Heat Transfer and Pressure

Drop of Liquids in Tubes

Data are presented on heat transfer to
three oils of widely different temperature
coefficients of viscosity in viscous flow
in a tube for both heating and cooling.
The results are correlated in terms of
main-stream fluid properties by a method
which takes into account the viscosity
gradient of the fluid in the tube by means
of the ratio p a / p w , where pa is the viscosity of the fluid at its main stream
temperature and pw is its viscosity at the
temperature of the tube wall. The data
for heating and cooling are brought
into agreement on this basis. The method
of correlation is then applied to data
in the literature on heat transfer and
pressure drop in both the viscous and
turbulent regions. Again heating and
cooling data are brought into substantial agreement.
Application of the
method of correlation to the design of
heat-exchanger equipment is discussed.

E. N. SIEDER AND G . E. "ATE

Foster Wheeler Corporation, New York, N. Y.
Because of the lack of an adequate theory incorporating
this modification of velocity distribution, empirical correlations using dimensionless groups are employed where designers
must consider a wide variety of fluids. The correlations a t
present available fall in general into two classes-one using
main stream properties and the other using film properties.
The former results in two curves, one for heating and another
for cooling, while other effects of temperature difference between tube wall and fluid are inadequately taken into account.
The use of film properties results in a single curve for heating
and cooling only if the data under consideration are taken
from fluids with nearly the same temperature coefficient of
viscosity. Therefore the method of determining this socalled film temperature should vary for each fluid of different
temperature coefficient of viscosity.
Colburn (3) obtained good correlation of data on heat
transfer and pressure drop by employing film properties,
but he had to supplement his correlation by introducing the
ratio p,/p,/. Not only did he combine this ratio with heat
transfer and pressure drop moduli to compensate for the ef-

ESIGKERS of heat transfer equipment

have long felt a need for heat transfer

and pressure drop correlations which
would give identical formulas for heating and cooling and
a t the same time evaluate the effect of temperature difference.
Furthermore, they would consider satisfactory only such correlations as use the same dimensionless groups computed
in the same way throughout all phases of flow. A method
of correlation is presented here which satisfies the above
conditions when applied to liquids. Originally devised for
the viscous region, it was found possible to extend it over the
turbulent. The proposed method offers a further simplification in that it makes use of main stream properties.
The theoretical formulas, such as those of Graetz (6)
and LBvhque ( l a ) for heat transfer to fluids in viscous flow
in tubes, do not take into consideration the effect of a radial
temperature gradient on the distribution of the axial and
radial components of velocity. Because of the magnitude
of the temperature coefficient of viscosity of many liquids, a
large radial viscosity gradient results, which effects a distribution of velocity considerably different from that occurring in isothermal viscous flow. The viscosity gradient has
opposite signs for heating and cooling. Heat transfer and
pressure drop data for such cases, when correlated by a means
which does not take into account this gradient, fall into two
plots. Furthermore, the spread of points for each plot is
unsatisfactory, all of which reveals the striking inadequacies
of these simplified theories. Hence, theories which neglect
this modification of velocity distribution are inadequate for
correlating heat tranqfer and pressure drop data,

FIGURE2
0 Oil A heating
0 Oil A: cooling

1429

Oil B, heating
Oil C. cooling

INDUSTRIAL AND ENGINEERING CHEMISTRY

1430

VOL. 28, NO. 12

,
TABLEI.
Run No.

HEAT

TRANSFER
DATAO S THREEOILS
hD

TI

Ti

tl

22

Tw

DO

Cr

pa

PU

545
535
345
348

0.109
0.112
0.076
0.074

12.1
10.9
15.2
14.2

7
g
12
13

1306
1330
1820
1388

136.85
138.0
160.45
160.25

135.15
136.2
158.5
157.9

0.463
0.464
0.477
0.477

1.05
1.1
1.8
1.55

469
460
500
471

Oil .4,21' A. P. I.
71.6
74.0
1.13
71.8
74.0
1.01
72.2
75.2
1.50
72.2
75.2
1.41

19
20
23
25
26

231
239
457
916
905

157.75
157.5
212.8
205.5
205.0

149.5
148.45
203.2
200.4
200.0

0.473
0.473
0,505
0.502
0.502

0.98
1.02
2.22
2.35
2.28

248
242
507
487
480

72.85
73.0
81.9
79.5
79.0

76.5
76.75
85.7
83.8
83.4

0.91
0.91
1.93
2.10
2.12

77.0
78.0
89.0
86.0
85.5

14.2
14.6
19.6
21.6
21.8

10.5
10.7
14.6
16.1
16.2

90.4
90.8
545
1005
978

385
390
151
167
168

0.094
0.101
0.052
0.050
0.050

9.3
9.4
19.3
20.5
20.7

27
28
29
30
37
39

1348
1360
1850
1860
229
885

206.35
207.6
206.9
207.0
141.6
140.35

202.9
204.0
203.7
204.0
134.65
138.05

0.503
0.504
0.504
0.504
0.465
0.466

2.34
2.46
2.98
2.81
0.70
1.35

428
411
466
451
247
477

78.5
78.2
77.8
77.8
77.8
76.2

83.3
83.7
83.7
83.9
81.0
78.2

2.27
2.27
2.75
2.75
0.79
0.96

87.5
87.5
88.5
90 0
82.5
77.0

23.4
23.1
28.4
28.7
17.1
18.5

17.4
17.3
21.2
21.4
12.5
13.6

1523
1560
2100
2110
67.2
269

160
158
159
159
520
507

0.051
0.051
0,053
0.056
0.161
0.121

22.a
22.5
30.4
30.1
9.3
10.6

42
43
45
46

1820
473
469
460

147.5
79.6
80.2
80.0

146.0
84.75
86.5
82.0

0.469
0.430
0.431
0,431

1.3
1.93
0.38
0.41

474
350
517

75.85
122.5
140.0
140.5

78.3
119.5
137.45
138.25

1.16
105
1.28
1.16

79.8
118.5
136.0
137.0

20.9
34.9
29.2
25.0

15.3
25 0
21.0
18.0

655
22.6
23.8
20.9

436
3,350
3,160
3,620

0.114
3.43
5.28
6.02

12.7
6.5
5.25
4.2

47
48
49
50
51
52

475
471
492
487
478
495

82.2
82.1
83.1
83.4
83.3
83.5

89.15
85.3
92.0
87.0
92.75
88.3

0.433
0.433
0.433
0.433
0.433
0.434

0.66
0.65
0.82
0.76
0.89
1.03

532
528
507
503
528
525

150.5
150.1
158.4
160.5
166.05
166.85

1.44
1.32
1.90
1.81
1.96
2.10

147.0
146.2
150.4
155.0
164.2
165.0

28.2
25.5
31.5
31.3
29.9
32.0

21.2
18.0
22.6
22.5
21.7
23.0

26.6
24.7
29.6
26.5
29.2
27.8

2,780
3,100
2,510
2,860
2,470
2,770

114
115
116
117

239
239
465
465

61.0
61.0
65.0
64.5

77.3
80.1
76.9
74.4

0.432
0.432
0.435
0.435

1.69
1.98
2.41
2.00

562
555
562
572

147.9
148.35
153.05
153.0

147.8
147,6
155.65
156.9
169.75
170.86
24.5' A. P.
144.9
144.8
148.75
149.5

1.69
1.97
2.42
2.00

143.5
144.0
147.0
146.8

27.4
32.3
38.3
31.2

19.2
22.6
26.8
21.8

61.5
64.7
128
121

567
540
532
560

5.05
4.80
4.95
5.25

8.65
10.3
12.2
9.8

118
119
120
121

916
916
lSC0
1x0

68.8
68.5
69.0
69.0

74.15
74.65
73.95
74.2

0.437
0.437
0.436
0,436

2.15
2.47
2.58
2.73

577
573
573
568

153.2
154.5
153.0
152.8

149.5
150.2
148.5
148.0

2.14
2.46
2.58
2.73

146.0
147.5
146.0
145.0

84.6
39.1
41.8
44.9

24.2
27.3
29.2
31.4

256
256
335
335

625
522
523
522

4.90
4.90
4.90
4.80

11.2
12.7
13.5
14.6

122
123
125
126
127

1857
1E57
2EG0
3280
3260

70.0
70.0
71.0
72.0
72.0

74.2
74.55
75.0
74.65
74.9

0.437
0.437
0.437
0,438
0,438

3.40
3.69
4.54
3.75
4.14

548
532
555
648
550

154.2
157.35
162.1
161.9
161.9

148.0
150.4
154.8
155.0
164.8

3.40
3.69
4.05
3.78
3.90

143.5
146.0
148.5
149.0
149.0

57.5
60.4
64.6
60.3
62.1

40.2
42.2
45.3
42.3
43.6

531
531
763
958
969

513
509
497
493
492

4.55
4.80
4.85
4.90
4.90

18.9
19.8
21.4
19.8
20.6

128
129
130
131

3C65
3C65
2655
2630

84.0
84.0
84.0
84.25

86.0
86.35
86.5
86.85

0.444
0.444
0.444
0.444

2.72
3.20
2.94
3.04

742
727
720
720

152.5
153.0
153.0
151.8

54.4
58.1
54.3
45.9

38.3
41.0
38.2
32.3

1255
1260
1090
1085

372
37 1
370
367

3.70
3.75
3.75
3.60

20.6
22.1
20.9
17.5

10
11

1800
1786
2590
2630
2680

275.0
275.5
276.0
276.65
235.8

272.3
272.35
273.7
274.0
234.1

0.526
0.526
0.526
0.527
0.505

3.1
2.95
3.3
3.65
1.75

301
301
405
403
371

159.0
159 0
159.0
158.5
16' A. P.
92.35
92.7
91.2
91.0
89.0

3.04
3.27
3.06
2.52

7
8

163.1
163.5
163.25
162.0
Oil C,
83 8
83.9
83.5
83.25
82.8

2.58
2.65
3.12
3.12
2.30

99.5
100.5
97.0
96.5
93.5

17.8
18.4
21.2
21.1
19.6

14.2
14.6
16.8
16.8
15.3

678
678
992
1030
476

523
520
512
507
1,050

0.0052
0.0057
0.0044
0.0042
0.0080

17.0
17.3
20.8
21.0
13.8

13
15
17
19
21

2010
1330
974
39 1
431

237.0
236. 9
236.5
222.3
210.5

234.9
234.1
233.1
215.55
204.8

0.506
0.506
0.504
0.496
0.488

1.85
1.6
2.1
1.4
2.45

362
364
361
359
375

82.8
82.6
82.6
82.6
82.2

88.7
87.8
87.25
86.25
85.4

2.14
1.89
1.68
1.31
1.20

92.8
91.0
89.3
86.5
88.5

18.6
15.8
13.9
11.9
12.1

14.6
12.3
11.9
9.3
9.3

356
238
173
49.0
40.6

1,050
1,040
1,050
1,490
1,920

0.0077
0.0089
0.0063
0.0077
0.0114

13.2
11.3
11.1
7.5
6.4

23
25
26
27
29

835
1270
1293
1815
3010

203.7
199. 2
199.9
199.8
202.0

200.45
196.7
198.1
197.95
200.65

0.486
0.485
0485
0.485
0.486

2.2
1.05
1.15
2.0
2.0

366
359
356
357
353

81.5
81.5
81.5
81.5
81.4

85.1
85.8
85.6
86.05
87.0

1.32
1.54
1.46
1.63
1.97

88.5
88.0
88.8
89.6
92.0

14.0
16.9
16.0
19.7
21.7

11.9
13.1
12.4
15.3
16.8

68.1
93.0
97.4
135
242

2,200
2,450
2,340
2,350
2,220

0.0131
0.0140
0.0143
0.0152
0.0166

7.5
7.7
7.9
9.7
10.1

31
33
35
36
37

1822
3550
1550
1573
792

171.7
170.45
171.5
172.9
173.6

170.15
169.45
168.85
171.1
170.55

0.470
0.469
0.470
0.470
0.470

1.3
1.65
1.19
1.35
1.13

290
292
290
290
290

80.8
80.6
80.3
80.2
80.1

85.3
86.3
84.8
84.8
84.0

1.31
1.66
1.30
1.33
1.13

91.0
93.0
91.0
90.5
89.5

19.7
26.1
18.1
19.8
16.4

15.1
20.0
13.8
15.1
12.6

58.0
111
49.0
49.8
26.6

5,100
5,170
5,150
5,150
4,900

0.0387
0.0438
0.0390
0.0380
0.0334

6.4
8.3
5.8
6.4
5.5

39
45
46
47
48

423
564
546
360
360

169.0
143.8
142.9
140.05
139.05

164.3
140.9
139.0
135.75
135.0

0.467
0,454
0.453
0,451
0.451

1.2
0.97
0.96
0.69
0.66

269
311
310
311
311

80.05
77.6
77.6
77.6
77.6

83.5
80.0
80.0
79.85
79.8

0.93
0.75
0.74
0.70
0.69

87.8
82.0
82.0
81.0
80.8

14.2
14.9
15.4
14.8
14.7

10.9
11.3
11.7
11.2
11.2

5,800
13,700
14,300
16,300
16,700

0.0366
0.0600
0.0666
0.0672
0.0694

4 4
3.2
3 25
3.1
2.9

6CO

Oil B,

fect of viscosity gradient on the distribution of velocity components, but also he used it to convert, in effect, his Reynolds number, which is calculated from film viscosity, back to
a main-stream basis near the transitional region. Hence, his
lines of constant p J p , in the transitional region are partly
corrections t o his method of calculating the Reynolds number. Furthermore, simplification of calculations using this
method of correlation is desirable, (1) because choice of a film
temperature is different for the turbulent and viscous regions,
(2) because the calculation of tube wall temperature by successive approximations involves successive recalculations of
both ordinate and abscissa, and (3) because the Reynolds
number must be calculated from main-stream viscosity to
determine the nature of the flow.

73.0
74.0
76.5
75.5

21.5
19.3
21.8
20.4

15.7
14.1
16.0
15.0

368
381
875
645

5.5
4.4
5.9
5.4
5.5
5.6

I.

I.

11.7
6.83
6.10
3.78
3.63

The authors felt that no correlations had been presented

which were entirely satisfactory for engineering calculations.
Because of the simplification of calculations when using mainstream properties, they sought to retain this advantage and at
the same time overcome the drawbacks previously mentioned. For this purpose a correlation using dimensionless
groups after the manner of Nusselt, including a group which
would serve t o evaluate the effect of the temperature coefficient of viscosity and temperature difference, seemed most
promising. As a measure of the interaction of the viscosity gradient of the temperature field on the velocity distribution, the ratio of the viscosity of the main stream to the
viscosity a t the temperature of the tube wall, pa/pw, was
chosen and included as a dimensionless group. Success of

DECEMBER, 1936

INDUSTRIAL AND ENGINEERING CHEMISTRY

1431

this method of correlation with the authors' data on heat

transfer in the viscous region led t o application of the method
to much of t h e data in the literature on heat transfer and
pressure drop, both in the viscous and turbulent regions.

temperature difference would correlate the data. For a first

approximation the function

Apparatus and Procedure

was plotted against temperature difference as shown in

Figure 2. As lack of agreement between heat transfer data
and the theory for isothermal streamline flow had been noted
by other workers (3,Q, 5 , 9), it was not surprising that the
points representing the heating data fell above, while the
points representing the cooling data fell below the value
$ = 1.5, derivable theoretically for zero temperature difference from the Graetz theory (3, 5 , 6) for values of D G / p
below 2100 and (4/.rr) (V'c/KL) above 40/.rr. Moreover,
the irregularity of the deviations was not ascribable to the
values used for the exponents of velocity and conductivity,
but indicated the need for evaluating the effect of temperature
difference in another manner.
As discussed earlier in this paper, the effect of temperature
difference may be rationally explained in terms of a velocity
gradient, dependent both on the magnitude of the temperature difference and the temperature coeffjcient of viscosity of
the fluid. It seemed worth while to examine the extent t o
which the ratio p,,/pu would serve to evaluate this effect
when employed in conjunction with the Nusselt, Prandtl, and
Reynolds dimensionless groups. The criterion for determining the exponents of these groups was that the average deviation of the experimental points from a formula under consideration should be a minimum. Since the combination of
the Prandtl (cp/K) and Reynolds ( D G / p ) groups into one
group-i. e., the Peclet group (4/a) (Wc/KL)-would have imposed a restriction on the analysis of the deviations, it seemed
preferable t o allow for the possibility that these two groups
might have different exponents.
As a first approximation, however, in introducing the
group, the exponents of the Prandtl and Reynolds groups
were both taken as -l/s in the expression for $. Then #
was plotted against pJpW as shown in the lower half of Figure
3. Data of other investigators are also shown in the upper
half of this figure. The equation of the empirical line shown
is :

The writers had available a mass of heat transfer data

taken some eight years ago by Sieder and Russell in the laboratory of the Foster Wheeler Corporation. The portion of
these data dealing with heating and cooling in viscous flow
covered a n extremely wide range of Reynolds numbers and,
what was more important for this investigation, was taken
with three oils of widely differing temperature coefficients of
viscosity. Figure 1 shows t h e viscosity-temperature characteristics of the three oils used. Data obtained for the
turbulent region are not included in this paper, since a considerable amount of literature is now available on the subject.
A description of the apparatus used in obtaining these data
is as follows:
A concentric-tube heat exchanger with 0.75-inch No. 16 Birmingham Wire Gage copper tube inside a 1.25-inch iron-pipe-size
brass pipe was used. The whole exchanger was heavily insulated.
The inner tube had an effective length of 5.1 feet. A calming
section of 2 feet and a mixing portion of 2 feet brought the overall length of the inner tube to 9.1 feet. A brass spiral retarder
inserted in the 2-foot section at the outlet served to mix the fluid
before the temperature was taken. The tube wall temperature
was determined by means of five thermocouples spaced along the
length of the tube and soldered in grooves in the wall. Calibrated thermometers were used to measure terminal temperatures. The flow rates were determined by measuring the time
required to fill a suitable size container. City water from a special stand pipe ensured a steady flow of a heating or cooling medium through the annular space. The use of water rather than a
condensing vapor as the heating medium made it possible to
obtain a wide range of tube wall temperatures. Furthermore,
the heat exchanger could be made t o act as a cooler or as a heater
at will, and a heat balance could be readily calculated for each
run.
When readings were taken, an effort was made to keep the
average temperature of the fluid constant for several values of
velocity so that its effect could be isolated from that of other
factors. When runs were made with different temperatures,
the fluid velocity was always adjusted to predetermined values
so that, when working up the results, the effect of changes in
fluid properties could be studied.

hD

cp

*=&)

-1f8

DG

),(

-113

113

(B)

hD 4 Wc
=-it(--)
K KL

-If3

Experimental Results
The results of the tests reported on heat tranafer in the
viscous region are listed in Table I. Specific heats were calculated from the Fortsch and Whitman equation, thermal
conductivities by the Bureau of Standards formula. The
viscosity-temperature characteristics of the three oils are
shown in Figure 1. The change of heat content of t h e water
was used for determining the value of t h e heat transferred in
each case in calculating h.

Method of Correlation
The data taken were plotted and used for design. No
attempt was made t o correlate them by means of theoretical
or dimensionless equations. I n 1928 comparison of the data
on turbulent flow was made with the curves of Morris and
Whitman ( I S ) , and good agreement was found. The data
for viscous flow were worked up by the method outlined
by Drew, Hogan, and McAdams ( 5 ) in 1931, and by Colburn's
method of correlation in 1933. The latter proved satisfactory
from the standpoint of accuracy but was found to be timeconsuming in design calculations. Hence a correlation was
sought which would be as accurate but simpler to use.
It had first been thought that, if a satisfactory formula were
obtained which would cover the variation of all factors except temperature difference, then a plot of i t as a function of

The terms were then rearranged to give

which was plotted against DG/pa as shown in Figure 4.

A second plot using -0.4 as the exponent of the Prandtl
group in the expression for D was made. The correlation was
inferior t o the first, for the average deviation of the points
was higher in this second case. A third plot using -0.25
as the exponent of the Prandtl group gave an average deviation nearly the same as in the first plot. However, the formula which then best satisfied the data had 0.3 for the exponent
of D G / p a , indicating a small net effect of viscosity. This
net effect could not be considered t o be established, hoJTever,
in view of such uncertainties as the true values of the thermal
conductivities of the three oils and such approximations as
using the arithmetic mean of the terminal fluid temperatures in calculating the mean temperature difference and
appropriate average fluid temperature. Thus the formulas:

INDUSTRIAL AND ENGIKEERING CHEMISTRY

1432

VOL. 28, NO. 12

+
A
6

3
++

Upper Plot
Velocite B oil ( I )
Velocite B oil (6)
Glycerol ( 4 )
Spindle oil (7)
Oils (10)
Oil average (16)
Oil and water (9)

Lower Plot

0 Oil A

Oil B
0 Oil C

0.01

or
$I = 1.86

tigators were plotted as shown in the upper half of Figure 4.

Drew (4) had pointed out that data on glycerol correlated
poorly with data on oil when using the Graetz empirical. We
might attribute this to the fact that the conductivity of
glycerol is quite different from that of oil and the exponent of
conductivity in the Graetz empirical equation was wrong, or
that some important function was missing. The data on
glycerol, however, correlated surprisingly well with the authors'
method, indicating that it did not have the weaknesses pointed
out for the Graetz empirical. The data of Holden call for
comment. ilt very low values of DG/pa his points fall off
towards the viscous asymptote [represented by 45' lines determined by values of ( c p / K X D / L ) 2 / 3 ] . At slightly higher
values of DGlp,, the points fall considerably higher than the
line. This may be due to free convection effects or to the
use of the arithmetic mean temperature difference. Use of
the temperature difference called for by the Graetz theory
brings the data into agreement with the line.
We now had a formula of the Susselt form for predicting
heat transfer coefficients in the viscous region. An examination of the formula indicated that it differed from the form
used by Morris and Whitman for plotting heat transfer data
in the turbulent region by the two ratios L I D and p,,/pw.
There was good evidence in favor of leaving in the latter ratio,

(E)""

were taken as representing the data within the range of significance of the deviations.
Equation 1 differs in three respects from the equation of
Colburn for heat transfer to fluids in viscous flow:

(z)

1 /3

$ = 1.5

(1

+ 0.015

Gr1'3)

(3)

First, Equations 1 and 2 employ the properties of the fluid

a t main-stream temperature, whereas Equation 3 employs
them at film temperatures. Secondly, Colburn used
(h/cG)( C ~ / K )instead
~ / ~ of ( h D / K )( c p / K )- l / 3 in his expression for the ordinate. Where the variations of K and c
with temperature are accurately known, either form is
equally convenient for calculating h, although h/cG has some
advantages. Thirdly, the authors use po/pr instead of
pQ/p, to evaluate the viscosity gradient throughout the range
of temperature, tube wall to fluid, This leads to the exponent 0.14 instead of 0.33 chiefly because lpw - po\ is
larger than Ip, - pal. Fourthly, the authors worked only
with liquids in small pipes, where the term containing the
Grashof grouz) varied but little, whereas Colburn considered
various &ids* and pipe sizes
where the G r a s h o f group
varied widely. When dealEO
ing with gases or liquids with
s m a11 temperature coefficients of viscosity, the pQ/pv
IO
group requires supplementation by the addition of the
G r a s h o f group. This is
5
b r o u g h t out in Table 11,
2
4
where the deviations of the
experimental data for water
are as high as ten times those
for oils. Correlation of the
I
data on fluids for which the
value of the Grashof group
e x c e e d s 25,000 may be
effected by multiplying the
2IY
values of h obtained from
Equation 2 by the term 0.8
(1 0.015 Grij3). Here the
G r a s h o f g r o u p is to be
evaluated using pa instead

"i--rr

&

of

Pj.

T o test the formula for

v a r i a t i o n s of L / D and of
thermal properties of other
fluids, the data of other inves-

FIGURE
4
0 Velooite B oil, heating
I

(1)
Yelocite B oil, cooling ( I )
Velocite B oil, heating (6)

+0

Glycerol heating (4)

Splndle &I, heating ( 7 )
Spindle oil, cooling ( 7 )

+!-

011 heating ( 9 )

X
0

W h e r heating (9)
Oils, hkrtting (10)

Oils, cooling ( 1 0 )
Oil, heating ( 1 5 )

INDUSTRIAL AND ENGINEERING CHEMISTRY

DECEMBER, 1936
T.4BLE

Fluid
Oil A
Oil B
Oil C

..

Type
Run0
H

..

Citatiori

(7

11. DEVIATIONS
OF EXPERIMENTAL
POIXTs
-100

nh

- 2
+15

5.5
15

27

15
11

+1;

13

+41

49
8

31
2

- 7
-11

10
11

43

+0.5
12
- 8
-10
13

14'
12
14
11

..

fl:

Velocite B

..
..

..

Velocite B
Spindle oil

Fi

..

C'

,.

14

ti9

..

16
11
1s
B

H
H
H

..
..
..

rr

74

H & C
H

88

- ..

= heating, C
cooling.
b n = number of runs, 6 = algebraic

14

..

..
+I2
+ 1

f10

..

33
12
I

..

Water
Water
Water
Water

20
32

..
..
..

. I

..

...

, .

..

..

..

..
..

16
16

..

..

11

I
I

Oil
Kerosene
Ben I en e

Water

27

15
15

Oils

I\

15'

+27
- 3

Oil
Oil

+'2

ii

10

13

ir

Velocite B
Glycerol
Oil

.i

...

..

24

..

..

..

78

...
...
10
...
,..

...
..,
...

17

19

17

..

mean deviation in per cent, A

201

93
300
17

IN

G < 2100< D-

..

+ 6

....
- 3
..
..
..

..

14
3 5

13

..
9
..
..
5
..
..

..
..
.*

+ i7

19

- 2

10

- 2
17

13
11
19

- 2

1433

FIGURES
4 AND 5
-2100
1

...
...
...

...

...
..
...

...
13
15
7
23
29

< DG
-<
ir

..
..

..
..
..
..

198
57
255
27

..

..

".

..

..

..

..

,.

..
..

..

..

..

..
..

..

..
..
..

..

..

..

,.

..

..

..

7
33
25

13
a5

..

+i4

17
29

34
29

+i 4
+10
+-10i 5

+63

63
29

33

12

- 7

..

+ 8

..9

76

+ 5

10
4 F,

++106 . 5

18
15

- 3

+ 2

- 18

.20. .
16
11

..

..
..

..

1-2
+21

+25

...
...

..

DG > 10,000-

7--

6
11
21
7
20

+ 4

-3
+33

16

..

103
16
13

10,000-

+11

i-29

..
..

+- 34 . 5
+ 4

f49

..

..
..

12

19

14

49

. I

..
..
..
..

25

..

..

..

..

..

..
..

..
..

14
15

..

26
10

..

10

17
10

mean absolute deviation in per cent:

and Fitzsimmons were correlated by Keevil and McAdams (8)

and the former could be taken care of in plotting by using it
by evaluating the viscosity a t the film temperatures:
as a parameter. Before the p a , p a ratio could be used in the
turbulent region, the value of its exponent was examined in
t/ = t,
'/& - t,)
the same manner as that described for finding its exponent
t/ = ta
l/4(fur - la)
in the viscous region. Data in the turbulent region show
In the viscous and far turbulent regions this procedure brings
little variation between pa and pw first, because fluids which
the friction factor for heating and cooling of the oils studied
give turbulent flow seldom have a large temperature coefinto substantial agreement with the isothermal friction lines.
ficient of viscosity and secondly, because the heat transfer
I n the transitional region there is a wide disagreement which
rates are high, preventing large temperature differences. A
careful study indicated that the exponent 0.14 of the pL,/pL, may be greatly reduced by using pa p/ lines as parameters,
as was shown by Colburn (3). However, use of film visgroup would bring the heating and cooling data into one band
cosity in the criterion of flow is an objectionable procedure.
of points. A plot of turbulent heat transfer data incorporatA plot of the ratio of the observed friction factor to the
ing (pn/p,Ja.l4is shown in Figure 5 , -here (hD/k')(c~/K)-~/~isothermal friction factor a t the same value of Re was made
(pajpw)-0.14 is plotted as ordinate and D G j p , as abscissa.
against the ratio palpw for the case of viscous flow, and is
Values of the ordinate for heating and cooling fall on either
shown in the lower curve of Figure 6. S o w when pJpY
side of the curve and have approximately the same distribuis 1, the flow must be isothermal and f fiaO. is 1. TJse of this
tion of deviations.
point as the origin gives two straight lines with slightly difI n plotting data by Sheru-ood, Kiley, and Mangsen (15)
ferent slopes for heating and cooling. However, a single line
for values of DG/p, between 2300 and 4000, it was noted that
of slope 0.25 was drawn so as to conform to the bulk of the
values for the same ratios of L I D fell on smooth curves, but
data, giving a value 0.9 to f,f,.o. a t ,ua'pm = 1. Thus for
that the curves were displaced for variations in this ratio.
simplicity it was assumed that the ratio mould have an exA separate plot was included on Figure 5 to indicate this condition, which is not altogether unexpected. As a matter of
ponent of 0.25 for both heating and cooling in viscous flow.
-4similar plot of f/f,ao. vs. pa/pu for the turbulent region
fact it would be surprising if the effect of length on heat transfer suddenly disappeared when passing from viscous to turbuindicated a value of 0.14 for the exponent and 0.98 for f/j,,,,,.
a t palpLto= 1. The friction factor multiplied by 1.1lent flow.
Designing heat transfer equipment calls for calculations of
(p,,/pw)o.25when DGjp, is less than 2100, and by 1.02pressure loss as well as heat transfer coefficients. We could
(pa/pcl,)aJ4
when DG/pL, is greater than 2100, was plotted
to effect
not be satisfied unless we could apply the ratio pL./pcL,
against DG/p, with the result shown by Figure 7.
a correlation of friction data when heat is being transferred to,
Thus by use of two figures (5 and 7) we can predict heat
or from, the tube wall. Pressure loss in isothermal viscous
transfer coefficients and friction factors quite simply. Diflow is proportional to the first power of both viscosity and
mensionless groups determined for heat transfer can be used
velocity. Therefore, we can predict that a given change in
for calculating the friction loss. There is little change in
the radial viscosity gradient of a fluid flowing through a tube
procedure when going from viscous to turbulent flow. Estiwould effect a greater change in pressure drop than i t would
mation of the tube wall temperature is simple, for when calin heat transfer, and conversely that any method which would
culating it by successive approximations only one of the
compensate for the effect of variations in the radial viscosity
groups making up the ordinate is affected, while the abscissa
gradient 011 fluid friction data would be adequate for use on
remains the same. Usually a first approximation suffices,
heat transfer data. Thus the critical test of suitability of
for an error in p w is reduced by one-fourth or one-seventh its
the ratio pa/pw was to apply it t o friction factor correlations.
size in the viscous region and by one-seventh its size in the
The data on fluid friction of Keevil, of White, and of Clapp
turbulent region by the exponent of the pJpw group.

+
+

1434

INDUSTRIAL AND ENGINEERING CHEMISTRY

0 Velocjte B oil, heating ( 1 )

0 Velocite B oil cooling ( 1 )
A Veloc:lte B oji, beavng (7)
A Velocite B oil. cooling (7)

0 Spindle oil heating ( 7 )

Spindle oil cooling( 7)
I3 Red oil, hlating (7)
I Red oil, cooling ( 7 )

Recommended Use in Heat-Exchanger Design

Figures 5 and 7 allow us to predict local values of the heat
transfer coefficient and friction factor inside tubes. For
problems in heat-exchanger design, where a fluid flows through
a bundle of tubes in contact with another fluid, an average
tube fluid temperature, different from the arithmetic except
when the temperature rise is quite small, should be used
for evaluating the fluid properties. Colburn (a)derived
a formula for the average temperature to be used in the case of
counterflow of fluids and drew a chart giving the average
temperature to be associated with an over-all coefficient of
heat transfer based on logarithmic mean temperature difference and the assumption:

YOL.28, NO. 12

= a(1 $- at)

In the case of noncounterflow Nagle (14) and Underwood (f7)

derived expressions for the mean temperature difference, on

the assumptions of a constant over-all coefficient and perfect

mixing across any cross section. A rational way of combining
the analyses of Colburn and Nagle is to assume

= 811

+ B(t - T ) ]

which, following an analysis outside the scope of the present

paper, leads to the mean temperature differences calculated
by Nagle and mean temperatures not far from those obtained
from Colburns chart. Hence, the use of Kagles curves and
of Colburns chart is recommended a t this time.
Pressure drops for turbulent flow can be calculated with
reasonable accuracy by using Colburns chart for the mean
temperature. However, in the viscous region, the following
formula is recommended (here F is obtained from Colburns
chart:
(1

- F) log 4

2 = log APC

*Ph

AQ h

DECEMBER, 1936

INDUSTRIAL AND EYGINEERING CHEMISTRY

Nomenclature
pressure drop calcd. with fluid properties taken at cold
terminal temp.
Aph = pressure drop calcd. with fluid properties taken at hot
terminal temp.
Apo = actual over-all pressure drop
h
= av. coefficient df heat trans'fer between inner tube surface and fluid, based on arithmeti: mean temp. difference, B. t. u./hr. X sq. ft. X F.
= friction factor in Fanning equation
= linear velocity of fluid, ft./sec.
= density of fluid, lb./cu. ft.
= acceleration due to gravity
= inside diam. of inner tube, ft.
= heated length of inner tube, ft.
= weight of fluid passing through inner tube, lb./hr.
= mass velocity of fluid passing through inner tube, lb./hr.
x sq. ft.
= outer pipe as in experimental data tables
= inner pipe as in experimental data tables
= av6 5mp. of fluid entering heating section of inner tube,
Ap,

- P.

av. temp. after mixing of fluid leaving heating section of

inner tube, O F.
(TI Tz),
B. t. u./lb. X ' F.
= sp. heat of fluid at
= heat given up to or by tube fluid, 1000 B. t. u./hr.
= weight of water passing through annular space of heat
exchanger, lb./hr.
= av. temp. of water entering annulus, F.
= av. temp. of water leaving annulus, F.
= heat given up by or t o water, 1000 B. t. ud/hr.
= av. temp. of inside surface of inner tube, F.
= thermal conductivity of fluid inside tub:, evaluated at
I/z (TI Tz), B. t. u./hr. X sq. ft. X
F./ft.
= Prandtl number, evaluated a t I/2 (TI
Tz)
= viscosity of tube fluid, evaluated a t 1/1 (TI
Tg),
Ib./hr. X ft.
= viscosity of tube fluid, evaluated a t I/z T ,
(Ti
T2)lb./hr. X ft.
= viscosity of tube fluid, evaluated a t T,, lb./hr. X ft.
=

(E)
(E)"'*
114

= 1.1
= 1.02

DG

below - = 2100
Pa

above

DG = 2100
Pa

Acknowledgment
The authors wish to acknowledge the kindness of W. H.
McAdams in making available to them the data of Keevil, of
Clapp and Fitzsimmons, and of White.

Literature Cited
Clapp and Fitzsimmons, Mass. Inst. Tech., Thesis, 1928.
Colburn, A. P., IND.
ENO.CHEM.,25, 873 (1933).
Colburn, A. P., Trans. Am. Inst. Chem. Engrs., 29, 174 (1933).
ENQ.CHEM.,24, 152 (1932).
Drew, IND.
Drew, Hogan, and McAdams (Holden and White data). Ibid., 23,
936 (1931).
(6) Graetz, L., Ann. Physik, 18,79 (1883); 25,337 (1885).
(7) Keevil, Mass. Inst. Tech., Thesis, 1930.
(8) Keevil and Mcridams, Chem. &. Met. Eng., 36,464 (1929)
ENQ.CREM.,23,625 (1931).
(9) Kirkbride and McCabe, IND.
(10) Kraussold, H., Forsch. Gebiete Ingenieurw. B 2, Forschudngheft
351, 1-20 (1931).
(11) Lawrence and Sherwood, IND.
ENG.CHEM.,23, 301 (1931).
(12) LQvLque, Ann. mines, [I21 13, 201, 305, 381 (1928).
(13) Morris and Whitman, IND. ENQ.CHEbr., 20, 234 (1928).
(14) Nagle, W. M., Ibid.,25, 604 (1933).
(15) Sherwood, Kiley, and Mangsen, Ibid., 24, 273 (1932).
(16) Sherwood and Petrie, Ibid.,24, 736 (1932).
(17) Underwood, A. J. V., J.Inst. Petroleum Tech., 20, 145 (1934).
(1)
(2)
(3)
(4)
(5)

RECEIVED
June 8,1930. Presented as part of the Heat Transfer Sympasium
held under the auspices of the Division of Industrial and Engineering
Chemistry of the American Chemical Society a t Yale University, New Haven,
Conn., December 30 and 31, 1935.

1435

VAPOR
RE=USEPROCESS
Separation of Mixtures
of Volatile Liquids
DONALD P.OTHMER
Polytechnic Institute, Brooklyn,N. Y.

HE heat requirements are considerable

for the separation by distillation of mixtures of two or more liquids in which the
less volatile is greatly predominant. By means of the processes described here, i t is possible to reduce this amount of
heat greatly, to reduce the size of distilling columns required
for the separation, and to make possible the operation of some
distillation processes without the use of water as a condensing
or cooling medium. While the process may be more readily
understood by reference to mixtures of two components, it
is applicable and shows the greatest merit when used with
more involved distillation projects, such as the production
of pure grain alcohol or pure absolute alcohol from a distiller's
beer.

Theory as Applied to Two Liquids

Many examples of pairs of liquids in which the more volatile
liquid is present in relatively small amount are met in industrial processes, and these separations are the ones in which
the Vulcan vapor re-use process shows the greatest savings.
One of these is the beer in alcohol production which usually
contains about fifteen times as much water as alcohol.
Others are the aqueous solutions of acetone or methyl alcohol
(or mixtures of the two) which are obtained by water washing of the vapors in air in various solvent recovery systems.
The resulting aqueous solution may have to be as dilute as
forty to fifty parts of water for one part of acetone, for example, in order to remove efficiently the vapors of the solvent
from the air.
The rectification of the alcohol in the first example, or of
the acetone in the second, from the water requires the expenditure of considerable heat because of the diluteness of
the solution. A distilling and rectifying column may be
regarded as a heat engine with poor thermal efficiency when
separating a very dilute solution such as these which are in
question. It is necessary to distill so much of the mixture even
with a theoretically perfect column in order to remove all
the solvent. Thus with a 2.5 per cent solution of acetone
in water, although the vapors arising are comparatively
strong (about 40 per cent acetone), there would necessarily
have to be distilled 60 per cent water or one and a &alf
times as much water as acetone. As the solution becomes
progressively weaker, i t is necessary to distill a much larger
amount of water to remove the acetone completely. Hambrand' (Principles and Practice of Industrial Distillation)
shows that to obtain 1 kg. of acetone as vapors of 99.75 per
cent purity from an acetone mixture of 2 per cent requires
almost 1000 kg.-cal. This large expenditure of heat shown
theoretically by Hausbrand is borne out in actual practice;
and in the recovery of acetone from a 2.5 per cent solution in

"Principles and Praotiae of Industrial Dietillation," New York, John

Wiley & Sone, 1928.