Chemical Engineering Science 188 (2018) 208–220

Contents lists available at ScienceDirect

Chemical Engineering Science

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

Hydrodynamic inhomogeneities in large scale stirred tanks – Influence

on mixing time
A. Rosseburg a,⇑, J. Fitschen a, J. Wutz b, T. Wucherpfennig b, M. Schlüter a
a
Institute of Multiphase Flows, Hamburg University of Technology, Hamburg, Germany
b
Late Stage USP Development, Bioprocess Development Biologicals, Boehringer Ingelheim Pharma GmbH & Co. KG, Germany

h i g h l i g h t s

Optical measurements of mixing time in a 15 000L acrylic glass reactor.

Bio processes in large scale operating in transition regime of loading/flooding.

Visualizations of two phase flow inhomogeneity by optical measurement method.

Global flow structure is dominated by impeller or buoyancy driven flow.

Mixing behaviour can be predicted by a modified correlation from literature.

a r t i c l e i n f o a b s t r a c t

Article history: Aerated stirred tank reactors are widely used in chemical industry and bioprocess engineering. One major
Received 22 December 2017 parameter to characterize the heat- and mass transfer performance of such aerated stirred tank reactors
Received in revised form 30 April 2018 is the mixing time necessary to homogenize the reactor volume. Despite its importance, the prediction of
Accepted 5 May 2018
the mixing time is still challenging due to the complex hydrodynamic inhomogeneities induced by the
Available online 10 May 2018
gaseous phase, which becomes strongly apparent in large scale systems. A precise measurement of the
two phase flow on the other hand requires a volumetric insight and is thus difficult to realize. To over-
Keywords:
come this problem a transparent stirred tank reactor on industrial scale has been erected at the Hamburg
Mixing time
Stirred tank reactor
University of Technology in cooperation with Boehringer Ingelheim Pharma GmbH & Co.KG. With the
Buoyancy driven flow decolouration method the temporal and spatial development of mixing can be taken into account. The
Flow pattern results indicate the importance of the local inhomogeneities on large scale. A first characterization can
be done by taking into account buoyancy driven flows superimposing the flow imposed by the impeller.
A correlation is presented to estimate the transition between a loading and flooding regime on large scale.
This correlation enables the calculation of mixing times for a wide range of stirrer frequencies and super-
ficial gas velocities. Furthermore, this publication emphasizes the challenges of scale-up on the basis of
laboratory experiments in small scale.
Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction the investigations have changed from global measurement in the

For many applications in chemical and biochemical engineer-
ing, efficient mixing of two phase flows with high heat- and mass
transfer performance is one of the most important challenges
(Middleton and Smith, 2003). For this purpose, aerated stirred tank
reactors are the most widely used apparatuses in industry due to
their simplicity, low investment costs and flexibility. Because of
their importance they have been investigated intensively in the
past decades to model and simulate these processes. Over the years
⇑ Corresponding author.
E-mail address: annika.rosseburg@tuhh.de (A. Rosseburg).
early days to more local experimental analysis in the last decades.
Whereas the mixing time, global flow structure and gas hold-up
has been analyzed first by overall measurements (e.g. Zlokarnik,
1967; Warmoeskerken and Smith, 1985; Haß and Nienow, 1989;
Nienow, 1998; Bouaifi et al., 2001; Kong et al., 2012) in the last
years more local measurements have been performed to determine
the bubble size distribution (Montante et al., 2008; Laakkonen
et al., 2005), the local gas hold-up (Busciglio et al., 2013; Kong
et al., 2012; Lee and Dudukovic, 2014) and the liquid and gas veloc-
ities (Chara et al., 2016; Montante et al., 2007). Yet reliable design
and scale-up is still challenging and a lot of open questions remain.
One reason is the limited amount of data about mixing

https://doi.org/10.1016/j.ces.2018.05.008
0009-2509/Ó 2018 Elsevier Ltd. All rights reserved.
 A. Rosseburg et al. / Chemical Engineering Science 188 (2018) 208–220 209

Nomenclature

List of symbols Vfill reactor volume, m3

A, a constants, – w0G superficial gas velocity, m s1
B, b constants, –
C, c constants, – Greek symbol
d impeller diameter, m eG gas hold-up, –
D tank diameter, m eT energy dissipation rate, W kg1
g acceleration due to gravity, m s1 q density, kg m3
Hliq liquid height, m m kinematic viscosity, m2 s1
H0 unaerated liquid height, m
HG aerated liquid height, m List of dimensionless relations
n stirrer frequency, s1 Fl flow number q 3
nS number of impeller, – nd

P0 unaerated power input, W Flc flow number at transition between flooding an loading
PG aerated power input, W FlP dimensionless pumping capacity qP3
2 nd
q gas flow rate, m3 s1 Fr Froude number ngd
qP pumping capacity, m3 s1 Po Power number P
qn3 d5
tmix mixing time, s 2
Re Reynolds number nd
m
tC circulation time, s

characteristics, especially in large scale reactors. The most effective

way to obtain a deeper insight into the complex interaction of local
flow pattern and mixing is the optical visualization on small and
large scales simultaneously with high spatial and temporal resolu-
tion. However, such optical investigation is difficult to realize due
to the lack of optical access in large scale applications (see Fig. 1).
To overcome the obstacle a large-scale acrylic glass reactor
with a volume of VR = 15 m3 and total optical access has been
erected at the Hamburg University of Technology together with
the Boeringer Ingelheim Pharma GmbH & Co.KG. This article gives
first insights into the global and local mixing performance as well
as local and global flow pattern within an aerated stirred tank
reactor of industrial scale and shows ways to a more reliable
design and scale-up.

2. Theory

Despite the fact that the complex two phase flow behavior of
aerated stirred tank reactors has been investigated intensively in
the last decades, there is still a lack of knowledge especially for
large scale applications (Nienow, 1998; Warmoeskerken and
Smith, 1985; Zlokarnik, 1967). For instance, during the process of
fermentation, fluids are added at regular intervals to the reactor
to regulate the pH value or to provide nutrients for the cells. In this
case a good mixing behavior is required to minimize local concen-
tration gradients (Warmoeskerken and Smith, 1985). Thus, as a
crucial parameter for the characterization of stirred tank reactors,
the mixing time is used that is necessary to achieve a certain
degree of homogenization after the addition of a tracer pulse into
the reactor (Ascanio, 2015; Nienow, 1997). Two principal measure-
ment techniques are common to determine the mixing time: The
first method is a local measurement where a fixed probe is used
that measures the concentration of a tracer which is added to the
reactor. For instance, conductivity probes are used to measure
the ion concentration of a salt that is mixed into an aqueous phase.
When the signal reaches a certain threshold of the end value of
conductivity, the mixing is defined as completed (Nienow, 1997).
Fig. 2 shows the typical signal for a homogenization experiment
with a conductivity probe. From such a graph, the defined circula-
tion time tc as well as the mixing time tmix can be determined. The
disadvantage of this method is that the mixing time can only be Fig. 1. 15kL acrylic glass reactor at TUHH.
210 A. Rosseburg et al. / Chemical Engineering Science 188 (2018) 208–220
Fig. 2. Mixing time tmix and circulation time tc from a transient response after a
pulse of tracer injection (Nienow, 1997).
measured locally so that no dead zones or flow structures can be
detected.
The second method is the decolouration of a dissolved dye, for
instance phenolphthalein. When a base is added to the aqueous
phenolphthalein solution, the shift in the pH value will change
its color from transparent to pink. By adding an acid afterwards,
the pH value is shifted backwards and the medium turns to color-
less again. The mixing time is defined as the time until the last
color within the reactor disappears. With this method, the global
flow structure as well as dead zones can be visualized and mea-
sured quantitatively. But a total optical access to the reactor is
needed which is often difficult to realize for reactors on industrial
scale. In addition to that, concentration gradients and local over-
shoots of the base and acid concentration cannot be detected with
this method (Nienow, 1997).
2.1. Bioreactors in single phase operation
To calculate or model the mixing time, a deep understanding of
the liquid flow pattern is required. The flow pattern is mainly influ-
enced by the impeller type as well as the gas flow rate. A lot of
investigations have been done to characterize the flow field of sin-
gle phase systems for various impeller and reactor types. The main
characterization is done by the Power number Po
P nolds number for three single stirrer types are shown. Zlokarnik
Po ¼
qn3 d5
which is calculated with the power input P, the density q, the stirrer
frequency n and the stirrer diameter d.
For many stirrer combinations, the Power number has already
been determined as a function of the Reynolds number
Fig. 3. Flow pattern of (a) a vertical down pumping pitched blade turbine with a dimensionless pumping capacity of Flp = 0.23 and (b) a radial pumping Rushton turbine with
a dimensionless pumping capacity of Flp = 0.24 (Nienow, 1998).
2
nd
Re ¼ ð2Þ
m
that takes into account the hydrodynamic conditions by means of
the stirrer frequency n, the stirrer diameter d and the kinematic vis-
cosity of the liquid phase m. For low Reynolds numbers the Power
number decreases with increasing Reynolds number. For high Rey-
nolds number Re > 105 the Power number can be assumed to be
constant (Zlokarnik, 2003).
Beside the Reynolds number, the stirrer frequency n can be
taken into account with the dimensionless Froude number accord-
ing to
n2  d
Fr ¼ ð3Þ
g
with the gravity constant g.
A further characterization of the hydrodynamics of stirred tank
reactors is done with the dimensionless pumping capacity Flp
qP
Flp ¼ 3
ð4Þ
nd
with the pumping capacity qP, that refers to the liquid flow that is
pumped by the impeller (see Fig. 3). The pumping capacity for dif-
ferent stirrer types has been investigated and presented by several
authors (Nienow, 1998; Wu and Patterson, 1989; Nienow, 1997;
Jaworski et al., 1996). It has been shown that the pumping capacity
is almost constant for high Reynolds numbers (Nienow, 1997) and
can assumed to be FlP = 0.23 for a down pumping pitched blade
and FlP = 0.24 for a Rushton turbine.
The most common stirrer types in bio processes are the Rushton
and the pitched blade turbine. When a gaseous phase is present a
Rushton turbine is often used due to its good dispersion ability
(Vrábel et al., 2000). On the other hand, the Rushton turbine is
known for the formation of large vortices with more or less closed
areas of back mixing (compartments) that are leading to an axial
flow barrier and thus to a limitation in homogenization (Vrábel
et al., 2000). Therefore, to reduce the flow barrier and enhance
the axial mixing, an axial down-pumping impeller is often used,
for instance a pitched blade turbine.
Zlokarnik investigated the mixing time t mix for single stirrers
intensively by using the dimensionless mixing time t mix  n
(Zlokarnik, 1967). In Fig. 4 the trends in dependency of the Rey-
ð1Þ divided the mixing characteristic into three different areas and
showed for the blade stirrer (d/D = 0.33; H/D = 1) that the dimen-
sionless mixing time is proportional to Re0.7 for 101 < Re < 103.
In the range of 103 < Re < 104, the dimensionless mixing time can
be assumed to be almost constant and for the range of Re > 104
the dimensionless mixing time is proportional to Re0.7. For
 A. Rosseburg et al. / Chemical Engineering Science 188 (2018) 208–220 211

Fig. 4. Mixing time characteristics for Rushton turbines, propeller stirrers and blade stirrers. All stirrers with the ratio d/D = 0.33; H/D = 1 (Zlokarnik, 2003).

Table 1
Overview of relevant mixing time correlations for stirred tank reactors with the unaerated dimensionless mixing time h0 .

Author Formula Comments

Single-phase
Nienow 1990 (Nienow et al., 1996)

d 3 1 (5)
t0mix  n ¼ 3:9 D Flp
Van’t Riet (Nienow et al., 1996)

d 3 1=3 (6)
t0mix n¼3 D Po
Grenville (Nienow et al., 1996)

d 1=3 (7)
t0mix  n ¼ 5:9D3 ðeT Þ2
2 1
D


Vasconcelos et al. (1995) t0mix  n ¼ 2:3 exp 0:68 Dd þ 0:83ns With ns = numbers of impeller (8)
Sieblist et al. (2016)
1
2
A = fitting constant (9)
t0mix  n ¼ e
A Dd  T 3 HD 3 ðns þ 1Þ 2=3

different stirrer types, the transition between these regimes might
change but always three different slopes of the curve can be found.
Zlokarnik stated that measurements in large scale reactors are
indispensable to detect the third zone correctly because for Re 
104 the mixing time becomes almost zero in lab scale reactors.
Further investigations have been conducted for single phase
mixing to develop more precise correlations, especially for multi-
ple impeller systems. Different assumptions have been made, to
derive correlations for the mixing time. Some of them are pre-
sented in Table 1. For instance, Eq. (5) is based on the circulation
time tc and assumes that after five circulation loops the mixing is
already completed. The correlation only depends on the ratio of
d/D and the pumping capacity Flp. Other correlations are based
on the energy dissipation rate eT (Eq. (7)) or are empirical correla-
tions based on the compartment model (Vasconcelos et al., 1995)
(Eq. (8)). All correlations assume that the dimensionless mixing
time is independent of the Reynolds number.
and Smith, 1985). In the loading regime the momentum induced
by the buoyancy driven flow is lower than the momentum induced
by the impeller, thus bubbles are mainly following the fluid phase
(Warmoeskerken and Smith, 1985) (see Fig. 5a). The homogeneous
dispersion of the gaseous phase within the reactor makes this
regime the preferred one in case of heat and mass transfer
(Nienow, 1998; Bombač and Žun, 2006). With increasing gas flow
rate or decreasing stirrer frequency, at some point the impeller is
not able to distribute the gaseous phase sufficiently anymore
(Nienow, 1998; Warmoeskerken and Smith, 1985; Bombač and
Žun, 2006). In this case, illustrated in Fig. 5b, the stirrer blades have
mainly contact with the gaseous phase, less momentum is trans-
ferred into the fluid, and thus the radial flow velocity decreases
(Gagnon et al., 1998; Warmoeskerken and Smith, 1985) which
leads to an undesirable condition (Nienow et al., 2013; Bombač
and Žun, 2006). At this point the momentum induced by the

2.2. Bioreactors in two-phase flow operation

Fermentations in bioreactors mostly take place under aerated

conditions and therefore the buoyancy driven flow induced by
the gaseous phase needs to be taken into account for the calcula-
tion of mixing time. To describe the influence of the gaseous phase,
the gas Flow number
q
Fl ¼ 3
ð10Þ
nd
is used with the aeration rate q.
For the characterization of mixing in aerated stirred tank reac-
tors several authors recommend to distinguish between the ‘‘load-
ing” and the ‘‘flooding” regime (Nienow, 1998; Warmoeskerken Fig. 5. Flow pattern at flooding and loading regime.
212 A. Rosseburg et al. / Chemical Engineering Science 188 (2018) 208–220

buoyancy driven flow exceeds the momentum induced by the

impeller and the main flow pattern is dominated by the upwards
rising gaseous phase (Warmoeskerken and Smith, 1985). In this
regime the power input by the stirrer as well as the gas hold up
decreases and a larger bubble size distribution can be expected.
Consequently the specific surface area decreases and thus to a
decrease in heat- and mass transfer performance can be expected.
To avoid a flooding of the impeller the transition point between
both regimes needs to be known (Warmoeskerken and Smith,
1985; Bombač and Žun, 2006).
Different techniques are presented in literature to determine
the transition from loading to flooding. The easiest one is the opti-
cal observation but it needs an optical access to the reactor and
additionally this procedure is very subjective. Another method is
the measurement of the overall gas hold-up or the power input
(Warmoeskerken and Smith, 1985; Bombač and Žun, 2006). For a
constant stirrer frequency (Froude number), the transition point
can be found if the gas hold-up drops due to flooding with increas-
ing superficial gas velocity (Flow number Fl). This point is called Fig. 6. Different correlations for the critical Flow number Flc as a function of the
‘‘critical Flow number” Flc and can be described as a function of Froude number Fr in gassed stirred tanks.
the Froude number. It has to be mentioned that different tech-
niques can lead to different results for this transition line. Never-
theless, the majority of authors found the critical Flow number
Flc to be a function of the Froude number Fr as well as the geomet-
rical factor D/d according to
 c
D
Fc ¼ a  Fr b ð11Þ
d

were a, b and c are fitted constants (Nienow, 1998; Warmoeskerken

and Smith, 1985; Bombač and Žun, 2006). With Eq. (11), the maxi-
mum gas flow rate at a fixed stirrer frequency can be estimated.
Table 2 gives a short overview of correlations for a Rushton turbine
in water/air. The differences in the correlation become apparent in
Fig. 6 were the results are compared for D/d = 3 and Froude num-
bers of 0.01 < Fr < 10. It can be seen that the correlation by Zlokarnik
(Zlokarnik, 1967) gives 100 times higher critical Flow numbers than
the correlation by Zwieting (Wiedmann, 1983). While Mikulcova,
1967 presented a range within the transition from loading to flood-
ing occurs.
Wiedmann et al. explained these differences in the investigated
Fig. 7. Critical Flow number Fc as function of the Froude number Fr for different
scales and presented results for three different vessels with the
vessel diameters (Bombač et al., 1997).
same geometrical ratios (D/d = 3 and h/d = 0.5) but different reactor
diameters. Their results are presented in Fig. 7 where the critical
Flow number Flc is shown for different vessel dimensions in depen-
dency of the Froude number. It can be seen that the larger vessel ative effect on mixing time. Vasconcelos et al. (1995) showed in
with a diameter of D = 1.5 m reaches higher Flow numbers before more detail that this is only the case when the gas phase reduces
flooding occurs than the smaller vessel with a diameter of D = the total energy input of the stirring. Shewale and Pandit (2006)
0.2 m. But it also becomes clear that the influence of size decreases and Alves and Vasconcelos (1995) also included the flooding
with increasing vessel diameter. regime into the discussion and showed that the aeration can have
Besides the investigation on the flow regime, a lot of investiga- both positive and negative effect on the mixing time depending on
tions have been done during the last decades on the influence of the flow regime within the reactor Shewale and Pandit stated that
the gas flow rate on the mixing time. It can be seen in Tables 3 the gaseous phase is distributed well for a large stirrer power input
and 4 that most of the investigations have been conducted in small and follows the impeller induced flow. Due to the reduction of the
scale systems under loading condition. Nienow (1997) and Saito power input in the presence of a gaseous phase, the flow pattern is
et al. (1992) showed that within this regime the aeration has a neg- weakened which results in an increased mixing time compared to
the ungassed case. On the other hand for a small stirrer power
input, Shewale and Pandit stated that the power induced by the
Table 2 gas phase is larger than the reduction of the power input of the
Correlations for the critical Flow number Flc as function of the Froude number Fr in stirrer. Due to the rise of the bubbles, the mixing between two
gassed stirred tanks.
adjacent compartments is increased which results in a shorter glo-
Author Formula Comment bal mixing time compared to the unaerated system.
Zwietering (Wiedmann, 1983) Flc = (D/d)3,3 (12) Investigations in large scale systems over Vfill = 5 m3 (Table 4)
Dickery (Wiedmann, 1983) Flc = 1.88 Fr 0,5(D/d)3,3 (13) are also available but are mostly conducted for conditions for bac-
Zlokarnik, 1967 Flc = 1.94 Fr0,75 (D/d) = 3.33 (14) terial fermentation with much higher energy input over 100 W/m3
Mikulcova (Wiedmann, 1983) Flc = 100/(6.11 ± 1.24) (15)
and gas flow rates over 1 vvm (Vrábel et al., 2000, 1999). Vrábel
Fr (D/d)3,43
et al. have validated a compartment model with reactors of Vfill
 A. Rosseburg et al. / Chemical Engineering Science 188 (2018) 208–220 213

Table 3
Investigation of the influence of the gas phase on the mixing time in small scale reactors.

Author Volume/L Stirrer Reynolds (P/V)/ Gassing Method Comment

type number Wm3 rate/vvm
Haß and Nienow 150 RT 1:3  105 20–500 0–1.5 Decolorisation Loading; Aeration: Increased mixing
(1989) 6:5  105
Saito et al. (1992) 180 RT 6  104 50–200 0–1 Decolorisation Loading; Aeration: Increased mixing
3  105
Vasconcelos et al. 60–260 2-RT 3  104 10–2000 0–0.6 Local Loading; Aeration: Decreased mixing if energy
(1995) 1  105 measurement input is not decreased
Alves and Vasconcelos 60 3-RT 3  104 50–2000 Local Flooding; Aeration: Decreased mixing
(1995) 1  105 measurement
Nienow et al. (1996) 70 RT 5:4  104  8  104 2–50 0–0.01 Decolorisation Loading; Aeration: Increased mixing
Machon and Jahoda 75 4-RT 2  104 20–500 0.2–0.5 Local Loading and Flooding; Aeration: Both effects
(2000) 6  104 measurement
Shewale and Pandit 70 3-RT 3  104 10–1000 0.3–1.5 Loading and Flooding
(2006) 8  104
Gabelle et al. (2011) 40–350 2RT 4  104 100– 0.4 Only loading; Aeration: Decreased mixing
2  105 1000
Montante and Paglianti 12 RT/PB 1  104 2–250 0.5–2 Local Loading and Flooding; Aeration: Decreased mixing
(2015) 5  104 measurement

Table 4
Investigation of the influence of the gas phase on the mixing time in large scale reactors.

Author Volume/L Stirrer Reynolds (P/V)/W/m3 Gassing rate/ Methode Comment

type number vvm
Nienow et al. (1996) 2000–8000 RT 1:1  105 2.0–50 0–0.01 Local Loading; Aeration: Decreased mixing
3  105 measurement
Vrábel et al. (1999) 22,000 4-RT 5:5  105 80–600 0–0.5 Local Loading; Aeration: No influence
11  105 measurement
Vrábel et al. (2000) 8000 3-RT 7  105 400–1500 0–2 Local Loading; Aeration: No influence
15  105 measurement
Guillard and Trädgårdh 1800– 2-RT to 3  105 80–2000 0–1 Local Loading and Flooding; Aeration: Both
(2003) 22,000 4-RT 14  105 measurement effects

= 8–22 m3 for the loading regime. For gas flow rates in the same
range of animal cell fermentation but much higher energy input
they didn’t find a strong influence of the gas phase on the mixing
time. Nienow et al. (1996) also investigated the mixing time in
large scale system (Vfill = 2–8 m3) with an energy input of (P/V) =
5–50 W/m3 and a gas flow rate up to q = 0.01 vvm. They showed
that the gas phase leads to an improvement of the mixing time
up to 50% and explained it with the buoyancy driven flow of the
gaseous phase.
Due to the complexity of the influence of the gas flow on the
flow pattern, only few correlations are available to describe the
aerated mixing time. Three of them are presented in Table 5.
Vasconcelos and Alves proposed two different correlations (Eqs.
(16) and (17)) for aerated systems which are obtained in an acrylic
glass reactor with a diameter of D = 0.3 m and for multiple impeller
systems. Eq. (16) was determined in the loading regime and
assumes that the influence of the aeration can be subtracted as a
factor from the unaerated dimensionless mixing time t 0mix  n. This
correlation gives the same trend for both, the unaerated and aer-
ated system. Eq. (17) was established in the flooding regime and
is an empirical correlation fitted to measurements. It is only a func-
tion of the factor Fl/Flc and three constants which are not described
in more detail.
From this literature survey it becomes clear, that aerated stirred
tanks have been investigated intensively in the past. Nevertheless,
there are still a lot of open questions especially for large scale
industrial applications concerning the interaction of the turbines
with the gaseous phase, the resulting gas hold-up distribution
and its influence on buoyancy driven flows and mixing. To answer
these open questions, a transparent acrylic glass reactor on indus-
trial scale has been designed and erected by the Institute of Mul-
tiphase Flows together with Boehringer Ingelheim Pharma GmbH
& Co.KG that enables deep insights into aerated stirred processes
by optical investigations.

Table 5
Correlations of dimensionless aerated mixing time for multiple impeller and baffled stirred tank reactors.

Author Formula System

Two-phase
 
Vasconcelos et al (1995) P
tGmix  n P0g ¼ t0mix  n  100½2ns  ðns  nÞFl (16) D = 0.3 m, D = 0.2 m; Water/air
H/D < 3
     
Alves and Vasconcelos (1995) tGmix
P
 n P0g ¼ A þ B exp C FlFlq  1 (17) D = 0.3 m, D = 0.2 m; Water/air
H/D < 3
Shewale and Pandit (2006)
0:127  0:496  1:297 1:756 (18) 0.07 m3 water/air
H eG
 ðn gDÞ
2
tGmix  n ¼ 453 w0G þ PG
Ri0:711
o
D P0
H/D  1
214 A. Rosseburg et al. / Chemical Engineering Science 188 (2018) 208–220

TIR
Gas

TIRC

TIRCH

PI

FIR
M PIR

Fig. 8. Schematic flow chart of the acrylic glass reactor at the Institute of Multiphase Flows.

3. Experimental setup and measurement procedure HG  H0

eG ¼ : ð19Þ
HG
The experiments are carried out in an acrylic glass reactor with
For estimating the mixing time, the decolouration method is
a diameter of D = 2 m and a volume of VR = 15 m3 (Fig. 8). The reac-
used. Therefore, 200 mL of a 5.8 vol% phenolphthalein solution is
tor is equipped with a bottom mounted magnetic agitator (ZETA
added to the reactor once. Phenolphthalein is colorless for pH val-
BMRF) with the possibility to utilize different stirrer types. Two
ues between 0 and 8.2 and is colored pink in case of pH values
different impeller types (Rushton turbine and pitched blade, (d/D
above 8.2. At the beginning of each measurement, the pH-value
= 0.33) are used. The investigated impeller set ups are presented
is shifted to a higher value by adding 500 mL of a 2 M NaOH solu-
in Table 6. An open tube sparger is used, which is located below
tion, which leads to a change in the color of the phenolphthalein
the Rushton turbine. The gas flow rate is controlled by a Proline
ion. After adding of HCl in a 1.5 times higher stoichiometric ratio
Promass 80F flowmeter. Additionally, three baffles of width LB =
(NHCl/NNaOH = 1.5), the pH-value is shifted backwards and the phe-
D/10 and a distance of 120° are mounted at the reactor wall. The
nolphthalein turns colorless. The mixing time is estimated by mea-
temperature within the reactor is monitored by two Thermocouple
suring the time that is needed from the addition of the HCl solution
Typ K temperature probes. With an external heat exchanger the
to reach a decolouration level of 95% (see Fig. 9).
temperature level of the reactor can be kept constant. During the
Five spotlights and a white diffuser foil at the backside of the
measurements the external loop is switched off. Due to the lid
reactor are providing a homogeneous background illumination.
and the good isolation characteristics of the acrylic glass, the tem-
For a quantitative evaluation of the mixing time, the decolouration
perature loss is only about 0.1 °C/h and therefore adiabatic condi-
is recorded by a CMOS camera with a sampling rate of 1 s1. The
tions can be assumed.
images are saved with a resolution of 4000  6000 pixels in the
Furthermore, a torque measuring cell (Lorenz DR-3000) is
native raw format. Afterwards, the time series are analyzed by
installed at the stirrer shaft to precisely measure the induced tor-
using the image analyzing software ImageJ that evaluates the
que by the impeller.
development of the mean grey value over time. The plotted mean
The gas hold-up eG is estimated by measuring the change in the
grey value for each time step is additionally normalized to the end
liquid height under aerated HG and unaerated H0 conditions
value of the corresponding measurement (see Fig. 10). The mixing
according to
time tmix is defined as the time from the addition of the acid until
the end value of 0.95 is reached.
Table 6
List of investigated stirrer combinations.

Nomenclature Rt-Pb-6kL Rt-Pb-12kL Rt-2Pb-15kL

4. Experimental results
Stirrer system Lower: Rushton Lower: Rushton Lower: Rushton
Turbine Turbine Turbine The following section provides the experimental results for
Upper: pitched Upper: pitched Middle: pitched
blade blade blade
mixing times that have been achieved in the 15 m3 acrylic glass
Upper: pitched reactor by means of the decolouration method. First the results
blade on industrial scale without aeration will be presented followed
Vfill/m3 6.5 12.5 15 by the results of flow patterns and mixing times for aerated oper-
H/D 1 2 2.7
ation conditions.
 A. Rosseburg et al. / Chemical Engineering Science 188 (2018) 208–220 215

t0 0.5 · tmix tmix

Fig. 9. Example of a decolouration time series.

combinations and filling levels H/D. As expected, the mixing time

Fig. 10. Example of the normalized grey value over the sample time to determine
the mixing time tmix.
4.1. Industrial scale bioreactor in single phase operation
In Fig. 11 the mixing time for the acrylic glass reactor is pre-
sented as a function of the Reynolds number for different impeller
decreases with increasing stirrer frequency due to the higher
energy dissipation rate and induced turbulence for all cases. Fur-
thermore, with increasing volume the mixing time is increasing,
even if an additional pitched blade turbine is added in the upper
part of the 15kL reactor to enhance the mixing performance in this
area. Because publications that are dealing with experimental
results on this large scale are scarce, a comparison with literature
is difficult. In most cases different geometries or stirrer combina-
tions have been used. Investigations on similar scale have been
published by Guillard and Trädgårdh (2003). They investigated
the mixing time on two different scales (D1 = 1.88 m/Vfill = 8 m3
and D2 = 1.4 m/Vfill = 1.8 m3) with multiple impeller systems (3
Rushton turbines and 2 Rushton turbines respectively) by measur-
ing the local conductivity. The comparison shows that the results
obtained in the acrylic glass reactor of Vfill = 6 m3 filling level (H/
D = 1, filled circle) are in a good agreement with the measurements
by Guillard for Vfill = 8 m3 (empty circle) and Vfill = 1.8 m3(empty

Fig. 11. Mixing time tmix for an unaerated stirred tank reactor as function of the Reynolds number for three different combinations of stirrers (Rt-2Pb-15kL/Rt-Pb-12kL/Rt-Pb-
6kL) [filled symbol] in comparison with data from Guillard and Trädgårdh (2003) [empty symbols].
216 A. Rosseburg et al. / Chemical Engineering Science 188 (2018) 208–220

Fig. 12. Dimensionless mixing time in dependency of the Reynolds number. Comparison between literature data and experimental results from the acrylic glass reactor.

diamond). With increasing volume and H/D respectively the mix- try and reactor aspect ratio and will provide data for more reliable
ing time increases. and transferable correlations in future.
To compare these results with common correlations, the dimen-
sionless mixing time is recommended in literature (Chmiel, 2011; 4.2. Industrial scale bioreactor in two-phase flow operation
Zlokarnik, 2003). Most authors predict a constant dimensionless
mixing time within the turbulent regime of Reynolds numbers In a second step the influence of the gaseous phase on hydrody-
Re > 104 (Kraume and Zehner, 2001; Henzler, 2005) which is also namics and mixing times has been investigated. Fig. 13 shows
reflected in the correlations in Table 1, where n  t mix is only a func- exemplarily the dispersion of the gaseous phase by the Rushton
tion of the impeller type and geometry of the reactor. turbine from a viewpoint below the stirrer. It can be seen clearly
The dimensionless mixing times for the acrylic glass reactor how the vortices behind the impeller blades are dispersing the gas-
with the three stirrer combinations (Rt-2Pb-15kL/Rt-Pb-12kL/Rt- eous phase over the whole cross section of the reactor. This leads to
Pb-6kL) from Fig. 11 are plotted in Fig. 12 as a function of the Rey- a homogeneous distribution of the gas bubbles with a narrow bub-
nolds number. Against most predictions the dimensionless mixing ble size distribution within the whole reactor volume. By increas-
time is not a constant for Re > 104 but increases with increasing ing the gas flow rate or decreasing the stirrer frequency the
Reynolds number. For all three combinations, a similar trend of induced momentum becomes insufficient to distribute the total
the dimensionless mixing time in dependency of the Reynolds amount of gas which is passing the impeller. Under this condition
number can be noticed: large bubbles can pass the impeller region and induce a heteroge-
neous bubbly flow due to their higher rise velocity. This strongly
Rt-Pb-6kL: tmix  n Re0:33 ;(H/D) = 1 influences the flow pattern within the reactor and thus the mixing
Rt-Pb-12kL: tmix  n Re0:47 ; (H/D) = 2 time as described in detail in the following section.
Rt-2Pb-15kL: t mix  n Re0:41 ; (H/D) = 2.4 The different flow regimes can be determined quantitatively
with the transition of loading and flooding and thus with the crit-
Similar experimental results with comparable trends have been ical Flow number Flc. To determine the transition between loading
reported by Gabelle et al. (2011) even for reactors of smaller vol- and flooding in the acrylic glass reactor the gas hold-up has been
ume but for high Reynolds numbers. Zlokarnik already mentioned
in 1967, that the mixing time for very high Reynolds numbers
tends to get constant because it is only limited by the constant dif-
fusion in this regime (Zlokarnik, 1967). This indicates, on the other
hand, that the dimensionless mixing time needs to increase at a
certain point. The dotted line in Fig. 12 shows this trend of the
dimensionless mixing time of a Rushton turbine proposed by
Zlokarnik (1967). This trend can be found in the data from Gabelle
as well as in the data from the industrial scale acrylic glass reactor.
Despite the small amount of measuring points the trend of sin-
gle phase mixing makes clear that the approximation of a constant
dimensionless mixing time needs to be treated with caution for
large scale reactors. Due to the large dimension of the stirrer the
Reynolds number is much higher even for very low stirrer frequen-
cies. Whereas in small scale reactors and moderate Reynolds num-
bers the assumption of a constant dimensionless mixing time is
useful, in large industrial scale reactors an error of at least 20%
needs to be taken into account.
However, the acrylic glass reactor enables a detailed analysis of
mixing times in dependency of the impeller combination, geome- Fig. 13. Dispersion effect of the Rushton turbine in the loading regime.
 A. Rosseburg et al. / Chemical Engineering Science 188 (2018) 208–220
Fig. 14. Gas hold-up in dependency of the Froude number for different exemplary superficial gas velocities in the Rt-Pb-12kL reactor.
measured for a wide range of parameters. In Fig. 14 the gas hold-up
eG is shown as a function of the Froude number for three different
exemplarily gas flow rates. As expected, with increasing superficial
gas velocity and increasing stirrer frequency, the gas hold-up
increases. However, the transition from loading to flooding can
be determined by a change in the slope of the graphs (red circles).
For higher stirrer frequencies the gas hold-up remains almost
constant because of the efficient dispersion of the gaseous phase
by the Rushton turbine. In this region a homogeneous bubbly flow
can be observed. With decreasing stirrer frequency a sudden drop
of the gas hold-up can be found because the momentum induced
by the turbine is not sufficient anymore to disperse the gaseous
phase completely. Bigger bubble and bubble agglomerates appear
which are passing the turbine with high rise velocity, leading to
a drop in gas hold-up and the transition to flooding. This can be
visualized clearly within the acrylic glass reactor at the upper
end of the rotating stirrer axis. Fig. 15a shows the heterogeneous
regime with a very broad bubble size distribution and gas hold-
up distribution whereas Fig. 15b, right shows the same area at a
higher Froude number and lower Flow number that causes a
homogeneous flow regime with a narrow bubble size distribution
as well as a homogeneous gas hold-up distribution over the reactor
a) Fl = 0.018
Fr = 0.017
Srrer
sha
Fig. 15. Exemplary bubbly flow of the inhomogeneous (a) and the homogeneous flow (b).
217
cross section. With optical measurement techniques the bubble
size distribution of the different flow regimes can be determined
precisely, but due to the inhomogeneity within the large scale
reactor, this requires great effort. Thus, the evaluation of the bub-
ble size distribution will be done separately in a further
publication.
In Fig. 16 the transition points detected in Fig. 14 are compared
with literature data. As can be seen, a good agreement between the
experimental data and the correlation of Mikulcova is archived.
From this investigation it can be concluded that the transition
between loading and flooding for the industrial scale acrylic glass
reactor is in the same range as presented by Mikulcova and can
be characterized with the critical Flow number
Flc ¼ 0:47 Fr ð20Þ
The mixing time in dependency of the Reynolds number for two
phase flow operation is shown in Fig. 17. As expected, at small Rey-
nolds numbers Re < 2  105 the mixing time decreases with increas-
ing Reynolds number, visible in Fig. 17 for the aerated condition. At
moderate Reynolds numbers 2  105 < Re < 6  105 the gassing still
has a positive effect on mixing time. However, the stirrer frequency
has a negative effect on the mixing time in case of higher gassing
b) Fl = 0.00085
Fr = 0.121
Srrer
sha
218 A. Rosseburg et al. / Chemical Engineering Science 188 (2018) 208–220

Fig. 16. Transition between loading and flooding - Comparison of flooding relations from literature with own data.

Fig. 17. Mixing time as a function of the Reynolds number in the acrylic glass reactor (Rt-Pb-12kL) for different superficial gas flow rates [symbols] and the fitted correlation
from Alves, Eq. (17) [A + B * exp(C * ((Fl/Flp)  1)) [lines] with A = 17; B = 20; C = 1.27.

rates. This behavior can be explained as follows: With increasing
stirrer frequency the momentum induced by the stirrer increases
and leads to a better mixing within the reactor. The gassing rate
has a positive effect as well due to the buoyancy induced flow that
leads to large scale secondary flows and a break up of stable fluid
structures. This inhomogeneity within the reactor induced by the
gaseous phase is homogenized by the Rushton turbine with
increasing stirrer frequency. Because the homogenization leads to
less large-scale vortices the mixing performance decreases with
increasing stirrer frequency in the region of lower Reynolds num-
bers. Especially this operation regime is of great importance for the
pharmaceutical industry because the volumetric power input is
restricted for this application. For higher Reynolds numbers the
momentum induced by the Rushton turbine is dominant and the
gaseous phase plays a minor role. Therefore, the mixing time is
almost independent on Reynolds number and superficial gas veloc-
ity under this condition.
As can be seen in Fig. 17, the correlation from Alves (Eq. (17)) is
in an acceptable agreement with the experimental data. For this
correlation, the critical Flow number Flc was calculated with the
equation by Milokova (Eq. (15)) and the assumption that the
influence of gassing on power input can be neglected (PG/P0 = 1).
This has been confirmed before by measurements of the power
input. Nevertheless, it has to be mentioned that an adjustment of
the parameter A, B and C of Eq. (17) by experiments is necessary.
The reliability of this correlation for different scales needs to be
proved in more detail, especially when the parameters are kept
constant.
The biggest advantage of the total optical access is possibility to
visualize the development of the mixing process in detail. In Fig. 18
two decolouration processes are shown. Both pictures were taken
after 50% of the total mixing time. In Fig. 18a it can be seen how
the compartments are formed in the lower, well mixed part and
the upper part of the reactor with a poor axial mixing. In this
regime, the bubbles are rising homogeneously dispersed over the
cross section of the reactor and are not able to break the barrier
of the compartments. The upper part of the reactor leads to a
longer mixing time. In Fig. 18b the same superficial gas velocity
but a decreased stirrer frequency had been adjusted. In this regime,
the induced momentum by the impeller is insufficient to disperse
 A. Rosseburg et al. / Chemical Engineering Science 188 (2018) 208–220
Fig. 18. Exemplary decolouration processes after t = tmix  0.5 for two different flow regime. (a) Loading. (b) Flooding.
and distribute the gas phase well over the cross section of the reac-
tor. This leads to a rise of large bubbles and thus to a strong buoy-
ancy driven flow. This flow breaks up the barrier and leads to a
better mass transfer to the upper region of the reactor which
results in a shorter global mixing time.
5. Conclusion
For industrial mixing processes in large scale aerated stirred
tank reactors, the knowledge how the mixing time depends on
the stirrer frequency and aeration rate is of great importance. Espe-
cially for pharmaceutical applications where the product quality
and yield is sensitive to deviations from the standard operation
conditions a deep understanding how the two-phase flow hydro-
dynamics influences the mixing is indispensable. To get a deep
understanding about the local flow pattern in large scale stirred
tank reactors, an optical access is useful which is not provided in
industry usually. From investigations on lab scale the different
influences of the gaseous phase on the flow structure and thus
on the mixing behavior are already known but it has not been
shown so far that those influences are also valid in large scale reac-
tors. With the help of the 15 m3 acrylic glass reactor at TUHH it is
possible for the first time to visualize the global flow pattern on
219
large scale and to examine the mixing characteristics for different
stirrer types and combinations.
Measurements in single phase operation for three different stir-
rer combinations and reactor volumes have shown that available
correlations assuming constant dimensionless mixing times are
only valid for small scales and limited Reynolds numbers. On large
scales, the Reynolds number exceeds large values even for small
stirrer frequencies because of the large characteristic length of
the stirrer diameter. This causes an increase of the dimensionless
mixing times even for large Reynolds numbers and deviations of
up to 20% by assuming a constant value.
For two-phase flow operations on industrial scale the hydrody-
namic inhomogeneities induced by the gaseous phase have been
analyzed in detail spatially and temporally by means of the optical
access. It could be shown that bioprocesses often operate in the
transition regime from loading to flooding. Therefore, the flow pat-
tern shows a high diversity where either the momentum induced
by the impeller or the momentum induced by the buoyancy driven
flow dominates the global flow pattern and thus the mixing time. It
has become evident that the homogeneity of the gaseous phase has
a strong influence on mixing time and needs to be taken into
account for calculations. Increasing the stirrer frequency within
the heterogeneous flooding regime might cause a transition to
the homogeneous loading regime with increasing mixing time. A
220 A. Rosseburg et al. / Chemical Engineering Science 188 (2018) 208–220
correlation from literature has been modified to predict this behav-
ior with acceptable accuracy.
With the 15 m3 acrylic glass reactor at TUHH a new facility is
available to study local hydrodynamic inhomogeneities on large
scales and their influences on mixing time, mass transfer perfor-
mance and shear stress. Further investigations are ongoing and will
be reported in this journal.
Acknowledgement
We grateful thank the ZETA company for providing the bottom-
mounted magnetic agitator and the images of the 15kL acrylic glass
reactor.
References
Alves, S.S., Vasconcelos, J.M., 1995. Mixing in gas-liquid contactors agitated by
multiple turbines in the flooding regime. Chem. Eng. Sci. 50 (14), 2355–2357.
https://doi.org/10.1016/0009-2509(95)00091-I.
Ascanio, G., 2015. Mixing time in stirred vessels: a review of experimental
techniques. Chin. J. Chem. Eng. 23 (7), 1065–1076. https://doi.org/10.1016/j.
cjche.2014.10.022.
Bombač, A., Žun, I., 2006. Individual impeller flooding in aerated vessel stirred by
multiple-Rushton impellers. Chem. Eng. J. 116 (2), 85–95. https://doi.org/
10.1016/j.cej.2005.10.009.
Bombač, A., Žun, I., Filipič, B., Žumer, M., 1997. Gas-filled cavity structures and local
void fraction distribution in aerated stirred vessel. AIChE J. 43 (11), 2921–2931.
https://doi.org/10.1002/aic.690431105.
Bouaifi, M., Hebrard, G., Bastoul, D., Roustan, M., 2001. A comparative study of gas
hold-up, bubble size, interfacial area and mass transfer coefficients in stirred
gas–liquid reactors and bubble columns. Chem. Eng. Process. Process Intensif.
40 (2), 97–111. https://doi.org/10.1016/S0255-2701(00)00129-X.
Busciglio, A., Grisafi, F., Scargiali, F., Brucato, A., 2013. On the measurement of local
gas hold-up, interfacial area and bubble size distribution in gas–liquid
contactors via light sheet and image analysis: imaging technique and
experimental results. Chem. Eng. Sci. 102, 551–566. https://doi.org/10.1016/j.
ces.2013.08.029.
Chara, Z., Kysela, B., Konfrst, J., Fort, I., 2016. Study of fluid flow in baffled vessels
stirred by a Rushton standard impeller. Appl. Math. Comput. 272, 614–628.
https://doi.org/10.1016/j.amc.2015.06.044.
Chmiel, H. (Ed.), 2011. Bioprozesstechnik. Spektrum Akademischer Verlag,
Heidelberg.
Gabelle, J.-C., Augier, F., Carvalho, A., Rousset, R., Morchain, J., 2011. Effect of tank
size on kLa and mixing time in aerated stirred reactors with non-newtonian
fluids. Can. J. Chem. Eng. 89 (5), 1139–1153. https://doi.org/10.1002/cjce.20571.
Gagnon, H., Lounès, M., Thibault, J., 1998. Power consumption and mass transfer in
agitated gas-liquid columns: a comparative study. Can. J. Chem. Eng. 76 (3),
379–389. https://doi.org/10.1002/cjce.5450760306.
Guillard, F., Trägårdh, C., 2003. Mixing in industrial Rushton turbine-agitated
reactors under aerated conditions. Chem. Eng. Process. Process Intensif. 42 (5),
373–386. https://doi.org/10.1016/S0255-2701(02)00058-2.
Haß, V.C., Nienow, A.W., 1989. Ein neuer, axial fördernder Rührer zum Dispergieren
von Gas in Flüssigkeiten. Chemie Ingenieur Technik 61 (2), 152–154. https://
doi.org/10.1002/cite.330610213.
Henzler, H.-J., 2005. Rührprobleme in der Biotechnologie. In: Kraume (Hg.) 2005 –
Mischen und Rühren, pp. 375–403.
Jaworski, Z., Nienow, A.W., Dyster, K.N., 1996. An LDA study of the turbulent flow
field in a baffled vessel agitated by an axial, down-pumping hydrofoil impeller.
Can. J. Chem. Eng. 74 (1), 3–15. https://doi.org/10.1002/cjce.5450740103.
Kong, L.-N., Li, W., Han, L.-C., Liu, Y.-J., Luo, H.-A., Al Dahhan, M., Dudukovic, M.P.,
2012. On the measurement of gas holdup distribution near the region of
impeller in a gas–liquid stirred Rushton tank by means of c-CT. Chem. Eng. J.
188, 191–198. https://doi.org/10.1016/j.cej.2012.02.023.
Kraume, M., Zehner, P., 2001. Experience with experimental standards for
measurements of various parameters in stirred tanks. Chem. Eng. Res. Des. 79
(8), 811–818. https://doi.org/10.1205/02638760152721316.
Laakkonen, M., Honkanen, M., Saarenrinne, P., Aittamaa, J., 2005. Local bubble size
distributions, gas–liquid interfacial areas and gas holdups in a stirred vessel
with particle image velocimetry. Chem. Eng. J. 109 (1–3), 37–47. https://doi.org/
10.1016/j.cej.2005.03.002.
Lee, B.W., Dudukovic, M.P., 2014. Determination of flow regime and gas holdup in
gas–liquid stirred tanks. Chem. Eng. Sci. 109, 264–275. https://doi.org/10.1016/
j.ces.2014.01.032.
Machon, V., Jahoda, M., 2000. Liquid homogenization in aerated multi-impeller
stirred vessel. Chem. Eng. Technol. 23 (10), 869–876. https://doi.org/10.1002/
1521-4125(200010)23:10<869:AID-CEAT869>3.0.CO;2-B.
Middleton, J.C., Smith, J.M., 2003. Gas-Liquid Mixing in Turbulent Systems. In: Paul,
E.L., Atiemo-Obeng, V.A., Kresta, S.M. (Eds.), Handbook of Industrial Mixing.
John Wiley & Sons Inc, Hoboken, NJ, USA, pp. 585–638.
Montante, G., Horn, D., Paglianti, A., 2008. Gas–liquid flow and bubble size
distribution in stirred tanks. Chem. Eng. Sci. 63 (8), 2107–2118. https://doi.org/
10.1016/j.ces.2008.01.005.
Montante, G., Paglianti, A., 2015. Gas hold-up distribution and mixing time in gas–
liquid stirred tanks. Chem. Eng. J. 279, 648–658. https://doi.org/10.1016/j.
cej.2015.05.058.
Montante, G., Paglianti, A., Magelli, F., 2007. Experimental analysis and
computational modelling of gas-liquid stirred vessels. Chem. Eng. Res. Des. 85
(5), 647–653. https://doi.org/10.1205/cherd06141.
Nienow, A.W., 1997. On impeller circulation and mixing effectiveness in the
turbulent flow regime. Chem. Eng. Sci. 52 (15), 2557–2565. https://doi.org/
10.1016/S0009-2509(97)00072-9.
Nienow, A.W., 1998. Hydrodynamics of stirred bioreactors. Appl. Mech. Rev. 51 (1),
3. https://doi.org/10.1115/1.3098990.
Nienow, A.W., Langheinrich, C., Stevenson, N.C., Emery, A.N., Clayton, T.M., Slater, N.
K., 1996. Homogenisation and oxygen transfer rates in large agitated and
sparged animal cell bioreactors: some implications for growth and production.
Cytotechnology 22 (1–3), 87–94. https://doi.org/10.1007/BF00353927.
Nienow, A.W., Rielly, C.D., Brosnan, K., Bargh, N., Lee, K., Coopman, K., Hewitt, C.J.,
2013. The physical characterisation of a microscale parallel bioreactor platform
with an industrial CHO cell line expressing an IgG4. Biochem. Eng. J. 76, 25–36.
https://doi.org/10.1016/j.bej.2013.04.011.
Saito, F., Nienow, A.W., Chatwin, S., Moore, I.P.T., 1992. Power, gas dispersion and
homogenisation characteristics of Scaba SRGT and rushton turbine impellers. J.
Chem. Eng. Japan/JCEJ 25 (3), 281–287. https://doi.org/10.1252/jcej.25.281.
Shewale, S.D., Pandit, A.B., 2006. Studies in multiple impeller agitated gas–liquid
contactors. Chem. Eng. Sci. 61 (2), 489–504. https://doi.org/10.1016/j.
ces.2005.04.078.
Sieblist, C., Jenzsch, M., Pohlscheidt, M., 2016. Equipment characterization to
mitigate risks during transfers of cell culture manufacturing processes.
Cytotechnology 68 (4), 1381–1401. https://doi.org/10.1007/s10616-015-9899-
0.
Vasconcelos, J.M., Alves, S.S., Barata, J.M., 1995. Mixing in gas-liquid contactors
agitated by multiple turbines. Chem. Eng. Sci. 50 (14), 2343–2354. https://doi.
org/10.1016/0009-2509(95)00090-R.
Vrábel, P., van der Lans, R., Cui, Y.Q., Luyben, K., 1999. Compartment model
approach. Chem. Eng. Res. Des. 77 (4), 291–302. https://doi.org/10.1205/
026387699526223.
Vrábel, P., van der Lans, R.G., Luyben, K.C., Boon, L., Nienow, A.W., 2000. Mixing in
large-scale vessels stirred with multiple radial or radial and axial up-pumping
impellers: modelling and measurements. Chem. Eng. Sci. 55 (23), 5881–5896.
https://doi.org/10.1016/S0009-2509(00)00175-5.
Warmoeskerken, M., Smith, J.M., 1985. Flooding of disc turbines in gas-liquid
dispersions: a new description of the phenomenon. Chem. Eng. Sci. 40 (11),
2063–2071. https://doi.org/10.1016/0009-2509(85)87023-8.
Wiedmann, J.A., 1983. Zum überflutungsverhalten zwei- und dreiphasig
betriebener Rührreaktoren. Chemie Ingenieur Technik 55 (9), 689–700.
https://doi.org/10.1002/cite.330550904.
Wu, H., Patterson, G.K., 1989. Laser-Doppler measurements of turbulent-flow
parameters in a stirred mixer. Chem. Eng. Sci. 44 (10), 2207–2221. https://doi.
org/10.1016/0009-2509(89)85155-3.
Zlokarnik, M., 1967. Eignung von Rührern zum Homogenisieren von
Flüssigkeitsgemischen. Chemie Ingenieur Technik 39 (9–10), 539–548.
https://doi.org/10.1002/cite.330390909.
Zlokarnik, M., 2003. Stirring. Wiley-VCH Verlag GmbH & Co. KGaA. https://doi.org/
10.1002/14356007.b02_25. URL: <http://onlinelibrary.wiley.com/doi/10.1002/
14356007.b02_25/full>.