Wat. Res. Vol. 34, No. 13, pp.

3460±3472, 2000
7 2000 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
PII: S0043-1354(00)00080-4 0043-1354/00/$ - see front matter

www.elsevier.com/locate/watres

CFD MODELLING OF A TWO-PHASE JET AERATOR

UNDER INFLUENCE OF A CROSSFLOW
J. MORCHAIN, C. MARANGES and C. FONADE*
DeÂpartement de GeÂnie Biochimique et Alimentaire, Centre de BioingeÂnierie Gilbert Durand,
Laboratoire Biotechnologie-BioproceÂdeÂs, UMR CNRS 5504, Institut National des Sciences AppliqueÂes,
135 Avenue de Rangueil, 31077, Toulouse, cedex 4, France

(First received 1 November 1998; accepted in revised form 1 December 1999)

AbstractÐThe coupling e€ect between the oxygen mass transfer, due to a two-phase jet aerator and the
recirculating ¯ow it generates when installed in a large vessel, is modelled using a Computational Fluid
Dynamic (CFD) approach. A semi-in®nite domain surrounding the jet aerator was built. Speci®c
boundary conditions were applied in order to represent a uniform, oxygen free, transverse ¯ow ¯ushing
the aerated zone. The motion of the gas phase was not taken into consideration and its in¯uence on
the motion of the liquid phase was neglected. The volume in which oxygen transfer occurs was
restricted to a subpart of the domain and it was characterized by a uniform mass transfer coecient
KLa. The ¯ow rate due to the aerator is 50 m3 hÿ1, that of the transverse current was varied from 700
to 3000 m3 hÿ1 and three values of KLa were investigated (360, 1000 and 3600 hÿ1). For each case, the
Navier±Stokes equations were solved to compute the velocity ®elds. Numerical results proved to be in
close agreement with published data of 3-D jets in a cross¯ow. The oxygen source term was then added
and the transport equation was time integrated until a steady state was reached. It was demonstrated
that the most in¯uential parameter on the amount of oxygen transferred was the transverse ¯ow rate.
When low, modifying the KLa value has almost no e€ect, whereas at high ¯ow rates a tenfold increase
in the KLa value only doubles the oxygen transfer value. An ideal reactor-based model was then used
to analyse the results. The behaviour of the aerated zone was found to be close to that of a plug ¯ow
reactor. Numerical values of the net mass ¯ux of oxygen are compatible with industrial reports (13
kgO2 hÿ1) and our modelling was able to account for the di€erences between laboratory and full-scale
experiments. 7 2000 Elsevier Science Ltd. All rights reserved

Key wordsÐaeration, lagoon, water treatment, oxygen transfer, CFD modelling, jet ¯ow

NOMENCLATURE

U0 velocity at the ejector exit, inlet velocity (m sÿ1)

Uf mean velocity of the cross-¯ow current (m sÿ1)
U, V, W=X, Y, Z components of the velocity vector (m sÿ1)
d diameter of the ejector nozzle (m)
D oxygen di€usivity (m2 sÿ1)
U non-dimensional velocity in the X-direction (U=U/U0)
V non-dimensional velocity in the Y-direction (V=V/U0)
W non-dimensional velocity in the Z-direction (W=W/U0)
k turbulent kinetic energy (m2 sÿ2)
k non-dimensional kinetic energy k ˆ k=U 02 )
p pressure (Pa)
q oxygen transfer ¯ux density (kgO2 mÿ3 sÿ1)
QA ¯ow rate through face A (mÿ3 sÿ1)
QB ¯ow rate through face B (mÿ3 sÿ1)
QJ ¯ow rate of the jet (mÿ3 sÿ1)
QT total ¯ow rate through the aerated zone (mÿ3 sÿ1)
F JET mass ¯ux of oxygen due to the jet (kgO2 sÿ1)
F AER mass ¯ux of oxygen due to aeration (kgO2 sÿ1)
F TOT total mass ¯ux of oxygen exiting the computational domain (kgO2 sÿ1)
QG gas ¯ow rate at the exit of the ejector (m3 sÿ1)
QL liquid ¯ow rate at the exit of the ejector (m3 sÿ1)
V volume of the aerated zone (m3)

*Author to whom all correspondence should be addressed. Tel.: +33-5-6155-9435; fax: +33-5-6155-9400; e-mail:
[email protected]

3460
 Two-phase jet aerator modelling 3461

C mass fraction of oxygen at saturation point

CA mean mass fraction of oxygen over face A
CB mean mass fraction of oxygen over face B
CJ inlet mass fraction of oxygen
CM mean mass fraction of oxygen over the aerated volume

Greek symbols
a velocity ratio U0/Uf
b jet spread angle
E turbulent dissipation rate (m2 sÿ3)
E' volume fraction
E non-dimensional dissipation E ˆ E d=U 30 †
r density (kg mÿ3)
m viscosity (Pa s)

INTRODUCTION
reports reveal that the in situ oxygenation capacity
of this hydro-ejector is up to three times higher
When looked at a bioreactor for wastewater treat- than that measured in a 100 m3 tank, which shows
ment, a lagoon can be de®ned as a large tank the importance of the ¯ow characteristics in the
designed to ensure a long residence time so that neighbourhood of the ejector.
microorganisms can achieve the degradation of In order to provide the required oxygen, numer-
complex organic pollutants. The ratio of the hori- ous hydro-ejectors are needed, which from a hydro-
zontal length to the depth of a lagoon varies dynamic point of view, results in a ¯ow generated
between 50 to 100 and the volume commonly in the large lagoon by several small jets issuing
reaches thousands of cubic meters. In an aerated from these units.
lagoon, a lot of oxygen is arti®cially supplied in So, as a ®rst step in a complete modelling of the
order to enhance the aerobic reactions. In such a hydrodynamic and biological operation of an aera-
system, the eciency of the degradation process ted lagoon, the present paper deals with the model-
depends both on the hydrodynamic behaviour of ling and computation of the momentum and
the pond and on the kinetics of biological reactions. oxygen transfer equations in the region close to one
Numerous studies have contributed to the determi- hydro-ejector subject to the in¯uence of the overall
nation of the parameter most in¯uential on the per- circulation in the lagoon. The following successive
formance of an aerated lagoon (Vassel et al., 1987; steps will be presented:
Slade, 1991; Maes, 1992; Dorego and Leduc, 1996;
1. computation of the 3-D velocity ®eld created by
Torres et al., 1997), but the operational design still
the ejector;
remains empirical and is based on certain overall
2. introduction of the recirculating ¯ow occurring
experimental correlations. In fact, because the oxy-
in a lagoon and acting transversally on the jets;
gen can only be provided in small parts of the
3. modelling of the oxygen transfer.
lagoon, the biodegradation process is heavily depen-
dent on the hydrodynamic structure of the ¯ow
which generally comes from the aeration devices.
In the past decades, a large variety of technical METHOD
solutions aimed at providing the required amount
of oxygen has been investigated. Among these, two- The ¯ow computations employed the 7.6 version of the
CFD code FIDAP on an HP 9000 workstation. This code
phase jets are the subject of an ever growing interest
due to their good transfer performance and because
their maintenance is easy and running costs are
very low (Jackson, 1964; Tojo and Miyanami, 1982;
Ogawa et al., 1983; Fonade et al., 1994). In the pre-
sent work we will refer to a particular type of ejec-
tor studied by Rainer et al. (1995) and Orfaniotis et
al. (1996). The scheme of the ¯ow generated by this
hydro-ejector is shown in Fig. 1: the transfer of
oxygen occurs in the ®rst zone, close to the ejector,
while a liquid jet ¯ows forward. In this way, the
same device provides simultaneously the aeration of
the broth and the required circulation in the whole
lagoon, a circulation which must prevent the
settling of the biomass and distribute the dissolved
oxygen into all parts of the lagoon. Industrial Fig. 1. Diagram of a hydro-ejector.
3462 J. Morchain et al.

Fig. 2. Geometry of the computational domain.

uses the Finite Element Method for the discretization of hydro-ejector. This geometry is shown in Fig. 2 and rep-
the Navier±Stokes equations. In our case, where turbulent resents only a small part of the lagoon in which the ejector
¯ows are considered, a decoupled solver was used to solve is installed. Dashed lines show the extension made in
the Reynolds Averaged Navier±Stokes equations along order to follow the jet curvature when the in¯uence of a
with the turbulent closure model equations. With this nu- recirculating ¯ow is introduced into the model.
merical method, computational work is reduced by com-
parison to fully coupled solvers but higher convergence
criteria have to be chosen. Typically, they were set to 10ÿ3 Hydrodynamic boundary conditions
for the ¯ow equations and to 10ÿ2 for the species equation The mean velocity at the ejector exit is set to 1.4 m sÿ1
when used. The maximum amount of memory required which corresponds to the actual liquid ¯ow rate passing
for solving the system was approximately 60 Mo. The through an ejector. The situation is therefore that of a jet
standard k±E model was used to account for turbulence ¯ow with a Reynolds number equal to 140,000. An identi-
e€ects. This model is robust, accurate and cost e€ective in cal set of boundary conditions is applied for all velocity
terms of computational time. It is implemented in most ®eld computations (see Table 1):
CFD codes and widely used in industrial situations.
. a slip wall condition is imposed on the free surface
(because the momentum is injected through submerged
Geometry of the ¯ow ®eld jets, the experience shows that the free surface is very
We will focus on the jet ¯ow region where momentum slightly disturbed. It is therefore assumed that the sur-
is supplied. An accurate knowledge of the ¯ow in this face remains planar);
region is needed because it is the source of the ¯ow . a standard law of wall is used close to solid boundaries;
throughout the whole lagoon. Depending on the length . a zero longitudinal gradient for all velocity components
scale of the hydro-ejector (1.5 m long and 0.1 m in diam- is applied on outlet boundaries to simulate an in®nite
eter), the computational domain is restricted to a hexahe- domain in the horizontal directions;
dron of dimensions 6  4  2 in meters (length  width  . inlet velocity pro®les (jets and transverse ¯ow) are
depth). This volume of ¯uid surrounds the submerged assumed to be uniform.

Table 1. Boundary type and particular conditions

Side wall number CFD simulation case

JF JCF OMT

1 Wall Wall Wall

2 Zero gradient Zero gradient Zero gradient
3 Wall Wall Wall
4 Zero gradient V = 1/a Zero gradient C = 10ÿ10
5 Slip wall Slip wall Slip wall
6 Zero gradient Zero gradient Zero gradient
 Two-phase jet aerator modelling 3463

Table 2. Inlet dimensions, velocities and turbulence characteristics for all cases

Parameter CFD simulation case

2-D 3-D JF JCF OMT

d (m) 0.04 0.04 0.1 0.1 0.1

U (m sÿ1) 5 5 1.4 1.4 1.4
V (m sÿ1) 0 0 0 0 0
W (m sÿ1) ± 0 0 0 0
k (m2 sÿ2) 2.5  10ÿ2 2.5  10ÿ2 1.96  10ÿ3 1.96  10ÿ3 1.96  10ÿ3
E (m2 sÿ3) 0.93 0.93 8.23  10ÿ3 8.23  10ÿ3 8.23  10ÿ3
r (kg mÿ3) 1000 1000 1000 1000 1000
m (kg mÿ1 sÿ1) 10ÿ3 10ÿ3 10ÿ3 10ÿ3 10ÿ3
X (kgO2 kgÿ1) ± ± ± ± 6.3  10ÿ6
Re 2000 2000 1.4  105 1.4  105 1.4  105

Non-dimensional variables will be used, leading to the fol- studied, some additional hypotheses will be made and dis-
lowing set of inlet values: cussed later, enabling the prediction of the size of this
transfer.
U  ˆ 1, V ˆW ˆ0 k ˆ 10 ÿ3 E ˆ 3  10 ÿ4
Three-dimensional model justi®cation
The actual numerical values corresponding to the di€erent Because of the speci®c geometry of a lagoon, the
situations which will be looked at are listed in Table 2. ¯ow structure is often seen as two-dimensional (2-D),
i.e. independent of the vertical coordinate. This situ-
Single-phase ¯ow simulation
ation is typical of shallow water ¯ows in which vertical
Although an ejector generates a two phase ¯ow, it will disturbances are quickly damped out, whereas horizon-
be assumed that the momentum transfer is not signi®- tal movements of the ¯uid can keep on interacting
cantly a€ected by the gas phase and the velocity ®eld will with each other over longer distances. With this
be computed using a single-phase model. In fact, the gas assumption, only the momentum balance in the hori-
phase slightly increases the vertical velocity component of zontal plane (x, y ) is used and the shear stresses on
the liquid particles, but when circulation in the lagoon is the free surface (wind e€ects) and on the bottom are
looked at, the main e€ect of the jet comes from its hori- taken into account through an ``integrated model''.
zontal momentum. When oxygen transfer performance is With such a simpli®cation, the diculty in constructing

Fig. 3. Non-dimensional speed contour plot, 2-D case.

3464 J. Morchain et al.

Fig. 4. Non-dimensional speed contour plot, 3-D case.

a suitable mesh grid, the memory requirements and the balance is quite di€erent and the jet actually impinges the
computational time are greatly reduced. Unfortunately wall. This behaviour was also observed by Wood et al.
convergence can be quite hard to obtain and results (1995) in their investigation about the in¯uence of the pos-
are often disappointing (Wood et al., 1998; Ta and ition of a ba‚e on the ¯ow structure in a lagoon with a 2-
Brignal, 1998). D model.
In our case, even if the structure looks like a 2-D ¯ow
in a great part of the lagoon, it is not true in the regions
close to the ejectors: because the jet diameter is small com- Mesh grid
pared to the height of liquid, transverse ¯ows can occur The structure of the ¯ow created by a hydro-ejector is
above and below the jet. The pressure ®eld around the jet very close to that of a turbulent free jet (Fig. 5) and is
is quite di€erent to that obtained with plane jet con®gur- characterized by high transverse velocity gradients. In
ation. As an example of this di€erence, the ¯ow compu- order to optimize the mesh grid, the longitudinal node
tation was achieved in a typical situation of the spacing was progressively re®ned until no further notice-
impingement of a jet on a wall. Figures 3 and 4 show iso- able improvement in the solution was obtained. The
velocity, respectively, when 2- and 3-D descriptions are observed characteristics were the existence of a potential
used. In the ®rst case, the jet acts as a barrier: the result- core and the jet centreline velocity decrease. Using this in-
ing pressure di€erence across the jet causes a great de¯ec- formation, a 41216 node mesh grid was built in order to
tion of the jet. With a 3-D description, the pressure compute the velocity ®eld.

Fig. 5. Description of a turbulent jet ¯ow structure.

 Two-phase jet aerator modelling 3465

equation of the regression line on Fig. 6 is

Y=sX+b, where s=ÿ1.004 and b = 0.8583 with a
standard deviation R 2=0.9993. The hyperbolic
decay of the centreline velocity is accurately calcu-
lated: the position of the virtual origin of the jet is
found to be Xp = 2.8 (MalmstroÈm et al., 1997 give
Xp $ [2.5; 3.2]), while the constant K takes the value
7.2, slightly higher than the usually obtained values.
In the same way, the analysis of transverse velocity
pro®le leads to the value 0.089 for tan b (Mal-
mstroÈm et al., 1997 give tan b $ [0.086; 0.102]). The
Fig. 6. Velocity pro®le along the jet axis. similarity of transverse velocity pro®les is well rep-
resented but the spreading rate of the jet is slightly
RESULTS overestimated by the simulation. Various authors
have reported that the isotropic k±E turbulence
Jet ¯ow simulation (JF) model was unable to accurately predict the spread-
This simulation was applied to the ¯ow domain ing rate of a 3-D free jet, mainly because turbulence
shown in Fig. 2 and boundary conditions are listed is generally non-isotropic but also because the con-
in Table 1. The zero velocity gradient is used to stants of the model are derived from boundary
represent a semi-in®nite physical domain surround- layer ¯ow experiments (Pope, 1978; Thies and Tam,
ing the computational domain. Considering the 1996).
depth of our domain (2 m) and the diameter of the
ejector (d = 0.1 m), we can expect the in¯uence of Three-dimensional jet in a cross¯ow (JFC)
the bottom and of the free surface to be negligible. In many practical cases, the ejector area is ¯ushed
Thus, the ¯ow pattern is expected to be very close by a recirculating current induced by geometry
to that of a free wall jet which will be characterized restrictions. In terms of ¯ow description, such a
through its centreline velocity decay and its trans- situation is very close to the problem of a free jet in
verse velocity pro®le. These characteristics are a cross¯ow, a situation which has been widely
usually written as follows (MalmstroÈm et al., 1997): investigated in the environmental domain (chimney
plume, discharge of an e‚uent into a river). The
U=U0 ˆ K d= X ÿ Xp† main parameter is the ratio a between the jet vel-
1† ocity U0 and that of the free stream Uf (a=U0/Uf ).
centreline velocity profile
Owing to the in¯uence of the transverse ¯ow, the
jet is de¯ected and the characteristic round shape of
Um=Umo ˆ exp ÿaZ 2 † with a ˆ ln 2 the jet is deformed into a kidney shape because of
2† turbulence mixing in the outer layers of the jet.
transverse velocity profile To compute the velocity ®eld in such a situation,
a rectangular zone (meshed with hexahedrons) was
where K is an empirical constant, U0 is the initial added on the right side to follow the jet curvature
velocity of the jet, d its diameter, Xp the abscissa of (Fig. 2). In order to represent the transverse free
the virtual origin of the jet (fully developed region), stream ¯owing through the side wall 4, a uniform
Um is the average longitudinal velocity (X-direc- velocity V=1/a is applied on that side of the com-
tion), Umo its centreline value, Z is the transverse putational domain. All other boundary conditions
non-dimensional ordinate Z ˆ Y=Y0:5 †, and Y0.5 is remain unchanged (see Table 1).
the ordinate where Um/Umo = 0.5. Five values of a were investigated: a=5, 10, 15,
The results are drawn in Figs 6 and 7. The 20, 30. They correspond to typical values encoun-
tered in lagoons. The curvature of the jet centreline
is presented for all values in Fig. 8. An equation of
the following form was used to compile the whole
data set:
 b
X Y
ˆ aa a ˆ 0:91 and b ˆ 0:39
d d

The values of the two parameters a and b are in

close agreement with reported values of various ex-
perimental studies (Rajaratnam, 1976, ``Turbulent
jets'', Developments in Water Science No. 5, Else-
vier, Amsterdam, pp. 184±192).
Fig. 7. Similarity of transverse velocity pro®les. Figure 9 shows the iso-velocity lines in the hori-
3466 J. Morchain et al.

Fig. 8. Curvature of the centreline of jets in a cross-¯ow.

zontal plane containing the jet axis and allows the (Rajaratnam, 1976). Since the transverse current
jet de¯ection to be evaluated visually. Figures 10 can ¯ow over and under the jet, the de¯ection is
and 11, respectively, show total pressure and vel- much less than when a 2-D description is used.
ocity contour plots in a transverse plane (normal to Moreover, an intense mixing occurs behind the
the jet axis) at X = 15 d for a=10. We can observe inner boundary of the jet due to the presence of a
the characteristic kidney shape of the jet, the global pair of vortices.
depression on the inner boundary which explains
the jet curvature, and two local depressions due to Oxygen transfer modelling (OTM)
counter-rotating vortices. The velocity contour plot In the two-phase jet issuing from the hydro-
illustrates the importance of a 3-D description: the ejector, the gas volume fraction de®ned as
asymmetry of the velocity pro®le along a vertical E '=QG/(QG+GL) reaches values up to 0.75. Such a
line is obvious and this particularly is likely to exert situation cannot be directly taken into account
an in¯uence far away from the jet ¯ow region because CFD codes generally apply when the

Fig. 9. Speed contour plot in horizontal plane Z = 0, a=15.

 Two-phase jet aerator modelling 3467

volume fraction of the dispersed phase is less than lowing relationship:

15%. In order to overcome this diculty, the fol-
lowing method will be used to compute the scalar
q ˆ rKL a C  ÿ C † 3†
®eld of the dissolved oxygen.
In their experimental study, Rainer et al. (1995)
observed that ``The structure of the submerged gas- where q is analogous to a ¯ux density (kgO2 mÿ3
liquid jet [created by a hydro-ejector] was shown to hÿ1), r is the density of the ¯uid, C is the mass
be emulsion-like with small bubbles following the fraction of oxygen in saturated water at 208C
jet liquid path''. For turbulent round bubbling jets, (C=9.0  10ÿ6 kgO2 kgÿ1 for calculations) and
Iguchi et al. (1997) con®rm that the mean velocity KLa is the overall volumetric transfer coecient
of the liquid phase is only slightly dependent on the (hÿ1). The velocity ®eld (Fig. 8) shows that the
gas fraction of the jet (E' ranging from 0 to 0.5), spreading of the jet is not signi®cantly a€ected by
whereas turbulence characteristics are signi®cantly its curvature: the gas bubbles will be dispersed in
modi®ed because the energy introduced by the gas practically the same volume and it can therefore be
phase is consumed to generate turbulence. expected that the KLa does not depend on the
So we will assume that the presence of bubbles transverse ¯owrate.
does not a€ect signi®cantly the velocities of the The shape and volume of the aerated zone were
liquid phase and that the previously computed vel- taken from experimental observation of a hydro-
ocity ®elds are representative of the motion of the ejector operating in a 100 m3 tank. For the sake of
liquid in the two-phase ¯ow. convenience, this zone is considered to be a collec-
With this assumption, the supply of oxygen is tion of two rectangular blocks located along the jet
simulated via a uniform source term in the subpart axis, its actual volume being 2.91 m3.
of the domain where gas bubbles are found (aerated In this study, the ¯ow is steady so that the vel-
zone of Fig. 12). The eventual action of the local ocity ®elds are those which were obtained pre-
turbulent intensities on the local transfer character- viously (JCF with a=5, 10, 15 and 20). They are
istics is not taken into account in order to only used to solve the transport equation in which the
retain the in¯uence of the transverse ¯ow. oxygen mass fraction C is the only time-dependent
The oxygen source is then described by the fol- variable:

Fig. 10. Speed contour plot in a vertical plane perpendicular to the curved jet axis.
3468 J. Morchain et al.

Fig. 11. Pressure contour plot in a vertical plane perpendicular to the curved jet axis.

 
@C @C @ 2C At time t = 0, the oxygen mass fraction is taken
r ‡ Ui ˆ rD 2 ‡ q 4†
@t @xj @xi to be 10ÿ10 everywhere in the domain (a non zero
value is chosen in order to avoid stability problems
where D is the oxygen di€usivity (D = 2.09  10ÿ9 when starting the integration). For t > 0 the mass
m2 sÿ1) and q the oxygen source term de®ned in fraction of oxygen at the exit of the hydro-ejector is
equation (3). set to 70% of the saturation value to simulate the

Fig. 12. Description of the aerated zone.

 Two-phase jet aerator modelling 3469

Table 3a. Numerical results of oxygen mass transfer simulations KLa = 360 hÿ1

Velocity ratio Mass ¯ow rates (m3 hÿ1) Oxygen mass ¯ow (kgO2 hÿ1) CM  10ÿ6 Transfer eciency (%)

a B facea A facea A face Outingb

20 728.28 91.221 0.313 5.78 3.99 55.6

15 958.6 192.73 0.285 6.25 3.42 62.0
10 1443.45 98.78 0.315 7.06 2.57 71.4
5 2940.33 149.68 0.498 8.32 1.40 84.5

a
A face, B face: see Fig. 12.
b
Outing refers to the total mass ¯ux of oxygen outing the computational domain.

fact that oxygen transfer has already taken place in ted zone is de®ned by QT=QA+QB. These results
this device (Rainer et al., 1995). This results in an show that the amount of transferred oxygen
oxygen mass ¯ux entering the domain increases with the transverse ¯ow rate ¯ushing the
F JET=rQJCJ. The cross¯ow current is oxygen free aerated zone. The higher the value of KLa, the
and the mass fraction of oxygen is taken to be more marked is this e€ect. However, it is seen that
10ÿ10. Equation (4) was time integrated using a a tenfold increase in the KLa value only roughly
backward Euler implicit method until the mass of doubles the transferred oxygen value. In fact, as
oxygen passing through the side walls 2 and 6 equation (3) shows, the transfer of oxygen depends
reaches a constant value (Fig. 12). The total mass on the KLa and driving force (CÿC ) values, and
¯ux of oxygen exiting the domain, FTOT, was com- these two parameters act in opposite ways.
puted by summing the values on all vertical outlet Figure 14, which displays the average concentration
boundaries. The ¯ow rate was also computed on of dissolved oxygen in the aerated zone, shows that
each side of the aerated volume in order to quantify this driving force always increases with the trans-
the exact ¯ow rate passing through this zone. It was verse ¯ow rate, but that it is much higher when the
found that the liquid enters the aerated zone only KLa is low. As a consequence, better oxygen trans-
through faces A and B (Fig. 12). The ¯ow rate fer eciency, de®ned by the ratio (CÿCM)/C, is
through face A, (QA), is essentially due to the jet obtained in systems having both low KLa and high
entrainment. The oxygen mass fraction of this cur- transverse ¯ow rate values (see Table 3a,b,c).
rent, CA, results from the dilution of the inlet mass
fraction. The ¯ow rate through face B, (QB), is Comparison with ideal reactor models
linked to the cross¯ow current; the associated oxy- The objective of this section is to check the val-
gen mass fraction, CB, is therefore zero. The mean idity of the oxygen transfer model implemented in
oxygen mass fraction in the aerated zone was also the CFD simulations. Even if the velocity ®eld in
computed, which enables the intensity of the oxygen the jet-transverse ¯ow con®guration is very com-
source term via equation (3) to be calculated. plex, it is interesting to compare the computed
As a result of the previous investigations car- results with the oxygen transfer obtained with ideal
ried out in the 100 m3 tank and in order to simu- behaviour-closed reactors, i.e. the perfectly-mixed
late extreme behaviours of the transfer reactor (PMR) and the plug-¯ow reactor (PFR).
phenomenon, three values of the volumetric transfer Indeed, even if the aerated domain shown in Fig. 12
coecient in the aerated zone were used: 360, 1000 seems like an open reactor, the Peclet number is
and 3600 hÿ1. large enough (1106) to make the di€usive transport
The results are given in Tables 3a,b,c. Figure 13 negligible.
shows the variation of the total oxygen mass ¯ux Writing the mass balance equations for each of
leaving the domain, as a function of the mass ¯ow these reactors (steady operation), the output con-
rate passing through the aerated zone for the three centration CE can be calculated and then the mass
values of KLa. The total ¯ow rate through the aera- ¯ow rate F AER of oxygen leaving the reactor

Table 3b. Numerical results of oxygen mass transfer simulations KLa = 1000 hÿ1

Velocity ratio Mass ¯ow rates (m3 hÿ1) Oxygen mass ¯ow (kgO2 hÿ1) CM  10ÿ6 Transfer eciency (%)

a B facea A facea A face Outingb

20 728.28 91.221 0.229 8.01 6.48 28

15 958.6 192.73 0.199 9.38 5.99 33.4
10 1443.45 98.78 0.172 12.12 5.01 44.3
5 2940.33 149.68 0.318 17.41 3.18 64.6

a
A face, B face: see Fig. 12.
b
Outing refers to the total mass ¯ux of oxygen outing the computational domain.
3470 J. Morchain et al.

Table 3c. Numerical results of oxygen mass transfer simulations KLa = 3600 hÿ1

Velocity ratio Mass ¯ow rates (m3 hÿ1) Oxygen mass ¯ow (kgO2 hÿ1) CM  10ÿ6 Transfer eciency (%)

a B facea A facea A face Outingb

20 728.28 91.221 0.789 173 8.27 8

15 958.6 192.73 0.589 210 8.07 10.3
10 1443.45 98.78 0.496 300 7.66 14.8
5 2940.33 149.68 0.59 29.28 6.37 29.2

a
A face, B face: see Fig. 12.
b
Outing refers to the total mass ¯ux of oxygen outing the computational domain.
   
(F AER=rQTCE). Remembering that the cross¯ow QA KL aV
F AER ˆ rQT C  ÿ CA 1 ÿ exp ÿ
current is oxygen free (CB=0), we can assume that QT QT
ideal reactors are fed with the ¯ow rate QT in
In both cases the total mass ¯ux of oxygen is
which the mean oxygen concentration is C=QACA/
expressed as the sum of two contributions:
QT. Then the following successive relations are
obtained: F TOT ˆ F JET ‡ F AER
. for the PMR:
These results are drawn in Fig. 15. They show that
mass balance equation: the real ¯ow, numerically computed, is of a plug
¯ow type and that the real amount of transferred
  oxygen is limited by the value of the transverse ¯ow
QA
rQT CE ÿ CA ˆ rKL aV C  ÿ CE † rate. It is noticeable that when this ¯ow rate is low
QT
(say under 300 m3 hÿ1) the value of the KLa has
hardly any e€ect on the amount of oxygen trans-
QA CA ‡ KL aVC  ferred.
output concentration:CE ˆ
QT ‡ KL aV
DISCUSSION

QA CA ‡ KL aVC The above results were obtained when the aera-
oxygen mass flux:F AER ˆ r
KL aV tion of the tank is achieved through the use of
1‡
QT hydro-ejectors. In that case, the oxygen transfer
occurs in a relatively small volume and the perform-
. for the PFR: ance has been shown to be heavily dependent on
the interaction between the two-phase jet and the
mass balance equation: transverse ¯ow. These results are consistent with
the conclusion drawn from the investigation of the
QT ‰C x ‡ dx† ÿ C x†Š ˆ KL a C  ÿ C †S dx oxygen transfer capacity of the same hydro-ejector
when operating inside a 100 m3 tank and in a
lagoon. These conclusions can, however, also be
 
C  ÿ CE KL aV applied to any technological aeration system.
output concentration: ˆ exp ÿ
C  ÿ C0 QT For any aeration device, the maximum transfer
performance is given by the value of the product
(KLaC), which represents the potential of the
oxygen mass flux: device. The amount of oxygen which is actually

Fig. 13. Variation in oxygen mass transfer with the ¯ow Fig. 14. Variation in mean oxygen mass fraction in the
rate through the aerated zone. In¯uence of the KLa par- aerated zone with the ¯ow rate through this zone.
ameter. In¯uence of the KLa parameter.
 Two-phase jet aerator modelling 3471

very dicult and tedious when experimental set-ups

are to be used. In another way, successful predic-
tion of the biodegradation performances needs to
associate the ¯ow characteristics and the biological
kinetics. To overcome such diculties, modelling
and computational ¯uid dynamics can be used.
The results presented in this paper deal with the
®rst step in such a methodology and more especially
with the determination of the ¯ow characteristics
Fig. 15. Variation in oxygen mass transfer with the mass and oxygen transfer performances in the region
¯ow rate through the aerated zone. Comparison between located close to a hydro-ejector subject to the in¯u-
CFD simulation and ideal reactor models.
ence of a transverse stream.
Velocity ®elds created by a submerged round jet
transferred (oxygen transfer capacity) is obviously were ®rstly computed using the CFD routine
lower but the above results show that it increases FIDAP. The 3-D Navier±Stokes equations were
with the liquid ¯ow rate passing through the aera- solved using the k±E model to take into account the
ted zone. It can be deduced that it is worthwhile to turbulent e€ects and the optimization of the mesh
try to enhance the transfer capacity of an aeration size was realized on the basis of the free jet charac-
device, i.e. its KLa, only if it can be ensured that teristics. The velocity ®elds obtained when a trans-
the aerated zone will be ¯ushed by a high transverse verse ¯ow de¯ects the jet issuing from the hydro-
¯ow rate of oxygen free liquid. So, particularly ejector were then used to calculate the amount of
where lagooning is concerned, this means that the transferred oxygen. Because the gas fraction is large
overall structure of the ¯ow around the aeration and the bubbles are distributed in a well-de®ned
devices strongly a€ects the oxygen transfer capacity small volume, a single-phase ¯ow was considered
of the device and therefore the energetic yield of the and the supply of oxygen modelled through an
oxygen supply. A better eciency can be achieved overall volumetric transfer coecient KLa. The
by increasing the mass ¯ow rate of fresh water results show that the amount of transferred oxygen
through the aerated zone: the higher the KLa of the depends on the value of KLa but is also strongly
device, the higher the transverse ¯ow rate must be dependent on the transverse ¯ow rate of fresh water
(see Fig. 16). ¯ushing the aeration domain: for an industrial sized
hydro-ejector, the mass of oxygen dissolved in the
CONCLUSIONS liquid can be about three times higher if this ¯ow
rate increases from 500 m3 hÿ1 up to 3000 m3 hÿ1:
The aerobic treatment of wastewaters requires such conditions are respectively encountered when
that a lot of oxygen is supplied to the liquid med- the hydro-ejector operates in a 100 m3 tank and in
ium and it is generally achieved in large tanks or a 20,000 m3 lagoon. So, it is especially important in
lagoons. To supply this oxygen, hydro-ejectors can the case of very large basins to optimize the overall
be distributed along the sides of the basin: the ad- structure of the ¯ow in order to increase the ¯ow
vantages of such a solution are the very low main- rate crossing through the bubbling regions.
tenance requirements of these devices, low running
costs and a greater freedom in the location of the
REFERENCES
hydro-ejectors. In this way, the aeration and circu-
lation in the basin can be simultaneously achieved. Dorego N. C. and Leduc R. (1996) Characterization of
However, because of this ¯exibility, the optimiz- hydraulic ¯ow patterns in facultative aerated lagoons.
Water Science Technology 34(11), 99±106.
ation of the arrangement of the hydro-ejectors is
Fonade C., Doubrovine N. and Lalane M. (1994) Aera-
tion and mixing engineering of lagoon used in biological
waste water treatment. In IAWQ, 17th Biennial. Intern.
Conf., Budapest, Hungary.
Iguchi M., Okita K., Nakatani T. and Kasai N. (1997)
Structure of turbulent round bubbling jet generated by
premixed gas and liquid injection. International Journal
of Multiphase Flow 23(2), 249±262.
Jackson M. L. (1964) Aeration in Bernouilli types of
devices. American Industrial Chemical Engineering Jour-
nal 10(6), 836±842.
Maes M. (1992) Au pays des lagunes . . . des lacunes.
L'Eau, l'Industrie, les Nuisances 159, 37±40.
MalmstroÈm T., Kirkpatrick A., Christensen B. and
Knappmiller K. (1997) Centerline velocity decay
measurements in low-velocity axisymmetric jets. Journal
of Fluid Mechanics 246, 363±377.
Fig. 16. Variation in modelled oxygen mass transfer with Ogawa S., Yamagushi H., Tone S. and Otake T. (1983)
the KLa. In¯uence of the circulating ¯ow rate. Gas±liquid mass transfer in the jet reactor with liquid
3472 J. Morchain et al.

jet ejector. Journal of Chemical Engineering of Japan lent axisymmetric and nonaxisymmetric jet ¯ows using
16(5), 419±425. the k±E model. AIAA Journal 34(2), 309±317.
Orfaniotis A., Lalane M., Doubrovine N., Fonade C., Tojo K. and Miyanami K. (1982) Oxygen transfer in jet
Mayr B. and Moser A. (1996) Oxygen transfer and mixers. Chemical Engineering Journal 24, 89±97.
scale-up in bioreactors using hydro-ejectors for gas± Torres J. J., Soler A., Saez J. and Ortuno J. F. (1997)
liquid contacting. Bioprocess Engineering 14, 211±218. Hydraulic performance of a deep wastewater stabiliz-
Pope S. B. (1978) An explanation of the turbulent round ation pond. Water Research 31(4), 679±688.
jet/plane jet anomaly. AIAA Journal 16(3), 279±281. Vassel J.-L., Mierge J.-J., Vander Borght P., Schrolbiltgen
Rainer B., Fonade C. and Moser A. (1995) Hydrodyn- P. and Walbrecq M.-P. (1987) Etude hydraulique et
amics of a new type of ejector. Bioprocess Engineering transfert d'oxygeÁne en lagunage aeÂreÂ, F.U.L., pp. 101±
13, 97±103. 117.
Rajaratnam N. (1976) Turbulent jets. In Developments in Wood M. G., Green®eld P. F., Howes T., Johns M. R.
Water Science Vol. 5. Elsevier, Amsterdam. and Keller J. (1995) Computational Fluid Dynamic
Slade A. H. (1991) Hydraulic modelling of an aerated modelling of wastewater ponds to improve design.
lagoon system. Appita 44(1), 25±28. Water Science Technology 31(12), 111±118.
Ta C. T. and Brignal W. J. (1998) Application of Compu- Wood M. G., Howes T., Keller J. and Johns M. R. (1998)
tational Fluid Dynamics technique to storage reservoir Two dimensional Computational Fluid Dynamic models
studies. Water Science Technology 37(2), 219±226. for waste stabilisation ponds. Water Research 32(3),
Thies A. T. and Tam C. K. (1996) Computation of turbu- 958±963.