Journal of Environmental Chemical Engineering 9 (2021) 104674

Contents lists available at ScienceDirect

Journal of Environmental Chemical Engineering

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

CFD study of the hydrodynamics and biofilm growth effect of an anaerobic

inverse fluidized bed reactor operating in the laminar regime
Jorge Ramírez-Muñoz a, Román Guadarrama-Pérez b, Alejandro Alvarado-Lassman c,
José J. Valencia-López d, Valaur E. Márquez-Baños e, *
a
Departamento de Energía, División de Ciencias Básicas e Ingeniería, Universidad Autónoma Metropolitana-Azcapotzalco, Av. San Pablo, 180, Col. Reynosa Tamaulipas,
C.P. 02200, CDMX, Mexico
b
Posgrado en Ciencias Naturales e Ingeniería, División de Ciencias Naturales Ingeniería, Universidad Autónoma Metropolitana-Cuajimalpa, Av. Vasco de Quiroga, 4871,
Col. Contadero, C.P. 05370, CDMX, Mexico
c
División de Estudios de Posgrado e Investigación, Tecnológico Nacional de México/Instituto Tecnológico de Orizaba, Av. Oriente 9, 852, Col. Emiliano Zapata, C.P.
94320, Orizaba, Mexico
d
Departamento de Procesos y Tecnología, División de Ciencias Naturales Ingeniería, Universidad Autónoma Metropolitana-Cuajimalpa, Av. Vasco de Quiroga, 4871, Col.
Contadero, C.P. 05370, CDMX, Mexico
e
Departamento de Procesos e Hidráulica, División de Ciencias Básicas e Ingeniería, Universidad Autónoma Metropolitana-Iztapalapa, Av. San Rafael Atlixco, 186, Leyes
de Reforma 1ra Secc, C.P. 09340, CDMX, Mexico

A R T I C L E I N F O A B S T R A C T

Editor: GL Dotto The unsteady laminar flow of an inverse fluidized bed reactor was addressed in this study using computational
fluid dynamics (CFD). Three-dimensional (3D) and two-dimensional (2D) simulations were carried out using the
Keywords: Eulerian-Eulerian approach with the Syamlal O’Brien and Gidaspow models for modeling the drag between the
Inverse fluidized bed reactor solid and liquid phases. Reported data of solids expansion and bed porosity were used to validate the simulations.
CFD simulations
Results show that 2D simulations can satisfactorily reproduce experimental data of fluidized bed reactors, which
Fluid-solid flow
are essentially equal to those obtained on 3D simulations. The model of Syamlal O’Brien was found to be in better
Biofilm growth
Laminar regime agreement with experiments than that of Gidaspow. It is shown that commonly used correlations for predicting
bed porosity fail to provide a reasonable overall description of the evaluated operating conditions. The flow
patterns for both phases in the bed section exhibit chaotic recirculation zones with high values of shear stress,
and after this, the flow quickly becomes fully developed. The effect of biofilm thickness increase on bed porosity,
under similar apparent densities of colonized particles reported for this system, is analyzed. Biofilm growth
generates a decrease in the Archimedes number (i.e., decreases the buoyancy of particles) and an increase in the
particle Reynolds numbers (Rep), affecting significantly the final porosity of the bed. However, this effect van­
ishes for Rep < 0.25. Results from this work can be used to develop scale-up criteria and to establish safe
operating conditions (e.g., maximum liquid velocity and biofilm thickness) to avoid loss of support particles.

eventually becomes colonized with biofilm and results in a biomass

concentration of approximately an order of magnitude greater than in
1. Introduction conventional suspended cell systems [5,6]. In fluidized bed reactors, if
the solid particle density is higher than the fluid density, the fluidization
Fluidized bed reactors are commonly used in a wide variety of is carried out by an upward flow. In contrast, if the solid particles are less
existing processes in the chemical and biotechnology industries [1]. dense than the fluid, then the bed should be fluidized by a downwards
They have significant advantages for wastewater treatment with respect flow, generally denoted as inverse fluidization.
to suspended cell systems, such as high resistance to system disturbances Inverse fluidized bed reactors (IFBRs) possess superior hydrody­
including the presence of inhibitors, sudden changes in pH and biomass namic performance and are more efficient than conventional fluidized
loss with the effluent [2], high mass transfer rates and uniform mixing bed reactors for wastewater treatment, at both laboratory and full scale
[3], as well as low operating costs [4]. In these systems, each particle

* Corresponding author at: Departamento de Procesos e Hidráulica, División de Ciencias Básicas e Ingeniería, Universidad Autónoma Metropolitana-Iztapalapa, Av.
San Rafael Atlixco, 186, Leyes de Reforma 1ra Secc, C.P. 09340, CDMX, Mexico.
E-mail address: [email protected] (V.E. Márquez-Baños).

https://doi.org/10.1016/j.jece.2020.104674
Received 7 September 2020; Received in revised form 20 October 2020; Accepted 24 October 2020
Available online 28 October 2020
2213-3437/© 2020 Elsevier Ltd. All rights reserved.
J. Ramírez-Muñoz et al. Journal of Environmental Chemical Engineering 9 (2021) 104674

Nomenclature y axial distance from the bottom [m]

Ar Archimedes number [-] Abbreviations

AL area occupied by the expanded bed [m2] CFD Computational Fluid Dynamics
C1 Campos et al. (2012) model parameter [-] DEM Discrete Element Method
CD drag coefficient [-] DPM Discrete Particle Model
Dc diameter of the column [m] FBR Fluidized Bed Reactor
Dls rate of energy exchange [kg/m·s3] IFBR Inverse Fluidized Bed Reactor
dp particle diameter [m] Greek letters
e coefficient of restitution [-] αl volume fraction of fluid [-]
fm virtual mass force [kg·m/s2] αs volume fraction of solids [-]
→g gravitational acceleration [m/s2] β interface momentum transfer coefficient [kg/m2·s2]
g0 radial distribution function [-] γs kinetic energy dissipation rate [kg/m·s3]
H0 initial height of solids [m] ε bed porosity [-]
H height of solids [m] ε0 initial bed porosity [-]
ks thermal diffusion coefficient [kg/m·s] θ granular temperature [m2/s2]
p fluid pressure [kg/m·s2] μl molecular viscosity of fluid [kg/m·s]
ps solids pressure [kg/m·s2] μs solid shear viscosity [kg/m·s]
Re Reynolds number [-] ρl density of liquid [kg/m3]
Res relative Reynolds number [-] ρs density of solids [kg/m3]
Rep particle Reynolds number [-] ξs bulk viscosity of particles [kg/m·s]
t time [s] τl stress tensor of liquid phase [kg/s2]
VL volume occupied by the expanded bed [m3] τs stress tensor of solid phase [kg/s2]
vr,s terminal velocity correlation [m/s] ϕs particle sphericity [-]
→ul liquid velocity [m/s]
Subscripts
→us solid velocity [m/s]
l liquid phase
Ul inlet liquid velocity [m/h]
s solid phase
x radial distance [m]

[7]. These systems are capable of achieving wastewater treatment with virtual mass models for modeling the solid-liquid hydrodynamic inter­
low retention times due to their high biomass concentrations in the action. These authors showed that CFD simulations can correctly ac­
reactor [6]. Additionally, this type of reactor is simpler in design, easier count for the flow behavior of the liquid and solid phases, as well as for
to control and has lower biomass production in comparison with aerobic the reported axial velocities of particles and bed expansion heights as a
systems [8]. Another advantage of IFBRs is that high fluidization rates function of liquid viscosity. For this same system, Wang et al. [27]
are not required in their operation, and thus, a reduction in operating carried out 2D flow simulations employing an Euler-Lagrangian
costs is achieved [9]. In recent decades, a large number of successful approach by using the Huilin-Gidaspow drag model, and lubrication
study cases of wastewater treatment using IFBRs has been reported in forces. They found that the behavior of the particles in an IFBR can be
both aerobic [10,11] and anaerobic [5,12,13] systems. divided into three stages: particles flow downward, particles flow up­
Most of the experimental and numerical research on IFBRs are car­ wards, and when the dynamic balance is reached in the bed.
ried out in turbulent conditions, the most relevant of which, in the Song et al. [28] performed a 3D CFD study in a full-scale IFBR
context of this work, will be mentioned below. Anaerobic IFBRs are operating in the turbulent regime using an Euler-Euler scheme coupled
liquid-solid systems whose study has been addressed, for example, by to the kinetic theory of granular flow, and using the Syamlal O’Brien
using experimental model reactors to determine bed expansion [14], drag model. They found a good agreement between their numerical
solids volume fraction profiles [15], minimum fluidization velocity results and experimental data of average solids holdup in the entire bed
[16], and pressure drop [17]. In other studies, mathematical modeling reported by Jaberi [29], and by Nan [30]. Furthermore, they investi­
of IFBRs has been carried out using empirical models to correlate these gated the effect of particle densities on the average solids expansion
variables [17–19]. along the reactor. Their results show that under this effect the average
In addition to the aforementioned experimental studies, CFD retention of solids in the cross-section along the axis was uniform, and
approach have nowadays become a powerful tool for modeling waste­ the radial flow structures at different bed heights were identical.
water treatment reactors because they provide deep insight into the On the other hand, some industrial applications of anaerobic IFBRs
existing local hydrodynamics of these systems at relatively very low can fall into the laminar regime due to the residence time used for
costs in comparison with experimental techniques [20–22]. Several certain processes, such as brewery wastewater treatment [12,31].
authors have reported that 2D models can adequately reproduce the However, CFD studies on these systems are relatively scarce. The main
hydrodynamics of fluidized bed reactors that occurs naturally in purpose of this work is to study and gain fundamental insight into the
three-dimensional spaces [23–25], however, these have not been tested hydrodynamics of an IFBR previously characterized experimentally in
for IFBRs operating in the laminar flow regime. 2D simulations signifi­ the laminar regime for the organic matter removal from brewery
cantly reduces the computational cost of transient simulations required wastewater (see Alvarado-Lassman et al. [12], “AL-2008′′ , from now
for modeling these systems. on). In order to simulate the effect of biofilm thickness increase on IFBR
Wang et al. [26] applied this modeling tool for simulating the par­ bed porosity and to ultimately ascertain the conditions that lead to
ticle flow behavior in an IFBR operating in turbulent conditions. They biomass loss, different particle sizes under similar apparent densities of
used a 2D model in an Euler-Euler scheme that includes the kinetic colonized particle supports reported for this system were evaluated. To
theory of granular flow, together with the Huilin-Gidaspow drag and the best of the authors’ knowledge, an analysis of the effect of the

2
J. Ramírez-Muñoz et al. Journal of Environmental Chemical Engineering 9 (2021) 104674

biofilm thickness growth on bed porosity using a CFD approach has each phase. The Eulerian-Lagrangian approach, which is also recognized
never been addressed elsewhere for an IFBR operating in the laminar as a discrete particle model (DPM) or discrete element method (DEM),
regime. regards the fluid as the continuous phase, while solid particles are the
dispersed phase [32]. The DPM uses the Eulerian framework to model
2. CFD simulations the continuous phase, and the trajectories of the particles are simulated
in the Lagrangian framework. Despite the advantage of having a detailed
2.1. Model geometry and computational mesh description of the movement of individual particles, Lagrangian models
require high computation power and long simulation times in order to
The inverse fluidized bed reactor is a cylindrical column with a obtain average values of the flow field [33]. For this reason, they are
height of 1.2 m, internal diameter (Dc) of 0.0448 m and a working rarely used in simulations of full-scale equipment or with a high con­
volume of 1.9 L. Its hydrodynamic performance and removal efficiency centration of particles [34].
were reported previously by our team (AL-2008), and therefore, its An alternative is to use models with the Eulerian-Eulerian approach,
detailed dimensions can be consulted there. also known as the two-fluid model. In this approach, both phases (solid
A schematic of the analyzed reactor and its computational meshes for and fluid) are taken as totally interpenetrating continua, and the con­
the 2D and 3D geometries are shown in Fig. 1. Both geometries were servation equations for the mass and the moment of both phases result
generated in ANSYS®-DesignModeler and the meshes were built in from the statistical average of the instantaneous local transport equa­
ANSYS®-Meshing. The geometries were mapped with dimensions equal tions [35–38]. This set of equations is solved by using constitutive re­
to the diameter and height of the working volume. Computational lationships obtained by granular kinetic theory, and the coupling is
meshes comprised with quadrilateral and hexahedral elements were achieved through the pressure and interface exchange coefficients. It has
built, for the 2D and 3D models, respectively. been reported that Eulerian-Lagrangian and Eulerian-Eulerian methods
produce comparable results between them with high concentration of
particles [39].
2.2. Eulerian two-phase model Due to the high solids concentration and the reactor size, the
Eulerian-Eulerian approach is employed in this work. This approach has
There are two main approaches for modeling solid-fluid systems by a lower computational cost, but its accuracy depends on the closure
using CFD tools: The Eulerian-Lagrangian and the Eulerian-Eulerian model used to determine the drag force between the two phases [40,41].
methods, which depend on the reference frame used for modeling For this reason, two closure models were tested in this study: The
Syamlal O’Brien and the Gidaspow drag models.

2.2.1. Governing equations

The model used to describe the solid-fluid hydrodynamics is based on
the Eulerian-Granular model of ANSYS Fluent v17.1. In this Eulerian-
Eulerian two-fluid model, all phases share the same pressure field ac­
cording to their respective volume fraction, and the laws of conservation
of mass and momentum are satisfied independently by each phase.
The models implemented involve a set of two equations of continuity
(Eq. 1 and 2) and momentum (Eq. 3 and 5). The set of momentum
constitutive equations are the liquid phase stress tensor (τl, Eq. 4), the
solid phase stress tensor (τs, Eq. 6), and the momentum exchange be­
tween the solid and liquid phases, which is a function of virtual mass
(fm), and drag force (β).
∂
(αl ρl ) + ∇⋅(αl ρl →
u l) = 0 (1)
∂t

∂
(α ρ ) + ∇⋅(αs ρs →
u s) = 0 (2)
∂t s s

∂
(α ρ →
u ) + ∇⋅(αl ρl →
u l→
u l ) = − αl ∇p + ∇(αl ⋅τl ) + αl ρl →
g + β(→
us− →
u l)
∂t l l l
(3)

[ ] 2
u l + (∇⋅→
τl = μl ∇⋅→ u l )T − μ (∇⋅→
u l )I (4)
3 l

∂
(α ρ →
u ) + ∇⋅(αs ρs →
u s→
u s ) = − αs ∇p − ∇ps + ∇(αs ⋅τs ) + αs ρs →
g + fm
∂t s s s
+ β(→
ul− →
u s)
(5)

[ ] 2
τs = μs ∇⋅→
u s + (∇⋅→
u s )T − μ (∇⋅→
u s )I + ξs ∇⋅→
u sI (6)
3 s
( )
dul dus
fm = − 0.5αs ρl − (7)
dt dt
The virtual mass force (Eq. 7) acts on a particle when a secondary
Fig. 1. Schematic drawing of the IFBR and its corresponding 2D and 3D
phase accelerates relative to the primary phase [42]. This effect is
computational mesh.

3
J. Ramírez-Muñoz et al. Journal of Environmental Chemical Engineering 9 (2021) 104674

significant when the solid phase density is much smaller than the liquid √̅̅̅
4 θ
phase density. ξs = α2s ρs dp g0 (1 + e) (21)
3 π
The accuracy of these models depends mainly on the selected drag
force model. Several drag models exist for the liquid-solid interface On the other hand, the solids shear viscosity model was selected
exchange coefficient. In this study, the performance of the drag models depending on the drag model used.
of Gidaspow [40], and Syamlal & O’Brien [43] were examined by using Gidaspow [40] solids shear viscosity
Eqs. (8)‒(10) and Eqs. (11)‒(14), respectively. √̅̅̅ √̅̅̅̅̅ [ ]2
Gidaspow [40] drag model: 4 θ 10ρs dp π θ 4
μs = α2s ρs dp g0 (1 + e) + 1 + αs g0 (1 + e) (22)
5 π 96g0 (1 + e) 5
⎧ 3αs αl ρl
⎪
⎪
⎪ C D |→
us− →
u l |α−l 2.65
Syamlal and O’Brien [43] solids shear viscosity
⎨ 4dp
, αl > 0.8
β= (8) √̅̅̅̅̅ [ ] √̅̅̅
⎪
⎪
⎪ αs (1 − αl )μl us− →
ρ l α s |→ u l | , αl ≤ 0.8 μs =
αs ρs dp πθ 2 4 2
1 + αs g0 (1 + e)(3e − 1) + αs ρs dp g0 (1 + e)
θ
(23)
⎩ 150 + 1.75 6(3 − e) 5 5 π
αl dp2 dp

24 [ ] 2.2.2. Initial and boundary conditions

CD = 1 + 0.15(αl Res )0.687 (9) The equations described above are numerically solved with appro­
αl Res
priate boundary and initials conditions. According to AL-2008, in their
Syamlal and O’Brien [43] drag model: (10)
bed expansion tests with Extendospheres™, the liquid phase was tap
( )
3αs αl ρ Res → → water, which was fed continuously to the reactor. Therefore, at the inlet
β = 2 l CD | u s − u l| (11) (top of the reactor), the liquid velocity with no solids entering the system
4vr,s dp vr,s
was established. At the outlet, the pressure is specified for both liquid
⎛ ⎞2 and solid phases. At the walls, the liquid tangential and normal veloc­
⎜ 4.8 ⎟ ities are set to zero. For solid particles, the normal velocity is set to zero,
CD = ⎜
⎝0.63 + √̅̅̅̅̅
⎟ (12) and for the tangential velocity, the partial slip boundary condition [46]
Res ⎠
vr,s is used. Initially, velocity is set at zero in the bed, for both the liquid
phases and the solid particles. In our simulations, different bed heights
( √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ) are analyzed and patched with an initial condition of porosity of 0.42.
vr,s = 0.5 A − 0.060Res + (0.06Res )2 + 0.12Res (2B − A) + A2 (13)
2.3. Study cases
{
0.8α1.28 , αl ≤ 0.85
A = α4.14
l ;B = l
(14) As a first study case, 2D and 3D CFD simulations corresponding to the
α2.65 , αl > 0.85
l
bed expansion tests reported in AL-2008 with Extendospheres™, at
different liquid velocities, were performed. The solid particles have an
Where Res is defined by Eq. 10
average diameter (dp) and density (ρs ) of 169 μm and 700.0 kg/m3,
The constitutive equations for the solids governing equation are
respectively. The liquid phase of the reactor is water at 35 ◦ C, with a
based on the kinetic theory of granular flow. In it, a granular tempera­
density (ρl ) of 994.08 kg/m3 and viscosity (μl ) of 0.000720 Pa∙s [47]. A
ture, θ, is introduced into the model, and appears in the expression for
list of different combinations of inlet liquid velocities (Ul) and initial bed
the solids pressure and viscosity [44]. The granular temperature is
heights are shown in Table 1. It is worth noting that it was not possible to
proportional to the kinetic energy of the random motion of the particles,
analyze all combinations because for high inlet liquid velocities or initial
and it is conserved and transported based on the conservation of solids
bed height a considerable loss of support particles was observed in
fluctuating energy equation (Eq. 15).
experiments.
[ ]
3 ∂ Also, the effect of biofilm growth on the Extendosphere™ support
(αs ρs θ) + ∇⋅(αs ρs θ→
u s ) = ( − ∇ps I + τs )
2 ∂t particles was studied. The diameter and density of the colonized parti­
: ∇→
u + ∇⋅(k ∇θ) − γ − 3βθ − D
s (15)s ls
cles were taken from AL-2008 and are shown in Table 2. This analysis
s
was performed considering an initial bed height of 0.24 m and inlet
Finally Eqs. 16–23 list additional constitutive equations required to liquid velocity ranging from 0.06–8.86 m/h.
close the fluctuation energy equation [45], which includes the thermal Particle properties result in an Archimedes number (Ar =
diffusion coefficient (ks), the solids pressure (ps), the kinetic energy d3p (ρl − ρs )ρl g/μ2l ) in the range of 1.86‒26.7. Operating conditions in the
dissipation rate (γs), the rate of energy exchange (Dls), and the radial column keep the reactor operating under the laminar flow regime, with
distribution function (g0). particle Reynolds number (Rep = dp Ul ρl /μl ) and column Reynolds
ps = αs ρs θ + 2ρs (1 + e)α2s g0 θ (16) number (Re = Dc Ul ρl /μl ) in the range of 0.06–1.22 and 15.5–214.8,
respectively.
√̅̅̅̅̅ [ ]2 √̅̅̅
25ρs dp πθ 6 θ
ks = 1 + αl g0 (1 + e) + 2α2s ρs dp g0 (1 + e) (17)
64(1 + e)g0 5 π
( √̅̅̅ )
( ) 4 θ
γs = 3 1 − e α ρ 2 2
s s dp g0 θ − ∇⋅→
us (18) Table 1
dp π
Combinations of the operating conditions simulated in the reactor (AL-2008).
( )2 Inlet liquid velocity [m/h] Initial bed height [m]
dp ρ 18μl
Dls = √̅̅̅̅̅s ul− →
|→ u s |2 (19) 0.12 0.24 0.36 0.48
4 π θg0 dp2 ρs
0.9 × × × ×
3.0 × × × ×
[ ( )1/3 ]− 1
5.3 × × ×
αs
g0 = 1 − (20) 8.0 × ×
αs,max 10.3 ×
12.5 ×

4
J. Ramírez-Muñoz et al. Journal of Environmental Chemical Engineering 9 (2021) 104674

Table 2
Particle properties of colonized Extendospheres™ (AL-2008).
Biofilm thickness [μm] 10 25 35 50 65 75 85 100
Particle diameter [μm] 189 219 239 269 299 319 339 369
Particle density [kg/m3] 778.2 871.6 906.1 940.5 961.1 971.3 979.4 988.9

2.4. Solver settings only a time step of 0.1 s. The results of this analysis are included in the
supplementary material section, where it can be noted that, under the
The set of governing equations for the 2D and 3D unsteady models same criteria used for the 2D model, the 3D geometry with 914,760
were solved with the finite volume method, performing a pressure- (Mesh 4) can be considered as the independent mesh.
velocity coupling and using the phase-coupled SIMPLE algorithm. The
partial differential equations are discretized using the QUICK scheme. 3.2. Comparison of simulation results with bed expansion experimental
First-order implicit time stepping was used as has been shown that data
second-order time discretization is only necessary for accurate solution
of fast-moving flows [48], which is not the case for this laminar study. A In order to validate 2D and 3D simulation results, bed expansion data
convergence criterion of 10− 3 (relative error) for all scaled residuals was reported by AL-2008, using the same studied system, are referred to in
established. To define the appropriate time step and element size in the this section. Fig. 3 shows the experimental data and numerical results for
computational mesh, a sensitivity analysis of these parameters was the bed height in fluidized state for the reactor operating at different
performed and it is presented in Section 3.1. inlet liquid velocities and initial bed heights shown in Table 1. For the
2D simulation results, predictions by using the drag models of Gidaspow
3. Results and discussion et al. [40] and Syamlal & O’Brien [43] at all operation conditions are
included. It is clear that at low inlet liquid velocities (i.e., <3 m/h) both
3.1. Mesh and time step independence analysis drag models adequately predict the final bed height; however, for higher
velocities, the discrepancies are larger for the model of Gidaspow. This
It is necessary to ensure that the results obtained from the simula­ error decreases for both drag models when the initial height of solids is
tions are independent of the mesh resolution and the time step selected. low as in the case of H0=0.12 m, i.e., with low solids loads.
The most common method to achieve mesh independence is to increase The above results are in agreement with those of Song et al. [28],
its resolution (usually by doubling the number of elements); when re­ who reported that CFD simulations employing the model of Syamlal &
sults do not change significantly and an established criterion is met, the O’Brien can adequately represent the distribution of solids in inverse
mesh with the least number of elements is chosen. fluidized bed reactors. On this basis, results presented for both 2D and
To this end, five meshes of decreasing element sizes were built. The 3D models from now on refer exclusively to numerical simulations made
number of elements of each mesh is shown in Table 3 for the 2D and 3D with the Syamlal & O’Brien’s drag model.
geometries. This analysis was carried out by using the Syamlal & Due to the fact that computational costs of 3D flow simulations are in
O’Brien [43] drag model, for an initial bed height of 0.12 m and with the between one and two orders of magnitude higher than 2D simulations,
highest inlet liquid velocity (12.5 m/h). Simulations were run until only an initial bed height of H0=0.12 with six inlet liquid velocities were
completing four residence times (θ=VR/Q, where VR is the reactor examined. It can be seen that both models (2D and 3D) can satisfactorily
working volume and Q is the inlet flow) in order to assure that the reproduce experimental data of IFBRs operating in the laminar regime.
system reached the steady state.
In this analysis, for the 2D CFD model, three time steps, namely 0.05, 3.3. Comparison between numerical bed porosity results and empirical
0.1, and 0.15 s were chosen to establish solution independence for this correlations
parameter. Fig. 2 shows simulation results of this analysis: Relative
expansion of the bed (Fig. 2a), and average velocity of the liquid The final porosity of the bed is directly related to the bed expansion.
(Fig. 2b) and solid particles (Fig. 2c). It can be seen in Fig. 2a that the It has been reported that this is a key parameter, as it is useful in the
change in the relative expansion of the bed between mesh 3 and 5 is only scaling and design of fluidized bed reactors [11,49]. Moreover, the
1%, while the three different time steps used do not influence signifi­ interfacial area and mass transfer parameters depend on it [50,51].
cantly this variable. Changes in bed porosity due to expansion affect the correct operation
The average velocity of liquid (Fig. 2b) and solid (Fig. 2c) phases and control of the reactor. For this reason, it is important to have a
show important differences for a time step of 0.15 compared to the reliable model to estimate the porosity of the expanding bed. The reactor
smaller time steps of 0.1 and 0.05 s. On the other hand, the average porosity (ε) at each inlet liquid velocity can be calculated from the initial
difference in velocity between the time step of 0.1 s and the smallest is porosity (ε0), initial height (H0) and final height (H) of the bed,
only 3% for all evaluated meshes. For these time steps, the average solids employing the following equation [17]:
velocity reaches its independence until Mesh 4, with a difference of only ( )
(1 − ε0 )
4% with respect to Mesh 5. From this analysis, it is clear that Mesh 4 ε = 1 − H0 (24)
H
(with 13,200 elements and a time step of 0.1 s) is enough to obtain in­
dependent numerical results for both parameters. By using CFD simulation results, the porosity can be calculated by
The same analysis was carried out for the 3D model, but considering extracting the average volume fraction of the liquid phase in the bed
region within the 2D model, as
∫∫
Table 3 1
Number of cells for the mesh independence analysis.
ε= αl dAL (25)
AL
AL
Mesh 2D model 3D model

1 2160 90,360 where AL is the area occupied by the expanded bed. While in the 3D
2 3300 228,690 model, it can be calculated as:
3 6000 457,380
4 13,200 914,760
5 54,000 1,864,200

5
J. Ramírez-Muñoz et al. Journal of Environmental Chemical Engineering 9 (2021) 104674

Fig. 2. Mesh- and time step independence analysis (2D model) for a) relative expansion of the bed, b) average liquid velocity, and c) average solid velocity.

( )0.34
C1 = 2.25 5.5/Rep (28)

Das et al. [19] developed an empirical correlation for predicting bed

expansion as a function of physical and dynamic variables of the system,
given by the following equation:
H ( / )0.4613
= 0.988Ar− 0.2124
Re0.4049 ϕ−s 1.804 Dc dp (29)
H0

where ϕs, is the particle sphericity. The final porosity from this equation
can be computed with Eq. (24).
Fig. 4 shows the values of bed porosities as a function of the particle
Reynolds numbers predicted by the empirical correlations of Campos-
Díaz et al. [49] and Das et al. [19], as well as 2D and 3D numerical
results from this work. For comparison purposes, experimental data
from AL-2008 are also included.

Fig. 3. Comparison of calculated bed heights with Syamlal and O’Brien and
Gidaspow drag models versus experiments (from AL-2008) as a function of inlet
liquid velocity.

∫∫∫
1
ε= αl dVL (26)
VL
VL

where VL is the volume occupied by the expanded bed.

Numerous empirical correlations have been proposed in the litera­
ture for predicting the porosity or expansion of the bed in IFBRs. Campos
et al. [49] proposed the following models to estimate the final bed
porosity as a function of particle Reynolds- and Archimedes numbers:
[ ( ) ]− 1
1.753ε− 3.807 = Ar 0.75 24Rep + C1 Re2p (27)

where
Fig. 4. Comparison of bed porosity predictions from existing empirical corre­
lations, experimental data, and numerical simulations from this work.

6
J. Ramírez-Muñoz et al. Journal of Environmental Chemical Engineering 9 (2021) 104674

It is evident that at Rep<0.2 the correlation of Campos-Díaz et al. and 12.5 m/h. Simulations were performed enabling support expansion
[49] adequately predicts the porosity of the bed, while at Rep>0.7 the until reaching equilibrium. Note that the bed expansion increases with
correlation of Das et al. [19] is more in agreement with experimental the liquid velocity and that the solid particles in the bed are distributed
data. However, both empirical correlations fail to provide a reasonable uniformly with a practically constant porosity without the presence of
overall description at all evaluated operating conditions, and this might solids clusters or void areas; this improves the homogeneity of the
be associated with the experimental conditions in which they were support particles and the hydrodynamics inside the reactor. AL-2008
fitted. On the other hand, for both 2D and 3D numerical predictions observed experimentally this same fluidization behavior of Extendo­
based on the drag model of Syamlal & O’Brien [43], these show an spheres™. They reported that under the same conditions, extendosphere
adequate tendency of the experimental data in the entire range of Rep, support presented excellent hydrodynamic behavior for fluidization,
which gives a comprehensive validation of the results of the present which favors surface colonization of particles and biofilm growth.
study.
Results suggest that 2D and 3D geometries with 13,200 and 914,760 3.5. Flow patterns and shear stress contours
cells, respectively, numerically reproduce bed expansion and bed
porosity outcomes that are nearly the same. Based on this, all the sim­ Fig. 7 shows flow patterns in the form of velocity vectors for both
ulations hereinafter in the present study are conducted with the 2D phases, and shear stress contours for the liquid phase, for the different
computational model. inlet liquid velocities analyzed. These simulations were carried out with
the reactor operating at an initial bed height of 0.12 m.
3.4. Solid phase distribution profiles In all cases, there are numerous recirculation zones inside and near
the solids bed region, and once the liquid passes through this region, the
Fig. 5 shows the axial profiles of solids volume fractions along the liquid flow returns to the characteristics pipe velocity profile. These
reactor center evaluated at different times and measured from the col­ chaotic recirculation regions might improve micromixing in the system,
umn bottom. These values were at an inlet liquid velocity of 3 m/h and reducing the existence of shortcut flow or stagnant volumes in the solids
different initial heights, i.e., H0 of 0.12 m (Fig. 5a), 0.24 m (Fig. 5b), bed regions. It is also evident that the flow patterns for both phases in the
0.36 m (Fig. 5c) and 0.48 m (Fig. 5d). It is clear that for an IFBR oper­ bed zone are complementary, showing that the solid particles also
ating in the laminar regime, the time required to obtain steady-state maintain important circulation throughout this region, which has been
conditions is about one residence time. After this, the profiles of the reported by Wang et al. [27].
solids volume fraction become flat, reaching equilibrium support Regarding shear stress contours, it can be seen that, for all cases, the
expansion and uniform distribution along the reactor. zones of maximum shear stress are the walls of the reactor in the final
Fig. 6 shows the contours of the solids volume fractions in the column section of the solids bed. Moreover, recirculation in this region also
at H0 = 0.12 m and inlet liquid velocities between 0 (stagnant liquid) generates zones of high shear stress values, which could cause biofilm

Fig. 5. Transient axial distributions of solids volume fractions at a liquid velocity of 3 m/h and different initial heights (2D model): a) 0.12 m, b) 0.24 m, c) 0.36 m,
d) 0.48 m.

7
J. Ramírez-Muñoz et al. Journal of Environmental Chemical Engineering 9 (2021) 104674

Fig. 6. SolidsvolumefractionprofilesatH0 = 0.12mfordifferentinletliquidvelocities(2Dmodel).

Fig. 7. Flow patterns and shear stress contours for an initial reactor height of 0.12 m at inlet liquid velocities between 0.9 and 12.5 m/h (2D model).

detachments [52–54]. excessive downwards liquid velocity causes overexpansion of the sup­
It is noteworthy from Fig. 7 that, at lower liquid velocities, lower port media with the inevitable bioparticle washout at the bottom of the
relative expansions of the solids bed are observed, i.e., there exists a high reactor [5,6].
concentration of particles in the bed near the column top. This causes an Fig. 8a shows the effect of biofilm growth on the final bed porosity
acceleration of the fluid flow passing this region (i.e., Venturi-like ef­ for an initial solids height of 0.12 m and different inlet liquid velocities
fects) and explains why higher local liquid and solids velocities, as well in the range of 0.06–8.54 m/h. These contour maps were built based on
as shear stresses, are induced at lower liquid flow rates. 42 simulation runs, which encompass all different sizes and densities of
colonized support particles reported by AL-2008, presented in Table 2.
Thus, these results represent the final porosity of the bed for Extendo­
3.6. Effect of biofilm thickness growth on bed porosity spheres™ at different colonization states.
In Fig. 8a the change in bed porosity exhibits a nonlinear behavior
Bed porosity is a key variable that determines the organic matter with respect to biofilm thickness and inlet liquid velocities. According to
residence time in the biocatalyst zone of IFBRs [55], and is directly these results, excessive growth of biofilm thickness can limit the oper­
related with the process pumping cost [17]. It has been reported that ation of the inverse fluidized bed reactor to very low inlet liquid
excessive biomass growth on the support particles coupled with

8
J. Ramírez-Muñoz et al. Journal of Environmental Chemical Engineering 9 (2021) 104674

Fig. 8. Bedporositycontours(2Dmodel):(a)asafunctionofbiofilmthicknesssizeandinletliquidvelocity,and(b)asafunctionofArchimedesandparticleReynoldsnumbers.

velocities in order to avoid the loss of support particles. For the analyzed relative expansions of the solids bed near the column top are observed,
column, the maximum porosity that allows the working height is 0.88 causing an acceleration of the fluid flow passing through this region and
(see Ec. (24) and the bold dotted lines in Figs. 8a and 8b). Therefore, inducing higher local liquid and solids velocities. Higher local shear
operating at inlet liquid velocities lower than 2 m/h might prevent stresses are induced at relatively lower liquid velocities, as well. Our
solids loss, since after the stabilization period there is a fully developed results show that biofilm growth generates a decrease in Archimedes
biofilm attached to the Extendospheres™, with an average thickness of number and an increase in particle Reynolds number, significantly
95 μm, as was reported by Alvarado-Lassman et al. [31] for the same influencing final porosity and the eventual loss of support particles in
IFBR. Simulations results from this study are also in agreement with the reactor. However, at low particle Reynolds number (i.e., Rep<0.25),
those reported by AL-2008. They reported that operating the reactor at final bed porosity is not affected by Archimedes number. It is shown that
maximum inlet liquid velocities of 0.2 m/h allows a biofilm thickness an excessive growth of biofilm can limit the operation of the IFBR to very
larger than 260 μm without loss of support particles due to low inlet liquid velocities in order to avoid the loss of support particles.
overexpansion. Our results suggest that 2D CFD models can be reliably used for
It is well known that dimensionless numbers reduce the number of reproducing observed experimental results in inverse fluidized bed re­
variables needed for description of a problem, and when dimensionless actors in the laminar regime, and hence aid in gaining insight into their
groups are properly formed, they improve the physical interpretation hydrodynamic performance and in establishing safe reactor operating
and contribute to the physical understanding of a specific phenomenon conditions (e.g., maximum inlet liquid velocities and biofilm thickness)
under study [56]. In addition, they can be used for estimating the in order to avoid loss of particles by fluid drag or settling.
behavior of a system and are useful for scaling geometrically similar
systems. Considering the advantages of using dimensionless numbers, CRediT authorship contribution statement
Fig. 8b shows bed porosity contours as a function of the Archimedes and
particle Reynolds numbers. It is clear that biofilm growth generates a Jorge Ramírez-Muñoz: Conceptualization, Investigation, Writing -
decrease in the Ar number and an increase in the Rep, i.e., the support review & editing. Román Guadarrama-Pérez: Formal analysis, Inves­
particle buoyancy decreases and the fluid drag force on them increases, tigation, Visualization. Alejandro Alvarado-Lassman: Conceptualiza­
leading altogether to the observed loss of support particles in the reactor tion, Validation, Supervision. José J. Valencia-López: Investigation,
(white region). At Rep smaller than 0.25, bed porosity is almost not Resources. Valaur E. Márquez-Baños: Methodology, Software, Writing
affected in the entire range of Archimedes number values evaluated. - original draft, Visualization.
However, at larger Rep, a slight change in the buoyancy of particles
generates a significant bed porosity increase. These results highlight the
Declaration of Competing Interest
practical application of our findings and open a new line of research for
scaling these systems, since IFBRs have been applied at the laboratory
The authors reported no declarations of interest.
level, but no fully-scale systems have been reported to date.

Acknowledgments
4. Conclusions

RGP thank Consejo Nacional de Ciencia y Tecnología (Mexico), for

Hydrodynamic parameters that influence the IFBR performance, i.e.,
their academic scholarship
flow patterns, shear stress, solids bed profiles, bed porosity, as well as
the effect of the biofilm thickness growth on the bed porosity, were
analyzed by using a CFD approach. Appendix A. Supplementary data
It was found that solid particles are uniformly distributed in the bed
with a nearly constant porosity without the presence of solids clusters or Supplementary material related to this article can be found, in the
void areas. The flow patterns for both phases in the column exhibit online version, at doi:https://doi.org/10.1016/j.jece.2020.104674.
recirculation zones in the bed section, and after this, the fluid flow
quickly becomes fully developed. These chaotic recirculation regions References
induce high values of shear stress, which could cause biofilm detach­
[1] M.M. Bello, A.A. Abdul Raman, M. Purushothaman, Applications of fluidized bed
ment from the particle support. At lower input liquid velocities, lower reactors in wastewater treatment – a review of the major design and operational

9
J. Ramírez-Muñoz et al. Journal of Environmental Chemical Engineering 9 (2021) 104674

parameters, J. Clean. Prod. 141 (2017) 1492–1514, https://doi.org/10.1016/j. [28] Y. Song, et al., Numerical study on liquid-solid flow characteristics in inverse
jclepro.2016.09.148. circulating fluidized beds, Adv. Powder Technol. 30 (2) (2019) 317–329, https://
[2] M.J. Brackin, et al., Laboratory-scale evaluation of fluidized bed reactor technology doi.org/10.1016/j.apt.2018.11.009.
for biotreatment of maleic anhydride process wastewater, J. Ind. Microbiol. [29] A. Jaberi, The Hydrodynamic Behavior of an Inverse Liquid-solid Circulating
Biotechnol. 16 (4) (1996) 216–223, https://doi.org/10.1007/BF01570024. Fluidized Bed, The University of Western Ontario, 2014.
[3] M. Andalib, et al., Performance of an anaerobic fluidized bed bioreactor (AnFBR) [30] T. Nan, Hydrodynamics of Liquid-Solid and Three-Phase Inverse Circulating
for digestion of primary municipal wastewater treatment biosolids and bioethanol Fluidized Beds, The University of Western Ontario, 2019.
thin stillage, Renew. Energy 71 (2014) 276–285, https://doi.org/10.1016/j. [31] A. Alvarado-Lassman, et al., Strategies for the startup of methanogenic inverse
renene.2014.05.039. fluidized-bed reactors using colonized particles, Water Environ. Res. 82 (5) (2010)
[4] M. Ahmadi, B. Ramavandi, S. Sahebi, Efficient degradation of a biorecalcitrant 387–391, https://doi.org/10.2175/106143009x12487095237233.
pollutant from wastewater using a fluidized catalyst-bed reactor, Chem. Eng. [32] M.J.H. Khan, et al., CFD simulation of fluidized bed reactors for polyolefin
Comm. 202 (8) (2015) 1118–1129, https://doi.org/10.1080/ production – a review, J. Ind. Eng. Chem. 20 (6) (2014) 3919–3946, https://doi.
00986445.2014.907567. org/10.1016/j.jiec.2014.01.044.
[5] W. Sokół, A. Ambaw, B. Woldeyes, Biological wastewater treatment in the inverse [33] V.V. Ranade, Computational Flow Modeling for Chemical Reactor Engineering,
fluidised bed reactor, Chem. Eng. J. 150 (1) (2009) 63–68, https://doi.org/ Elsevier, 2001.
10.1016/j.cej.2008.12.021. [34] H. Pan, et al., CFD simulations of gas–liquid–solid flow in fluidized bed reactors —
[6] W. Sokół, W. Korpal, Aerobic treatment of wastewaters in the inverse fluidised bed a review, Powder Technol. 299 (2016) 235–258, https://doi.org/10.1016/j.
biofilm reactor, Chem. Eng. J. 118 (3) (2006) 199–205, https://doi.org/10.1016/j. powtec.2016.05.024.
cej.2005.11.013. [35] M.F. Al-Adel, D.A. Saville, S. Sundaresan, The effect of static electrification on
[7] D.G. Karamanev, L.N. Nikolov, Application of inverse fluidization in wastewater gas− Solid flows in vertical risers, Ind. Eng. Chem. Res. 41 (25) (2002) 6224–6234,
treatment: from laboratory to full-scale bioreactors, Environ. Prog. Sustain. Energy https://doi.org/10.1021/ie010982w.
(1996) 194–196, https://doi.org/10.1002/ep.670150319. Medium: X; Size. [36] Y.T. Shih, D. Gidaspow, D. Wasan, Hydrodynamics of electroluidization: separation
[8] H.-D. Han, et al., Phase hold-up and critical fluidization velocity in a three-phase of pyrites from coal, AIChE J. 33 (8) (1987) 1322–1333, https://doi.org/10.1002/
inverse fluidized bed, Korean J. Chem. Eng. 20 (1) (2003) 163–168, https://doi. aic.690330809.
org/10.1007/BF02697203. [37] H. Enwald, E. Peirano, A.E. Almstedt, Eulerian two-phase flow theory applied to
[9] L.-S. Fan, K. Muroyama, S.-H. Chern, Hydrodynamic characteristics of inverse fluidization, Int. J. Multiph. Flow. 22 (1996) 21–66, https://doi.org/10.1016/
fluidization in liquid—solid and gas—liquid—solid systems, Chem. Eng. J. 24 (2) S0301-9322(96)90004-X.
(1982) 143–150, https://doi.org/10.1016/0300-9467(82)80029-4. [38] D.A. Drew, L.A. Segel, Averaged equations for two-phase flows, Stud. Appl. Math.
[10] D.H. Sur, M. Mukhopadhyay, Process aspects of three-phase inverse fluidized bed 50 (3) (1971) 205–231, https://doi.org/10.1002/sapm1971503205.
bioreactor: a review, J. Environ. Chem. Eng. 5 (4) (2017) 3518–3528, https://doi. [39] R.G. Patel, et al., Verification of Eulerian–Eulerian and Eulerian–Lagrangian
org/10.1016/j.jece.2017.06.052. simulations for turbulent fluid–particle flows, AIChE J. 63 (12) (2017) 5396–5412,
[11] A.K. Swain, et al., Industrial wastewater treatment by aerobic inverse fluidized bed https://doi.org/10.1002/aic.15949.
biofilm reactors (AIFBBRs): a review, J. Water Process Eng. 23 (2018) 61–74, [40] D. Gidaspow, Multiphase Flow and Fluidization: Continuum and Kinetic Theory
https://doi.org/10.1016/j.jwpe.2018.02.017. Descriptions, Academic press, 1994.
[12] A. Alvarado-Lassman, et al., Brewery wastewater treatment using anaerobic [41] E. Cruz, F.R. Steward, T. Pugsley, New closure models for CFD modeling of high-
inverse fluidized bed reactors, Bioresour. Technol. 99 (8) (2008) 3009–3015, density circulating fluidized beds, Powder Technol. 169 (3) (2006) 115–122,
https://doi.org/10.1016/j.biortech.2007.06.022. https://doi.org/10.1016/j.powtec.2006.08.005.
[13] L.C. Reyes-Alvarado, et al., Lignocellulosic biowastes as carrier material and slow [42] D.A. Drew, R.T. Lahey, The virtual mass and lift force on a sphere in rotating and
release electron donor for sulphidogenesis of wastewater in an inverse fluidized straining inviscid flow, Int. J. Multiph. Flow. 13 (1) (1987) 113–121, https://doi.
bed bioreactor, Environ. Sci. Pollut. Res. 25 (6) (2018) 5115–5128, https://doi. org/10.1016/0301-9322(87)90011-5.
org/10.1007/s11356-017-9334-5. [43] M. Syamlal, T.J. O’Brien, Simulation of granular layer inversion in liquid fluidized
[14] D.G. Karamanev, L.N. Nikolov, Bed expansion of liquid-solid inverse fluidization, beds, Int. J. Multiph. Flow. 14 (4) (1988) 473–481, https://doi.org/10.1016/0301-
AIChE J. 38 (12) (1992) 1916–1922, https://doi.org/10.1002/aic.690381208. 9322(88)90023-7.
[15] T. Renganathan, K. Krishnaiah, Voidage characteristics and prediction of bed [44] T.B. Anderson, R. Jackson, Fluid mechanical description of fluidized beds.
expansion in liquid–solid inverse fluidized bed, Chem. Eng. Sci. 60 (10) (2005) Equations of motion, Ind. Eng. Chem. Fundamen. 6 (4) (1967) 527–539, https://
2545–2555, https://doi.org/10.1016/j.ces.2004.11.034. doi.org/10.1021/i160024a007.
[16] A.C. Vijaya Lakshmi, et al., Minimum fluidization velocity and friction factor in a [45] C.K.K. Lun, et al., Kinetic theories for granular flow: inelastic particles in Couette
liquid-solid inverse fluidized bed reactor, Bioprocess Eng. 22 (5) (2000) 461–466, flow and slightly inelastic particles in a general flowfield, J. Fluid Mech. 140
https://doi.org/10.1007/s004490050759. (2006) 223–256, https://doi.org/10.1017/S0022112084000586.
[17] R.J. Femin Bendict, G. Kumaresan, M. Velan, Bed expansion and pressure drop [46] P.C. Johnson, R. Jackson, Frictional–collisional constitutive relations for granular
studies in a liquid-solid inverse fluidised bed reactor, Bioprocess Eng. 19 (2) (1998) materials, with application to plane shearing, J. Fluid Mech. 176 (2006) 67–93,
137–142, https://doi.org/10.1007/s004490050494. https://doi.org/10.1017/S0022112087000570.
[18] N. Ulaganathan, K. Krishnaiah, Hydrodynamic characteristics of two-phase inverse [47] R. Perry, D. Green, J. Maloney, Perry’s Chemical Engineers’ Handbook, McGraw-
fluidized bed, Bioprocess Eng. 15 (3) (1996) 159–164, https://doi.org/10.1007/ Hill, New York, 1997.
BF00369620. [48] S. Cloete, S. Amini, S.T. Johansen, On the effect of cluster resolution in riser flows
[19] B. Das, et al., Holdup prediction in inverse fluidization using non-Newtonian on momentum and reaction kinetic interaction, Powder Technol. 210 (1) (2011)
pseudoplastic liquids: empirical correlation and ANN modeling, Powder Technol. 6–17, https://doi.org/10.1016/j.powtec.2011.02.003.
273 (2015) 83–90, https://doi.org/10.1016/j.powtec.2014.12.034. [49] K.E. Campos-Díaz, E.R. Bandala-González, R. Limas-Ballesteros, Fluid bed porosity
[20] R.W. Samstag, et al., CFD for wastewater treatment: an overview, Water Sci. mathematical model for an inverse fluidized bed bioreactor with particles growing
Technol. 74 (3) (2016) 549–563, https://doi.org/10.2166/wst.2016.249. biofilm, J. Environ. Manage. 104 (2012) 62–66, https://doi.org/10.1016/j.
[21] C.D.’ Bastiani, et al., Three-phase hydrodynamic simulation and experimental jenvman.2012.03.019.
validation of an upflow anaerobic sludge blanket reactor, Comput. Math. Appl. [50] Y.J. Cho, et al., Heat transfer and hydrodynamics in two- and three-phase inverse
(2020), https://doi.org/10.1016/j.camwa.2020.02.017. fluidized beds, Ind. Eng. Chem. Res. 41 (8) (2002) 2058–2063, https://doi.org/
[22] I. Nopens, et al., Water and wastewater CFD and validation: are we losing the 10.1021/ie0108393.
balance? Water Sci. Technol. 81 (8) (2020) 1636–1645, https://doi.org/10.2166/ [51] O. Sánchez, et al., Liquid mixing and gas–liquid mass transfer in a three-phase
wst.2020.181. inverse turbulent bed reactor, Chem. Eng. J. 114 (1) (2005) 1–7, https://doi.org/
[23] L. Yu, et al., Multiphase modeling of settling and suspension in anaerobic digester, 10.1016/j.cej.2005.08.009.
Appl. Energy 111 (2013) 28–39, https://doi.org/10.1016/j.apenergy.2013.04.073. [52] B.M. Peyton, W.G. Characklis, A statistical analysis of the effect of substrate
[24] S. Das, S. Sarkar, S. Chaudhari, Modification of UASB reactor by using CFD utilization and shear stress on the kinetics of biofilm detachment, Biotechnol.
simulations for enhanced treatment of municipal sewage, Water Sci. Technol. 77 Bioeng. 41 (7) (1993) 728–735, https://doi.org/10.1002/bit.260410707.
(3) (2017) 766–776, https://doi.org/10.2166/wst.2017.584. [53] Y. Liu, J.-H. Tay, The essential role of hydrodynamic shear force in the formation of
[25] H. Pan, et al., CFD optimization of the baffle angle of an expanded granular sludge biofilm and granular sludge, Water Res. 36 (7) (2002) 1653–1665, https://doi.org/
bed reactor, J. Environ. Chem. Eng. 5 (5) (2017) 4531–4538, https://doi.org/ 10.1016/S0043-1354(01)00379-7.
10.1016/j.jece.2017.07.050. [54] H. Horn, H. Reiff, E. Morgenroth, Simulation of growth and detachment in biofilm
[26] S. Wang, et al., Numerical simulation of flow behavior of particles in an inverse systems under defined hydrodynamic conditions, Biotechnol. Bioeng. 81 (5) (2003)
liquid–solid fluidized bed, Powder Technol. 261 (2014) 14–21, https://doi.org/ 607–617, https://doi.org/10.1002/bit.10503.
10.1016/j.powtec.2014.04.017. [55] V. Diez Blanco, P.A.G. Encina, F. Fdz-Polanco, Effects of biofilm growth, gas and
[27] S. Wang, et al., Numerical simulation of flow behavior of particles in an inverse liquid velocities on the expansion of an anaerobic fluidized bed reactor (AFBR),
liquid–solid fluidized bed with a jet using CFD-DEM, J. Taiwan Inst. Chem. Eng. 82 Water Res. 29 (7) (1995) 1649–1654, https://doi.org/10.1016/0043-1354(95)
(2018) 214–225, https://doi.org/10.1016/j.jtice.2017.08.049. 00001-2.
[56] M.C. Ruzicka, On dimensionless numbers, Chem. Eng. Res. Des. 86 (8) (2008)
835–868, https://doi.org/10.1016/j.cherd.2008.03.007.

10