Article 2017 Effects of Geometry and Internals of A Continuous Gravity Settler On Liquid Liquid Separation PDF

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Article

Cite This: Ind. Eng. Chem. Res. 2017, 56, 13929-13944 pubs.acs.org/IECR

Effects of Geometry and Internals of a Continuous Gravity Settler on


Liquid−Liquid Separation
Saroj K. Panda and Vivek V. Buwa*
Department of Chemical Engineering, Indian Institute of Technology Delhi, New Delhi 110016, India

ABSTRACT: Continuous gravity settlers are widely used for


liquid−liquid separations in solvent extraction processes. In
the present work, the effects of settler design [geometry,
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

settling area (A), locations of inlet and outlet] and internals


(baffles, picket fence, end-plate) on the separation perform-
ance were investigated. An experimentally validated Eulerian
CFD model implemented in OpenFOAM was used. For a
fixed flow rate of dispersion (Qt), an increase in the settler
Downloaded via AALTO UNIV on January 30, 2023 at 09:13:33 (UTC).

length led to a reduction in the dispersion-band thickness. For


settlers with length-to-width ratios (L/W) of <1.5, the settler
performance was found to be improved by combined use of
baffle and picket fence. The organic-to-aqueous phase ratio
(αorg/αaq), end-plate height, and aqueous outlet location were
found to influence the phase separation significantly. An empirical correlation was developed to predict the dispersion-band
thickness as a function of Qt, A, ρorg/ρaq, inlet baffle opening slot position, and αorg/αaq. The present work will be useful for the
design of optimal settler configurations.

1. INTRODUCTION process requires a longer time for phase separation. Despite this
The separation of liquid−liquid dispersions is an important limitation, continuous gravity settlers are often chosen where
process that is used in many industries, such as in the oil/ the phases are easily separable because of differences in density
petrochemical industry for the processing of crude oils, in the and because of their other advantages, such as low maintenance
metal processing industry for the extraction of metals from their costs and uncomplicated parts, compared to other process
ores, and in the chemical industry for operations involving equipment.
liquid−liquid reactions/extractions. The types of industrial Understanding of the flow and separation of dispersion
equipment used most frequently for liquid-phase disengage- inside the vessel; for example, due to buoyancy and convective
ment processes include mixer-settlers, stirred vessels, rotating forces acting on drops, drop−drop (binary) and drop−interface
disk contactors (RDCs), centrifugal contactors, and pulsed/ (interfacial) coalescence; can assist in enhancing the perform-
sieve-plate columns. Mixer-settlers consist of a small mixing ance of the settler. In addition to the aforementioned flow
chamber followed by a large gravity settling vessel. The processes, the settler design (i.e., geometry, settling area,
dispersion of liquid and the extraction process are performed in locations of inlet and outlet) and internals (baffle, picket fence,
the mixer using an impeller, and the dispersion is fed to the end plate, and guide plate) also influence the separation
gravity settler for phase separation. RDCs are most suitable for performance. It is therefore essential to understand the effects
systems in which the interfacial tension between the phases is of design parameters and internals on the separation perform-
low, which assists the generation of small drop sizes and ance. For example, a smaller settler leads to flooding of the
thereby increases the interfacial area for mass transfer. The dispersion and high entrainment rates (aqueous phase coming
performance is improved by agitation provided by discs, but the out of the organic outlet and organic phase coming out of the
efficiency is affected by the entrainment of smaller droplets and
aqueous outlet), whereas an oversized settler leads to the
increased axial mixing.1,2 Compared to the other equipment,
locking of large quantities of valuable solvents. The latter also
pulsed columns are relatively tall, which, in turn, increases the
head space. The liquid phases are fed to the column in a contributes to excessive operating costs and risks inside the
countercurrent fashion, and pressurized air is used to pulse the plant. The purpose of the present work is to investigate the
liquid inside the column, which enhances the mass transfer.3 In effects of the aforementioned design parameters and settler
the centrifugal contactor, a spinning rotor is used that mixes the internals on the separation performance through computational
liquid phases extensively. Phase separation occurs inside the
rotor under the action of a high centrifugal force.4 Received: September 10, 2017
The aforementioned equipment such as pulsed-sieve-plate Revised: October 13, 2017
columns, centrifugal contactors, and RDCs require shorter Accepted: October 17, 2017
residence times, whereas gravity settlers are used when the Published: October 17, 2017

© 2017 American Chemical Society 13929 DOI: 10.1021/acs.iecr.7b03756


Ind. Eng. Chem. Res. 2017, 56, 13929−13944
Industrial & Engineering Chemistry Research Article

fluid dynamics (CFD) simulations. In view of this objective, a importance of continuous gravity settlers, only a few
brief review of the relevant literature is provided next. experimental22−27 and modeling24,25 studies of liquid−liquid
Ample studies were carried out in the past on the separation disengagement in continuous gravity settlers have been
of liquid−liquid dispersions through small-scale batch settling published. In most of these investigations, the steady-state
experiments. The measurements in these experiments were dispersion-band thickness (positions of the active and passive
often carried out with the aim of designing continuous settlers interfaces) was measured along the length of the settler. For
using batch settling data.5,6 In addition, several other example, Eckert and Gormely23 investigated the dispersion-
experimental and mathematical modeling studies were carried band thickness as a function of flow rate and phase ratio and
out on batch settling to understand phase separation in detail reported that the dispersed drop size was strongly dependent
(e.g., see refs 7−9). In such works, the researchers successfully on the phase ratio. For a fixed flow rate and phase ratio, the
calculated the positions of coalescence and the sedimentation introduction of dispersion near the active interface led to poor
profiles using measurements followed by simple mathematical separation of the phases.23 However, Eckert and Gormely23
modeling. For example, Hartland and Jeelani5,10−12 performed observed that the level at which the dispersion is introduced in
rigorous modeling to determine the coalescence and the settler is not important for settlers with larger settling areas
sedimentation profiles using experimental data. They reported but is significant for those with smaller settling areas. Further,
that coalescence events occur at the drop−drop (binary) and Jeelani and Hartland24 and Padilla et al.26 performed experi-
drop−interface (interfacial) levels. Therefore, their modeling ments to study the volume fraction of the dispersed phase, drop
considered two events separately, and they developed size distribution at different positions in the dispersion band,
corresponding profiles. The dispersed drops in the sedimenta- drop velocity in the sedimentation zone (by identifying larger
tion zone rise due to buoyancy, enter the dense-packed zone, drops or clusters of drops that are moving), positions of the
and undergo interfacial coalescence. Also, they calculated the active and passive interfaces, and effects of the dispersion
volume rate of coalescence by considering the drop volume entrance level on phase separation. In another experimental
fraction, disengaging area, and drop coalescence time at the investigation of a mixer-settler, Gharehbagh and Mousavian27
interface. Apart from sedimentation/coalescence profiles, the used tributyl phosphate (TBP) with kerosene as the dispersed
effects of the initial dispersion height, phase ratio, fluid phase and HNO3 with zirconium oxide as the continuous
properties (density, viscosity, interfacial tension), and initial phase. They investigated the dispersion characteristics such as
drop size distribution (determined by impeller speed) on phase drop size distribution, mean drop size, and dispersed-phase
separation were also studied.11,12 However, the corresponding volume fraction at different impeller speeds in the mixer and
mathematical models were tuned with empirical parameters and dispersion-band thickness in the settler.
constants to improve the agreement of the positions of the Different mathematical models24,25 have been used to predict
active and passive interfaces (dispersion-band thickness), the time for binary (drop−drop) coalescence, the interfacial
dispersed-phase volume fraction, coalescence rate, and drop (drop−interface) coalescence rate, and the variations of the
size distribution with the measurements. drop diameter and drop velocity along the length of the settler
In research efforts over the past several years, numerous at steady state. The interfacial coalescence rate was calculated
studies based on CFD simulations were performed on liquid− using information such as the time taken by the drops for
liquid flows in various types of process equipment used for coalescence, the diameter of the drops, and the dispersed-phase
solvent extraction, for example, stirred vessels,13−15 pulsed/ volume fraction at the coalescing interface. Similarly, the
sieve-plate columns,16,17 RDCs,18−20 and centrifugal contac- variation in drop velocity along the length of the settler was
tors.21 In these investigations, an Eulerian−Eulerian approach determined from the dispersed-phase flow rate, the volume
was used to simulate the liquid−liquid flows in these types of fraction of the dispersed phase, and the dispersion-band
equipment. Most of the aforementioned works involved thickness along the length of the settler. Most of these models
predictions of the volume fractions and velocity distributions used to predict the aforementioned parameters are lower-order
of the dispersed and continuous phases inside the different models and rely on simplified assumptions and several
types of equipment using two-fluid CFD models. Whereas adjustable constants. Their predictive abilities are often limited
some of the researchers (e.g., Laurenzi et al.,14 Cheng et al.,15 to the settler geometry/configuration for which the measure-
Drumm and Bart19) considered constant drop sizes, others ments were performed and cannot be generalized. Therefore,
(e.g., Wang and Mao,13 Yadav and Patwardhan16) calculated the applicability of these models is rather limited, and these
the drop size using mathematical correlations. In simulations of models cannot be used to predict the separation performances
a stirred vessel, Wang and Mao13 reported the importance of for varying positions of settler internals (baffles, picket fence,
drag force over other forces such as lift and added mass forces. and end plate).
In their study, the radial dispersed-phase volume fraction was Based on experiments26,28 on a continuous gravity settler,
overpredicted in comparison to the measurements at low two different zones, namely, the sedimentation zone (which
impeller speed, and they emphasized the need for coalescence forms in a narrow band above the passive interface) and the
and breakage models to be incorporated into the CFD model. dense-packed zone (which forms above the sedimentation
Drumm and Bart19 considered a constant drop size in their zone), are formed within the dispersion band. Nonuniformly
Eulerian two-fluid simulations of a RDC and selected the drag sized and loosely packed drops move randomly in the
force as the only interaction force between the phases. Later, sedimentation zone. The velocity of the drops is high near
Ghaniyari-Benis et al.20 performed Eulerian multifluid simu- the inlet and decreases toward the end of the settler. Whereas
lations of liquid−liquid flows in a RDC and reported a some drops move horizontally due to convective forces toward
satisfactory agreement between the predicted and measured the end of the settler, others undergo binary coalescence, grow
overall volume fractions of the dispersed phase. in size, and rise vertically due to buoyancy. Larger drops are
Analysis of the present literature shows that liquid−liquid accumulated in the dense-packed zone, and interfacial
flows in batch settlers have been widely studied. Despite the coalescence takes place at the interface of the clear organic
13930 DOI: 10.1021/acs.iecr.7b03756
Ind. Eng. Chem. Res. 2017, 56, 13929−13944
Industrial & Engineering Chemistry Research Article

phase and the dense-packed zone. As interfacial coalescence module of the open-source CFD code OpenFOAM was
proceeds, clear organic phase is formed continuously. Further, developed further for the simulation of aqueous- and organic-
several studies29−32 were carried out to separate liquid−liquid phase disengagement in a continuous gravity settler with a
dispersion using phase inversion (with less settling area and freeboard region at the top. Interphase momentum exchange
high throughput). For example, Hadjiev and Paulo30,31 was taken into account only through the drag force acting on
successfully improved separation efficiency by using a smaller the drops. The flow was assumed to be Newtonian, laminar,
settling area and allowing high throughput. They varied the and incompressible. The continuity and momentum equations
phase ratio in a vertical settler to achieve an increase in the were solved with appropriate interphase momentum exchange
coalescence rate between dispersed drops in a small settling models. The following mass conservation equations were
area. In our previous work,33 phase inversion was studied by solved for the aqueous (a), organic (o), and gas (g) phases.
changing the fluid properties, namely, the difference in density

between the phases. When the density of the organic phase was (αoρo ) + ∇·(αoρo Vo⃗ ) = 0
very close to the density of the aqueous phase, poor separation ∂t (1)
was observed. However, with a further increase in the organic ∂
density (ρo > ρa), phase inversion occurred, as expected. (αaρa ) + ∇·(αaρa Va⃗ ) = 0
∂t (2)
The analysis of the literature presented above indicates that a
few experimental studies are available on continuous gravity ∂
settlers and on the development of lower-order models that rely (αgρg ) + ∇·(αgρg Vg⃗ ) = 0
∂t (3)
on simplified assumptions and several adjustable constants. For
the design and scaleup of continuous gravity settlers, it is Similarly, the following momentum conservation equations
important to develop experimentally validated comprehensive were solved for the aqueous (a), organic (o), and gas (g)
CFD models that are capable of predicting the effects of inlet phases.
flow rates, settler geometry and internals (baffles, picket fence, ∂
and end plate), phase ratio and drop size distribution at the (αoρo Vo⃗ ) + ∇·(αoρo Vo⃗ Vo⃗ ) = −αo∇P − ∇·αoτo̿ + αoρo g ⃗
inlet, and physical properties. For the development of such ∂t
comprehensive CFD models, it is necessary to account for + M⃗ o − a + M⃗ o − g (4)
changes in the drop size distribution caused by binary
coalescence processes occurring at different rates in the ∂
sedimentation and densely packed regions, simultaneous (αaρa Va⃗ ) + ∇·(αaρa Va⃗ Va⃗ ) = −αa∇P − ∇·αaτa̿ + αaρa g ⃗
∂t
liquid−liquid separation (or disengagement) caused by
interfacial coalescence, the effects of drop transport due to + M⃗ a − g − M⃗ o − a (5)
convective flow in the setter and the simultaneous rise due to
buoyancy, and so on. To make progress in this direction, the ∂
(αgρg Vg⃗ ) + ∇·(αgρg Vg⃗ Vg⃗ ) = −αg∇P − ∇·αgτg̿ + αgρg g ⃗
present work is focused on the development of an Eulerian ∂t
two-fluid model using the open-source CFD solver Open- − M⃗ o − g − M⃗ a − g (6)
FOAM and its use to investigate the effects of the
aforementioned operating, design, and physical parameters on Because the flow was assumed to be Newtonian and laminar,
the separation performance. the stress tensor (τ)̿ was calculated as
In the present work, we have performed simulations of
liquid−liquid flow in a continuous gravity settler using τ ̿ = −μ(∇V̅ + ∇V̅ T)
OpenFOAM. In our previous work,33 we reported the for all of the phases. As mentioned earlier, the interphase
development of an Eulerian−Eulerian model in OpenFOAM momentum exchange was taken into account only through the
and validation of the numerical results using measured drag force and was calculated as
dispersion-band thicknesses and organic-phase volume frac-
tions. The predicted results showed a satisfactory agreement 3 C Do−a ⃗
with the measured data. The model was further validated using M⃗ o − a = Ko − aαaαo(Vo⃗ − Va⃗ ), Ko − a = ρ |Vo − Va⃗ |
4 a do
experiments performed in settlers with different lengths. The (7)
effects of different drag corrections that mimic the presence of
neighboring drops in the dense region were also investigated. In 3 C Do−g ⃗
this work, the experimentally validated computational model M⃗ o − g = Ko − gαoαg(Vg⃗ − Vo⃗ ), Ko − g = ρ |Vg − Vo⃗ |
4 o dg
was used to investigate the effects of settler geometry (length
and width, locations of the organic- and aqueous-phase outlets) (8)
and internals (picket fences, end plate) used in the settler on
the liquid−liquid separation. An empirical correlation was 3 C Da−g ⃗
M⃗ a − g = K a − gαaαg(Vg⃗ − Va⃗ ), Ka− g = ρ |Vg − Va⃗ |
developed to predict the dispersion-band thickness as a 4 a dg
function of flow rate, settling area, density ratio, inlet baffle (9)
opening slot position, and organic-to-aqueous phase ratio.
The drag coefficient was determined using the following
2. CFD MODEL model34
⎧ ⎫
2.1. Governing Equations. For simulations of liquid− ⎪ 24 8 ⎛⎜ Eoï − j ⎞⎪
⎟⎟⎬
liquid flow in a continuous gravity settler, the Eulerian− C Di−j = max⎨ (1.0 + 0.15Re 0.687
),
3 ⎜⎝ Eoï − j +
⎪ i − j ⎪
⎩ Rei − j 4 ⎠⎭
Eulerian (E−E) approach, based on the assumption of
interpenetrating continua, was used. The twoPhaseEulerFoam (10)

13931 DOI: 10.1021/acs.iecr.7b03756


Ind. Eng. Chem. Res. 2017, 56, 13929−13944
Industrial & Engineering Chemistry Research Article

Figure 1. (a) Schematic of the settler with dimensions, (b) boundary conditions and different parts of the settler (1, inlet; 2, baffle; 3, picket fence; 4,
bottom surface; 5, end plate; 6, guide plate; 7, top surface; 8, organic outlet; 9, aqueous outlet), and (c) schematic of the computational mesh.

Table 1. Boundary Conditions


inlet organic outlet aqueous outlet top surface inner walls outer walls
P zero gradient fixed value of 0 fixed value of 0 fixed value of 0 zero gradient zero gradient
Vo fixed valuea zero gradient zero gradient fixed value of 0 fixed value of 0 fixed value of 0
Va fixed valuea zero gradient zero gradient fixed value of 0 fixed value of 0 fixed value of 0
Vg fixed value of 0 fixed value of 0 fixed value of 0 zero gradient fixed value of 0 fixed value of 0
αo fixed value of 0.5 zero gradient zero gradient fixed value of 0 zero gradient zero gradient
αa fixed value of 0.5 zero gradient zero gradient fixed value of 0 zero gradient zero gradient
αg fixed value of 0 zero gradient zero gradient fixed value of 1 zero gradient zero gradient
a
Calculated on the basis of the inlet flow rate.

where Re is the Reynolds number, defined as 2.2. Solution Domain and Boundary Conditions.
Detailed dimensions of the continuous gravity settler
→ → considered for the simulations are shown in Figure 1a. The
Rei − j = di| Vi − Vj |ρj /μj
overall dimensions of the settler were 0.59 m (length), 0.25 m
and Eö is the Eötvös number, defined as (width), and 0.33 m (height), where an additional 0.1 m height
was kept as the freeboard region. Separate outlets were
g ⃗(ρi − ρj )di 2 provided for the exits of the organic and aqueous phases after
Eoï − j = the separation. An interface controller connected to the
σi − j aqueous-phase outlet was used to maintain the desired
hydrostatic head in the actual operation of the continuous
The indices i−j indicate the phase pairs o−a, o−g, and a−g. settler. However, in the present work, because the height of the
13932 DOI: 10.1021/acs.iecr.7b03756
Ind. Eng. Chem. Res. 2017, 56, 13929−13944
Industrial & Engineering Chemistry Research Article

interface controller was maintained at 0.23 m from the bottom


of the settler in the experiments, the aqueous outlet was
positioned at the same height. The boundary conditions and
computational mesh used for the simulations of the continuous
gravity settler are shown in panels b and c, respectively, of
Figure 1. A uniform structured mesh was prepared using third-
party mesh generation software and was exported to Open-
FOAM. The effects of the grid resolution on the predicted
dispersion-band thickness are discussed in the following
section.
For the aqueous- and organic-phase flow rates used in the
present work, the liquid−liquid flow in the continuous settler
was laminar. The “velocity inlet” boundary condition was
specified at the inlet, whereas the “pressure outlet” boundary
condition was specified at the top surface of the settler. All walls
including baffles, picket fence, end plate, and guide plate were
specified as “no-slip” boundary conditions, whereas the aqueous
and organic outlets were specified as “zero gradient”. The
boundary conditions used in the present work are summarized Figure 2. Snapshot of the simulated distribution of the organic-phase
in Table 1. To discretize the temporal and spatial derivatives, volume fraction in a continuous gravity settler (L = Lref, Qt = 400 L/h,
different discretization schemes were used, as described in αo,inlet = αa,inlet = 0.5, dd = 100 μm).
Table 2. A convergence criterion of 1 × 10−3 and an adjustable

Table 2. Discretization Schemes


term scheme accuracy
time scheme Euler first-order, implicit
gradient Gauss second-order
scheme
divergence GaussLimitedLinearV1, second-order,
scheme GaussLimitedLinearV, central
GaussLinear 1 differencing
Laplacian GaussLinearCorrected second-order
scheme
interpolation Linear second-order,
scheme central
differencing

time step were used for all simulations. The user-defined drag
correlation34 was implemented in the twoPhaseEulerFoam
module of OpenFOAM (2.2.2) to simulate the three-phase (air,
aqueous phase, and organic phase) flow.

3. RESULTS AND DISCUSSION


3.1. Validation of Numerical Results. Preliminary
simulations of the liquid−liquid flow and disengagement in
the continuous gravity settler were carried out at Qt = 400 L/h
using water (aqueous phase) (μa = 0.00116 kg/ms, ρa = 1033
kg/m3) as the continuous phase and dodecane (organic phase)
(μo = 0.00246 kg/ms, ρo = 800 kg/m3) as the dispersed phase.
A representative snapshot of the simulated organic-phase
volume fraction distribution is shown in Figure 2. Because
the snapshot is provided for the organic-phase volume fraction,
the dark blue color on the scale indicates zero for both the
aqueous (bottom) and air (top) phases. The interface between
the organic phase and the freeboard air is marked by line 1,
whereas the active interface (AI) and the passive interface (PI)
are marked by lines 2 and 3, respectively. The dark blue color at
the bottom indicates pure aqueous phase, the red color
indicates pure organic phase, the variation from blue to red in
the middle indicates the dispersion band, and the dark blue at Figure 3. Effects of grid resolution on the positions of the active
the top (above the organic phase) indicates the freeboard air interface (AI) and passive interface (PI) along the length of the settler
phase. at steady state for different grids: (a) L = Lref and (b) L = 2Lref (Qd =
Before performing the actual simulation and analysis of the 200 L/h, αo,inlet = αa,inlet = 0.5, dd = 100 μm).
continuous settler, we carried out a detailed study of grid
13933 DOI: 10.1021/acs.iecr.7b03756
Ind. Eng. Chem. Res. 2017, 56, 13929−13944
Industrial & Engineering Chemistry Research Article

Figure 4. Comparison of measured28 and predicted dispersion-band thicknesses at a Qt value of 600 L/h (dd = 138 μm, αo,inlet = αa,inlet = 0.5) for
different positions of the end plate along the length of the settler: (a) L = Lref, (b) L = 0.67Lref, and (c) L = 0.56Lref.

resolution to ascertain the accuracy of the predicted results. area is greater and a narrow dispersion band is formed.
The effects of grid resolution were investigated for settler However, at L = 0.67Lref and L = 0.56Lref, because of the
length of L = Lref by performing simulations using coarse reduced settling areas, the residence times of the dispersion are
(37413 cells), medium (77000 cells), fine (103461 cells), and shorter, leading to thicker dispersion bands. An increased
very fine (212134 cells) grids at Qt = 400 L/h. The predicted residence time provides more time for binary and interfacial
positions of the active and passive interfaces (dispersion-band coalescence and results in a reduction of the dispersion-band
thickness) are shown in Figure 3a. It can be observed that the thickness.
dispersion-band thickness changed substantially with grid Figure 5 shows that the predicted organic-phase volume
refinement from the coarse to the medium level, whereas the fraction was in a satisfactory agreement with the measurements
difference for the refinement from medium to fine and very fine of Thaker et al.,28 except for some minor variations in some
grid resolutions was marginal. Therefore, considering computa- regions. The maximum and minimum errors between the
tional costs, the medium grid with 77000 cells was chosen for predicted and measured organic-phase volume fractions were
all further simulations for L = Lref. Further, to carry out studies found to be +14.08% and −2.24% for L = Lref, +15.41% and
on longer settlers (L = 1.5Lref and L = 2Lref), additional grid- −0.65% for L = 0.67Lref, and +14.49% and −1.12% for L =
resolution effects were studied for the settler with L = 2Lref. 0.56Lref. At a settling length of L = Lref, owing to the thin
Three different grid resolutions were chosen, and the dispersion band, it was possible to measure the organic-phase
corresponding total numbers of cells represented coarse volume fraction for two rows of ports for Qt = 600 L/h.
(56000 cells), medium (112368 cells), and fine (222678 However, for other settling areas (L = 0.67Lref and L =
cells) grids. Similarly to the earlier case for L = Lref, a substantial 0.56Lref), the measurements were performed for three rows
change in the positions of the AI and PI from coarse to medium because of the thick dispersion band. For all of the settling areas
grid resolution was observed, whereas a marginal difference was considered in the present work at the constant value of Qt =
observed from medium to fine grid resolution (see Figure 3b). 600 L/h, the measured organic-phase volume fraction changed
Hence, the medium grid with 112368 cells was used for further marginally along the length of the settler. The difference
simulations of the settler with L = 2Lref. between the simulated and measured organic-phase volume
Liquid−liquid flow and disengagement experiments were fractions can be attributed to the two-fluid model used for the
carried out in a laboratory-scale continuous gravity settler made simulations. It should be noted that the two-fluid model with
of polymethyl methacrylate (PMMA).33 The dispersion-band the measured average drop size (dd = 138 μm) was used in the
thickness and dispersed-phase volume fraction were measured simulations and that changes in the drop size distribution due
for different flow conditions and for Qt in the range of 400−800 to binary and interfacial coalescence were ignored in the
L/h. We validated (Qt = 400 and 800 L/h) the predictions of present work. Further, the experimentally verified computa-
the dispersion-band thickness and organic-phase volume tional model was used to investigate the effects of settler length;
fraction against experimental measurements.33 Further, addi- settler internals such as inlet baffle, picket fence, and end-plate
tional experiments28 and corresponding simulations were height; aqueous outlet position; and inlet volume fraction on
performed for a Qt value of 600 L/h at three different end- the liquid−liquid flow and separation performance, and the
plate positions (hence three different settling areas). The results are discussed in the following sections.
measured and predicted dispersion-band thicknesses (positions 3.2. Effects of Drag Corrections. The interphase
of AI and PI) and organic-phase volume fractions were momentum exchange occurring only through the drag force
compared (see Figures 4 and 5). In Figure 4, for L = Lref, the was considered in the present work.34 To account for the
dispersion has more time to reach the end plate as the settling effects of the presence of neighboring drops (swarms of drops)
13934 DOI: 10.1021/acs.iecr.7b03756
Ind. Eng. Chem. Res. 2017, 56, 13929−13944
Industrial & Engineering Chemistry Research Article

along the vertical distance in the dispersion band, the drag


coefficient was corrected as35
C Do−a = C DOo−a(1 − αo) p (11)
where the subscripts o and a denote the organic and aqueous
phases, respectively
The effects of the drag-coefficient correction parameter (p,
considering p = 1, 0.5, and 0.1) on the dispersion-band
thickness (positions of AI and PI) and the variations in the
organic-phase volume fraction distribution along the length of
the settler were investigated at a Qt value of 800 L/h (see
Figure 6). It was observed that a fluctuating dispersion band
was obtained using p = 1, whereas no fluctuations in AI and PI
were observed at p = 0.5 and 0.1. The dispersion-band
thickness and organic-phase volume fraction predicted using p
= 0.1 agreed well with the measurements. As the value of p was
increased from 0.1 to 0.5 and then to 1, the values of Ko−a
(calculated using eqs 7 and 10) averaged over the entire
domain volume were found to decrease; for example, Ko−a =
10.08 × 105 kg/(m3 s) at p = 0.1, Ko−a = 7.14 × 105 kg/(m3 s)
at p = 0.5, and Ko−a = 5.29 × 105 kg/(m3 s) at p = 1. Such
decrease in the value of Ko−a led to an increase in the dispersed-
phase velocities; specifcally, the magnitudes of the organic-
phase velocity averaged over entire domain were 0.0172,
0.0305, and 0.0546 m/s for p = 0.1, 0.5, and 1, respectively.
This correction factor mimics the effects of the presence of
neighboring drops, in particular in the densely packed region.
3.3. Effects of Settler Length. To understand the effects
of the settler length on the separation performance, simulations
were performed for settlers with different lengths, namely, 59
cm (hereafter called the reference length, Lref, as it was used for
the experiments), 88.5 cm (1.5Lref), and 118 cm (2Lref), at 400
L/h (constant drop size of 100 μm). Snapshots of the predicted
steady-state distributions of the organic-phase volume fractions
for settlers of different lengths are shown in Figures 2 and 7.
The effects of the settler length on the steady-state dispersion-
band thickness are shown in Figure 8. With increasing settler
length, the dispersion-band thickness decreased. Also with
increasing settler length, the residence time of the dispersion
flowing inside the settler increased, and this led to a narrow
dispersion band. The residence time of the dispersion in the
settler was calculated as tr = L/Vt, where L is the length of the
settler and Vt is the velocity of the dispersion through the cross-
sectional area (ACS) of the settler and can be estimated as Vt =
Qt/ACS. The residence time was found to decrease with an
increase in Qt and to increase with an increase in the settler
length (see Figure 9). Whereas the incoming dispersion flows
along the length of the settler, drops move upward due to
buoyancy and coalesce at the active interface. Therefore, for
settlers with shorter lengths, fewer drops coalesce at the
interface as the dispersion travels faster toward the end plate,
leading to a thick dispersion band. In the case of smaller settling
areas, large numbers of drops are packed close together near
the coalescence front and form a thick dispersion band.23 To
understand the relationship between the flow rate and the
settler length, a few additional simulations were carried out at
higher flow rates (600−1000 L/h with dd = 100 μm) for a
settler with a length of Lref, and flooding was observed inside
the settler, which, in turn, led to a poor separation. With all
Figure 5. Comparison of the predicted organic-phase volume fractions other parameter being the same, the length of the settler was
with the measurements of Thaker et al.28 at a Qt value of 600 L/h (dd increased to 1.5Lref and 2Lref in the present study (Figure 10).
= 138 μm, αo,inlet = αa,inlet = 0.5) for different positions of the end plate: Because of the increasing settling area/length, good separations
(a) L = Lref, (b) L = 0.67Lref, and (c) L = 0.56Lref. were observed instead of flooding inside the settler. At a fixed
13935 DOI: 10.1021/acs.iecr.7b03756
Ind. Eng. Chem. Res. 2017, 56, 13929−13944
Industrial & Engineering Chemistry Research Article

Figure 6. Comparison of measured33 and predicted (a) dispersion-band thicknesses and (b) organic-phase volume fractions obtained using no drag
correction and using drag corrections of 0.1, 0.5, and 1 at a Qt value of 800 L/h (dd = 129 μm, αo,inlet = αa,inlet = 0.5, L = Lref). Symbols and lines
indicate experimental and simulated results, respectively. (NC, no drag correction; WC, with drag correction.)

settler length, an increase in Qt led to a reduction in the 3.4. Effects of Picket Fence. A schematic of the baffle and
residence time, and therefore, the dispersion-band thickness picket fence is shown in Figure 1a. The simulations were
was found to increase for all settler lengths considered in the performed with the exact dimensions and positions of the slots
present work. for baffles and picket fences used in the experiments. In our
13936 DOI: 10.1021/acs.iecr.7b03756
Ind. Eng. Chem. Res. 2017, 56, 13929−13944
Industrial & Engineering Chemistry Research Article

Figure 9. Effects of flow rate and settler length on the residence time
(αo,inlet = αa,inlet = 0.5, dd = 100 μm).
Figure 7. Predicted steady-state distributions of the organic-phase
volume fraction for settler lengths of (a) 88.5 cm (1.5Lref) and (b) 118
cm (2Lref) (Qt = 400 L/h, αo,inlet = αa,inlet = 0.5, dd = 100 μm).

Figure 8. Effects of settler length on the steady-state dispersion-band


thickness for settler lengths of (a) 59 cm (Lref), (b) 88.5 cm (1.5Lref),
and (c) 118 cm (2Lref) (Qt = 400 L/h, αo,inlet = αa,inlet = 0.5, dd = 100 Figure 10. Predicted steady-state distributions of the organic-phase
μm). volume fraction for a Qt value of 600 L/h at (a) L = Lref, (b) L =
1.5Lref, and (c) L = 2Lref (αo,inlet = αa,inlet = 0.5, dd = 100 μm).
previous work,33 the effects of the baffle opening position were
studied, and the predictions were compared qualitatively with dispersion-band thickness was observed between the cases with
the measurements of Padilla et al.;26 satisfactory agreement was no picket fence and with the picket fence in the second slot.
found. Whereas the position of the baffle opening near the To understand the effects of the picket fence further,
settler inlet resulted in a poor settler performance, positions of additional simulations were carried out with a reduced settler
the baffle opening in the middle or near the bottom of the length (0.75Lref) and an increased settler width such that the
settler resulted in better separation performances.23 In the settling area was kept constant and equal to that of the base
present work, all of the simulations were carried out with the case (with a length of L = Lref). Snapshots of the time evolution
baffle placed in the first slot and the other slots used for picket of the dispersion-band thickness with and without picket fences
fences. Simulations were performed in which the distance for L = 0.75Lref are shown in Figure 12. It can be observed that
between the baffle and picket fence was varied by inserting the the inclusion of the picket fence reduced the dispersion-band
picket fence in the (i) second slot, (ii) fourth slot, and (iii) sixth thickness significantly. The average organic-phase velocity was
slot inside the settler (see Figure 1a). A comparison of the monitored on y−z (cross-sectional) planes located at different x
dispersion-band thicknesses for these arrangements of baffles positions along the length of the settler for lengths of Lref and
and picket fences is shown in Figure 11. The influence of the 0.75Lref for cases without and with a picket fence in the second
position of the picket fence on the thickness of the dispersion slot. Three different positions were chosen for Lref: x = 0.085 m
band was marginal using the experimental settler length (Lref) (region before the picket fence), x = 0.135 m (region after the
at a Qt value of 400 L/h. However, a visible difference in picket fence), and x = 0.257 m (middle region of the settling
13937 DOI: 10.1021/acs.iecr.7b03756
Ind. Eng. Chem. Res. 2017, 56, 13929−13944
Industrial & Engineering Chemistry Research Article

In our previous work,33 with an increase in the drop diameter


(e.g., dd = 200 μm), significant fluctuations were seen within the
dispersion band. Therefore, the effects of a picket fence were
investigated further for the case of dd = 200 μm (Qt = 400 L/h,
L = Lref) by considering no picket fence and also by considering
a picket fence in the second slot, in the fourth slot, and in the
second and fifth slots (two picket fences in a row). The effects
of the picket-fence position on the organic-phase volume
fraction distribution and isosurfaces are shown in Figure 13.
The dispersion band (see Figure 13ia) and isosurface (see
Figure 13iia) were found to fluctuate substantially when no
picket fence was considered in the simulation. As can be seen in
Figure 13ib,iib, the picket fence in the second slot with the
baffle led to a reduction of the fluctuations in the dispersion
band. The snapshots and isosurfaces of the organic-phase
volume fraction in Figure 13c,d shows that, for the picket-fence
arrangements in the fourth slot and in the second and fifth
Figure 11. Effects of the picket-fence position on the dispersion-band slots, the dispersion-band fluctuations were stabilized in both
thickness (Qt = 400 L/h, αo,inlet = αa,inlet = 0.5, dd = 100 μm, L = Lref, arrangements and even more so in the latter arrangement
baffle position = first slot). involving two picket fences.
The stabilization of the dispersion-band fluctuations and the
distribution of the organic-phase volume fraction were analyzed
quantitatively using the time histories of the organic-phase
volume fraction recorded at points before and after the picket
fence and also in the middle of the settler length (x = 0.275 m)
(see Figure 12). In the case of no picket fence (Figure 14a), the
organic-phase volume fraction was seen to fluctuate signifi-
cantly at all points (x = 0.085, 0.125, and 0.275 m) for dd = 200
μm. The corresponding standard deviations (SDs) at these
points were 0.071, 0.071 and 0.069, respectively. The amplitude
of the fluctuations was found to decrease when a picket fence
was inserted in the second slot at x = 0.085 m (SD = 0.067)
and to decrease even more when the picket fence was inserted
at x = 0.125 m (Figure 14b). The fluctuations were reduced
further for the picket fence at x = 0.275 m (SD = 0.024), and
the organic-phase volume fraction was increased. The volume
fraction was recorded at four different points for the case of
picket fence in the fourth slot (Figure 14c). It was found that
the amplitude of the organic-phase fluctuations in the
beginning (x = 0.085 m) of the settler was similar to that
observed for the case of no picket fence (SD = 0.069).
However, the fluctuations were reduced at other points (x =
0.185, 0.225, and 0.275 m) (SD at x = 0.275 was 0.039), and it
Figure 12. Effects of a picket fence on the simulated time evolutions of was found that the magnitude of the organic-phase volume
the organic−aqueous phase separation and dispersion-band thickness fraction was similar. Further, the organic-phase volume fraction
at (i) 60, (ii) 180, and (iii) 250 s: (a) without a picket fence and (b) was recorded at various points for the case of two picket fences
with a picket fence (Qt = 400 L/h, αo,inlet = αa,inlet = 0.5, dd = 100 μm, L in a row (second and fifth slots simultaneously) (see Figure
= 0.75Lref, W = 0.334 m). 14d), and it was found that the fluctuations were reduced along
the length of the settler. In addition, the SD was reduced from
area). The corresponding positions for 0.75L ref were x = 0.085, 0.043 (at x = 0.085) to 0.021 (at x = 0.275 m). Whereas the
0.135, and 0.195 m, respectively, for the calculation of the amplitude of the fluctuations decreased, the volume fraction of
average organic-phase velocity on the corresponding y−z the organic phase was found to increase from the picket fence
planes. Whereas the velocities calculated at these three in the second slot to that in the fifth slot. The predictions show
positions for Lref considering the absence and presence of a that the use of multiple picket fences in a row is more effective
picket fence (in the second slot) showed a marginal difference, than the use of one picket fence inserted away from the inlet.
the average velocities for 0.75Lref showed substantial differences 3.5. Effects of End-Plate Height and Aqueous-Phase
for the picket fence in the second slot position. It was observed Outlet Position on Phase Separation. In a continuous
that the organic-phase velocity for 0.75L ref decreased from gravity settler, the organic phase flows over the end plate after
0.0153 m/s (x = 0.085 m) to 0.0116 m/s (x = 0.195 m) for the the phase separation. Whereas the position of the dispersion
case where no picket fence was used. However, when the picket band is maintained with an interface controller, the thickness of
fence was used in the second slot, the organic-phase velocity the dispersion band in the settler depends on the height of the
was found to decrease from 0.012 m/s (x = 0.085 m) to end plate. The end plate, which prevents the liquid dispersion
0.00688 m/s (x = 0.195 m). from overflowing along with pure organic phase needs to be
13938 DOI: 10.1021/acs.iecr.7b03756
Ind. Eng. Chem. Res. 2017, 56, 13929−13944
Industrial & Engineering Chemistry Research Article

Figure 13. Effects of a picket fence for dd = 200 μm on the (i) phase separation and (ii) organic-phase isosurface: (a) without a picket fence and (b−
d) with a picket fence in (b) the second slot, (c) the fourth slot, and (d) the second and fifth slots (Qt = 400 L/h, L = Lref).

Figure 14. Comparison of the fluctuation time series of the organic-phase volume fraction along the length of the settler: (a) without a picket fence
and (b−d) with a picket fence in (b) the second slot, (c) the fourth slot, and (d) the second and fifth slots (dd = 200 μm, Qt = 400 L/h) (PF, picket
fence).

13939 DOI: 10.1021/acs.iecr.7b03756


Ind. Eng. Chem. Res. 2017, 56, 13929−13944
Industrial & Engineering Chemistry Research Article

Figure 16. Effects of the distance of the aqueous-phase outlet from the
bottom of the settler on the separation of the organic and aqueous
phases and the dispersion-band thickness at Qt = 400 L/h: (a) 0.23,
(b) 0.2, (c) 0.189, and (d) 0.167 m (L = Lref, αo,inlet = αa,inlet = 0.5, dd =
100 μm).

Table 4. Effects of the Aqueous Outlet Position on the


Separation Performance
flow rate at
flow rate at aqueous outlet flow rate at organic
inlet (L/h) (L/h) outlet (L/h)
aqueous outlet
position from aqueous +
bottom (m) organic aqueous organic aqueous organic
0.23 400 200 0 0 200
0.2 400 197.27 25.02 0 179.78
Figure 15. Effects of the end-plate height from the bottom on the 0.189 400 201.88 198.76 0 0
instantaneous distribution of the organic-phase volume fraction: (a) 0.167 400 200.07 203.11 0 0
0.25, (b) 0.23, and (c) 0.21 m (Qt = 400 L/h, L = 1.5Lref, αo,inlet =
αa,inlet = 0.5, dd = 100 μm).
15 for end-plate heights of 0.23 and 0.21 m, respectively),
Table 3. Effects of the End-Plate Height on the Separation where the corresponding differences in the levels of the baffle
Performance opening and the end-plate height were reduced, poor
separations were observed.
flow rate at The aqueous outlet position also has an influence on phase
flow rate at aqueous outlet flow rate at organic
inlet (L/h) (L/h) outlet (L/h) separation. In the actual operation of a settler, an interface
controller is used to vary the aqueous outlet location and,
end-plate height aqueous +
from bottom (m) organic aqueous organic aqueous organic therefore, to control the dispersion-band thickness and the
overflow of the liquid dispersion from the settler. The role of an
0.25 400 200 0 0 200
interface controller was mimicked by varying the location of the
0.23 400 181.23 0 18.37 202.18
aqueous-phase outlet in the simulations. Figure 16 shows the
0.21 400 167.83 0 34.212 197.69
steady-state separations of the organic and aqueous phases for
different positions of the aqueous-phase outlet. When the
designed adequately for the particular size of the settler. location of the aqueous outlet was changed from 0.23 to 0.2 m
Therefore, the effects of the height of the end plate on the (from the bottom), the hydrostatic head at the aqueous outlet
phase separation were investigated by varying the end-plate increased, which consequently resulted in a higher outward
height from 0.25 to 0.21 m for a settler with a constant length flow rate. The corresponding steady-state mass flow rates are
of L = 1.5Lref. Typical steady-state snapshots of the simulated summarized in Table 4 for both the organic- and aqueous-phase
organic-phase volume fractions for different end-plate heights outlets. The snapshot (see Figure 16b) and the corresponding
are shown in Figure 15. Better phase separation was obtained mass flow rates provided in Table 4 indicate that some amount
when the end-plate height was maintained at 0.25 m (see of organic liquid flowed along with the clear aqueous phase
Figure 15a). When the end-plate height was reduced to 0.23 or through the aqueous outlet. After the position of the aqueous
0.21 m, the overflow of the liquid dispersion along with the outlet had been changed to 0.189 and 0.167 m, the dispersion
organic phase was observed. The effects of the end-plate height level fell below the end-plate height (Figure 16c,d), and the
on the flow rates of the two phases through the aqueous and dispersion (aqueous + organic) flowed only through the
organic outlets are reported in Table 3. It was observed that the aqueous outlet (also see Table 4).
difference between the upper height of the baffle opening and 3.6. Effects of Inlet Volume Fraction. In the simulations
the end plate was greater for the end plate with a height of 0.25 performed to investigate the effects of various internals on
m. Therefore, in this case, the dispersion has more time for settler performance, an inlet dispersion with αo,inlet = αa,inlet =
phase separation after entering through the inlet baffle. 0.5 was used. Further, the effects of the inlet phase volume
However, in the other two cases (see panels b and c of Figure fractions on the performance of the settler were investigated by
13940 DOI: 10.1021/acs.iecr.7b03756
Ind. Eng. Chem. Res. 2017, 56, 13929−13944
Industrial & Engineering Chemistry Research Article

of the active and passive interfaces are shown in Figures 17 and


18, respectively. In batch settling experiments, Henschke et al.7
showed experimentally that, with decreasing organic-phase
volume fraction, the drop size decreased. Even though the drop
size is small, because of the low volume fraction, the dispersion
is separated more rapidly from the continuous phase. Because
the volume fraction of the organic phase was further reduced
(see Figure 17b,c), the separation was faster, and this led to a
thinner dispersion band than in the base case. However, the
coalescence front was sharp in all cases (Figure 17a−c),
whereas the settling front in Figure 17c is not sharp, and this
phenomenon arose because of the change in phase ratio.23
Eckert and Gormely23 further explained that the settling front is
easily distinguishable for concentrated dispersions, whereas it
becomes hazy in the case of dilute dispersions. It can also be
seen in Figure 17 that the increase in αa,inlet from 0.5 to 0.8 did
not significantly influence the position of the active interface. In
addition, the effects of the organic-to-aqueous phase ratio on
the dispersion-band thickness are compared quantitatively and
discussed in the following section.
3.7. Correlation for the Prediction of Dispersion-Band
Thickness. A few correlations23,26 have been proposed to
calculate the dispersion-band thickness in a continuous gravity
settler. The correlation used by Eckert and Gormely23 and
Padilla et al.,26 which was initially proposed by Ryon et al.,36
relates the dispersion-band thickness to the specific settling rate
(Qd/A) as log(ΔH) = log k + ylog(Qd/A), where the value of y
is typically in the range from 1 to 7 depending on the system.
The values of k and y were found to change with the phase
ratio. However, the effects of the organic (dispersed)-to-
aqueous (continuous) phase density ratio (ρo/ρa), baffle
opening position (Hb/Ht), and organic-to-aqueous phase ratio
at the inlet (αo/αa) are not considered in the available
correlations. Also, it is necessary to include all of these
parameters to develop a correlation that predicts the
Figure 17. Computational snapshots showing the separation of the dispersion-band thickness for a wide range of flow rates, baffle
organic−aqueous phases and the dispersion-band thickness at organic- positions, phase properties, and phase ratios at the inlet and to
to-aqueous ratios (αo,inlet/αa,inlet) of (a) 0.5:0.5 (b) 0.4:0.6, and (c)
0.2:0.8 (Qt = 400 L/h, L = Lref, dd = 100 μm).
verify the predictive capability of the resulting correlation for a
large number of experimental and simulation data sets.
Using the present simulated data, the following functional
form was developed in this work

⎛ Q d ⎞ ⎛ ρo ⎞ 2 ⎛ Hb ⎞ 3⎛ αo ⎞
C C
H = C1⎜ ⎜ ⎟
⎟ + ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ + C4
⎝ A ⎠ ⎝ ρa ⎠ ⎝ Ht ⎠ ⎝ αa ⎠ (12)
Using the simulation data sets, the best values of the constants
were found to be C1 = 121.3, C2 = 6.5, C3 = 3.5, and C4 =
−0.021 (with all of the parameters expressed in SI units).
Figure 19 shows a comparison of the simulated (using CFD)
and predicted (using eq 12) dispersion-band thicknesses for
different ranges of conditions, that is, for different values of
(ρo/ρa), (Hb/Ht), (αo/αa), and (Qd/A). The correlation gives
satisfactory predictions for the above-mentioned parameters.
Finally, a parity plot (see Figure 20) of dispersion-band
thickness was obtained using the simulated and predicted (eq
12) data points. The errors in the parity plot were found to be
within ±14%.
Figure 18. Effects of αo,inlet/αa,inlet on the dispersion-band thickness (Qt
= 400 L/h, L = Lref, dd = 100 μm). 4. CONCLUSIONS
In the present work, the effects of settler geometry (settling
performing simulations for (i) αo,inlet = 0.4, αa,inlet = 0.6, and (ii) area, location of the aqueous-phase outlet, position of the end
αo,inlet = 0.2, αa,inlet = 0.8. The steady-state distributions of the plate) on the liquid−liquid separation in a continuous gravity
organic-phase volume fraction and the corresponding positions settler were investigated through numerical simulations. An
13941 DOI: 10.1021/acs.iecr.7b03756
Ind. Eng. Chem. Res. 2017, 56, 13929−13944
Industrial & Engineering Chemistry Research Article

Figure 19. Variations in the dispersion-band thickness with varying (a) baffle opening, (b) organic-phase density, (c) organic-to-aqueous phase ratio
at the inlet, and (d) specific settling area: simulated results versus predictions based on eq 12.

measurements of dispersion-band thickness and organic-phase


volume fraction performed by Thaker et al.28 for different
settler lengths.
For a given flow rate of the dispersion, an increase in the
settler length (for a fixed width) led to an increased settling
area and also to an increased residence time. This led to a
reduction of the dispersion-band thickness. For a constant
settling area, an increase in the total flow rate leads to a
decrease of the residence time, and this, in turn, leads to an
increase in the dispersion-band thickness or flooding inside the
setter.
For settlers with length-to-width ratios of L/W > 1.5, the
insertion of a picket fence did not improve the separation
performance. However, for smaller values of L/W, the insertion
of a picket fence was found to lead to a reduction in the
Figure 20. Comparison of simulated and predicted (eq 12) values of dispersion-band thickness. Both the end-plate height and the
the dispersion-band thickness. location of the aqueous-phase outlet were found to influence
the phase separation significantly. A reduction in the end-plate
Eulerian multifluid model developed using the open-source height led the dispersion to overflow and to exit from the
CFD solver OpenFOAM, that has been validated experimen- organic outlet, thus leading to a poor separation performance.
tally,33 was used to perform the numerical simulations. The When the height of the aqueous outlet was decreased beyond a
computational model was further validated by comparing the certain level, no phase separation was achieved. For a fixed Qt
13942 DOI: 10.1021/acs.iecr.7b03756
Ind. Eng. Chem. Res. 2017, 56, 13929−13944
Industrial & Engineering Chemistry Research Article

value and settler configuration, with increasing inlet aqueous- Vt = settling velocity, m/s
phase volume fraction, the dispersion-band thickness was found Greek Letters
to decrease. A new correlation was developed in the present
α = phase hold-up/volume fraction
work for the prediction of the dispersion-band thickness that
μ = dynamic viscosity, N s/m2
successfully predicts the effects of the dispersion flow rate,
ρ = fluid density, kg/m3
settling area, inlet baffle opening position, inlet phase ratio, and
σ = interfacial tension, mN/m
density ratio. The results reported and the correlation
τ ̿ = stress tensor, Pa
developed in the present work will be useful for the design of
optimal settler configurations. Subscript/Superscript
The drag-coefficient correction factor that was introduced to a = aqueous phase
account for effects of neighboring drops led to improvements in g = gas
the predictions of the dispersion-band thickness and dispersed- o = organic phase


phase volume fraction. These predictions can be further
improved using the population balance model incorporated REFERENCES
within the CFD model that accounts for binary coalescence.
(1) Moris, M. A.; Diez, F. V.; Coca, J. Hydrodynamics of a Rotating
This also necessitates the implementation of models for
Disc Contactor. Sep. Purif. Technol. 1997, 11, 79.
interfacial coalescence at the active interface, and efforts are (2) Wang, Y.; Fei, W.; Sun, J.; Wan, Y. Hydrodynamics and Mass
underway to develop such comprehensive CFD models using Transfer Performance of a Modified Rotating Disc Contactor
OpenFOAM.


(MRDC). Chem. Eng. Res. Des. 2002, 80, 392.
(3) Venkatanarasaiah, D.; Varma, Y. Dispersed Phase Holdup and
AUTHOR INFORMATION Mass Transfer in Liquid Pulsed Column. Bioprocess Eng. 1998, 18, 119.
Corresponding Author (4) Wardle, K. E. Open-source CFD Simulations of Liquid−Liquid
*Tel.: +91 11 2659 1027. Fax: +91 11 2658 1120. E-mail: Flow in the Annular Centrifugal Contactor. Sep. Sci. Technol. 2011, 46,
2409.
vvbuwa@iitd.ac.in.
(5) Jeelani, S. A. K.; Hartland, S. Prediction of Steady State
ORCID Dispersion Height from Batch Settling Data. AIChE J. 1985, 31, 711.
Vivek V. Buwa: 0000-0003-2335-6379 (6) Nadiv, C.; Semiat, R. Batch Settling of Liquid-liquid dispersion.
Notes Ind. Eng. Chem. Res. 1995, 34, 2427.
(7) Henschke, M.; Schlieper, L. H.; Pfennig, A. Determination of a
The authors declare no competing financial interest.


Coalescence Parameter from Batch-Settling Experiments. Chem. Eng. J.
2002, 85, 369.
ACKNOWLEDGMENTS (8) Yu, G. Z.; Mao, Z. S. Sedimentation and Coalescence Profiles in
The authors acknowledge scientific discussions with Mr. Liquid-Liquid Batch Settling Experiments. Chem. Eng. Technol. 2004,
Abhijeet H. Thaker (IIT Delhi) and Dr. K. K. Singh (BARC 27, 407.
Mumbai). The suggestions provided by Mr. Thaker on the (9) Ruiz, M.; Padilla, R. Determination of Coalescence Functions in
development of the correlation for the prediction of dispersion- Liquid−Liquid Dispersions. Hydrometallurgy 2005, 80, 32.
band thickness are gratefully acknowledged. Both authors (10) Hartland, S.; Jeelani, S. A. K. Prediction of Sedimentation and
gratefully acknowledge the financial support provided by the Coalescence Profiles in a Decaying Batch Dispersion. Chem. Eng. Sci.
1988, 43, 2421.
Government of India (Project 2013/36/01-BRNS/0579).


(11) Jeelani, S.; Hartland, S. Dynamic Response of Gravity Settlers to
Changes in Dispersion Throughput. AIChE J. 1988, 34, 335.
NOMENCLATURE (12) Jeelani, S.; Hartland, S. Effect of Dispersion Properties on the
A = settling area, m2 Separation of Batch Liquid-Liquid Dispersions. Ind. Eng. Chem. Res.
ACS = cross-sectional area of the settler, m2 1998, 37 (2), 547.
CD = interfacial drag coefficient (13) Wang, F.; Mao, Z.-S. Numerical and Experimental Investigation
CDO = corrected interfacial drag coefficient of Liquid-Liquid Two-Phase Flow in Stirred Tanks. Ind. Eng. Chem.
dd = diameter of drop, μm Res. 2005, 44, 5776.
(14) Laurenzi, F.; Coroneo, M.; Montante, G.; Paglianti, A.; Magelli,
Eö = Eötvös number
F. Experimental and Computational Analysis of Immiscible Liquid−
g = gravitational constant, m/s2 Liquid Dispersions in Stirred Vessels. Chem. Eng. Res. Des. 2009, 87,
H = dispersion-band thickness, m 507.
Hb = baffle opening height, cm (15) Cheng, D.; Cheng, J.; Yong, Y.; Yang, C.; Mao, Z. S. CFD
Ht = height of the end plate, cm Prediction of the Critical Agitation Speed for Complete Dispersion in
K = liquid−liquid momentum exchange coefficient, kg/(m3 Liquid-Liquid Stirred Reactors. Chem. Eng. Technol. 2011, 34, 2005.
s) (16) Yadav, R. L.; Patwardhan, A. W. CFD Modeling of Sieve and
L = settling length, m Pulsed-Sieve Plate Extraction Columns. Chem. Eng. Res. Des. 2009, 87,
M⃗ = interphase momentum transfer, N/m3 25.
P = pressure, N/m2 (17) Din, G. U.; Chughtai, I. R.; Inayat, M. H.; Khan, I. H.; Qazi, N.
p = drag correction factor K. Modeling of a Two-phase Countercurrent Pulsed Sieve Plate
Extraction Column- A Hybrid CFD and Radiotracer RTD Analysis
Q = flow rate, L/h
Approach. Sep. Purif. Technol. 2010, 73, 302.
Qt = total flow rate, L/h (18) Rieger, R.; Weiß, C.; Wigley, G.; Bart, H.-J.; Marr, R.
Qd = dispersed-phase flow rate, L/h Investigating the Process of Liquid-Liquid Extraction by Means of
Re = Reynolds number Computational Fluid Dynamics. Comput. Chem. Eng. 1996, 20, 1467.
t = time, s (19) Drumm, C.; Bart, H. J. Hydrodynamics in a RDC Extractor:
tr = residence time, s Single and Two-Phase PIV Measurements and CFD Simulations.
V = phase velocity, m/s Chem. Eng. Technol. 2006, 29, 1297.

13943 DOI: 10.1021/acs.iecr.7b03756


Ind. Eng. Chem. Res. 2017, 56, 13929−13944
Industrial & Engineering Chemistry Research Article

(20) Ghaniyari-Benis, S.; Hedayat, N.; Ziyari, A.; Kazemzadeh, M.;


Shafiee, M. Three-Dimensional Simulation of Hydrodynamics in a
Rotating Disc Contactor using Computational Fluid Dynamics. Chem.
Eng. Technol. 2009, 32, 93.
(21) Li, S.; Duan, W.; Chen, J.; Wang, J. CFD Simulation of Gas−
Liquid−Liquid Three-Phase Flow in an Annular Centrifugal
Contactor. Ind. Eng. Chem. Res. 2012, 51, 11245.
(22) Horng, J. S.; Maa, J. R. The Effect of a Surfactant on Mixer
Settler Operation. J. Chem. Technol. Biotechnol. 1986, 36, 15.
(23) Eckert, N.; Gormely, L. Phase Separation in an Experimental
Mixer-Settler. Chem. Eng. Res. Des. 1989, 67, 175.
(24) Jeelani, S.; Hartland, S. The Continuous Separation of Liquid/
Liquid Dispersions. Chem. Eng. Sci. 1993, 48, 239.
(25) Lee, H. Y.; Oh, J. K.; Lee, D. H. Interpretation of Continuous
Settling Behavior from Batch Settling Data in a Versatic Acid 10-Water
System. Hydrometallurgy 1993, 32 (3), 273.
(26) Padilla, R.; Ruiz, M.; Trujillo, W. Separation of Liquid-Liquid
Dispersions In a Deep-Layer Gravity Settler: Part I. Experimental
Study of the Separation Process. Hydrometallurgy 1996, 42, 267.
(27) Gharehbagh, F. S.; Mousavian, S. M. A. Hydrodynamic
Characterization of Mixer-Settlers. J. Taiwan Inst. Chem. Eng. 2009,
40, 302.
(28) Thaker, A. H.; Singh, K. K.; Buwa, V. V. Experimental
Investigations of Liquid−Liquid Disengagement in a Continuous Gravity
Settler; Internal Report; Indian Institute of Technology Delhi: New
Delhi, 2017.
(29) Reeve, R.; Godfrey, J. Phase Inversion During Liquid−Liquid
Mixing in Continuous Flow, Pump−Mix, Agitated Tanks. Chem. Eng.
Res. Des. 2002, 80, 864.
(30) Hadjiev, D.; Paulo, J. Extraction Separation in Mixer−Settlers
Based on Phase Inversion. Sep. Purif. Technol. 2005, 43, 257.
(31) Paulo, J. B.; Hadjiev, D. E. Mixer-Settler Based on Phase
Inversion: Design of the Mixing Zone. Ind. Eng. Chem. Res. 2006, 45,
3821.
(32) Ngan, K. H.; Ioannou, K.; Rhyne, L. D.; Wang, W.; Angeli, P. A
Methodology for Predicting Phase Inversion During Liquid−Liquid
Dispersed Pipeline Flow. Chem. Eng. Res. Des. 2009, 87, 318.
(33) Panda, S. K.; Singh, K.; Shenoy, K.; Buwa, V. V. Numerical
Simulations of Liquid-Liquid Flow in a Continuous Gravity Settler
using OpenFOAM and Experimental Verification. Chem. Eng. J. 2017,
310, 120.
(34) Tsuchiya, K.; Furumoto, A.; Fan, L.-S.; Zhang, J. Suspension
Viscosity and Bubble Rise Velocity in Liquid-Solid Fluidized Beds.
Chem. Eng. Sci. 1997, 52, 3053.
(35) Ishii, M.; Zuber, N. Drag Coefficient and Relative Velocity in
Bubbly, Droplet or Particulate Flows. AIChE J. 1979, 25, 843.
(36) Ryon, A.; Daley, F.; Lowrie, R. Design and Scaleup of Mixer-
Settlers for the DAPEX Solvent Extraction Process; Oak Ridge National
Laboratory; Oak Ridge, TN, 1960.

13944 DOI: 10.1021/acs.iecr.7b03756


Ind. Eng. Chem. Res. 2017, 56, 13929−13944

You might also like