- Email: [email protected]
CFD-aided modelling of activated sludge systems – A critical review
Accepted Manuscript CFD-Aided Modelling of Activated Sludge Systems – A Critical Review Anna M. Karpinska, John Bridgeman PII:
S0043-1354(15)30336-5
DOI:
10.1016/j.watres.2015.11.008
Reference:
WR 11635
To appear in:
Water Research
Received Date: 25 June 2015 Revised Date:
1 November 2015
Accepted Date: 2 November 2015
Please cite this article as: Karpinska, A.M., Bridgeman, J., CFD-Aided Modelling of Activated Sludge Systems – A Critical Review, Water Research (2015), doi: 10.1016/j.watres.2015.11.008. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT 1
Title: CFD-Aided Modelling of Activated Sludge Systems –
2
A Critical Review.
3 4
Authors: Anna M. Karpinskaa, John Bridgemana
5
a
6
Edgbaston, Birmingham, B15 2TT, United Kingdom
7
Email
8
[email protected]
9
Corresponding
RI PT
[email protected];
author:
Anna
([email protected])
SC
addresses:
M.
Karpinska
M AN U
10
– School of Civil Engineering, University of Birmingham,
11 Abstract
13
Nowadays, one of the major challenges in the wastewater
14
sector is the successful design and reliable operation of
15
treatment
16
efficiencies to comply with effluent quality criteria, while
17
keeping the investment and operating cost as low as possible.
18
Although conceptual design and process control of activated
TE D
12
which
guarantee
high
treatment
AC C
EP
processes,
19
sludge plants are key to ensuring these goals, they are still
20
based
21
experience, dominated often by rule of thumb. This review
22
paper discusses the rationale behind the use of Computational
23
Fluid Dynamics (CFD) to model aeration, facilitating
24
enhancement of treatment efficiency and reduction of energy
25
input. Several single- and multiphase approaches commonly
on
general
empirical
1
guidelines
and
operators’
ACCEPTED MANUSCRIPT used in CFD studies of aeration tank operation, are
27
comprehensively described, whilst the shortcomings of the
28
modelling assumptions imposed to evaluate mixing and mass
29
transfer in AS tanks are identified and discussed. Examples and
30
methods of coupling of CFD data with biokinetics, accounting
31
for the actual flow field and its impact on the oxygen mass
32
transfer and yield of the biological processes occurring in the
33
aeration tanks, are also critically discussed. Finally, modelling
34
issues, which remain unaddressed, (e.g. coupling of the AS
35
tank with secondary clarifier and the use of population balance
36
models to simulate bubbly flow or flocculation of the activated
37
sludge), are also identified and discussed.
M AN U
SC
RI PT
26
38 Keywords
40
Activated Sludge; aeration; Computational Fluid Dynamics;
41
hydrodynamics; multiphase models; turbulence
Nomenclature
AC C
43
EP
42
TE D
39
2
ACCEPTED MANUSCRIPT mass fraction of phase [-]
sum of interfacial forces shared by the phases [N]
sum of interfacial forces between the continuous and dispersed phases [N] diffusion flux of ′ species [kg m-2 s-1]
turbulent kinetic energy, [m2 s-2]
M AN U
SC
RI PT
volumetric mass transfer coefficient [h-1]
mass transfer from phase to [kg s-1]
mass transfer from phase to [kg s-1]
TE D
pressure [Pa]
̅
averaged pressure field [Pa]
AC C
EP
net rate of production of ′ species due to chemical
Reynolds number [-]
reaction [kg m-3 s-1]
3
ACCEPTED MANUSCRIPT
source term representing rate of creation of ′ species from dispersed phase and any user-defined sources [kg
time [s]
particle relaxation time [s]
velocity [m s-1]
, ,
drift velocity for phase [m s-1]
M AN U
,
SC
RI PT
m-3 s-1]
velocity components [m s-1]
averaged velocity term [m s-1]
′
fluctuating velocity term [m s-1]
velocity of mixture [m s-1]
AC C
EP
TE D
̅
velocity of the particle [m s-1]
velocity of phase [m s-1]
velocity of phase relative to velocity of phase [m s-1]
4
ACCEPTED MANUSCRIPT velocity of phase [m s-1]
distance [m]
,
direction) [m] mass fraction of the ’-th species [-]
SC
!
spatial coordinates (displacement in the streamwise
Greek letters
"∗ $
damping function coefficient [-]
turbulent kinetic energy dissipation rate [m2 s-3] dynamic viscosity of the fluid [Pa s]
EP
%
phasic volume fraction [-]
TE D
"
M AN U
44 45
RI PT
,
dynamic viscosity of the mixture [Pa s]
AC C
% % %& '
'
dynamic viscosity of the phase [Pa s] turbulent viscosity [Pa s]
fluid density [kg m-3] density of the mixture [kg m-3]
5
ACCEPTED MANUSCRIPT '
density of the particle [kg m-3] density of the phase [kg m-3]
(
specific turbulence dissipation (frequency)
'
density of the phase [kg m-3]
[s-1]
)*
′ +
,
refers to drift velocity index or counter species index
index or counter
refers to turbulence kinetic energy index or counter
refers to mixture
EP
-
M AN U
Indices
TE D
47
SC
46
refers to particle
refers to phase
refers to turbulence
refers to phase
AC C
RI PT
'
48 49 50
Abbreviations
51
ADV – Acoustic Doppler Anemomentry
52
AS – Activated Sludge 6
54
ASM1 – Activated Sludge Model No. 1
55
ASP – Activated Sludge Plant
56
BOD – Biochemical Oxygen Demand
57
CFD – Computational Fluid Dynamics
58
COD – Chemical Oxygen Demand
59
CPU – Central Processing Unit
60
CSTR – Continuous Stirred Tank Reactor
61
DO – Dissolved Oxygen
62
LDV – Laser Doppler Velocimetry
63
LES – Large |Eddy Simulation
64
MDV– Mono-directional Velocimetry
65
MLSS – Mixed Liquor Suspended Solids
66
MRF – Multiple Reference Frames
67
PBM – Particle Balance Model
68
PDA – Particle Dynamic Analysis
69
PFR – Plug Flow Reactor
70
PID – Proportional-Integral-Derivative
SC
ASM – Activated Sludge Model
AC C
EP
TE D
M AN U
53
RI PT
ACCEPTED MANUSCRIPT
71
QMOM – Quadrature Method of Moments
72
RAM – Random Access Memory
73
RANS – Reynolds Averaged Navier- Stokes Simulation
74
RSM – Reynolds Stress Model
75
RTD – Residence Time Distribution
76
SCADA – Supervisory Control and Data Acquisition
77
SGS – Sub-grid Scale
7
ACCEPTED MANUSCRIPT SMM – Standard Method of Moments
79
SRT – Solids Retention Time
80
TSS – Total Suspended Solids
81
URANS – Unsteady Reynolds Averaged Navier- Stokes
82
Simulation
83
WWTP – Wastewater Treatment Plant
84
RI PT
78
1. Introduction
86
To comply with global water policy focussed on responsible
87
management of water resources and protection of public health,
88
wastewater collected from municipalities and communities
89
must be treated to achieve levels imposed by discharge permits
90
and maximum daily loads, allowing its subsequent return to
91
receiving water bodies, or to the land or even to be reused. In
92
the last century, the application of scientific knowledge and
93
engineering practice led to significant developments in the
94
wastewater
95
treatment based on aerobic biological methods, specifically the
96
activated sludge (AS) process (Ardern and Lockett 1914),
97
which is now a well-documented standard for many wastewater
98
treatment utilities. The objectives of secondary biological
99
wastewater treatment in the AS process have also expanded
100
from an early emphasis on high levels of Biological Oxygen
101
Demand (BOD) and Total Suspended Solids (TSS) removal to
102
cover enhanced nutrients (N and P) removal, as the process
TE D
M AN U
SC
85
particularly in
biological
secondary
AC C
EP
sector,
8
ACCEPTED MANUSCRIPT itself has flexibility and numerous modifications can be tailored
104
to meet specific requirements.
105
Undoubtedly, one of the foremost challenges in the wastewater
106
sector is the successful design and reliable operation of
107
treatment plant, which guarantee high treatment efficiencies in
108
order to meet the effluent quality standards defined by
109
regulators, while keeping the investment and operating cost as
110
low as possible (Brouckaert and Buckley 1999, Do-Quang et al.
111
1999). One of the characteristic features of ASP is continuous
112
operation of the aeration process and sludge and nitrates
113
recycling, and thus the process performance relies on a steady
114
energy supply for operation of air blowers and sludge and
115
mixed liquor recirculation pumps. Aeration accounts for the
116
largest fraction of a total wastewater treatment plant’s (WWTP)
117
energy bill, ranging from 45 to 75% (Reardon 1995, Rieger et
118
al. 2006) and in extreme cases of stringent effluent nitrogen
119
criterion requiring enhanced nitrification, even up to 85%.
120
Thus, aeration has a significant effect on the operation and
AC C
EP
TE D
M AN U
SC
RI PT
103
121
maintenance budget of water utilities (WEF 2009). As a
122
consequence of the global emphasis on the water-energy-food-
123
climate change nexus there is an urgent need to reduce energy
124
usage
125
conservation measures, engineering practices and management
126
programs. According to guidelines (EPA 2013, WEF 2009)
127
opportunities for improving energy efficiency in wastewater
at
WWTPs
by
imposing
9
cost-effective
energy
ACCEPTED MANUSCRIPT treatment utilities can be obtained by optimizing aeration
129
processes or equipment upgrades, which focus on replacing
130
items such as blowers with more efficient models; replacing the
131
whole aeration system with less energy intensive systems (e.g.
132
replacement of the surface aeration system by porous diffusers
133
in full floor coverage configuration); and operational
134
modifications, involving reduction of the energy requirements
135
to perform specific functions by modification of the aeration
136
control systems, e.g. on-off operation allowing formation of the
137
anoxic conditions for denitrification. The last option facilitates
138
greater savings than equipment upgrades, and may not require
139
capital investment. Nevertheless, current best available aeration
140
technology (i.e. membrane diffusers supplied by atmospheric
141
air) are characterized by relatively low Standard Oxygen
142
Transfer Efficiencies of around 40 up to 60% (EPA 1989,
143
Mueller et al. 2002, Taricska et al. 2009), which reduce with
144
time as a result of fouling and scaling. Consequently, a wide
145
range of multidisciplinary approaches contribute to current
AC C
EP
TE D
M AN U
SC
RI PT
128
146
research aiming to improve the development, troubleshooting
147
and management of aeration systems.
148
2. Current Trends in Engineering Practice
149
2.1. Aeration Control and Design Assumptions
150
When investigating performance and energy expenditure of AS
151
plants it is clear that dissolved oxygen (DO) concentration is a
152
key process variable, which controls both, nutrient removal (in 10
ACCEPTED MANUSCRIPT the case of biological nutrient removal plants) and thus effluent
154
quality, and the operating cost of the utility. While operating
155
DO profiles and nitrogen patterns in the AS system are usually
156
obtained from biokinetics modelling with Activated Sludge
157
Model No. 1- ASM1 (Henze et al. 2000) employed by
158
supervisory control and data acquisition (SCADA) systems, the
159
robust modelling of the optimization strategies requires
160
implementation of complex computational algorithms and off-
161
line optimization techniques, which allow for coupling of the
162
biological process with pre-defined control variables such as
163
minimal DO concentration in the aeration tank or effluent
164
nitrogen criterion (Åmand and Carlsson 2012, Chachuat et al.
165
2005, Cristea et al. 2011, De Araújo et al. 2011, Fernández et
166
al. 2011, Fikar et al. 2005, Holenda et al. 2007, Holenda et al.
167
2008).
168
Although hydraulic design of wastewater treatment systems is a
169
crucial step to assure reliable and energy-optimised operation
170
of the process, it is usually labelled as a “low-tech” task, which
AC C
EP
TE D
M AN U
SC
RI PT
153
171
is based on empirical guidelines without sound theoretical basis
172
(Bosma and Reitsma 2007, Pereira et al. 2012, Stamou 2008).
173
Thus, in the majority of designs, flow behaviour in the unit
174
process tanks is predicted via the ideal reactor model while the
175
actual reactor hydrodynamics are not taken into account
176
(Stamou 2008). The classic example for Activated Sludge
177
Plants (ASPs) is the assumption of a completely mixed flow 11
ACCEPTED MANUSCRIPT regime in aerobic, anoxic and anaerobic tanks. Other
179
commonly practised rules of thumb are (Samstag and Wicklein
180
2012, Tchobanoglous et al. 2003) the assumption that in AS
181
basins equipped with diffused aeration systems, the air
182
requirement to ensure good mixing will vary from 1.2 – 1.8 m3
183
h-1 / m3 of tank volume; and typical power requirements for
184
maintaining a completely mixed flow regime with mechanical
185
aerators varies from 13 to 26 W per m3 of tank volume. None
186
of these assumptions consider the impact of tank hydraulics
187
(cross-section, depth or presence of baffles), energy input, or
188
any variable affecting mixing, such as local density gradients
189
due to solids transport. Furthermore, none of these guidelines
190
defines clearly hydraulic features and performance of
191
“completely mixed” wastewater treatment systems. The one
192
commonly used criterion for “good mixing” in AS process
193
control is that variations of solids concentration across the
194
complete mixed tank profile should be less than 10% (Samstag
195
and Wicklein 2014) However, the proper design of such “well
AC C
EP
TE D
M AN U
SC
RI PT
178
196
mixed” AS systems requires a sophisticated analysis of flow
197
behaviour accounting for mixing patterns in the tank,
198
distribution of the oxygen, solids and determination of local
199
densities.
200
3. Dynamic Behaviour of AS Tanks
201
Dynamic modelling of wastewater treatment plants has been
202
shown to be a powerful tool providing detailed insight into the 12
ACCEPTED MANUSCRIPT 203
unit process and system behaviour, useful for optimization
204
studies
205
evaluation, operational optimization, or controller design) and
206
model-based process control (Langergraber et al. 2004).
207
Nonetheless, the majority of the ASP design procedures are still
208
limited to the empirical principles and static models, such as
209
the ‘classic’ ATV-A-131 guideline (ATV-DVWK 2000), while
210
control strategies are based on a simple or cascade
211
proportional-integral-derivative (PID) controllers for DO and
212
ammonia. While not often used for design and control of ASP
213
due to its complexity, the systemic approach based on well-
214
established ASM models focuses mainly on the reactions of
215
biochemical conversion within the ideal reactors: usually one or
216
a cascade of Continuous Stirred Tank Reactors (CSTRs) or
217
Plug Flow Reactor (PFR) (Abusam and Keesman 1999, Le
218
Moullec et al. 2011, Makinia and Wells 2000, Pereira et al.
219
2009). Such approaches enable only quantitative prediction of
220
the oxygen consumption and nutrients removal and qualitative
conceptual
process
design,
performance
AC C
EP
TE D
M AN U
SC
RI PT
(e.g.
221
assessment of biomass growth and decay. Nevertheless, as the
222
overall biochemical conversion reactions occurring in AS are of
223
orders greater than zero, the wastewater treatment efficiency in
224
such non-ideal systems will depend on the bioreactor’s
225
hydrodynamics (Le Moullec et al. 2008), spatial distribution of
226
oxygen, and temperature. Therefore, a more accurate forecast
227
of the actual local scale phenomena occurring within the tank,
13
ACCEPTED MANUSCRIPT (such as DO patterns and solids profiles) is of crucial
229
importance to enhance system design, process efficiency and
230
energy requirements, to predict failures of existing aeration
231
systems and finally, to design upgrades and enhance control
232
strategy, such as fine-tuning blower operation.
233
The physics of typical AS systems is complex, not only due to
234
presence of the multiphase (gas-liquid-solid) flow, comprising
235
mixed liquor and air/oxygen, but also due to the different
236
length scales between the sludge flocs, bubbles and tank
237
geometry; and furthermore, different velocities of the phases
238
imparted by mixers and aerators yielding turbulent Reynolds
239
numbers (Karpinska Portela 2013, Pereira et al. 2012). The
240
fluid flow in such bioreactors is governed by the vessel
241
geometry, physical properties of its contents (phase, density,
242
viscosity) and operating condition variables (flow rates and
243
concentrations). At the same time, fluid flow governs the local
244
concentration of components, interphase contact and mass
245
transfer, reaction conversions and performance (Nopens and
AC C
EP
TE D
M AN U
SC
RI PT
228
246
Wicks 2012).
247
In engineering practice, whilst assessment of the flow regime,
248
and thus overall mixing phenomena in AS bioreactors, can be
249
achieved via experimental assessment of local flow velocities
250
through a tracer technique, the dimensions of full scale units
251
generally render this unfeasible (Pereira et al. 2012, Stamou
14
ACCEPTED MANUSCRIPT 2008), making simulation an attractive alternative. Several
253
factors contribute to the increasing popularity of modelling in
254
engineering practice, as it is a cost and time efficient solution,
255
which allows evaluation of process performance (whether a
256
new unit or modification will operate properly); prediction of
257
consequences
258
quantification of bottlenecks in liquid or solid handling lines in
259
the AS system. Thus, prediction of Residence Time
260
Distributions (RTDs) of settled sewage is a fundamental tool to
261
help understand and analyse a flow system providing realistic
262
information on tank hydrodynamics (Danckwerts 1953,
263
Levenspiel 1999, Nauman 2007). Furthermore, the RTD yields
264
information about the macromixing within the reactor, allowing
265
recognition of mixing behaviours that are not plug flow or
266
complete mixing regimes, and that are usually described by
267
tank-in-series models or even more complex arrangements of
268
the unit reactors (Pereira et al. 2012). Consequently, successful
269
biological
implementation;
isolation
and
EP
TE D
M AN U
SC
before
RI PT
252
treatment
modelling
combining
AC C
wastewater
270
hydrodynamics, mass transfer and biochemical reactions
271
kinetics remains one of the major goals in wastewater
272
engineering (Le Moullec et al. 2010a, b, Morchain et al. 2014,
273
Pereira et al. 2012).
274
4. Computational Fluid Dynamics
275
Since the 1970s, increasing computational power has been
276
accompanied by a rapid development in the software intended 15
ACCEPTED MANUSCRIPT 277
for solution of fluid flow problems. Nowadays, a wide range of
278
software suites intended for the solution of complex fluid flow
279
problems is commercially available. Initially, CFD was almost
280
exclusively
281
industries allowing simulation of the processes occurring in
282
combustion chambers of rocket engines; physico-chemical
283
processes in the flow around rocket airframe and supersonic
284
aircrafts. Subsequently CFD found applications with chemical
285
engineers, mainly for design of reaction vessels, and in the last
286
two decades, application of CFD has been extended to the civil
287
and environmental engineering sectors (Kochevsky 2004).
288
Recent developments in multiphase flow research have seen a
289
steady growth in the application of CFD modelling in
290
wastewater treatment, with a focus on the design of pumping
291
stations, headworks, screens, grit chambers, flow splitters, AS
292
tanks, clarifiers and digesters. Undoubtedly, one of the great
293
opportunities of CFD modelling of ASPs is the analysis of the
294
multiphase flow behaviour and prediction of the impact of a
with
aerospace
and
mechanical
AC C
EP
TE D
M AN U
SC
RI PT
associated
295
wide range of operating parameters on the local scale
296
phenomena, such as flow field coupled with interfacial mass
297
transfer and chemical reaction. In addition to that, CFD has
298
gained popularity over traditional wastewater treatment
299
modelling approaches, as it is a high-precision technique
300
allowing evaluation of the engineering systems, which are
301
expensive, difficult or even dangerous to reproduce in
16
ACCEPTED MANUSCRIPT laboratory-scale, pilot-scale or field conditions. Therefore,
303
CFD-aided modelling can be used as a robust tool for the
304
design of a new facility or the optimization or retrofitting of
305
existing ASPs, leading to enhanced performance and energy-
306
optimised operation of the utility, facilitating time, economic
307
cost and manpower savings (De Gussem et al. 2014, Do-Quang
308
et al. 1999, Essemiani et al. 2004, Guimet et al. 2004, Laurent
309
et al. 2014).
310
4.1. Hydrodynamics – RANS and URANS
311
Various options exist for the numerical simulation of the
312
turbulent flow with CFD codes.
313
In most engineering practice, time-averaged properties of the
314
flow are able to provide the required information. Steady
315
Reynolds Averaged Navier-Stokes (RANS) simulations and
316
unsteady RANS (URANS) focus on the representation of the
317
effects of turbulence on the mean flow properties by solving
318
transport equations for the averaged flow quantities with whole
319
range of the turbulent scales being modelled. Thus this
AC C
EP
TE D
M AN U
SC
RI PT
302
320
modelling approach greatly reduces required computational
321
effort and resources, and is widely adopted for practical
322
engineering applications, including the wastewater sector
323
(Karpinska Portela 2013).
324
In RANS and URANS the flow patterns within the AS tank are
325
obtained from the solution of nonlinear partial differential
326
equations, expressing balances of mass and momentum. 17
ACCEPTED MANUSCRIPT 327
Therefore, the flow is governed by the following mass
328
conservation equation: .' =0 .
RI PT
329
(1)
and momentum conservation equation, which for RANS is:
330
(2)
M AN U
. + <−' = .
SC
. .̅ . . . 2 . 1' 2 = − + 5% 6 + − 9 :; . . . . . 3 .
and for time dependent, transient flow (URANS) is:
TE D
.̅ . + 1' 2 . .
=−
.̅ . . . 2 . + 5% 6 + − 9 :; . . . . 3 . (3)
. <−' = .
EP AC C
+
331
332 333 334
where ̅ is the averaged pressure field, ' and % are the fluid
density and viscosity, respectively, is the time,
,
and
are the spatial coordinates, , and are the velocity components and 9 is the Kronecker delta.
18
ACCEPTED MANUSCRIPT 335
Here, the turbulent mean velocity field is described by
336
Reynolds decomposition of the velocity using time averaged
337
term, ̅ , and a fluctuating term, : = ̅ +
RI PT
(4)
RANS and URANS equations are linearized and solved. The
339
flow structures originating from the momentum transfer by the
340
fluctuating velocity field, which are smaller than the numerical
341
grid discretization and represented by the term
342
namely Reynolds stresses, are unclosed, and thus they must be
343
modelled.
344
Turbulence models
345
The
346
commercial CFD software suites, which are used for Reynolds
347
stresses closure are: standard, renormalized group (RNG) and
M AN U models
included
in
popular
TE D
turbulence
realizable − $ models; standard and shear stress transport
(SST) − ( models; and Reynolds stress model (RSM)
AC C
349
principal
−' ,
EP
348
SC
338
350
(Bridgeman et al. 2009).
351
The two-equation models (i.e. − $ and − (), solve
352
additional transport equations for two turbulence quantities:
353 354 355 356
velocity scale- turbulent kinetic energy, , and turbulence length scale- either its dissipation rate, $, or the specific
frequency, ( (Pope 2000). From among the two-equation
models, the standard − $ (> − $) has found the broadest 19
ACCEPTED MANUSCRIPT range of applicability in both, academia and industrial sectors,
358
due to its robustness, relatively low computational requirements
359
and satisfactory accuracy. The characteristic feature of all two-
360
equation models is that modelling of the Reynolds stresses
361
employs the Boussinesq hypothesis relating these stresses to the
362
mean deformation rates and thus mean velocity gradients, and
363
assuming locally isotropic turbulence (Rodi 1993):
RI PT
357
SC
. . 2 . −' = %& 6 + : − ?' + %& @9 . . 3 .
365
where %& is the turbulent viscosity computed as a function of
M AN U
364
(5)
and $:
TE D
%& = 'AB
C $
(6)
where AB is a model constant (variable function in a different
367
turbulence model).
368
A comprehensive description of the > − $ and the relatively
369
recently developed improved models, namely Renormalized
AC C
EP
366
370
Group (RNG) − $ and realizable − $, can be found in the
371
literature (Launder and Spalding 1974, Orszag et al. 1996, Shih
372
et al. 1995).
373
The second widely used two-equation model, introduced by
374
Wilcox, is the − ( model, also based on the isotropic eddy
20
ACCEPTED MANUSCRIPT
376 377
viscosity hypothesis described by Equation (5), but where the
turbulent viscosity is computed as a function of and (.
Knowing, that ( = $ ⁄ , Equation (6) takes the following form: %& = '" ∗
(
(7)
RI PT
375
where " ∗ is a coefficient related to the use of functions
379
damping the turbulent viscosity causing a low-Reynolds-
380
number correction.
381
Contrary to the − $ model, the standard − ( model uses
382
enhanced wall treatment to solve low-Re-number flows in the
383
viscous layer in near-wall region. Its improved variant, SST
M AN U
SC
378
− ( model, considered to be the most accurate from two-
385
equation eddy viscosity models, provides modified turbulent
386
viscosity formulation which account for the transport of the
387
principal turbulent shear stress.
TE D
384
A wider discussion on the
standard and SST − ( models can be found in Wilcox
389
(Wilcox 1998, Wilcox and Traci 1976) and Menter (Menter
390
1993, 1994).
391
The Reynolds Stress Model (RSM) (Launder et al. 1975,
392
Versteeg and Malalasekera 1995) is the most elaborate and
393
complex turbulence model, referred as the Second Order
394
Closure. In RSM the isotropic eddy viscosity hypothesis is
395
discarded and the RANS equations are closed by solving
396
transport equations for the Reynolds stresses, together with an
AC C
EP
388
21
ACCEPTED MANUSCRIPT equation for the dissipation rate, yielding seven additional
398
transport equations to be solved in a 3D scheme. A detailed
399
description of the RSM turbulent closure development can be
400
found in the literature (Launder et al. 1975), whilst a
401
summarised description of the turbulence models commonly
402
used for RANS closure has been summarized in Table 1
403
(Bridgeman et al. 2009).
AC C
EP
TE D
M AN U
SC
RI PT
397
22
ACCEPTED MANUSCRIPT
Table 1. Comparison of turbulence models used in ASP modelling. Adapted from Bridgeman et al. (2009) with permission from Taylor &
405
Francis Ltd (www.tandfonline.com/) Standard E − F (GE − F)
Comments
Renormalized Group (RNG) E−F
• Based on the statistical methods, not observed
Advantages
• Semi-empirical modelling of and $. • Valid only for fully developed turbulent flow cores (molecular diffusion ignored).
• Simplest and complete turbulence model. • Excellent performance for many flows. • Well established in academia and industry. • Robust, economic in terms of computational effort and satisfactory accuracy in diverse turbulent flow issues.
• • •
TE D
• Improved performance for swirling
and high-strained flows compared to > − $.
EP
•
fluid behaviour. Mathematics is highly abstruse. Texts only quote model equations which result from it. Effects of small-scale turbulence represented by means of random forcing function in NavierStokes equations. Procedure systematically removes small scales of motion by expressing their effects in terms of larger scale motions and a modified viscosity. Similar in form to > − $, but modified $ equation to describe high-strain flows better. Differential equation solved for %& (changes AB from 0.09 to 0.0845 at high Re).
AC C
•
M AN U
SC
Model
RI PT
404
406
23
Disadvantages • Poor performance in some scenarios (strong streamline curvature, vortices, rotating flows, flow separation, adverse pressure gradients). • Assumes locally isotropic turbulence. • Poor prediction of the lateral expansion in 3D wall jets.
• Less stable than > − $.
ACCEPTED MANUSCRIPT
Table 1. (Continued). Advantages • Suited for planar and rounded jets, swirling and separating flows and wall-bounded flows with strong adverse pressure gradients.
Standard E − H
• Recent development. In highly strained flows, the normal Reynolds stresses become negative (unrealizable condition), so %& uses variable AB . • AB is function of local strain rate and fluid rotation. • Different source and sink terms in transport equations for eddy dissipation. • Good for spreading rate of round jets.
• Specific dissipation rate is ( = $/ . • > − $ solves for dissipation of turbulent kinetic energy, − ( solves for rate at which dissipation occurs. • Resolves near wall region without wall functions so can be applied through boundary layer.
408
SC
M AN U
• As − ( except from gradual change from
− ( to inner region of boundary layer to high Re version of > − $ in outer part. • Modified %& formulation to account for transport effects of principal turbulent shear stresses.
• Not recommended to use with multiple reference frames.
• Valid throughout to boundary layer, subject to fine grid resolution. • Accounts for the stream-wise pressure gradients. • Applicable for detached, separated flows and fully turbulent flows.
• Pressure induced separation is typically predicted to be excessive and early.
• The most accurate from twoequation eddy viscosity models. • Suitable for adverse pressure gradients and pressure-induced flow separation. • Accounts for the transport of the principal shear stresses.
• Less suitable for free shear flows.
TE D
Shear Stress Transport (SST) E − H
Disadvantages
RI PT
Realizable E − F
Comments
EP
Model
AC C
407
409 410 24
ACCEPTED MANUSCRIPT
Table 1. (Continued). Comments
Advantages
Reynolds Stress Model (RSM)
• The most general and complex of all models solving transport of the Reynolds stresses- so called seven-equation model • Isotropic eddy viscosity hypothesis is discarded.
• Accurate calculation of the mean flow properties and all Reynolds stresses. • Accounts for the streamline curvature, rotation and rapid changes in strain rate yielding superior results to two-equation models for complex flows, e.g. with stagnation points.
412
SC
AC C
EP
TE D
413 414
Disadvantages
RI PT
Model
M AN U
411
25
• Computationally expensive. • Not always more accurate than two-equation models. • Harder to obtain converged result. • Reliability of RSM predictions are still limited by the closure assumptions employed to model pressure-strain and dissipationrate terms.
ACCEPTED MANUSCRIPT 4.1.1. Alternative approaches - LES and DNS
416
Large Eddy Simulation (LES) is an alternative approach in
417
hydrodynamic modelling where large, three-dimensional
418
unsteady scale motions affected by the flow geometry, i.e. large
419
eddies, are directly and explicitly solved in time-dependent
420
simulation using space-filtered Navier-Stokes equations. LES is
421
one of the most expensive simulation options and requires
422
a refined grid to accurately resolve eddies in the boundary
423
layer. A filtering operation, analogous to the Reynolds
424
decomposition in RANS, is based on the decomposition of the
SC
M AN U
425
RI PT
415
velocity into the resolved (filtered) component ̅ J , K and the residual, so called subgrid-scale (SGS) component J , K
427
(Pope 2000). The accuracy of the LES model is the result of
428
modelling only the SGS motions- the smallest eddies, which
429
tend to have more universal properties (Karpinska Portela
430
2013). The model commonly used to model small eddies is the
431
Smagorinsky SGS model (Smagorinsky 1963).
432
LES has been most successful for high-end applications where
AC C
EP
TE D
426
433
the RANS models fail to meet the required goals, e.g.
434
modelling of combustion and mixing. Although it provides
435
improved accuracy, wide application of LES approach to solve
436
flow related issues is still limited due to the large mesh
437
requirements and high computational costs.
438
Direct Numerical Simulation (DNS) is a high-fidelity tool
439
offering the explicit solution of the whole range of turbulent 26
ACCEPTED MANUSCRIPT time and length scales, from the Kolmogorov scales to large
441
motion scales transporting most of the kinetic energy within the
442
domain (Orszag 1970). As a result, the computational cost of
443
DNS is extremely high even at low Reynolds numbers,
444
preventing this approach from being used as a general-purpose
445
design tool, and making it impractical for most industrial flow
446
conditions, especially large multiphase AS systems. Hence the
447
CFD modelling of the AS tanks requires more computationally
448
efficient and thus simplified methods.
449
4.2. Multiphase Modelling
450
A robust understanding of the physics and biochemical
451
processes in the gas-liquid-solid environment of the AS tank
452
relies on accurate assessment of the transport phenomena and
453
character of interactions between the phases. Numerical
454
simulation of the multiphase flow in AS system is usually
455
enabled by Euler-Euler or Euler-Lagrange approaches. Table 2
456
outlines the relevant characteristics of the multiphase models
457
used in modelling of the AS systems.
AC C
EP
TE D
M AN U
SC
RI PT
440
458
In Eulerian (Eulerian-Eulerian or two-fluid) model, the phases
459
are treated mathematically as separate and interpenetrating
460
continua, hence the introduction of the phasic volume fractions-
461
" . The sum of volume fractions is equal to unity
27
ACCEPTED MANUSCRIPT M
L " = 1
(8)
NO
Each phase is governed by a set of continuity and momentum
463
conservation equation, which have similar structure for all
464
phases.
Thus,
Q-phase
considering
system,
RI PT
462
the
mass
conservation equation for phase is (Joshi 2001, Ratkovich
466
2010): M
SC
465
M AN U
. 1" ' 2 + ∇1" ' S 2 − L1 − 2 = 0 . NO
467 468
(9)
where ' is density and the term " ' is the effective density of the phase , S denotes its velocity, and are mass
transfer mechanisms from phase to and from to ,
470
respectively.
471
The generalized momentum conservation equation for phase
472
can be written in the simplified form as (Azzopardi et al. 2011,
AC C
EP
TE D
469
473
Ratkovich 2010): . 1" ' S 2 + ∇1" ' S S 2 .
= −" ∆ + ∇% 1∇" S + ∇" SU 2 + ' VS + S
28
(10)
ACCEPTED MANUSCRIPT
474
where is pressure shared by all phases, % denotes shear viscosity of phase , S represents the sum of interfacial forces
476
between the continuous and dispersed phases.
477
Considering gas-liquid system, the fluid flow around the
478
bubbles is characterized by the occurrence of relative motion
479
between the phases yielding local pressure and shear stress
480
gradients. As a result, relative motion of the bubbles will be
481
affected by a drag force which is predominant in conditions of
482
the uniform flow. In case of non-uniform bubble motion, the
SC
RI PT
475
concept of interfacial forces S needs to account for drag and
484
various non-drag forces, such as virtual mass force, lateral lift
485
force, wall lubrication force, turbulent dispersion force, Basset
486
force and momentum transfer associated with mass transfer.
487
Accordingly, the closure of Equation (11) requires correct
488
assessment of the interfacial forces, typically using analytical
489
models, empirical correlations and coefficients (e.g. drag
490
coefficient), described comprehensively in Clift et al. (1978),
491
Joshi (2001), Azzopardi et al. (2011) and Yang and Mao
492
(2014). Many of these relations are case-specific and based on
493
limited data what yields difficulties in exploring complex
494
multiphase systems (Azzopardi et al. 2011, Manninen et al.
495
1996) and influencing the outcomes of the CFD simulations.
496
The Eulerian model is commonly used in multiphase systems,
497
where momentum exchange between the phases is significant,
AC C
EP
TE D
M AN U
483
29
ACCEPTED MANUSCRIPT e.g. gas-liquid flow in aeration tanks or solid-liquid flow in AS
499
systems, especially in the case of zones with high solids
500
concentrations, where solid-solid interactions are relevant;
501
transitional zones with steep solids concentrations gradients
502
(from high to low concentration), where momentum of the solid
503
phase is relevant to model its dissipation (Samstag et al. 2015).
504
The Volume of Fluid (VOF) model is a single-fluid approach
505
based on a surface tracking technique that solves the
506
momentum equations for the continuous phase while the
507
dispersed phase follows the closure conditions of the volume
508
fraction for the incompressible flow (Wang et al. 2013). While
509
the phases are immiscible, the fields for all variables and
510
properties are shared and represented by volume-average
511
values. Therefore, the continuity equation has the following
512
form (Ratkovich 2010, Vedantam et al. 2006):
TE D
M AN U
SC
RI PT
498
EP
." + ∇1" S 2 = 0 .
(11)
where the computation of the primary-phase volume fraction is
514
based on the constraint defined by Equation (9).
515
The single momentum equation, which is solved throughout the
516
domain yielding velocity field shared by the phases, is
517
expressed as:
AC C
513
30
ACCEPTED MANUSCRIPT . J'SK + ∇J'SSK = −∆ + ∇W%J∇S + ∇S U KX + 'VS + S (12) .
518
The properties ' and % appearing in the transport equations are
519
determined by the presence of the component phases in each
control volume. Considering two-phase ( and ) system if the
RI PT
520
volume fraction of phase is being tracked, the density in each
522
cell is given by (Ratkovich 2010, Vedantam et al. 2006):
523
M AN U
' = " ' + 11 − " 2'
SC
521
(13)
The relationship described by Equation (13) is based on the
524
fact, that for an Q-phase system, the volume-fraction-averaged
525
density is
(14)
TE D
' = L " '
All other properties (e.g volume-fraction-averaged viscosity, %)
527
are computed in the same manner.
528
VOF is designed for modelling multiphase systems where the
AC C
EP
526
529
position of the interface between the immiscible fluids is of
530
interest. This approach has found application in modelling of
531
stratified and free-surface flows, filling, sloshing, and motion
532
of large bubbles (slug flow). Examples of VOF application to
533
model Membrane Bioreactors (MBRs) can be found in the
534
literature (Andersson et al. 2011, Ratkovich et al. 2009, Wang
535
et al. 2013). 31
ACCEPTED MANUSCRIPT The Eulerian-algebraic (slip mixture or algebraic slip) model is
537
a simplified multiphase model used for two or more phases,
538
treated as interpenetrating continua. This single-fluid approach
539
can be used to model phases moving at different velocities- by
540
using concept of slip (or drift) velocities.
541
Here, the continuity and momentum equations are solved for
542
the mixture and algebraic equations are used to solve relative
543
velocities to describe the dispersed phases. Thus, the continuity
544
equation for the mixture is (Manninen et al. 1996):
M AN U
SC
RI PT
536
.' + ∇J' S K = 0 .
545
where the mixture density, ' is defined as: M
TE D
' = L " ' NO
AC C 547
(16)
and the mass-averaged mixture velocity S is:
EP
546
(15)
M
1 S = L " ' S ' NO
(17)
The mass fraction of phase is defined as: =
" ' '
(18)
548
The momentum equation for the mixture is obtained from the
549
following formula (Ratkovich 2010):
32
ACCEPTED MANUSCRIPT . J' S K + ∇J' S S K .
U KX = −∆ + ∇W% J∇S + ∇S + ' VS + S
NO
550
where % is mixture viscosity equal to: M
% = L " %
M AN U
NO
SC
+ ∇ YL " ' S, S, Z
(19)
RI PT
M
(20)
551
and S, is drift velocity for the secondary phase expressed
552
as follows:
(21)
TE D
S, = S − S
The slip velocity of the secondary phase (K relative to the
554
velocity of the primary phase (K is computed as follows:
AC C
EP
553
555
S = S − S
(22)
The relation between drift and slip velocities can be written as: M
S, = S − L S NO
(23)
556
The basic assumption of the algebraic slip mixture model is that
557
a local equilibrium between the phases should be reached over
558
a short spatial length scale. For the phases moving with the 33
ACCEPTED MANUSCRIPT same velocity, the mixture approach can be used to model
560
homogeneous multiphase flow (Manninen et al. 1996, Wicklein
561
and Samstag 2009). However, this approach is usually not
562
recommended in cases when the interface laws are not well
563
known (Ratkovich 2010). This model has been successfully
564
used to simulate sediment-induced density currents, and thus
565
solid-liquid mixing in AS tanks, and turbulent transport of
566
suspended solids in secondary clarifiers (Samstag et al. 2015,
567
Wicklein and Samstag 2009, Yeoh and Tu 2009), but also a
568
gas-liquid turbulent flow in closed loop bioreactors (Xu et al.
569
2010, Yang et al. 2011).
570
Another method to deal with multiphase flow in AS systems is
571
via the modelling of turbulent transport of the secondary phase
572
within a continuous primary phase. The transported phase
573
(species) is treated as an active or passive scalar (density,
574
viscosity, temperature, dissolved oxygen or solids), hence the
575
effects of its gradients across the domain can be either coupled
576
to the momentum equation as an equation of state or treated as
AC C
EP
TE D
M AN U
SC
RI PT
559
577
a passive property transported by the fluid flow.
578
Therefore prediction of local mass fraction of each species, ! ,
579
is modelled by solving conservation equation describing
580
convection and diffusion of the ′-th species (Cartland Glover et
581
al. 2000):
34
ACCEPTED MANUSCRIPT . J'! K + ∇J'S! K = −∇S + + .
582 583
(24)
where S is the ′-th diffusive mass flux of species; is net
rate of production of ′-th specie due to chemical reaction; is
the source term which injects ′-th specie into the domain by
585
addition from the dispersed phase plus any user-defined
586
sources.
587
The active/passive scalar approach is used to model solid-liquid
588
interactions in the systems with lower solids concentrations
589
(Combest et al. 2011, De Clercq 2003, Samstag et al. 2015,
590
Wicklein and Samstag 2009) and to model oxygen mass
591
transfer in aerated tanks (Fayolle et al. 2007).
592
In the Lagrangian model, the governing phase (fluid) is treated
593
as a continuum by solving time-averaged Navier-Stokes
594
equations (Eulerian reference frame), while the behaviour of
595
secondary phase (e.g. solids) is predicted by tracking a large
596
number of particles using random-walk Lagrangian trajectory
AC C
EP
TE D
M AN U
SC
RI PT
584
597
calculations for dispersed phase through the flow field of the
598
continuous phase. Here, the particles can exchange momentum,
599
mass, and energy with the fluid phase. The particles’
600
trajectories are predicted by integrating the force balance on the
601
single particle and recording the particle position. This force
602
balance, which equates the particle inertia with the forces
603
acting on it is defined as (Bridgeman et al. 2009): 35
ACCEPTED MANUSCRIPT .S S − S VSJ' − 'K = + + S . '
(25)
604
where S is particle velocity and S is fluid phase velocity, the
605
term
606
particle relaxation time, ' and ' are particle and fluid
_`
represents drag force per unit particle mass, is
RI PT
\S][ \S^ [
densities, respectively; and S relates to additional forces that
608
are relevant under special circumstances; for example virtual
609
mass and pressure gradient forces, or forces on particles that
610
arise due to rotation of the reference frame.
611
Additionally, particle-particle interactions, such as collisions
612
and momentum exchange can be enabled through the
613
introduction of drag coefficients (Bridgeman et al. 2009,
614
Karpinska Portela 2013). This approach is suitable to simulate
615
gas-liquid flow in aeration tanks, however should not be used
616
when the volume fraction of the secondary (particulate) phase
617
exceeds 10-12% (Sokolichin et al. 1997), as e.g. solid-liquid
618
flow in complete mixed AS systems.
619
Apart from multiphase modelling, particle tracking within the
620
Lagrangian reference frame is a useful tool to simulate RTDs in
621
process tanks (Danckwerts 1953, Levenspiel 1999), and so to
622
assess the macromixing, established by convective flow
623
patterns (Karpinska Portela 2013, Le Moullec et al. 2008).
624
However, the main limitation here is the number of particles
625
being tracked within the simulated system, and in the case of a
AC C
EP
TE D
M AN U
SC
607
36
ACCEPTED MANUSCRIPT 626
full-scale AS tank, tracking too many particles will result in
627
extensive computational
628
workability of different approaches in multiphase modelling,
629
the Lagrangian approach is the most expensive, involving long
630
computational times and requiring a large number of CPUs
631
(Central Processing Units), so limiting its popularity in the
632
simulation of wastewater treatment tanks.
when
comparing
SC
633
Thus,
RI PT
times.
AC C
EP
TE D
M AN U
634
37
ACCEPTED MANUSCRIPT Table 2. Summary of approaches to multiphase modelling. Concept
Modelling
Applicability
Issues
Eulerian
Each phase modeled as a separate fluid.
A set of averaged, volume fractionweighted NavierStokes (NS) equations per phase. Momentum transfer terms and constitutive equations to be modelled.
Theoretically, every type of flow; depending on the additional terms’ modelization.
Additional terms’ modelization is determinant, but their modelling is difficult.
VOF
Each phase modeled as a separate fluid. The interface between the phases is tracked.
Interface tracked via a continuity equation and the domains of the single phases are defined. A set of phase-specific NS equations with momentum exchange terms is solved for each domain.
Flows where the interphase surface is clearly defined (e.g. a single, large bubble inside a liquid).
Inapplicable if the interphase surface is too complex (e.g. bubbly flow where bubble dimensions are smaller than single cell size).
Mixture
Both phases treated as a whole.
Single set of NS equations.
Homogeneous fluids or nonhomogeneous fluids that are treated as homogeneous.
Inapplicable in every case in which there is clear distinction between the phases.
Flows where at least one phase is clearly dispersed into a principal phase. Particles smaller than mesh cell size.
Computational expense not a priori computable, and proportional to the particles number. May be prohibitive for high particle numbers.
SC
M AN U
TE D
Liquid phase treated with Eulerian approach. Every particle (bubble) tracked along trajectory.
AC C
EulerianLagrangian
RI PT
Model
Effective mixture density and viscosity to be modelled.
EP
635
A set of averaged NS equations for the liquid phase. A set of Newton’s 2nd Law (N2L) equations applied to all particles. Momentum transfer terms in both NS and N2L to be modeled.
636
38
ACCEPTED MANUSCRIPT 637
5. Application of CFD in ASP
638
Biological wastewater treatment systems require enhanced
639
transfer of oxygen into the AS tanks to maintain aerobic
640
processes
641
Enhanced oxygen mass transfer in most common ASP
642
configurations i.e. channel or closed-loop bubbly flow AS
643
reactors aerated by diffusers, can be achieved by maximisation
644
of the surface area of the interface between the dispersed phase
645
(air/oxygen bubbles) and the continuous phase (mixed liquor).
646
Nonetheless, in most wastewater utilities, designs of AS tank
647
and aeration system, and furthermore process control are based
648
on static models and simple PID controllers, and so a major
649
concern
650
hydrodynamics and its impact on DO profiles within the basin.
651
Therefore,
652
efficiencies, hydraulic and process design, and reliable
653
operation of AS systems rely on an improved understanding of
654
the bioreactors’ behaviour, with emphasis on micro- and
in
biodegrading-nitrifying
biomass.
the
accurate
TE D
remains
M AN U
SC
RI PT
occurring
of
biochemical
of
tank
conversion
AC C
EP
improvement
determination
655
macro-scale mixing and, in particular, on the analysis of the
656
interactions between the mixed liquor circulation imparted by
657
mechanical agitation and the fluid and sludge flocs motions
658
induced by the air bubbles. The key advantage of CFD as a
659
virtual modelling technique is powerful visualization capability
660
allowing detailed characterization of the local-scale phenomena
661
in varying operating conditions. Hence, CFD can be used as a 39
ACCEPTED MANUSCRIPT robust tool for the design of either new efficient and energy
663
optimised unit processes at WWTPs or for “tune for benefit”
664
optimization of performance and even retrofitting of existing
665
AS systems.
666
The complete CFD simulation of a biologically active gas-
667
liquid-solid AS system is challenging, due to the complex
668
hydrodynamics and biochemical reactions of conversions
669
involved, resulting in massive computational resources required
670
in terms of RAM and CPU usage and with long computational
671
times involved. Furthermore, increasing complexity of the
672
model involved, mesh resolution and solution accuracy may
673
lead distinctly higher costs of the CFD analysis, yet lower than
674
a capital investment in a new layout. To avoid overly long and
675
complex CFD runs (Karpinska Portela 2013, Le Moullec et al.
676
2010b), common engineering practice has been to model the
677
AS processes separately (e.g. aeration system performance or
678
flow field within the basin) and afterwards to couple the results
679
(Pereira et al. 2012). In addition to that, depending on the
AC C
EP
TE D
M AN U
SC
RI PT
662
680
purpose of the CFD analysis, data collection for model
681
calibration and validation using advanced measurement
682
techniques, may be absolutely required and yet time and
683
resource consuming (Nopens et al. 2012).
684
It is worth mentioning, that despite the current successful
685
spread and usage of CFD modelling, potential risk of its misuse
40
ACCEPTED MANUSCRIPT due to poor model choice, wrong setup assumptions or
687
interpretation of the results, has been also recognized (Nopens
688
et al. 2012).
689
5.1. Development of New Aeration Devices
690
CFD simulations have been used as a high-tech design tool in
691
the development of new aeration devices and optimisation of
692
their spatial arrangement in wastewater treatment tanks.
693
A computationally inexpensive modelling procedure, based on
694
3D steady-state and single-phase flow simulations with the
SC
RI PT
686
> − $ turbulence model was used by Morchain et al. (2000) to
696
study the impact of the spatial distribution of the cross-flow
697
hydro-ejectors on the recirculation and the oxygen mass
698
transfer in a large tank for wastewater treatment. Although the
699
cross-flow hydro-ejectors generate two-phase flow, it was
700
shown experimentally that the momentum transfer is not
701
significantly affected by the presence of bubbles and thus the
702
velocity field could be obtained using a single-phase model,
703
while the oxygen transfer in the tank was enabled by the
AC C
EP
TE D
M AN U
695
704
introduction of transport species.
705
The same single-flow modelling scheme was applied to modify
706
geometry and operating scenarios of a curved blade mechanical
707
aerator intended for use in oxidation ditches (Bhuyar et al.
708
2009). The results, which were in good agreement with the
709
experimental data, allowed optimization of blade design and
710
the aerators’ submergence, and reductions to rotational speed 41
ACCEPTED MANUSCRIPT 711
range, yielding enhanced aeration efficiencies when compared
712
with conventional mechanical devices.
713
Despite higher computational costs, transient gas-liquid simulations involving URANS with the > − $ turbulence
715
model and one of the Euler-Euler multiphase models have
716
found applications in the development of new aeration
717
techniques, as they provide direct and more accurate analysis of
718
the multiphase reactor systems. Xu et al. (2010) simulated an
719
oxidation ditch equipped with cylindrical airlift aerators. Here
720
the CFD simulations of the fluid flow in an airlift oxidation
721
ditch served to verify the feasibility of the preliminary design
722
and to assess its applicability for municipal wastewater
723
treatment. The results from the simulations (which were
724
validated in a bench-scale ditch) emphasized the suitability of
725
the
726
configurations.
727
Karpinska Portela (2013) used CFD to focus on overcoming
728
oxygen transfer efficiency limitations of membrane diffusers
729
via the introduction of an independent external aeration unit,
730
designed as a continuous flow component included in the
731
mixed liquor recirculation loop. Similar to the previous work,
732
CFD
733
configurations provided an excellent design tool for selection of
734
the most efficient geometry, characterized by the highest value
TE D
M AN U
SC
RI PT
714
aeration
system
for
deep
tank
ditch
AC C
EP
proposed
simulations
of
aeration
42
within
several
device
ACCEPTED MANUSCRIPT of SOTE. The aeration process parameters determined from
736
pilot-scale aeration tests matched the predictions obtained from
737
the CFD simulations.
738
A more detailed description of the CFD approaches used in the
739
development of aeration devices is shown in Table 3.
RI PT
735
740
AC C
EP
TE D
M AN U
SC
741
43
ACCEPTED MANUSCRIPT Table 3. CFD studies of new aeration devices. Author
Code
Device
Scale
Dim.
Turbulence Extra model model
Bhuyar et al. (2009)
Fluent
Curved blade mechanical aerator
Lab
3D
RANS + > − $
Xu et al. (2010)
Fluent
Airlift aerator
Lab
3D
URANS + > − $
Karpinska Portela (2013)
Fluent
Pressurized Lab Aeration Chamber
3D
URANS + > − $
743 744 745 746
750 751 752 753 754 755
EP
749
AC C
748
TE D
747
44
Validation measurements
RI PT
Transport species + chemical reaction Mixture
SC
M AN U
742
Mixture
Particle Dynamic Analysis (PDA)
measurements
ACCEPTED MANUSCRIPT 756 757
5.2. Evaluation of the performance of the existing aeration systems Robust and energy efficient operation of complex AS reactor
759
systems relies on an understanding of the multiphase nature of
760
the AS tank and on the assessment of the impact of the aeration
761
system on the mixing, biological treatment efficiency and
762
associated energy expenditure, while taking into account the
763
number, type and spatial distribution of the aeration and mixing
764
devices. This section considers CFD approaches used to
765
determine the dynamic behaviour of aeration tanks, and for
766
which descriptions are summarized in Table 4.
767
5.2.1. Gas-liquid Models
768
When considering fluid flow within an aeration tank, the use of
769
gas-liquid
770
straightforward prediction of the multiphase velocity field
771
induced by the aerators and mixers, bubble sizes and local gas
772
hold-up, i.e. process parameters that have significant impact on
773
the oxygen mass transfer in bioreactors, but either require time-
774
and resource-expensive experimental techniques or are not
775
taken into account in the analysis of the experimental results
776
(Fayolle et al. 2007, Gillot et al. 2005, Gillot and Héduit 2000,
777
Le Moullec et al. 2008, Vermande et al. 2007).
778
Hence the focus of early CFD works was on a better
779
understanding of mass transfer phenomena in full scale closed-
TE D
M AN U
SC
RI PT
758
models
enables
relatively
fast
and
AC C
EP
CFD
45
ACCEPTED MANUSCRIPT loop aeration tanks equipped with membrane diffusers and
781
agitated by means of horizontal slow speed impellers (Cockx et
782
al. 2001, Do-Quang et al. 1999). 2D two-phase flow in the
783
aeration tank was simulated using an Eulerian model for
784
imposed, constant gas bubble size. To simplify numerical
785
simulations, an equivalent uniform liquid velocity field was
786
imposed in the section of the impellers. The oxygen mass
787
transfer was obtained via introduction of the global volumetric
RI PT
780
mass transfer coefficient, , determined experimentally, as a
789
transport source term. Here, the correct estimation of the
790
absolute values of oxygen transfer requires assessment of the
791
most uncertain aeration process parameter, i.e. the alpha-factor,
792
quantifying the impact of the contaminants and process
793
conditions on the oxygen transfer rates- in this work- the
794
impact of the imposed constant bubble diameter on the gas-
795
liquid interface surface area (Nopens et al. 2015, Rosso et al.
796
2011, Rosso and Stenstrom 2006).
797
The
EP
TE D
M AN U
SC
788
from
the
simulations
showed
that
the
AC C
results
798
hydrodynamics of these tanks was controlled by mutual
799
competition between vertical gas plume and horizontal fluid
800
flow currents. Thus, in the absence of the horizontal flow
801
motion, the interactions between gas plumes released by
802
diffusers generate vertical, massive liquid loop circulations, i.e.
803
spiral flow, yielding low gas hold-up. Contrary to that, under
804
conditions of horizontal liquid motion imparted by e.g. rotating
46
ACCEPTED MANUSCRIPT impellers,
the
increase
in
wastewater
velocity causes
806
neutralization of the spiral flow patterns, inclination of the
807
vertical gas plume, and dispersion of the bubbles and, as a
808
consequence, longer contact times between the phases yielding
809
better gas retention in the tank. As a result, the increase of the
810
global mass transfer coefficient in the closed-loop tanks was
811
found to be associated exclusively with fluid circulation.
812
The impact of the fluid velocity on oxygen mass transfer was
813
later confirmed in experimental studies (Abusam et al. 2002). It
814
was reported that even small changes in the axial velocity may
815
have a dramatic effect on the oxygen profiles, and thus on the
816
ammonia removal in oxidation ditches.
817
closed-loop AS systems, horizontal fluid velocity should be
818
treated as an important process variable to control total nitrogen
819
removal efficiency.
820
A clear shortcoming of the discussed work (Cockx et al. 2001,
821
Do-Quang et al. 1999) is that the 2D modelling is insufficient
822
to represent mutual interactions between the neighbouring gas
M AN U
SC
RI PT
805
AC C
EP
TE D
Therefore in these
823
plumes causing formation of the more complex, three-
824
dimensional flow structures. Moreover, the impact of the flow
825
circulation on the size of the bubbles released by membrane
826
diffusers was neglected; yet this plays a key role in the oxygen
827
mass transfer within the aeration tank. Nevertheless, despite
828
these weaknesses and the lack of experimental validation,
829
conclusions from this work provided valuable research
47
ACCEPTED MANUSCRIPT background for many successive numerical and experimental
831
studies.
832
Subsequent studies (Fayolle et al. 2007) addressed development
833
of a more complete numerical tool, to predict oxygen mass
834
transfer in a number of pilot- and full-scale oxidation ditches of
835
various configurations and where aeration is dissociated from
836
mixing by the introduction of membrane diffusers and slow
837
speed mixers. The aeration was simulated using Eulerian model
838
for water and air. Mixing was modelled using fixed value
839
method by imposing flow characteristics induced by agitation
840
on the grid zones corresponding to mixers location. This
841
method gives good prediction of the average experimental axial
842
velocity. Similar to earlier CFD works (Cockx et al. 2001, Do-
843
Quang et al. 1999), Fayolle et al. (2007) showed that oxygen
844
mass transfer in closed-loop AS basins aerated with diffusers is
845
linked to the number, type, spatial distribution and performance
846
of the agitators (Fayolle et al. 2007, Fayolle et al. 2010, Fayolle
847
et al. 2006, Vermande et al. 2007),
The impact of the
AC C
EP
TE D
M AN U
SC
RI PT
830
848
ascending bubbles released by the diffusers on the velocity
849
profile along the oxidation ditch can be seen in Figure 1. It was
850
also showed that implementation of more complex modelling
851
approach accounting for the impact of the bubble size on the
852
axial fluid velocities and on the global gas hold-up profile
853
within the tank, was able to reproduce precisely the values of
854
with high accuracy (≤ 5%). Nonetheless, it was found, that 48
ACCEPTED MANUSCRIPT the predicted accuracy depends on the assumed value of inlet
856
bubble size, emphasizing the necessity either to measure in situ
857
bubble diameter or apply an appropriate numerical model
858
estimating bubble sizes at the diffuser level. A shortcoming of
859
Fayolle et al. (2007) is that although the proposed CFD
860
protocol appears to be suitable for the prediction and
861
optimization of oxygenation capacities in a full scale AS tank,
862
it is based on clean water-air simulations. However, for robust
863
analysis of agitation system performance, the impact of the
864
velocity on the local solids content and density gradients should
865
be considered.
866
Yang et al. (2011) focused on predicting the flow pattern and
867
oxygen mass transfer in a multichannel oxidation ditch aerated
868
with horizontal rotors and agitated with submerged mixers (a
869
Carrousel). Experimental field data showed that under existing
870
operational conditions, DO concentration in the anoxic zone
871
exceeded 1.0 mg/L, inhibiting the denitrification process. Thus,
872
the aim of the CFD study was to determine an operational
AC C
EP
TE D
M AN U
SC
RI PT
855
873
regime for the surface aerators, which would lead to formation
874
of stronger DO gradients in the anoxic zone, yielding a
875
reduction in energy expenditure for aeration while maintaining
876
high treatment efficiencies. The modelling procedure involved
877
a two-phase mixture approach, enhanced with additional
878
transport equations related to the transfer of oxygen introduced
879
by rotors and biochemical reaction of oxygen consumption for 49
ACCEPTED MANUSCRIPT BOD removal. To reduce computational expenses related to the
881
modelling of a number of rotating devices, the large surface
882
aerators were simulated using moving wall model, while
883
actuator disk (fan) approach was used to simulate submerged
884
mixers. Although an energy efficient operating scheme for the
885
ditches was developed, this study was based on average flow
886
and constant influent quality parameters, thus the validity of the
887
optimal operating conditions deduced from the steady-state is
888
limited.
889
More recent CFD studies on the aeration of conventional AS
890
tanks (Gresch et al. 2011) provided further and more detailed
891
analysis of the flow field induced by porous diffusers. The CFD
892
approach proposed in this work consisted of the gas-liquid
893
Eulerian multiphase model, where the physical properties of the
894
continuous phase were approximated to those of activated
895
sludge. Moreover, instead of solving additional transport
896
equations for oxygen transfer, hydrodynamic simulations were
897
enhanced with a biokinetic model for nitrification. The results
AC C
EP
TE D
M AN U
SC
RI PT
880
898
from CFD studies validated experimentally in a full-scale plug
899
flow AS tank complemented conclusions from earlier work
900
(Cockx et al. 2001, Do-Quang et al. 1999) and provided a
901
comprehensive description of the spiral flow generated by
902
changing diffuser layout, its impact on the velocity field, air
903
hold-up (Figure 2), and ammonia degradation. It was also
904
shown that, contrary to the closed-loop basins, in plug flow AS 50
ACCEPTED MANUSCRIPT lanes without flow boosters, the flow field determines the
906
aeration efficiency and the intensity of the longitudinal mixing,
907
which are significantly reduced by either occurrence of non-
908
aerated zones at the sidewalls or the rolling motion of the fluid
909
generated by changes in diffusers layout.
910
5.2.2. Density-coupled Models
911
In recent years, multiphase modelling of AS tanks based on
912
gas-liquid neutral density simulations has become common
913
practice for evaluation of both aeration and mixing, mainly
914
because it enables faster setup of the lab-scale validation,
915
usually involving use of tap water and air only. However, CFD
916
studies on a sequencing batch reactor equipped with
917
conventional jet aeration system (Samstag et al. 2012) showed
918
that use of neutral density may lead to over-prediction of the
919
degree of mixing. In this work, a density-coupled CFD model
920
incorporating solids settling and transport was calibrated to
921
field data and used to evaluate capacity of the jet aeration
922
system in keeping the solids suspended and to determine power
AC C
EP
TE D
M AN U
SC
RI PT
905
923
consumption for pumping considering two operating modes:
924
mixing with and without air. The authors found that in order to
925
generate complete information about activated sludge mixing,
926
complementary
927
modelling are required to predict local density gradients due to
928
the impact of the flow regime on solids transport. While this
929
work underlines the importance of the analysis of the aeration
CFD
studies
51
based
on
density-couple
ACCEPTED MANUSCRIPT systems with respect to the activation sludge mixing, the
931
approach is rarely used by the wastewater modelling
932
community for AS tanks, and only limited works can be found
933
in the literature (Jensen et al. 2006, Laursen 2007, Samstag et
934
al. , Samstag et al. 1992, Samstag et al. 2012, Wicklein and
935
Samstag 2009).
936
Recently Xie et al. (2014) considered prediction of the mixing
937
and suspended solids distribution in a full-scale Carrousel ditch
938
equipped with surface aerators and submerged impellers. The
939
simulation procedure involved a solid-liquid mixture model,
940
where the sludge settling was coupled through the slip velocity.
941
Similar to the earlier work by Yang et al. (2011), surface rotors
942
and submerged mixers were simulated using moving wall and
943
fan models. The resulting mixing patterns and solids profiles
944
throughout the ditch, shown in Figure 3, facilitated
945
identification of the stagnation regions affected by the sludge
946
settling, and resulted in optimization of the operation scenario.
947
Previously, Fan et al. (2010) simulated a single-channel
948
oxidation ditch aerated by means of inversed umbrella surface
949
aerators using a solid-liquid two-fluid Eulerian model. The
950
operating aerators were simulated using Multiple Reference
951
Frames (MRF) approach. This work also focused on a detailed
952
description of the mixing regime induced by the operating
953
aeration system, aiming to select the optimal range of aerator
AC C
EP
TE D
M AN U
SC
RI PT
930
52
ACCEPTED MANUSCRIPT speed ensuring the most uniform distribution of the solids
955
within the ditch, and thus prevents solids settling.
956
Thus, the literature shows that the use of density-coupling is
957
well founded to assess sludge mixing in agitated AS systems,
958
however its applicability to study bubbly bioreactors is still
959
uncertain.
960
5.2.3. Residence Time Distribution of AS tanks
961
Although rarely used to simulate the aeration process itself,
962
CFD modelling based on the Lagrangian multiphase approach
963
with particle tracking can be used as a tool for evaluation of the
964
aeration system performance through the assessment of the
965
macromixing, and thus the overall transport phenomena in a
966
bioreactor. The literature offers numerous examples of studies
967
on aeration and/or agitation systems responsible for different
968
flow patterns within the tank, and thus having an impact on the
969
mixing at micro- and macro-scale in AS systems. Considering
970
different advection paths of the fluid elements within an aerated
971
continuous flow channel or closed-loop bioreactors for
AC C
EP
TE D
M AN U
SC
RI PT
954
972
wastewater
treatment,
RTD
973
simulations can be used successfully to predict such adverse
974
phenomena as segregation of the flow resulting in short-
975
circuiting (channelling), recycling of the flow, or formation of
976
dead zones. RTD data can also be used to assess the mixing
977
time, which can be of crucial importance for oxygen transfer
978
and impact on chemical and biochemical reaction yield. Data 53
data
obtained
from
CFD
ACCEPTED MANUSCRIPT can also be used for troubleshooting of the reactor and
980
improved design of the future vessels (Brannock 2003,
981
Karpinska et al. 2015, Karpinska Portela 2013, Kjellstrand
982
2006).
983
Glover et al. (2006) considered oxidation ditches aerated by
984
bottom diffusers and agitated by slow-speed rotors and
985
determined reactor model structure by implementation of a
986
CFD-generated RTD curve in Fluent into WEST® software
987
environment based on the systemic approach. The fluid flow in
988
the ditch was obtained from the Eulerian gas-liquid model, and
M AN U
SC
RI PT
979
the turbulence simulated with > − $ model. It was found that
990
the RTDs of the ditch behaviour can be approximated by 90
991
CSTRs (plug flow) for non-aerated conditions and by 20
992
CSTRs (perfectly mixed) for aerated conditions (Glover et al.
993
2006). The work showed that closed-loop reactors with
994
different internal recycling rates, such as oxidation ditches, can
995
be modelled by one or series of CSTRs, the number of which
996
can be determined from CFD-RTD studies. In this way, a
AC C
EP
TE D
989
997
suitable reactor model structure for ASM studies can be
998
predicted.
999
Le Moullec et al. (2008) considered the assessment of RTDs in
1000
a cross-flow channel reactor aerated with porous tube, in which
1001
the hydrodynamics was modelled using a gas-liquid Eulerian
1002
model and the turbulence models used were > − $ and RSM. 54
ACCEPTED MANUSCRIPT It was shown that in the channel-type reactors for wastewater
1004
treatment, turbulence has a dominant role in axial dispersion
1005
(90%), while dispersion due to convection is negligible.
1006
Furthermore it was shown that only the RTDs obtained from
1007
the RSM model are in good agreement with the experimental
1008
data treated by curve fitting for a plug flow with axial
RI PT
1003
dispersion model, while the > − $ model underestimated the
1010
value of dispersion coefficient by around 50%, as seen in
1011
Figure 4. This particular case is the effect of the default
1012
constants in Lagrangian stochastic particle motion model,
1013
which are determined by different closure assumptions. Thus,
M AN U
SC
1009
both models, > − $ and RSM, will use different AB which are
1015
directly linked to the modelling of the turbulent dispersion,
1016
what emphasizes the need for careful attribution of values to
1017
parameters, through the proper model calibration.
1018
Macromixing assessment is a useful tool to describe actual
1019
reactor performance. CFD studies assessing the impact of
1020
turbulence model selection on the RTDs of oxidation ditches
AC C
EP
TE D
1014
1021
aerated with slot injectors (Karpinska Portela 2013, Pereira et
1022
al. 2012), considered the fluid flow simulated with RANS and
1023
URANS with the > − $ model and also LES with
1024
Smagorinsky’s SGS model. This work shows the limitations of
1025
some approaches in computation of the RTD, as the turbulence
1026
model involved to simulate hydrodynamics has a large impact
1027
on prediction of the tracer’s trajectories. The RTD simulations 55
ACCEPTED MANUSCRIPT based on the average flow field, RANS and URANS, led to
1029
overestimation of channelling effects. The flow dynamics
1030
underlies mixing at all scales, both macro- and micro-, and thus
1031
this should always be accounted for. Furthermore, is has been
1032
shown that CFD data for AS modelling must align with all
1033
dynamic components of the flow, thus LES simulations, which
1034
demand more computational resource, should be used for some
1035
specific purposes.
1036
5.2.4. Full model: CFD-ASM
1037
Few workers have focused on the development of a complete
1038
three-phase CFD model to simulate actual physical-chemical-
1039
biological processes within different AS system configurations
1040
(Glover et al. 2006, Le Moullec et al. 2010a, b, Le Moullec et
1041
al. 2011, Lei and Ni 2014). Here, the integration of
1042
hydrodynamics with biokinetics is accomplished by embedding
1043
an ASM model (e.g. ASM1) in a CFD model through
1044
introduction of the additional species of transport with source
1045
terms comprising bioreaction rates. The CFD-ASM simulations
AC C
EP
TE D
M AN U
SC
RI PT
1028
1046
permit the quantification of interactions and transport
1047
phenomena between the water-gas, water-sludge and gas-
1048
sludge phases occurring within the AS reactor and accounting
1049
for the flow field, oxygen mass transfer, growth, decay and
1050
metabolic activity of the AS biomass, and sludge settling.
1051
Therefore it is possible to predict simultaneously the system
1052
hydrodynamics and its impact on the reactions occurring in 56
ACCEPTED MANUSCRIPT nitrifying-denitrifying-biodegrading biomass and represented
1054
by the concentration profiles, in order to optimize aeration,
1055
mixing or system layout leading to efficient process operation
1056
with reduced exploitation costs.
1057
The first attempt at CFD-ASM1 coupling reported in the
1058
literature (Glover et al. 2006), based on the Eulerian gas-liquid
1059
approach, facilitated prediction of poor performance of the
1060
diffused aeration system in a full scale oxidation ditch system,
1061
giving low nitrifying capacity of the sludge. These studies
1062
demonstrated the robustness of the CFD-ASM1 approach;
1063
however it was emphasized, that the validity of this modelling
1064
procedure requires further work to study different AS systems
1065
and aeration/agitation scenarios.
1066
Le Moullec et al. (2010a) provided a robust discussion on the
1067
modelling issues concerning simulations of the hydrodynamics
1068
and biokinetics in a channel AS tank aerated by porous tube
1069
using Eulerian two-fluid model coupled to the ASM1. The
1070
authors discussed the trade-off between a model which can be
1071
implemented and run in the realistic time, and the extent of
1072
simplifications required to facilitate this. The authors assumed
1073
that the channel reactor was in a pseudo steady-state with
1074
constant biomass content; air bubbles with constant diameter;
1075
and the activated sludge flocs being perfectly soluble in the
1076
liquid phase. These simplifications gave rise to a series of
AC C
EP
TE D
M AN U
SC
RI PT
1053
57
ACCEPTED MANUSCRIPT 1077
errors in model output; specifically, incorrect estimation of the
1078
local
1079
overestimation of the ammonia values (wrong estimation of the
1080
autotrophic fraction of biomass in the system).
1081
The conclusions from this work emphasized the importance of
1082
the bubble size distribution for the oxygen mass transfer and for
1083
the molar diffusivity of oxygen in mixed liquor. Furthermore, a
1084
new concept for the approximation of AS floc properties
1085
(highly hydrated solid phase) to a liquid phase, was proposed.
1086
The authors also highlighted that, as the hydrodynamics-
1087
biokinetics coupling requires high number of CPUs and long
1088
computational times, a compromise between mesh resolution
1089
and solution accuracy has to be found.
1090
In more recent work on three-phase CFD-ASM1 simulations of
1091
a Carrousel oxidation ditch aerated and agitated by means of
1092
mechanical aerators and mixers, Lei and Ni (2014) treated the
1093
AS flocs as a pseudo-solid phase. Here, the solid-liquid-gas
1094
flow field was obtained with a three-phase mixture model. The
underestimation
of
the
nitrate
and
AC C
EP
TE D
M AN U
SC
RI PT
value;
1095
results from the simulations, (i.e. velocity, DO, organics and
1096
nutrients profiles) are in good agreement with the validation
1097
data. Crucially, this modelling approach allowed the prediction
1098
of physical kinematics of sludge settling. Concentration maps
1099
of the solids (near the bottom), DO, COD, ammonia and
1100
nitrates (in the mid-depth) in the horizontal cross-section
1101
through the oxidation ditch, are shown in Figure 5. 58
ACCEPTED MANUSCRIPT It can be concluded, that the CFD-ASM1 simulations provide
1103
more reliable results than those obtained from oversimplified
1104
ASM models, which fail to represent the flow dynamics within
1105
a system, particularly regarding DO profiles along an AS basin,
1106
which have an impact on the biological nutrient removal
1107
(Pereira et al. 2009). Thus, the CFD-ASM data can be
1108
considered more suitable for the design and scale-up of
1109
bioreactors. However, besides high computational costs,
1110
another emerging drawback of CFD-ASM modelling approach
1111
is its limited feasibility due to stringent convergence criteria
1112
and equilibrium solution (Nopens and Wicks 2012). As a result,
1113
the procedures of coupling hydrodynamics data obtained from
1114
CFD simulations with the ASM simulations have been also
1115
intensively studied (Glover et al. 2006, Karpinska Portela 2013,
1116
Le Moullec et al. 2010b, Pereira et al. 2012). Here the RTD
1117
curves, actual HRT values, corrected recycle ratio, local
1118
velocities and other hydrodynamics characteristics obtained
1119
from the CFD simulations of an analysed AS system can be
AC C
EP
TE D
M AN U
SC
RI PT
1102
1120
used to generate a suitable reactor model, i.e. in terms of
1121
number of CSTRs in series, recirculation rate and the flow
1122
pattern between each of the reactors, where the ASM can be
1123
implemented.
1124
Nowadays, an alternative modelling approach using a
1125
compartmental model, based on the description of the AS
1126
reactor system using a network of interconnected sections, i.e. 59
ACCEPTED MANUSCRIPT compartments, is emerging. The bi-directional flow rates
1128
between the compartments are computed from the flow field
1129
obtained in the CFD simulations and accounting for the local
1130
values of velocities and mixing due to turbulence. When
1131
comparing with CFD models, compartmental models are
1132
inexpensive in terms of RAM and CPU usage. Nevertheless, as
1133
they are derived from steady-state CFD simulations, the results
1134
must be experimentally validated. In addition, it is still
1135
necessary to develop a more detailed biokinetic model to apply
1136
this approach in modelling of the full-scale industrial
1137
bioreactors, including ASPs (Le Moullec et al. 2010b, Le
1138
Moullec et al. 2011, Nopens and Wicks 2012, Pereira et al.
1139
2012).
M AN U
SC
RI PT
1127
AC C
EP
TE D
1140
60
ACCEPTED MANUSCRIPT
Table 4. CFD modelling of aeration in Activated Sludge tanks. Reference
Code
Scale
Dim. Mesh size
Model
Multiphase model
Extra model
Validation
Numerical modelling of an oxidation ditch aerated with diffusers and agitated by impellers.
Cockx et al. (2001), Do-Quang et al. (1999)
Astrid
Full
2D
n/a
RANS + > − $
Gas-liquid Eulerian
Transport eq. for oxygen
-
CFD studies of the oxidation ditches- coupling of hydrodynamic with biokinetics modelling.
Glover et al. (2006)
Fluent Lab pilot, full
3D
n/a
RANS + > − $
Gas-liquid Eulerian
DO, COD, NH4, NO3 measurements
29k452k
URANS + > − $
Gas-liquid Eulerian
Lagrangian, transport eq. for oxygen + complete biokinetics (ASM1) Transport eq. for oxygen (each phase)
Prediction of the oxygen transfer in oxidation ditches.
Fayolle et al. (2007)
Fluent Pilot, full
3D
RTD of an aerated channel bioreactor.
Le Moullec et al. (2008)
Fluent Lab
3D
350k
Gas-liquid Eulerian
Lagrangian
LDV, tracer experiments
3D
50k
RANS + > − $/RSM
Simulation of the hydrodynamics and reactions in activated sludge channel reactor.
Le Moullec et al. (2010a), b), Le Moullec et al. (2011)
Fluent Lab
Gas-liquid Eulerian
Transport eq. for oxygen + biokinetics (ASM1)
LDV, bubble size measurements
SC
M AN U
TE D
EP
RI PT
Aim
AC C
1141
61
RANS + > − $
MDV, DO measuremets
ACCEPTED MANUSCRIPT
Reference
Code
Impact of the surface aeration on the solids distribution in the oxidation ditch.
Fan et al. (2010)
Fluent Lab
3D
70,3k
RANS + > − $
Effect of aeration patterns on the flow field in the conventional AS tank.
Gresch et al. (2011)
CFX
Full
3D
400k
URANS + SST
−(
Modification of the operation conditions in the oxidation ditch to enhance energy efficiency.
Yang et al. (2011)
Fluent Full
3D
Evaluation of jet aeration and mixing in sequence batch reactors.
Samstag et al. (2012)
Fluent Full
3D
RTD of an oxidation ditch aerated with hydrojets.
Karpinska Portela (2013); Karpinska et al. (2015)
Fluent Pilot
3D
Multiphase model
Extra model
Validation
Solid-liquid Eulerian
-
PDA
Gas-liquid Eulerian
Biokineticsconsumption of NH4
ADV, reactive tracer experiments
Gas-liquid mixture
Transport eq. for oxygen + biokineticsconsumption of BOD
MDV, DO measurements
URANS + > − $
Gas-liquid mixture
Density-coupled model- solids settling
MLSS measurements
RANS + > − $, URANS + > − $, LES + Smagorinsky SGS
-
Lagrangian
-
1,66M RANS + > − $
TE D
EP
Model
SC
Scale Dim. Mesh size
RI PT
Aim
M AN U
Table 4. Continuation.
600k
AC C
1142
62
ACCEPTED MANUSCRIPT
1143
Aim
Reference
Code
Model
Studies on flow field and sludge settling in Carrousel oxidation ditch.
Xie et al. (2014)
Fluent Full
3D
1,53M URANS + > − $
Fluent Pilot
3D
Multiphase model
Extra model
Validation
Solid-liquid mixture
Slip velocity- sludge settling
MDV, MLSS measurements
Gas-liquidsolid mixture
Transport eq. for oxygen + biokinetics- (ASM1)
ADV, DO, COD, MLSS, NH4, NO3 measurements
SC
Scale Dim. Mesh size
RI PT
Table 4. Continuation.
M AN U
1144
A complete model to predict Lei and Ni (2014) hydrodynamics, oxygen transfer and biokinetic reactions in an oxidation ditch.
162k
RANS + > − $
ADV - Acoustic Doppler Velocimetry, MDV - Mono-directional Velocimetry, PDA - Particle Dynamic Analysis, LDV - Laser Doppler
1146
Velocimetry.
AC C
EP
TE D
1145
63
ACCEPTED MANUSCRIPT 6. Unaddressed Issues in CFD Modelling of AS Tanks
1148
Despite the fact that work has taken place on the CFD
1149
modelling of AS systems for over 15 years now, there are a
1150
number of issues which remain unaddressed and which, if
1151
successfully overcome, would enhance model fidelity and
1152
robustness further.
1153
6.1. Secondary Settler
1154
In engineering practice, the behaviour of the AS tanks is
1155
inseparably linked to the performance of the secondary settler,
1156
as the efficient removal of BOD and nutrients in AS process
1157
depends on the Solid Retention Times (SRTs), concentration of
1158
Mixed Liquor Suspended Solids (MLSS), and the composition
1159
of biomass. The rationale behind the dynamic modelling of a
1160
coupled aeration tank- clarifier system is the prediction of
1161
interconnected flow and mass transport in unsteady flow
1162
loading conditions, and thus evaluation of the impact of
1163
dynamic changes in return sludge flow rates on the
1164
concentration patterns (Patziger et al. 2012).
AC C
EP
TE D
M AN U
SC
RI PT
1147
1165
Secondary settling is referred to in the literature as “the most
1166
sensitive and complicated process in activated sludge plants”
1167
(Ji et al. 1996). In fact, the complete CFD modelling of a
1168
clarifier is not a feasible task, since many physico-biochemical
1169
phenomena
1170
hydrodynamics, turbulence, flocculation, sludge rheology,
1171
settling characteristics, heat exchange and temperature, and
must
be
considered
64
simultaneously,
i.e.
ACCEPTED MANUSCRIPT finally, biokinetics (Plósz et al. 2012). The most critical issue in
1173
the modelling of clarifiers is the inherent unpredictability of
1174
activated sludge settleability and missing data regarding the
1175
particulate fraction. Additionally, even the most advanced
1176
models for clarifiers are still based on empirical equations
1177
describing sludge settling. The variability of settling behaviour
1178
which is not predicted by models affects the actual SRTs of AS
1179
system, yielding a potential source of error in wastewater
1180
treatment plant models (Plósz et al. 2011, Plósz et al. 2012).
1181
Consequently, there is always a risk that the simultaneous
1182
modelling of clarifiers may affect the results obtained for the
1183
AS tank. Nonetheless, a CFD simulation of a simplified 2D AS
1184
tank- clarifier system has been reported in the literature
1185
(Patziger et al. 2012), where the focus was on the solids
1186
transport between both units in conditions of variable inflow
1187
and wet weather, but neglecting biokinetics and oxygen mass
1188
transfer. The necessity for further model enhancement and
1189
validation was also emphasized.
AC C
EP
TE D
M AN U
SC
RI PT
1172
1190
When considering the complexity of CFD models predicting
1191
clarifier behaviour simultaneously with a complete solid-liquid-
1192
gas model of the AS tank, computing time, availability of CPU
1193
resources and predictive capability must be recognised as
1194
confounding issues. As a result, common modelling practice is
1195
to avoid the interference between the two unit processes, and
1196
assume constant separation efficiency for the clarifier
65
ACCEPTED MANUSCRIPT performance (Le Moullec et al. 2010a, b, Le Moullec et al.
1198
2011) predicted in agreement with the guideline (Copp 2001).
1199
Moreover, in engineering practice, the design of clarifiers is
1200
based on assumption of certain sludge properties, hence the
1201
CFD models of settlers only have been frequently used for
1202
optimization of design and retrofitting purposes (De Clercq
1203
2003, Stamou et al. 2009). However, it will be worthwhile to
1204
pursue CFD analysis of integrated aeration tank- secondary
1205
settler systems, when models capable of predicting the
1206
character of activated sludge flocculation are regularly
1207
incorporated into sludge settling models.
1208
6.2. Population Balance Model
1209
The AS system can be described as an ensemble of populations
1210
of individual entities (bubbles, flocs, biomass cells), having
1211
specific properties (size, density, viscosity, enzymatic activity).
1212
In such a system, two kinds of behaviour may be recognized:
1213
interactions of the individual entities with the environment (e.g.
1214
interfacial oxygen transfer, shear induced break-up), and
AC C
EP
TE D
M AN U
SC
RI PT
1197
1215
mutual interactions between the individual entities (e.g.
1216
coalescence, aggregation). The character of these interactions is
1217
a function of one or more properties of the entities, which vary
1218
within the population. Thus, it is more correct to refer to this
1219
variation as “distributed properties”, as they can be represented
1220
by a distribution instead of a scalar (Nopens et al. 2015).
1221
Nevertheless, in order to reduce overall complexity of the 66
ACCEPTED MANUSCRIPT model framework associated with aeration tanks, a common
1223
procedure is to assume non-distributed scalar properties
1224
implying that all individuals behave in exactly the same way
1225
(Nopens et al. 2015, Sobremisana et al. 2011). The classic
1226
example is the assumption of fixed bubble diameter (Fayolle et
1227
al. 2007, Glover et al. 2006, Gresch et al. 2011, Le Moullec et
1228
al. 2010a) and uniform floc size (Fan et al. 2010).
1229
When considering an AS tank, recent experimental studies have
1230
demonstrated the influence of floc size on the flocculation
1231
behaviour and thus settleability of the activated sludge (Nopens
1232
et al. 2015), as well as on the biokinetic reaction environment
1233
within the AS floc (Sobremisana et al. 2011). Therefore
1234
estimation of floc size distribution may provide further insight
1235
into modelling of the activated sludge mixing (Samstag et al.
1236
2012) and assessment of aeration tank settling, an important
1237
process parameter allowing the prediction of hydraulic capacity
1238
and treatment efficiency during high hydraulic loads associated
1239
with wet weather flows (Nielsen et al. 2000, Sharma et al.
AC C
EP
TE D
M AN U
SC
RI PT
1222
1240
2013).
1241
Depending on the bubble flow regime in an AS tank, the
1242
intensity
1243
deformation promotes a large diversity in the shapes and sizes
1244
of the bubbles (Karpinska Portela 2013, Shaikh and Al-Dahhan
1245
2007, Takács 2005). At the same time, the impact of the bubble
1246
of
collisions,
agglomerations,
breakups
and
size on , superficial gas velocity, and thus gas hold-up and 67
ACCEPTED MANUSCRIPT oxygen mass transfer in aeration tanks has been also recognized
1248
(Fayolle et al. 2006, Vermande et al. 2007). However, in some
1249
cases involving modelling of aeration in conventional AS
1250
tanks, the assumption of non-distributed scalar properties is still
1251
predominant, while in others, i.e. modelling of clarifiers or
1252
bubble columns, it may lead to predictions that diverge
1253
significantly from the real systems (Sobremisana et al. 2011).
1254
Consequently,
1255
bubble/particle size distributions require the use of population
1256
balance models (PBM) to describe variations in populations of
1257
entities. An assessment of several solution methods for PBM,
1258
namely the discrete class size method (Hounslow et al. 1988),
1259
the standard method of moments- SMM (Randolf and Larson
1260
1971) and the quadrature method of moments- QMOM
1261
(Marchisio et al. 2003), can be found in the literature
1262
(Bridgeman et al. 2009). It should be highlighted, that although
1263
development of PBMs is at advanced stage and the coupling of
1264
QMOM with hydrodynamics requires only a small number of
models
which
incorporate
AC C
EP
TE D
M AN U
SC
multiphase
RI PT
1247
1265
computationally inexpensive scalar equations to be tracked, its
1266
application in modelling of aeration systems with suspended
1267
solids is still scarce and limited to very few examples
1268
(Bridgeman et al. 2009, Nopens et al. 2015, Sobremisana et al.
1269
2011), associated almost exclusively with secondary settlers
1270
(Griborio and McCorquodale 2006, Nopens et al. 2005).
1271
However, PBM-CFD coupling has been successfully exploited
68
ACCEPTED MANUSCRIPT 1272
by the chemical engineering community to study bubble
1273
columns, airlift and stirred bioreactors (Dhanasekharan et al.
1274
2005, Morchain et al. 2014, Wang 2011).
1275 7. Conclusions
1277
In the last few years, as a result of increasing availability and
1278
accessibility of commercial and open-source software suites,
1279
the use of CFD has evolved into a robust and precise technique
1280
for design, optimization and control of the AS systems. The
1281
following key conclusions can be put forward from this review:
SC
•
M AN U
1282
RI PT
1276
The complete CFD simulation of the complex multiphase flow in AS tanks remains a challenge, due
1284
to the high CPU and RAM requirements and limited
1285
feasibility resulting from the imposed convergence
1286
criteria. Although there is still no unequivocal protocol
1287
on CFD methodology, the most computationally
turbulence model has been adopted as the standard for
AC C
1289
efficient scenario, RANS/URANS closed by > − $
EP
1288
TE D
1283
1290
1291
the modelling of AS tanks;
•
Different CFD models serve different applications.
1292
Examples from the literature have demonstrated the
1293
potential and robustness of single flow simulations in
1294
design and optimization of the AS systems equipped
1295
with mechanical and jet aeration systems;
69
ACCEPTED MANUSCRIPT 1296
•
The Euler-Euler approach has been extensively
1297
explored for the prediction and optimization of
1298
oxygenation capacities in AS tanks equipped with
1299
diffused aeration systems; •
Lagrangian approach with particle tracking has been
RI PT
1300
used to determine RTDs of the tanks- a tool for
1302
evaluation of the aeration and mixing performance and
1303
troubleshooting of the reactor design;
1304
•
SC
1301
The neutral density modelling of AS tanks became common practice used for design and evaluation of
1306
aeration and mixing systems, despite leading to over-
1307
prediction of the degree of mixing. Hence the necessity
1308
to include density-coupled modelling of aeration tanks
1309
to predict accurately local density gradients due to the
1310
impact of the flow regime on solids transport; A small number of works concerns a complete threephase CFD model coupled with ASM1 aiming to quantify the transport and mutual interactions between
AC C
1313
TE D
1312
•
EP
1311
M AN U
1305
1314
water-gas-sludge phases within the AS reactor. Despite
1315
providing more reliable results than those obtained
1316
from ASM, this approach relies on the trade-off
1317
between a model which can be implemented and run in
1318
a realistic timeframe, and the extent of simplifications
1319
and the solution accuracy, which may lead to a series of
1320
output errors;
70
ACCEPTED MANUSCRIPT 1321
•
Potential possibilities of coupling of the CFD data with
1322
ASM based codes have been explored in order to
1323
generate a suitable tank-in-series model, where the
1324
biokinetic model can be implemented; •
There are several areas in modelling practice, which
RI PT
1325
remain unaddressed and require further study, e.g.
1327
modeling of coupled aeration tank- clarifier system to
1328
predict interconnected flow and mass transport in
1329
unsteady flow conditions and CFD-PBM coupling to
1330
assess the impact of the AS floc/air bubble size on
1331
mixing, settling, gas hold-up and oxygen mass transfer
1332
in the aeration tank.
M AN U
SC
1326
TE D
1333 Acknowledgements
1335
The research work of Dr. Anna M. Karpinska Portela was
1336
funded by the College of Engineering and Physical Sciences,
1337
University of Birmingham, UK.
EP
1334
AC C
1338 1339
References
1340 1341 1342 1343
Abusam, A. and Keesman, K.J. (1999) Effect of number of CSTR’s on the modelling of oxidation ditches; steady state and dynamic analysis . Proceedings of 13th Forum for Applied Biotechnology, pp. 91-94, Sept. 22-23, Gent, Belgium.
1344 1345 1346 1347
Abusam, A., Keesman, K.J., Spanjers, H., Van Straten, G. and Meinema, K. (2002) Effect of oxidation ditch horizontal velocity on the nitrogen removal process. European Water Management Online European Water Association (EWA).
71
ACCEPTED MANUSCRIPT Åmand, L. and Carlsson, B. (2012) Optimal aeration control in a nitrifying activated sludge process. Water Research 46(7), 21012110.
1351 1352 1353 1354
Andersson, B., Andersson, R., Håkansson, L., Mortensen, M., Sudiyo, R. and Van Wachem, B. (2011) Computational fluid dynamics for engineers, Cambridge: Cambridge University Press, Cambridge.
1355 1356 1357
Ardern, E. and Lockett, W.T. (1914) Experiments on the oxidation of the sewage without the aid of filters. Journal of the Society of Chemical Industry 33(10), 523-539.
1358 1359 1360
ATV-DVWK (2000) Standard ATV-DVWK- A 131E Dimensioning of Single-Stage Activated Sludge Plants, German Association for Water, Wastewater, Waste, Hennef, Germany.
1361 1362 1363
Azzopardi, B.J., Mudde, R.F., Lo, S., Morvan, H., Yan, Y. and Zhao, D. (2011) Hydrodynamics of gas-liquid reactors: normal operation and upset conditions, John Wiley & Sons, Ltd, Chichester, UK.
1364 1365 1366 1367 1368
Bhuyar, L.B., Thakre, S.B. and Ingole, N.W. (2009) Design characteristics of curved blade aerator w.r.t. aeration efficiency and overall oxygen transfer coefficient and comparison with CFD modeling. International Journal of Engineering, Science and Technology 1(1), 1-15.
1369 1370 1371
Bosma, A.J. and Reitsma, B. (2007) Hydraulic design of activated sludge tanks with CFD. Proceedings of the COMSOL Users Conference, October 23-24, Grenoble, France.
1372 1373 1374
Brannock, M. (2003) Computational Fluid Dynamics tools for the design of mixed anoxic wastewater treatmnent vessels. PhD Thesis, University of Queensland, Brisbane, Australia.
1375 1376 1377 1378
Bridgeman, J., Jefferson, B. and Parsons, S.A. (2009) Computational Fluid Dynamics modelling of flocculation in water treatment: A review. Engineering Applications of Computational Fluid Mechanics 3(2), 220-241.
AC C
EP
TE D
M AN U
SC
RI PT
1348 1349 1350
1379 1380 1381 1382
Brouckaert, C.J. and Buckley, C.A. (1999) The Use of Computational Fluid Dynamics for improving the design and operation of water and wastewater treatment plants Water Science & Technology 40(4-5), 81-89.
1383 1384 1385
Cartland Glover, G.M., Generalis, S.C. and Thomas, N.H. (2000) CFD and bubble column reactors: simulation and experiment. Chemical papers 54(6A), 361-369.
1386 1387 1388 1389
Chachuat, B., Roche, N. and Latifi, M.A. (2005) Long-term optimal aeration strategies for small-size alternating activated sludge treatment plants. Chemical Engineering and Processing: Process Intensification 44(5), 591-604.
72
ACCEPTED MANUSCRIPT Clift, R., Grace, J.R. and Weber, M.E. (1978) Bubbles, drops, and particles, Academic Press Inc., London, UK.
1392 1393 1394 1395
Cockx, A., Do-Quang, Z., Audic, J.M., Liné, A. and Roustan, M. (2001) Global and local mass transfer coefficients in waste water treatment process by computational fluid dynamics. Chemical Engineering and Processing 40(2), 187-194.
1396 1397 1398 1399
Combest, D.P., Ramachandran, P.A. and Dudukovic, M.P. (2011) On the gradient diffusion hypothesis and passive scalar transport in turbulent flows. Industrial & Engineering Chemistry Research 50(15), 8817-8823.
1400 1401 1402
Copp, J.B. (2001) The COST Simulation Benchmark: Description and simulator manual. Office for Official Publications of the European Community, Luxembourg.
1403 1404 1405
Cristea, S., de Prada, C., Sarabia, D. and Gutiérrez, G. (2011) Aeration control of a wastewater treatment plant using hybrid NMPC. Computers & Chemical Engineering 35(4), 638-650.
1406 1407
Danckwerts, P.V. (1953) Continuous flow systems. Chemical Engineering Science 2(1), 1-13.
1408 1409 1410 1411
De Araújo, A.C.B., Gallani, S., Mulas, M. and Olsson, G. (2011) Systematic approach to the design of operation and control policies in activated sludge systems. Industrial and Engineering Chemistry Research 50(14), 8542-8557.
1412 1413 1414 1415
De Clercq, B. (2003) Computational Fluid Dynamics of settling tanks:Development of experiments and rheological, settling, and scraper submodels. PhD Thesis, University of Ghent, Ghent, Belgium.
1416 1417 1418 1419
De Gussem, K., Fenu, A., Wambecq, T. and Weemaes, M. (2014) Energy saving on wastewater treatment plants through improved online control: case study wastewater treatment plant AntwerpSouth. Water Science & Technology 65(5), 1074-1079.
1420 1421 1422 1423
Dhanasekharan, K.M., Sanyal, J., Jain, A. and Haidari, A. (2005) A generalized approach to model oxygen transfer in bioreactors using population balances and computational fluid dynamics. Chemical Engineering Science 60(1), 213-218.
1424 1425 1426 1427
Do-Quang, Z., Cockx, A., Liné, A. and Roustan, M. (1999) Computational fluid dynamics applied to water and wastewater treatment facility modeling. Environmental Engineering and Policy 1(3), 137-147.
1428 1429 1430
EPA (1989) Design Manual: Fine pore aeration systems, United States Environmental Protection Agency, Research and Development - Risk Reduction Engineering Laboratory, Cincinnati, OH
1431 1432
EPA (2013) Energy efficiency in water and wastewater facilities - A guide to developing and implementing greenhouse gas reduction
AC C
EP
TE D
M AN U
SC
RI PT
1390 1391
73
ACCEPTED MANUSCRIPT 1433 1434
programs United Washington, DC.
1435 1436 1437 1438
Essemiani, K., Vermande, S., Marsal, S., Phan, L. and Meinhold, J. (2004) Optimization of WWTP units using CFD - a tool grown for real scale application. van Loosdrecht, M. and Clement, J. (eds), pp. 183-192, IWA Publishing.
1439 1440 1441
Fan, L., Xu, N., Wang, Z. and Shi, H. (2010) PDA experiments and CFD simulation of a lab-scale oxidation ditch with surface aerators. Chemical Engineering Research and Design 88(1), 23-33.
1442 1443 1444
Fayolle, Y., Cockx, A., Gillot, S., Roustan, M. and Héduit, A. (2007) Oxygen transfer prediction in aeration tanks using CFD. Chemical Engineering Science 62(24), 7163-7171.
1445 1446 1447 1448
Fayolle, Y., Gillot, S., Cockx, A., Bensimhon, L., Roustan, M. and Heduit, A. (2010) In situ characterization of local hydrodynamic parameters in closed-loop aeration tanks. Chemical Engineering Journal 158(2), 207-212.
1449 1450 1451 1452 1453
Fayolle, Y., Gillot, S., Cockx, A., Roustan, M. and Heduit, A. (2006) In situ local parameter measurements for CFD modelling to optimize aeration. Proceedings of WEFTEC 06 - 79th Annual WEF Technical Exhibition and Conference, pp. 3314-3325, Oct. 21-25, Dallas, TX, USA.
1454 1455 1456 1457
Fernández, F.J., Castro, M.C., Rodrigo, M.A. and Cañizares, P. (2011) Reduction of aeration costs by tuning a multi-set point on/off controller: A case study. Control Engineering Practice 19(10), 12311237.
1458 1459 1460
Fikar, M., Chachuat, B. and Latifi, M.A. (2005) Optimal operation of alternating activated sludge processes. Control Engineering Practice 13(7), 853-861.
1461 1462 1463 1464
Gillot, S., Capela-Marsal, S., Roustan, M. and Héduit, A. (2005) Predicting oxygen transfer of fine bubble diffused aeration systems Model issued from dimensional analysis. Water Research 39(7), 1379-1387.
1465 1466 1467
Gillot, S. and Héduit, A. (2000) Effect of air flow rate on oxygen transfer in an oxidation ditch equipped with fine bubble diffusers and slow speed mixers. Water Research 34(5), 1756-1762.
1468 1469 1470 1471
Glover, G.C., Printemps, C., Essemiani, K. and Meinhold, J. (2006) Modelling of wastewater treatment plants - how far shall we go with sophisticated modelling tools? Water Science & Technology 53(3), 79-89.
1472 1473 1474
Gresch, M., Armbruster, M., Braun, D. and Gujer, W. (2011) Effects of aeration patterns on the flow field in wastewater aeration tanks. Water Research 45(2), 810-818.
Environmental
Protection
Agency,
AC C
EP
TE D
M AN U
SC
RI PT
States
74
ACCEPTED MANUSCRIPT Griborio, A. and McCorquodale, J.A. (2006) Optimum design of your center well: use of a CFD model to understand the balance between flocculation and improved hydrodynamics. Proceedings of WEFTEC 06 - 79th Annual WEF Technical Exhibition and Conference, pp. 5735-5746, Oct. 21-25, Dallas, TX, USA.
1480 1481 1482 1483
Guimet, V., Savoye, P., Audic, J.-M. and Do-Quang, Z. (2004) Advanced CFD tool for Wastewater: today complex Modelling & tomorrow easy-to-use Interface. Van Loosdrecht, M. and Clement, J. (eds), pp. 165-172, IWA Publishing.
1484 1485 1486
Henze, M., Gujer, W., Mino, T. and van Loosdrecht, M.C.M. (2000) Activated Sludge Models ASM1, ASM2, ASM2d and ASM3. IWA Scientific and Technical Report No. 9, London, UK.
1487 1488 1489 1490
Holenda, B., Domokos, E., Rédey, Á. and Fazakas, J. (2007) Aeration optimization of a wastewater treatment plant using genetic algorithm. Optimal Control Applications and Methods 28(3), 191208.
1491 1492 1493 1494
Holenda, B., Domokos, E., Rédey, Á. and Fazakas, J. (2008) Dissolved oxygen control of the activated sludge wastewater treatment process using model predictive control. Computers & Chemical Engineering 32(6), 1270-1278.
1495 1496 1497
Hounslow, M.J., Ryall, R.L. and Marshall, V.R. (1988) A discretized population balance for nucleation, growth, and aggregation. AIChE Journal 34(11), 1821-1832.
1498 1499 1500 1501
Jensen, M.D., Ingildsen, P., Rasmussen, M.R. and Laursen, J. (2006) Computational Fluid Dynamics modelling of hydraulics and sedimentation in process reactors during aeration tank settling. Water Science & Technology 53(12), 257-264.
1502 1503 1504
Ji, Z., McCorquodale, J.A., Zhou, S. and Vitasovic, Z. (1996) A dynamic solids inventory model for activated sludge systems. Water Environment Research 68(3), 329-337.
1505 1506 1507
Joshi, J.B. (2001) Computational flow modelling and design of bubble column reactors. Chemical Engineering Science 56(21–22), 5893-5933.
1508 1509 1510 1511
Karpinska, A.M., Dias, M.M., Boaventura, R., A. and Santos, R.J. (2015) Modeling of the hydrodynamics and energy expenditure of oxidation ditch aerated with hydrojets using CFD codes. Water Quality Research Journal of Canada 50(1), 83-94.
1512 1513 1514
Karpinska Portela, A.M. (2013) New design tools for Activated Sludge process. PhD Thesis, FEUP, University of Porto, Porto, Portugal.
1515 1516 1517 1518
Kjellstrand, R. (2006) Hydraulic Behaviour in an Activated Sludge tank. From tracer test through hydraulic modelling to full-scale implementation. Lic. of Eng. Thesis, Chalmers University of Technology, Göteborg, Sweden
AC C
EP
TE D
M AN U
SC
RI PT
1475 1476 1477 1478 1479
75
ACCEPTED MANUSCRIPT Kochevsky, A.N. (2004) Possibilities of simulation of fluid flows using the modern CFD software tools. arXiv preprint physics/0409104.
1522 1523 1524 1525
Langergraber, G., Rieger, L., Winkler, S., Alex, J., Wiese, J., Owerdieck, C., Ahnert, M., Simon, J. and Maurer, M. (2004) A guideline for simulation studies of wastewater treatment plants. Water Science & Technology 50(7), 131-138.
1526 1527 1528
Launder, B.E., Reece, G.J. and Rodi, W. (1975) Progress in the development of a Reynolds-stress turbulent closure. Journal of Fluid Mechanics 68, 537-566.
1529 1530 1531
Launder, B.E. and Spalding, D.B. (1974) The numerical computation of turbulent flows. Computer Methods in Applied Mechanics and Engineering 3(2), 269-289.
1532 1533 1534 1535 1536
Laurent, J., Samstag, R.W., Ducoste, J., Griborio, A., Nopens, I., Batstone, D.J., Wicks, J., Saunders, S. and Potier, O. (2014) A protocol for the use of computational fluid dynamics as a supportive tool for wastewater treatment plant modelling. Water Science & Technology 70(10), 1757-1584.
1537 1538 1539 1540
Laursen, J. (2007) WWTPprocess tank modelling - The integration of advanced hydrodynamic and microbiological models. PhD Thesis, Department of Civil Engineering, Aalborg University, Aalborg, Denmark.
1541 1542 1543 1544
Le Moullec, Y., Gentric, C., Potier, O. and Leclerc, J.P. (2010a) CFD simulation of the hydrodynamics and reactions in an activated sludge channel reactor of wastewater treatment. Chemical Engineering Science 65(1), 492-498.
1545 1546 1547 1548
Le Moullec, Y., Gentric, C., Potier, O. and Leclerc, J.P. (2010b) Comparison of systemic, compartmental and CFD modelling approaches: Application to the simulation of a biological reactor of wastewater treatment. Chemical Engineering Science 65(1), 343-350.
1549 1550 1551 1552
Le Moullec, Y., Potier, O., Gentric, C. and Leclerc, J.P. (2011) Activated sludge pilot plant: Comparison between experimental and predicted concentration profiles using three different modelling approaches. Water Research 45(10), 3085-3097.
1553 1554 1555 1556
Le Moullec, Y., Potier, O., Gentric, C. and Pierre Leclerc, J. (2008) Flow field and residence time distribution simulation of a cross-flow gas-liquid wastewater treatment reactor using CFD. Chemical Engineering Science 63(9), 2436-2449.
1557 1558 1559 1560
Lei, L. and Ni, J. (2014) Three-dimensional three-phase model for simulation of hydrodynamics, oxygen mass transfer, carbon oxidation, nitrification and denitrification in an oxidation ditch. Water Research 53(0), 200-214.
1561 1562
Levenspiel, O. (1999) Chemical Reaction Engineering, Wiley & Sons, New York, NY.
AC C
EP
TE D
M AN U
SC
RI PT
1519 1520 1521
76
ACCEPTED MANUSCRIPT Makinia, J. and Wells, S.A. (2000) A general model of the activated sludge reactor with dispersive flow - I. Model development and parameter estimation. Water Research 34(16), 3987-3996.
1566 1567
Manninen, M., Taivassalo, V. and Kallio, S. (1996) On the mixture model for the multiphase flow. VTT Publications 288, 1-67.
1568 1569 1570
Marchisio, D.L., Vigil, R.D. and Fox, R.O. (2003) Quadrature method of moments for aggregation–breakage processes. Journal of Colloid and Interface Science 258(2), 322-334.
1571 1572 1573 1574
Menter, F.R. (1993) Zonal two equation k-w turbulence models for aerodynamic flows. Proceedings of 23rd Fluid Dynamics, Plasmadynamics and Lasers Conference, July 6-9, Orlando, FL, USA.
1575 1576
Menter, F.R. (1994) Two-equation eddy-viscosity turbulence models for engineering applications. AIAA Journal 32(8), 1598-1605.
1577 1578 1579 1580
Morchain, J., Gabelle, J.-C. and Cockx, A. (2014) A coupled population balance model and CFD approach for the simulation of mixing issues in lab-scale and industrial bioreactors. AIChE Journal 60(1), 27-40.
1581 1582 1583
Morchain, J., Maranges, C. and Fonade, C. (2000) CFD modelling of a two-phase jet aerator under influence of a crossflow. Water Research 34(13), 3460-3472.
1584 1585
Mueller, J.A., Boyle, W.C. and Pöpel, H.J. (2002) Aeration: principles and practice, CRC Press, Boca Raton, FL, USA.
1586 1587
Nauman, E.B. (2007) Residence Time Distributions, pp. 535-574, John Wiley & Sons, Inc., New Jersey, NJ, USA.
1588 1589 1590
Nielsen, M.K., Bechmann, H. and M., H. (2000) Modelling and test of aeration tank settling (ATS). Water Science & Technology 41(9), 179-184.
1591 1592 1593 1594 1595
Nopens, I., Batstone, D.J., Griborio, A., Samstag, R.W., Wicklein, E. and Wicks, J. (2012) Computational Fluid Dynamics (CFD): What is Good CFD-Modeling Practice and what can be the added value of CFD models to WWTP modeling?, pp. 7400-7405, Sept. 29-Oct. 3, New Orleans, LA, USA.
1596 1597 1598
Nopens, I., Koegst, T., Mahieu, K. and Vanrolleghem, P.A. (2005) PBM and activated sludge flocculation: From experimental data to calibrated model. AIChE Journal 51(5), 1548-1557.
1599 1600 1601 1602
Nopens, I., Torfs, E., Ducoste, J., Vanrolleghem, P.A. and Gernaey, K.V. (2015) Population balance models: a useful complementary modelling framework for future WWTP modelling. Water Science & Technology 71(2), 159-167.
1603 1604
Nopens, I. and Wicks, J. (2012) Integration of CFD and Biokinetics. CFD Workshop- Presentation. 3rd IWA/WEF Wastewater Treatment
AC C
EP
TE D
M AN U
SC
RI PT
1563 1564 1565
77
ACCEPTED MANUSCRIPT 1605 1606
Modelling Canada.
1607 1608
Orszag, S.A. (1970) Analytical theories of turbulence. Journal of Fluid Mechanics 41(2), 363-386.
1609 1610 1611 1612
Orszag, S.A., Staroselsky, I., Flannery, W.S. and Zhang, Y. (eds) (1996) Simulation and modeling of turbulent flows. Chapter 4: Introduction to Renormalization Group Modeling of Turbulence, Oxford University Press Inc., New York, NY, USA.
1613 1614 1615 1616
Patziger, M., Kainz, H., Hunze, M. and Józsa, J. (2012) Influence of secondary settling tank performance on suspended solids mass balance in activated sludge systems. Water Research 46(7), 24152424.
1617 1618 1619 1620
Pereira, J.P., Karpinska, A., Gomes, P.J., Martins, A.A., Dias, M.M., Lopes, J.C.B. and Santos, R.J. (2012) Activated Sludge Models coupled to CFD simulations. Dias, R., Lima, R., Martins, A.A. and Mata, T.M. (eds), pp. 153-173, Bentham Science Publishers Ltd.
1621 1622 1623
Pereira, J.P., Karpinska, A.M., Dias, M.M. and Santos, R.J. (2009) The dissolved oxygen profile in oxidation ditches: a numerical study on its effect on nutrient removal, pp. 97-104, Krakow, Poland.
1624 1625 1626 1627 1628
Plósz, B.G., De Clercq, J., Nopens, I., Benedetti, L. and Vanrolleghem, P.A. (2011) Shall we upgrade one-dimensional secondary settler models used in WWTP simulators? – An assessment of model structure uncertainty and its propagation. Water Science & Technology 63(8), 1726–1738.
1629 1630 1631 1632 1633 1634
Plósz, B.G., Nopens, I., Rieger, L., Griborio, A., De Clercq, J., Vanrolleghem, P.A., Daigger, G.T., Takács, I., Wicks, J. and Ekama, G.A. (2012) A critical review of clarifier modelling: State-of-the-art and engineering practices. Proceedings of WWTmod2012 - 3rd IWA/WEF Wastewater Treatment Modelling Seminar, pp. 27-30, Feb. 26-28, Mont-Sainte-Anne, Québec, Canada.
1635
Pope, S.B. (2000) Turbulent Flows, New York, NY, USA.
Mont-Sainte-Anne,
QC,
AC C
EP
TE D
M AN U
SC
RI PT
Seminar-WWTmod2012,
1636 1637
Randolf, A.D. and Larson, M.A. (1971) Theory of particulate processes, Academic Press, New York, NY, USA.
1638 1639 1640 1641 1642
Ratkovich, N. (2010) Understanding hydrodynamics in Membrane Bioreactor systems for wastewater treatment: Two-phase empirical and numerical modelling and experimental validation. PhD thesis, Faculty of Bioscience Engineering, Ghent University, Ghent, Belgium.
1643 1644 1645 1646 1647
Ratkovich, N., Hunze, M., Futselaar, H. and Nopens, I. (2009) Use of CFD to model and optimize the hydrodynamics of an airlift MBR side-stream module. Proceedings of WEFTEC 09 - 82th Annual WEF Technical Exhibition and Conference, pp. 2812-2818, Oct. 1014, Orlando, FL, USA
78
ACCEPTED MANUSCRIPT Reardon, D.J. (1995) Turning down the power. Civil Engineering 65(8), 54-56.
1650 1651 1652
Rieger, L., Alex, J., Gujer, W. and Siegrist, H. (2006) Modelling of aeration systems at wastewater treatment plants. Water Science & Technology 53(4-5), 439-447.
1653 1654
Rodi, W. (1993) Turbulence models and their application in hydraulics., IAHR Monograph, Delft, The Netherlands.
1655 1656 1657 1658
Rosso, D., Lothman, S.E., Jeung, M.K., Pitt, P., Gellner, W.J., Stone, A.L. and Howard, D. (2011) Oxygen transfer and uptake, nutrient removal, and energy footprint of parallel full-scale IFAS and activated sludge processes. Water Research 45(18), 5987-5996.
1659 1660
Rosso, D. and Stenstrom, M.K. (2006) Surfactant effects on α-factors in aeration systems. Water Research 40(7), 1397-1404.
1661 1662 1663 1664
Samstag, R.W., Ducoste, J.M., Griborio, A., Nopens, I., Batstone, D.J., Wicks, J.D., Saunders, S., Wicklein, E., Kenny, G. and Laurent, J. (2015) CFD for wastewater treatment: A State of the Art review. Submitted manuscript.
1665 1666 1667
Samstag, R.W., McCorqudale, J.A. and Zhou, S.P. (1992) Prospects for Transport Modelling in Process Tanks. Water Science & Technology 26(5-6), 1401-1410.
1668 1669 1670 1671
Samstag, R.W. and Wicklein, E. (2012) Application of CFD to Activated Sludge mixing analysis. CFD Workshop- Presentation. 3rd IWA/WEF Wastewater Treatment Modelling SeminarWWTmod2012, February 26-28 Mont-Sainte-Anne, QC, Canada.
1672 1673 1674 1675
Samstag, R.W. and Wicklein, E. (2014) A Protocol for optimization of Activated Sludge mixing. Proceedings of WEFTEC - 87th Annual WEF Technical Exhibition and Conference, Sept. 27-Oct. 1, New Orleans, LA, USA.
1676 1677 1678 1679 1680
Samstag, R.W., Wicklein, E., Reardon, R.D., Leetch, R.J., Parks, R.M. and Groff, C.D. (2012) Field and CFD analysis of jet aeration and mixing. Proceedings of WEFTEC 2012 - 85th Annual WEF Technical Exhibition and Conference, pp. 4113-4139, Sept. 29-Oct. 3, New Orleans, LA, USA.
1681 1682 1683
Shaikh, A. and Al-Dahhan, M.H. (2007) A Review on flow regime transition in bubble columns. International Journal of Chemical Reactor Engineering 5(1), 1-68.
1684 1685 1686 1687 1688
Sharma, A.K., Guildal, T., Thomsen, H.A., Mikkelsen, P.S. and Jacobsen, B.N. (2013) Aeration tank settling and real time control as a tool to improve the hydraulic capacity and treatment efficiency during wet weather: results from 7 years' full-scale operational data. Water Science & Technology 67(10), 2169-2176.
1689 1690
Shih, T.-H., Liou, W.W., Shabbir, A., Yang, Z. and Zhu, J. (1995) A New K-epsilon eddy viscosity model for high Reynolds number
AC C
EP
TE D
M AN U
SC
RI PT
1648 1649
79
ACCEPTED MANUSCRIPT turbulent flows: Model development and validation. Computers & Fluids 24(3), 227-238.
1693 1694
Smagorinsky, J. (1963) General circulation experiments with the primitive equations. Monthly Weather Review 93(3), 99-164.
1695 1696 1697 1698 1699 1700
Sobremisana, A.P., Ducoste, J.J. and de los Reyes III, F.L. (2011) Combining CFD, floc dynamics and biological reaction kinetics to model carbon and nitrogen removal in an activated sludge system. Proceedings of WEFTEC 2011 - 84th Annual WEF Technical Exhibition and Conference, pp. 3272-3282, Oct. 15-19, Los Angeles, CA, USA.
1701 1702 1703 1704
Sokolichin, A., Eigenberger, G., Lapin, A. and Lübert, A. (1997) Dynamic numerical simulation of gas-liquid two-phase flows Euler/Euler versus Euler/Lagrange. Chemical Engineering Science 52(4), 611-626.
1705 1706 1707
Stamou, A.I. (2008) Improving the hydraulic efficiency of water process tanks using CFD models. Chemical Engineering and Processing: Process Intensification 47(8), 1179-1189.
1708 1709 1710
Stamou, A.I., Theodoridis, G. and Xanthopoulos, K. (2009) Design of secondary settling tanks using a CFD model. Journal of Environmental Engineering 135(7), 551-561.
1711 1712
Takács, G. (2005) Gas Lift: Manual, PennWell Books, Tulsa, OK, USA.
1713 1714 1715 1716
Taricska, J.R., Huang, J.Y.C., Chen, P., Hung, Y.-T. and Zou, S.-W. (2009) Biological treatment processes. Wang, L.K., Pereira, N.C., Hung, Y.-T. and Shammas, N.K. (eds), The Humana Press, Totowa, NJ.
1717 1718 1719
Tchobanoglous, G., Burton, F.L., Stensel, H.D. and Eddy, M. (eds) (2003) Wastewater engineering: treatment and reuse, McGraw-Hill New York, NY.
1720 1721 1722
Vedantam, S., Joshi, J.B. and Koganti, S.B. (2006) Threedimensional CFD simulation of stratified two-fluid Taylor-Couette flow. The Canadian Journal of Chemical Engineering 84(3), 279-288.
1723 1724 1725 1726
Vermande, S., Simpson, K., Essemiani, K., Fonade, C. and Meinhold, J. (2007) Impact of agitation and aeration on hydraulics and oxygen transfer in an aeration ditch: Local and global measurements. Chemical Engineering Science 62(9), 2545-2555.
1727 1728 1729
Versteeg, H.K. and Malalasekera, W. (1995) An introduction to Computational Fluid Dynamics. The Finite Volume Method, Pearsons Prentice Hall, Harlow, Essex, UK.
1730 1731 1732
Wang, T. (2011) Simulation of bubble column reactors using CFD coupled with a population balance model. Frontiers of Chemical Science and Engineering 5(2), 162-172.
AC C
EP
TE D
M AN U
SC
RI PT
1691 1692
80
ACCEPTED MANUSCRIPT Wang, Y., Waite, T.D. and Leslie, G.L. (2013) Handbook of membrane reactors. Basile, A. (ed), pp. 532-568, Woodhead Publishing, Cambridge, UK.
1736 1737 1738
WEF (2009) Energy conservation in water and wastewater treatment facilities, Water Environment Federation Manual of Practice No. 32; McGraw-Hill, Inc., New York, NY.
1739 1740 1741 1742
Wicklein, E. and Samstag, R.W. (2009) Comparing commercial and transport CFD models for secondary sedimentation. Proceedings of WEFTEC 09 - 82th Annual WEF Technical Exhibition and Conference, pp. 6066-6081, Oct. 10-14, Orlando, FL, USA
1743 1744 1745
Wilcox, D.C. (1998) Re-assessment of the scale-determining equation for advanced turbulence models. AIAA Journal 26(11), 1299-1310.
1746 1747 1748
Wilcox, D.C. and Traci, R.M. (1976) A complete model of turbulence. Proceedings of 9th Fluid and PlasmaDynamics Conference, July 14-16, San Griego, CA, USA.
1749 1750 1751 1752
Xie, H., Yang, J., Hu, Y., Zhang, H., Yang, Y., Zhang, K., Zhu, X., Li, Y. and Yang, C. (2014) Simulation of flow field and sludge settling in a full-scale oxidation ditch by using a two-phase flow CFD model. Chemical Engineering Science 109(0), 296-305.
1753 1754 1755 1756
Xu, N., Fan, L., Pang, H. and Shi, H. (2010) Feasibility study and CFD-aided design for a new type oxidation ditch based on airlift circulation. The Canadian Journal of Chemical Engineering 88(5), 728-741.
1757 1758
Yang, C. and Mao, Z.-S. (2014) Numerical simulation of multiphase reactors with continuous liquid phase, Academic Press, Oxford, UK.
1759 1760 1761 1762
Yang, Y., Yang, J., Zuo, J., Li, Y., He, S., Yang, X. and Zhang, K. (2011) Study on two operating conditions of a full-scale oxidation ditch for optimization of energy consumption and effluent quality by using CFD model. Water Research 45(11), 3439-3452.
1763 1764
Yeoh, G.H. and Tu, J. (2009) Computational techniques for multiphase flows, Elsevier.
AC C
EP
TE D
M AN U
SC
RI PT
1733 1734 1735
1765 1766
81
ACCEPTED MANUSCRIPT Figure captions Figure 1. Streamlines coloured by liquid velocity magnitude ܷ in tank (a) without aeration (ܷ = 0.35 ms−1) (b) without aeration (ܷ = 0.27 ms−1) and (c) with aeration (ܷ = 0.23 ms−1). Reprinted from Fayolle et al. (2007). Copyright (2007) with permission from Elsevier.
RI PT
Figure 2. Snapshot of air volume fractions at two cross sections with different diffuser patterns. Reprinted from Gresch et al. (2011). Copyright (2011) with permission from Elsevier.
Figure 3. Contours of volume fraction of solid phase at different height: (A) top of the tank, (B) middle of the tank, and (C) bottom of the tank. Reprinted from Xie et al. (2014). Copyright (2014) with permission from Elsevier.
M AN U
SC
Figure 4. Comparison between experimental and simulated RTD obtained with the RSM and the k – ε turbulence models and the particle tracking method for a liquid flowrate of 3.6 L min-1 and a gas flowrate of 15 L min-1. Reprinted from Le Moullec et al. (2008). Copyright (2008) with permission from Elsevier.
AC C
EP
TE D
Figure 5. The contour plots of: (a) MLSS, (b) DO, (c) COD, (d) ammonia and (e) nitrate concentration distribution, where “+” indicates measured data values at the sampling points in a Carrousel oxidation ditch. Reprinted from Lei and Ni (2014). Copyright (2014) with permission from Elsevier.
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT
Highlights
EP
TE D
M AN U
SC
RI PT
CFD analysis is a robust tool for evaluation and optimization of the AS systems. Turbulence models and multiphase approaches used to study AS tanks are presented. Pitfalls of modelling assumptions and simplifications are identified. Methods and examples of coupling of the CFD data with biokinetics are discussed. Unaddressed challenges in modelling of the AS systems are critically discussed.
AC C
• • • • •
S0043-1354(15)30336-5
DOI:
10.1016/j.watres.2015.11.008
Reference:
WR 11635
To appear in:
Water Research
Received Date: 25 June 2015 Revised Date:
1 November 2015
Accepted Date: 2 November 2015
Please cite this article as: Karpinska, A.M., Bridgeman, J., CFD-Aided Modelling of Activated Sludge Systems – A Critical Review, Water Research (2015), doi: 10.1016/j.watres.2015.11.008. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT 1
Title: CFD-Aided Modelling of Activated Sludge Systems –
2
A Critical Review.
3 4
Authors: Anna M. Karpinskaa, John Bridgemana
5
a
6
Edgbaston, Birmingham, B15 2TT, United Kingdom
7
8
[email protected]
9
Corresponding
RI PT
[email protected];
author:
Anna
([email protected])
SC
addresses:
M.
Karpinska
M AN U
10
– School of Civil Engineering, University of Birmingham,
11 Abstract
13
Nowadays, one of the major challenges in the wastewater
14
sector is the successful design and reliable operation of
15
treatment
16
efficiencies to comply with effluent quality criteria, while
17
keeping the investment and operating cost as low as possible.
18
Although conceptual design and process control of activated
TE D
12
which
guarantee
high
treatment
AC C
EP
processes,
19
sludge plants are key to ensuring these goals, they are still
20
based
21
experience, dominated often by rule of thumb. This review
22
paper discusses the rationale behind the use of Computational
23
Fluid Dynamics (CFD) to model aeration, facilitating
24
enhancement of treatment efficiency and reduction of energy
25
input. Several single- and multiphase approaches commonly
on
general
empirical
1
guidelines
and
operators’
ACCEPTED MANUSCRIPT used in CFD studies of aeration tank operation, are
27
comprehensively described, whilst the shortcomings of the
28
modelling assumptions imposed to evaluate mixing and mass
29
transfer in AS tanks are identified and discussed. Examples and
30
methods of coupling of CFD data with biokinetics, accounting
31
for the actual flow field and its impact on the oxygen mass
32
transfer and yield of the biological processes occurring in the
33
aeration tanks, are also critically discussed. Finally, modelling
34
issues, which remain unaddressed, (e.g. coupling of the AS
35
tank with secondary clarifier and the use of population balance
36
models to simulate bubbly flow or flocculation of the activated
37
sludge), are also identified and discussed.
M AN U
SC
RI PT
26
38 Keywords
40
Activated Sludge; aeration; Computational Fluid Dynamics;
41
hydrodynamics; multiphase models; turbulence
Nomenclature
AC C
43
EP
42
TE D
39
2
ACCEPTED MANUSCRIPT mass fraction of phase [-]
sum of interfacial forces shared by the phases [N]
sum of interfacial forces between the continuous and dispersed phases [N] diffusion flux of ′ species [kg m-2 s-1]
turbulent kinetic energy, [m2 s-2]
M AN U
SC
RI PT
volumetric mass transfer coefficient [h-1]
mass transfer from phase to [kg s-1]
mass transfer from phase to [kg s-1]
TE D
pressure [Pa]
̅
averaged pressure field [Pa]
AC C
EP
net rate of production of ′ species due to chemical
Reynolds number [-]
reaction [kg m-3 s-1]
3
ACCEPTED MANUSCRIPT
source term representing rate of creation of ′ species from dispersed phase and any user-defined sources [kg
time [s]
particle relaxation time [s]
velocity [m s-1]
, ,
drift velocity for phase [m s-1]
M AN U
,
SC
RI PT
m-3 s-1]
velocity components [m s-1]
averaged velocity term [m s-1]
′
fluctuating velocity term [m s-1]
velocity of mixture [m s-1]
AC C
EP
TE D
̅
velocity of the particle [m s-1]
velocity of phase [m s-1]
velocity of phase relative to velocity of phase [m s-1]
4
ACCEPTED MANUSCRIPT velocity of phase [m s-1]
distance [m]
,
direction) [m] mass fraction of the ’-th species [-]
SC
!
spatial coordinates (displacement in the streamwise
Greek letters
"∗ $
damping function coefficient [-]
turbulent kinetic energy dissipation rate [m2 s-3] dynamic viscosity of the fluid [Pa s]
EP
%
phasic volume fraction [-]
TE D
"
M AN U
44 45
RI PT
,
dynamic viscosity of the mixture [Pa s]
AC C
% % %& '
'
dynamic viscosity of the phase [Pa s] turbulent viscosity [Pa s]
fluid density [kg m-3] density of the mixture [kg m-3]
5
ACCEPTED MANUSCRIPT '
density of the particle [kg m-3] density of the phase [kg m-3]
(
specific turbulence dissipation (frequency)
'
density of the phase [kg m-3]
[s-1]
)*
′ +
,
refers to drift velocity index or counter species index
index or counter
refers to turbulence kinetic energy index or counter
refers to mixture
EP
-
M AN U
Indices
TE D
47
SC
46
refers to particle
refers to phase
refers to turbulence
refers to phase
AC C
RI PT
'
48 49 50
Abbreviations
51
ADV – Acoustic Doppler Anemomentry
52
AS – Activated Sludge 6
54
ASM1 – Activated Sludge Model No. 1
55
ASP – Activated Sludge Plant
56
BOD – Biochemical Oxygen Demand
57
CFD – Computational Fluid Dynamics
58
COD – Chemical Oxygen Demand
59
CPU – Central Processing Unit
60
CSTR – Continuous Stirred Tank Reactor
61
DO – Dissolved Oxygen
62
LDV – Laser Doppler Velocimetry
63
LES – Large |Eddy Simulation
64
MDV– Mono-directional Velocimetry
65
MLSS – Mixed Liquor Suspended Solids
66
MRF – Multiple Reference Frames
67
PBM – Particle Balance Model
68
PDA – Particle Dynamic Analysis
69
PFR – Plug Flow Reactor
70
PID – Proportional-Integral-Derivative
SC
ASM – Activated Sludge Model
AC C
EP
TE D
M AN U
53
RI PT
ACCEPTED MANUSCRIPT
71
QMOM – Quadrature Method of Moments
72
RAM – Random Access Memory
73
RANS – Reynolds Averaged Navier- Stokes Simulation
74
RSM – Reynolds Stress Model
75
RTD – Residence Time Distribution
76
SCADA – Supervisory Control and Data Acquisition
77
SGS – Sub-grid Scale
7
ACCEPTED MANUSCRIPT SMM – Standard Method of Moments
79
SRT – Solids Retention Time
80
TSS – Total Suspended Solids
81
URANS – Unsteady Reynolds Averaged Navier- Stokes
82
Simulation
83
WWTP – Wastewater Treatment Plant
84
RI PT
78
1. Introduction
86
To comply with global water policy focussed on responsible
87
management of water resources and protection of public health,
88
wastewater collected from municipalities and communities
89
must be treated to achieve levels imposed by discharge permits
90
and maximum daily loads, allowing its subsequent return to
91
receiving water bodies, or to the land or even to be reused. In
92
the last century, the application of scientific knowledge and
93
engineering practice led to significant developments in the
94
wastewater
95
treatment based on aerobic biological methods, specifically the
96
activated sludge (AS) process (Ardern and Lockett 1914),
97
which is now a well-documented standard for many wastewater
98
treatment utilities. The objectives of secondary biological
99
wastewater treatment in the AS process have also expanded
100
from an early emphasis on high levels of Biological Oxygen
101
Demand (BOD) and Total Suspended Solids (TSS) removal to
102
cover enhanced nutrients (N and P) removal, as the process
TE D
M AN U
SC
85
particularly in
biological
secondary
AC C
EP
sector,
8
ACCEPTED MANUSCRIPT itself has flexibility and numerous modifications can be tailored
104
to meet specific requirements.
105
Undoubtedly, one of the foremost challenges in the wastewater
106
sector is the successful design and reliable operation of
107
treatment plant, which guarantee high treatment efficiencies in
108
order to meet the effluent quality standards defined by
109
regulators, while keeping the investment and operating cost as
110
low as possible (Brouckaert and Buckley 1999, Do-Quang et al.
111
1999). One of the characteristic features of ASP is continuous
112
operation of the aeration process and sludge and nitrates
113
recycling, and thus the process performance relies on a steady
114
energy supply for operation of air blowers and sludge and
115
mixed liquor recirculation pumps. Aeration accounts for the
116
largest fraction of a total wastewater treatment plant’s (WWTP)
117
energy bill, ranging from 45 to 75% (Reardon 1995, Rieger et
118
al. 2006) and in extreme cases of stringent effluent nitrogen
119
criterion requiring enhanced nitrification, even up to 85%.
120
Thus, aeration has a significant effect on the operation and
AC C
EP
TE D
M AN U
SC
RI PT
103
121
maintenance budget of water utilities (WEF 2009). As a
122
consequence of the global emphasis on the water-energy-food-
123
climate change nexus there is an urgent need to reduce energy
124
usage
125
conservation measures, engineering practices and management
126
programs. According to guidelines (EPA 2013, WEF 2009)
127
opportunities for improving energy efficiency in wastewater
at
WWTPs
by
imposing
9
cost-effective
energy
ACCEPTED MANUSCRIPT treatment utilities can be obtained by optimizing aeration
129
processes or equipment upgrades, which focus on replacing
130
items such as blowers with more efficient models; replacing the
131
whole aeration system with less energy intensive systems (e.g.
132
replacement of the surface aeration system by porous diffusers
133
in full floor coverage configuration); and operational
134
modifications, involving reduction of the energy requirements
135
to perform specific functions by modification of the aeration
136
control systems, e.g. on-off operation allowing formation of the
137
anoxic conditions for denitrification. The last option facilitates
138
greater savings than equipment upgrades, and may not require
139
capital investment. Nevertheless, current best available aeration
140
technology (i.e. membrane diffusers supplied by atmospheric
141
air) are characterized by relatively low Standard Oxygen
142
Transfer Efficiencies of around 40 up to 60% (EPA 1989,
143
Mueller et al. 2002, Taricska et al. 2009), which reduce with
144
time as a result of fouling and scaling. Consequently, a wide
145
range of multidisciplinary approaches contribute to current
AC C
EP
TE D
M AN U
SC
RI PT
128
146
research aiming to improve the development, troubleshooting
147
and management of aeration systems.
148
2. Current Trends in Engineering Practice
149
2.1. Aeration Control and Design Assumptions
150
When investigating performance and energy expenditure of AS
151
plants it is clear that dissolved oxygen (DO) concentration is a
152
key process variable, which controls both, nutrient removal (in 10
ACCEPTED MANUSCRIPT the case of biological nutrient removal plants) and thus effluent
154
quality, and the operating cost of the utility. While operating
155
DO profiles and nitrogen patterns in the AS system are usually
156
obtained from biokinetics modelling with Activated Sludge
157
Model No. 1- ASM1 (Henze et al. 2000) employed by
158
supervisory control and data acquisition (SCADA) systems, the
159
robust modelling of the optimization strategies requires
160
implementation of complex computational algorithms and off-
161
line optimization techniques, which allow for coupling of the
162
biological process with pre-defined control variables such as
163
minimal DO concentration in the aeration tank or effluent
164
nitrogen criterion (Åmand and Carlsson 2012, Chachuat et al.
165
2005, Cristea et al. 2011, De Araújo et al. 2011, Fernández et
166
al. 2011, Fikar et al. 2005, Holenda et al. 2007, Holenda et al.
167
2008).
168
Although hydraulic design of wastewater treatment systems is a
169
crucial step to assure reliable and energy-optimised operation
170
of the process, it is usually labelled as a “low-tech” task, which
AC C
EP
TE D
M AN U
SC
RI PT
153
171
is based on empirical guidelines without sound theoretical basis
172
(Bosma and Reitsma 2007, Pereira et al. 2012, Stamou 2008).
173
Thus, in the majority of designs, flow behaviour in the unit
174
process tanks is predicted via the ideal reactor model while the
175
actual reactor hydrodynamics are not taken into account
176
(Stamou 2008). The classic example for Activated Sludge
177
Plants (ASPs) is the assumption of a completely mixed flow 11
ACCEPTED MANUSCRIPT regime in aerobic, anoxic and anaerobic tanks. Other
179
commonly practised rules of thumb are (Samstag and Wicklein
180
2012, Tchobanoglous et al. 2003) the assumption that in AS
181
basins equipped with diffused aeration systems, the air
182
requirement to ensure good mixing will vary from 1.2 – 1.8 m3
183
h-1 / m3 of tank volume; and typical power requirements for
184
maintaining a completely mixed flow regime with mechanical
185
aerators varies from 13 to 26 W per m3 of tank volume. None
186
of these assumptions consider the impact of tank hydraulics
187
(cross-section, depth or presence of baffles), energy input, or
188
any variable affecting mixing, such as local density gradients
189
due to solids transport. Furthermore, none of these guidelines
190
defines clearly hydraulic features and performance of
191
“completely mixed” wastewater treatment systems. The one
192
commonly used criterion for “good mixing” in AS process
193
control is that variations of solids concentration across the
194
complete mixed tank profile should be less than 10% (Samstag
195
and Wicklein 2014) However, the proper design of such “well
AC C
EP
TE D
M AN U
SC
RI PT
178
196
mixed” AS systems requires a sophisticated analysis of flow
197
behaviour accounting for mixing patterns in the tank,
198
distribution of the oxygen, solids and determination of local
199
densities.
200
3. Dynamic Behaviour of AS Tanks
201
Dynamic modelling of wastewater treatment plants has been
202
shown to be a powerful tool providing detailed insight into the 12
ACCEPTED MANUSCRIPT 203
unit process and system behaviour, useful for optimization
204
studies
205
evaluation, operational optimization, or controller design) and
206
model-based process control (Langergraber et al. 2004).
207
Nonetheless, the majority of the ASP design procedures are still
208
limited to the empirical principles and static models, such as
209
the ‘classic’ ATV-A-131 guideline (ATV-DVWK 2000), while
210
control strategies are based on a simple or cascade
211
proportional-integral-derivative (PID) controllers for DO and
212
ammonia. While not often used for design and control of ASP
213
due to its complexity, the systemic approach based on well-
214
established ASM models focuses mainly on the reactions of
215
biochemical conversion within the ideal reactors: usually one or
216
a cascade of Continuous Stirred Tank Reactors (CSTRs) or
217
Plug Flow Reactor (PFR) (Abusam and Keesman 1999, Le
218
Moullec et al. 2011, Makinia and Wells 2000, Pereira et al.
219
2009). Such approaches enable only quantitative prediction of
220
the oxygen consumption and nutrients removal and qualitative
conceptual
process
design,
performance
AC C
EP
TE D
M AN U
SC
RI PT
(e.g.
221
assessment of biomass growth and decay. Nevertheless, as the
222
overall biochemical conversion reactions occurring in AS are of
223
orders greater than zero, the wastewater treatment efficiency in
224
such non-ideal systems will depend on the bioreactor’s
225
hydrodynamics (Le Moullec et al. 2008), spatial distribution of
226
oxygen, and temperature. Therefore, a more accurate forecast
227
of the actual local scale phenomena occurring within the tank,
13
ACCEPTED MANUSCRIPT (such as DO patterns and solids profiles) is of crucial
229
importance to enhance system design, process efficiency and
230
energy requirements, to predict failures of existing aeration
231
systems and finally, to design upgrades and enhance control
232
strategy, such as fine-tuning blower operation.
233
The physics of typical AS systems is complex, not only due to
234
presence of the multiphase (gas-liquid-solid) flow, comprising
235
mixed liquor and air/oxygen, but also due to the different
236
length scales between the sludge flocs, bubbles and tank
237
geometry; and furthermore, different velocities of the phases
238
imparted by mixers and aerators yielding turbulent Reynolds
239
numbers (Karpinska Portela 2013, Pereira et al. 2012). The
240
fluid flow in such bioreactors is governed by the vessel
241
geometry, physical properties of its contents (phase, density,
242
viscosity) and operating condition variables (flow rates and
243
concentrations). At the same time, fluid flow governs the local
244
concentration of components, interphase contact and mass
245
transfer, reaction conversions and performance (Nopens and
AC C
EP
TE D
M AN U
SC
RI PT
228
246
Wicks 2012).
247
In engineering practice, whilst assessment of the flow regime,
248
and thus overall mixing phenomena in AS bioreactors, can be
249
achieved via experimental assessment of local flow velocities
250
through a tracer technique, the dimensions of full scale units
251
generally render this unfeasible (Pereira et al. 2012, Stamou
14
ACCEPTED MANUSCRIPT 2008), making simulation an attractive alternative. Several
253
factors contribute to the increasing popularity of modelling in
254
engineering practice, as it is a cost and time efficient solution,
255
which allows evaluation of process performance (whether a
256
new unit or modification will operate properly); prediction of
257
consequences
258
quantification of bottlenecks in liquid or solid handling lines in
259
the AS system. Thus, prediction of Residence Time
260
Distributions (RTDs) of settled sewage is a fundamental tool to
261
help understand and analyse a flow system providing realistic
262
information on tank hydrodynamics (Danckwerts 1953,
263
Levenspiel 1999, Nauman 2007). Furthermore, the RTD yields
264
information about the macromixing within the reactor, allowing
265
recognition of mixing behaviours that are not plug flow or
266
complete mixing regimes, and that are usually described by
267
tank-in-series models or even more complex arrangements of
268
the unit reactors (Pereira et al. 2012). Consequently, successful
269
biological
implementation;
isolation
and
EP
TE D
M AN U
SC
before
RI PT
252
treatment
modelling
combining
AC C
wastewater
270
hydrodynamics, mass transfer and biochemical reactions
271
kinetics remains one of the major goals in wastewater
272
engineering (Le Moullec et al. 2010a, b, Morchain et al. 2014,
273
Pereira et al. 2012).
274
4. Computational Fluid Dynamics
275
Since the 1970s, increasing computational power has been
276
accompanied by a rapid development in the software intended 15
ACCEPTED MANUSCRIPT 277
for solution of fluid flow problems. Nowadays, a wide range of
278
software suites intended for the solution of complex fluid flow
279
problems is commercially available. Initially, CFD was almost
280
exclusively
281
industries allowing simulation of the processes occurring in
282
combustion chambers of rocket engines; physico-chemical
283
processes in the flow around rocket airframe and supersonic
284
aircrafts. Subsequently CFD found applications with chemical
285
engineers, mainly for design of reaction vessels, and in the last
286
two decades, application of CFD has been extended to the civil
287
and environmental engineering sectors (Kochevsky 2004).
288
Recent developments in multiphase flow research have seen a
289
steady growth in the application of CFD modelling in
290
wastewater treatment, with a focus on the design of pumping
291
stations, headworks, screens, grit chambers, flow splitters, AS
292
tanks, clarifiers and digesters. Undoubtedly, one of the great
293
opportunities of CFD modelling of ASPs is the analysis of the
294
multiphase flow behaviour and prediction of the impact of a
with
aerospace
and
mechanical
AC C
EP
TE D
M AN U
SC
RI PT
associated
295
wide range of operating parameters on the local scale
296
phenomena, such as flow field coupled with interfacial mass
297
transfer and chemical reaction. In addition to that, CFD has
298
gained popularity over traditional wastewater treatment
299
modelling approaches, as it is a high-precision technique
300
allowing evaluation of the engineering systems, which are
301
expensive, difficult or even dangerous to reproduce in
16
ACCEPTED MANUSCRIPT laboratory-scale, pilot-scale or field conditions. Therefore,
303
CFD-aided modelling can be used as a robust tool for the
304
design of a new facility or the optimization or retrofitting of
305
existing ASPs, leading to enhanced performance and energy-
306
optimised operation of the utility, facilitating time, economic
307
cost and manpower savings (De Gussem et al. 2014, Do-Quang
308
et al. 1999, Essemiani et al. 2004, Guimet et al. 2004, Laurent
309
et al. 2014).
310
4.1. Hydrodynamics – RANS and URANS
311
Various options exist for the numerical simulation of the
312
turbulent flow with CFD codes.
313
In most engineering practice, time-averaged properties of the
314
flow are able to provide the required information. Steady
315
Reynolds Averaged Navier-Stokes (RANS) simulations and
316
unsteady RANS (URANS) focus on the representation of the
317
effects of turbulence on the mean flow properties by solving
318
transport equations for the averaged flow quantities with whole
319
range of the turbulent scales being modelled. Thus this
AC C
EP
TE D
M AN U
SC
RI PT
302
320
modelling approach greatly reduces required computational
321
effort and resources, and is widely adopted for practical
322
engineering applications, including the wastewater sector
323
(Karpinska Portela 2013).
324
In RANS and URANS the flow patterns within the AS tank are
325
obtained from the solution of nonlinear partial differential
326
equations, expressing balances of mass and momentum. 17
ACCEPTED MANUSCRIPT 327
Therefore, the flow is governed by the following mass
328
conservation equation: .' =0 .
RI PT
329
(1)
and momentum conservation equation, which for RANS is:
330
(2)
M AN U
. + <−' = .
SC
. .̅ . . . 2 . 1' 2 = − + 5% 6 + − 9 :; . . . . . 3 .
and for time dependent, transient flow (URANS) is:
TE D
.̅ . + 1' 2 . .
=−
.̅ . . . 2 . + 5% 6 + − 9 :; . . . . 3 . (3)
. <−' = .
EP AC C
+
331
332 333 334
where ̅ is the averaged pressure field, ' and % are the fluid
density and viscosity, respectively, is the time,
,
and
are the spatial coordinates, , and are the velocity components and 9 is the Kronecker delta.
18
ACCEPTED MANUSCRIPT 335
Here, the turbulent mean velocity field is described by
336
Reynolds decomposition of the velocity using time averaged
337
term, ̅ , and a fluctuating term, : = ̅ +
RI PT
(4)
RANS and URANS equations are linearized and solved. The
339
flow structures originating from the momentum transfer by the
340
fluctuating velocity field, which are smaller than the numerical
341
grid discretization and represented by the term
342
namely Reynolds stresses, are unclosed, and thus they must be
343
modelled.
344
Turbulence models
345
The
346
commercial CFD software suites, which are used for Reynolds
347
stresses closure are: standard, renormalized group (RNG) and
M AN U models
included
in
popular
TE D
turbulence
realizable − $ models; standard and shear stress transport
(SST) − ( models; and Reynolds stress model (RSM)
AC C
349
principal
−' ,
EP
348
SC
338
350
(Bridgeman et al. 2009).
351
The two-equation models (i.e. − $ and − (), solve
352
additional transport equations for two turbulence quantities:
353 354 355 356
velocity scale- turbulent kinetic energy, , and turbulence length scale- either its dissipation rate, $, or the specific
frequency, ( (Pope 2000). From among the two-equation
models, the standard − $ (> − $) has found the broadest 19
ACCEPTED MANUSCRIPT range of applicability in both, academia and industrial sectors,
358
due to its robustness, relatively low computational requirements
359
and satisfactory accuracy. The characteristic feature of all two-
360
equation models is that modelling of the Reynolds stresses
361
employs the Boussinesq hypothesis relating these stresses to the
362
mean deformation rates and thus mean velocity gradients, and
363
assuming locally isotropic turbulence (Rodi 1993):
RI PT
357
SC
. . 2 . −' = %& 6 + : − ?' + %& @9 . . 3 .
365
where %& is the turbulent viscosity computed as a function of
M AN U
364
(5)
and $:
TE D
%& = 'AB
C $
(6)
where AB is a model constant (variable function in a different
367
turbulence model).
368
A comprehensive description of the > − $ and the relatively
369
recently developed improved models, namely Renormalized
AC C
EP
366
370
Group (RNG) − $ and realizable − $, can be found in the
371
literature (Launder and Spalding 1974, Orszag et al. 1996, Shih
372
et al. 1995).
373
The second widely used two-equation model, introduced by
374
Wilcox, is the − ( model, also based on the isotropic eddy
20
ACCEPTED MANUSCRIPT
376 377
viscosity hypothesis described by Equation (5), but where the
turbulent viscosity is computed as a function of and (.
Knowing, that ( = $ ⁄ , Equation (6) takes the following form: %& = '" ∗
(
(7)
RI PT
375
where " ∗ is a coefficient related to the use of functions
379
damping the turbulent viscosity causing a low-Reynolds-
380
number correction.
381
Contrary to the − $ model, the standard − ( model uses
382
enhanced wall treatment to solve low-Re-number flows in the
383
viscous layer in near-wall region. Its improved variant, SST
M AN U
SC
378
− ( model, considered to be the most accurate from two-
385
equation eddy viscosity models, provides modified turbulent
386
viscosity formulation which account for the transport of the
387
principal turbulent shear stress.
TE D
384
A wider discussion on the
standard and SST − ( models can be found in Wilcox
389
(Wilcox 1998, Wilcox and Traci 1976) and Menter (Menter
390
1993, 1994).
391
The Reynolds Stress Model (RSM) (Launder et al. 1975,
392
Versteeg and Malalasekera 1995) is the most elaborate and
393
complex turbulence model, referred as the Second Order
394
Closure. In RSM the isotropic eddy viscosity hypothesis is
395
discarded and the RANS equations are closed by solving
396
transport equations for the Reynolds stresses, together with an
AC C
EP
388
21
ACCEPTED MANUSCRIPT equation for the dissipation rate, yielding seven additional
398
transport equations to be solved in a 3D scheme. A detailed
399
description of the RSM turbulent closure development can be
400
found in the literature (Launder et al. 1975), whilst a
401
summarised description of the turbulence models commonly
402
used for RANS closure has been summarized in Table 1
403
(Bridgeman et al. 2009).
AC C
EP
TE D
M AN U
SC
RI PT
397
22
ACCEPTED MANUSCRIPT
Table 1. Comparison of turbulence models used in ASP modelling. Adapted from Bridgeman et al. (2009) with permission from Taylor &
405
Francis Ltd (www.tandfonline.com/) Standard E − F (GE − F)
Comments
Renormalized Group (RNG) E−F
• Based on the statistical methods, not observed
Advantages
• Semi-empirical modelling of and $. • Valid only for fully developed turbulent flow cores (molecular diffusion ignored).
• Simplest and complete turbulence model. • Excellent performance for many flows. • Well established in academia and industry. • Robust, economic in terms of computational effort and satisfactory accuracy in diverse turbulent flow issues.
• • •
TE D
• Improved performance for swirling
and high-strained flows compared to > − $.
EP
•
fluid behaviour. Mathematics is highly abstruse. Texts only quote model equations which result from it. Effects of small-scale turbulence represented by means of random forcing function in NavierStokes equations. Procedure systematically removes small scales of motion by expressing their effects in terms of larger scale motions and a modified viscosity. Similar in form to > − $, but modified $ equation to describe high-strain flows better. Differential equation solved for %& (changes AB from 0.09 to 0.0845 at high Re).
AC C
•
M AN U
SC
Model
RI PT
404
406
23
Disadvantages • Poor performance in some scenarios (strong streamline curvature, vortices, rotating flows, flow separation, adverse pressure gradients). • Assumes locally isotropic turbulence. • Poor prediction of the lateral expansion in 3D wall jets.
• Less stable than > − $.
ACCEPTED MANUSCRIPT
Table 1. (Continued). Advantages • Suited for planar and rounded jets, swirling and separating flows and wall-bounded flows with strong adverse pressure gradients.
Standard E − H
• Recent development. In highly strained flows, the normal Reynolds stresses become negative (unrealizable condition), so %& uses variable AB . • AB is function of local strain rate and fluid rotation. • Different source and sink terms in transport equations for eddy dissipation. • Good for spreading rate of round jets.
• Specific dissipation rate is ( = $/ . • > − $ solves for dissipation of turbulent kinetic energy, − ( solves for rate at which dissipation occurs. • Resolves near wall region without wall functions so can be applied through boundary layer.
408
SC
M AN U
• As − ( except from gradual change from
− ( to inner region of boundary layer to high Re version of > − $ in outer part. • Modified %& formulation to account for transport effects of principal turbulent shear stresses.
• Not recommended to use with multiple reference frames.
• Valid throughout to boundary layer, subject to fine grid resolution. • Accounts for the stream-wise pressure gradients. • Applicable for detached, separated flows and fully turbulent flows.
• Pressure induced separation is typically predicted to be excessive and early.
• The most accurate from twoequation eddy viscosity models. • Suitable for adverse pressure gradients and pressure-induced flow separation. • Accounts for the transport of the principal shear stresses.
• Less suitable for free shear flows.
TE D
Shear Stress Transport (SST) E − H
Disadvantages
RI PT
Realizable E − F
Comments
EP
Model
AC C
407
409 410 24
ACCEPTED MANUSCRIPT
Table 1. (Continued). Comments
Advantages
Reynolds Stress Model (RSM)
• The most general and complex of all models solving transport of the Reynolds stresses- so called seven-equation model • Isotropic eddy viscosity hypothesis is discarded.
• Accurate calculation of the mean flow properties and all Reynolds stresses. • Accounts for the streamline curvature, rotation and rapid changes in strain rate yielding superior results to two-equation models for complex flows, e.g. with stagnation points.
412
SC
AC C
EP
TE D
413 414
Disadvantages
RI PT
Model
M AN U
411
25
• Computationally expensive. • Not always more accurate than two-equation models. • Harder to obtain converged result. • Reliability of RSM predictions are still limited by the closure assumptions employed to model pressure-strain and dissipationrate terms.
ACCEPTED MANUSCRIPT 4.1.1. Alternative approaches - LES and DNS
416
Large Eddy Simulation (LES) is an alternative approach in
417
hydrodynamic modelling where large, three-dimensional
418
unsteady scale motions affected by the flow geometry, i.e. large
419
eddies, are directly and explicitly solved in time-dependent
420
simulation using space-filtered Navier-Stokes equations. LES is
421
one of the most expensive simulation options and requires
422
a refined grid to accurately resolve eddies in the boundary
423
layer. A filtering operation, analogous to the Reynolds
424
decomposition in RANS, is based on the decomposition of the
SC
M AN U
425
RI PT
415
velocity into the resolved (filtered) component ̅ J , K and the residual, so called subgrid-scale (SGS) component J , K
427
(Pope 2000). The accuracy of the LES model is the result of
428
modelling only the SGS motions- the smallest eddies, which
429
tend to have more universal properties (Karpinska Portela
430
2013). The model commonly used to model small eddies is the
431
Smagorinsky SGS model (Smagorinsky 1963).
432
LES has been most successful for high-end applications where
AC C
EP
TE D
426
433
the RANS models fail to meet the required goals, e.g.
434
modelling of combustion and mixing. Although it provides
435
improved accuracy, wide application of LES approach to solve
436
flow related issues is still limited due to the large mesh
437
requirements and high computational costs.
438
Direct Numerical Simulation (DNS) is a high-fidelity tool
439
offering the explicit solution of the whole range of turbulent 26
ACCEPTED MANUSCRIPT time and length scales, from the Kolmogorov scales to large
441
motion scales transporting most of the kinetic energy within the
442
domain (Orszag 1970). As a result, the computational cost of
443
DNS is extremely high even at low Reynolds numbers,
444
preventing this approach from being used as a general-purpose
445
design tool, and making it impractical for most industrial flow
446
conditions, especially large multiphase AS systems. Hence the
447
CFD modelling of the AS tanks requires more computationally
448
efficient and thus simplified methods.
449
4.2. Multiphase Modelling
450
A robust understanding of the physics and biochemical
451
processes in the gas-liquid-solid environment of the AS tank
452
relies on accurate assessment of the transport phenomena and
453
character of interactions between the phases. Numerical
454
simulation of the multiphase flow in AS system is usually
455
enabled by Euler-Euler or Euler-Lagrange approaches. Table 2
456
outlines the relevant characteristics of the multiphase models
457
used in modelling of the AS systems.
AC C
EP
TE D
M AN U
SC
RI PT
440
458
In Eulerian (Eulerian-Eulerian or two-fluid) model, the phases
459
are treated mathematically as separate and interpenetrating
460
continua, hence the introduction of the phasic volume fractions-
461
" . The sum of volume fractions is equal to unity
27
ACCEPTED MANUSCRIPT M
L " = 1
(8)
NO
Each phase is governed by a set of continuity and momentum
463
conservation equation, which have similar structure for all
464
phases.
Thus,
Q-phase
considering
system,
RI PT
462
the
mass
conservation equation for phase is (Joshi 2001, Ratkovich
466
2010): M
SC
465
M AN U
. 1" ' 2 + ∇1" ' S 2 − L1 − 2 = 0 . NO
467 468
(9)
where ' is density and the term " ' is the effective density of the phase , S denotes its velocity, and are mass
transfer mechanisms from phase to and from to ,
470
respectively.
471
The generalized momentum conservation equation for phase
472
can be written in the simplified form as (Azzopardi et al. 2011,
AC C
EP
TE D
469
473
Ratkovich 2010): . 1" ' S 2 + ∇1" ' S S 2 .
= −" ∆ + ∇% 1∇" S + ∇" SU 2 + ' VS + S
28
(10)
ACCEPTED MANUSCRIPT
474
where is pressure shared by all phases, % denotes shear viscosity of phase , S represents the sum of interfacial forces
476
between the continuous and dispersed phases.
477
Considering gas-liquid system, the fluid flow around the
478
bubbles is characterized by the occurrence of relative motion
479
between the phases yielding local pressure and shear stress
480
gradients. As a result, relative motion of the bubbles will be
481
affected by a drag force which is predominant in conditions of
482
the uniform flow. In case of non-uniform bubble motion, the
SC
RI PT
475
concept of interfacial forces S needs to account for drag and
484
various non-drag forces, such as virtual mass force, lateral lift
485
force, wall lubrication force, turbulent dispersion force, Basset
486
force and momentum transfer associated with mass transfer.
487
Accordingly, the closure of Equation (11) requires correct
488
assessment of the interfacial forces, typically using analytical
489
models, empirical correlations and coefficients (e.g. drag
490
coefficient), described comprehensively in Clift et al. (1978),
491
Joshi (2001), Azzopardi et al. (2011) and Yang and Mao
492
(2014). Many of these relations are case-specific and based on
493
limited data what yields difficulties in exploring complex
494
multiphase systems (Azzopardi et al. 2011, Manninen et al.
495
1996) and influencing the outcomes of the CFD simulations.
496
The Eulerian model is commonly used in multiphase systems,
497
where momentum exchange between the phases is significant,
AC C
EP
TE D
M AN U
483
29
ACCEPTED MANUSCRIPT e.g. gas-liquid flow in aeration tanks or solid-liquid flow in AS
499
systems, especially in the case of zones with high solids
500
concentrations, where solid-solid interactions are relevant;
501
transitional zones with steep solids concentrations gradients
502
(from high to low concentration), where momentum of the solid
503
phase is relevant to model its dissipation (Samstag et al. 2015).
504
The Volume of Fluid (VOF) model is a single-fluid approach
505
based on a surface tracking technique that solves the
506
momentum equations for the continuous phase while the
507
dispersed phase follows the closure conditions of the volume
508
fraction for the incompressible flow (Wang et al. 2013). While
509
the phases are immiscible, the fields for all variables and
510
properties are shared and represented by volume-average
511
values. Therefore, the continuity equation has the following
512
form (Ratkovich 2010, Vedantam et al. 2006):
TE D
M AN U
SC
RI PT
498
EP
." + ∇1" S 2 = 0 .
(11)
where the computation of the primary-phase volume fraction is
514
based on the constraint defined by Equation (9).
515
The single momentum equation, which is solved throughout the
516
domain yielding velocity field shared by the phases, is
517
expressed as:
AC C
513
30
ACCEPTED MANUSCRIPT . J'SK + ∇J'SSK = −∆ + ∇W%J∇S + ∇S U KX + 'VS + S (12) .
518
The properties ' and % appearing in the transport equations are
519
determined by the presence of the component phases in each
control volume. Considering two-phase ( and ) system if the
RI PT
520
volume fraction of phase is being tracked, the density in each
522
cell is given by (Ratkovich 2010, Vedantam et al. 2006):
523
M AN U
' = " ' + 11 − " 2'
SC
521
(13)
The relationship described by Equation (13) is based on the
524
fact, that for an Q-phase system, the volume-fraction-averaged
525
density is
(14)
TE D
' = L " '
All other properties (e.g volume-fraction-averaged viscosity, %)
527
are computed in the same manner.
528
VOF is designed for modelling multiphase systems where the
AC C
EP
526
529
position of the interface between the immiscible fluids is of
530
interest. This approach has found application in modelling of
531
stratified and free-surface flows, filling, sloshing, and motion
532
of large bubbles (slug flow). Examples of VOF application to
533
model Membrane Bioreactors (MBRs) can be found in the
534
literature (Andersson et al. 2011, Ratkovich et al. 2009, Wang
535
et al. 2013). 31
ACCEPTED MANUSCRIPT The Eulerian-algebraic (slip mixture or algebraic slip) model is
537
a simplified multiphase model used for two or more phases,
538
treated as interpenetrating continua. This single-fluid approach
539
can be used to model phases moving at different velocities- by
540
using concept of slip (or drift) velocities.
541
Here, the continuity and momentum equations are solved for
542
the mixture and algebraic equations are used to solve relative
543
velocities to describe the dispersed phases. Thus, the continuity
544
equation for the mixture is (Manninen et al. 1996):
M AN U
SC
RI PT
536
.' + ∇J' S K = 0 .
545
where the mixture density, ' is defined as: M
TE D
' = L " ' NO
AC C 547
(16)
and the mass-averaged mixture velocity S is:
EP
546
(15)
M
1 S = L " ' S ' NO
(17)
The mass fraction of phase is defined as: =
" ' '
(18)
548
The momentum equation for the mixture is obtained from the
549
following formula (Ratkovich 2010):
32
ACCEPTED MANUSCRIPT . J' S K + ∇J' S S K .
U KX = −∆ + ∇W% J∇S + ∇S + ' VS + S
NO
550
where % is mixture viscosity equal to: M
% = L " %
M AN U
NO
SC
+ ∇ YL " ' S, S, Z
(19)
RI PT
M
(20)
551
and S, is drift velocity for the secondary phase expressed
552
as follows:
(21)
TE D
S, = S − S
The slip velocity of the secondary phase (K relative to the
554
velocity of the primary phase (K is computed as follows:
AC C
EP
553
555
S = S − S
(22)
The relation between drift and slip velocities can be written as: M
S, = S − L S NO
(23)
556
The basic assumption of the algebraic slip mixture model is that
557
a local equilibrium between the phases should be reached over
558
a short spatial length scale. For the phases moving with the 33
ACCEPTED MANUSCRIPT same velocity, the mixture approach can be used to model
560
homogeneous multiphase flow (Manninen et al. 1996, Wicklein
561
and Samstag 2009). However, this approach is usually not
562
recommended in cases when the interface laws are not well
563
known (Ratkovich 2010). This model has been successfully
564
used to simulate sediment-induced density currents, and thus
565
solid-liquid mixing in AS tanks, and turbulent transport of
566
suspended solids in secondary clarifiers (Samstag et al. 2015,
567
Wicklein and Samstag 2009, Yeoh and Tu 2009), but also a
568
gas-liquid turbulent flow in closed loop bioreactors (Xu et al.
569
2010, Yang et al. 2011).
570
Another method to deal with multiphase flow in AS systems is
571
via the modelling of turbulent transport of the secondary phase
572
within a continuous primary phase. The transported phase
573
(species) is treated as an active or passive scalar (density,
574
viscosity, temperature, dissolved oxygen or solids), hence the
575
effects of its gradients across the domain can be either coupled
576
to the momentum equation as an equation of state or treated as
AC C
EP
TE D
M AN U
SC
RI PT
559
577
a passive property transported by the fluid flow.
578
Therefore prediction of local mass fraction of each species, ! ,
579
is modelled by solving conservation equation describing
580
convection and diffusion of the ′-th species (Cartland Glover et
581
al. 2000):
34
ACCEPTED MANUSCRIPT . J'! K + ∇J'S! K = −∇S + + .
582 583
(24)
where S is the ′-th diffusive mass flux of species; is net
rate of production of ′-th specie due to chemical reaction; is
the source term which injects ′-th specie into the domain by
585
addition from the dispersed phase plus any user-defined
586
sources.
587
The active/passive scalar approach is used to model solid-liquid
588
interactions in the systems with lower solids concentrations
589
(Combest et al. 2011, De Clercq 2003, Samstag et al. 2015,
590
Wicklein and Samstag 2009) and to model oxygen mass
591
transfer in aerated tanks (Fayolle et al. 2007).
592
In the Lagrangian model, the governing phase (fluid) is treated
593
as a continuum by solving time-averaged Navier-Stokes
594
equations (Eulerian reference frame), while the behaviour of
595
secondary phase (e.g. solids) is predicted by tracking a large
596
number of particles using random-walk Lagrangian trajectory
AC C
EP
TE D
M AN U
SC
RI PT
584
597
calculations for dispersed phase through the flow field of the
598
continuous phase. Here, the particles can exchange momentum,
599
mass, and energy with the fluid phase. The particles’
600
trajectories are predicted by integrating the force balance on the
601
single particle and recording the particle position. This force
602
balance, which equates the particle inertia with the forces
603
acting on it is defined as (Bridgeman et al. 2009): 35
ACCEPTED MANUSCRIPT .S S − S VSJ' − 'K = + + S . '
(25)
604
where S is particle velocity and S is fluid phase velocity, the
605
term
606
particle relaxation time, ' and ' are particle and fluid
_`
represents drag force per unit particle mass, is
RI PT
\S][ \S^ [
densities, respectively; and S relates to additional forces that
608
are relevant under special circumstances; for example virtual
609
mass and pressure gradient forces, or forces on particles that
610
arise due to rotation of the reference frame.
611
Additionally, particle-particle interactions, such as collisions
612
and momentum exchange can be enabled through the
613
introduction of drag coefficients (Bridgeman et al. 2009,
614
Karpinska Portela 2013). This approach is suitable to simulate
615
gas-liquid flow in aeration tanks, however should not be used
616
when the volume fraction of the secondary (particulate) phase
617
exceeds 10-12% (Sokolichin et al. 1997), as e.g. solid-liquid
618
flow in complete mixed AS systems.
619
Apart from multiphase modelling, particle tracking within the
620
Lagrangian reference frame is a useful tool to simulate RTDs in
621
process tanks (Danckwerts 1953, Levenspiel 1999), and so to
622
assess the macromixing, established by convective flow
623
patterns (Karpinska Portela 2013, Le Moullec et al. 2008).
624
However, the main limitation here is the number of particles
625
being tracked within the simulated system, and in the case of a
AC C
EP
TE D
M AN U
SC
607
36
ACCEPTED MANUSCRIPT 626
full-scale AS tank, tracking too many particles will result in
627
extensive computational
628
workability of different approaches in multiphase modelling,
629
the Lagrangian approach is the most expensive, involving long
630
computational times and requiring a large number of CPUs
631
(Central Processing Units), so limiting its popularity in the
632
simulation of wastewater treatment tanks.
when
comparing
SC
633
Thus,
RI PT
times.
AC C
EP
TE D
M AN U
634
37
ACCEPTED MANUSCRIPT Table 2. Summary of approaches to multiphase modelling. Concept
Modelling
Applicability
Issues
Eulerian
Each phase modeled as a separate fluid.
A set of averaged, volume fractionweighted NavierStokes (NS) equations per phase. Momentum transfer terms and constitutive equations to be modelled.
Theoretically, every type of flow; depending on the additional terms’ modelization.
Additional terms’ modelization is determinant, but their modelling is difficult.
VOF
Each phase modeled as a separate fluid. The interface between the phases is tracked.
Interface tracked via a continuity equation and the domains of the single phases are defined. A set of phase-specific NS equations with momentum exchange terms is solved for each domain.
Flows where the interphase surface is clearly defined (e.g. a single, large bubble inside a liquid).
Inapplicable if the interphase surface is too complex (e.g. bubbly flow where bubble dimensions are smaller than single cell size).
Mixture
Both phases treated as a whole.
Single set of NS equations.
Homogeneous fluids or nonhomogeneous fluids that are treated as homogeneous.
Inapplicable in every case in which there is clear distinction between the phases.
Flows where at least one phase is clearly dispersed into a principal phase. Particles smaller than mesh cell size.
Computational expense not a priori computable, and proportional to the particles number. May be prohibitive for high particle numbers.
SC
M AN U
TE D
Liquid phase treated with Eulerian approach. Every particle (bubble) tracked along trajectory.
AC C
EulerianLagrangian
RI PT
Model
Effective mixture density and viscosity to be modelled.
EP
635
A set of averaged NS equations for the liquid phase. A set of Newton’s 2nd Law (N2L) equations applied to all particles. Momentum transfer terms in both NS and N2L to be modeled.
636
38
ACCEPTED MANUSCRIPT 637
5. Application of CFD in ASP
638
Biological wastewater treatment systems require enhanced
639
transfer of oxygen into the AS tanks to maintain aerobic
640
processes
641
Enhanced oxygen mass transfer in most common ASP
642
configurations i.e. channel or closed-loop bubbly flow AS
643
reactors aerated by diffusers, can be achieved by maximisation
644
of the surface area of the interface between the dispersed phase
645
(air/oxygen bubbles) and the continuous phase (mixed liquor).
646
Nonetheless, in most wastewater utilities, designs of AS tank
647
and aeration system, and furthermore process control are based
648
on static models and simple PID controllers, and so a major
649
concern
650
hydrodynamics and its impact on DO profiles within the basin.
651
Therefore,
652
efficiencies, hydraulic and process design, and reliable
653
operation of AS systems rely on an improved understanding of
654
the bioreactors’ behaviour, with emphasis on micro- and
in
biodegrading-nitrifying
biomass.
the
accurate
TE D
remains
M AN U
SC
RI PT
occurring
of
biochemical
of
tank
conversion
AC C
EP
improvement
determination
655
macro-scale mixing and, in particular, on the analysis of the
656
interactions between the mixed liquor circulation imparted by
657
mechanical agitation and the fluid and sludge flocs motions
658
induced by the air bubbles. The key advantage of CFD as a
659
virtual modelling technique is powerful visualization capability
660
allowing detailed characterization of the local-scale phenomena
661
in varying operating conditions. Hence, CFD can be used as a 39
ACCEPTED MANUSCRIPT robust tool for the design of either new efficient and energy
663
optimised unit processes at WWTPs or for “tune for benefit”
664
optimization of performance and even retrofitting of existing
665
AS systems.
666
The complete CFD simulation of a biologically active gas-
667
liquid-solid AS system is challenging, due to the complex
668
hydrodynamics and biochemical reactions of conversions
669
involved, resulting in massive computational resources required
670
in terms of RAM and CPU usage and with long computational
671
times involved. Furthermore, increasing complexity of the
672
model involved, mesh resolution and solution accuracy may
673
lead distinctly higher costs of the CFD analysis, yet lower than
674
a capital investment in a new layout. To avoid overly long and
675
complex CFD runs (Karpinska Portela 2013, Le Moullec et al.
676
2010b), common engineering practice has been to model the
677
AS processes separately (e.g. aeration system performance or
678
flow field within the basin) and afterwards to couple the results
679
(Pereira et al. 2012). In addition to that, depending on the
AC C
EP
TE D
M AN U
SC
RI PT
662
680
purpose of the CFD analysis, data collection for model
681
calibration and validation using advanced measurement
682
techniques, may be absolutely required and yet time and
683
resource consuming (Nopens et al. 2012).
684
It is worth mentioning, that despite the current successful
685
spread and usage of CFD modelling, potential risk of its misuse
40
ACCEPTED MANUSCRIPT due to poor model choice, wrong setup assumptions or
687
interpretation of the results, has been also recognized (Nopens
688
et al. 2012).
689
5.1. Development of New Aeration Devices
690
CFD simulations have been used as a high-tech design tool in
691
the development of new aeration devices and optimisation of
692
their spatial arrangement in wastewater treatment tanks.
693
A computationally inexpensive modelling procedure, based on
694
3D steady-state and single-phase flow simulations with the
SC
RI PT
686
> − $ turbulence model was used by Morchain et al. (2000) to
696
study the impact of the spatial distribution of the cross-flow
697
hydro-ejectors on the recirculation and the oxygen mass
698
transfer in a large tank for wastewater treatment. Although the
699
cross-flow hydro-ejectors generate two-phase flow, it was
700
shown experimentally that the momentum transfer is not
701
significantly affected by the presence of bubbles and thus the
702
velocity field could be obtained using a single-phase model,
703
while the oxygen transfer in the tank was enabled by the
AC C
EP
TE D
M AN U
695
704
introduction of transport species.
705
The same single-flow modelling scheme was applied to modify
706
geometry and operating scenarios of a curved blade mechanical
707
aerator intended for use in oxidation ditches (Bhuyar et al.
708
2009). The results, which were in good agreement with the
709
experimental data, allowed optimization of blade design and
710
the aerators’ submergence, and reductions to rotational speed 41
ACCEPTED MANUSCRIPT 711
range, yielding enhanced aeration efficiencies when compared
712
with conventional mechanical devices.
713
Despite higher computational costs, transient gas-liquid simulations involving URANS with the > − $ turbulence
715
model and one of the Euler-Euler multiphase models have
716
found applications in the development of new aeration
717
techniques, as they provide direct and more accurate analysis of
718
the multiphase reactor systems. Xu et al. (2010) simulated an
719
oxidation ditch equipped with cylindrical airlift aerators. Here
720
the CFD simulations of the fluid flow in an airlift oxidation
721
ditch served to verify the feasibility of the preliminary design
722
and to assess its applicability for municipal wastewater
723
treatment. The results from the simulations (which were
724
validated in a bench-scale ditch) emphasized the suitability of
725
the
726
configurations.
727
Karpinska Portela (2013) used CFD to focus on overcoming
728
oxygen transfer efficiency limitations of membrane diffusers
729
via the introduction of an independent external aeration unit,
730
designed as a continuous flow component included in the
731
mixed liquor recirculation loop. Similar to the previous work,
732
CFD
733
configurations provided an excellent design tool for selection of
734
the most efficient geometry, characterized by the highest value
TE D
M AN U
SC
RI PT
714
aeration
system
for
deep
tank
ditch
AC C
EP
proposed
simulations
of
aeration
42
within
several
device
ACCEPTED MANUSCRIPT of SOTE. The aeration process parameters determined from
736
pilot-scale aeration tests matched the predictions obtained from
737
the CFD simulations.
738
A more detailed description of the CFD approaches used in the
739
development of aeration devices is shown in Table 3.
RI PT
735
740
AC C
EP
TE D
M AN U
SC
741
43
ACCEPTED MANUSCRIPT Table 3. CFD studies of new aeration devices. Author
Code
Device
Scale
Dim.
Turbulence Extra model model
Bhuyar et al. (2009)
Fluent
Curved blade mechanical aerator
Lab
3D
RANS + > − $
Xu et al. (2010)
Fluent
Airlift aerator
Lab
3D
URANS + > − $
Karpinska Portela (2013)
Fluent
Pressurized Lab Aeration Chamber
3D
URANS + > − $
743 744 745 746
750 751 752 753 754 755
EP
749
AC C
748
TE D
747
44
Validation measurements
RI PT
Transport species + chemical reaction Mixture
SC
M AN U
742
Mixture
Particle Dynamic Analysis (PDA)
measurements
ACCEPTED MANUSCRIPT 756 757
5.2. Evaluation of the performance of the existing aeration systems Robust and energy efficient operation of complex AS reactor
759
systems relies on an understanding of the multiphase nature of
760
the AS tank and on the assessment of the impact of the aeration
761
system on the mixing, biological treatment efficiency and
762
associated energy expenditure, while taking into account the
763
number, type and spatial distribution of the aeration and mixing
764
devices. This section considers CFD approaches used to
765
determine the dynamic behaviour of aeration tanks, and for
766
which descriptions are summarized in Table 4.
767
5.2.1. Gas-liquid Models
768
When considering fluid flow within an aeration tank, the use of
769
gas-liquid
770
straightforward prediction of the multiphase velocity field
771
induced by the aerators and mixers, bubble sizes and local gas
772
hold-up, i.e. process parameters that have significant impact on
773
the oxygen mass transfer in bioreactors, but either require time-
774
and resource-expensive experimental techniques or are not
775
taken into account in the analysis of the experimental results
776
(Fayolle et al. 2007, Gillot et al. 2005, Gillot and Héduit 2000,
777
Le Moullec et al. 2008, Vermande et al. 2007).
778
Hence the focus of early CFD works was on a better
779
understanding of mass transfer phenomena in full scale closed-
TE D
M AN U
SC
RI PT
758
models
enables
relatively
fast
and
AC C
EP
CFD
45
ACCEPTED MANUSCRIPT loop aeration tanks equipped with membrane diffusers and
781
agitated by means of horizontal slow speed impellers (Cockx et
782
al. 2001, Do-Quang et al. 1999). 2D two-phase flow in the
783
aeration tank was simulated using an Eulerian model for
784
imposed, constant gas bubble size. To simplify numerical
785
simulations, an equivalent uniform liquid velocity field was
786
imposed in the section of the impellers. The oxygen mass
787
transfer was obtained via introduction of the global volumetric
RI PT
780
mass transfer coefficient, , determined experimentally, as a
789
transport source term. Here, the correct estimation of the
790
absolute values of oxygen transfer requires assessment of the
791
most uncertain aeration process parameter, i.e. the alpha-factor,
792
quantifying the impact of the contaminants and process
793
conditions on the oxygen transfer rates- in this work- the
794
impact of the imposed constant bubble diameter on the gas-
795
liquid interface surface area (Nopens et al. 2015, Rosso et al.
796
2011, Rosso and Stenstrom 2006).
797
The
EP
TE D
M AN U
SC
788
from
the
simulations
showed
that
the
AC C
results
798
hydrodynamics of these tanks was controlled by mutual
799
competition between vertical gas plume and horizontal fluid
800
flow currents. Thus, in the absence of the horizontal flow
801
motion, the interactions between gas plumes released by
802
diffusers generate vertical, massive liquid loop circulations, i.e.
803
spiral flow, yielding low gas hold-up. Contrary to that, under
804
conditions of horizontal liquid motion imparted by e.g. rotating
46
ACCEPTED MANUSCRIPT impellers,
the
increase
in
wastewater
velocity causes
806
neutralization of the spiral flow patterns, inclination of the
807
vertical gas plume, and dispersion of the bubbles and, as a
808
consequence, longer contact times between the phases yielding
809
better gas retention in the tank. As a result, the increase of the
810
global mass transfer coefficient in the closed-loop tanks was
811
found to be associated exclusively with fluid circulation.
812
The impact of the fluid velocity on oxygen mass transfer was
813
later confirmed in experimental studies (Abusam et al. 2002). It
814
was reported that even small changes in the axial velocity may
815
have a dramatic effect on the oxygen profiles, and thus on the
816
ammonia removal in oxidation ditches.
817
closed-loop AS systems, horizontal fluid velocity should be
818
treated as an important process variable to control total nitrogen
819
removal efficiency.
820
A clear shortcoming of the discussed work (Cockx et al. 2001,
821
Do-Quang et al. 1999) is that the 2D modelling is insufficient
822
to represent mutual interactions between the neighbouring gas
M AN U
SC
RI PT
805
AC C
EP
TE D
Therefore in these
823
plumes causing formation of the more complex, three-
824
dimensional flow structures. Moreover, the impact of the flow
825
circulation on the size of the bubbles released by membrane
826
diffusers was neglected; yet this plays a key role in the oxygen
827
mass transfer within the aeration tank. Nevertheless, despite
828
these weaknesses and the lack of experimental validation,
829
conclusions from this work provided valuable research
47
ACCEPTED MANUSCRIPT background for many successive numerical and experimental
831
studies.
832
Subsequent studies (Fayolle et al. 2007) addressed development
833
of a more complete numerical tool, to predict oxygen mass
834
transfer in a number of pilot- and full-scale oxidation ditches of
835
various configurations and where aeration is dissociated from
836
mixing by the introduction of membrane diffusers and slow
837
speed mixers. The aeration was simulated using Eulerian model
838
for water and air. Mixing was modelled using fixed value
839
method by imposing flow characteristics induced by agitation
840
on the grid zones corresponding to mixers location. This
841
method gives good prediction of the average experimental axial
842
velocity. Similar to earlier CFD works (Cockx et al. 2001, Do-
843
Quang et al. 1999), Fayolle et al. (2007) showed that oxygen
844
mass transfer in closed-loop AS basins aerated with diffusers is
845
linked to the number, type, spatial distribution and performance
846
of the agitators (Fayolle et al. 2007, Fayolle et al. 2010, Fayolle
847
et al. 2006, Vermande et al. 2007),
The impact of the
AC C
EP
TE D
M AN U
SC
RI PT
830
848
ascending bubbles released by the diffusers on the velocity
849
profile along the oxidation ditch can be seen in Figure 1. It was
850
also showed that implementation of more complex modelling
851
approach accounting for the impact of the bubble size on the
852
axial fluid velocities and on the global gas hold-up profile
853
within the tank, was able to reproduce precisely the values of
854
with high accuracy (≤ 5%). Nonetheless, it was found, that 48
ACCEPTED MANUSCRIPT the predicted accuracy depends on the assumed value of inlet
856
bubble size, emphasizing the necessity either to measure in situ
857
bubble diameter or apply an appropriate numerical model
858
estimating bubble sizes at the diffuser level. A shortcoming of
859
Fayolle et al. (2007) is that although the proposed CFD
860
protocol appears to be suitable for the prediction and
861
optimization of oxygenation capacities in a full scale AS tank,
862
it is based on clean water-air simulations. However, for robust
863
analysis of agitation system performance, the impact of the
864
velocity on the local solids content and density gradients should
865
be considered.
866
Yang et al. (2011) focused on predicting the flow pattern and
867
oxygen mass transfer in a multichannel oxidation ditch aerated
868
with horizontal rotors and agitated with submerged mixers (a
869
Carrousel). Experimental field data showed that under existing
870
operational conditions, DO concentration in the anoxic zone
871
exceeded 1.0 mg/L, inhibiting the denitrification process. Thus,
872
the aim of the CFD study was to determine an operational
AC C
EP
TE D
M AN U
SC
RI PT
855
873
regime for the surface aerators, which would lead to formation
874
of stronger DO gradients in the anoxic zone, yielding a
875
reduction in energy expenditure for aeration while maintaining
876
high treatment efficiencies. The modelling procedure involved
877
a two-phase mixture approach, enhanced with additional
878
transport equations related to the transfer of oxygen introduced
879
by rotors and biochemical reaction of oxygen consumption for 49
ACCEPTED MANUSCRIPT BOD removal. To reduce computational expenses related to the
881
modelling of a number of rotating devices, the large surface
882
aerators were simulated using moving wall model, while
883
actuator disk (fan) approach was used to simulate submerged
884
mixers. Although an energy efficient operating scheme for the
885
ditches was developed, this study was based on average flow
886
and constant influent quality parameters, thus the validity of the
887
optimal operating conditions deduced from the steady-state is
888
limited.
889
More recent CFD studies on the aeration of conventional AS
890
tanks (Gresch et al. 2011) provided further and more detailed
891
analysis of the flow field induced by porous diffusers. The CFD
892
approach proposed in this work consisted of the gas-liquid
893
Eulerian multiphase model, where the physical properties of the
894
continuous phase were approximated to those of activated
895
sludge. Moreover, instead of solving additional transport
896
equations for oxygen transfer, hydrodynamic simulations were
897
enhanced with a biokinetic model for nitrification. The results
AC C
EP
TE D
M AN U
SC
RI PT
880
898
from CFD studies validated experimentally in a full-scale plug
899
flow AS tank complemented conclusions from earlier work
900
(Cockx et al. 2001, Do-Quang et al. 1999) and provided a
901
comprehensive description of the spiral flow generated by
902
changing diffuser layout, its impact on the velocity field, air
903
hold-up (Figure 2), and ammonia degradation. It was also
904
shown that, contrary to the closed-loop basins, in plug flow AS 50
ACCEPTED MANUSCRIPT lanes without flow boosters, the flow field determines the
906
aeration efficiency and the intensity of the longitudinal mixing,
907
which are significantly reduced by either occurrence of non-
908
aerated zones at the sidewalls or the rolling motion of the fluid
909
generated by changes in diffusers layout.
910
5.2.2. Density-coupled Models
911
In recent years, multiphase modelling of AS tanks based on
912
gas-liquid neutral density simulations has become common
913
practice for evaluation of both aeration and mixing, mainly
914
because it enables faster setup of the lab-scale validation,
915
usually involving use of tap water and air only. However, CFD
916
studies on a sequencing batch reactor equipped with
917
conventional jet aeration system (Samstag et al. 2012) showed
918
that use of neutral density may lead to over-prediction of the
919
degree of mixing. In this work, a density-coupled CFD model
920
incorporating solids settling and transport was calibrated to
921
field data and used to evaluate capacity of the jet aeration
922
system in keeping the solids suspended and to determine power
AC C
EP
TE D
M AN U
SC
RI PT
905
923
consumption for pumping considering two operating modes:
924
mixing with and without air. The authors found that in order to
925
generate complete information about activated sludge mixing,
926
complementary
927
modelling are required to predict local density gradients due to
928
the impact of the flow regime on solids transport. While this
929
work underlines the importance of the analysis of the aeration
CFD
studies
51
based
on
density-couple
ACCEPTED MANUSCRIPT systems with respect to the activation sludge mixing, the
931
approach is rarely used by the wastewater modelling
932
community for AS tanks, and only limited works can be found
933
in the literature (Jensen et al. 2006, Laursen 2007, Samstag et
934
al. , Samstag et al. 1992, Samstag et al. 2012, Wicklein and
935
Samstag 2009).
936
Recently Xie et al. (2014) considered prediction of the mixing
937
and suspended solids distribution in a full-scale Carrousel ditch
938
equipped with surface aerators and submerged impellers. The
939
simulation procedure involved a solid-liquid mixture model,
940
where the sludge settling was coupled through the slip velocity.
941
Similar to the earlier work by Yang et al. (2011), surface rotors
942
and submerged mixers were simulated using moving wall and
943
fan models. The resulting mixing patterns and solids profiles
944
throughout the ditch, shown in Figure 3, facilitated
945
identification of the stagnation regions affected by the sludge
946
settling, and resulted in optimization of the operation scenario.
947
Previously, Fan et al. (2010) simulated a single-channel
948
oxidation ditch aerated by means of inversed umbrella surface
949
aerators using a solid-liquid two-fluid Eulerian model. The
950
operating aerators were simulated using Multiple Reference
951
Frames (MRF) approach. This work also focused on a detailed
952
description of the mixing regime induced by the operating
953
aeration system, aiming to select the optimal range of aerator
AC C
EP
TE D
M AN U
SC
RI PT
930
52
ACCEPTED MANUSCRIPT speed ensuring the most uniform distribution of the solids
955
within the ditch, and thus prevents solids settling.
956
Thus, the literature shows that the use of density-coupling is
957
well founded to assess sludge mixing in agitated AS systems,
958
however its applicability to study bubbly bioreactors is still
959
uncertain.
960
5.2.3. Residence Time Distribution of AS tanks
961
Although rarely used to simulate the aeration process itself,
962
CFD modelling based on the Lagrangian multiphase approach
963
with particle tracking can be used as a tool for evaluation of the
964
aeration system performance through the assessment of the
965
macromixing, and thus the overall transport phenomena in a
966
bioreactor. The literature offers numerous examples of studies
967
on aeration and/or agitation systems responsible for different
968
flow patterns within the tank, and thus having an impact on the
969
mixing at micro- and macro-scale in AS systems. Considering
970
different advection paths of the fluid elements within an aerated
971
continuous flow channel or closed-loop bioreactors for
AC C
EP
TE D
M AN U
SC
RI PT
954
972
wastewater
treatment,
RTD
973
simulations can be used successfully to predict such adverse
974
phenomena as segregation of the flow resulting in short-
975
circuiting (channelling), recycling of the flow, or formation of
976
dead zones. RTD data can also be used to assess the mixing
977
time, which can be of crucial importance for oxygen transfer
978
and impact on chemical and biochemical reaction yield. Data 53
data
obtained
from
CFD
ACCEPTED MANUSCRIPT can also be used for troubleshooting of the reactor and
980
improved design of the future vessels (Brannock 2003,
981
Karpinska et al. 2015, Karpinska Portela 2013, Kjellstrand
982
2006).
983
Glover et al. (2006) considered oxidation ditches aerated by
984
bottom diffusers and agitated by slow-speed rotors and
985
determined reactor model structure by implementation of a
986
CFD-generated RTD curve in Fluent into WEST® software
987
environment based on the systemic approach. The fluid flow in
988
the ditch was obtained from the Eulerian gas-liquid model, and
M AN U
SC
RI PT
979
the turbulence simulated with > − $ model. It was found that
990
the RTDs of the ditch behaviour can be approximated by 90
991
CSTRs (plug flow) for non-aerated conditions and by 20
992
CSTRs (perfectly mixed) for aerated conditions (Glover et al.
993
2006). The work showed that closed-loop reactors with
994
different internal recycling rates, such as oxidation ditches, can
995
be modelled by one or series of CSTRs, the number of which
996
can be determined from CFD-RTD studies. In this way, a
AC C
EP
TE D
989
997
suitable reactor model structure for ASM studies can be
998
predicted.
999
Le Moullec et al. (2008) considered the assessment of RTDs in
1000
a cross-flow channel reactor aerated with porous tube, in which
1001
the hydrodynamics was modelled using a gas-liquid Eulerian
1002
model and the turbulence models used were > − $ and RSM. 54
ACCEPTED MANUSCRIPT It was shown that in the channel-type reactors for wastewater
1004
treatment, turbulence has a dominant role in axial dispersion
1005
(90%), while dispersion due to convection is negligible.
1006
Furthermore it was shown that only the RTDs obtained from
1007
the RSM model are in good agreement with the experimental
1008
data treated by curve fitting for a plug flow with axial
RI PT
1003
dispersion model, while the > − $ model underestimated the
1010
value of dispersion coefficient by around 50%, as seen in
1011
Figure 4. This particular case is the effect of the default
1012
constants in Lagrangian stochastic particle motion model,
1013
which are determined by different closure assumptions. Thus,
M AN U
SC
1009
both models, > − $ and RSM, will use different AB which are
1015
directly linked to the modelling of the turbulent dispersion,
1016
what emphasizes the need for careful attribution of values to
1017
parameters, through the proper model calibration.
1018
Macromixing assessment is a useful tool to describe actual
1019
reactor performance. CFD studies assessing the impact of
1020
turbulence model selection on the RTDs of oxidation ditches
AC C
EP
TE D
1014
1021
aerated with slot injectors (Karpinska Portela 2013, Pereira et
1022
al. 2012), considered the fluid flow simulated with RANS and
1023
URANS with the > − $ model and also LES with
1024
Smagorinsky’s SGS model. This work shows the limitations of
1025
some approaches in computation of the RTD, as the turbulence
1026
model involved to simulate hydrodynamics has a large impact
1027
on prediction of the tracer’s trajectories. The RTD simulations 55
ACCEPTED MANUSCRIPT based on the average flow field, RANS and URANS, led to
1029
overestimation of channelling effects. The flow dynamics
1030
underlies mixing at all scales, both macro- and micro-, and thus
1031
this should always be accounted for. Furthermore, is has been
1032
shown that CFD data for AS modelling must align with all
1033
dynamic components of the flow, thus LES simulations, which
1034
demand more computational resource, should be used for some
1035
specific purposes.
1036
5.2.4. Full model: CFD-ASM
1037
Few workers have focused on the development of a complete
1038
three-phase CFD model to simulate actual physical-chemical-
1039
biological processes within different AS system configurations
1040
(Glover et al. 2006, Le Moullec et al. 2010a, b, Le Moullec et
1041
al. 2011, Lei and Ni 2014). Here, the integration of
1042
hydrodynamics with biokinetics is accomplished by embedding
1043
an ASM model (e.g. ASM1) in a CFD model through
1044
introduction of the additional species of transport with source
1045
terms comprising bioreaction rates. The CFD-ASM simulations
AC C
EP
TE D
M AN U
SC
RI PT
1028
1046
permit the quantification of interactions and transport
1047
phenomena between the water-gas, water-sludge and gas-
1048
sludge phases occurring within the AS reactor and accounting
1049
for the flow field, oxygen mass transfer, growth, decay and
1050
metabolic activity of the AS biomass, and sludge settling.
1051
Therefore it is possible to predict simultaneously the system
1052
hydrodynamics and its impact on the reactions occurring in 56
ACCEPTED MANUSCRIPT nitrifying-denitrifying-biodegrading biomass and represented
1054
by the concentration profiles, in order to optimize aeration,
1055
mixing or system layout leading to efficient process operation
1056
with reduced exploitation costs.
1057
The first attempt at CFD-ASM1 coupling reported in the
1058
literature (Glover et al. 2006), based on the Eulerian gas-liquid
1059
approach, facilitated prediction of poor performance of the
1060
diffused aeration system in a full scale oxidation ditch system,
1061
giving low nitrifying capacity of the sludge. These studies
1062
demonstrated the robustness of the CFD-ASM1 approach;
1063
however it was emphasized, that the validity of this modelling
1064
procedure requires further work to study different AS systems
1065
and aeration/agitation scenarios.
1066
Le Moullec et al. (2010a) provided a robust discussion on the
1067
modelling issues concerning simulations of the hydrodynamics
1068
and biokinetics in a channel AS tank aerated by porous tube
1069
using Eulerian two-fluid model coupled to the ASM1. The
1070
authors discussed the trade-off between a model which can be
1071
implemented and run in the realistic time, and the extent of
1072
simplifications required to facilitate this. The authors assumed
1073
that the channel reactor was in a pseudo steady-state with
1074
constant biomass content; air bubbles with constant diameter;
1075
and the activated sludge flocs being perfectly soluble in the
1076
liquid phase. These simplifications gave rise to a series of
AC C
EP
TE D
M AN U
SC
RI PT
1053
57
ACCEPTED MANUSCRIPT 1077
errors in model output; specifically, incorrect estimation of the
1078
local
1079
overestimation of the ammonia values (wrong estimation of the
1080
autotrophic fraction of biomass in the system).
1081
The conclusions from this work emphasized the importance of
1082
the bubble size distribution for the oxygen mass transfer and for
1083
the molar diffusivity of oxygen in mixed liquor. Furthermore, a
1084
new concept for the approximation of AS floc properties
1085
(highly hydrated solid phase) to a liquid phase, was proposed.
1086
The authors also highlighted that, as the hydrodynamics-
1087
biokinetics coupling requires high number of CPUs and long
1088
computational times, a compromise between mesh resolution
1089
and solution accuracy has to be found.
1090
In more recent work on three-phase CFD-ASM1 simulations of
1091
a Carrousel oxidation ditch aerated and agitated by means of
1092
mechanical aerators and mixers, Lei and Ni (2014) treated the
1093
AS flocs as a pseudo-solid phase. Here, the solid-liquid-gas
1094
flow field was obtained with a three-phase mixture model. The
underestimation
of
the
nitrate
and
AC C
EP
TE D
M AN U
SC
RI PT
value;
1095
results from the simulations, (i.e. velocity, DO, organics and
1096
nutrients profiles) are in good agreement with the validation
1097
data. Crucially, this modelling approach allowed the prediction
1098
of physical kinematics of sludge settling. Concentration maps
1099
of the solids (near the bottom), DO, COD, ammonia and
1100
nitrates (in the mid-depth) in the horizontal cross-section
1101
through the oxidation ditch, are shown in Figure 5. 58
ACCEPTED MANUSCRIPT It can be concluded, that the CFD-ASM1 simulations provide
1103
more reliable results than those obtained from oversimplified
1104
ASM models, which fail to represent the flow dynamics within
1105
a system, particularly regarding DO profiles along an AS basin,
1106
which have an impact on the biological nutrient removal
1107
(Pereira et al. 2009). Thus, the CFD-ASM data can be
1108
considered more suitable for the design and scale-up of
1109
bioreactors. However, besides high computational costs,
1110
another emerging drawback of CFD-ASM modelling approach
1111
is its limited feasibility due to stringent convergence criteria
1112
and equilibrium solution (Nopens and Wicks 2012). As a result,
1113
the procedures of coupling hydrodynamics data obtained from
1114
CFD simulations with the ASM simulations have been also
1115
intensively studied (Glover et al. 2006, Karpinska Portela 2013,
1116
Le Moullec et al. 2010b, Pereira et al. 2012). Here the RTD
1117
curves, actual HRT values, corrected recycle ratio, local
1118
velocities and other hydrodynamics characteristics obtained
1119
from the CFD simulations of an analysed AS system can be
AC C
EP
TE D
M AN U
SC
RI PT
1102
1120
used to generate a suitable reactor model, i.e. in terms of
1121
number of CSTRs in series, recirculation rate and the flow
1122
pattern between each of the reactors, where the ASM can be
1123
implemented.
1124
Nowadays, an alternative modelling approach using a
1125
compartmental model, based on the description of the AS
1126
reactor system using a network of interconnected sections, i.e. 59
ACCEPTED MANUSCRIPT compartments, is emerging. The bi-directional flow rates
1128
between the compartments are computed from the flow field
1129
obtained in the CFD simulations and accounting for the local
1130
values of velocities and mixing due to turbulence. When
1131
comparing with CFD models, compartmental models are
1132
inexpensive in terms of RAM and CPU usage. Nevertheless, as
1133
they are derived from steady-state CFD simulations, the results
1134
must be experimentally validated. In addition, it is still
1135
necessary to develop a more detailed biokinetic model to apply
1136
this approach in modelling of the full-scale industrial
1137
bioreactors, including ASPs (Le Moullec et al. 2010b, Le
1138
Moullec et al. 2011, Nopens and Wicks 2012, Pereira et al.
1139
2012).
M AN U
SC
RI PT
1127
AC C
EP
TE D
1140
60
ACCEPTED MANUSCRIPT
Table 4. CFD modelling of aeration in Activated Sludge tanks. Reference
Code
Scale
Dim. Mesh size
Model
Multiphase model
Extra model
Validation
Numerical modelling of an oxidation ditch aerated with diffusers and agitated by impellers.
Cockx et al. (2001), Do-Quang et al. (1999)
Astrid
Full
2D
n/a
RANS + > − $
Gas-liquid Eulerian
Transport eq. for oxygen
-
CFD studies of the oxidation ditches- coupling of hydrodynamic with biokinetics modelling.
Glover et al. (2006)
Fluent Lab pilot, full
3D
n/a
RANS + > − $
Gas-liquid Eulerian
DO, COD, NH4, NO3 measurements
29k452k
URANS + > − $
Gas-liquid Eulerian
Lagrangian, transport eq. for oxygen + complete biokinetics (ASM1) Transport eq. for oxygen (each phase)
Prediction of the oxygen transfer in oxidation ditches.
Fayolle et al. (2007)
Fluent Pilot, full
3D
RTD of an aerated channel bioreactor.
Le Moullec et al. (2008)
Fluent Lab
3D
350k
Gas-liquid Eulerian
Lagrangian
LDV, tracer experiments
3D
50k
RANS + > − $/RSM
Simulation of the hydrodynamics and reactions in activated sludge channel reactor.
Le Moullec et al. (2010a), b), Le Moullec et al. (2011)
Fluent Lab
Gas-liquid Eulerian
Transport eq. for oxygen + biokinetics (ASM1)
LDV, bubble size measurements
SC
M AN U
TE D
EP
RI PT
Aim
AC C
1141
61
RANS + > − $
MDV, DO measuremets
ACCEPTED MANUSCRIPT
Reference
Code
Impact of the surface aeration on the solids distribution in the oxidation ditch.
Fan et al. (2010)
Fluent Lab
3D
70,3k
RANS + > − $
Effect of aeration patterns on the flow field in the conventional AS tank.
Gresch et al. (2011)
CFX
Full
3D
400k
URANS + SST
−(
Modification of the operation conditions in the oxidation ditch to enhance energy efficiency.
Yang et al. (2011)
Fluent Full
3D
Evaluation of jet aeration and mixing in sequence batch reactors.
Samstag et al. (2012)
Fluent Full
3D
RTD of an oxidation ditch aerated with hydrojets.
Karpinska Portela (2013); Karpinska et al. (2015)
Fluent Pilot
3D
Multiphase model
Extra model
Validation
Solid-liquid Eulerian
-
PDA
Gas-liquid Eulerian
Biokineticsconsumption of NH4
ADV, reactive tracer experiments
Gas-liquid mixture
Transport eq. for oxygen + biokineticsconsumption of BOD
MDV, DO measurements
URANS + > − $
Gas-liquid mixture
Density-coupled model- solids settling
MLSS measurements
RANS + > − $, URANS + > − $, LES + Smagorinsky SGS
-
Lagrangian
-
1,66M RANS + > − $
TE D
EP
Model
SC
Scale Dim. Mesh size
RI PT
Aim
M AN U
Table 4. Continuation.
600k
AC C
1142
62
ACCEPTED MANUSCRIPT
1143
Aim
Reference
Code
Model
Studies on flow field and sludge settling in Carrousel oxidation ditch.
Xie et al. (2014)
Fluent Full
3D
1,53M URANS + > − $
Fluent Pilot
3D
Multiphase model
Extra model
Validation
Solid-liquid mixture
Slip velocity- sludge settling
MDV, MLSS measurements
Gas-liquidsolid mixture
Transport eq. for oxygen + biokinetics- (ASM1)
ADV, DO, COD, MLSS, NH4, NO3 measurements
SC
Scale Dim. Mesh size
RI PT
Table 4. Continuation.
M AN U
1144
A complete model to predict Lei and Ni (2014) hydrodynamics, oxygen transfer and biokinetic reactions in an oxidation ditch.
162k
RANS + > − $
ADV - Acoustic Doppler Velocimetry, MDV - Mono-directional Velocimetry, PDA - Particle Dynamic Analysis, LDV - Laser Doppler
1146
Velocimetry.
AC C
EP
TE D
1145
63
ACCEPTED MANUSCRIPT 6. Unaddressed Issues in CFD Modelling of AS Tanks
1148
Despite the fact that work has taken place on the CFD
1149
modelling of AS systems for over 15 years now, there are a
1150
number of issues which remain unaddressed and which, if
1151
successfully overcome, would enhance model fidelity and
1152
robustness further.
1153
6.1. Secondary Settler
1154
In engineering practice, the behaviour of the AS tanks is
1155
inseparably linked to the performance of the secondary settler,
1156
as the efficient removal of BOD and nutrients in AS process
1157
depends on the Solid Retention Times (SRTs), concentration of
1158
Mixed Liquor Suspended Solids (MLSS), and the composition
1159
of biomass. The rationale behind the dynamic modelling of a
1160
coupled aeration tank- clarifier system is the prediction of
1161
interconnected flow and mass transport in unsteady flow
1162
loading conditions, and thus evaluation of the impact of
1163
dynamic changes in return sludge flow rates on the
1164
concentration patterns (Patziger et al. 2012).
AC C
EP
TE D
M AN U
SC
RI PT
1147
1165
Secondary settling is referred to in the literature as “the most
1166
sensitive and complicated process in activated sludge plants”
1167
(Ji et al. 1996). In fact, the complete CFD modelling of a
1168
clarifier is not a feasible task, since many physico-biochemical
1169
phenomena
1170
hydrodynamics, turbulence, flocculation, sludge rheology,
1171
settling characteristics, heat exchange and temperature, and
must
be
considered
64
simultaneously,
i.e.
ACCEPTED MANUSCRIPT finally, biokinetics (Plósz et al. 2012). The most critical issue in
1173
the modelling of clarifiers is the inherent unpredictability of
1174
activated sludge settleability and missing data regarding the
1175
particulate fraction. Additionally, even the most advanced
1176
models for clarifiers are still based on empirical equations
1177
describing sludge settling. The variability of settling behaviour
1178
which is not predicted by models affects the actual SRTs of AS
1179
system, yielding a potential source of error in wastewater
1180
treatment plant models (Plósz et al. 2011, Plósz et al. 2012).
1181
Consequently, there is always a risk that the simultaneous
1182
modelling of clarifiers may affect the results obtained for the
1183
AS tank. Nonetheless, a CFD simulation of a simplified 2D AS
1184
tank- clarifier system has been reported in the literature
1185
(Patziger et al. 2012), where the focus was on the solids
1186
transport between both units in conditions of variable inflow
1187
and wet weather, but neglecting biokinetics and oxygen mass
1188
transfer. The necessity for further model enhancement and
1189
validation was also emphasized.
AC C
EP
TE D
M AN U
SC
RI PT
1172
1190
When considering the complexity of CFD models predicting
1191
clarifier behaviour simultaneously with a complete solid-liquid-
1192
gas model of the AS tank, computing time, availability of CPU
1193
resources and predictive capability must be recognised as
1194
confounding issues. As a result, common modelling practice is
1195
to avoid the interference between the two unit processes, and
1196
assume constant separation efficiency for the clarifier
65
ACCEPTED MANUSCRIPT performance (Le Moullec et al. 2010a, b, Le Moullec et al.
1198
2011) predicted in agreement with the guideline (Copp 2001).
1199
Moreover, in engineering practice, the design of clarifiers is
1200
based on assumption of certain sludge properties, hence the
1201
CFD models of settlers only have been frequently used for
1202
optimization of design and retrofitting purposes (De Clercq
1203
2003, Stamou et al. 2009). However, it will be worthwhile to
1204
pursue CFD analysis of integrated aeration tank- secondary
1205
settler systems, when models capable of predicting the
1206
character of activated sludge flocculation are regularly
1207
incorporated into sludge settling models.
1208
6.2. Population Balance Model
1209
The AS system can be described as an ensemble of populations
1210
of individual entities (bubbles, flocs, biomass cells), having
1211
specific properties (size, density, viscosity, enzymatic activity).
1212
In such a system, two kinds of behaviour may be recognized:
1213
interactions of the individual entities with the environment (e.g.
1214
interfacial oxygen transfer, shear induced break-up), and
AC C
EP
TE D
M AN U
SC
RI PT
1197
1215
mutual interactions between the individual entities (e.g.
1216
coalescence, aggregation). The character of these interactions is
1217
a function of one or more properties of the entities, which vary
1218
within the population. Thus, it is more correct to refer to this
1219
variation as “distributed properties”, as they can be represented
1220
by a distribution instead of a scalar (Nopens et al. 2015).
1221
Nevertheless, in order to reduce overall complexity of the 66
ACCEPTED MANUSCRIPT model framework associated with aeration tanks, a common
1223
procedure is to assume non-distributed scalar properties
1224
implying that all individuals behave in exactly the same way
1225
(Nopens et al. 2015, Sobremisana et al. 2011). The classic
1226
example is the assumption of fixed bubble diameter (Fayolle et
1227
al. 2007, Glover et al. 2006, Gresch et al. 2011, Le Moullec et
1228
al. 2010a) and uniform floc size (Fan et al. 2010).
1229
When considering an AS tank, recent experimental studies have
1230
demonstrated the influence of floc size on the flocculation
1231
behaviour and thus settleability of the activated sludge (Nopens
1232
et al. 2015), as well as on the biokinetic reaction environment
1233
within the AS floc (Sobremisana et al. 2011). Therefore
1234
estimation of floc size distribution may provide further insight
1235
into modelling of the activated sludge mixing (Samstag et al.
1236
2012) and assessment of aeration tank settling, an important
1237
process parameter allowing the prediction of hydraulic capacity
1238
and treatment efficiency during high hydraulic loads associated
1239
with wet weather flows (Nielsen et al. 2000, Sharma et al.
AC C
EP
TE D
M AN U
SC
RI PT
1222
1240
2013).
1241
Depending on the bubble flow regime in an AS tank, the
1242
intensity
1243
deformation promotes a large diversity in the shapes and sizes
1244
of the bubbles (Karpinska Portela 2013, Shaikh and Al-Dahhan
1245
2007, Takács 2005). At the same time, the impact of the bubble
1246
of
collisions,
agglomerations,
breakups
and
size on , superficial gas velocity, and thus gas hold-up and 67
ACCEPTED MANUSCRIPT oxygen mass transfer in aeration tanks has been also recognized
1248
(Fayolle et al. 2006, Vermande et al. 2007). However, in some
1249
cases involving modelling of aeration in conventional AS
1250
tanks, the assumption of non-distributed scalar properties is still
1251
predominant, while in others, i.e. modelling of clarifiers or
1252
bubble columns, it may lead to predictions that diverge
1253
significantly from the real systems (Sobremisana et al. 2011).
1254
Consequently,
1255
bubble/particle size distributions require the use of population
1256
balance models (PBM) to describe variations in populations of
1257
entities. An assessment of several solution methods for PBM,
1258
namely the discrete class size method (Hounslow et al. 1988),
1259
the standard method of moments- SMM (Randolf and Larson
1260
1971) and the quadrature method of moments- QMOM
1261
(Marchisio et al. 2003), can be found in the literature
1262
(Bridgeman et al. 2009). It should be highlighted, that although
1263
development of PBMs is at advanced stage and the coupling of
1264
QMOM with hydrodynamics requires only a small number of
models
which
incorporate
AC C
EP
TE D
M AN U
SC
multiphase
RI PT
1247
1265
computationally inexpensive scalar equations to be tracked, its
1266
application in modelling of aeration systems with suspended
1267
solids is still scarce and limited to very few examples
1268
(Bridgeman et al. 2009, Nopens et al. 2015, Sobremisana et al.
1269
2011), associated almost exclusively with secondary settlers
1270
(Griborio and McCorquodale 2006, Nopens et al. 2005).
1271
However, PBM-CFD coupling has been successfully exploited
68
ACCEPTED MANUSCRIPT 1272
by the chemical engineering community to study bubble
1273
columns, airlift and stirred bioreactors (Dhanasekharan et al.
1274
2005, Morchain et al. 2014, Wang 2011).
1275 7. Conclusions
1277
In the last few years, as a result of increasing availability and
1278
accessibility of commercial and open-source software suites,
1279
the use of CFD has evolved into a robust and precise technique
1280
for design, optimization and control of the AS systems. The
1281
following key conclusions can be put forward from this review:
SC
•
M AN U
1282
RI PT
1276
The complete CFD simulation of the complex multiphase flow in AS tanks remains a challenge, due
1284
to the high CPU and RAM requirements and limited
1285
feasibility resulting from the imposed convergence
1286
criteria. Although there is still no unequivocal protocol
1287
on CFD methodology, the most computationally
turbulence model has been adopted as the standard for
AC C
1289
efficient scenario, RANS/URANS closed by > − $
EP
1288
TE D
1283
1290
1291
the modelling of AS tanks;
•
Different CFD models serve different applications.
1292
Examples from the literature have demonstrated the
1293
potential and robustness of single flow simulations in
1294
design and optimization of the AS systems equipped
1295
with mechanical and jet aeration systems;
69
ACCEPTED MANUSCRIPT 1296
•
The Euler-Euler approach has been extensively
1297
explored for the prediction and optimization of
1298
oxygenation capacities in AS tanks equipped with
1299
diffused aeration systems; •
Lagrangian approach with particle tracking has been
RI PT
1300
used to determine RTDs of the tanks- a tool for
1302
evaluation of the aeration and mixing performance and
1303
troubleshooting of the reactor design;
1304
•
SC
1301
The neutral density modelling of AS tanks became common practice used for design and evaluation of
1306
aeration and mixing systems, despite leading to over-
1307
prediction of the degree of mixing. Hence the necessity
1308
to include density-coupled modelling of aeration tanks
1309
to predict accurately local density gradients due to the
1310
impact of the flow regime on solids transport; A small number of works concerns a complete threephase CFD model coupled with ASM1 aiming to quantify the transport and mutual interactions between
AC C
1313
TE D
1312
•
EP
1311
M AN U
1305
1314
water-gas-sludge phases within the AS reactor. Despite
1315
providing more reliable results than those obtained
1316
from ASM, this approach relies on the trade-off
1317
between a model which can be implemented and run in
1318
a realistic timeframe, and the extent of simplifications
1319
and the solution accuracy, which may lead to a series of
1320
output errors;
70
ACCEPTED MANUSCRIPT 1321
•
Potential possibilities of coupling of the CFD data with
1322
ASM based codes have been explored in order to
1323
generate a suitable tank-in-series model, where the
1324
biokinetic model can be implemented; •
There are several areas in modelling practice, which
RI PT
1325
remain unaddressed and require further study, e.g.
1327
modeling of coupled aeration tank- clarifier system to
1328
predict interconnected flow and mass transport in
1329
unsteady flow conditions and CFD-PBM coupling to
1330
assess the impact of the AS floc/air bubble size on
1331
mixing, settling, gas hold-up and oxygen mass transfer
1332
in the aeration tank.
M AN U
SC
1326
TE D
1333 Acknowledgements
1335
The research work of Dr. Anna M. Karpinska Portela was
1336
funded by the College of Engineering and Physical Sciences,
1337
University of Birmingham, UK.
EP
1334
AC C
1338 1339
References
1340 1341 1342 1343
Abusam, A. and Keesman, K.J. (1999) Effect of number of CSTR’s on the modelling of oxidation ditches; steady state and dynamic analysis . Proceedings of 13th Forum for Applied Biotechnology, pp. 91-94, Sept. 22-23, Gent, Belgium.
1344 1345 1346 1347
Abusam, A., Keesman, K.J., Spanjers, H., Van Straten, G. and Meinema, K. (2002) Effect of oxidation ditch horizontal velocity on the nitrogen removal process. European Water Management Online European Water Association (EWA).
71
ACCEPTED MANUSCRIPT Åmand, L. and Carlsson, B. (2012) Optimal aeration control in a nitrifying activated sludge process. Water Research 46(7), 21012110.
1351 1352 1353 1354
Andersson, B., Andersson, R., Håkansson, L., Mortensen, M., Sudiyo, R. and Van Wachem, B. (2011) Computational fluid dynamics for engineers, Cambridge: Cambridge University Press, Cambridge.
1355 1356 1357
Ardern, E. and Lockett, W.T. (1914) Experiments on the oxidation of the sewage without the aid of filters. Journal of the Society of Chemical Industry 33(10), 523-539.
1358 1359 1360
ATV-DVWK (2000) Standard ATV-DVWK- A 131E Dimensioning of Single-Stage Activated Sludge Plants, German Association for Water, Wastewater, Waste, Hennef, Germany.
1361 1362 1363
Azzopardi, B.J., Mudde, R.F., Lo, S., Morvan, H., Yan, Y. and Zhao, D. (2011) Hydrodynamics of gas-liquid reactors: normal operation and upset conditions, John Wiley & Sons, Ltd, Chichester, UK.
1364 1365 1366 1367 1368
Bhuyar, L.B., Thakre, S.B. and Ingole, N.W. (2009) Design characteristics of curved blade aerator w.r.t. aeration efficiency and overall oxygen transfer coefficient and comparison with CFD modeling. International Journal of Engineering, Science and Technology 1(1), 1-15.
1369 1370 1371
Bosma, A.J. and Reitsma, B. (2007) Hydraulic design of activated sludge tanks with CFD. Proceedings of the COMSOL Users Conference, October 23-24, Grenoble, France.
1372 1373 1374
Brannock, M. (2003) Computational Fluid Dynamics tools for the design of mixed anoxic wastewater treatmnent vessels. PhD Thesis, University of Queensland, Brisbane, Australia.
1375 1376 1377 1378
Bridgeman, J., Jefferson, B. and Parsons, S.A. (2009) Computational Fluid Dynamics modelling of flocculation in water treatment: A review. Engineering Applications of Computational Fluid Mechanics 3(2), 220-241.
AC C
EP
TE D
M AN U
SC
RI PT
1348 1349 1350
1379 1380 1381 1382
Brouckaert, C.J. and Buckley, C.A. (1999) The Use of Computational Fluid Dynamics for improving the design and operation of water and wastewater treatment plants Water Science & Technology 40(4-5), 81-89.
1383 1384 1385
Cartland Glover, G.M., Generalis, S.C. and Thomas, N.H. (2000) CFD and bubble column reactors: simulation and experiment. Chemical papers 54(6A), 361-369.
1386 1387 1388 1389
Chachuat, B., Roche, N. and Latifi, M.A. (2005) Long-term optimal aeration strategies for small-size alternating activated sludge treatment plants. Chemical Engineering and Processing: Process Intensification 44(5), 591-604.
72
ACCEPTED MANUSCRIPT Clift, R., Grace, J.R. and Weber, M.E. (1978) Bubbles, drops, and particles, Academic Press Inc., London, UK.
1392 1393 1394 1395
Cockx, A., Do-Quang, Z., Audic, J.M., Liné, A. and Roustan, M. (2001) Global and local mass transfer coefficients in waste water treatment process by computational fluid dynamics. Chemical Engineering and Processing 40(2), 187-194.
1396 1397 1398 1399
Combest, D.P., Ramachandran, P.A. and Dudukovic, M.P. (2011) On the gradient diffusion hypothesis and passive scalar transport in turbulent flows. Industrial & Engineering Chemistry Research 50(15), 8817-8823.
1400 1401 1402
Copp, J.B. (2001) The COST Simulation Benchmark: Description and simulator manual. Office for Official Publications of the European Community, Luxembourg.
1403 1404 1405
Cristea, S., de Prada, C., Sarabia, D. and Gutiérrez, G. (2011) Aeration control of a wastewater treatment plant using hybrid NMPC. Computers & Chemical Engineering 35(4), 638-650.
1406 1407
Danckwerts, P.V. (1953) Continuous flow systems. Chemical Engineering Science 2(1), 1-13.
1408 1409 1410 1411
De Araújo, A.C.B., Gallani, S., Mulas, M. and Olsson, G. (2011) Systematic approach to the design of operation and control policies in activated sludge systems. Industrial and Engineering Chemistry Research 50(14), 8542-8557.
1412 1413 1414 1415
De Clercq, B. (2003) Computational Fluid Dynamics of settling tanks:Development of experiments and rheological, settling, and scraper submodels. PhD Thesis, University of Ghent, Ghent, Belgium.
1416 1417 1418 1419
De Gussem, K., Fenu, A., Wambecq, T. and Weemaes, M. (2014) Energy saving on wastewater treatment plants through improved online control: case study wastewater treatment plant AntwerpSouth. Water Science & Technology 65(5), 1074-1079.
1420 1421 1422 1423
Dhanasekharan, K.M., Sanyal, J., Jain, A. and Haidari, A. (2005) A generalized approach to model oxygen transfer in bioreactors using population balances and computational fluid dynamics. Chemical Engineering Science 60(1), 213-218.
1424 1425 1426 1427
Do-Quang, Z., Cockx, A., Liné, A. and Roustan, M. (1999) Computational fluid dynamics applied to water and wastewater treatment facility modeling. Environmental Engineering and Policy 1(3), 137-147.
1428 1429 1430
EPA (1989) Design Manual: Fine pore aeration systems, United States Environmental Protection Agency, Research and Development - Risk Reduction Engineering Laboratory, Cincinnati, OH
1431 1432
EPA (2013) Energy efficiency in water and wastewater facilities - A guide to developing and implementing greenhouse gas reduction
AC C
EP
TE D
M AN U
SC
RI PT
1390 1391
73
ACCEPTED MANUSCRIPT 1433 1434
programs United Washington, DC.
1435 1436 1437 1438
Essemiani, K., Vermande, S., Marsal, S., Phan, L. and Meinhold, J. (2004) Optimization of WWTP units using CFD - a tool grown for real scale application. van Loosdrecht, M. and Clement, J. (eds), pp. 183-192, IWA Publishing.
1439 1440 1441
Fan, L., Xu, N., Wang, Z. and Shi, H. (2010) PDA experiments and CFD simulation of a lab-scale oxidation ditch with surface aerators. Chemical Engineering Research and Design 88(1), 23-33.
1442 1443 1444
Fayolle, Y., Cockx, A., Gillot, S., Roustan, M. and Héduit, A. (2007) Oxygen transfer prediction in aeration tanks using CFD. Chemical Engineering Science 62(24), 7163-7171.
1445 1446 1447 1448
Fayolle, Y., Gillot, S., Cockx, A., Bensimhon, L., Roustan, M. and Heduit, A. (2010) In situ characterization of local hydrodynamic parameters in closed-loop aeration tanks. Chemical Engineering Journal 158(2), 207-212.
1449 1450 1451 1452 1453
Fayolle, Y., Gillot, S., Cockx, A., Roustan, M. and Heduit, A. (2006) In situ local parameter measurements for CFD modelling to optimize aeration. Proceedings of WEFTEC 06 - 79th Annual WEF Technical Exhibition and Conference, pp. 3314-3325, Oct. 21-25, Dallas, TX, USA.
1454 1455 1456 1457
Fernández, F.J., Castro, M.C., Rodrigo, M.A. and Cañizares, P. (2011) Reduction of aeration costs by tuning a multi-set point on/off controller: A case study. Control Engineering Practice 19(10), 12311237.
1458 1459 1460
Fikar, M., Chachuat, B. and Latifi, M.A. (2005) Optimal operation of alternating activated sludge processes. Control Engineering Practice 13(7), 853-861.
1461 1462 1463 1464
Gillot, S., Capela-Marsal, S., Roustan, M. and Héduit, A. (2005) Predicting oxygen transfer of fine bubble diffused aeration systems Model issued from dimensional analysis. Water Research 39(7), 1379-1387.
1465 1466 1467
Gillot, S. and Héduit, A. (2000) Effect of air flow rate on oxygen transfer in an oxidation ditch equipped with fine bubble diffusers and slow speed mixers. Water Research 34(5), 1756-1762.
1468 1469 1470 1471
Glover, G.C., Printemps, C., Essemiani, K. and Meinhold, J. (2006) Modelling of wastewater treatment plants - how far shall we go with sophisticated modelling tools? Water Science & Technology 53(3), 79-89.
1472 1473 1474
Gresch, M., Armbruster, M., Braun, D. and Gujer, W. (2011) Effects of aeration patterns on the flow field in wastewater aeration tanks. Water Research 45(2), 810-818.
Environmental
Protection
Agency,
AC C
EP
TE D
M AN U
SC
RI PT
States
74
ACCEPTED MANUSCRIPT Griborio, A. and McCorquodale, J.A. (2006) Optimum design of your center well: use of a CFD model to understand the balance between flocculation and improved hydrodynamics. Proceedings of WEFTEC 06 - 79th Annual WEF Technical Exhibition and Conference, pp. 5735-5746, Oct. 21-25, Dallas, TX, USA.
1480 1481 1482 1483
Guimet, V., Savoye, P., Audic, J.-M. and Do-Quang, Z. (2004) Advanced CFD tool for Wastewater: today complex Modelling & tomorrow easy-to-use Interface. Van Loosdrecht, M. and Clement, J. (eds), pp. 165-172, IWA Publishing.
1484 1485 1486
Henze, M., Gujer, W., Mino, T. and van Loosdrecht, M.C.M. (2000) Activated Sludge Models ASM1, ASM2, ASM2d and ASM3. IWA Scientific and Technical Report No. 9, London, UK.
1487 1488 1489 1490
Holenda, B., Domokos, E., Rédey, Á. and Fazakas, J. (2007) Aeration optimization of a wastewater treatment plant using genetic algorithm. Optimal Control Applications and Methods 28(3), 191208.
1491 1492 1493 1494
Holenda, B., Domokos, E., Rédey, Á. and Fazakas, J. (2008) Dissolved oxygen control of the activated sludge wastewater treatment process using model predictive control. Computers & Chemical Engineering 32(6), 1270-1278.
1495 1496 1497
Hounslow, M.J., Ryall, R.L. and Marshall, V.R. (1988) A discretized population balance for nucleation, growth, and aggregation. AIChE Journal 34(11), 1821-1832.
1498 1499 1500 1501
Jensen, M.D., Ingildsen, P., Rasmussen, M.R. and Laursen, J. (2006) Computational Fluid Dynamics modelling of hydraulics and sedimentation in process reactors during aeration tank settling. Water Science & Technology 53(12), 257-264.
1502 1503 1504
Ji, Z., McCorquodale, J.A., Zhou, S. and Vitasovic, Z. (1996) A dynamic solids inventory model for activated sludge systems. Water Environment Research 68(3), 329-337.
1505 1506 1507
Joshi, J.B. (2001) Computational flow modelling and design of bubble column reactors. Chemical Engineering Science 56(21–22), 5893-5933.
1508 1509 1510 1511
Karpinska, A.M., Dias, M.M., Boaventura, R., A. and Santos, R.J. (2015) Modeling of the hydrodynamics and energy expenditure of oxidation ditch aerated with hydrojets using CFD codes. Water Quality Research Journal of Canada 50(1), 83-94.
1512 1513 1514
Karpinska Portela, A.M. (2013) New design tools for Activated Sludge process. PhD Thesis, FEUP, University of Porto, Porto, Portugal.
1515 1516 1517 1518
Kjellstrand, R. (2006) Hydraulic Behaviour in an Activated Sludge tank. From tracer test through hydraulic modelling to full-scale implementation. Lic. of Eng. Thesis, Chalmers University of Technology, Göteborg, Sweden
AC C
EP
TE D
M AN U
SC
RI PT
1475 1476 1477 1478 1479
75
ACCEPTED MANUSCRIPT Kochevsky, A.N. (2004) Possibilities of simulation of fluid flows using the modern CFD software tools. arXiv preprint physics/0409104.
1522 1523 1524 1525
Langergraber, G., Rieger, L., Winkler, S., Alex, J., Wiese, J., Owerdieck, C., Ahnert, M., Simon, J. and Maurer, M. (2004) A guideline for simulation studies of wastewater treatment plants. Water Science & Technology 50(7), 131-138.
1526 1527 1528
Launder, B.E., Reece, G.J. and Rodi, W. (1975) Progress in the development of a Reynolds-stress turbulent closure. Journal of Fluid Mechanics 68, 537-566.
1529 1530 1531
Launder, B.E. and Spalding, D.B. (1974) The numerical computation of turbulent flows. Computer Methods in Applied Mechanics and Engineering 3(2), 269-289.
1532 1533 1534 1535 1536
Laurent, J., Samstag, R.W., Ducoste, J., Griborio, A., Nopens, I., Batstone, D.J., Wicks, J., Saunders, S. and Potier, O. (2014) A protocol for the use of computational fluid dynamics as a supportive tool for wastewater treatment plant modelling. Water Science & Technology 70(10), 1757-1584.
1537 1538 1539 1540
Laursen, J. (2007) WWTPprocess tank modelling - The integration of advanced hydrodynamic and microbiological models. PhD Thesis, Department of Civil Engineering, Aalborg University, Aalborg, Denmark.
1541 1542 1543 1544
Le Moullec, Y., Gentric, C., Potier, O. and Leclerc, J.P. (2010a) CFD simulation of the hydrodynamics and reactions in an activated sludge channel reactor of wastewater treatment. Chemical Engineering Science 65(1), 492-498.
1545 1546 1547 1548
Le Moullec, Y., Gentric, C., Potier, O. and Leclerc, J.P. (2010b) Comparison of systemic, compartmental and CFD modelling approaches: Application to the simulation of a biological reactor of wastewater treatment. Chemical Engineering Science 65(1), 343-350.
1549 1550 1551 1552
Le Moullec, Y., Potier, O., Gentric, C. and Leclerc, J.P. (2011) Activated sludge pilot plant: Comparison between experimental and predicted concentration profiles using three different modelling approaches. Water Research 45(10), 3085-3097.
1553 1554 1555 1556
Le Moullec, Y., Potier, O., Gentric, C. and Pierre Leclerc, J. (2008) Flow field and residence time distribution simulation of a cross-flow gas-liquid wastewater treatment reactor using CFD. Chemical Engineering Science 63(9), 2436-2449.
1557 1558 1559 1560
Lei, L. and Ni, J. (2014) Three-dimensional three-phase model for simulation of hydrodynamics, oxygen mass transfer, carbon oxidation, nitrification and denitrification in an oxidation ditch. Water Research 53(0), 200-214.
1561 1562
Levenspiel, O. (1999) Chemical Reaction Engineering, Wiley & Sons, New York, NY.
AC C
EP
TE D
M AN U
SC
RI PT
1519 1520 1521
76
ACCEPTED MANUSCRIPT Makinia, J. and Wells, S.A. (2000) A general model of the activated sludge reactor with dispersive flow - I. Model development and parameter estimation. Water Research 34(16), 3987-3996.
1566 1567
Manninen, M., Taivassalo, V. and Kallio, S. (1996) On the mixture model for the multiphase flow. VTT Publications 288, 1-67.
1568 1569 1570
Marchisio, D.L., Vigil, R.D. and Fox, R.O. (2003) Quadrature method of moments for aggregation–breakage processes. Journal of Colloid and Interface Science 258(2), 322-334.
1571 1572 1573 1574
Menter, F.R. (1993) Zonal two equation k-w turbulence models for aerodynamic flows. Proceedings of 23rd Fluid Dynamics, Plasmadynamics and Lasers Conference, July 6-9, Orlando, FL, USA.
1575 1576
Menter, F.R. (1994) Two-equation eddy-viscosity turbulence models for engineering applications. AIAA Journal 32(8), 1598-1605.
1577 1578 1579 1580
Morchain, J., Gabelle, J.-C. and Cockx, A. (2014) A coupled population balance model and CFD approach for the simulation of mixing issues in lab-scale and industrial bioreactors. AIChE Journal 60(1), 27-40.
1581 1582 1583
Morchain, J., Maranges, C. and Fonade, C. (2000) CFD modelling of a two-phase jet aerator under influence of a crossflow. Water Research 34(13), 3460-3472.
1584 1585
Mueller, J.A., Boyle, W.C. and Pöpel, H.J. (2002) Aeration: principles and practice, CRC Press, Boca Raton, FL, USA.
1586 1587
Nauman, E.B. (2007) Residence Time Distributions, pp. 535-574, John Wiley & Sons, Inc., New Jersey, NJ, USA.
1588 1589 1590
Nielsen, M.K., Bechmann, H. and M., H. (2000) Modelling and test of aeration tank settling (ATS). Water Science & Technology 41(9), 179-184.
1591 1592 1593 1594 1595
Nopens, I., Batstone, D.J., Griborio, A., Samstag, R.W., Wicklein, E. and Wicks, J. (2012) Computational Fluid Dynamics (CFD): What is Good CFD-Modeling Practice and what can be the added value of CFD models to WWTP modeling?, pp. 7400-7405, Sept. 29-Oct. 3, New Orleans, LA, USA.
1596 1597 1598
Nopens, I., Koegst, T., Mahieu, K. and Vanrolleghem, P.A. (2005) PBM and activated sludge flocculation: From experimental data to calibrated model. AIChE Journal 51(5), 1548-1557.
1599 1600 1601 1602
Nopens, I., Torfs, E., Ducoste, J., Vanrolleghem, P.A. and Gernaey, K.V. (2015) Population balance models: a useful complementary modelling framework for future WWTP modelling. Water Science & Technology 71(2), 159-167.
1603 1604
Nopens, I. and Wicks, J. (2012) Integration of CFD and Biokinetics. CFD Workshop- Presentation. 3rd IWA/WEF Wastewater Treatment
AC C
EP
TE D
M AN U
SC
RI PT
1563 1564 1565
77
ACCEPTED MANUSCRIPT 1605 1606
Modelling Canada.
1607 1608
Orszag, S.A. (1970) Analytical theories of turbulence. Journal of Fluid Mechanics 41(2), 363-386.
1609 1610 1611 1612
Orszag, S.A., Staroselsky, I., Flannery, W.S. and Zhang, Y. (eds) (1996) Simulation and modeling of turbulent flows. Chapter 4: Introduction to Renormalization Group Modeling of Turbulence, Oxford University Press Inc., New York, NY, USA.
1613 1614 1615 1616
Patziger, M., Kainz, H., Hunze, M. and Józsa, J. (2012) Influence of secondary settling tank performance on suspended solids mass balance in activated sludge systems. Water Research 46(7), 24152424.
1617 1618 1619 1620
Pereira, J.P., Karpinska, A., Gomes, P.J., Martins, A.A., Dias, M.M., Lopes, J.C.B. and Santos, R.J. (2012) Activated Sludge Models coupled to CFD simulations. Dias, R., Lima, R., Martins, A.A. and Mata, T.M. (eds), pp. 153-173, Bentham Science Publishers Ltd.
1621 1622 1623
Pereira, J.P., Karpinska, A.M., Dias, M.M. and Santos, R.J. (2009) The dissolved oxygen profile in oxidation ditches: a numerical study on its effect on nutrient removal, pp. 97-104, Krakow, Poland.
1624 1625 1626 1627 1628
Plósz, B.G., De Clercq, J., Nopens, I., Benedetti, L. and Vanrolleghem, P.A. (2011) Shall we upgrade one-dimensional secondary settler models used in WWTP simulators? – An assessment of model structure uncertainty and its propagation. Water Science & Technology 63(8), 1726–1738.
1629 1630 1631 1632 1633 1634
Plósz, B.G., Nopens, I., Rieger, L., Griborio, A., De Clercq, J., Vanrolleghem, P.A., Daigger, G.T., Takács, I., Wicks, J. and Ekama, G.A. (2012) A critical review of clarifier modelling: State-of-the-art and engineering practices. Proceedings of WWTmod2012 - 3rd IWA/WEF Wastewater Treatment Modelling Seminar, pp. 27-30, Feb. 26-28, Mont-Sainte-Anne, Québec, Canada.
1635
Pope, S.B. (2000) Turbulent Flows, New York, NY, USA.
Mont-Sainte-Anne,
QC,
AC C
EP
TE D
M AN U
SC
RI PT
Seminar-WWTmod2012,
1636 1637
Randolf, A.D. and Larson, M.A. (1971) Theory of particulate processes, Academic Press, New York, NY, USA.
1638 1639 1640 1641 1642
Ratkovich, N. (2010) Understanding hydrodynamics in Membrane Bioreactor systems for wastewater treatment: Two-phase empirical and numerical modelling and experimental validation. PhD thesis, Faculty of Bioscience Engineering, Ghent University, Ghent, Belgium.
1643 1644 1645 1646 1647
Ratkovich, N., Hunze, M., Futselaar, H. and Nopens, I. (2009) Use of CFD to model and optimize the hydrodynamics of an airlift MBR side-stream module. Proceedings of WEFTEC 09 - 82th Annual WEF Technical Exhibition and Conference, pp. 2812-2818, Oct. 1014, Orlando, FL, USA
78
ACCEPTED MANUSCRIPT Reardon, D.J. (1995) Turning down the power. Civil Engineering 65(8), 54-56.
1650 1651 1652
Rieger, L., Alex, J., Gujer, W. and Siegrist, H. (2006) Modelling of aeration systems at wastewater treatment plants. Water Science & Technology 53(4-5), 439-447.
1653 1654
Rodi, W. (1993) Turbulence models and their application in hydraulics., IAHR Monograph, Delft, The Netherlands.
1655 1656 1657 1658
Rosso, D., Lothman, S.E., Jeung, M.K., Pitt, P., Gellner, W.J., Stone, A.L. and Howard, D. (2011) Oxygen transfer and uptake, nutrient removal, and energy footprint of parallel full-scale IFAS and activated sludge processes. Water Research 45(18), 5987-5996.
1659 1660
Rosso, D. and Stenstrom, M.K. (2006) Surfactant effects on α-factors in aeration systems. Water Research 40(7), 1397-1404.
1661 1662 1663 1664
Samstag, R.W., Ducoste, J.M., Griborio, A., Nopens, I., Batstone, D.J., Wicks, J.D., Saunders, S., Wicklein, E., Kenny, G. and Laurent, J. (2015) CFD for wastewater treatment: A State of the Art review. Submitted manuscript.
1665 1666 1667
Samstag, R.W., McCorqudale, J.A. and Zhou, S.P. (1992) Prospects for Transport Modelling in Process Tanks. Water Science & Technology 26(5-6), 1401-1410.
1668 1669 1670 1671
Samstag, R.W. and Wicklein, E. (2012) Application of CFD to Activated Sludge mixing analysis. CFD Workshop- Presentation. 3rd IWA/WEF Wastewater Treatment Modelling SeminarWWTmod2012, February 26-28 Mont-Sainte-Anne, QC, Canada.
1672 1673 1674 1675
Samstag, R.W. and Wicklein, E. (2014) A Protocol for optimization of Activated Sludge mixing. Proceedings of WEFTEC - 87th Annual WEF Technical Exhibition and Conference, Sept. 27-Oct. 1, New Orleans, LA, USA.
1676 1677 1678 1679 1680
Samstag, R.W., Wicklein, E., Reardon, R.D., Leetch, R.J., Parks, R.M. and Groff, C.D. (2012) Field and CFD analysis of jet aeration and mixing. Proceedings of WEFTEC 2012 - 85th Annual WEF Technical Exhibition and Conference, pp. 4113-4139, Sept. 29-Oct. 3, New Orleans, LA, USA.
1681 1682 1683
Shaikh, A. and Al-Dahhan, M.H. (2007) A Review on flow regime transition in bubble columns. International Journal of Chemical Reactor Engineering 5(1), 1-68.
1684 1685 1686 1687 1688
Sharma, A.K., Guildal, T., Thomsen, H.A., Mikkelsen, P.S. and Jacobsen, B.N. (2013) Aeration tank settling and real time control as a tool to improve the hydraulic capacity and treatment efficiency during wet weather: results from 7 years' full-scale operational data. Water Science & Technology 67(10), 2169-2176.
1689 1690
Shih, T.-H., Liou, W.W., Shabbir, A., Yang, Z. and Zhu, J. (1995) A New K-epsilon eddy viscosity model for high Reynolds number
AC C
EP
TE D
M AN U
SC
RI PT
1648 1649
79
ACCEPTED MANUSCRIPT turbulent flows: Model development and validation. Computers & Fluids 24(3), 227-238.
1693 1694
Smagorinsky, J. (1963) General circulation experiments with the primitive equations. Monthly Weather Review 93(3), 99-164.
1695 1696 1697 1698 1699 1700
Sobremisana, A.P., Ducoste, J.J. and de los Reyes III, F.L. (2011) Combining CFD, floc dynamics and biological reaction kinetics to model carbon and nitrogen removal in an activated sludge system. Proceedings of WEFTEC 2011 - 84th Annual WEF Technical Exhibition and Conference, pp. 3272-3282, Oct. 15-19, Los Angeles, CA, USA.
1701 1702 1703 1704
Sokolichin, A., Eigenberger, G., Lapin, A. and Lübert, A. (1997) Dynamic numerical simulation of gas-liquid two-phase flows Euler/Euler versus Euler/Lagrange. Chemical Engineering Science 52(4), 611-626.
1705 1706 1707
Stamou, A.I. (2008) Improving the hydraulic efficiency of water process tanks using CFD models. Chemical Engineering and Processing: Process Intensification 47(8), 1179-1189.
1708 1709 1710
Stamou, A.I., Theodoridis, G. and Xanthopoulos, K. (2009) Design of secondary settling tanks using a CFD model. Journal of Environmental Engineering 135(7), 551-561.
1711 1712
Takács, G. (2005) Gas Lift: Manual, PennWell Books, Tulsa, OK, USA.
1713 1714 1715 1716
Taricska, J.R., Huang, J.Y.C., Chen, P., Hung, Y.-T. and Zou, S.-W. (2009) Biological treatment processes. Wang, L.K., Pereira, N.C., Hung, Y.-T. and Shammas, N.K. (eds), The Humana Press, Totowa, NJ.
1717 1718 1719
Tchobanoglous, G., Burton, F.L., Stensel, H.D. and Eddy, M. (eds) (2003) Wastewater engineering: treatment and reuse, McGraw-Hill New York, NY.
1720 1721 1722
Vedantam, S., Joshi, J.B. and Koganti, S.B. (2006) Threedimensional CFD simulation of stratified two-fluid Taylor-Couette flow. The Canadian Journal of Chemical Engineering 84(3), 279-288.
1723 1724 1725 1726
Vermande, S., Simpson, K., Essemiani, K., Fonade, C. and Meinhold, J. (2007) Impact of agitation and aeration on hydraulics and oxygen transfer in an aeration ditch: Local and global measurements. Chemical Engineering Science 62(9), 2545-2555.
1727 1728 1729
Versteeg, H.K. and Malalasekera, W. (1995) An introduction to Computational Fluid Dynamics. The Finite Volume Method, Pearsons Prentice Hall, Harlow, Essex, UK.
1730 1731 1732
Wang, T. (2011) Simulation of bubble column reactors using CFD coupled with a population balance model. Frontiers of Chemical Science and Engineering 5(2), 162-172.
AC C
EP
TE D
M AN U
SC
RI PT
1691 1692
80
ACCEPTED MANUSCRIPT Wang, Y., Waite, T.D. and Leslie, G.L. (2013) Handbook of membrane reactors. Basile, A. (ed), pp. 532-568, Woodhead Publishing, Cambridge, UK.
1736 1737 1738
WEF (2009) Energy conservation in water and wastewater treatment facilities, Water Environment Federation Manual of Practice No. 32; McGraw-Hill, Inc., New York, NY.
1739 1740 1741 1742
Wicklein, E. and Samstag, R.W. (2009) Comparing commercial and transport CFD models for secondary sedimentation. Proceedings of WEFTEC 09 - 82th Annual WEF Technical Exhibition and Conference, pp. 6066-6081, Oct. 10-14, Orlando, FL, USA
1743 1744 1745
Wilcox, D.C. (1998) Re-assessment of the scale-determining equation for advanced turbulence models. AIAA Journal 26(11), 1299-1310.
1746 1747 1748
Wilcox, D.C. and Traci, R.M. (1976) A complete model of turbulence. Proceedings of 9th Fluid and PlasmaDynamics Conference, July 14-16, San Griego, CA, USA.
1749 1750 1751 1752
Xie, H., Yang, J., Hu, Y., Zhang, H., Yang, Y., Zhang, K., Zhu, X., Li, Y. and Yang, C. (2014) Simulation of flow field and sludge settling in a full-scale oxidation ditch by using a two-phase flow CFD model. Chemical Engineering Science 109(0), 296-305.
1753 1754 1755 1756
Xu, N., Fan, L., Pang, H. and Shi, H. (2010) Feasibility study and CFD-aided design for a new type oxidation ditch based on airlift circulation. The Canadian Journal of Chemical Engineering 88(5), 728-741.
1757 1758
Yang, C. and Mao, Z.-S. (2014) Numerical simulation of multiphase reactors with continuous liquid phase, Academic Press, Oxford, UK.
1759 1760 1761 1762
Yang, Y., Yang, J., Zuo, J., Li, Y., He, S., Yang, X. and Zhang, K. (2011) Study on two operating conditions of a full-scale oxidation ditch for optimization of energy consumption and effluent quality by using CFD model. Water Research 45(11), 3439-3452.
1763 1764
Yeoh, G.H. and Tu, J. (2009) Computational techniques for multiphase flows, Elsevier.
AC C
EP
TE D
M AN U
SC
RI PT
1733 1734 1735
1765 1766
81
ACCEPTED MANUSCRIPT Figure captions Figure 1. Streamlines coloured by liquid velocity magnitude ܷ in tank (a) without aeration (ܷ = 0.35 ms−1) (b) without aeration (ܷ = 0.27 ms−1) and (c) with aeration (ܷ = 0.23 ms−1). Reprinted from Fayolle et al. (2007). Copyright (2007) with permission from Elsevier.
RI PT
Figure 2. Snapshot of air volume fractions at two cross sections with different diffuser patterns. Reprinted from Gresch et al. (2011). Copyright (2011) with permission from Elsevier.
Figure 3. Contours of volume fraction of solid phase at different height: (A) top of the tank, (B) middle of the tank, and (C) bottom of the tank. Reprinted from Xie et al. (2014). Copyright (2014) with permission from Elsevier.
M AN U
SC
Figure 4. Comparison between experimental and simulated RTD obtained with the RSM and the k – ε turbulence models and the particle tracking method for a liquid flowrate of 3.6 L min-1 and a gas flowrate of 15 L min-1. Reprinted from Le Moullec et al. (2008). Copyright (2008) with permission from Elsevier.
AC C
EP
TE D
Figure 5. The contour plots of: (a) MLSS, (b) DO, (c) COD, (d) ammonia and (e) nitrate concentration distribution, where “+” indicates measured data values at the sampling points in a Carrousel oxidation ditch. Reprinted from Lei and Ni (2014). Copyright (2014) with permission from Elsevier.
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT
Highlights
EP
TE D
M AN U
SC
RI PT
CFD analysis is a robust tool for evaluation and optimization of the AS systems. Turbulence models and multiphase approaches used to study AS tanks are presented. Pitfalls of modelling assumptions and simplifications are identified. Methods and examples of coupling of the CFD data with biokinetics are discussed. Unaddressed challenges in modelling of the AS systems are critically discussed.
AC C
• • • • •