Professional Documents
Culture Documents
A generalized model has been developed for the prediction of pressure drop and flooding in packed
columns in which gas and liquid flow countercurrently. The model has been validated for a wide variety
of packings, both random and structured. A single mathematical expression is used to describe all flow
regimes: dry gas, irrigated gas flow below the load point, loading region, and flooding. The approach to
the model development is fundamental in character and is an improvement over models published
earlier.
Nomenclature
a Specific surface area of packing S Cross-sectional area of column (m2)
(m2 me3) v, Superficial gas velocity through a
c Exponent in Equation (10) packed bed (m s-i)
C,, c2, c3 Constants in Equation (7) l-J, Superficial liquid velocity through a
4 Nominal diameter of packing element packed bed (m s-‘)
(mm) u, Velocity for suspending a single particle
Particle diameter, dP = 6(1 - &)/a (m) (m s-‘)
4
Particle diameter including surface us Velocity for fluidizing a bed of particles
4
liquid (m) (m s-‘)
Column diameter (m) Z Total height of packing (m)
f” Friction factor for flow through a
packed bed Greek letters
.A Friction factor for Ergun equation Pressure drop (N mm2)
AP
fo Friction factor for flow past a single Pressure drop through an unirrigated
APd
particle (dry) bed (N mm2)
fs Friction factor for flow past a particle Pressure drop through an irrigated bed
APin
in a bed
CN mm2)
Frr Froude number for liquid [Equation & Bed void fraction (porosity) (m3 mm3)
(1411 Y Absolute viscosity (kg (ms)-‘)
g Gravitational constant (m se2) (3 Surface tension (N m-‘)
h Liquid hold-up in a packed bed Density (kg mm3)
P
(m3 mm3)
h, Liquid hold-up below the loading point Subscripts
(m3 me3)
n Exponent in Equation (1) f Flooding
Pressure (N rnd2 or bar) g Gas
P
Reynolds number for the L Liquid
R%
gas = dr Ugpglclg
Numerous attempts have been made to describe the have ranged from the very empirical to the semi-empirical
hydrodynamic behaviour of packed columns operating as and have achieved moderate success for some applications
countercurrent gas/liquid contactors. These attempts within certain limited ranges of operating conditions. In
0950-4214/89/010019-10$03.00
0 1989 Butterworth 8 Co (Publishers) Ltd.
Gas Separation 8 Purification 1989 Vol 3 March 19
General model for pr~djctio~ of pressure drop: J. Stichlmajr et al.
the porosity does not depend on loading. Such is also the The change in bed void fraction can be expressed as
case for a packed column operating under dry (unirrigated)
conditions. This can be demonstrated by comparison of c’=.s-h
Equation (5) with the large collection of pressure drop
data presented by Coulson and Richardson*. This or
comparison is shown in Figure 2 for beds of spheres. In
addition, Rumpf and Gupte’ studied pressure drop in c’ = ~(1 - h/E) (8)
fixed beds with a systematic variation of porosity from
0.35 to 0.70, and found that the correction to the friction where h is the operating hold-up (volume liquid/volume
factor for a single particle was (1 - s)s-4.55which compares total bed). The change in particle diameter can be
favourably with the form of Equation (5). described by
The Ergun4 equation, generally accepted as descriptive
of the pressure drop of a single phase fluid flowing (1 - c’)/d; = (1 - .c)/d;
through a fixed bed is, for the gas
or
A&Z =fe [Cl - W3] p&Id, (6)
d; = d&l - ~(1 - hk)]l(l - E)}I’~ (9)
The main difference between Equations (5) and (6) is the
exponent on the porosity term. In the Ergun equation the The friction factor for a single wet particle will be
porosity term results from an oversimplified model never different from that of a dry particle since the actual gas
validated experimentally, since porosity is a constant in a Reynolds number depends on the actual wet particle
given packed bed. In Equation (5) however, the porosity diameter. Equation (7) can be simplified (see Appendix B)
term results from a large number of experiments in to
fluidized beds having a wide variation in porosity.
Because of the improved porosity term in Equation (5) f,=Rei (10)
the pressure drop in a bed of particles can be calculated
from a knowledge of the friction factor of a single particle. where
Thus, there is no need for a friction factor of the bed.
F&re 2 represents an example of how the data can be [-C,/Re, - C2/(2Rei”)]
c=
correlated by a relationship of the following type Al
fo = C,/Re, + C,IReiJ2 + C, (7) then the change in the friction factor will be
lo5
Structured packings:
Montz Bl 300 300 0.97 2 3 0.9 18
Bl 200 200 0.98 2 4 1 .o 18
Bl 100 100 0.99 3 7 1 .o 18
:z 165
139 0.76
0.773 1 : 1.1
1.4 f ;:
Reflux Rings 50 120 0.78 75 15 1.6 25
Hiflow Rings 20 291 0.75 10 5 1.1 27
The ratio of Equation (12) to Equation (5) gives, in Comparison of Equation (13) with similar ones by
combination with Equations (8) (9) and (11) Bemer and Kalis’, Billet and Mackowiak2, and Bravo et
d3 clearly shows its advantages. First, there is no need for
Api=/Apd = {[ 1 - E (1 - h/&)1/( 1 - E)}‘*’ cK3(1 - h/~)-~.~~ empirical packing correlation factors since the first term
(13) on the right-hand side of Equation (13) should account for
all packing differences. Second, the influence of the
which describes the increase in pressure drop in an effective void fraction on the pressure drop follows the
irrigated packing as a function of the dry packing void same functionality for the wet and dry cases.
fraction and the liquid hold-up. This expression should be Equation (13) should be valid below as well as above
valid for any type of packing so long as the single particle the loading point provided that the liquid hold-up is
dry friction factor can be described by Equation (10). known or can be accurately predicted. The excellent tit of
Figure 4 shows how this expression describes various sets this model exhibited in F’i&z.ue4 demonstrates this fact; for
of experimental data for random and structured packings all the points plotted there was a corresponding experi-
in the turbulent regime, where c = 0. mental liquid hold-up value. These experimental values
0.2
t 0.1
0.06
CJl
2 $ 0.04
$g 0.02
aI
$ 0.01
: 0.006
0.004 1.10-Z
0.1 0.2 0.L 0.6 0.8 1.0 1.5 2 3 m/s
0.003
0.2 0.4 0.6 1.0 2 4 m/s Superficial gas velocity U, -
Gas velocity U, --D Figure 5 Hold-up of Bialecki rings, conditions as for Figure 32
0.2
f
0.1
r"
5
u
z
= 0.0
1
0.01 0.02 0.04 0.06 01 0.2 0.4 0.6 ld5
Hold up h/E -
Figure 4 Validation of Equation (13). using different types of
packings*
Figure 6 Correlation of liquid hold-up below the loading point
(data by BilletlO)
were the ones used in Equation (13) to plot the line
representing the predicted values.
FrL E u2 a
Lg&4.65
Liquid hold-up
The correlation does not take into account any properties
The typical behaviour of liquid hold-up in a packed
column for different liquid and gas loadings is shown in of the liquid and has been validated for air/water only.
The definition of the Froude number includes the term
Figure 5 for a random packing. Below the loading point,
the hold-up is a function only of the liquid rate; above the .s4.65because it intuitively follows Equation (5). It was
loading point the hold-up also depends on the gas rate. found that this definition provides an excellent correlation
The region where there is an influence of gas rate is of the hold-up data.
Figure 7 shows that Equation (14) is applicable for
commonly known as the loading region.
liquid viscosities up to about 5 centipoises, based on the
data of Buchanan”. The influence of higher viscosities on
Hold-up below the loading point hold-up is not fully accounted for. Nevertheless, most
Numerous attempts have been made to describe the distillation, absorption and stripping applications of
dependence of hold-up on liquid velocity below the commercial importance exhibit viscosities in the range of
loading point. Hold-up measurements by Billet”, for eight applicability of Equation (14).
different packings, have been plotted in Figure 6 and may Still another effect that is not taken into account in
be correlated by Equation (14) is surface tension of the irrigating liquid.
Mersmann and Deixler12 developed a correlation for
ho = 0.555Fr;’ (14) small Raschig rings that incorporates the effect of surface
tension as indicated in Figure 8. It would appear that
where the Froude number is defined as surface tension is important at low liquid loadings but
T
Lo
loo
4
n Pressure drop AP;,, /( z.‘?ig) ---D
I”
Figure 9 New plot of liquid hold-up; data according to Figure 5
dAPd
-= 0
Figure 8 Correlation of liquid hold-up below the loading point, for
small Raschig rings12 aAPim
Differentiation of Equation (16) then yields Equations (15) and (16) were used to determine irrigated
pressure drops. Comparisons between prediction and
2+ch
40-- measurement were made in plots, such as those shown in
3 O Figures 10-12. Because of the huge amount of data, a single
1 - E + hoi1 + 20 [ApirJ(ZPLg)l fI
parity plot was not feasible.
Figures lo-22 demonstrate the ability of the model to
186 ho predict the wet pressure drop for a variety of packing types.
W-9 The comparison makes it clear that the pressure drop in
- E - ho{1 + 2O[Api,l(Zp,g)]f] =O
the loading region can be predicted satisfactorily. Figures
13-26 are parity plots for flooding conditions. The
The solution of this equation yields the pressure drop at
agreement between experiment and model is satisfactory
the flood point. In most cases this pressure drop has a
for all four data groups. The structured packings have
value between [Api,/(Zprg)]r = 0.1-0.3, which is in good
larger deviations, particularly at combinations of high gas
agreement with practical experience.
rates and low liquid rates. However, it is in this region that
The essential quantity for the calculation of the flood
flooding velocities are difficult to measure and thus the
point is liquid hold-up. The data published thus far
basic data tend to be unreliable.
(mostly for air/water) are not sufficient to develop a
The validation of the model will continue as more data
correlation valid for all types of packing and all systems.
become available. Future adjustments may well be
In particular, experimental hold-up data are required for
required. Still, the model currently represents the best
systems with low surface tensions and low viscosities.
400
Validation of the model mm Water
It seems clear that for the complex contacting mechanics m
in an irrigated packed column, the most meaningful
method of validation is to compare the predicted para-
100
meters against those measured under test conditions.
P
Liquid hold-up is not normally reported, and few reliable
data are available; thus for this parameter the empirical
correlation [Equation (14)] is used to support estimates of
pressure drop and flooding.
Many pressure drop data have been reported, often for
wide ranges of liquid and gas rates, and these permit
comparisons between dry and irrigated pressure drops, as
shown in Figure 3. When the loadings have been carried to
high values, flooding conditions are obtainable from the
same sets of data. The data bank that we have used for
validation is shown in Table I. It contains entries for 4
structured as well as random packings. Data sources are 0.1 0.2 0.4 1.0 2.0 4.0 10.0 m/s
given as part of the table.
For dry pressure drop, data were chosen carefully from Superficial gas velocity U, ---Lz
graphed results unless specific values were reported. Figure 11 Comparison of experimental (points) and predicted
(lines) pressure drop values, for 35 mm ceramic Pall rings
40.0
mbar 6.0
m in Water Cascade Mini Rings # 1
ft
10.0 10.0
m/S m/s
f 4.0 P 4.0
2 5
x 2.0 x 2.0
.)1 c
.z 5
$! 1.0 9 1.0
:: ::
G-l 0)
; 0.4
3
ii 0.2
s
0.1
0.1 0.2 0.4 1.0 2.0 4.0 10.0 m/s 0.1 0.2 0.4 1.0 2.0 4.0 10.0 m/s
10.0 10.0
m/s m/s
f 4.0 f 4.0
2 3
x 2.0 )r 2.0
.Y c
.E
$
ul
1.0 L---L _!
$ 1.0
::
z 0)
; 0.4
3
z 0.2
z
I 0.1
0.1 0.2 0.4 1.0 2.0 4.0 10.0m/s 0.1 0.2 0.4 1.0 2.0 4.0 10.0m/s
Figure 14 Parity plot for gas velocities at the flood point; dumped Figure 16 Parity plot for gas velocities at the flood wint: dumoed
ceramic packings (see Table 1) plastic packings (see Table 7)
1 or
fo = AReC (2)
Taking logarithms
- 186 h,
ln~~=lnA+clnRe (3)
h--ho[l+2*(~)~] =O
Taking the derivative of Equation (3)
The procedure for calculating the flood point is as
follows:
d Info
dInc (4)