terials, but the data apply only to

Flooding Velocities in air andwater. Sinceindustrialab-

sorption processes involve many
systems other than air and water,
it was felt that a determination of
the effect of the physical proper-
ties of the two fluids would be
Packed Columns useful and desirable.

Experimental Procedure (3)

An investigation of the effect on
flooding velocities of the physical
T. K. SHERWOOD, G. H. SHIPLEY, properties of the gas and the liquid
A N D F. A. L. HOLLOWAY was made in a small tower with a
Massachusetts Institute of Technology, single packing. Dry 0.5-inch (12.7-
mm.) carbon Raschig rings were
Cambridge, Mass. dumped into a %inch (5.1-em.) glass
tower, to a height of approximately
4 feet (122 cm.). The tower and
auxiliary equipment are shown dia-
grammatically in Figure l.
Carbon dioxide, hydrogen,
and air were passed countercurrent
HEWAS the height of an absorption tower determines with water to determine the influence of the gas density upon
its absorption efficiency, the cross section determines flooding velocities. Another series of runs was made with
aqueous glycerol and air to determine the effect of liquid vis-
the capacity or the quantity of gas which can be cosity. The viscosity of the liquid was measured by calibrated
treated. For design purposes it is obviously important to capillary-tube viscometers. Investigation of the surface tension
have information on the allowable gas and liquor rates per effect was carried out by adding butyric acid to water, causing a
square foot of tower cross section. twofold variation in the surface tension of the solution. Various
mixtures of methanol and water were used in which the surface
I n packed towers the gas and liquid rates are limited by tension varied threefold. Half-inch Berl saddles were also used
the tendency of the column to flood. As either liquid or gas in place of the original Raschig r h g packing, and flooding ve-
velocity is increased, the liquid holdup in the packing in- locities were determined with air and water.
creases, the free area for gas flow decrea,ses, and the pressure The actual experimental procedure was carried out in a step-
wise manner which was carefully followed for each flooding ve-
drop through the column increases. A point is finally reached locity measurement. A liquid rate was arbitrarily established
when the gas bubbles violently through the liquid, the pres- in the tower with zero gas velocity. The gas velocity was then
sure drop rises extremely sharply with the slightest increase increased from zero to the flooding velocity in small increments.
in gas velocity, and much liquid is carried off mechanically
by the gas leaving the top of the packing. This point is
called the flooding point and is determined by both gas
and liquid rates. At any given gas rate there is a definite
liquid rate above which the column will flood. Similarly
at any given liquid rate there is a definite gas rate above Data are presented showing the effects
which the column will flood. of gas density and liquid viscosity on the
(Loading point and load point are sometimes used flooding velocities in an experimental
synonymously with flooding point. White ( 7 ) , however, packed column. Data for a number of
defines the loading point as the gas velocity a t which, for a gases and liquids obtained with 12.7-mm.
fixed liquid rate, the pressure drop first becomes proportional
to an exponent greater than 2.0 on the gas velocity, or the Raschig rings in the small experimental
logarithmic curve of pressure vs. gas velocity first deviates column were used to develop a new
from a slope of approximately 2. As the gas rate is increased method of correlating flooding velocity
above the loading point, a point is finally reached a t which the measurements. Surface tension of t h e .
same curve turns abruptly almost vertically upward. The liquid was found to have no appreciable
latter point is the limiting condition of operation or flooding
point, and the best condition of operation is considerably effect.
below the flooding point. The best operation should be Using the new method of correlation,
determined by an economic balance, since power for the data of a number of investigators on
blower, 8,s well as height and cross section of the tower, are all flooding velocitieg for Haschig rings vary-
involved. ing in size from 12.7 to 35 mm. in large
Although the best condition of operation is determined by
a n economic balance, a knowledge of flooding velocities is columns were plotted and found to be in
extremely useful in first determining the limiting gas and excellent agreement. Flooding velocities
liquid rates above which operation is not possible, and in on stacked rings are much higher than
estimating the optimum liquid and gas rates where there for dumped rings. The general relation
are not enough data to make a n exact economic balance of obtained for flooding velocities in dumped
operating and fixed costs.
Flooding velocities are a function of the liquid and gas Raschig rings was found to be a good ap-
rates, the physical properties of the two fluids, and the char- proximation for many dumped packing
acteristics of the packing material. There are some data in materials.
the literature (I, 6, 7) on flooding velocities over a range of
liquid and gas velocities and,for a variety of packing ma-
765
766 INDUSTRIAL AND ENGINEERING CHEMISTRY VOL. 30, NO. 7
The flooding condition was ascertained by visual observation of
the liquid flowing over the packing and down the walls of the
tower. Results were readily reproducible. Liquid rates were 2 00
determined by weighing a measured volume over an increment
of time. The gas rate was measured by a calibrated orifice.
At the flooding velocity readings were taken of the pressure drop IO0
across the orifice, static pressure downstream, gas temperature, 80
liquid rate, and liquid temperature. All experiments were car- 60
ried out at atmospheric pressure.

Results and Discussion

The results of these measurements are presented in Table
I. Correlation of the data was attempted in a manner re-
cently suggested and known to be of considerable value for
the system air and water. The suggested method is to plot

FIGURE2. DATAON EFFECTOF GAS

DENSITY

and carbon dioxide. When the data are plotted in the man-
ner suggested, Figure 2 shows that the points fall on three
well-defined lines. The only factor whic) may be held to
account for the separation of the data is gas density. Cor-

I.
TABLE EXPERIMENTAL
DATA
L UO
L b . / ( h r . )(sq. ft.) Ft./sec.
0.5-IN. CARBON RASCHIG RINQS;S = 92; F = 0.8 (AS U S ~ D ;S )a = 0.0107
Water, air 13,800 0.84
pL = 62.4; p G = 0.074 lb./cu. f t . 9,450 1.32
t = 23' C. (73.4O F.) 6,280 1.62
Surface tenslon = 73 dynedcm. 4,630 1.95
p = 0.94 centipoises 3,180 2.29
7,580 1.48
11,100 1.07
FIGURE
1. SKETCH
OF APPARATUS 4,660 1.97
14,800 0.88
8,620 1.38
5,220 1.80
as abscissa the volumetric ratio of liquid to gas rates (&/uo) Water, hydrogen 16,000 2.65
and as ordinate the square of the gas velocity based on free pL = 62.4; pG = 0.0051 Ib./cu. ft. 14,900 2.82
t = 24' C . (75.2" F.) 14,600 3.00
area of the packing, divided by the mean hydraulic radius Surface tension = 73 dynes/cm. 13,900 3.30
p = 0.91 9,400 4.45
of the packed section. The ordinate is written as uo2 S/F3, 4,830 6.80
dr as the dimensionless ratio (uO2S/gF3), derived as follows: 9,220
6,680
4.90
5.80
11.500 3.42
U3-4
gas velocity based on free area = - = % Water, carbon dioxide 13,900 0.72
AF F p L = 62.4; pG = 0.113 lb./cu. ft. 10,900 0.95
t = 23O C. 8,260 1.20
vol. of space filled with fluid 5,570 1.48
hydraulic mean radius = Surface tension = 73 dynes/cm.
wetted surface p = 0.94 3,240
5,570
1.84
1.45
= hAF
_ _ = _ F
Aqueous glycerol I , air 15,600 0.81
ShA S p L = 68.5; pG = 0.074 Ib./cu. ft. 10,000 1.23
.
* '
(gas velocity based on free area)z - ( u o / F )
hydraulic mean radius F/S
-~ e
F3
t = 24O C .
Surface tension = 73 dynes/om.
6,450
4,130
14,700
1.81
2.14
0.82
p ="3.2
Aqueous glycerol 11, air 14,800 0.81
There is some theoretical justification for the relation sug- pL = 67.3; pG = 0.074 Ib./cu. ft. 14,100 0.89
t = 23' C. 11,700 1.03
gested above (6),and the groups involved may be predicted Surface tension = 73 dynes/cm. 8,900 1.37
from dimensional analysis. For values of the factors S and p = 2.3 7,000
4,500
1.65
2.10
F , tabulations ( 2 ) are available for average characteristics of
Aqueous glycerol 111, air 14,200 0.53
typical packing materials, but the values will change with pL = 72.9; pG = 0.074 lb./cu. ft. 10.200 0.96
each dump, depending principally upon the number of pack- t = 23' C . 8,500
7,200
1.10
1.25
Surface tension = 72 dynes/cm.
ing units per unit volume. The method of evaluation adopted p = 11.6 5,230 1.62
in the experimental procedure was to determine the sur- 4.620 1.73
~~~~

face area, Sc,and water displacement volume, Bo,of a typical *4queous glycerol I V , air
p L = 74.1; pG = 0.074 lb./cu. f t .
12,800
9,780
0.56
0.85
packing unit, and to count the number, N , of packing units t = 23' C. 8,100 1.03
5,800 1.33
per unit volume for each dump. The product, So N , i s the Surface tension = 70 dynes/cm.
p = 25.1 4,130 1.55
factor 8, and the difference, 1- V oN , is the factor F . 2,970 1.87
The first data to be analyzed by the suggested method of (Continued on page 767)
correlation were those for water flowing against air, hydrogen,
JULY, 1938 INDUSTRIAL AND ENGINEERINQ CHEMISTRY % 767

rection of the data by some function of the gas density will By multiplying either the abscissa or the ordinate of Figure
bring them into agreement, but it is believed that a more 2 by some power of the density ratio, the lines can be
logical correction factor than the gas density alone would be brought together. However, such an adjustment as could
the ratio of the gas density to the liquid density ( p Q / p L ) , be made, for instance, by plotting (uo2S/gF3) (pa/p4)O+6
since both factors are involved when employing fluids ;of vs. (&/uo) will cause the position of the three lines to coincide,
but their curvatures will not coincide. It is believed that
this curvature is sufficiently important to warrant more care-
0.04 ful correlation; therefore by first correcting the abscissa by
(pQ/pL)-"' and then correcting the ordinate by (PG/pL)0'g7,
the data may be brought into excellent agreement. The
0.02 method of correlation is, therefore, to plot as ordinate (uo2S/
gF3) ( p G / p L ) , and as abscissa (&/uo) (pG/pL)-"'. The ab-
scissa is the same as ( L / G ) (pc/p1,)0.5, and the latter form is
0.0 I used in Figure 3 since the weight ratio of fluid rates is a more
0.008 familiar term than the volumetric ratio. The points plotted
0.00 6 in Figure 3 are in good agreement, and there is no trend of
any individual system to follow any curvature other than
OD04 that of the average.
The next data to be studied were those for aqueous glycerol
and air, in which the viscosity factor varies almost thirty
0.002 I I fold. If the data are plotted in the same manner as in Figure
C 0 0.20 0.40 0.60 0.80 1.0 2.0 3, the points fall on a series of parallel curves. Since
FIGURE3. CORRELATION
OF DATA
ON the surface tension factor is nearly constant, viscosity is the
THREEGASES
0 .I
.08
different densities, and by dimensional analysis the two will
occur as a ratio. Hence it is assumed that the missing factor .06
in Figure 2 is the fluid density ratio.
.O 4

TABLEI. EXPERIMENTAL
DATA(Continued)
L UQ
L b . / ( h r . )(sg. It.) Ft./sec.
0.5-IN. CARBON RASCHIG RINGS;S = 92; F = 0.8 (as USED); SO= 0.0107
Aqueous butyric acid I, air 14,750 0.50
pL = 64.3; pG 0.074 lb./cu. ft. 13,500 0.68
t = 23' C . 10,400 0 .91 .006
Surface tension = 42 dynes/cm. 6,830 1.48
p = 0.94 3,750 2 . 00
8,150 1.32 .004
Aqueous butyric acid 11, air 14,800 0.63
p L = 62.4; pG = 0.074 lb./cu. f t . 14,000 0.70
t = 23' C. 11,300 0.95
Surface tension = 47 dynes/cm. 8,170 1.30 .a0 2
p = 0.94
5.770 1.80
5~3,3.0 ~ . 2 ni
11,500 1.00
5,950 1.80
3,100 2.45 -00I
11,100 0.67
0.I 0.2 0.4 0.6 0.8 10. 2.0 4.0 6.0
Methanol, ax
p L = 49.9; pG = 0.075 lb./cu. f t . 10,500 0.75
t l = 23O C . 6,590 1.23
Surface tension = 26 dynes/cm. 3,790 1.73
J, = 0.55 1,750 2.40
5,460 1.48 FIGURE4. CORRELATION IN WHICH
OF DATAFOR TESTS
3,060 1.92 WAS.VARIED
LIQUIDVISCOSITY
50y0 methanol
pL = 57.1; pG
+ 50Y0 water, air
= 0.074 lb./cu. ft.
t = 230 c.
Surface tension = 33 dynes/cm. only factor which mill account for the deviation. By mul-
p = 1.55
tiplying the ordinate by the viscosity in centipoises raised t o
25% methanol -I-75% water, air the 0.2 power, the data for aqueous glycerol may be brought
pL = 59.9; pG 0.074 lb./cu. ft. together and agree very well with the data for water with air,
t = 23O C.
Surface tension = 44 dynes/cm. hydrogen, and carbon dioxide. This is shown graphically in
p = 1.45
Figure 4.
The remaining data for aqueous methanol and aqueous
0 . 6 , . 1 ~BERL
. SlDDLES; 8 ~ 7 7 F; S 0 . 7 4 (AS USED); 8 0 30.0108 butyric acid against air were plotted as in Figure 4, and
Water, air 14,200 0.70
p L = 62.4: pG 0.074 lb./ou. Et. 12,100 0.85 Figure 5 shows that there is no serious trend of points away
t = 220 c. 10,400 0.91 from the curve establi'shed in Figure 4. Since the surface
Surface tension = 73 dynes/cm. 8,550 1.08
P = 96 6,450 1.42 tension varied threefold in these experiments, i t is concluded
4,330 1.85
3,070 2.25 that surface tension has no effect on flooding velocities within
11.800 0.73 the range of the variables employed. The data obtained for
8,400 1.05
6,560 1.38 0.5-inch Berl saddles are also presented in Figure 5 and are
3,680 1.90
8,050 1.10 in good agreement with the points representing 0.5-inch
Raschig rings.
 768 INDUSTRIAL AND ENGINEERING CHEMISTRY VOL. 30, NO. 7

rings in a 2-inch (5.1-em.) tower. The general line lies above

the data of Figure 4 by a factor of about 2. This might
0.04
have been predicted, since the data of White show some
evidence of a wall effect (2), the flooding point being reached
a t higher flow rates in large than in small columns with the

-
N
0.
9
0.02 same size packing.

Application of Dumped Ring Correlation to

Packing Materials
Besides the data for Raschig rings, Baker, Chilton, and
Vernon (1) reported the air rates a t which flooding occurs in
0.004 various packings over which water is circulated a t the rate
of 500 pounds per (hour)(square foot)[2l.l kg. per (hour)
(square meter)]. These data for miscellaneous packing ma-
0.002
0.I 0.2 0.4 0.6 0.8 10
. 2.0 4.0 terials are shown in Figure 7, augmented by the unpublished
data of Uchida (4) on three sizes of broken solids. The line
for dumped Raschig rings was transferred from Figure 6 to
Figure 7 and represents the data for miscellaneous packing
5. DATAON EFFECT
FIGURE OF SURFACE
TENSION materials surprisingly well. Hence it is concluded that the
line for dumped Raschig rings may be used to estimate flood-
ing velocities for any dumped packing material for which
The final correlation of the effect of fluid properties on there are no data.
flooding velocities is that used in Figure 4,
(uo2S/gF3) ( P ~ / P ~vs.
) PW. I~
G) (P,/P,)~.~ Illustrative Calculation of Flooding Velocity
All of the ratios involved are dimensionless, with the exception The use of Figures 6 and 7 for the calculation of the gas
of the viscosity factor ~ 0 . 2 . Applying dimensional analysis flooding velocity is straightforward if the ratio L/G is known,
to the seven variables selected ( u o / F ,Q / F , p,, p L , p, F / S , and but is necessarily by trial and error if only L is known. If
g) we can show that the gas rate is specified, the calculation of the limiting liquor
rate involves no trial and error.
(dSlgF3) = +[ ( L / G ) , (PG/PL), (hI~5)l As an example, suppose it is desired to determine the tower
cross section corresponding to the flooding velocity with air
This indicates that the missing dimensionless
ratio is (pLs/L)o2, which, if included in the
ordinate, would put a net exponent on factor S
of 1.2. There are not sufficient data available
at present to determine whether 1.0 or 1.2 is
the correct exponent on factor S , and to avoid
u n d u e c a l c u l a t i o n in the use of the final
correlation, the effect of the liquid viscosity
is represented only by $. Should subse-
quent experimental data justify an exponent
on factor X of 1.2, the correlation can be made
entirely dimensionless.

0eneral Correlation for Raschig

Rings
Data are available in the literature (1, 5 , 7 )
on flooding velocities of Raschig rings, in towers
sufficiently large so that the wall effect is not
important, for dumped rings varying in size
from 0.5 to 1.4 inches (1.3 to 3.6 cm.). These
data were augmented considerably through the
courtesy of S. Uchida, who furnished un-
p u b l i s h e d d a t a f o r t h r e e sizes of rings
(dumped) a t high liquor rates and for stacked
rings. The flooding velocities are presented
g r a p h i c a l l y i n F i g u r e 6 by means of the
method of correlation developed previously.
The points fall in two distinct lines, one for
d u m p e d r i n g s a n d o n e for stacked rings.
Flooding velocities in stacked rings are evidently
considerably higher than those for dumped rings.
The line representing the general correlation
for dumped Raschig rings in Figure 6 was
transferred to Figure 4, for comparison with
t h e d a t a o b t a i n e d f o r 0.5-inch (12.7-mm.) FIGURE O F DATA
6. CORRELATION ON RINGS
IN LARGETOWERS
JULY, 1938 INDUSTRIAL AND ENGINEERING CHEMISTRY 769

0.8 3. Liquid surface tension has a negligible

0.6 effect upon flooding velocities within the range
0.4 of 73 to 26 dynes per em.
4. The available flooding velocity data for
0.2 Raschig rings in towers sufficiently large to elimi-
nate the wall effect correlate very well by the
0.1
0.0 K graphical method suggested above. Points for
0.OK stacked rings lie above those for dumped rings,
0.04 indicating that limiting velocities are much higher
for stacked rings.
0.02
5. The available flooding velocity data for
0.0 I
miscellaneous packing materials fit the general
0.006 correlation for dumped Raschig rings quite well,
0.006 indicating that the dumped ring correlation
0.004 may be used to estimate flooding velocities for
any dumped packing material for which there
0.002
are no data.
0,001
0.0008 Nomenclature
0.0006
0.0004 Either the English or the centimeter-gram-
second system of units may be employed, except
0.0002 that p must be expressed in centipoises when
using the plots shown.
0.0001
0.00008
0.00006 A = cross-sectional area of column, sq. f t .
0.00004 F = fractioh of free volume in packing, cu. ft./
(cu. f t . tower volume)
ODOK 0.01 0.02 0.040.06 0.1 a2 0.4 0.6 10. 2.0 4.0 6.0 10 20 40 KO 100 g = acceleration of gravity, ft./(sec.) (sec.)
G = superficial mass velocity of gas =
uOPG, 1b.l
FIGURE,7. DATAON MISCELLANEOUS PACKINGMATERIALS
COMPARED (sec.)(sq. ft.)
WITH CORRELATION
FOR RINGS(SOLID
LINE) h = tower height. ft.
L = superficiarmass velocity of liquid = QPL, lb./
(sec.) (sq. ft.)
and water in 1.0-inch Berl saddles. The water rate is to be N = No. of packing units per cu. ft. of tower volume, l/(cu.
li \
IL.)
10,000 pounds per hour, and the gas rate 1900 cubic feet per Q superficial liquid velocity, ft./sec.
=
minute a t 68 F. and one atmosphere. S surface area of packing, sq. ft./(cu. f t . tower volume)
=
Under these conditions p = 1.0; pG = 0.075; p L = 62.4; So = surface area of one packing unit, sq. ft.
S = 79; F = 0.75. The abscissa is uo = superficial gas velocity (based on empty tower), ft./sec.
V o = water displacement volume of one pecking unit, cu. f t ,
p = viscosity of liquid, centipoise8
PL = density of liquid, lb./(cu. ft.)
PG = density of gas, Ib./(cu. ft.)

From Figure 7 the ordinate is 0.15, whence Acknowledgment

32.2 X 0.753 X 62.4 = 21,4
The authors wish to thank S. Uchida of Tokyo University
uo2 = 0.15 of Engineering for placing a t their disposal his unpublished
79 X 0.075 X 1.0.2
data on flooding velocities of Raschig rings and broken solids.
The allowable gas velocity, uo,is 4.6 feet per second, and the
minimum tower cross section is Literature Cited
Baker, T. C., Chilton, T. H., and Vernon, H. C., Trans. Am. Inst.
= 6.9 square feet Chem. Engrs., 31, 296 (1935).
60 X 4.6 Sherwood, T. K., Absorption and Extraction, 1st ed., New
York, McGraw-Hill Book Co., 1937.
Shipley, G. H., Jr., S.M. thesis in chemical engineering, Mass.
corresponding to a diameter of approximately 3.0 feet. A4-foot Inst. Tech., 1937.
tower would avoid flooding, but the best diameter would be Uchida, S., private communication, 1937.
determined by an economic balance involving pressure drop. Uchida, S., and Fujita, S., J. SOC.Chem. Ind. (Japan), 39, 886
(1936).
Walker, W. H., Lewis, W. K., McAdams, W. H., and Gilliland,
Summary E. R., Principles of Chemical Engineering, 3rd ed., New
York, McGraw-Hill Book Co., 1937.
1. The utility of general flooding velocity data lies first in White, A. M., Trans. Am. Inst. Chem. Engrs., 31, 390 (1935).
determining the limiting gas and liquid rates above which RECEIVED
April 5. 1938.
operation is not possible, and in estimating the optimum
liquid and gas rates where there are not sufficient data to make
an exact economic balance of the costs involved.
2. Fluid density and liquid viscosity have an effect upon
flooding velocities in packed columns, best correlated by a
plot of