1 a y e r, which states
that the principal change of the relative velocity from the value
at the surface of the body to the value in the undisturbed flow takes
place over a distance which is small compared with the dimensions
of the body, allows the hydrodynamical equations to be simplified.
In many cases the boundary layer theory has given correct results
regarding the frictional resistance in a forced flow, and it has proved
to represent the actual conditions very well at large Re numbers.
In this region it also applies to the distribution of temperature (or
concentration).
Many authors have made use of the boundary layer theory in the
treatment of transfer problems, but quantitative results are scarce,
because these calculations remain complicated. E c k e r t 12) and
F r 6 s z 1 i n g 11) have shown that in cases, where this theory may
be assumed to apply (i.e. at high Rc values), the rate of transfer of
heat will be proportional to (RE)* (PY)“, where n is very near to 1/3.
This is in accordance with the experimental results available. Both
authors have indicated rather complicated methods for the quan-
titative calculation of heat transfer from spheres and cylinders in
this region.
Its results would still be somewhat uncertain, because on the
offstream side of spheres and cylinders the boundary layer is
disturbed by whirls, so that the calculations cannot be extended
over this region. But as the principal amount of heat is transferred
at the ,,front” side of these bodies, the general dependence on Re
and Pr will remain valid.
Summarizing we may state that, although solutions based on the
theory of potential flow have been given, these do not agree with
he experimental results, neither quantitatively nor with respect to
80 H. KRAMERS, HEAT TRANSFER FROM SPHERES TO FLOWING MEDIA

the dependence on Pr number, because the type of flow assumed


here does not correspond to the actual velocity distribution. The
boundary layer theory, on the other hand, gives results which are
correct, at least in a qualitative way and for relatively high Re
numbers. As yet no theoretical explanation has been found for the
experimental fact that at very low Re numbers N% increases more
quickly than at higher Re values, so this region, which is also of
practical interest, is still open to theoretical investigation.

In conclusion, the author wishes gratefully to remember the


late Prof. Dr. E. C. W i e r s m a, who encouraged these investig-
ations greatly by his original suggestions and stimulating interest.
Further the author is very much indebted to Prof. Dr. J. d e
B o e r, Amsterdam, for his valuable advice in the preparation of
this paper, and to the ,,Technisch Physische Dienst van T.N.O.
en T.H.“, Delft, and the N.V. ,,Chamotte Unie”, Geldermalsen,
who kindly permitted the publication of these experiments, which
were carried out at the Laboratory for Technical Physics, Delft, in
the end of 1943.

Received March 20th, 1946

REFERENCES

1) H. F. Johns t o nr, R. L. Pig f ord and J, H. C h a pi n, Trans. Am. Sot.


them. Eng. :$7, 95, 1941.
2) D. N. L j n c h o w s k i, J. techn. Phys., Leningrad, 10, 999, 1940.
3) H. G r 6 be r and S. E r k, ,,Die Grundgesetze der Wgrmeiibertragung”, Berlin,
1933, p. 188.
4) M. D i vi 1 k o v s k i, J. Phys. USSR, 1, 471, 1939.
5) L. P r a n d t I and 0. Tie t j e n s, ,, Hydro- und Aeromechanik” II, Berlin 1931,
p. 27 and further.
6) W. E 1 e n b a a s, Physica 9, 285, 1942.
7) D. V y r o u b o w, J. techn. Phys., Leningrad, 9, 1923, 1939.
8) N. F r 6 s z 1 i n g, Gerl. Beitr. zur Geophysik, BZ, 170, 1938.
9) J. U 1 s a m e r, Forschung, 3, 94, 1932.
10) N. F u c h s, Phys. 2. Sowjet Union, (i, 224, 1934.
11) J, B o u s s i n e s q, Liouville Journal, 1, 285, 1905.
12) L. V. 1~ i n g, Phil. Trans. (A), ?l/+, 373, 1914.
13) E. E c k e r t, Forschung 1942, Forschungsheft 416.
14) N. Frb;szling, Lunds Univ. Arssltr., ~6, nr. 4, 194~.

Das könnte Ihnen auch gefallen