308 Chem. Eng. Technol.

19 (1996) 308-314

Simulation of Hydrodynamic Behaviour and Mass Transfer Efficiency

of Reduced Pressure Distillation

Belkacem Benadda, Pierre Moszkowicz, Michel Prost, Michel Otterbein,

and Kamal Mohammedi *

A random packing hydrodynamic simulator is designed specially to carry out experiments

under reduced pressures with a counter-current flow air/water system. The simulator re-
sults (air/water system) compared to those obtained in a real packed distillation column
(benzylchloride/ethylbenzene system) show a good agreement between hydrodynamic pa-
rameters like pressure drop, flooding flow rate, and total liquid hold-up. Empirical rela-
tionships derived from the simulator experimental results are proposed allowing the
calculation of the pressure drop in a packed distillation column, operating in the pressure
range between 10 kPa and atmospheric pressure. The influence of reduced pressure on the
HETP in the distillation column and interfacial area in the simulator was also investigated.

1 Introduction with air/water system under reduced pressure, to those ob-

In distillation under reduced pressure, counter-current ran-
dom packed columns allow the separation of corrosive or
thermosensitive materials and the recovery of chemicals
dangerous to the environment.
A reliable design of packed column depends primarily on
the knowledge of hydrodynamic behaviour of packing ele-
ments through the whole load range of operating pressures.
This requires the prediction of the effect of the operating
parameters on pressure drop, flooding flow rate, liquid and
gas hold-ups. The existing correlations often cover only the
experimental range conditions on which they were based.
Few published or still unpublished works deal with the hy-
drodynamics of packed columns operating under reduced
pressure [I]. All of these works showed a different behav-
iour when compared to that observed under atmospheric
pressure. Extrapolation of the available empirical or semi-
empirical correlations developed for atmospheric pressure
is not always recommended without prior experimental ver-
ification.
To our knowledge, nothing has been reported in the litera-
ture to date concerning the influence of reduced pressure on
the interfacial area a .
The aim of this work is twofold:
Firstly: to determine, at different reduced pressures, the
hydrodynamic characteristics of a distillation column of a
given binary mixture from simple trials on an air/water sim-
ulator. This can be reached by comparing simulator results
tained in a packed distillation column with benzylchloride
ethylbenzene system. A good correlation between simulator
values and real distillation values at different pressures
shows the interest of such a n investigation. Secondly: to
study the distillation column efficiency under reduced pres-
sure by means of the height equivalent to a theoretical plate
(HETP). This leads to interfacial area determinations in the
simulator under reduced pressure. This last work represents
in itself another original feature of this investigation.
2 Experimental Apparatus
The distillation column and simulator are represented in
Figs. 1 and 2, respectively. They are mainly composed of
packed columns in stainless steel for distillation and glass
for simulation column. In both cases, the height of packing
and diameter of columns are 0.8 and 0.1 m, respectively.
Two dumped packings were used: lox 10 mm ceramic or
glass Raschig rings and l o x 10 mm ceramic Berl saddles. A
vacuum system composed of a vacuum pump, a vacuum
tank and a pressure regulator can lower the operating pres-
sure to 1.104 Pa. Both devices were widely described in
Refs. [2] and [3]. Aside from these references, nothing has
been cited in literature concerning the use of reduced pres-
sure simulator. We would like to state that all the experi-
ments have been carried out using both kinds of packings
as well as for hydrodynamics and mass transfer. In order to
simplify details, only the results obtained with one kind of
packing were presented here.
3 Hydrodynamic Study

For both experimental apparatus, all pressure drops were

* Dr. B. Benadda, Dr. P. Moszkowicz, Dr. M. Prost, Prof. Dr. M. Ot-
terbein (author to whom correspondence should be addressed),
LCPAE, INSA de Lyon, 20 Av. A. Einstein, F-69621 Villeurbanne,
France, K. Mohammedi, INGM Departement d’EnergCtique,
DZ-35000, Boumerdes, Algeria.
measured with a differential manometer located at the top
and the bottom of the packing. A liquid tank gauge and two
electrically-operated valves were used to measure the total
liquid hold-up by gravimetric method [4].

0 VCH Verlagsgesellschaft mbH, D-69451 Weinheim, 1996 0930-7516/96/0408-0308 $10.00+ .25/0

Chem. Eng. Technol. 19 (1996) 308-314 309

Vacuum
system

Manometer

iquid
qowmeter

3 1

Figure 2. Experimental set-up: distillation column.

Figure 1. Experimental set-up: air/water simulator.

3.1 Air/ Water Simulator

3.1.1 Flooding Flow rates

I
For pressure values ranging from atmospheric pressure
down to 1.3.104 Pa, the flooding air mass flow rates Jan
1.4 - c.l

were determined whereas the liquid mass flow rate Jw in-

creased from 0.7 to 1.9 k g m P 2 s - ' . Using Raschig rings,
when Jw is equal to 0.7 k g m - 2 s - 1 the corresponding 1-
flooding flow rates decrease from 1.4 to 0.52 kg m - 2 s - '
and when Jw is equal to 1 . 4 k g m - 2 s - 1 they vary from
1.25 to 0.35 k g m - 2 s p 1 (s. Fig. 3). When Berl Saddles
packing is used in the same pressure range and Jwis equal
0.6 -
to 0 . 7 k g m p 2 s - ' , the Jaflvalues decrease from 1.46 to
0.64 kg m - 2 s p '. For both types of packings, at a fixed liq-
uid flow rate, flooding is first observable at low pressures.
At a fixed operating pressure, flooding air flow rate increas-
es when liquid flow rate decreases.
Oa2 1 P (kPa)

0 20 40 60 80 100 120
Figure 3. Flooding flow rate vs. total pressure (Raschig rings, air/water
3.1.2 Liquid Hold-ups
simulator).

The average experimental values of static hold-up obtained

with the air/water system are about 4 and 2% for Raschig
rings and Berl saddles, respectively. For dynamic hold-ups
&, in counter-current flow of gas and liquid phases and at
fixed values of water and air flow rates, when the total pres-
sure decreases, the values of a d decrease for Berl saddles
packing whereas they increase for Raschig rings as shown in
Fig. 4. The corresponding values of the factor d(1npd)/dP
are found to be +4.1OP6 (Raschig rings) and -6.10-6
(Berl saddles), respectively, in the range investigated. This
opposite behaviour of the liquid dynamic hold-ups, when
the packing is changed, is probably due to the different geo-
metric shapes of both packing.
 310 Chem. Eng. Technol. 19 (1996) 308-314

12
I I 4
*
n

a
0

A
13.3kPa
40kPa
67 kPa
n
E
2
a 101.3 kPa 3
W

8 N
a
\
a
a
2

4
1

Jair (kg.m-2s-1)
0 I I
0-
0 1 2 3 10 -l 10 O 10
Figure 4. Dynamic hold-up vs. water flow rate (Raschig rings, air/wa- J .(kg.m2 s - ' )
ter simulator). air
Figure5. Pressure drop vs. air mass flow rate at different pressures
(Raschig rings, air/water simulator).
3.1.3 Pressure Drops

Fig. 5 shows the variations of pressure drop versus air flow 2000
rate and operating pressure for Raschig rings. The wa-
ter-specific flow remains constant at 1.4 kg m P 2 s - ' . Pres-
sure drops per meter of packed height dP/dz may be ex-
pressed as: ')
i3
n

dP 4
-= K
dz
J b JZr P" &
W

1000
The values of parameters K, m, n, q are very similar for
both packings and are given in Tab. 1.

Inside the foreloading zone the average reliabilities between 500

experimental values and those obtained from both correla-
tions are found to be 4% (s. Fig. 6).

3.2 Distillation Column 0

0 500 3000 1500 2000
Liquid hold-up, pressure drop and flooding vapor flow rate
have been measured in a randomly packed column under to- Figure 6. Experimental pressure drop vs. calculated pressure drop val-
ues (air/water simulator).
tal reflux conditions ( J L = JG) using benzylchloride/
ethylbenzene binary mixture. Fig. 7 shows the variation of
pressure drop per meter of packing versus the specific mass
flow rate of vapor JG for pressure values ranging from at- mospheric pressure to 6.7 kPa. For the same JG values, the
pressure drop increases when pressure decreases. The flood-
ing flow rate variations versus pressure are represented in
Table 1. Values of parameters K, m, n and q (Eq. (1)). Fig. 8. The same remarks as for simulation may be done
~~~ ~

about corresponding distillation experiments, e.g., the in-

Packing K [ x LO*] m n 4 crease of flooding flow rate versus JG. Dynamic liquid
hold-up variations (in To of packed volume) versus vapor
Raschig rings 1.3 2.00 0.44 - 1.00
Berl saddles 1.3 2.03 0.42 - 1.00 flow rates were represented in Fig. 9. The obtained values
seem slightly lower when pressure is reduced. These curves
also allow the determination of flooding flow rates in agree-
1) List of symbols at the end of the paper. ment with the pressure drop variation curves.
Chem. Eng. Technol. 19 (1996) 308-314 31 1

4 6.7 kPa
n 133 kPa
+ 13.3 kPa
E A 26.6 kPa
a + 533 kPa 26.6 kPa
3 3
W
A 80 kPa
N
z 30-
%
2
20 -

10 -
1

n-
v

0 0.5 1 1.5 2
0 I
Figure 9. Dynamic hold-up vs. gas flow rate (Berl saddles, distillation
10 -l 10 O 10 column).

least from the quantitative point of view. For this reason we

Figure7. Pressure drop vs. gas mass flow rate at different pressures distinguish in two aspects.
(Berl saddles, distillation column).

3.3.1 Qualitative Aspect

3
As said before, distillation column pressure drops were de-
termined under total reflux conditions (JL= JG), whereas
under the different pressures used with the simulator JG is
progressively increased at fixed JL values. However, in
both cases, the corresponding curves dP/dz = f(JG) appear
2 similar and especially for the same mass flow rate JG. For
example, if JLis taken equal to JG it is always observed
that the pressure drop increases when the pressure decreas-
es. As for dynamic hold-ups, they increase slightly with va-
por mass flow rate before the loading zone and suddenly be-
fore the flooding zone. Furthermore, they decrease when
1 pressure increases for real distillation as well as for the sim-
ulator. In the same way flooding flow rates give similar
variations in both cases.
-P
P (kPa)
3.3.2 Quantitative Aspect
0 l . l . i ' l m l -

Whatever the binary mixture used, it may be observed that

Figure 8. Flooding flow rate vs. total pressure (distillation column, RR: the energy loss at flooding is always of the same order [ S ] .
Raschig rings, BS: Berl saddles, g: glass, c: ceramic). If two binary mixtures are considered, the corresponding
vapor phases are i and j , the following expression for a giv-
en packing can be written:

3.3 Comparison of Real Distillation

and Simulation

It is obvious that the values of hydrodynamic characteristics

obtained with an airlwater system in the simulator and
those obtained with a benzylchloride/ethylbenzene system where K , represents the pressure drop coefficient for a giv-
Y

in the distillation column cannot be directly compared, at en packing, let:

312 Chem. Eng. Technol. 19 (1996) 308-314

HETP = N1,2
-
Z

To compare two mixtures, if air is the reference gas phase, z being the packed height and N 1 , 2 the number of
the equivalent velocity UGes of the gas phase will be writ- theoretical stages between two levels obtained from the Fen-
ten: ske relation [7]:

1 /2 You, 1 -Yin
nl-
UGeq = Uair (4) 1 -You Yin
N1.2 = (7)
In a
or
in which a is the relative volatility, yinand yo,, the respec-

JGeq = Jair [$1 1/2
@G and e a i r are the specific gravities of the gas phase (or-
ganic vapor phase) and the air phase, respectively, under
corresponding conditions (temperature and pressure). As-
suming that differences between other physical properties
of the compounds are not significant influences and may be
neglected, the comparison between certain hydrodynamic
characteristics will be carried out when only taking into ac-
count the specific gravities of gas phases [6]. Tab. 2 shows
that for the three chosen pressure values and for the same
values of water mass flow rate Jwin the simulator and liq-
uid mass flow rate JLin the distillation column, observed
values for vapor flooding flow rate JGfl are close to equiva-
lent flooding flow rates JGeqfl calculated by Eq. (5).
Concerning pressure drops, before reaching the loading
zone and for the two packings, they are calculated for distil-
lation with an accuracy better than 10% (Fig. 10) using the
correlation (1) in which JGis substituted by JGeq (Eq. (5))
and where JL= Jw.
tive molar ratios of benzylchloride in the vapor phase at the
packing inlet and outlet. These ratios were determined by
gas chromatography. Fig. 11 gives the variation of HETP
versus mass flow rate J G at different pressures for Berl sad-
dles. It can be noted that HETP decreases when JGincreas-
es before flooding occurs and it is lower under reduced pres-
sure than at atmospheric pressure at a same JG.The col-
umn is therefore more efficient under reduced pressure.
Consequently, when pressure decreases or flow rate increas-
es, the separation is improved and mass transfer is facilitat-
ed. One of the key parameters of mass transfer which may
vary with pressure is the interfacial area. It would be inter-
esting to determine the variation of this parameter versus
flow rates and reduced pressure. Absorption gadliquid
with chemical reaction is one of determination methods of
interfacial area. But the design of our distillation column
does not allow application of this method. For this reason
and assuming that interfacial area was independent of the
chemical system involved [8], the interfacial areas of
Raschig rings and Berl saddles were determined in the simu-

4 Mass Transfer Study

4.1 Real Distillation -

Height Equivalent to a Theoretical Plate (HETP)

To estimate the efficiency of the total reflux distillation col-

umn using the binary mixture benzylchloride/ethylbenzene,
the HETP has been determined by the following expression:

Table 2. Comparison between flooding flow rate calculated and gas

flooding flow rate obtained in distillation column (RR: Raschig rings,
BS: Berl saddles).

Total pressure Packing Simulator Calculated Distillation

[Pal Jairfl flooding JGfl
flow rate JG,,,[kgm-2~-']

100000 RR 1.14 1.82 1.8

BS 1.18 1.87 1.9
19950 RR 0.58 1.05 1.08
BS 0.66 1.18 1.15 0 500 1000 1500
13300 RR 0.48 0.9 0.92
Figure 10. Experimental pressure drop vs. calculated pressure drop
BS 0.56 1.04 1.06
values (distillation column).
Chem. Eng. Technol. 19 (1996) 308-314 313

0
h
Berl saddles + 40kPa
E A 13.3 kPa
. a (m-') 101.3kPa
1
W
0
a
b A 40 kPa
500 I 59.3kPa
E 0 101.3kPa

0.3
400
\A' -0-0-

Oa2
0.1 1 I I
300

200

0 1 2
Figure 11. HETP vs. gas flow rate at different pressures.
3 100 : I I I

Figure 12. Interfacial area vs. liquid flow rate at different pressures.
4.2 Simulator - Interfacial Area

The chemical technique, with the absorption of oxygen into

sodium sulfite solutions catalysed by cobalt ions, was used
to measure the gadliquid interfacial area a of packed simu-
lator column.
The concentration of liquid phase was analysed by iodomet-
ric titration. The corresponding method for determination
of absorbed oxygen rate and interfacial area was previously
described in [3, 91. The interfacial area was determined in
the rapid pseudo m-nth order regime from the measurement
of the absorption rate @:
7 20070, the total pressure seems to have no significant effect
The dissolved gas undergoes a pseudo m-nth order reaction
when the concentration of reactant in the neighbourhood of
the surface is very little different from that in the bulk, and
if the reaction is rapid, the condition to be satisfied is:
3sHa<l+--- DB CBO
(9)
2D02 G,
where Ha is the Hatta number
Absorption rates were measured in solutions having cobalt
sulphate concentrations 5 . m o l l - ' and sodium
sulphite concentrations from 0.4 to 0.8 moll-' at p H = 8
and at 293 K. The kinetics parameters were determined with
the same chemical regimes in laboratory apparatus in which
interfacial area was known. This was evaluated by observ-
ing by stroboscopy and photographing the gas bubbles
moving in plug flow in 3 mm diameter capillary tube [3, 101.
The results for Berl saddles are represented in Fig. 12 and
show that the interfacial area a increases with liquid flow
rate for given values of gas flow rate and pressure. At at-
mospheric pressure and when the liquid flow rate JLin-
creases from 1.14 to 4.56 kg m - 2 s - l , interfacial areas of
Berl saddles and Raschig rings represent about 40 to 100%
and 15 to 100% of the geometric superficial area of each
packing, respectively. This confirms the general trend ob-
served with randomly dumped packings which are more ef-
ficient at high liquid flow rates.
Taking into account the reproducibility and the precision of
the measured values of interfacial area, estimated at about
on this parameter in the studied pressure range. We must
also add that the results are not reproducible at very low
pressures, we therefore limited our experiments to 40 kPa.
5 Conclusion
Pressure drops, flooding flow rates and even the variations
of hydrodynamic hold-ups occurring during the reduced
pressure distillation process with Raschig rings and Berl
saddles packings can be estimated accurately. Hydrody-
namic air/water simulations give qualitative and quantita-
tive parameters for distillation unit design.
The essential conclusion of the mass transfer study is to
state the similar behaviour of HETP and interfacial area
versus liquid flow rate and reduced pressure in the range we
investigated. As a particular observation, interfacial area
increasing with liquid flow rate may partially explain the
corresponding increase of efficiency with the same flow rate
values. The same observation can also be made concerning
the independence of HETP and interfacial area to pressure
only in the range comprised between 40 and 100 kPa.
314 Chem. Eng. Technol. 19 (1996) 308-314

Subscripts
Received: May 18, 1995 [CET 7511

eq equivalent
fl flooding
Symbols used G gas
L liquid
a interfacial area per unit volume of reactor P packing
cBO concentration of SO:- within the liquid V vapor
C02* concentration of oxygen in equilibrium W water
with the interface gas
Do2 diffusivity of oxygen in the liquid
DB diffusivity of SO:- in the liquid
References
dP pressure drop
Ha Hatta number
J fluid mass flow rate [ I ] Bressat, R., Goury, D., Otterbein, M., Vincent, M., Inf. Chim.
K constant of Eq. ( 1 ) 289 (1987) pp. 153- 158.
pressure drop coefficient [2] Messaoudene, A., Thesis, INSA Lyon 1989.
k,
Nl,2 number of theoretical stages between two [3] Mohammedi, K., Thesis, INSA Lyon 1992.
levels of packing [4] Charpentier, J . C., Van Swaaij, W., Le Goff, P., Chim. Ind. Ge-
P pressure nie Chim. 24 (1969) pp. 1083- 1095.
r rate of reaction [5] Bressat, R., Otterbein, M., Vincent, M., Znf. Chim. 274 (1986)
(I fluid velocity pp. 119-125.
Y molar ratio of benzylchloride [6] Mohammedi, K . , Bressat, R., Otterbein, M., Moszkowicz, P.,
Z packed height Prost, M., 4th World Congr. Chem. Eng.,Karlsruhe, Germany,
16-21 June 1991.
[7] Fenske, M. R . , Tongberg, C. O., Quiggle, D., Ind. Eng. Chem. 26
(1934) pp. 1167-1177.
Greek symbols 181 Laurent, A., Charpentier, J.C., Chem. Eng. J. 8 (1974)
pp. 85 - 101.
a [-I relative volatility [9] Andrieu, J . , Claudel, B., Chem. Eng. Sci. 29 (1974)
pd [m3/m31 dynamic liquid hold-up pp. 1263-1271.
0 [kmol m-3 s-'] absorption rate [lo] Benadda, B., Prost, M., Ismaily, S., Bressat, R., Otterbein, M.,
e specific gravity Chem. Eng. Process. 33 (1994) pp. 55 - 59.