Vous êtes sur la page 1sur 11

w a t e r r e s e a r c h 6 8 ( 2 0 1 5 ) 2 6 2 e2 7 2

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.elsevier.com/locate/watres

A model of particle removal in a dissolved air


flotation tank: Importance of stratified flow and
bubble size

B. Lakghomi a,*, Y. Lawryshyn b, R. Hofmann a


a
Department of Civil Engineering, University of Toronto, 35 St. George Street, Toronto, ON M5S 1A4, Canada
b
Department of Chemical Engineering, University of Toronto, 200 College Street, Toronto, ON M5S 3E5, Canada

article info abstract

Article history: An analytical model and a computational fluid dynamic model of particle removal in
Received 21 April 2014 dissolved air flotation were developed that included the effects of stratified flow and
Received in revised form bubble-particle clustering. The models were applied to study the effect of operating con-
20 August 2014 ditions and formation of stratified flow on particle removal. Both modeling approaches
Accepted 21 September 2014 demonstrated that the presence of stratified flow enhanced particle removal in the tank. A
Available online 13 October 2014 higher air fraction was shown to be needed at higher loading rates to achieve the same
removal efficiency. The model predictions showed that an optimum bubble size was pre-
Keywords: sent that increased with an increase in particle size.
Dissolved air flotation (DAF) Crown Copyright 2014 Published by Elsevier Ltd. All rights reserved.
Particle removal
CFD
Stratified flow
Bubbles
Multiphase flow

Edzwald, 2004; Edzwald, 2007). Stratified flow, shown in the


1. Introduction experimental pilot-scale measurements of Lundh et al. (2001),
is defined as water traveling in a horizontal layer from the top
Dissolved air flotation (DAF) systems for drinking water of the inlet baffle towards the far wall of the separation zone,
treatment are currently being operated at surface loading and then traveling back to the baffle in a horizontal second
rates as high as 20e40 m/h (Edzwald, 2007). These rates are layer below the first (Fig. 1). Lakghomi et al. (2012) investigated
higher than those predicted to be possible by simple vertical the effects of stratified flow on bubble removal using a
plug-flow hydrodynamic models, demonstrating that more computational fluid dynamics (CFD) model of two-phase flow
complex DAF models are needed if they are to be used as (air and water, with no particles). By including the effect of
optimization tools. Some researchers have attributed the bubble aggregation for the first time in a CFD model of DAF,
discrepancy between the current high surface loading rates the model was able to identify conditions under which strat-
and the earlier model predictions to the presence of stratified ified flow occurs, and to quantitatively predict the benefit of
flow in the separation zone of a DAF tank (Haarhoff and the stratified flow in terms of better bubble removal. The

* Corresponding author. Tel.: 1 416 978 5974.


E-mail addresses: b.lakghomi@utoronto.ca, lakghomi@gmail.com (B. Lakghomi).
http://dx.doi.org/10.1016/j.watres.2014.09.053
0043-1354/Crown Copyright 2014 Published by Elsevier Ltd. All rights reserved.
w a t e r r e s e a r c h 6 8 ( 2 0 1 5 ) 2 6 2 e2 7 2 263

DAF, models solely based on turbulent mixing are likely not


able to fully represent bubble-particle attachment in DAF
(Edzwald, 2010). In addition, it was assumed that bubble-
particle aggregates included a single particle and a number
of bubbles attached to it, neglecting the attachment between
aggregates containing multiple particles, referred to as clus-
tering, that is described to have a significant effect on particle
removal (Leppinen and Dalziel, 2004). CFD has also been pre-
viously applied to include the effects of hydrodynamics on
particle removal, but it has been based on simplifying as-
sumptions to reduce the numerical cost. Kostoglou et al. (2007)
developed a three-phase CFD model of DAF by incorporating
the flotation kinetics developed by Koh and Schwarz (2003),
however, their model assumed that solid particles follow the
liquid flow path and are completely removed from the domain
as they attach to the bubbles, neglecting the effect of particles
and bubble-particle aggregates on the flow pattern. The effect
of particles and aggregates on the flow pattern can be signif-
icant and needs to be included at a reasonable computational
cost.
In this study, the effect of stratified flow, bubble size, and
Fig. 1 e The conceptual model for stratified flow in the
air fraction on particle removal in a DAF system is evaluated
separation zone.
using two different approaches: analytical and CFD models.
The developed approaches account for bubble-particle
attachment and clustering in both contact and separation
zones. The analytical approach assumes a simplified flow
results of the model were in good agreement with the exper- pattern in the tank, i.e., stratified or non-stratified flow, and
imental measurements of Lundh et al. (2001). The inclusion of can evaluate the effect of input parameters independently
only two phases (air and water), however, was a limitation in and at essentially no computational cost. The CFD model, on
the model, as it could not be used to directly extrapolate the the other hand, provides a more realistic representation of the
impact of the stratified flow on particle removaldwhich is the flow, but has a significant computational cost. First, the
ultimate goal of DAF treatment. While it may be intuitively analytical model is used to show the theoretical effect of
predicted that the same trends in particle removal would be stratified flow on particle removal under simplified flow con-
observed as for bubble removal given that particles are ditions. Then, the CFD model is applied, to study the effect of
removed from DAF systems by being attached to bubbles, bubble size, and air fraction on bubble and particle removal
particle removal will also be affected by bubble-particle and to verify if the presence of stratified flow will promote
attachment in the contact and separation zones. In addition, particle removal under more realistic conditions. It should be
the introduction of particles to the system and the formation emphasized that both the analytical and CFD models are ex-
of bubble-particle aggregates could conceivably change flow tensions of the model of Lakghomi et al. (2012) where, now,
behavior enough to warrant an independent investigation. the effect of particles and bubble-particle aggregation is also
Therefore, one objective of this work is to build a model that implemented. The physics of the aggregation is mainly based
accounts for bubble-particle attachment, and includes air, on the previous works of Kostoglou et al. (2007) and Rollie et al.
water, particle and aggregate phases. (2009). It was not within the scope of this research to validate
This is not the first modeling attempt to include the fate of the models at pilot- or full-scale. Instead, the aim was to
particles in a DAF system. Both analytical approaches and CFD provide a better theoretical understanding of the effect of
have been applied in the past for this purpose, but most of the stratified flow and key parameters in a general sense. Such
studies have focused on analytical models based on the plug understanding can help to direct future pilot- or full-scale
flow assumption in the contact and separation zones and studies to validate the developed theory.
have not considered the effects of stratified flow and bubble-
particle attachment in the separation zone (Haarhoff and
Edzwald, 2004). Moreover, most of these models have 2. Methodology
focused on the efficiency of the contact zone (Edzwald et al.,
1991; Edzwald, 1995; Matsui et al., 1998) and separation zone 2.1. An analytical hetero-aggregation model for particle
(Haarhoff and Edzwald, 2001) independently, and have not removal
calculated overall particle removal efficiency in the system.
Leppinen et al. (2001) was the first to calculate the overall Lakghomi et al. (2012) applied an analytical flow model in the
particle removal efficiency of DAF by adopting the population separation zone to explain higher bubble removal in the
balance model of Matsui et al. (1998), however, the bubble- presence of stratified flow, in a two-phase (air and water)
particle attachment in their model was based only on a tur- system. In the current study, particles and aggregates are
bulent mixing mechanism. Given the low turbulence levels in added to the previous model and bubble-particle attachment
264 w a t e r r e s e a r c h 6 8 ( 2 0 1 5 ) 2 6 2 e2 7 2

is included to confirm, theoretically, whether the same outgoing aggregate numbers in a group. The incoming parti-
beneficial impact of stratified flow exists for particle removal, cles can be formed from aggregation of a variety of different
as was previously demonstrated for bubble removal. sources. For example, a product aggregate N3, 4 with i 3
In previous bubble-particle attachment models, it has been bubbles and j 4 particles can result from different aggrega-
assumed that bubbles attach to a single particle until the tion scenarios such as from N2, 4 and N1, 0, or N2, 2 and N1, 2.
particle is saturated with bubbles and a population balance The attachment frequency can be calculated by multi-
model has been reported for particles based on this assump- plying collision frequency, K[i, j, m, l], with attachment effi-
tion (Matsui et al., 1998; Leppinen et al., 2001). The effect of ciency, a[i, j, m, l]:
clustering (formation of aggregates with multiple bubbles and
particles) was not considered. In addition, the presence of ki; j; m; l Ki; j; m; lai; j; m; l: (4)
back and forth horizontal layers and bubble-particle attach- The collision frequency is calculated by extending the
ment in the separation zone were not included. In this study, model of Kostoglou et al. (2007), based on the assumption of
the analytical model was based on a population balance spherical aggregates, as the linear addition of the collision
model that allows clustering of multiple particles and bubbles frequencies from buoyancy and gravity (KBG), and turbulence
into aggregates in the contact and separation zones. The (KT):
change in the concentration of each group of aggregates was
calculated based on their attachment rate from turbulence, K KBG KT (5)
buoyancy and gravity mechanisms using the population bal- The collision frequency due to buoyancy and gravity can be
ance model. The concentration of different aggregate groups calculated as follows.
leaving the contact zone was then applied as the concentra-
p h i  2
tion of aggregates entering the separation zone. A conceptual KBG p i; j; m; l vij  vml  dij dml (6)
4
flow model was applied assuming uniform plug flow in the
contact zone, and two horizontal stratified flow layers fol where dij is the diameter of an aggregate with i bubbles and j
lowed by vertical plug flow in the separation zone (Fig. 1). The particles, and vij is the aggregate rise velocity. The term p[i, j,
average velocity in the contact zone and in the horizontal m, l] is the collision efficiency and can be calculated from the
layers is obtained as follows: model of Nguyen (1999) based on the flow around a spherical
bubble (aggregate in this case):
Q
Vc vc (1)
wLc   2
4 dml
pi; j; m; l 1:5 Re0:72 ; (7)
15 d2ij
Q
Uu (2) where Re is the aggregate Reynolds number.
wH
The turbulent collision frequency, KT, can be calculated
where Q is the flow rate, w is the tank width, Lc is the contact
based on the SaffmaneTurner relation (Saffman and Turner,
zone length, and H is the thickness of the horizontal layer.
1956):
The change in the concentration of aggregates was calcu-
lated based on the population balance in each zone depending " # 1=2
3  3
on the removal of the aggregates and their attachment to KT pp i; j; m; l dij dml ; (8)
2 15m
other aggregates.
The applied bivariate population balance model was where is the turbulent dissipation rate, and m is the water
similar to the model used by Rollie et al. (2009) for binary viscosity. The turbulent dissipation rate is estimated based on
particle mixtures to allow hetero-aggregation of bubbles and the k-epsilon model at the inlet (Launder and Spalding, 1974),
particles and formation of aggregates with different sizes. The
general form of the population balance equation is: 0:1643k1:5
; (9)
l
dNi;j imaxm X
X jmaxl where l 0.07 Lc, k 3=2ul2 , and Lc is the length of the inlet
 ki; j; m; lNi;j Nm;l
dt m0 l0
(contact zone).
(3)
The term a[i, j, m, l] is the attachment efficiency between
1X i X j
km; l; i  m; j  lNm;l Nim;jl two aggregates and depends on the balance of the microscopic
2 m0 l0
forces between the aggregates. The parameter a represents
where Ni,j is the number concentration (the number of ag- the number of attachments per number of collisions, and
gregates per unit volume) of an aggregate with i bubbles and j depends on the coagulation conditions and water character-
particles, k[i, j, m, l] is the attachment frequency between ag- istics. Haarhoff and Edzwald (2004) reported that under
gregates Ni, j and Nm, l, and is and jmax are the upper limit of favorable coagulation conditions, a values of 0.5 and 1 be-
the number of bubbles and particles forming an aggregate. tween a bubble and a particle provided fairly reasonable
The first double sum on the right side is the sink term repre- agreement with the available experimental data. Given that
senting the reduction in the number concentration of aggre- the main purpose of this research was to investigate the effect
gates Ni ,j as a result of attachment to other aggregates, and of different physical parameters on the removal efficiency, the
the second term is the source term representing the formation study did not focus on the effect of physicochemical factors on
rate of aggregate Ni, j by attachment of smaller aggregates. the attachment efficiency, and a constant attachment effi-
Aggregations between groups will lead to incoming and ciency of 1 was assumed between any two aggregates.
w a t e r r e s e a r c h 6 8 ( 2 0 1 5 ) 2 6 2 e2 7 2 265

v
First, the population balance was written for the contact layer from the lower layer will be Hij N*i;j . If dt is replaced by dx
U in
zone as. the population balance, the total rate of change in the aggre-
gate number concentration in the first layer can be written as.
dNi;j imaxm X
X jmaxl
vij  ki; j; m; lNi;j Nm;l X
X jmaxl
dy dNi;j imaxm
m0 l0
(10) U ki;j;m;lNi;j Nm;l
dx
1X i X j m0 l0
km; l; i  m; j  lNm;l Nim;jl (16)
1X i Xj
2 m0 l0 vij vij
km;l;im;jNm;l Nim;jl  Ni;j Ni;j ;
2 m0 l0 H H
where vij is the aggregate rise velocity for each group of ag-
gregates that can be calculated based on the aggregate size, dij, where N*i;j is the number concentration of aggregate in the
and density, rij, estimated similar to the model of Edzwald second horizontal layer.
et al. (1991): For the second horizontal layer the population balance
 1=3 equation will be similar, with the difference that the direction
dij id3p jd3b ; (11) of the horizontal flow is reversed, so U will have a negative
sign. In addition, the term for flow of the aggregates from the
lower layer will be omitted. Therefore, the population balance
irp d3p jrb d3b equations in the second layer can be written as:
rij ; (12)
id3p jd3b

where i and j are the number of attached bubbles and parti- dNi;j X jmaxl
imaxm X
U  km; l; i; jNi;j Nm;l
cles, and dp and db are the particle and bubble diameters. dx m0 l0
(17)
At the inlet (y 0), N1,0 and N0,1 are equal to the inlet 1X i X j
vij
concentration of bubbles and particles, calculated as: km; l; i  m; j  lNm;l Nim;jl  Ni;j :
2 m0 l0 H
6fb Assuming that all of the aggregates that are not removed
N1;0 in Nb0 ; (13)
pd3b
from the first layer will be uniformly mixed at the beginning of
the second layerN*i;j Ni;j xLs ; the above coupled differential
6fp equations can be converted to a system of algebraic equations
N1;0 in Np0 ; (14)
pd3p by dividing the horizontal layers to numerical grids from x 0
where fb and fp are bubble and particle volume fractions, and to x Ls, and solved to obtain the number of particles leaving
db and dp are the bubble and particle diameters. the system.
Equation (10) can be integrated numerically from y 0 to
y h to obtain the concentration of aggregates leaving the
contact zone and entering the separation zone 2.2. CFD model
 
(Ni;j yh contact zone Ni;j x0 separation zone. Aggregates
that enter the top horizontal layer in the separation zone can The two-dimensional two-phase (air/water) CFD model
be removed from the top of the tank. Similar to bubbles, the developed by Lakghomi et al. (2012) was extended to include
removal ratio of aggregates from the top layer will depend on particles and bubble-particle aggregates, allowing bubble-
their rise velocity. The removal efficiency of aggregate group particle attachment to be modeled. Water, bubbles and ag-
Ni,j from the top layer (with length Ls and thickness of H) can gregates were modeled as three different phases using the
be obtained similar to bubble removal in Lakghomi et al.
(2012):

     
Ls vij Ls vij
Aggregate removal efficiency
U H Q=wH H
 
Lw vij
vij ;
Q Vs
(15)

where U is the horizontal velocity, vij is the aggregate rise


velocity, w is the tank width, Q is the flow rate and Vs is the
surface loading rate.
Based on the conceptual flow model (Fig. 1), the population
balance for aggregates in the first horizontal layer can be
written similar to Equation (3), but two additional terms need
to be added for the aggregates that leave through the top
surface, or enter from the layer underneath. The fraction of
aggregates (group ij) removed from the top in an element with
vij
the width dx can be calculated as dx U H by replacing L with dx in
Equation (15). As a result, the rate of removal from the top
v
layer will be Hij Ni;j , and the rate of aggregates entering the top Fig. 2 e Configuration of the modeled DAF system.
266 w a t e r r e s e a r c h 6 8 ( 2 0 1 5 ) 2 6 2 e2 7 2

Fig. 3 e The effect of air fraction and bubble size on particle removal calculated from the analytical model, a) in the absence
of stratified flow, b) in the presence of stratified flow. Particle size 10 mm, particle density 1100 kg/m3, particle volume
fraction 0.0001, loading rate 23.6 m/hr.

ANSYS FLUENT mixture model (ANSYS, 2012a), and particle The average air volume fraction and particle concentration
concentration was modeled as a user-defined scalar. were monitored at the pressure outlet over time and overall
The geometry of the simulated system can be observed in convergence was assumed when a quasi-steady-state
Fig. 2. The system was divided into 76,570 structured rectan- behavior was observed. At this point, the temporal fluctua-
gular elements using ANSYS ICEM (ANSYS, 2012b). The two tion of the velocity magnitude in the recirculation zones was
dimensional model did not include the recycle air/water in- observed to be less than 2% of its average value.
jection system, so the inlet position of the bubble-water The particle number concentration was modeled as a
mixture was defined as the bottom of the contact zone to passive scalar to reduce the computational demand. The flux
minimize deviations from plug flow. The inlet boundary of a scalar (particle number concentration) into or out of a
condition was modeled as a velocity inlet. The collection control volume can occur based on convection or diffusion
system was modeled as a uniform pressure outlet with at- mechanisms. In this case, particles were converted to aggre-
mospheric pressure. The top surface was assumed to be a gates following attachment to bubbles. As a result, the popu-
symmetry boundary condition (zero shear stress) where the lation balance equation for the solid particles, including
removal of the air and aggregate phases at a rate equal to their convection, turbulent diffusion, and bubble-particle attach-
flux across the water surface was modeled via a user defined ment, can be written as follows:
function (Lakghomi et al., 2012).
 

The standard k-epsilon model was employed to account for V$ ! u !


u p np Dt Vnp knp nb (18)
turbulence. A transient solver was used, as the flow in the
separation zone exhibited temporal behavior. Model conver- where ! u is the mixture velocity,! u p is the particle settling
gence for each time step was assumed to be achieved when velocity and ! u !u p represents particle total velocity. Dt is
the normalized residuals decreased to less than 1  104. turbulent diffusion, k is the collision frequency (calculated
Convergence was also evaluated by checking the overall sys- similar to the theoretical model) and np and nb are particle and
tem mass balance for each phase to be less than 1  106 kg/s. bubble number concentrations. The focus of the present study
was to investigate the effect of hydrodynamic conditions on
particle removal. Therefore, the effect of different coagulation
scenarios on particle removal was not considered in this
context, and, similar to the analytical model, it was assumed
that all of the particles colliding with the bubbles formed
bubble-particle aggregates (a 1).
The bubble-particle aggregates were modeled as the ter-
tiary phase in the mixture model. Two mass source terms
were defined to calculate the exchange rate between the
bubble (secondary) and aggregate (tertiary) phases, as bubbles
are converted to aggregates according to the bubble-particle
attachment frequency based on Equations (5), (6) and (8).
The maximum number of bubbles that could attach to a par-
ticle was calculated as.

Fig. 4 e Loading rates at 60% particle removal in the  2


absence and presence of stratified flow. Bubble size 40 mm, dp
Nmax 4p ; (19)
particle size 10 mm, particle density 1100 kg/m3. db
w a t e r r e s e a r c h 6 8 ( 2 0 1 5 ) 2 6 2 e2 7 2 267

Fig. 5 e The effect of air fraction and bubble size on particle removal from the theoretical model, a) loading rate 11.8 m/hr, b)
loading rate 23.6 m/hr. Particle size 10 mm, particle density 1100 kg/m3, particle volume fraction 0.0001, in presence of
stratified flow.

assuming that each bubble occupies a db  db square on a separation zone of DAF systems leads to higher bubble
particle's surface area, modified by a packing factor of removal. Intuitively, one would predict that the same phe-
4 (Haarhoff and Edzwald, 2001). Two user-defined functions nomenon would be observed for a DAF system consisting of
were defined to update the aggregate phase size and density both bubbles and particles, but for the sake of completeness,
according to Equations (11) and (12). this paper aims to confirm this assumption. It also in-
Under the general mixing conditions in DAF systems, tur- vestigates whether different limiting conditions might apply
bulence is much smaller compared to mixed flotation and to particles compared to the bubbles. To this end, an analytical
flocculation systems that have an external mixer, and as a model was developed to predict, using first principles and in a
result, breakage (detachment) of bubble-particle aggregates is simplified DAF system, the impact of horizontal stratified flow
reportedly not significant compared to attachment rates on particle removal as a function of bubble size and air
(Kostoglou et al., 2007). As such, aggregate breakage was fraction.
neglected. The analytical model was first used to evaluate the effect of
The effect of solid particles on the flow pattern was stratified flow on particle removal. The model used a 10 mm
included through a gravity source term in the vertical mo- particle size, bubble sizes between 20 and 120 mm, along with
mentum mixture equation similar to the approach used by air fractions ranging from 0.005 to 0.02. These are all typical
Lakehal et al. (1999) for secondary clarifiers. This term repre- values for drinking water DAF systems (Edzwald, 2010). Fig. 3
sents buoyancy effects due to a change in mixture density as a shows a comparison of the model results for particle
result of the presence of particles. The effect of particles on removal in the absence and presence of stratified flow,
mixture viscosity was neglected. demonstrating an average increase of 52% in particle removal
Once the particle and bubble-particle aggregate distribu- for the stratified flow condition. With the addition of the
tion in the system was calculated, the total particle removal second horizontal layer, the chance for formation of aggre-
efficiency was calculated by subtracting the total rate of par- gates with larger rise velocity in the separation zone in-
ticles leaving the system (in the form of particles or bubble- creases, increasing particle removal efficiency. As a result, the
particle aggregates) from the total rate of particles entering analytical model confirms that the addition of a second hori-
the system, and dividing by the rate of particles entering the zontal layer enhances particle removaldsimilar to bubble
system: removal, which was previously reported by Lakghomi et al.
Z (2012).
  !
np;out npb;out !u $ n dAout The analytical model can also be used to predict the
Particle removal 1  Z ; (20)
 ! ! maximum limiting loading rate for a DAF system (to avoid
np;in u $ n dAin
particle washout) in the presence/absence of stratified flow.
where !
n is the unit direction vector. Fig. 4 shows loading rates at which 60% of particles (an arbi-
trarily selected value, for illustration) will be removed in the
absence and presence of stratified flow for a bubble size of
3. Results and discussions 40 mm and air fractions ranging from 0.005 to 0.02. The results
demonstrate that between 26 and 130% higher loading rates
3.1. Analytical results can be achieved in the presence of stratified flow depending
on the air fraction. It should be mentioned, however, that the
Previous work by Lakghomi et al. (2012) demonstrated that a presence of stratified flow itself depends on the loading rate,
horizontally-stratified back-and-forth flow pattern in the bubble size and air fraction as shown in Lakghomi et al. (2012),
268 w a t e r r e s e a r c h 6 8 ( 2 0 1 5 ) 2 6 2 e2 7 2

Fig. 6 e The effect of particle and bubble size on particle removal, a) loading rate 11.8 m/hr, air fraction 0.005, b) loading rate
11.8 m/hr, air fraction 0.01, c) loading rate 23.6 m/hr, air fraction 0.005, d) loading rate 23.6 m/hr, air fraction 0.01 (in presence
of stratified flow layer).

and that this comparison using the analytical model is only whereas an increase in air fraction is shown to always in-
made to conceptually show the impact of stratified flow. crease particle removal (Fig. 5). The optimum bubble size was
The analytical model was then used to explore the trends generally between 30 and 40 mm. The analytical model can be
in particle removal as a function of bubble size and air frac- used to explain the reasoning behind these observations. To
tion. Particle removal was modeled at air fractions and bubble be removed, particles would first need to attach to the bubbles
sizes similar to Fig. 3 at loading rates of 11.8 m/hr and 23.4 m/ and form aggregates in the contact zone, and then the ag-
hr. The modeling results demonstrated that there is an opti- gregates would need to rise to the surface and be removed at
mum bubble size for a given loading rate and air fraction, the top of the tank. The removal of particles, therefore, de-
below and above which particle removal deteriorates, pends on both contact and separation zone parameters: the

Fig. 7 e Velocity vectors (0e0.1 m/s) at different air fractions, a) 0.008, b) 0.01, c) 0.02. Loading rate 11.8 m/hr, bubble size
40 mm, particle size 10 mm, particle density 1100 kg/m3, particle volume fraction 0.0001.
w a t e r r e s e a r c h 6 8 ( 2 0 1 5 ) 2 6 2 e2 7 2 269

Fig. 8 e The velocity vectors (0e0.1 m/s) at different bubble sizes, a) 40 mm, b) 80 mm, c) 120 mm. Air fraction 0.008, loading
rate 11.8 m/hr, particle size 10 mm, particle density 1100 kg/m3, particle volume fraction 0.0001.

number of available bubbles for collision, contact time, colli- The results of the analytical model therefore confirmed
sion frequency, and the rise velocity of the formed aggregates. that, from a fundamental perspective, stratified flow allows
Bubbles that are too small are calculated to have lower colli- for higher surface loading rates to remove particles. The
sion frequencies with particles in the contact zone (Equations model also provided a physical explanation for the existence
(5)e(8)) and when they do collide, their rise velocities are small of optimum bubble sizes and particle sizes. While the
and they are therefore washed downwards and out in the analytical model allows for a good qualitative understanding
separation zone. When bubbles become too large, there are of such phenomena, it does not allow for accurate quantita-
fewer bubbles available for a given air fraction in the contact tive predictions to be made about DAF performance, since the
zone (Equation (13)), and bubbles that are too large escape to flow pattern is dictated and fixed within the model. In reality,
the top of the DAF system too quickly and do not have enough the flow pattern will be variable and will be influenced by
contact time with the particles to form bubble-particle bubble and particle characteristics. CFD modeling was there-
aggregates. fore undertaken to explore the same phenomena but under
The analytical model was then used to predict the impact presumably more realistic conditions.
of particle size on particle removal. As shown in Fig. 6, the
optimum particle size varied as a function of bubble size. In
3.2. Numerical (CFD) results
general, for bubble sizes of 40e60 mm that are typically present
in DAF, an optimum particle size of 25e50 mm was obtained.
A previous two-phase (air/water) CFD model by Lakghomi
Smaller bubble sizes led to smaller optimum particle sizes,
et al. (2012) suggested that an increase in air fraction can
while lager bubble sizes led to larger optimum particle sizes.
help to promote flow stratification and therefore enhance
The analytical model demonstrated that as bubble sizes are
bubble removal. The current CFD model builds on this previ-
small, the light weight of a small particle in terms of the
ous model by adding particle and aggregate phases, and is
resulting rise velocity of the bubble-particle aggregate and
used to evaluate the effect of air fraction and bubble size on
availability of more particles for collision are more important
the formation of stratified flow and particle removal in the
than the smaller bubble-particle collision frequency when
presence of particles and aggregates.
both bubbles and particles are small. Conversely, when bubble
The CFD model was first used to explore the effect of air
sizes are large, they have a larger collision frequency with
fraction and bubble size on formation of stratified flow in the
larger particles and can lift the weight of such larger particles,
presence of particles. Fig. 7 shows the flow pattern at different
leading to higher overall removal efficiency despite the
air fractions for a loading rate of 11.8 m/hr, bubble size of
availability of fewer bubbles and particles.
40 mm, particle size of 10 mm, and particle density of 1100 kg/
m3. The results show that the flow pattern moved from a
short-circuiting regime toward stratified flow with an increase
Table 1 e Particle removal efficiency for varying air in air fraction from 0.008 to 0.02. For the same loading rate and
fractions and bubble sizes. Loading rate 11.8 m/hr, at an air fraction of 0.008, an increase in bubble size from
particle size 10 mm, particle density 1100 kg/m3, particle 40 mm to 80 mm led to the flow pattern becoming closer to
volume fraction 0.0001. stratified, whereas a further increase of bubble size to 120 mm
Bubble size/ 0.005 0.008 0.01 0.02 caused disruption to the stratified flow pattern (Fig. 8). The
Air fraction CFD model demonstrated that when bubbles are too large, the
20 mm 0.08 0.10 0.27 0.35 bubbles and aggregates are removed quickly from the top of
40 mm 0.33 0.36 0.38 0.57 the contact zone and do not allow enough bubble-particle
80 mm 0.55 0.70 0.40 0.22 contact in the contact zone. Moreover, the large bubbles do
120 mm 0.23 0.28 0.21 0.19
not penetrate into the separation zone where they can help to
270 w a t e r r e s e a r c h 6 8 ( 2 0 1 5 ) 2 6 2 e2 7 2

Fig. 9 e The effect of air fraction and loading rate on particle removal from the CFD model, a) bubble size 40 mm, b) bubble
size 80 mm. Particle size 10 mm, particle density 1100 kg/m3, particle volume fraction 0.0001.

form stratified flow (bubbles are needed to create stratified summarized in Table 1. For a bubble size of 40 mm, for
flow since the stratified flow is caused by flow density differ- example, the creation of stratified flow due to an increase in
ences, as explained in Lundh et al. (2001)). In general, there- air fraction from 0.01 to 0.02 (Fig. 7) led to a 50% improvement
fore, an increase in bubble size and air fraction up to a certain in particle removal efficiency (Table 1).
point promotes stratification of the flow, whereas an increase The CFD model can be applied to find the required air
beyond this bubble size and air fraction can lead to the quick fraction to achieve a target particle removal with an increase
removal of all of the bubbles from top of the tank, thereby in loading rate. Contours of particle removal efficiency at
preventing the formation of stratified flow in the separation different air fractions and loading rates are shown in Fig. 9 for
zone. two bubble sizes. For both cases, with an increase in the
The CFD model can also be applied to show that the pres- loading rate, a higher air fraction is required to achieve the
ence of stratified flow can enhance particle removal, as same particle removal. At a higher loading rate, a higher air

Fig. 10 e The effect of air fraction and bubble size on particle and bubble removal from the CFD model, a) bubble removal,
loading rate 11.8 m/hr, b) bubble removal 23.6 m/hr, c) particle removal 11.8 m/hr, d) particle removal 23.6 m/hr. Particle size
10 mm, particle density 1100 kg/m3, particle volume fraction 0.0001.
w a t e r r e s e a r c h 6 8 ( 2 0 1 5 ) 2 6 2 e2 7 2 271

that in the contact zone. Table 3 shows the percentage of


Table 2 e Particle removal efficiency for varying air
particle aggregation (particles converted to aggregates) that
fractions and bubble sizes. Loading rate 23.6 m/hr,
particle size 10 mm, particle density 1100 kg/m3, particle occurred in the separation zone at different air fractions and
volume fraction 0.0001. bubble sizes. The results show that for all of the cases, most
of the aggregation happens in the contact zone. However, a
Bubble size/ 0.005 0.008 0.01 0.02
Air fraction higher percentage of aggregation is shown to occur in the
separation zone at the optimum bubble sizes (Table 1),
20 mm 0.01 0.02 0.09 0.23
40 mm 0.14 0.21 0.23 0.32
where the bubbles penetrate into the separation zone and
80 mm 0.02 0.30 0.67 0.24 the formation of stratified flow promotes further bubble-
120 mm 0.00 0.25 0.48 0.20 particle attachment.
The CFD model also demonstrated the presence of an
optimum air fraction for enhanced particle removal. Particle
removal was at first enhanced by an increase in air fraction,
fraction is required to form stratified flow, and more bubbles
but decreased when the air fraction increased beyond a
are needed to compensate for the shorter contact times in the
certain point (Fig. 10 c and Fig. 10 d). At the loading rate of
contact zone, therefore a higher air fraction is required.
11.8 m/hr, maximum removal was observed at an air frac-
The CFD model was then applied to find the optimum
tion of 0.008 and bubble size of 80 mm. At the loading rate of
bubble size for particle removal. Contours of particle and
23.6 m/hr, particle removal was greatest at air fractions
bubble removal efficiency at different air fractions and bubble
between 0.01 and 0.012 and bubble sizes of 80e100 mm. At
sizes are shown in Fig. 10 for loading rates of 11.8 m/hr and
the optimum air fraction, the bubble layer was able to
23.6 m/hr. The data for particle removal is also provided in
penetrate into the separation zone where bubble-particle
Tables 1 and 2. Fig. 10 a and Fig. 10 b demonstrate that an
attachment was enhanced by the presence of stratified
increase in bubble size increased bubble removal, in agree-
flow. Beyond the maximum air fraction, bubbles formed too
ment with the results of Lakghomi et al. (2012). Fig. 10 c and
quickly and were therefore removed in the top of the con-
Fig. 10 d show that particle removal increased up to a certain
tact zone without allowing further bubble-particle collision
bubble size and decreased beyond that point, which is in
in the separation zone. In addition, due to the higher
agreement with the trend from the analytical model. For air
number of bubbles present in the contact zone, bubbles
fractions higher than 0.01, optimum bubble size was close to
were more likely to attach to other bubbles than to particles
40 mm, similar to the predictions of the analytical model,
and to quickly rise to the surface and be removed from the
whereas for air fractions smaller than 0.01, optimum bubble
system.
size was close to 80 mm. The observed reduction in particle
The CFD model was then applied to predict the impact of
removal with increased bubble size beyond an optimum point
particle size on particle removal. As shown in Fig. 11, the
depends on both contact and separation zone mechanisms.
optimum particle size increased with an increase in bubble
With an increase in the bubble size, fewer bubbles are avail-
size similar to the prediction of the analytical model. Up to
able in the contact zone and lower bubble-particle attachment
the bubble size of 80 mm, the optimum particle size was in the
can occur due to the quick removal of bubbles from top of the
range of 30e70 mm, but with a further increase in bubble size,
contact zone. The bubbles under these conditions have a large
particles larger than 80 mm provided better removal effi-
rise velocity and are removed from the system quickly, so
ciency. When bubble sizes are large, they have a larger
bubble removal is enhanced as shown in Fig. 10 a and Fig. 10 b.
collision frequency with larger particles and can lift the
However, there are fewer bubbles available with lower contact
weight of such larger particles, leading to higher overall
times; therefore, there is less chance for collision among the
removal efficiency despite the availability of fewer bubbles
bubbles and particles in the contact zone. In addition, as the
and particles.
larger bubbles do not penetrate into the separation zone,
stratified flow cannot be formed, which also reduces the op-
portunity for further bubble-particle attachment in the sepa-
ration zone.
The CFD model was then applied to evaluate the signifi-
cance of aggregation in the separation zone compared to

Table 3 e Percentage of particle aggregation occurring in


the separation zone at different air fractions and bubble
sizes. Loading rate 11.8 m/hr, particle size 10 mm, particle
density 1100 kg/m3, particle volume fraction 0.0001.
Bubble size/ 0.005 0.008 0.01 0.02
Air fraction
20 mm 22% 13% 16% 18% Fig. 11 e The effect of particle and bubble size on particle
40 mm 11% 16% 22% 33% removal from the CFD model, loading rate 11.8 m/hr, air
80 mm 42% 33% 11% 18% fraction 0.01, particle density 1100 kg/m3, particle volume
120 mm 17% 24% 10% 12%
fraction 0.0001.
272 w a t e r r e s e a r c h 6 8 ( 2 0 1 5 ) 2 6 2 e2 7 2

Edzwald, J.K., 2007. Developments of high rate dissolved air


4. Summary and conclusions flotation for drinking water treatment. J. Water Supply: Res.
Technol. Aqua 56 (6), 399e409.
An analytical and a CFD model of DAF that included all the Edzwald, K., 2010. Dissolved air flotation and me. Water Res. 44
phases (air, water, particles, and aggregates) were applied to (7), 2077e2106.
Haarhoff, J., Edzwald, J.K., 2001. Modelling of floc-bubble
evaluate the effect of stratified flow, bubble size, air fraction
aggregate rise rates in dissolved air flotation. Water Sci.
and particle size on particle removal. The analytical model
Technol. 43 (8), 175e184.
showed general trends that were similar to the CFD model, Haarhoff, J., Edzwald, J.K., 2004. Dissolved air flotation modelling:
and provided a good semi-quantitative understanding of the insights and shortcomings. J. Water Supply: Res. Technol.
effect of operating parameters without the need for AQUA 53 (3), 127e150.
computationally-intensive CFD. Koh, P.T.L., Schwarz, M.P., 2003. CFD modeling of bubble-particle
The following conclusions can be drawn from both collision rates and efficiencies in a flotation cell. Mineral. Eng.
16 (11), 1055e1059.
modeling approaches:
Kostoglou, M., Karapantsios, T.D., Matis, K.A., 2007. CFD model
for the design of large scale flotation tanks for water and
 The presence of stratified flow improves particle removal. wastewater treatment. Ind. Eng. Chem. Res. 46 (20),
 Higher air fractions are required to achieve the same par- 6590e6599.
ticle removal at a higher loading rate. Lakehal, D., Krebs, P., Krijgsman, J., Rodi, W., 1999. Computing
 An optimum bubble size is present for maximum particle shear flow and sludge blanket in secondary clarifiers. J.
removal depending on the air fraction and particle size. Hydraul. Eng. 125 (3), 253e262.
Launder, B.E., Spalding, D.B., 1974. The numerical computation of
 An increase in air fraction enhances flow stratification and
turbulent flows. Comput. Methods Appl. Mech. Eng. 3 (2),
particle removal up to a certain point, beyond which par- 269e289.
ticle removal deteriorates. Lakghomi, B., Lawryshyn, Y., Hofmann, R., 2012. Importance of
 The optimum particle size increases with an increase in flow stratification and bubble aggregation in the separation
bubble size. zone of a dissolved air flotation tank. Water Res. 46 (14),
4468e4476.
Leppinen, D.M., Dalziel, S.B., Linden, P.F., 2001. Modelling the
global efficiency of dissolved air flotation. Water Sci. Technol.
43 (8), 159e166.
Acknowledgments Leppinen, D.M., Dalziel, S.B., 2004. Bubble size distribution in
dissolved air flotation tanks. J. Water Supply: Res. Technol.
AQUA 53 (8), 531e543.
This work was funded in part by Corix Water Systems and the
Lundh, M., Jonsson, L., Dahlquist, J., 2001. The flow structure in
Natural Sciences and Engineering Research Council of Canada the separation zone of a DAF pilot plant and the relation with
(NSERC) (CRDPJ 446248 - 12) CRD program. bubble concentration. Water Sci. Technol. 43 (8), 185e194.
Matsui, Y., Fukushi, K., Tambo, N., 1998. Modelling, simulation
and operational parameters of dissolved air flotation. J. Water
references Supply: Res. Technol. AQUA 47 (1), 9e20.
Nguyen, A.V., 1999. Hydrodynamics of liquid flows around air
bubbles in flotation: a review. Int. J. Mineral Process. 56 (1),
ANSYS, 2012. ANSYS FLUENT 13.0 Manual and User's Guide. 165e205.
ANSYS Inc., Canonsburg, PA. Rollie, S., Briesen, H., Sundmacher, K., 2009. Discrete bivariate
Edzwald, J.K., Malley, J.P., Yu, C., 1991. A conceptual model for population balance modeling of heteroaggregation processes.
dissolved air flotation in water treatment. Water Supply 9 (1), J. Colloid Interface Sci. 336 (2), 551e564.
141e150. Saffman, P.G., Turner, J.S., 1956. On the collision of droplets in
Edzwald, J.K., 1995. Principles and applications of dissolved air turbulent clouds. J. Fluid Mech. 1 (1), 16e30.
flotation. Water Sci. Technol. 31 (3), 1e23.

Vous aimerez peut-être aussi