Anda di halaman 1dari 9

Chemical Engineering Science, 1966,Vol. 21, pp. 87-95. Pergamon Press Ltd., Oxford. Printed in Great Britain.

The efl.ect of viscosity on the flooding conditions in wetted wall columns

R. CLIFT, C. L. PRITCHARD and R. M. NEDDERMAN


Department of Chemical Engineering, University of Cambridge, Cambridge

(Received 12 November 1964; in revised form 18 January 1965)

Abstract--The gas velocitiesfor flooding in a wetted wall column have been measured as a function of
liquid viscosity and flow rate. The flooding velocities were found to be somewhat higher than those
predicted by DAVIDSONand SrmA~R. Measurements have also been made of the liquid downflowafter
the column has flooded.

INTRODUCTION and a is the surface tension at the gas liquid inter-


face. m is the thickness of the undisturbed film
A WETTED wall column is a device often used for
given by the Nusselt equation:
contacting a gas with a liquid. It consists of a
vertical tube down which the liquid flows freely as m--- [ 3#Q,/1/3
an annular film and through which the gas passes,
generally counter-current to the liquid. Under
certain conditions the counter-current flow be- According to the theory We, R e and Z are uniquely
comes unstable and part of the liquid passes co- related at the point where the standing waves
current with the gas. This phenomenon is known become very large. SrIEARER and DAVtDSON
as " f l o o d i n g " plotted a family of curves for W e against R e with
The formation of standing waves on a liquid various values of Z as a parameter. The curves
film due to gas flowing upwards over it has been appeared to agree fairly well with published values
considered theoretically by SHEARER and DAVID- of the flooding point in wetted wall columns.
SON [1]. They found that for every value of liquid Therefore they suggested that the phenomenon of
flow there is a critical gas velocity beyond which flooding was due to wave formation on the liquid
the wave amplitude became very large. Three film, and that their theoretical curves could be
dimensionless groups emerge naturally from the used to predict flooding.
theory. The results with which the SHEARERand DAWD-
SON theory was correlated covered only a small
(1) Weber number W e = PgU2m range of Z values. This work was therefore under-
(7 taken to provide information over a wide range
of Z. It will be noted that variation in Z results
4 PzQI almost entirely from variation in the viscosity of
(2) Reynolds number Re = the liquid since most commonly encountered
l.t ~z D
liquids have more or less the same density (0-5-2)
and surface tension (30-80 dyn]cm) but have a
(3) " L i q u i d property Z = [tr3 /91/1/3
wide variation in viscosity (0.01-1 P).
parameter" | p4 g )
A similar study has been undertaken by WALLIS
[2] but since there is reason to suppose that end
where U is the critical gas velocity, Qe is the liquid effects were important in WALLIS' apparatus it
volume flow rate, pg and p~ are the densities of the was thought worthwhile repeating the work. The
gas and liquid respectively, p is the liquid velocity present results are compared with WALLIS' below.

87
R. CLIFT, C. L. PRITCHARD and R. M. NEDDERMAN

EXPERIMENTAL PROCEDURE
Consttanntkhead
The object of this work was to investigate the
effect of viscosity on the flooding in a wetted wall
column. Water and aqueous glycerol solutions
were therefore used as these always show New-
tonian behaviour and it is possible to change the
viscosity over a wide range without appreciably
affecting either the density or surface tension.
Details of the solutions used are given in Table 1. I
I I
Sintered[-~
sectionL_~ .

1
Table 1

Viscosity* Density Surface Z*


Tension RotameterJ~
(cP) (g/cma) (dyn/cm) I
I
1° 82 ~ Glycerol 82-5 1.21 65-0 9.1
2. 77~ Glycerol 46-0 1.20 65"8 20
3. 70~o Glycerol 23.4 1.18 66"4 50
4. 59~ Glycerol 10.4 1.15 68.0 147
Air
5. 25~ Glycerol 2"18 1-06 70-2 1200 supply ,[--'~manometer
6. Water 1.32 1 "00 72'0 2310

* These values varied somewhat from day to day depen-


ding on the ambient temperature.

Storage tank
In principle the apparatus required was simple,
consisting of a vertical column with means for F[o. 1
introducing air and liquid, and controlling and
measuring their flow rates. Steps were taken to end of the pipe. In this way the airflow was dis-
ensure that smooth entry conditions were main- tributed across the width of the entry compartment
tained in both fluids. The apparatus is shown and prevented from passing straight up the column
diagrammatically in Fig. 1. as a jet. A smooth bell-shaped section turned
The column consisted of two 6 ft lengths of from a block of wood served to accelerate the air
perspex tubing, the nominal i.d. being 1¼ in. stream to the velocity with which it would pass up
The sections were checked to ensure that their the column. It was anticipated that this device
internal dimensions were matched to within fairly would calm the air flow and reduce its turbulence
small tolerances. The perspex sections were in- at entry to the column. This calming section was
terspaced by an annular section of sintered bronze, similar to that used by BASHFORTHet al. [3] and
approximately 3 in. long and machined so that NICKLIN and DAVIDSON [4] before them. The air
its i.d. was the same as that of the perspex tube. pressure in the chamber was measured by either a
The whole column was set up with its axis ver- mercury manometer or an inclined water mano-
tical. meter to give a wide range of sensitivity.
Air could be introduced at the base of the column The sintered bronze section already referred to
from the low pressure laboratory supply via a was used to ensure that the liquid flow was com-
rotameter. The horizontal 1 in. pipe from the pletely smooth at entry. It was thought that the
rotameter turned through a right angle inside the standing disturbance set up at entry when a liquid
compartment at the lower end of the column, so is simply poured down a vertical surface (see
that the air stream passed vertically upwards to Fig. 2) would be largely eliminated by admitting
impinge on a horizontal circular disc fixed to the the liquid through a porous section. To ensure

88
The effectof viscosityon the floodingconditionsin wettedwallcolumns
solution by measuring the volume of liquid passed
in a given time. The air rotameter was calibrated

Liquid~~/'-~l~' inJ/ Lqiu~~


by passing the air through it and then through

~l di a standard gas meter. A calibration curve was drawn


up for air at 16.7°C and 78.0 cm H g under which
conditions its density is 1-25 x 10 -3 g/cc.
In SHEARER and DAVIDSON'S predictions the gas
velocity and density appear only in the combina-
tion U2pg a quantity which is directly related to
the reading on a rotameter since the volume of
gas passed at a given reading is inversely propor-
tional to the square root of the gas density. Hence
[a] (b) in this work it was never necessary to know either
Fla. 2 the gas flow rate or density separately since the
group U2pgcould be calculated directly from the
steady flow the liquid was supplied to the porous rotameter reading. Henceforth all air velocities
section from a constant head tank via a rotameter mentioned are hypothetical velocities calculated on
and control valve. The liquid leaving the top or the assumption that the air was at its " s t a n d a r d "
b o t t o m of the column was returned to a storage density of 1.25 x 10 - a g/cc.
tank from which it was pumped to the head tank. Most of the experimental work was concerned
Since the liquid was introduced half way up the with obtaining pairs of values of the liquid and
column instead of at the top as is usual in such gas flows at which the column would just flood
work it was possible to measure the division of with the flows increasing or would just cease to
the flow into components passing co- and counter- flood with the flows decreasing. In these deter-
current to the gas flow. This division occurs at minations it was found simpler to operate the
flow rates higher than the flooding rates. column at constant input of liquid and to change
A disentrainment device was fitted to the top of the air flow slowly, rather than vice versa. Occa-
the column to prevent the droplets of liquid formed sional checks showed that the flooding points
under flooding conditions from spraying over the determined in this way were identical with deter-
laboratory. minations made at constant air flow. This is
inevitable since if a column is just about to flood
a minute increase in either flow rate will cause
METHOD OF OPERATION
flooding.
Since the viscosity of aqueous glycerol solutions At a later stage in the work, measurements
is very sensitive to variations in composition and were made of the quantity of liquid flowing down
temperature, it was measured frequently during the column under certain conditions. The tech-
the course of each run by a portable Ferranti nique adopted was to maintain a constant level in
viscometer (Couette type). The other relevant the funnel which formed the base of the column
properties surface tension and density are less by a suitable setting of the outlet valve and to
affected by changes in temperature or composi- pass the outflow into a measuring cylinder for a
tion and it was thought sufficient to measure the definite time.
density at the end of each run and to calculate the
surface tension from this using published data.
QUALITATIVEDESCRIPTIONOF THE PHENOMENA
Since the surface tension varied by only 15 per
OBSERVED
cent throughout the experiments, this technique
must have been accurate to within a few per cent. At low or zero air flow, the liquid flows as a
The calibration of the liquid rotameter was film down the walls of the column. With the
sensitive to the viscosity so this was checked for each more viscous solutions (20 cP and above) the

89
R. Cur-r, C. L. PRITCHARDand R. M. NEDDERMAN

~
o

QL=33-1 crn3/sec

F /f
<~

' I ~ ,~
I
0 I 3 4 5 6 7 8 9 I0
Qg, L/sec

FIG. 3

liquid film was disturbed by waves which became length of liquid filled column. Therefore at the
visible some distance below the sinter and increased flooding point the pressure drop would increase
in amplitude as they travelled to the bottom of with time as the top half of the column filled with
the column. As the air flow was increased, these liquid.
waves grew in amplitude and the point of wave Once the column was flooded, the liquid below
inception crept up the column until they were the sinter appeared to be in a much less chaotic
visible right from the liquid entry. At the higher state than that immediately before flooding. Slugs
Reynolds numbers these waves were replaced by still formed across the column, but much less
the smaller more random ripples commonly found frequently.
on turbulent films. However as the gas flow was As the air flow was increased the pressure drop
increased waves were formed, but with the least across the column passed through a maximum
viscous solution these were only observed at gas (see Fig. 3). The subsequent fall off of pressure
velocities close to flooding. appeared to be due to considerable reduction in
When the flooding point was almost reached the the hold up in the lower part of the column. At
pressure drop across the column began to increase a higher value of the air flow Qc the downflow of
from its previous very small value (see Fig. 3) and liquid ceased altogether and the lower part of the
at the same time the waves on the film were seen column began to dry out. This point appears to
to have become so large as to form thin bridges or correspond roughly to an inflexion in the curve of
slugs across the air stream. These bridges were pressure drop against gas flow and was indepen-
quickly broken up by the air flow and the droplets dent of total liquid flow rate.
resulting from their collapse were carried up by Once co-current flow has been fully established
the gas stream. in the upper part of the column, the air supply
Further increase in either liquid or gas flow could be reduced below the value required to
increased the bridging and pressure drop across initiate flooding before the column returned to
the column until a critical point was reached at purely counter-current flow. The point at which
which the liquid slugs became larger and were no the air ceased to carry any liquid upwards was as
longer dispersed immediately by the air stream. definite as the point at which co-current flow started
The flow pattern in the tube was now chaotic and and was also characterised by a discontinuity in
the liquid slugs were carried above the sinter, the pressure drop curve. It will be referred to as
thereby resulting in a division of the liquid flow the flooding point with decreasing gas flow or the
into components passing co- and counter-current de-flooding point.
with the gas. The increase in pressure drop was For the higher viscosity solutions, the flooding
augmented at the flooding point by the increased and de-flooding points were not very different but

90
The effect of viscosity on the flooding conditions in wetted wall columns

,.4~--" /×/× \ Eq.9


1"2

I'O

0"8 _ . " / Z=50 j × f >~\ \


~ ' - ~ \ \ ~x~×X Z=1200
, . × / //"Z=I50 \ \ ,,,,.q,.x x Z=~310
0'6 - - ~ ~ ' - ~ ~.~x .....x~

_ .,
0.4

0.2 •"/Z=3360 "~.


I ] '
3 ,o so ,oo soo ioo~-
Re
Fro. 4

with the less viscous liquids the two points were HEWITT and WALLIScorrelations are in terms of
some way apart. That these two points are not the quantities Vg and Vt defined as
coincident has already been remarked by HEWITT
and WALLIS[5] in the course of some experiments V./ P' t ~/2
on flooding with the air-water system. In fact it V* = .[Dg(p t _ P.)] (2)
is hardly surprising that they are not the same
since they represent quite different conditions: the v: =
{'D.(fL Pg)//'/2
flooding point is the upper limit of counter-current
flow, the de-flooding point is the lower limit of where Vg and Vl are the superficial gas and liquid
co-current flow. velocities respectively. No theoretical basis is
given for the use of these groups. HEW[TT and
RESULTS WALUS correlated their results with the expressions

(a) Flooding point determination


,/(v?) + , / ( v * ) = 0.88 (4)

The flooding point was determined by the pro- for low liquid flows and
cedure outlined above. From the measured values
V(v, ) + x/(v~*) = 2.0 (5)
of the flow rates and fluid properties the results
were converted to a relation between Re, We and for high liquid flows.
Z, allowance being made for the fraction of the tube It will be seen from Fig. 5 that our results lie
occupied by the undisturbed film in calculating U. almost entirely in the transition region between
These values are shown in Fig. 4 along with the these two correlations. Since the results cover
predictions of SHEARERand DAVIDSON. It will be the range of liquid Reynolds number from 100 to
seen that though the predicted Weber Numbers 1100, the transition region is clearly extensive and
are roughly the same as those found experimentally, of major importance. If a correlation in the
there is not close agreement between theory and HEWITT and WALL[S tradition is required the
experiment. In particular the maxima in the following would be of more use.
theoretical curves occur at considerably lower
Reynolds numbers than those found by experi- x/(V*) + 0.34x/(V* ) = 0.79 (6)
ment. The correlation
The present results for water are consistent with
those of HEWITT and WALLIS (see Fig. 5). The ~/(V¢) + 0.55x/(V~) = 0.65 (7)

91
R. CLIFT, C. L. PRITCHARD and R. M. NEDDERMAN

I.(3
experimental results. KAPITZA'S equation is of
interest in that it predicts an infinite flooding
velocity at zero liquid flow. SHEARERand DAVIDSON'S
~Eq.5 theory and HEWITT and WALUS' correlation suggest
0,8 that the flooding velocity tends to a finite value at
zero liquid flow. The experimental evidence suggests
that the latter trend is correct. Furthermore,
KAPITZA predicts that the relationship between We
0.6 and Re should be independent of Z. This is n o t
in agreement with experiment.
In one respect the Re-We-Z plot is a little
misleading. The use of Re (~ #- 1) and We (a #~)
0,4
ensures that the maxima in the flooding curves
shall become higher and further to the left as # is
increased and Z thereby decreased. As a result,
0"2
Hewittand Wallis Ec~4~ the effect of viscosity on the flooding point is made
o Present work to appear much more critical than it really is.
The flooding point for three of the experimental
solutions are plotted on Fig. 6 as flooding velocity
I I t as against liquid flow rate with viscosity as a
02 0'4 0.6
parameter. The effect of viscosity on flooding
conditions is seen to be quite small: the change in
FIG. 5
the flooding velocity is of the order of 20 per cent
for a 70 fold change in viscosity. Figure 6 also shows
given by StoRES, PICKERING and BLACKER [6] is that the column floods more easily with increased
not in agreement with either the present results liquid viscosity, a fact not obvious from the
or those of HEWITr and WALUS. Presumably this Re-We-Z representation. VERSCHOOR [7] has
is because of different entry conditions. pointed out that the liquid surface tension affects
It will of course be realised that SHEARER and the flooding point critically; and this is consistent
DAVIDSON'S correlation involves the gas flow rate with SHEARER and DAVIDSON'S curves, although
squared, whereas the HEWITT and WALLIS' corre- again the method of plotting conceals it.
lations involve the square root of the gas flow rate. WhELm' results for the 9 cP solution are shown
It is therefore not surprising that there should in Fig. 6. The flooding velocities are much less
appear to be better agreement by the latter method. than those found in these experiments, indicating
KAPITZA [7] has predicted the flooding condi- that end effects were important in WALLIS' experi-
tions in a wetted wall column and gives the rela- ments.
tionship
U- k = 3"5[Tzopgl/2~ l/3 ( t7t l/2 (b) Conditions when the column is flooded
\ p,(2, ] (8)
Liquid downflow. In order to investigate the
where k is the velocity of a surface wave on the behaviour of the column under flooded conditions,
film." Since k ~ U it can be neglected and equation a series of runs was performed in which the airflow
(8) can therefore be rearranged to give was kept constant and the nett downflow of liquid
71 (Qd) was measured as a function of the total flow
We = - - . (9) of liquid (Qt) to the column. Figures 7 and 8 show
Re
the results as Qt was increased from zero to a value
This is plotted on Fig. 4 and it can be seen that it considerably in excess of the flooding value and
does not give an accurate prediction of the present then reduced again to zero.

92
The effect of viscosity on the flooding conditions in wetted wall columns

95C

90C

~- 85C
E
u

soc 0"°,2 , o . , o m

75C

70C ~./.,c~9Cp 0"=0:3 uyneS.C,,,

,r ".~ ~' r ,E I ,,,


I 2 3 4 5

Q__L cmZ/sec
lrD '

FIG. 6

When the column was not flooded, the liquid difference represents the upflow into the top
downflow was of course identical to the liquid input section of the column.
so the condition in the column moved along the If the total flow is increased the pressure drop
line Qd = Qt (marked OA on Fig. 7). As the flow increases slightly (B'C') but the liquid downflow
was increased there was a small increase in pressure remains constant. Under flooded conditions the
drop (marked O 'A' on Fig. 8). The flooding point liquid passed down the column in discrete slugs.
is characterised by the usual discontinuity in the This appears to account for the scatter of the
pressure drop curve (A'B'). The results show that experimental points about the line DBC. The
there is also a discontinuity in the Qa-Qt plot downflow was measured by collecting the liquid
(from A to B) as the downflow drops to a value over an interval of time, so the collection of one
which is sometimes as low as a fifth of the down- slug more or less made an appreciable difference
flow immediately before flooding. Clearly the to the measured value of Qa. Nevertheless, the
results show quite clearly that the liquid downflow
in a flooded column is independent of the total
× Flow increosing liquid input.
t21
0 Flow decreosing Once the column was flooded Qt could be re-
duced below the flooding value without altering
IO

× Flow increasing
0 Flow decreasing
m ,×C t

u 6 E' ~ . . . , ~ - ~ ' 5 " ~ ~


D~.o~- u iB

0
4 E
~J
xO C

(3.
<3
O' 0 LD" xi×J A"
I I I I I . T
4 8 12 16 20 4 8 t2 16
Or, cm3/sec QI , cm3/sec

Flo. 7 FIG. 8

93
R. CLIFT,C. L. PmTCHARDand R. M. NEDDERMAN

Qd. The column did not deflood until Qt had without causing any co-current flow. Thus to pass
been reduced to the value of Qa (point D, D'), from a point on D A to a point on BD or vice
that is to say until the upflow has been reduced versa, it is normally necessary to pass clockwise
to zero. At this point there was a sudden reduction round the triangle DAB.
in the pressure drop as the top part of the column Since the downflow in a f o o d e d column is inde-
emptied. pendent of total liquid input it must be a function
The triangle DAB on the Qt-Qd plot and the of gas velocity and liquid properties only. Also it
area D " A ' B ' D ' on the Q r A P plot represent regions must be equal to the de flooding flow rate. Figure 9
over which the liquid downflow is dependent on shows the gas velocities as a function of liquid
the history of the flow conditions. To investigate rate for the flooding, de-flooding and downflow of
whether there was any tendency to pass spon- the 10cP solution. It can be seen that the two
taneously from DB to DA, the column was held latter effects lie on the same line.
at the conditions represented by E and E' for over Figure 9 shows that there is a critical gas velocity
an hour. N o change in the conditions could be at which the downflow becomes zero. This presum-
observed. An attempt was also made to initiate ably represents the deflooding rate at zero liquid
flooding from a point near A by shaking the flow but there is no fundamental reason why it
column. A very hard blow would sometimes cause should also represent the flooding rate. However,
a slug to form but this was quickly dispersed it will be seen that the de-flooding and flooding
rates approach each other closely as the flow rate
is decreased, and it is therefore reasonable to
suppose that the flooding and de-flooding points
I,OOC
are the same for zero liquid flow.
Values of the critical gas flow for zero liquid
downflow, Qc, were determined for various solu-
o Flooding
tions and the results shown in Table 2. While there
:~ Deflooding
90C
~ Nett liquid downflow Table 2

Viscosity of Liquid(cP) Qe (1./sec)


1.3 8-06
80G
2.2 7"92
11-8 7.92
E 23-4 7.77
44.8 7"84

70C appears to be a gradual decrease in Qc as the


viscosity is increased, this only amounts to about
3 per cent change for a 40 fold change in viscosity.
Thus Qc is effectively independent of viscosity and
a mean value of 7.901./sec is taken. This corres-

\
60C
ponds to a gas velocity of 10.1 m/see which is in
good agreement with the value of 9.5 m/sec given
by BASHFORTH et aL [3]. This then represents the
gas flow rate at which no counter-current flow can
f I I I i r . exist, i.e. the socalled " d r y o u t " point.
0 I 2 3 4 5 6
OL The values for the de-flooding conditions for
crnZ/sec
TrD ' the 10cP solution are presented on the We-Re-Z
FxG. 9 plot (Fig. 10) for comparison with the predictions

94
The effect of viscosity on the flooding conditions in wetted wall columns

agree well with those of HEwITr and WALLIS, who


I'0 used similar apparatus. However, the range investi-
gated (Reynolds numbers from 100 to 1100) is not
covered by the WALLIS correlations.
(3) When a wetted wall column floods, the quantity
O'e
of liquid flowing down the column immediately
decreases. In general, the less viscous the liquid,
the greater the decrease. Further increase in the
0 . 6 mA
total liquid flow does not affect the downflow.
x (4) A column " d e f l o o d s " when the total liquid
Shearer and Davidson" flow equals the downflow which the gas flow will
permit in a flooded column.
0"4-- (5) At, and above, a certain critical gas velocity
there is no nett flow of liquid down the column.
This critical gas velocity is close to the value given
0 Flooding by BASm:ORTHet aL for the lowest air velocity which
02-- will support a standing film of water. It is virtually
x Deflooding
independent of liquid viscosity.
Nett liquid downflow
NOTATION
I 1 I I I # Acceleration due to gravity
0 50 I00 150 200 250 D* k Wavevelocity
Re m Liquid film thickness
F~. I0 D Tube diameter
Qc Critical volumetric gas flow
of SHEARER and DAVIDSON. It can be seen that Qa Volumetric liquid downflow
at high Reynolds numbers the experimental results Q~ Volumetric gas flow
Qt Volumetric liquid flow
for flooding are in fair agreement with the pre- Qt Total volumetric liquid flow
dictions. This is perhaps fortuitous since the agree- U Mean gas velocity
ment is not so good at low Reynolds numbers. Vu Superficialgas velocity
Vg* Dimensionless gas velocity
Vt Superficialliquid velocity
CONCLUSIONS Vt* Dimensionless liquid velocity
(1) Neither SItEARER and DAVIDSON nor KAPIrZA Z Liquid property parameter
Re Reynoldsnumber
adequately predict the onset of flooding in a wetted We Weber number
wallcolumn. SrmA_r~ERand DAVIDSONat least predict /z Liquid viscosity
the general form of the dependence of flooding on po Gas density
p~ Liquid density
gas and liquid flows. KAeITZA fails even in this. ~r Surfacetension
(2) The present experimental results for water AP Overall pressure drop
REFERENCES
[1] SrmAReRC. J. and DAVmSONJ. F., J. Fluid Mech. (In Press)
[2] W~d~taSG. B., General Electric Co., Schenectady, New York. Report No. 62 GL 132 1962.
[3] BASHFORTHE. Q. FRASERJ. B. P. HUCrnSONH. P. and NEDD~RMANR. M., Chem. Engng ScL 1963 18 41.
[4] NICKI.rND. J. and DAVIDSONJ. F., Proc. of Two Phase Fluid Flow Syrnp., Inst. Mech. Eng., London, 1962 p. 29.
[5] H~wrrr G. F. and WAtus G. B., A.E.R.E.--R4022 1963.
[6] SHn~sG. L. PICKEeal~GA. R. and BLACKERP. T., A.E.E.W.--R343 1964.
[7] KAPITZAP. L., Zh. Exp. i. Teor. Fiz., 1948 18 19.
[8] VERSCHOORH., Trans. Inst. Chem. Enyng, 1938 16 66.

Zusammenfassung--In einer NaBwandkolonne wurden die Flutgeschwindigkeiten des Gases in


Abh~ingigkeitvon Z[ihigkeitund Durchsatz der Fliissigkeit gemessen. Die Flutgeschwindigkeitenlagen
etwas h6her als nach der Voraussagevon DAVIDSONund SI~E~R. Auch die Rieselgeschwindigkeit der
Fliissigkeit am Flutpunkt wurde bestimmt.

95

Anda mungkin juga menyukai