Hydrodynamics and mass transfer in a packed column:

Case of toluene absorption with a viscous absorbent

Frederic Heymes, Peggy Manno Demoustier, Françoise Charbit, Jean-Louis
Fanlo, Philippe Moulin

To cite this version:

Frederic Heymes, Peggy Manno Demoustier, Françoise Charbit, Jean-Louis Fanlo, Philippe Moulin.
Hydrodynamics and mass transfer in a packed column: Case of toluene absorption with a viscous
absorbent. Chemical Engineering Science, 2006, 61 (15), pp.5094 - 5106. �10.1016/j.ces.2006.03.037�.
�hal-01916631�

HAL Id: hal-01916631

https://hal.science/hal-01916631
Submitted on 8 Jun 2021

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
 Hydrodynamics and mass transfer in a packed column: Case of toluene
absorption with a viscous absorbent
Frederic Heymes a , Peggy Manno Demoustier a , Françoise Charbit b , Jean Louis Fanlo a ,
Philippe Moulin b,∗
a LGEI, Ecole des Mines d’Alès, 6 avenue de Clavières, Alès 30319, France
bLPPE-UMR 6181, Université Paul Cezanne, Batiment Laennec, Hall C, Europole de l’arbois, BP 80, Aix en Provence 13545, France

Abstract
Volatile organic compounds (VOCs) cause nuisance to humans and the environment. Recent legislation encourages industrialists to set up
equipment for treating their VOC-loaded gaseous effluents. This piece of research studies the absorption process, using a viscous organic
absorbent (di(2-ethylhexyl) adipate = DEHA) to treat a toluene-loaded vent gas, in terms of hydrodynamics and mass transfer. It is shown that
DEHA does not lead to an excessive pressure drop. Correlations predicting hydrodynamic parameters from previous literature are summarised
and tested against experimental results. It is shown that acceptable prediction accuracy can be achieved for counter-current pressure drop and
liquid hold-up. Treatment efficiency for the toluene-loaded vent gas is shown to be very good. Calculation of mass transfer constants (kL a)
enables to test literature correlations against the experimental results. The mass transfer is supposed to be limited by the liquid-side resistance.
Our experimental results showed that the kL a of the system depends on the liquid velocity but also on the gas velocity. This behaviour has
also been observed by the few authors who have used viscous fluids in their experiments, but is contrary to all the authors who have work
on low-viscosity fluids. It is therefore clear that the influence of viscosity on the phenomenon is considerable. Not one current correlation is
currently accurate in the case of a viscous absorbent.

Keywords: Absorption; Hydrodynamics; Mass transfer; Environment; Chemical processes

1. Introduction This paper focuses on the absorption method. The most im-
portant point concerning this process is the choice of a suitable
Atmospheric volatile organic compounds (VOCs) are in- absorbent liquid. In the case of hydrophobic VOCs, water can-
volved in photochemical reactions that create ground-level not be used therefore other kinds of absorbents are required:
ozone (urban smog) and entail environmental threats such as water–oil emulsions, water–solid suspensions or liquids with
global warming, acid rain and air-borne toxics. In order to high boiling points. Several authors studied the possibility of
reduce anthropic emissions, legislation such as the Clean Air using such absorbents for industrial air cleaning processes. For
Act Amendments (1990) and EEC Directives (1999, 2001) toluene absorption, experiments carried out in a bubbling de-
specify maximum concentration levels for industrial emissions vice (Heymes et al., 2006) have shown that di(2-ethylhexyl)
into the atmosphere. Various processes are available for VOC adipate (DEHA) seems to be an efficient absorbent due to its
abatement, such as thermal or catalytic oxidation, adsorption, good toluene affinity, acceptable viscosity, high diffusion coef-
condensation, absorption, membrane permeation and biolog- ficient, low vapour pressure, and because it is non-toxic, non-
ical treatments. No single method can be used in all cases: explosive and inexpensive (Table 1).
most of them are very specific in nature. However, viscosity of DEHA is more than 10 times greater
than that in case of water (14.4 mPa s at 20 ◦ C). This point is a
∗ Corresponding author. Tel.: +33 4 42 90 85 05; fax: +33 4 42 90 85 15. major concern, since viscosity plays an important role in hy-
E-mail address: philippe.moulin@univ.u-3mrs.fr (P. Moulin). drodynamics and mass transfer kinetics of an absorption tower.
Table 1
Properties of pure DEHA and DEHA/toluene mixtures

Pure DEHA properties Toluene/DEHA binary mixture

Formula C22 H42 O4

Molar weight 350.8 g mol−1 Activity coefficient (
) 25 ◦ C 0.50
Density 930 kg m−3 Henry’s law constant (H ) 25 ◦ C 1.98 kPa
Fusion temperature −76 ◦ C Viscosity (20 ◦ C) 12.5 mPa s
Vaporisation temperature 210 ◦ C 8.65 × 10−10
Vapour pressure (20 ◦ C) 3.25 × 10−6 Pa Diffusion coefficient of toluene m2 s−1
in DEHA at infinite dilution (25 ◦ C)
Cost < 3¥ L−1

The first part of this work studies hydrodynamics (pressure Numerous attempts were made to model hydrodynamics by
drop, liquid hold-up and flooding point) of counter-current op- means of dimensionless parameters taking into account the
eration. The influences of gas and liquid flow rates are evalu- properties of the system together with the gas and liquid flow
ated. Correlations from previous literature are summarised and rates. These relationships are indispensable for scaling the col-
tested against our results. umn and fluid flow apparatus in industrial processes. Table 2
The second part studies toluene removal performance and summarises a non-exhaustive list of recent works on hydrody-
mass transfer kinetics for DEHA flowing in a closed loop. namics in packed columns. All authors constituted a large data
Overall mass transfer constants KL a and KG a are calculated set of several thousands experimental points and most of them
and presented. It is shown that there is a lack of knowledge included viscous absorbents (engine oil, etc.). Previous system-
about predicting transfer constants exists in case of a viscous atic evaluations of the several correlations reached the conclu-
absorbent. sion that the Billet (1990) and Billet and Schultes (1991, 1992,
1995), the Mackowiak (1990, 1991) and the generalised pres-
2. Theoretical considerations sure drop correlation (GPDC) (Leva, 1992) are recommended
for predicting hydrodynamics in randomly packed columns pro-
2.1. Hydrodynamics of a counter-current gas–liquid flow in a viding the packing constants (packing factor). If packing con-
packed column stants are not available, the Miyahara correlation (1992) can be
used. The statistical method of Piché et al. (2001) is a recent
Optimising the operation of a packed column gas absorption
alternative to predict hydrodynamics in a packed column. Piché
system requires a thorough understanding of hydrodynamic pa-
et al. used artificial neural networks and dimensional analysis
rameters (pressure drop and liquid hold-up), which depend on
(ANN-DA) to model both hydrodynamics and mass transfer in
many variables such as the flow rates and characteristics of
case of counter-current flows, with the advantage of not requir-
the moving fluids under the operating conditions (temperature,
ing any packing constant.
pressure, physico-chemical properties) together with the type
By comparing the accuracy of the different correlations,
of column packing (size, shape, porosity). A very large num-
it appears that the Piché and Billet correlations forecast the
ber of studies have provided, in different cases, either semi-
loading and flooding points, pressure drop and liquid hold-up
theoretical or totally empirical relationships.
with a precision higher than 20% (Piché et al., 2001). The
Various authors have tried to explain the experimental re-
Mackowiak formula presents a prediction accuracy of 20% on
sults by a phenomenological approach. Indeed, three different
his data set whereas Miyahara obtains a precision of 30% with
hydrodynamic regimes can be observed in the counter-current
his data. The GPDC is accurate in the case of aqueous solu-
flow: (i) below the loading point, where gas and liquid flow
tions but deviates for non-aqueous systems (average absolute
without noticeable interaction (the liquid hold-up is indepen-
error of 60%). In order to give recommendations for indus-
dent of the gas velocity); (ii) the loading zone, where the shear
trialists interested in air treatment with high boiling point oil,
stress between gas and liquid increases first slowly then much
it seemed relevant to compare the prediction accuracy of the
faster until the flooding point is reached (the liquid hold-up in-
Takahashi and Miyahara, Billet and Mackowiak correlations on
creases with the gas flow rate) and (iii) the flooding point which
our measurements. The neural correlation of Piché was not in-
corresponds to a critical hydrodynamic state when, at a given
vestigated because of the required and non-published weights
liquid throughput, the gas flow rate becomes high enough to
of the neural network. The GPDC was not tested.
trigger liquid overflow from the column.
From an industrial point of view, it is of interest to predict
the pressure drop per unit of height
P /Z (Pa m−1 ) and the 2.1.1. The work of Takahashi et al. (1979) and Miyahara
total liquid hold-up hL (simply referred to as the liquid hold-up) et al. (1992)
whatever the hydrodynamic regime is. Moreover, it is important The correlations of Miyahara and Takahashi are summarised
to calculate the flooding point (defined by the values of the gas in Table 3. The method of Zenz and Eckert is used to predict
velocity at flooding point UG,F for each liquid velocity UL ) flooding and loading points with a single experimental constant
which should not be reached during normal operation. . For the loading point, Miyahara gives  = 0.547; for the
Table 2
Literature review about hydrodynamics in a packed column

Year Author Predicted variables Range of viscosity (mPa s)

2003 Iliuta et al. (2003) Pressure drop, liquid hold-up, wetted area ≈1
2001 Piché et al. Pressure drop, liquid hold-up, flooding point, wetted area 0.09–50
1993 Billet et al. Pressure drop, liquid hold-up 0.26–185
1992 Leva Pressure drop, flooding point 0.2–20
1992 Miyahara et al. Pressure drop, liquid hold-up 1.0–17.0
1991 Mackowiak et al. Pressure drop, liquid hold-up, flooding point 0.2–8

Table 3
Formulae of Miyahara and Takahashi

Flooding and loading points √ =  with

X 0.5 + Y 0.5

X = UL,F Px S and Y = UG,F Py S
and
0,2 0,2
   a
Px = L Py = G L S= 3
g L g 
 0.75
1.53 × 10−4 L
Liquid hold-up hL = + 1.42 × 10−5 0.66 dp−1.2
dp1..2 H 2 O
 
Z −0.35 0.8 −0.16
+ 4.11 × 101 Ca M FrG (1)
dp

P
P 0
Pressure drop Miyahara: = +  L g hL (2)
Z Z

 2

P
P 0 UG
Takahashi: = + k h3L (3)
Z Z  − hL

flooding point Miyahara gives  = 0.629 whereas Takashi from simple experiments. Billet considers that the liquid hold-
suggests  = 0.552. The liquid retention is divided into three up in the loading zone follows a polynomial behaviour versus
contributions (static hold-up, dynamic hold-up and operating gas velocity, with boundary conditions corresponding to the
dynamic hold-up). The pressure drop in counter-current flow is flooding point and the flow without gas. He showed that the
deduced from the liquid retention by adding the dry pressure polynomial is a 13 degree equation. Billet then calculates the
drop to the wet pressure drop. Miyahara and Takahashi do not total pressure drop by estimating the ratio of the wet column
suggest the same model for the dry pressure drop: Miyahara pressure drop (
P ) over the dry column pressure drop (
P0 ).
considers that the wet pressure drop is proportional to the This ratio depends on the hydrodynamic liquid flow represented
liquid retention whereas Takashi considers the pressure drop by the parameter f (S) and by the interaction between gas and
as a function of h3L /( − hL )2 . Both require an experimental liquid represented by the /( − hL ) term.
constant k or .
2.1.3. The work of Mackowiak (1990, 1991)
2.1.2. The work of Billet (1990) and Billet and Schultes Mackowiak uses a model of bed of suspended drops to pre-
(1991, 1992, 1995) dict the velocity of the gas at the flooding point (Table 5). The
Billet starts from phenomenological observations to develop model consists of two linked equations enabling to calculate
his theory (Table 4). He does not differentiate between static both the flooding gas velocity and flooding liquid hold-up. Liq-
and dynamic hold-up and models directly the total liquid hold- uid hold-up before the loading point depends on its velocity
up by dissociating the behaviour of the column according to the and enables the calculation of the liquid hold-up in the load-
operating zones and the liquid flow regime (Reynolds number). ing zone by a polynomial function of the gas velocity. Pressure
Before the loading point, Billet considers the liquid hold-up as drop can be calculated from the liquid hold-up by way of a
a function of the liquid velocity and the hydraulic surface ah . drag coefficient G,L which depends on the liquid Reynolds
This surface is calculated by considering inertia, viscosity and number. Mackowiak’s theory requires a flooding factor CF l and
gravity forces acting on the liquid. A packing parameter Ch is a packing shape factor M . The author gives many values for
required and is available in the literature or it can be calculated CF l and M according to the packing used.
Table 4
Formulae of Billet and Schultes
 1/3
12L UL a 2  a 2/3
h
Liquid retention before loading point hL,S = (4)
g L a

where
ah
= Ch Re0.15
L FrL
0.1 if ReL < 5
a

and
ah
= 0.85Ch Re0.25
L FrL
0.1 if ReL
5
a

 0.05
L H 2 O
Liquid retention at flooding point hL,F = 2.2hL,S (5)
H 2 O L

 n
UG
Liquid retention at loading zone hL = hL,S + (hL,F − hL,S ) (6)
UG,F
 1,5  
Pressure drop before loading point
P  ReL
= fB where fB = exp (7)

P 0  − hL 200

Pressure drop at loading zone  1,5    

P  hL 0,3 ReL
= fB where fB = exp (8)

P 0  − hL hL,S 200

Table 5
Formulae of Mackowiak

     
Flooding point dh 1/4 dT L g 1/2 hL,F 7/2
UG,F = CF l 1.2 1− (9)
dT G 

L
where dT =
( L − G ) g

For ReL > 2 hL,F = [(1.44 20 + 0.8 0 (1 − 0 ))2 − 1.2 0 ]
0.4(1 − 0 )

Liquid retention at flooding point For ReL < 2 hL,F = [(1.254 20 + 0.48 0 (1 − 0 ))2 − 1.12 0 ]
0.24(1 − 0 )
UL
where 0 =
UG,F
 1/3

L UL 1 − 
Liquid retention up to loading point hL,S = 2.2 BL where BL = (10)
L g 2 3 d p

  1/2
(FG /FG,F ) − 0.65
Loading zone hL = hL,F − (hL,F − hL,S ) 1 − (11)
0.35

  
Pressure drop
P G,L h h −3
= 1+ L 1− L (12)

P 0  1− 

2.1.4. Conclusion different due to the initial assumptions and subsequent de-
The principal models available for predicting pressure drop velopments. Takahashi and Billet separately calculate step by
and liquid hold-up were presented. While the parameters taken step the different hydrodynamic parameters of interest (flood-
into account are similar, their formulations are completely ing point, liquid retention, pressure drop) while Mackowiak
proposes a different approach where flooding point and liq- measured under the conditions of 46 different systems. The
uid hold-up are calculated together according to a suspended equations define and take into account the interfacial area ac-
droplet bed approach. tually involved in the transfer:
The different authors calculate the wet pressure drop from awet
the dry pressure drop. Miyahara and Takahashi add a term = 1.5(adh )−0.5 Re−0.2 0.75 −0.45
L WeL FrL , (16)
a
reflecting the liquid hold-up effect on gas pressure drop. Billet
   
and Mackowiak consider that the influence of liquid flow on  g 1/6 DL 1/2 2/3 1/3  awet 
the gas pressure drop can be represented by multiplying the kL awet = CL∗ G a UL , (17)
G dh a
dry pressure drop with corrective terms. More details about the
different correlations were not described in this paper, but the  3/4
∗ 1 a 3/2 U G G
references were listed. kG awet = CG D
1/2 G
( − hT )1/2 dh aG
 1/3 
2.2. Mass transfer in a counter-current flow G awet 
× . (18)
G D G a
Delaloye (1991) published a summary of the work carried
out in the modelling of mass transfer coefficients. It appears that CL∗ , CV∗ are experimental constants of the Billet equations.
even the most recent of the studies listed date from more than Billet proposes the values CL∗ = 1.577 and CV∗ = 0.390 for
20 years ago and the authors devoted their efforts to systems Hiflow rings of a nominal diameter of 25 mm.
whose viscosity is close to that of water (1 mPa s). Delaloye The approaches of these authors are similar in terms of mod-
(1991) highlights the inaccuracy of all these correlations as soon elling. They correlate the real wetted interfacial area awet with
as the viscosity exceeds 10 mPa s. DEHA, which is used in this the physico-chemical properties of the system and deduce the
study, is therefore outside the previously researched domains. mass transfer coefficients for the gas from the velocity of the
Since this work, only Billet and Schultes (1993, 1999) has gas and the mass transfer coefficients of the liquid from the
endeavoured to improve the modelling of the mass transfer velocity of the liquid. Only the forms of the equations differ.
coefficients, and he did not focus his work on viscous fluids.
Two approaches will be presented and tested in this paper: one 3. Materials and methods
is old but much used: Onda, whereas the other is that of Billet.
Hydrodynamics and mass transfer experiments were per-
2.2.1. The Onda model (1968) formed with di(2-ethylhexyl) adipate supplied by BASF.
This is the model most commonly used for predicting the
surface area of interfacial exchange and the mass transfer con- 3.1. Hydrodynamics
stants for both gas and liquid. It introduces the distinction be-
tween the total surface area provided by the packing and the Fig. 1 shows the pilot unit used for this study. It is made of
surface area actually wetted by the liquid, which is where the Schott glassware parts. The column has an internal diameter
mass transfer takes place. This wetted surface area awet is a
function of the flow rate and the properties of the liquid and
the packing. Onda proposes the equation: gas
  0.75 
awet c −0.05 2
= 1 − exp −1.45 Re0.1
L FrL We0.2 . (13) 2
a L L
8

Re, Fr and We are the dimensionless numbers of Reynolds, liquid

Froude and Weber. The formula considers the ratio of packing
and liquid surface tensions c /L . The local transfer coeffi-
cients for the liquid (kL ) and gas (kG ) are then calculated from
the following relationships giving the Sherwood numbers: 1

4/3 −1/2 10 3
ShL = 0.0051ReL,wet ScL (adp )0.4 , (14)
FI 5
1/3 −2.0
ShG = 5.23Re0.7
G,wet ScG (adp ) . (15)
TI FI
2.3. The Billet model (1993, 1999) 4
7 6
11 TI
Billet et al. developed a model enabling mass transfer co-
efficients to be determined for both the gas and the liquid. 9 7
The proposed correlations were established from a bank of
experimental results containing more than 3500 sets of data, Fig. 1. Flow sheet of hydrodynamics experimental setup.
 measurement of the pressure drop undergone by the gas current
circulating alone or counter-current to the liquid, and the liquid
hold-up in the column according to the following experimental
protocol. The complete setup was filled with absorbent. The
pump was stopped and, after a while, the mass of liquid on the
balance (11) was measured. Then, the pump was switched on
at a given flow rate. After stabilisation, the total hold-up was
determined by the difference in weight: the missing liquid in
the tank was held by the column. For each liquid flow rate,
the gas flow was increased incrementally up to flooding point.
At each incremental step, we waited until the hydrodynamic
SCALE
equilibrium was reached because the sudden increase in the
(mm) 0 10 20 30 40 50 gas flow perturbs the system, which then gradually returns to
a state of the hydrodynamic equilibrium in which the liquid
Fig. 2. Hiflow rings. hold-up and pressure drop are constant. Experimental investi-
gation at that point showed that waiting 5 min was sufficient to
obtain stable and accurate data. The range of fluid flow rates
of D = 0.1 m and a total height of 2.5 m. The packing (1) was L = 2.9–9.8 kg m−2 s−1 and G = 0–3.5 kg m−2 s−1 . Since
consists of a stack of Hiflow rings (Techim, France) to a height the height of packing was 1 m,
P /Z was directly read on the
of Z=1 m. The Hiflow rings (Fig. 2) are made of polypropylene U tube and hL was calculated by dividing the missing volume
(c =0.03 N m−1 ) wires and are a highly porous type of packing of absorbent by the cross-section of the column.
(void fraction  = 0.92). Because of this high porosity, Hiflow The accuracy of the different correlations to assess our ex-
rings are suitable for high gas flow rates and are therefore well perimental results was evaluated by the statistical indicator av-
suited for the treatment of industrial gaseous effluents. The erage absolute relative error (AARE) defined as follows:
particle diameter is dp = 17 mm and the specific area is a =
275 m2 m−3 . The ratio D/dp equals 5.9, but according to Baker 1
n
X theory − X experimental
this ratio is not sufficient and wall flow could occur. However, AARE = for n data.
n X experimental
such an effect was not observed experimentally. The packing i=1
was held in place top and bottom by grids whose porosity was
close to 100%. 3.2. Mass transfer
The column was topped with a droplet collector (2) fabri-
cated in the laboratory. This device was made of many grid The experimental apparatus is represented in Fig. 3. Toluene-
layers in a cylindrical shape collecting the droplets. Two pres- charged gas was generated by means of a micrometric pump
sure taps situated under and over the retainer grids enabled (3) which injected toluene into a current of compressed air (2).
the pressure drop created by the packing to be measured. The mixture was pumped into a heated flask to ensure that the
These two taps were connected to a vertical U-shaped tube toluene was completely evaporated. The flow of polluted air
filled with water (3). The apparatus was fed with air by means was injected into the main air inlet current, whose flow rate
of a 1.5 kW fan (4) (LAFERT, France). The maximum load was measured using a hot-wire anemometer (5). The mixed gas
was 2.8 m of water and the maximum flow rate 180 m3 h−1 . flow was homogenised by means of a static mixer (L = 1 m)
The air flow was regulated by means of a butterfly valve (Durapack, Schott, France) before being introduced into the
(5) placed after the fan. The velocity was measured using a column.
hot-wire anemometer (Kimo, France). The temperatures of The aim was to study the mass transfer of toluene present in
the gas and liquid (7) were measured by means of a thermo- the gaseous phase to the DEHA. The setup operated in open
couple (Kimo, France). The column was fed with washing loop for the gas and closed loop for the liquid. This implied
liquid through a conical sprinkler (8) by means of a cen- that the concentration in the liquid was not constant at the
trifugal pump (SALMSON 700 W, France) with a maximum inlet of the column. Sampling points for the liquid at the top
of flow of 3.6 m3 h−1 and a maximum load of 38 m water. and the bottom of the column (1) allowed determination of the
The flow rate through the column was regulated by means toluene concentration in DEHA by means of a UV spectrometer
of two valves situated on the recirculation by-pass and the (Secomam, France) at a wavelength of =259.9 nm. Air sample
column input. The liquid flow rate was measured using a flows were pumped at the head and foot of the column and were
water-meter type volumetric counter (10) (M2RSP-1H, Mac- analysed each second using a chromatograph (6, HP 5890).
naught, Australia). A 10 L liquid reservoir was placed on a Results were averaged to obtain a point per minute.
scale (Mettler PJ 6000, France) (11). The protocol for measur- A known volume of solvent was introduced into the ther-
ing the static hold-up was as follows: the initially dry packing mostated balloon flask and moved through the by-pass by the
was wet with the absorbent and after a sufficient lapse of centrifugal pump. The coil of water in the balloon flask bal-
time, the volume of undrained liquid corresponds to the static anced out the heat imparted to the fluid by the centrifugal pump
hold-up. The experimental procedure enabled simultaneous so as to work at constant temperature (20 ◦ C). The micrometric
 6

7
FI

TI FI
FI
2 5
TI
4
1 3

Fig. 3. Flow sheet and picture of mass transfer experimental setup.

pump was set to inject the correct amount of toluene for the de- QG (CG,in − CG,out (t))
sired air flow and concentration. The column outlet was moni- (CG,in /H −CL,out (t))−(CG,out /H −CL,in (t))
tored by means of FID to ensure that the toluene concentration = KL aV C   ,
CG,in /m − CL,out (t)
in the gas was stable and equal to the desired value. Time t = 0 Ln
corresponds to the opening of the valve to introduce the DEHA CG,out /H − CL,in (t)
into the top of the column. The flow rate was quickly set to (20)
the desired value by means of this valve and that of the by- QG (CG,in − CG,out (t))
pass. From that moment, the absorption started and the toluene (CG,in −H C L,out (t))−(CG,out −H C L , in(t))
concentration in the liquid rose little by little. The toluene con- = KG aV C .
Ln((CG,in −H C L,out (t))/(CG,out (t)−H C L,in (t)))
centration at the gas outlet fell quickly because of the absorp-
(21)
tion by the DEHA and rose after a while to its initial value
when DEHA was not able to absorb any more toluene. Then Since the liquid flowed in a closed loop, the inlet liquid
the experiment was over. Experiments could last up to 40 h. concentration varied with time. This unsteady state required a
They were carried out in batches using 10 L of DEHA. The mathematical treatment of the experimental points to calculate
temperature of the DEHA was kept constant throughout the ex- KL a and KG a by linear regression. H is the equilibrium con-
periments. The temperature of the air varied between 22 and stant (Henry’s law constant) between the liquid and gas phases
31 ◦ C. This variation of the air temperature did not influence defined as CG = H C ∗L .
the absorption results since the liquid temperature was stable.
The toluene concentrations and air and liquid flow rates tested 4. Results and discussion
were as follows:
4.1. Experimental results
(i) toluene concentrations in gas phase: 500; 1000; 5000
mg m−3 ; Experimental points at constant liquid flow rate are plotted
(ii) liquid flow rates L = 6.58; 9.87; 13.16 kg m−2 s−1 ; in Fig. 4. Pressure drop [A] and liquid hold-up [B] are given in
(iii) gas velocities: 0.42; 1; 1.5 m s−1 (G = 0.51; 1.21 and log–log scale. Flooding data were determined and is indicated
1.81 kg m−2 s−1 ). in the chart [C] included in Fig. 4. The experimental points
for the pressure drop under the flooding point can be well
This corresponds to L/G values ranging from 3.6 to 25.9.
represented by a line with a slope equal to 2.26 (corresponding
The gas and liquid flow rates were chosen from the results of the
to the exponent of the gas velocity). Concerning liquid hold-up,
hydrodynamic study in order to explore an interesting range of
the resulting curves correspond to the classical behaviour, as
L/G values. The values of the gas phase toluene concentration
described by Billet. The Mackowiak assumption, which states
at the column outlet CG,S (t) were used at 1-min intervals in
that the loading zone begins at 65% of the flooding point, is
the following equations. The toluene removal performance is
quite fairly respected.
defined as 100 × (CG,in − CG,out (t))/CG,in . The mass balance
Fig. 5 shows the experimental results for toluene recov-
in the column was verified by the equations:
ery performance with three different gas flow rates. It appears
that the ranges of flow rates and concentrations investigated
QG (CG,in − CG,out (t)) = QL (CL,out (t) − CL,in (t)), (19) give an initial toluene removal performance close to 100%,
 1000 toluene removal according to the gas or liquid flow rates used.
(A)
The influence of the gas and liquid flow rates versus time is
L (kg.m-2.s-1) 2.9 4.2 6.5 9.8
-1
very marked, depending on the L/G ratio.
UG.E (m.s ) 2.9 4.2 6.5 9.8
3 -3
hD,S (m .s ) 0.157 0.181 0.198 0.245
However, these qualitative results cannot be extrapolated. It
(C) 9.87 kg.m-2 .s-1
is therefore necessary to calculate the gas and liquid transfer
6.58 kg.m-2 .s-1
-2 -1 constants KL a and KG a. For that purpose, it was necessary
4.28 kg.m-2.s-1
2.96 kg.m .s
dry L = 0 to calculate the Henry’s law constant H between the liquid
Pressure drop (Pa.m-1)

and gas phases. Since all experiments were performed until the
equilibrium state between the gas and liquid was reached, the
equilibrium constant was determined by plotting CG versus CL∗ .
0 1 It was found that H = 1.96 × 10−6 at 20 ◦ C. The values of the
Liquid holedup (m3.m-3)

(B)
global mass transfer coefficients were then calculated using the
method indicated previously and they are listed in Table 6.

-2 -1 4.2. Discussion
L = 9.87 kg.m-2.s-1
L = 6.58 kg.m-2.s-1
L = 4.28 kg.m-2.s-1
L = 2.96 kg.m .s
0.1
4.2.1. Hydrodynamics
0.5 1
-1
5 The prediction of dry pressure drop is not presented here be-
Superficial gas velocity(m.s )
cause of its low interest. The Ergun formula is accurate, but
10 constants hK =1682 and hB =1.665 have to be used. Miyahara,
0.5 1 5
Billet and Mackowiak formulae were tested with our experi-
Superficial gas velocity (m.s-1)
mental data. Very good accuracy was observed (3–5%). For wet
Fig. 4. Pressure drop, liquid hold-up and flooding points as a function of gas pressure drop and liquid hold-up, the correlations presented in
and liquid flow rates. the theoretical part of this paper were compared with our ex-
perimental results.

100% 4.2.1.1. Miyahara and Takashi modelling. In flooding point

1 UG = 0.4 m.s-1 L/G = 12.9 prediction a significant error appeared when using Miyahara
90% constants. The closest prediction of the flooding point is ob-
2 UG = 1.0 m.s-1 L/G = 5.4
served when the value of  is taken to be 0.69. The AARE is
80%
3 UG = 1.5 m.s-1 L/G = 3.6 then 1.5%. This value is higher than those given by Miyahara
Toluene removal performance

70%
and Takahashi (resp.  = 0.629 and  = 0.552) for the packing
systems they used. This results from the fact that our system
60% floods at higher gas and liquid flow rates, which confirms the
favourable hydrodynamic behaviour of Hiflow rings.
50% Takahashi divides liquid hold-up into three terms: static, dy-
namic and operating dynamic liquid hold-up. Static hold-up
40% is very poorly predicted, since the correlation gives a value
1
hs = 0.020 m3 m−3 while the experimental measurement is
30%
0.008 m3 m−3 . This is not surprising since Miyahara’s formula
2
20% does not take into account the surface tension and density of
the liquid. Concerning the dynamic and operating hold-up, the
10% predictions are more accurate. The relationship predicting the
3 dynamic liquid hold-up has an AARE of 11%, whereas the re-
0% lationship predicting the operating liquid hold-up has an AARE
0 10 20 of up to 22%. The author did not specify the range of use, but it
Time (hours) seems that the proposed correlation while simple to use, is not
Fig. 5. Toluene removal performance as a function of time and gas superficial
appropriate for high liquid flow rates. Concerning the pressure
velocity (L = 6.58 kg s−1 m−2 ). drop, Miyahara and Takahashi propose adding a liquid term to
the dry pressure drop. The inaccuracy concerning the prediction
of the operating liquid hold-up has an impact on the prediction
whatever the configuration. This confirms the interest of DEHA of the wet pressure drop. Takahashi correlation is reasonably
as an efficient absorbent for treating a gas loaded with toluene, satisfactory (AARE = 21%).
with concentrations varying between 500 and 5000 mg m−3 .
Nonetheless, this performance decreases very quickly when the 4.2.1.2. Billet modelling. The correlations proposed by
liquid becomes enriched with toluene. Fig. 5 shows this drop in Billet proved very accurate for predicting our results
Table 6
Experimental overall mass transfer constants KL a and KG a

Experiment number [toluene] (mg m−3 ) L (kg m−2 s−1 ) G (kg m−2 s−1 ) KL a (s−1 ) KG a (s−1 )

1 4990 13.16 0.51 2.52 × 10−4 2.75

2 1205 13.16 0.51 2.51 × 10−4 0.79
3 1193 9.87 0.51 3.69 × 10−4 1.89
4 5038 6.58 0.51 3.23 × 10−4 1.34
5 520 6.58 1.16 2.28 × 10−3 13.5
6 1005 6.58 1.21 2.38 × 10−3 13.8
7 1005 6.58 1.81 2.43 × 10−3 12.1

900 900
L = 9.8 kg.m-2.s-1
800 L = 9.8 kg.m-2.s-1 800
L = 6.5 kg.m-2.s-1
∆P experimental [Pa.m-1]

∆P experimental [Pa.m-1]
L = 6.5 kg.m-2.s-1
700 700 L = 4.2 kg.m-2.s-1
L = 4.2 kg.m-2.s-1
L = 2.9 kg.m-2.s-1
600 L = 2.9 kg.m-2.s-1 600
500 500
400 400
300 300
200 200
Z

Z
100 100
0 0
0 100 200 300 400 500 600 700 800 900 0 100 200 300 400 500 600 700 800 900
∆P calculated [Pa.m-1] ∆P calculated [Pa.m-1]
Z Z

Fig. 6. Parity graph of Billet’s formulae. Fig. 7. Parity graph of Mackowiak’s formulae.

concerning the loading zone. In fact, Billet’s correlation gives Mackowiak’s work enabled the total pressure drop to be es-
a satisfactory prediction for the liquid hold-up (AARE = 6%). timated with an error of 15% (Fig. 7).
In the loading zone, a divergence was observed when the liquid
flow rate increased (AARE from 5% to 19% for liquid flow 4.2.1.4. Influence of viscosity. The Miyahara, Billet and Mack-
rates of 3.2 × 10−3 to 10.6 × 10−3 , respectively). These diver- owiak formulae presented good prediction accuracy. Our data
gences are mainly due to the zone close to the flooding point. can be completed with the work of Cotte (1996), who stud-
Fig. 6 highlights the good predictions for pressure drops given ied the hydrodynamics of air/water/Hiflow ring and air/PEG
by Billet’s relation provided the system is not too close to 400/Hiflow ring systems. All data could not be compared rigor-
flooding. ously because of the different gas and liquid flow rates used by
the author. Nonetheless, examining the whole set of values re-
vealed similarity of the reported points (Table 7). The three ex-
4.2.1.3. Mackowiak modelling. Iteration of the Mackowiak perimental points correspond to the beginning of loading zone.
equations enables the calculation of the velocities and liquid As certain authors have observed (Billet and Schultes, 1991;
hold-up at the flooding point. The most accurate predictions Cotte, 1996), the viscosity of the liquid considerably influences
were obtained by taking a droplet diameter dT = 2.5 mm. The the hydrodynamic behaviour of the column. Indeed, the viscos-
velocities at the flooding point were predicted at 4% and the ity of the three liquids varies by a factor of 80. The Reynolds
liquid hold-up at 8%. Concerning liquid hold-up at zero gas number was calculated according to the definition of Billet and
flow rate, Mackowiak’s equation gives approximate results Mackowiak.
with the coefficient proposed by the author. A deviation of Table 8 presents the predicted results of Billet and Mack-
37% was observed. By modifying the empirical coefficients in owiak for liquid hold-up and pressure drop in comparison with
the correlation, an accuracy of 6% could be achieved. The new experimental data. It is obvious that Billet’s formulae predict
correlation for predicting the total dynamic hold-up at zero gas results with a good accuracy (9% in case of liquid hold-up,
flow rate should be written as 18% in case of pressure drop). The Mackowiak formula gives
 1/3 bad predictions for liquid retention (AARE = 54%) and pres-

L UL 1 −  sure drop. With the modified constant, the result is better (25%)
hL,S = 3.61
 BL with BL = .
L g 2 2 dp but still insufficient to have an accurate prediction about pres-
modified value sure drop. This reveals that Mackowiak’s formulae require a
Table 7
Comparison of experimental data with literature

Author Liquid UG (m s−1 ) L (kg m−2 s−1 )  (25 ◦ C) (Pa s) Re

Cotte (1996) Water 1.0 4.2 0.0010 15

This work DEHA 1.0 4.2 0.0144 1.0
Cotte (1996) PEG 400 1.0 3.9 0.0820 0.17

Table 8
Comparison of Billet and Mackowiak predictions versus experimental results

Pressure drop (Pa m−1 ) Liquid hold-up (m3 m−3 )

Experimental Billet Mackowiak Experimental Billet Mackowiak Mackowiak (modified)

Water 77 86 322 0.081 0.073 0.049 0.080

DEHA 107 93 1257 0.197 0.186 0.080 0.131
PEG 400 150 105 2595 0.259 0.290 0.092 0.151

new consideration in the case of viscous liquids. Billet’s corre- 5000 mg m−3 ). The experimental pilot used in this study could
lation gives accurate results, by considering Eqs. (4) and (10), therefore fulfil the requirements to treat a small flow indus-
it appears that Billet models the liquid hold-up with a power of trial vent gas. Since DEHA is a viscous liquid and that the
viscosity equal to 0.267 (Re < 5) while Mackowiak proposes Henry’s law constant is not too high, it can be assumed that
a power of viscosity equal to 0.16. A regression on the results the main transfer resistance is located in the liquid phase. In
of the table gives a power of viscosity equal to 0.269. The ex- fact, diffusion coefficients of toluene are 8.6 × 10−6 cm2 s−1 in
ponent proposed by Billet fits well with our data. DEHA (Heymes, 2006) and 8.7 × 10−2 cm2 s−1 in air. Values
of KL a (Table 6) are low in comparison with literature data
4.2.1.5. Conclusions about hydrodynamics. Experiments per- about classical mass transfer constants (up to 0.1 m s−1 ). This
formed in a pilot scale setup revealed that using DEHA does not reflects the negative effect of viscosity on absorption kinetics,
create excessive pressure drops. When comparing our results in particular on the diffusion coefficient and the liquid bound-
with experiments performed with water in the same column, ary layer thickness. Low values of KL a involve a high height
it appears that the viscosity of DEHA creates a supplementary of transfer units (HTU) increasing the necessary total height
pressure drop lower than 30% of the water pressure drop. This of an industrial column.
pressure drop is acceptable for industrial purposes. Concerning
modelling, the most significant recent works were studied and 4.2.2.2. kL a considerations. Since the mass transfer resistance
compared with our experimental results. The authors’ excellent is mainly located in the liquid boundary layer, local liquid mass
predictions for dry bed pressure drops (3–5%) are worse when transfer constant kL a are equal to overall liquid mass transfer
the column is wetted (20–25%): this comes from the increased constant KL a. A difference in order of a factor of 10 between
complexity of the triphasic gas–liquid–solid system. Both the the different values of kL a can be observed. To explain this ob-
old approach of Eckert, updated by Miyahara and the new ap- servation, it is useful to link mass transfer to the hydrodynam-
proach of Mackowiak modelled the flooding points fairly well. ics of the column. Fig. 8 is a development of the liquid hold-up
Thus, the hydrodynamic point where this chaotic state occurs curves presented in the hydrodynamics section. The zones rep-
can be accurately predicted. However, if under and in the load- resenting the loading regime and flooding point of the column
ing zone the hydrodynamic parameters (
P /Z and hL ) are were added in grey. The hatched area corresponds to the range
sufficiently predicted, when approaching the flooding point the of operating conditions used for the mass transfer experiments.
error increases significantly. Overall results and observations Experiments 1–7 have been added to the figure to show at a
about the influence of viscosity lead us to consider that Bil- glance the differences in calculated kL a values depending on
let’s correlations should be recommended for scaling industrial the hydrodynamic regime in the column. In fact, points 1–4,
equipment. Results are accurate and the formulae are simple- which are not located in the loading regime, have the lowest
to-use. The mass transfer in the column will now be considered. kL a values (around 10−4 ), while the points located in the load-
ing zone have kL a values around 10−3 . It is well-known that
4.2.2. Mass transfer the loading zone corresponds to efficient hydrodynamic condi-
4.2.2.1. Air treatment performance considerations. Experi- tions which enhance mass transfer kinetics.
mental results show that the toluene-charged vent gas can Experiments 1–4 were led with the same gas flow rate below
be cleaned efficiently (Fig. 5). The legal limit of toluene re- the loading zone. Results show no significant influence of liquid
lease into the atmosphere is 110 mg m−3 (EEC directive). flow rates on liquid mass transfer constants. The liquid hold-up
This corresponds to removal performances of 78%, 89% and rises from 0.20 to 0.30 m3 m−3 and the kL a from 2.5 × 10−4
98% according to the experimental concentrations (500, 1000, to 3.7 × 10−4 s−1 . This confirms that interactions between the
 0.6 enabled modern packing systems to provide much more exten-
Flooding zone sive interfacial areas due to the numerous droplets and trick-
0.5 les. Billet’s correlation, established using more recent packing
Loading zone systems, is thus more appropriate for the packing used in this
Liquid holdup [m2.m-3]

0.4
study.
Table 10 compares the predictions of Onda and Billet with
the experimental results. Onda’s simple-to-use correlations give
0.3
1 2
values very far from the experimental results, which is to be
3
4
expected given the above explanation. Billet’s more recent cor-
0.2 relations give predictions that are closer to reality but still too
5 6 7
inaccurate and therefore not valid for industrial utilisation. This
0.1 inaccuracy can be explained by the fact that the viscosity of
Mass transfer experimental range 4 Experience number
DEHA does not fall within the range of values used by Billet
0.0 to adjust his correlations. Table 11 summarises the ranges of
0.0 0.5 1.0 1.5 2.0 2.5 3.0 experimental operating parameters used by Billet and by this
Superficial gas velocity [m.s-1] study: it shows that our experimental data are included within
the values investigated by Billet et al., except that the kinetic
Fig. 8. Mass transfer experimental conditions and hydrodynamic zones.
viscosity of DEHA is much higher than that of the liquids used
by Billet. Therefore, this explains the predictions of the Billet
Table 9
correlation, which overestimate the liquid-side mass transfer
Comparison of this work results and literature data coefficients by a factor of 10–50.
The two correlations from the literature are not satisfactory
Absorbent Packing kL a (m s−1 ) (×104 ) Reference for modelling our system. Analysis of experimental data shows
PEG 400 1.4–5.5 Cotte (1996) that the kL a varies versus the gas velocity and rather little in
DEHA Hiflow rings 2.5–24 This work function of liquid velocity. This observation is not much de-
Water 78–83 Cotte (1996) scribed in the literature. In fact, most authors agree in describ-
ing mass transfer as a value dependent on the liquid velocity
and independent of the gas velocity. This is the case of the Bil-
liquid and gas phases are small. In the loading zone, changing let and Onda relationships used in this work. Nonetheless, this
the gas velocity from 0.42 to 1.50 increases the kL a value from unexpected behaviour of our system has also been observed by
0.3×10−3 to 2.4×10−3 s−1 . The gas interaction with the liquid two recent authors. Cotte (1996) used PEG 400 ( = 80 mPa s)
can also be observed by considering the liquid hold-up. to absorb various VOCs (acetone, toluene, dichloromethane)
It is interesting to compare the kL a values obtained with data and observed that the kL a values deduced from his experiments
from other literature (Table 9). With the same packing (Hiflow obeyed the law kL a =7.5×10−4 UG +5.4×10−5 . Cotte did not
rings), it appears that DEHA occupies an intermediate position vary the liquid flow rate in his experiments and therefore did
between PEG 400 and water. DEHA is thus more effective than not take that parameter into account. Delaloye (1991) studied
PEG 400 for absorbing toluene. aqueous solutions with algynate, glycerol or PEG 3000 added
to produce solutions whose viscosity reached 9.63 mPa s. He
4.2.2.3. Modelling of experimental results. Onda and Billet noted two interesting phenomena. The first result he high-
proposed correlations to model the wetted area aw . The predic- lighted is that the gas velocity has an influence on the kL a in
tions of these authors differ considerably. This big difference the loading zone, contrary to previously published studies. The
results from developments in packing systems between 1968 second observation is that an increase of the viscosity leads to a
and 1993. The higher void fraction and improved shapes have reduction of the liquid velocity’s influence. He showed that the

Table 10
Comparison of Billet and Onda predictions versus experimental data

Experiment Toluene in gas feed Liquid velocity Gas velocity Experimental Billet Onda
number (g m−3 ) (m s−1 ) kL a (s−1 ) (m s−1 ) kL a (s−1 ) kL a (s−1 )

1 4.990 0.014 0.422 2.5 × 10−4 5.0 × 10−2 6.4 × 10−1

2 1.205 0.014 0.422 2.5 × 10−4 5.0 × 10−2 6.4 × 10−1
3 1.193 0.011 0.422 3.7 × 10−4 4.1 × 10−2 5.4 × 10−1
4 5.038 0.007 0.422 3.2 × 10−4 3.0 × 10−2 4.2 × 10−1
5 0.520 0.007 0.964 2.3 × 10−3 3.0 × 10−2 4.2 × 10−1
6 1.005 0.007 1.005 2.4 × 10−3 3.0 × 10−2 4.2 × 10−1
7 1.005 0.007 1.502 2.4 × 10−3 3.0 × 10−2 4.2 × 10−1
Table 11
Parameters ranges of Billet’s work and this work

Variable Range investigated by Billet (1990) and Range investigated in this work
Billet and Schultes (1991, 1992)

Gas capacity factor FG (Pa1/2 ) 0.003–2.77 0.510–2.18

Superficial liquid velocity UL (m s−1 ) (×103 ) 0.071–32.77 4.24–14.1
Liquid density L (kg m−3 ) 361–1237 930
Liquid kinetic viscosity L (m2 s−1 ) (×106 ) 0.14–1.66 15.5
Diffusion coefficienta DL (m2 s−1 ) (×10−9 ) 0.29–6.50 0.86
Superficial tension L (N m−1 ) (×103 ) 0.7–74.0 31
Gas density G (kg m−3 ) 0.07–97 1.18
Gas diffusion coefficienta DG (m2 s−1 ) (×106 ) 0.29–87.4 7.6
Gas dynamic viscosity G (Pa s) (×106 ) 0.14–126 15.4
a Diffusion coefficient of toluene into the considered liquid.

exponent of the UL term decreases as the viscosity increases. Our experimental results showed that the kL a of the system
All these observations show the lack of knowledge regarding the depends on the liquid velocity but also on the gas velocity. This
modelling of mass transfer in viscous systems. Complementary behaviour has also been observed by the few authors who have
work is needed to be carried out in order to provide a correlation used viscous fluids in their experiments, but is contrary to all
that enables accurate prediction of the results. the authors who have work on low-viscosity fluids: generally,
these authors express kL a using a term that increases with liq-
uid velocity and does not take into account the gas velocity.
5. Conclusion It is therefore clear that the influence of viscosity on the phe-
nomenon is considerable and that further research is necessary
This study examined hydrodynamics and mass transfer of a in order to better describe this area of study.
packed column fed with DEHA to remove the toluene from a This work contributes to the development of hybrid
medium-concentration gaseous effluent (0.5–5 g m−3 ). absorption–pervaporation processes for treating gases con-
The hydrodynamic study showed that the viscosity of DEHA taining hydrophobic compounds. The aim of this study was
was not a technical obstacle to its implementation in an indus- to evaluate the potential of using a heavy solvent (DEHA) to
trial column. The pressure drops caused remain within accept- absorb such compounds in a packed column and, based on a
able levels. From a modelling point of view, the bibliographical bibliographical summary, to propose empirical correlations for
summary dealt with numerous works and described the more both determining the hydrodynamic behaviour and estimating
recent of them. The correlations given proved satisfactory for mass transfer coefficients with a view to scaling up the sys-
the prediction of hydrodynamic values (dry column pressure tem for industrial applications. It is important to note that the
drops, liquid hold-up, wet column pressure drops). We recom- economic cost of the solvent is not covered in this work but
mend using the correlations proposed by Billet because they was taken into consideration for the laboratory-scale tests. The
are accurate and quite simple to use. recovery of hydrophobic VOCs and the regeneration of the
The study of the absorption of toluene by DEHA showed the solvent by pervaporation will be presented in a future paper.
efficiency of this process (initial toluene recovery performance
close to 100% under our experimental conditions). Nonethe-
less, this efficiency decreases quickly when the washing liquid
becomes loaded with toluene. An effective regeneration process Notation
is therefore essential in order to make the process technically
viable. a specific surface area of packing, m2 m−3
Concerning mass transfer modelling, the mass transfer is sup- ah hydraulic area, m2 m−3
posed to be limited by the liquid-side resistance, which seems awet wetted area of packing, m2 m−3
logical since DEHA is a viscous absorbent. The bibliographi- CF l packing constant for flooding (Mackowiak,
cal review showed that few recent works have been devoted to 1990)
the modelling of gas–liquid mass transfer in packed columns. CG gas concentration, mol m−3
Billet and Schultes (1999) is the only author to have recently Ch packing constant for hold-up (Billet, 1990; Billet
tried elaborating a new theory for predicting mass transfer co- and Schultes, 1991,1992)
efficients. He based his model on a data bank created from a CL liquid concentration, mol m−3
great deal of data from the literature, which enabled him to CL∗ , CV∗ packing constant for hold-up (Billet and
propose a more accurate correlation than those given in previ- Schultes, 1993, 1999)
ous works. Nonetheless, few authors have investigated viscous dh hydraulic diameter, m
fluids and Billet’s correlations proved relatively inaccurate for dp packing size, m
predicting toluene transfer coefficients in the case of DEHA. dT droplet size, m
DG , DL gas and liquid diffusivities, m2 s−1 U 
ReL = L aL Reynolds number of liquid
FG , FL gas or liquid capacity factor FG = 
L
Sc =  D Schmidt number of liquid
G UG , kg1/2 m−1/2 s−1
1/2 L
dh
F (S) friction factor Sh = Sherwood number
g gravitational acceleration, m s−2 L UL2
WeL = Weber number of liquid
hL liquid hold-up, m3 m−3 L a
hL,F flooding liquid hold-up, m3 m−3
hK , hB Kozeny and Burke-Plummer constants Acknowledgements
H Henry constant, dimensionless
k hydrodynamic parameter (Takahashi) This work was carried out with the financial support of
kG , kL partial mass transfer coefficients, m s−1 the ADEME (Agence de l’Environnement et de la Maitrise
Kp wall factor de l’Energie, France) and the Europe Environnement company
KG , KL global mass transfer coefficients, m s−1 (Vieux Thann, France).
L, G specific liquid or gas flow rates, kg m−2 s−1
QG , QL gas or liquid flow rate, m3 h−1 References
S column section, m2
t time, h Billet, R., 1990. Contribution to design and scale up of packed columns. Fat
U superficial liquid velocity, m s−1 Science and Technology 9, 361–370.
Billet, R., Schultes, M., 1991. Modelling of pressure drop in packed columns.
VC column volume, m3 Chemical Engineering and Technology 14 (2), 89–95.
x, y liquid or gas mass fraction Billet, R., Schultes, M., 1992. Relation between pressure drop and mass-
Z packing height, m transfer in two-phase countercurrent-flow columns. Fat Science and
Technology 4 (94), 138–144.
Greek letters Billet, R., Schultes, M., 1993. Predicting mass transfer in packed columns.
Chemical Engineering and Technology 16, 1–9.
 experimental constant (Miyahara Billet, R., Schultes, M., 1995. Influence of phase ratio on packing efficiency
in columns for mass transfer processes. Chinese Journal of Chemical of
et al., 1992)
Engineering 5 (2), 117–126.
 liquid hold-up contribution factor (Miyahara Billet, R., Schultes, M., 1999. Prediction of mass transfer columns with
et al., 1992) dumped and arranged packings. Transactions of the Institution of Chemical

P0 dry pressure drop, Pa Engineers 77, 498–504.

Pwet wet pressure drop, Pa Cotte, F., 1996. Absorption en colonne garnie et en tour à atomisation.

P /Z total pressure drop, Pa m−1 Application au traitement de composés organiques volatils. Thesis,
Université de Pau et des Pays de l’Adour, Alès.
 fractional void space of packed bed Delaloye, M., 1991. The influence of viscosity on the liquid-phase mass
 dynamic viscosity, Pa s transfer resistance in packed columns. Chemical Engineering Journal 47,
M packing shape factor (Mackowiak, 1990, 1991) 51–61.
 density, kg m−3 Heymes, F., Manno-Demoustier, P., Charbit, F., Fanlo, J.L., Moulin, P., 2006.
A new efficient absorption liquid to treat exhaust air loaded with toluene.
c packing surface tension, N m−1 Chemical Engineering Journal 115 (3), 225–231.
L liquid surface tension, N m−1 Iliuta, I., Grandjean, B.P.A., Piché, S., Larachi, F., 2003. Two fluid
0 resistance coefficient of dry packing model for counter-current dumped packing-containing columns. Chemical
G,L resistance coefficient of trickle packing Engineering Science 58, 1373–1380.
Leva, M., 1992. Reconsider packed-tower pressure-drop correlations.
Subscripts Chemical Engineering Progress 65–72.
Mackowiak, J., 1990. Determination of flooding gas velocity and liquid
F flooding hold-up at flooding in packed columns for gas/liquid systems. Chemical
Engineering and Technology 13, 184–196.
G gas Mackowiak, J., 1991. Pressure drop in irrigated packed columns. Chemical
in, out inlet, outlet Engineering Progress 29, 93–105.
L liquid Miyahara, T., Ogawa, K., Hirade, A., Takahashi, T., 1992. Flow dynamics in
low height packed columns having large fractional void space. Chemical
Dimensionless numbers Engineering Science 47 (13/14), 3323–3330.
Onda, U., Takeuchi, H., Okumoto, Y., 1968. Mass transfer coefficients between
 u
Ca = LL L Capillary number of liquid gas and liquid phases in packed columns. Journal of Chemical Engineering
of Japan 1 (56),
u2G Piché, S., Iliuta, I., Grandjean, B.P.A., Larachi, F., 2001. A unified approach
Fr = gd 0 Froude number of gas
to hydraulics and mass transfer in randomly packed towers. Chemical
g 4L Engineering Science 56, 6003–6013.
M= L 3L
Morton number of liquid Takahashi, T., Akagi, Y., Ueyama, K., 1979. A new correlation for pressure
U  drop in packed columns. Journal of Chemical Engineering of Japan 12
ReG = G aG Reynolds number of gas (5), 341–346.
G