Chemical Engineering Journal 243 (2014) 5159

Contents lists available at ScienceDirect

Chemical Engineering Journal

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

Oxygen transfer model development based on activated sludge and clean

water in diffused aerated cylindrical tanks
Erika Pittoors a,b, Yaping Guo a,, Stijn W.H. Van Hulle b,c,d,
a

College of Biological and Environmental Engineering, Zhejiang University of Technology, 18 Chaowang Road, Hangzhou Zhejiang, PR China
ENBICHEM Research Group, University College West Flanders, Graaf Karel de Goedelaan 5, 8500 Kortrijk, Belgium
c
BIOMATH, Department of Mathematical Modelling, Statistics and Bioinformatics, Ghent University, Coupure Links 653, 9000 Gent, Belgium
d
LIWET, Department of Industrial Biological Sciences, Ghent University, Graaf Karel de Goedelaan 5, 8500 Kortrijk, Belgium
b

h i g h l i g h t s
 The volumetric oxygen mass transfer kLa was measured under different operational conditions.
 Experiments in clean water and with activated sludge were done.
 The experimental results were used to develop a high t empirical model.
 The airow rate was the main factor affecting the kLa.

a r t i c l e

i n f o

Article history:
Received 4 July 2013
Received in revised form 20 September 2013
Accepted 22 December 2013
Available online 10 January 2014
Keywords:
Dissolved oxygen (DO)
Activated sludge process (ASP)
Diffused aeration
Dimensional analysis
Volumetric mass transfer coefcient (kLa)
Static method

a b s t r a c t
The oxygen mass transfer kLa is generally studied under non-reactive conditions, leaving out the most
fundamental operational condition in activated sludge processes (ASPs). Existing oxygen transfer models,
used in wastewater treatment plant design and optimizations, have therefore a major shortcoming. More
accurate kLa models lead to improved system analysis and knowledge. This work studied the volumetric
oxygen mass transfer kLa in an ASP, under varying operational conditions. An empirical correlation for kLa
versus nine studied variables (tank volume (Vt), height (Ht), diameter (Dt), surface area (At), airow rate
(Qa), diffusers surface area (Ad) and depth (hd), bubble size (db) and dynamic viscosity (l)) for clean water
(kLaCW) and for activated sludge (kLaAS) in a diffused aerated cylindrical batch reactor is created. The
experimental results were used to develop a high t empirical model for kLaAS (R2 = 0.96) and kLaCW
(R2 = 0.95). The following equations were obtained (kLa in s1):

!0:017
 0:291  0:554  0:135  0:321  0:086
D2t kL aCW
db
Ht
Ad
Dt
Ht
Vt
0:030Re1:718 Fr0:709
D
hd
Dt
At
hd
hd
A1:5
d
!0:01
 0:23  0:120  0:326  0:164  0:173
2
Dt kL aAS
db
Ht
Ad
Dt
Ht
Vt
0:060Re1:906 Fr0:631
:
D
hd
Dt
At
hd
hd
A1:5
d





a
were used. The coefThe Reynolds Re mmL QDagq and the (adapted) Froude number Fr pm pQ
5
Lg

Dt g

cients for clean water and activated sludge varied up to 66% for the same base model but show similar
trends and effects for different hydrodynamic, physicochemical and geometrical parameters. The airow
rate was the main factor affecting both kLaAS and kLaCW. Next were diffusers depth and bubble size. Airow
rate and diffusers surface area had a signicantly larger impact in the presence of biomass, since it promotes bubble distribution, mixing of the solution and an improved oxygen transfer, therefor demonstrating the need for an adapted model for ASPs.
2014 Elsevier B.V. All rights reserved.

1. Introduction
Corresponding authors. Tel.: +86 13858043191 (Y. Guo). Address: ENBICHEM
Research Group, University College West Flanders, Graaf Karel de Goedelaan 5, 8500
Kortrijk, Belgium. Tel.: +32 56241251; fax: +32 56241224 (S. Van Hulle).
E-mail addresses: erika.pittoors@howest.be (E. Pittoors), luckyyaping@zjut.
edu.cn (Y. Guo), stijn.vanhulle@ugent.be (S.W.H. Van Hulle).
1385-8947/$ - see front matter 2014 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.cej.2013.12.069

The oxygen mass transfer, kLa, in wastewater is often studied

under nonreactive conditions, even though it is known that biomass signicantly impacts the oxygen transfer in activated sludge
systems (ASPs) [13]. Existing oxygen transfer models, used among

52

E. Pittoors et al. / Chemical Engineering Journal 243 (2014) 5159

Nomenclature

a
l
m
q
h
A/A/O
Ad
At
AS(P)
Bt
CO2
CO2
D
db
deq
Dt
DO
Fr
g
Ht
hd
HRT

alpha factor ()
dynamic viscosity (kg/m/s)
kinematic viscosity (m2/s)
density (kg/m3)
temperature correction factor ()
anaerobicanoxicoxic
total coverage area of the diffusers (m2)
total area of the tank (m2)
activated sludge (process)
width of the tank (m)
concentration of dissolved oxygen (mgO2/l)
oxygen saturation concentration (mgO2/l)
diffusion coefcient (m2/s)
bubble diameter (m)
equivalent bubble diameter (m)
tank diameter (m)
dissolved oxygen (mg/l)
Froude number ()
acceleration due to gravity (=9.8 m/s2)
height of tank (m)
diffuser submergence (m)
hydraulic retention time (h)

kLa
kLaAS
kLaCW
L
MLSS
OTR
OUR
Q
Qa
SRT
T
TIC
V
ycalc
yexp

Table 1
Empirical correlations for kLa prediction for diffused aeration systems.
Empirical correlation
L2 k L a
D

1:46

0:033Re

kL a

 0:73  
 0:24
H 1:77 A

Fr0:49 dhb

At

ph0:44
1:3 0:93 W 0:49 H 1:63 p0:074
h
6:86dB
H
L
ph 1

 
 0:5  D  At 0:72
kL a 49Re Dm
2
D2
t

kL a

 2 1=3
m

hd

7:77  105

At

Table 2
Ranges of the operational variables and derived (non-dimensional) variables.
Reference

Variable

Name

Unit

Range or value

[12]

Qa
hd
Vt
Ht
Dt
At
db
Ad
MLSS

lCW

Airow rate
Diffusers depth
Tank volume
Tank height
Tank diameter
Tank surface area
Bubble diameter
Diffusers surface area
Mixed liquor Suspended Solids
Dynamic viscosity CW

0.240.60
0.010.24
2.79.3
0.130.53
0.100.30
1.8  103 7.4  103
0.005
7  104
9182543
1.1  103 9.8  104

lAS

Dynamic viscosity AS

Density
Oxygen diffusion coefcient
Reynolds CW
Reynolds AS
Froude

l/min
m
l
m
m
m2
m
m2
mg/l
kg/
ms
kg/
ms
kg/m3
m2/s

[34]
[10]

 0:24  0:15  0:13

Ap
Ap
Dt
Ad

volumetric mass transfer coefcient (h1, unless otherwise specied)

volumetric mass transfer coefcient in activated sludge
tank (h1, unless otherwise specied)
volumetric mass transfer coefcient in clean water tank
(h1, unless otherwise specied)
length of the tank (m)
mixed liquor suspended solids (mg/l)
oxygen transfer rate (mgO2/l/d)
oxygen up-take rate by microorganisms (mgO2/l/d)
volumetric wastewater ow rate (m3/s unless otherwise
stated)
airow rate (m3/s)
sludge retention time (d)
temperature (C)
Theils inequality coefcient ()
working volume reactor (m3)
calculated values
data points obtained through the experiments

[6]

others in wastewater treatment plant (WWTP) optimizations, have

therefore a major shortcoming as they are not based on the most
fundamental operational condition in ASPs. Often oxygen mass
transfer is measured to check the performance of ASPs before
start-up of the WWTP or during design of a new one. These tests
are mainly done in clean water, following the ASCE [4] and NFEN
[5] standard. This leads to signicant inaccuracies in oxygenation
performance prediction of the full-scale system, since many factors
in the wastewater affect the oxygen transfer [6]. Physicochemical
(solution composition, biomass, viscosity, pH, TSS, dissolved oxygen (DO), etc.), geometrical (aerator submergence, length and
width of the tank, total tank area, bubble diameter, diffusers total
coverage area, reactors working volume, etc.) and dynamical
parameters (airow rate, water density, surface tension, kinematic
viscosity, airow velocity, etc.) and aerator type contribute to aeration and oxygen mass transfer all to a different extent depending
on the wastewater type, treatment system and equipment used.
Rening the oxygen mass transfer prediction will lead to an improved optimized system, meaning reduced costs and increased
effectiveness of ASPs and even WWTPs. This remark counts particularly for medium-sized plants, as operational inspections are
more difcult to accomplish systematically [6].
Diffused aeration (subsurface or submerged bubble aeration) is
dened as the injection of air or oxygen enriched air under pressure below a liquid surface [7]. Air is blown into the water by
means of diffusers or mechanical agitators. This contribution deals
with submerged ne pore diffusers (db < 5 mm). These release air
via porous media or nozzles at increased depths [8]. The ne-pore

D
ReCW
ReAS
Fr
db/h
Ht/Dt
Ad/At
Dt/hd
Ht/hd
V t =A1:5
d

1.4  103 1.9  103

998
1.86  109
13105
1454
2.6  105 1.1  103
0.0210.76
0.443.8
0.010.096
0.5746
220
145494

aerators are different from coarse bubble aerators, owing to their

ow regime. The former has low interfacial gas velocities, hence induces low ow regimes at gasliquid interfaces. Due to this ow
regime (and the added decrease through scaling and fouling over
time) the alpha factor (a, [9]) is substantially lower compared to
the wide variety of aeration systems, leading to a lowered kLa. In
contrast coarse-bubbles produce greater velocity gradients at the
gasliquid interface [8]. However the kLa of coarse bubble aerators
is about six times smaller in contrast to the ne bubble KLa. Nonetheless when demanding an equal kLa, 3 times more diffusers are
needed for ne bubble diffusers as these have the highest Ad, but
demand a lower airow rate, hence a higher power consumption
(at 10 m tank depth +40%) [10]. Diffusers are still the most

53

E. Pittoors et al. / Chemical Engineering Journal 243 (2014) 5159

tested combinations of operational parameters such as tank geometry, airow rate and diffuser surface and depth can be found in Table 2. A 3 cm diameter round gas sparger was utilized, which
created creating ne bubbles. The air went rst through a calibrated ow meter connected to the diffusers. The set-up is shown
in Fig. 1. In total 47 aeration experiments in clean water and AS
were performed.
2.2. Experimental determination of kLaCW and kLaAS
The dissolved oxygen was measured under equal changing
operational variables according to the static ASCE Standard method
[4] for both clean water and activated sludge. The Oxygen Uptake
Rate was assumed to be low during the tests, although possible
oxygen consumption was accounted for by estimating the actual
oxygen saturation concentration as described below. As such the
kLa can be determined by integration of Eq. (1), which is based
on the oxygen saturation level and the DO variation over time.

OTR

Fig. 1. Cylindrical batch reactor used as experimental set-up for kLa determination
(inuent and efuent connections are foreseen, but were not used in this study).

common aeration system in ASPs in developed countries and provide the extra benet of keeping the solution well-mixed [11].
As such the aim of this work is the study of the volumetric oxygen mass transfer kLa of diffused aerators in an ASP, under varying
operational conditions, by measurement of the dissolved oxygen
(DO). An empirical correlation for kLa versus nine studied variables
(Table 1; tank volume (Vt), height (Ht), diameter (Dt) & surface area
(At), airow rate (Qa), diffusers surface area (Ad) & depth (hd), bubble size (db) and dynamic viscosity (l)) for clean water and for activated sludge in a diffused aerated cylindrical batch reactor will be
created. Via the static method (ASCE [5]) the oxygen mass transfer
from clean water and AS samples will be tested under equal varying operational conditions. An adapted model (based on Al-Ahmady [12] for rectangular tanks) for diffused aeration in wastewater,
similar to the many existing models (although based on clean
water, Table 1) will be created via the dimensionless analysis.

2. Materials and methods

dCO2
kL aCO2  CO2
dt

With the help of Na2SO3, dependent on the expected theoretical

oxygen demand and the catalyst CoCl26H2O, the DO in the tank is
sequestered. To decrease DO by 1 mgO2/l, 7.9 mg/l Na2SO3 is
needed. When reaching an oxygen concentration of 0 mgO2/l in
the tank, the aeration is turned on. A typical DO prole for both
the test with clean water and the tests with activated sludge is given in Fig. 2.
The addition of Na2SO3 with the aeration system off may induce
mixing problems and potentially settling of the sludge. This was
observed in some (not all) experiments. For example in Fig. 2 it
can be seen that a delay in the oxygen prole occurs during the
rst minutes. This may be linked to a non-uniform distribution
of the added chemicals in the reactor and introduce uncertainty
in the obtained results. In order the account for this delay, the rst
3 min of each data set were not used for kLa determination.
The test duration is 30 min. After each test the activated sludge
is discarded, the batch reactor cleaned and similar fresh sludge is
added. As such the impact of adding the chemicals is limited. To
make sure the physicalchemical parameters of the sludge stayed
the same, the TSS of the reactor was closely monitored and the
average MLSS was kept at around 2.5 g/l.
Using a correction factor, the oxygen transfer coefcient values
were standardized to a temperature of 20 C with the help of Eq.
(2) [4,11,13].

2.1. Aerated tanks

kL aT h20T kL a20  C

All the experiments have been carried out in lab-scale cylindrical batch reactors of different sizes (volume range: 39 l). The

For clean water the temperature correction factor h equals

1.024 according to the ASCE standard [4]. This factor is also used
in diffused aeration [11] and was therefore also used for the activated sludge experiments.

Table 3
Characteristics of the two Sequencing Batch Reactors (SBRs) from the the Haining
Wastewater Treatment Plant of the Zhejiang Province (China) from which the sludge
for the experiments was sampled.

Fig. 2. Typical DO graph for the batch reactor, obtained with the method for the
determination of kLaCW and kLaAS with Qa = 0.6 l/min, V = 2.7 l and TSS = 2.5 g/l
(during the activated sludge test).

Variable

Unit

Range

Average inuent ow rate (Q)

HRT (1 cycle, 6 cycles/day)
SRT
Tank volume
Inuent COD concentration
Inuent total nitrogen concentration
Inuent total phosphorous concentration
%COD removal
%N removal

m3/d
h
d
m3
mgO2/l
mgN/l
mgP/l
%
%

100,000
2
12
50,160
500
25
4
86
84

54

E. Pittoors et al. / Chemical Engineering Journal 243 (2014) 5159

For the oxygen saturation concentration (CO2 ) a temperature

dependent [4] equation could be used (Eq. (3)). However, this
equation is valid in clean water conditions but not under process
conditions. During the experiments the actual oxygen saturation
concentration can change because of variation in ionic strength
and endogenous respiration of the activated sludge. As in both
types of experiments (clean water and activated sludge) the ionic
strength was changed by adding chemicals, Eq. (3) was not used
but instead an oxygen saturation concentration was estimated
based on all the obtained oxygen proles. As such an oxygen saturation concentration of 6.1 mgO2/l was used for the clean water
experiments and an oxygen saturation concentration of 5.8
mgO2/l was used for the experiments with activated sludge.

CO2 14:65  0:41T 7:99  103 T 2  7:78  105 T 3

(Fig. 3). The ellipsoid bubble shape is calculated according to Eq. (4)
for ellipsoid volumes with an equivalent diameter deq. In this equation a stands for the major axis and b for the minor. Next the Sauter
diameter db of minimum 100 bubbles is calculated according to Eq.
(5).

deq

3
a2 b

PN

3
i1 deq
2
i1 deq

db PN

During all experiments the bubble diameter did not vary significantly and averaged 5 mm.

2.3. Activated sludge collection

2.5. Physicalchemical measurements

The Haining Wastewater Treatment Plant of the Zhejiang Province in China supplied the activated sludge for the experiments
from their sequencing batch reactor (characteristics in Table 3)
and only good quality sludge with excellent settle-ability was used.
The AS was concentrated by settling from 4 g/l to around 14 g/l
MLSS for easy transportation. The AS was used for the tests within
48 h of sampling, so all the tests are under equal physicochemical
conditions and there was no need for aeration or feeding. In order
to obtain the correct concentration for testing (+/2.5 g/l) the
sludge was diluted with ltered waste water efuent. As such real
wastewater characteristics were used as this research precedes a
model development research for pilot-scale SBRs (Sequencing
Batch Reactors) and A/A/O ASPs (Anaerobic/Anoxic/Oxic reactor).

Inuent and efuent quality were measured following the Standard Methods [15]. The dissolved oxygen and temperature were
monitored during the aeration cycles with an oxygen selective
electrode from INESA, model JPB-607A with a response time less
than 30 s. Immediately transferring the probe from an oxygen saturated solution to a solution with zero oxygen, gives the response
time of the probe [16]. The response time of the DO electrode was
assumed to be fast enough, without affecting the measuring accuracy. The temperature uctuated around 1522 C throughout the
entire testing period.
Dynamic viscosity (l), diffusion coefcient (D) and the density
(q) of clean water are based on Tchobanoglous et al. [11] (Table 2).
Apparent viscosity was measured at 25 C by means of a viscometer from Hangping MC, Model NDJ-1, using spindle no1. Notice that
this equipment is for Newtonian uids, while activated sludge is
known as non-Newtonian with ever varying rheological characteristics [2]. As such some deviation might occur. As the AS in this re-

2.4. Bubble diameter determination

The bubble diameter can be measured with an easy and fast
method developed by Fayolle et al. [14]. A ruler placed in the water
in front of the camera lens allows the measurement of the bubbles

Fig. 3. Typical picture obtained in the bubble plume for bubble diameter
determination.

Table 4
Overview of the combinations of operational variables used during the experiments
with clean water and resulting kLaCW.
Ht
m

Dt
M

Qa
l/min

hd
m

At
m2

V
l

Tcw
C

lCW
kg/ms

kLaCW
1/h

0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.53
0.53
0.53
0.53
0.53
0.53
0.53
0.53
0.53
0.13
0.13
0.13
0.13
0.13
0.37
0.37
0.37
0.37
0.37
0.37

0.10
0.10
0.10
0.10
0.10
0.10
0.10
0.10
0.10
0.15
0.15
0.15
0.15
0.15
0.15
0.15
0.15
0.15
0.30
0.30
0.30
0.30
0.30
0.15
0.15
0.15
0.15
0.15
0.15

0.60
0.40
0.24
0.60
0.40
0.24
0.60
0.40
0.24
0.60
0.40
0.24
0.60
0.40
0.24
0.60
0.40
0.24
0.60
0.40
0.24
0.60
0.60
0.60
0.40
0.24
0.60
0.60
0.24

0.02
0.02
0.02
0.08
0.08
0.08
0.17
0.17
0.17
0.03
0.03
0.03
0.11
0.11
0.11
0.24
0.24
0.24
0.01
0.01
0.01
0.03
0.03
0.02
0.02
0.02
0.08
0.17
0.17

7.4  103
7.4  103
7.4  103
7.4  103
7.4  103
7.4  103
7.4  103
7.4  103
7.4  103
1.8  102
1.8  102
1.8  102
1.8  102
1.8  102
1.8  102
1.8  102
1.8  102
1.8  102
7.1  102
7.1  102
7.1  102
7.1  102
7.1  102
1.8  102
1.8  102
1.8  102
1.8  102
1.8  102
1.8  102

2.7
2.7
2.7
2.7
2.7
2.7
2.7
2.7
2.7
9.3
9.3
9.3
9.3
9.3
9.3
9.3
9.3
9.3
9.3
9.3
9.3
9.3
9.3
6.5
6.5
6.5
6.5
6.5
6.5

21.1
21.1
20.2
21.0
15.9
16.0
20.5
15.9
16.0
19.1
18.1
16.0
18.8
17.9
16.0
18.2
17.8
19.1
18.1
17.4
17.8
18.0
18.4
18.3
18.3
18.4
17.8
18.7
18.1

9.8  104
9.8  104
1.0  103
9.8  104
1.1  103
1.1  103
9.9  104
1.1  103
1.1  103
1.0  103
1.1  103
1.1  103
1.0  103
1.1  103
1.1  103
1.1  103
1.1  103
1.0  103
1.1  103
1.1  103
1.1  103
1.1  103
1.0  103
1.0  103
1.0  103
1.0  103
1.1  103
1.0  103
1.1  103

8.37
4.47
2.32
6.33
3.65
2.55
6.27
3.39
2.46
3.48
2.20
1.19
3.25
1.68
0.94
1.51
1.60
1.05
2.56
1.98
1.30
2.55
2.02
4.00
3.04
1.78
3.68
2.72
1.39

55

E. Pittoors et al. / Chemical Engineering Journal 243 (2014) 5159

search has a low viscosity (l < 1.9  103 kg/ms, Table 1), l is assumed constant for model development.
2.6. Non-linear regression analysis and statistical analysis
The non-linear regression analysis was conducted with the help
of the solver function in Microsoft Excel according to Brown [17]
by minimizing the quadratic difference between measured and calculated values by changing the parameters values b1 through b9
(see below). The correlation index (R2) was calculated and also a
two tailed t-test was performed. The null hypothesis of this t-test
states that there is a small difference between ycalc and yexp, where
yexp represents the data points obtained through the experiments
and ycalc stands for the values calculated via the developed model
(see below). If the statistic t-value is closer to zero than the critical
t value (95% condence interval), than the null hypothesis to be accepted. And it can be concluded that the kLa can be predicted by
the model (Eq. (8) below).
As an extra test, a Theils inequality coefcient (TIC) was calculated (Eq. (6)). When TIC is lower than 0.3, a good agreement can
be found between the calculated and measured data [1820].

P i
2
i
i ycalc  yexp
TIC q
q
P i2
P i2
i ycalc
i yexp

As such, different methods are used to demonstrate the agreement between the experimental data and the empirical model.
3. Results and discussion
3.1. Determination of kLaCW and kLaAS
In Tables 4 and 5 the resulting kLa values determined in clean
water kLaCW and in activated sludge (kLaAS) are given these values
were further used for dimensionless analysis and based on this
analysis the most important design and operational settings are
discussed below.
Al-Ahmady [12] reported a kLaCW range of 2.2106 l/h for a rectangular batch reactor, based on his work and literature. Gillot et al.
[6] mentioned a kLaCW range of 313 l/h. Both ranges are comparatively higher than the 0.510 1/h range found in this research.
Causes are mainly due to differences in geometrical and hydrodynamical parameters. E.g. a smaller Ad/At ratio of 0.040.1 was used,
compared to those mentioned in Al-Ahmady [12] and Gillot et al.

[6] (resp. a range of 0.10.9 and 0.040.14). The bubble size db of

0.005 m is larger than the 0.00250.011 m of Al-Ahmady [12].
Smaller bubbles are known for enhancing the oxygen transfer.
The largest difference is found in the airow rate Qa. The airow
rate of Al-Ahmady [12] (43312,284 l/min) and Gillot et al. [6]
(2551230 l/min) are considerably higher than the airow rate in
this paper: 0.240.60 l/min. All these variables have a large impact
on kLa, as discussed below.
3.2. Dimensional analysis
Table 2 summarizes the design and operating parameters characteristic for the diffused aerated cylindrical tanks, while Tables 4
and 5 give the actual experimental values. Mass transfer correlations can be easily made with a dimensional analysis. The method
is thoroughly explained in Cussler [21]. The effects of the liquid
lm coefcient kL and the interfacial area a on the oxygen concentration are determined together, being the volumetric oxygen mass
transfer kLa. The operational parameters inuencing kLa (Table 2)
are shown in Eq. (7).

kL a f V t ; Ht ; Dt ; At ; Q a ; Ad ; db ; hd ; q; g; D; g

In this contribution the liquid viscosity (l) has limited inuence

on the oxygen transfer and is seen as a constant because of the low
shear rate of the bubbles. Liquids with high viscosity require a different dimensional analysis. The transport of oxygen is dissimilar
compared to low viscous systems [8]. The operational parameters
Ad, db, q and D are also seen as constants, as they have a very small
operational range (Table 2).
Grouping the geometrical and dynamical variables from Eq. (7),
according to popular dimensionless parameters for ASPs, a dimensionless function can be created following the p-theorem Eq. (8).

0
1bs

b
 b  b  b  b
D2t kL aCW
Q a q 2 B Q a C db 4 Ht 5 Ad 6 Dt 7
b1
@qA
D
Dt g
hd
Dt
At
hd
D5t g
!b 9
 b8
Ht
Vt

8
hd
A1:5
t
This function is an adapted version of a dimensionless equation
used for rectangular tanks [12]. Compared to the research performed on rectangular tanks, the length of the tank in this research
was substituted by the diameter and three additional dimensionless numbers were implemented and investigated: Dt/hd, Ht/hd

Table 5
Overview of the combinations of operational variables used during the experiments with activated sludge and resulting kLaAS.

lAS

Ht
m

Dt
m

Qa
l/min

hd
m

At
m2

V
l

Tas
C

kg/ms

TSS
mg/l

kLaAS
1/h

0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.53
0.53
0.53
0.53
0.53
0.53
0.53
0.53
0.53

0.10
0.10
0.10
0.10
0.10
0.10
0.10
0.10
0.10
0.15
0.15
0.15
0.15
0.15
0.15
0.15
0.15
0.15

0.60
0.40
0.24
0.60
0.40
0.24
0.60
0.40
0.24
0.60
0.40
0.24
0.60
0.40
0.24
0.40
0.24
0.60

0.02
0.02
0.02
0.08
0.08
0.08
0.17
0.17
0.17
0.03
0.03
0.03
0.11
0.11
0.11
0.24
0.24
0.03

7.4  103
7.4  103
7.4  103
7.4  103
7.4  103
7.4  103
7.4  103
7.4  103
7.4  103
1.8  102
1.8  102
1.8  102
1.8  102
1.8  102
1.8  102
1.8  102
1.8  102
1.8  102

2.7
2.7
2.7
2.7
2.7
2.7
2.7
2.7
2.7
9.3
9.3
9.3
9.3
9.3
9.3
9.3
9.3
9.3

17.0
16.5
17.2
17.0
16.6
16.8
17.0
16.5
17.2
16.7
17.1
17.7
16.4
17.7
17.3
18.0
18.0
17.1

1.9  103
1.9  103
1.9  103
1.9  103
1.9  103
1.9  103
1.9  103
1.9  103
1.9  103
1.9  103
1.9  103
1.9  103
1.5  103
1.9  103
1.9  103
1.9  103
1.9  103
1.4  103

2525
2510
2209
2540
2257
2174
2418
2507
2182
2447
2354
2421
1010
2147
2270
2404
2543
918

6.35
4.10
2.69
6.74
3.18
1.56
4.88
2.87
1.47
2.18
1.02
0.56
3.17
1.44
0.31
1.16
0.76
3.46

56

E. Pittoors et al. / Chemical Engineering Journal 243 (2014) 5159

(rewritten) equation and corresponding parameters values (b1

through b9) were obtained for clean water.

!
0:03D

kL aCW

D2t

Fig. 4. Comparison between experimental and calculated values for the oxygen
mass transfer in clean water (kLaCW).

the hydrodynamic inuences of the system, represented respec


tively by the Reynolds Re mmL QDagq and the (adapted) Froude


a
number Fr pm pQ
. The last 7 dimensionless numbers on
5
Lg

Dt g

the right side represent the geometrical inuences on kLa caused

by the tank and diffusers.
3.3. Prediction of the oxygen mass transfer based on the dimensional
analysis
The data represented in Table 4 were used to estimate the
parameters in Eq. (8) for the test with clean water. The following

Fig. 5. Comparison between experimental and calculated values for the oxygen
mass transfer in activated sludge (kLaAS).

 0:291  0:554  0:135

db
Ht
Ad
hd
Dt
At

!0:017
 0:086
Dt
Vt
hd
A1:5
d

In Fig. 4 the resulting t between the experimental and calculated values for clean water is demonstrated. A R2 of 0.95 can be
found for the model of kLaCW (Eq. (9)), meaning that 95% of the variation of the independent tested variables can be explained by the
change of the dependent variable kLa [17]. The t-test demonstrated
that the null hypothesis can be accepted as the statistic t-value of
0.32 is closer to zero than the critical t-value of 1.7 (28 degrees of
freedom). As such it can be concluded that kLaCW can be accurately
predicted by Eq. (9). An average relative difference smaller than 1%
was noticed between the experimental and calculated kLaCW.
The data represented in Table 5 were used to estimate the
parameters in Eq. (8) for the test with activated sludge. The following (rewritten) equation and corresponding parameters values (b1
through b9) were obtained for activated sludge.

and V t =A1:5
d . Other dimensionless combinations of the operational
variables were also tested (e.g. Q 2a /VtD2) but proved to have no signicant impact on both kLaCW and kLaAS. As such, the same model
will be used for clean water and activated sludge, but with different coefcients. The left part of Eq. (8) shows the impact of the tank
diameter and kLa, represented by the dimensionless Sherwood


D2 k a
number Sh t DL CW . The rst two terms of right part stand for

Re1:718 Fr0:709

kL aAS

!
0:06D

Re1:906 Fr 0:631

 0:23  0:120  0:326

db
Ht
Ad
hd
Dt
At
!0:01
Vt

D2t
 0:164  0:173
Dt
Ht

hd
hd
A1:5
d

10

In Fig. 5 the resulting t between the experimental and calculated values for activated sludge is demonstrated A R2 of 0.96
was obtained for the kLaAS-model (Eq. (10)). An average relative
difference of 5% was noticed between the experimental and calculated kLaAS. kLaAS can be accurately predicted by Eq. (10), as the statistic t-value of 0.07 is closer to zero than the critical t-value of
1.74 (17 degrees of freedom). The empirical model leads to slight
deviation in measured and calculated values due to lack of all the
inuencing variables (e.g. there are no parameters for TSS, aerated
and diffusers perforated area in the model).
The TIC-values for both kLaCW and kLaAS are below 0.3 (both
0.06), showing a good agreement between measured and calculated data. Both models for clean water and activated sludge tanks
show that it is better to construct a tank with an increased diameter or smaller height and increased diffuser submergence for a
higher oxygen transfer, as some negative exponents demonstrate
in Eqs. (9) and (10).
When comparing the correlations that were obtained in other
studies, no good results are obtained. The model by Al Ahmady

Fig. 6. Coefcients values of the kLaCW and kLaAS models (Eq. (9) and (10)).

E. Pittoors et al. / Chemical Engineering Journal 243 (2014) 5159

[12] results in calculated kLaAS values that are between 0.03 1/h
and 0.5 1/h and are as such an order of magnitude lower than
the experimental values obtained in this study. The model of Schierholz et al. [10] results in calculated kLa values that are above
100 1/h and are as such too high. As such it can be concluded that
a suitable model was developed for calculating the volumetric oxygen mass transfer kLa in an ASP under varying operational
conditions.
Further, the he impact of activated sludge on aeration performance can be quantied by calculating the a-factor (ratio of kLaAS
and kLaCW). Formerly many plants were designed with an a-factor
of 0.8, which was considered as a universal a for all types of
aeration systems [8,9]. Currently full-scale eld measurements
suggest that the a-value should be in the range of 0.20.8 [22].
The average a-factor for the experiments presents in this study
is 0.78.
3.4. Effect of design and operational conditions
Fig. 6 summarizes the coefcients for the kLaAS and kLaCW models. The ratio Ad/At has a positive coefcient (resp. 0.326 and 0.135)
for both activated sludge and clean water. These results conrm
those obtained by Chavan and Mukherji [23] (rotating biological
contactor), Gillot et al. [6] (cylindrical tank) and Al-Ahmady [12]
(rectangular tank). Increasing the diffusers area will lead to increased oxygen transfer as this aids bubble distribution and improves contact between the gas and liquid interfaces [24].
Activated sludge is more sensitive to the diffusers area since activated sludge solutions are harder to mix and consume a lot more
oxygen compared to clean water. Increasing Ad ameliorates the distribution of oxygen, explaining the higher coefcient for the
dimensionless parameter.
The negative coefcients for the ratio db/hd for both activated
sludge (0.23) and clean water (0.291) demonstrate that increasing diffusers depth leads to improved oxygen transfer considering
the increasing bubbles pathway and contact between the phases
[14]. Decreased bubble size enhances bubble stability and the contact surface area between the gas and liquid phase on account of
increased interfacial area a [8,12,14,25]. Fine bubbles have a
db < 5 mm [8,26]. Note that increasing the airow rate Qa leads to
a linear increase in bubble size [26]. Chern and Yang [27] and Schierholz et al [10] concluded that the hydrostatic pressure increases
with increasing water depth, stimulating oxygen transfer and
thereby kLa. In view of the negative coefcient, there is an optimal
ratio of tank height and depth.
Previous studies [12,28] have shown that increasing the tank
height (Ht) will increase the bubbles pathways prior to the surface
and increase contact time between the phases. This is conrmed by

Fig. 7. Impact of varying tank diameter Dt on kLaCW and kLaAS.

57

Fig. 8. Impact of varying airow rate Qa on kLaCW and kLaAS.

the positive coefcients for the ratio Ht/hd (0.086 for clean water
and 0.173 for activated sludge) in this research.
The coefcients of the ratio Dt/hd (0.321 for clean water and
0.164 for activated sludge) are in agreement with the research
from Gillot et al. [6]. Again reactors with activated sludge are more
sensitive to these effects, seen the higher coefcient. The value of
Dt stands in close relation with Ad as often the entire oor of diffused tanks are lled. Increasing the diameter of the tank or the
depth of the diffusers will lead to a better contact between the biomass or clean water and oxygen coming from the diffusers. If the
activated sludge is mixed more efciently then the biomass cannot
settle at the bottom and oxygen can more easily be absorbed leading to improved kLaAS. In this research increasing Dt has a negative
impact on kLa because Ad remained constant throughout the entire
research, leading to an inefcient bubble distribution and contact
between the gas and liquid phase (Fig. 7).
Many researchers such as Casey [13] and Chavan and Mukherji
[23] noticed to the positive effect on oxygen transfer of increased
working volumes (V), since it increases the hydrostatic pressure
because of the higher gas hold-up. This is however not demonstrated this research as a negative coefcient for V=Ad1:5 is obtained
for clean water (0.017) and activated sludge (0.01). However,
the contribution of V=A1:5
is rather insignicant as the respective
d
b-coefcient is lower than 0.05, which is the cut-off value for an
insignicant coefcient [6].
The largest impact on the model is from the Reynolds number
with a coefcient of 1.718 (clean water) and 1.906 (activated
sludge). The hydrodynamic parameter airow rate Qa is known to
have a quantitative positive impact on oxygen transfer (Fig. 8)
and can easily be changed to improve the system performance, corroborating several studies which have established these positive
effects [10,12,13,24,25,2830]. The cause is a complex interaction
between the interfacial area a, air-hold up, driving force Eq. (1)
and bubble density. Galaction et al. [31] validated that the positive
impact of the airow rate is decreased with increasing biomass
concentration, similar to this research. The range of kLaAS is significantly lower than kLaCW. Even so, increasing the airow rate to
excessive rates leads to low driving force and bubble residence
time and therefor worsens oxygen transfer [32].
Although dynamic viscosity l is presumed constant in this paper, it still has a large impact on the oxygen transfer rate compared
to the clean water tests and is closely connected to the biomass
concentration (see differences kLaAS and kLaCW in gures above).
The maximum Reynolds number for activated sludge is quit lower
than the one for clean water tests (Table 2). This result is in agreement with the trends recorded by other researchers, although none
have researched the difference between clean water and AS
models. Summarized, biomass and viscosity have the following
impact on the oxygen transfer. Biomass increase leads to an

58

E. Pittoors et al. / Chemical Engineering Journal 243 (2014) 5159

exponential decrease in kLa [2] as it hinders oxygen uxes due to

microorganisms adsorbing at the bubble surface, reinforces buildup of extracellular material, inhibits and disrupts bubble transport,
turbulence and the dispersion-coalescence activity [31]. Viscosity
is directly proportional to biomass which on his turn inuences
the rheological properties and thereby oxygen transfer
[1,3,31,3336]. At excessive sludge concentrations (>17.9 g/l) the
positive effect of airow rate gets altered [2]. The sludge concentration can easily be changed by regulating sludge wastage,
although this will also affect the performance of the waste water
treatment plant.

3.5. Scale-up considerations

In this contribution a general equation was developed for calculating the kLa of activated sludge tanks on lab-scale. Extrapolation
to full-scale should be possible although perhaps not straightforward as it might be necessary to re-estimate the empirical coefcients. This can be done by the procedure described above using
larger scale test reactors. Nevertheless, the developed equations
can be used as a rst estimate of the kLa at full-scale. This can be
illustrated by a using a WWTP design based on Qasim [37]. If an
airow rate (Qa) of 1 m3/s is supplied to a cylindrical reactor of
10,000 m3 (Vt) with a height of 5 m (Ht) and a diameter of 50 m
(Dt). Further, it can be assuming that 5% of the total area is covered
with diffusers (compared to 10% in the experiments performed in
this study), that the diffusers are at 4.5 m depth (hd) and that the
sludge has a dynamic viscosity of 2  103 kg/ms. When all other
operational variables (bubble size (db), water density (q) and oxygen diffusion coefcient (D)) are assumed the same as in this study,
a kLa value of 23 1/h or 550 1/d is obtained. This value is at the high
end of values used in activated sludge simulation [3840], but is
still realistic. This demonstrates the ability to use the developed
equations as a rst estimate on full-scale.

4. Conclusions
Up until now no research has developed a model specically
based and tested on ASPs in the presence of biomass, taking into
account geometrical, hydrodynamical and physicochemical
parameters and then compared it to a similarly tested clean water
model.
The inuence of 12 different parameters were researched;
cylindrical tank volume (Vt), height (Ht), diameter (Dt) and surface
area (At), airow rate (Qa), diffusers surface area (Ad) and depth (hd)
and bubble size (db), water density (q), dynamic viscosity (l),
oxygen diffusion coefcient (D) and the gravitational (g). A
relation between the measured volumetric oxygen mass transfer
coefcient kLa in a cylindrical batch reactor and twelve operational
variables is developed for activated sludge (kLaAS) and clean water
(kLaCW).
Airow rate was the main parameter affecting the kLa for both
clean water and the activated sludge reactor. Next were the bubble
size and diffusers depth. Airow rate and diffusers area had a signicantly bigger impact in the presence of biomass, since it promotes bubbles distribution, mixing of the solution and an
improved oxygen transfer.
There are differences in coefcients of up to 66% between kLaAS
and kLaCW for the same operational conditions but generally the
same trends and effects of different hydrodynamic, physicochemical and geometrical parameters is seen. An average a-factor for the
experiments presents in this study of 0.78 was calculated. This paper showed an easy and quick procedure for measuring different
parameters and developing AS models via a dimensional analysis.

Acknowledgements
Special thanks are extended to Fu XianWei for his help and
hints during the research. The author acknowledges the nancial
support obtained through the scholarship of Zhejiang Province
and by the Key Project of Science and Technology of Zhejiang Province (Grant No. 2011C14023), the National Natural Science Foundation of China (Grant No. 51108416) and the International
Cooperation Project of Ministry of Science and Technology of China
(Grant No. 2010DFA92050).

References
[1] F.A. Rodriguez, P. Reboleiro- Rivas, F. Osorio, M.V. Martinez-Toledo, E.
Hontoria, J.M. Poyato, Inuence of mixed liquid suspended solids and
hydraulic retention time on oxygen transfer efciency and viscosity in a
submerged membrane bioreactor using pure oxygen to supply aerobic
conditions, Biochem. Eng. J. 60 (2012) 135141.
[2] E. Germain, F. Nelles, A. Drews, P. Pearce, M. Kraume, E. Reid, S. Judd, T.
Stephenson, Biomass effects on oxygen transfer in membrane bioreactors,
Water Res. 41 (2007) 10381044.
[3] T. Jiang, G. Sin, H. Spanjers, I. Nopens, M.D. Kennedy, W. van der Meer, H.
Futselaar, G. Amy, P.A. Vanrolleghem, Comparison of the modeling approach
between membrane bioreactor and conventional activated sludge processes,
Water Environ. Res. 81 (2009) 432440.
[4] American Society of Civil Engineers (ASCE), Standard Guidelines for in-Process
Oxygen Transfer Testing, 1996.
[5] Association Francaise de Normalisation (NFEN), Wastewater treatment Plants
Part 15: Measurements of the Oxygen Transfer in Clean Water in Aeration
Tanks of Activated Sludge Plants, NFEN-12255 15, 2004.
[6] S. Gillot, S. Capela-Marsal, M. Roustan, A. Heduit, Predicting oxygen transfer of
ne bubble diffused aeration systems: model issued from dimensional
analysis, Water Res. 39 (2005) 13791387.
[7] J. Mueller, W. Boyle, J. Popel, Aeration: Principles and Practice, Volume 11. CRC
Press, 2002.
[8] D. Rosso, M. Stenstrom, Surfactant effects on alpha factors in aeration systems.
Water Res. 40 (2006) 13971404 (10).
[9] M.K. Stenstrom, R. Gilbert, Effects of alpha, beta and theta factor upon the
design, specication and operation of aeration systems, Water Res. 15 (1981)
643654.
[10] E. Schierholz, J.S. Gulliver, S.C. Wilhelms, H.E. Henneman, Gas transfer from air
diffusers, Water Res. 40 (2006) 10181026.
[11] G. Tchobanoglous, F. Burton, D. Stensel, Wastewater Engineering Treatment
and Reuse, fourth ed., Metcalf & Eddy, Inc., 2003.
[12] K.K. Al Ahmady, Mathematical model for calculating oxygen mass
transfer coefcient in diffused air systems, Al Rafdain Eng. J. 19
(2011) 4354.
[13] T. Casey, Diffused Air Aeration Systems for the Activated Sludge Process,
Technical Report, Aquaverrra Research Publications, 2009.
[14] Y. Fayolle, S. Gillot, A. Cockx, L. Bensimhon, M. Roustanb, A. Heduit, In situ
characterization of local hydrodynamic parameters in closed loop aeration
tanks, Chem. Eng. J. 158 (2010) 207212.
[15] APHA, Standard Methods for the Examination of Water and Wastewater, 20th
ed., American Public Health Association, 1999.
[16] A. Damayanti, Z. Ujang, M. Salim, G. Olsson, A. Sulaiman, Respirometric
analysis of activated sludge models from palm oil mill efuent, Bioresour.
Technol. 101 (2010) 144149.
[17] A.M. Brown, A step-by-step guide to non-linear regression analysis of
experimental data using a Microsoft Excel spreadsheet, Comput. Methods
Programs Biomed. 65 (2001) 191200.
[18] S. Van Hulle, J. Verstraete, J. Hogie, P. Dejans, A. Dumoulin, Modelling and
simulation of a nitrication biolter for drinking water purication, Water SA
32 (2006) 257264.
[19] W.T.M. Audenaert, M. Callewaert, I. Nopens, J. Cromphout, R. Vanhoucke, A.
Dumoulin, P. Dejans, S.W.H. Van Hulle, Full-scale modelling of an ozone
reactor for drinking water treatment, Chem. Eng. J. 157 (2010) 551557.
[20] A. Vandekerckhove, W. Moerman, S.W.H. Van Hulle, Full-scale modelling of a
food industry wastewater treatment plant in view of process upgrade, Chem.
Eng. J. 135 (2008) 185194.
[21] E. Cussler, Diffusion: Mass Transfer in Fluid Systems, Cambridge University
Press, 2003.
[22] D. Rosso, M.K. Stenstrom, Alpha factors in full scale wastewater aeration
systems, in: Proceedings WEFTEC, 2006, pp. 48534863.
[23] A. Chavan, S. Mukherji, Dimensional analysis for modeling oxygen
transfer in rotating biological contactor, Bioresour. Technol. 99 (2008)
37213728.
[24] K.K. Al Ahmady, Analysis of oxygen transfer performance on subsurface
aeration systems, Int. J. Environ. Res. Public Health 3 (2006) 301308.
[25] S. Vermande, K. Simpson, K. Essemiani, C. Fonade, J. Meinhold, Impact of
agitation and aeration on hydraulics and oxygen transfer in an aeration ditch:
local and global measurements, Chem. Eng. Sci. 62 (2007) 25452555.

E. Pittoors et al. / Chemical Engineering Journal 243 (2014) 5159

[26] D.F. McGinnis, J.C. Little, Predicting diffused bubble oxygen transfer rate using
the discrete bubble model, Water Res. 36 (2002). 4627-1635.
[27] J. Chern, S. Yang, Oxygen transfer rate in a coarse bubble diffused aeration
system, Ind. Eng. Chem. Res. 42 (2003) 66536660.
[28] K.K. Al Ahmady, Effect of airow rate and submergence of diffusers on oxygen
transfer capacity of diffused aeration systems, Al Rafdain Eng. J. 14 (2006) 27
38.
[29] M. Oliveira, A. Franca, Simulation of oxygen mass transfer in aeration systems,
Int. Commun. Heat Mass Transfer 25 (1998) 853862.
[30] M.H. Kim, J.F. Kragi, R.G. Thompson, Oxygen mass transfer coefcient (kla)
consideration for scale-up fermentation systems at the Biotechnology
Laboratory U.S. Army Edgewood Chemical Biological Center, Technical
Report, Research and technology directorate, 2011.
[31] A. Galaction, D. Cascaval, S. Camarut, Analysis of distribution of oxygen
transfer rate in stirred bioreactors for yeasts broths, Romanian Biotechnol.
Lett. 15 (2010) 54235435.
[32] A. Thunberg, A. Sundin, B. Carlsson, Energy optimization of the aeration
process at Kappala wastewater treatment plant, in: Proceedings 10th IWA
Conference on Instrumentation, Control & Automation, 2009.
[33] F. Garcia-Ochoa, E.G. Castro, V. Santos, Oxygen transfer and uptake rates
during xanthan gum production, Enzyme Microbial Technol. 27 (2000) 680
690.

59

[34] H. Djelal, F. Larher, G. Martin, A. Amrane, Effect of the dissolved oxygen on the
bioproduction of glycerol and ethanol byhansenula anomala growing under
salt stress conditions, J. Biotechnol. 125 (2006) 95103.
[35] P.F. Amaral, M.G. Freire, M.H.M. RochaLeao, I.M. Marrucho, J.A. Coutinho,
M.A.Z. Coelho, Optimization of oxygen mass transfer in a multiphase
bioreactor with peruorodecalin as a second liquid phase, Biotechnol.
Bioeng. 99 (2008) 588598.
[36] M. Bayramoglu, A. Cakici, T. Tekin, Modelling of oxygen transfer rate in
diffused air aeration tanks, Process Saf. Environ. Prot. 78 (2000) 209212.
[37] S.R. Qasim, Wastewater Treatment Plants: Planning, Design and Operation,
second ed., CRC Press, Boca Raton, 1999.
[38] Corominas. Ll, G. Sin, S. Puig, Balaguer M. Dolors, P.A. Vanrolleghem, J. Colprim,
Modied calibration protocol evaluated in a model-based testing of SBR
exibility, Bioprocess Biosyst. Eng. 34 (2011) 205214.
[39] I. Nopens, L. Benedetti, U. Jeppsson, M.-N. Pons, J. Alex, J.B. Copp, K.V. Gernaey,
C. Rosen, J.-P. Steyer, P.A. Vanrolleghem, Benchmark simulation model No 2:
nalisation of plant layout and default control strategy, Water. Sci. Technol. 62
(9) (2010) 19671974.
[40] L. Guo, J. Porro, K.R. Sharma, Y. Amerlinck, L. Benedetti, I. Nopens, A. Shaw,
S.W.H. Van Hulle, Z. Yuan, P.A. Vanrolleghem, Towards a benchmarking tool
for minimizing wastewater utility greenhouse gas footprints, Water. Sci.
Technol. 66 (11) (2012) 24832495.