Anda di halaman 1dari 20

Chemical Engineering Science 81 (2012) 231–250

Contents lists available at SciVerse ScienceDirect

Chemical Engineering Science


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

Flow regimes and surface air entrainment in partially filled stirred vessels for
different fill ratios
Shilan Motamedvaziri, Piero M. Armenante n
Otto H. York Department of Chemical, Biological and Pharmaceutical Engineering, New Jersey Institute of Technology Newark, 323 M. L. King Boulevard,
Newark, NJ 07102-1982, USA

H I G H L I G H T S

c The hydrodynamics changes when the fill ratio H/T is less than one in stirred vessels.
c A critical impeller submergence ratio, Sbcr/D, exists.
c For Sb/D o Sbcr/D the flow pattern transitions from ‘‘double-loop’’ to ’’single-loop up-pumping’’ flow regime.
c The critical Sbcr/D ratio is not affected by agitation speed.
c Impeller flooding can occur only if Sb/D o Sbcr/D and, at the same time, if the Froude Number is above a critical value, Frcr.

a r t i c l e i n f o abstract

Article history: In many industrial applications, stirred vessels have a liquid height-to-tank diameter ratio (fill ratio),
Received 27 March 2012 H/T, equal to, or larger than, 1. However, there are many instances when H/To 1, as when a vessel is
Received in revised form emptied or filled. When the impeller submergence is reduced as a result of lowering the liquid level, the
24 May 2012
fluid dynamics of even a single-phase stirred liquid can become quite complex. The objective of this
Accepted 28 May 2012
Available online 14 June 2012
work was to study the hydrodynamic changes that occur when H/T is decreased, and to determine the
conditions for which adequate mixing can still be achieved. A baffled vessel equipped with a disk
Keywords: turbine was used. Particle Image Velocimetry (PIV) was used for the experimental determination of the
Mixing velocity profiles, impeller pumping capacities and Pumping Numbers. A strain gage system was used to
Hydrodynamics
measure power dissipation. Computational Fluid Dynamics (CFD) was used to predict velocity profiles,
Simulation
Power Numbers, and Pumping Numbers using a multiple reference frame (MRF) model coupled, when
Aeration
Agitation needed, with a Volume of Fluid (VOF) model. Results show that a critical impeller submergence ratio,
Impeller submergence Sbcr/D, exists below which: (1) the macroscopic flow pattern generated by the impeller changes
substantially, transitioning from a ‘‘double-loop’’ recirculation flow to a ’’single-loop up-pumping’’
recirculation flow; (2) the Power Number and radial Pumping Number drop significantly; (3) vortex
formation occurs, air entrainment is greatly facilitated, and impeller flooding typically results. The
critical Sbcr/D ratio was not affected by agitation speed. Impeller flooding was observed only when the
new regime was established and the vortex depth reached the impeller disk. This phenomenon was
correlated to a critical value of the Froude Number, Frcr. These results show that stirred vessels can be
effectively operated only within certain ranges of the operating variables without compromising their
mixing effectiveness. This knowledge can help practitioners avoid operating their equipment in regions
where the desired mixing effects are not achievable.
& 2012 Elsevier Ltd. All rights reserved.

1. Introduction as well as in many other industries. The development and


utilization of mixing processes involving mixing tanks is critical
Mixing of liquids and dispersed-phase systems in mixing tanks to a variety of industrial applications, since different types of
and reactors stirred by a centrally mounted impeller is a very mixing-related phenomena are often the preliminary requirement
common operation encountered in the pharmaceutical industry for, or an integral component of, other operations, such as
reaction or crystallization (Paul et al., 2004).
Typically, mixing vessels and reactors are utilized at full
n
Corresponding author. Tel.: þ1 973 596 3548; fax: þ 1 973 596 8436. capacity, which implies that the liquid level is the maximum
E-mail address: piero.armenante@njit.edu (P.M. Armenante). practically achievable in the system without overflowing it.

0009-2509/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ces.2012.05.050
232 S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250

In vessels in which the tank diameter, T, is similar in size to the the distance between the impeller and the free surface is reduced as
vessel height, the liquid height, H, is typically equal to, or larger a result.
than, the tank diameter, i.e., H/T¼1 or H/T41. It is therefore not Therefore, the objective of this work was to study in some
surprising that systems for which H/T¼1 have been extensively detail the hydrodynamic changes that occur in a mixing tank
studied to determine, among the others, the velocity distribution when the fill ratio H/T is decreased, and determine the role that
of the liquid inside the vessel, the power dispersed by the different variables, such as agitation speed and impeller submer-
impeller, the pumping capacity of the impeller, and the minimum gence, and hence liquid level, play in the establishment of
agitation speed for complete dispersion of a second phase in the possible different mixing regimes. An additional focus of this
liquid, including finely dispersed solids, immiscible liquids, and work was to characterize the different fluid flow regimes that can
sparged gases. Thus, a significant body of literature exists on be observed in such systems when the impeller submergence is
many such aspects of mixing in single and multiphase systems for reduced as a result of lowering the liquid level and how these
which H/T¼1. regimes affect the mixing performance of the agitation system.
However, there are many instances where this ‘‘fill ratio’’, H/T, The overarching objective of this work was to determine the
is lower than 1, as in all those cases in which the vessel is either operating regions in which stirred tanks can be safely operated
emptied or filled. This is often the case for stirred reactors and at reduced H/T levels while still maintaining adequate mixing
vessels operating in a batch mode, which have been, and still are, intensity for industrial processes to be effectively conducted.
very commonly used in industry (Horowitz, 2010), and especially The approach that was used in this work to achieve this goal
in the fine chemical and pharmaceutical industry. For example, the consisted primarily in using a number of techniques to study the
syntheses of intermediates and final products for the manufactur- velocity distributions in agitation vessels including experimental
ing of Active Pharmaceutical Ingredients (API’s) is typically con- methods such as Particle Image Velocimetry (PIV) and computa-
ducted in a series of batch processes in stirred reactors, with each tional methods such as Computational Fluid Dynamics (CFD).
step being described in a predetermined engineering flowsheet These methods were essential to investigate a number of complex
detailing the sequence of operations and the time involved (e.g., phenomena in mixing vessels in the past (Yianneskis et al., 1987;
‘‘charge vessel with initial amount of solvent’’, ‘‘add solid reac- Armenante and Chou, 1996; Armenante et al., 1997; Sheng et al.,
tant’’, ‘‘raise the temperature to set value,’’ ‘‘stir at set agitation 2000; Montante et al., 2001a, 2001b; Campolo et al., 2002; Akiti
speed and let system react for predetermined time’’, ‘‘add remain- and Armenante, 2004; Yeoh et al., 2004; Deglon and Meyer, 2006;
ing solvent and stir,’’ etc.). Because of the nature of such a process, Murthy and Joshi, 2008).
the reactor volume often changes and so does the fill ratio H/T.
When the impeller submergence is reduced as a result of lowering
the liquid level, the fluid dynamics of even a single-phase stirred 2. Background
liquid can become quite complex, with different regimes possibly
existing depending on the geometric characteristics of the system Modifying the fill ratio H/T can, in principle, have an impact
(such as impeller clearance, liquid height, and liquid submergence on some of the macroscopic flow characteristics in the mixing
above the impeller). However, even when H/To1, it would be system, such as flow regimes, as well as other key process
clearly desirable to maintain a sufficient agitation level in order to variables such as power dissipation by the impeller and impeller
attain the desired process objectives. If this is not the case, the pumping rate. Therefore, other cases where such changes have
impeller could lose its ability to discharge its mixing duties (e.g., been shown to occur in mixing vessels are briefly reviewed here.
liquid blending, solid suspension) even before it becomes exposed Different macroscopic flow patterns (flow regimes) are com-
to air, i.e., when the liquid drops below it. In addition, undesired monly observed in stirred tanks depending on the impeller type
effects such as surface air entrainment could appear as the liquid and a number of geometric parameters. The occurrence of a given
level is being lowered, with unintended consequences for the flow pattern in stirred tanks has a significant impact on the
process (e.g., liquid splashing, solids deposition on the vessel process carried out in the mixing system. For instance, a typical
surface, oxidation of product, incorporation of bubbles in the radial impeller placed in a cylindrical vessel filled with a low
product, etc.). Therefore, it is critical to maintain the proper level viscosity liquid typically produces a ‘‘double-loop’’ recirculation
of liquid and adequate intensity of agitation at all times in order to flow, where the fluid flow forms two loops, one above and the
satisfy the mixing requirements of the process or, at least, to be other below the impeller. However, if the impeller clearance off
able to determine the operating regions in which the vessel can the bottom of the tank is reduced, a flow pattern transition
perform satisfactorily as far as mixing is concerned. eventually (and suddenly) occurs at a critical impeller clearance
Despite its industrial relevance little attention has been paid value, and the lower recirculation loop is suppressed as a result.
so far to mixing phenomena in stirred vessels with a fill ratio lower This double-loop to single-loop transition was first described by
than 1, especially when the impeller submergence, i.e., the distance Nienow (1968) and its importance in processes carried out in
between the impeller and the free surface, becomes small enough stirred tanks, such as solid suspension behavior, was investigated
for other phenomena to become important. Some investigators have by a number of investigators (Ibrahim and Nienow, 1996, 1999;
examined a related issue, i.e., the effect of the vertical location of the Armenante and Uehara Nagamine, 1998; Armenante et al., 1998;
impeller in a regular mixing tank with H/T¼1 on phenomena such Montante et al., 2001a, 2001b; Ochieng et al., 2008; Li et al., 2011).
as surface aeration, and have included the impeller clearance ratio The overall flow pattern in mixing tanks and the changes that
C/T in their correlations for minimum agitation speed for surface air are introduced when geometric and operating variables are varied
entrainment (Bhattacharya et al., 2007; Mali and Patwardhan, 2009). also have a strong impact on different mixing phenomena and
Some of these studies have also looked at the corresponding changes mixing parameters, which, in turn, have a significant effect on the
in liquid velocity near the air–liquid interface. However, none of process carried out in the mixing system. One of the most
these studies was focused specifically on the hydrodynamic changes important variables in the description of mixing phenomena is
that can be expected to occur when the liquid level drops rather the power P dissipated by the impeller and the non-dimensional
than when the impeller is placed closer to the surface. Significant Power Number, Po, defined as:
work still remains to do to understand how mixing vessels operate
at low fill ratios. There is clearly the need to examine the mixing P
Po ¼ ð1Þ
phenomena that occur in stirred tanks when the H/T is lowered and rN3 D5
S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250 233

The Power Number can be shown to be constant for impellers up-pumping flow have a stronger tendency to incorporate air
operating in baffled tanks under turbulent regime, once the geo- bubbles from the air–liquid interface by producing stronger
metry of the system is fixed. Typically, the Power Number changes if velocities close to air–liquid interface. For a constant flow pattern,
the geometry of the system changes (e.g., by varying the impeller researchers have reported that when the impeller submergence
clearance), especially if the flow regime changes as a result. For decreases surface air entrainment is observed at a lower agitation
example, the turbulent Power Number for a typical disk turbine speed (Bhattacharya et al., 2007). While air entrainment is
operating in the double-loop regime in a typical mixing configura- beneficial to some processes like wastewater treatment, it can
tion is about 5 (Wu and Patterson, 1989; Dyster et al., 1993; Lane produce serious problems in others, such as suspension polymer-
and Koh, 1997; Armenante et al., 1992, 1998; Armenante and ization. Therefore, NE, the minimum agitation speed capable of
Uehara Nagamine, 1998; Montante et al., 2001a, 2001b; Murthy entraining bubbles from the free surface into the liquid, is an
and Joshi, 2008; Yang and Takahashi, 2010). However, this value important operational limit. One of the mechanisms that has been
drops to about 4 if the impeller is positioned low enough in the reported in the literature to be responsible for the gas entrain-
vessel for the lower recirculation loop to be suppressed (Nienow, ment, and an important one for the work conducted here, is the
1968; Armenante and Uehara Nagamine, 1998; Armenante et al., engulfment of gas bubble caused by turbulent velocity fluctua-
1998; Montante et al., 2001a, 2001b). The power dissipated by the tions at the free interface between the stirred liquid and the gas
impeller is fundamental to quantify a number of mixing phenom- phase above it (Bhattacharya et al., 2007).
ena, such as mass transfer to a dispersed phase, since energy is
needed to homogenize the vessel content, disperse immiscible
phases, suspend solids, increase mass transfer, and, in general,
produce a number of the desired mixing effects. 3. Experimental apparatus, materials and methods
The impeller pumping capacity, i.e., the flow rate generated by
the impeller during its operation, Q, is another important char- 3.1. Mixing system
acteristic of the impeller. Usually the pumping capacity is
expressed in terms of the non-dimensional Pumping Number The experimental apparatus used in this work, shown in Fig. 1,
(or Flow Number) NQ defined as: consisted of a baffled, cylindrical, transparent, Plexiglas mixing
tank with a flat bottom, having an internal diameter, T equal to
Q
NQ ¼ ð2Þ 0.246 m. The tank was provided with four baffles having a width
ND3
of T/10, spaced 901 apart and extending all the way to the bottom
Different types of impellers produce different pumping actions, of the tank. The tank was equipped with a single 6-blade disk
which cause the establishment of different fluid flow circulation turbine (DT) impeller having a diameter, D equal to 0.0765 m, and
patterns inside the tank. a blade height of 0.0153 m (¼0.2 D). The thickness of the impeller
Power consumption and pumping capacity of the impeller blade and disk was 1.21 mm. The impeller was mounted on a
depend not only on the type of impeller used, the agitation speed, centrally placed shaft (shaft diameter: 1.27 cm) coupled to an
and the physical properties of fluid, but also on the geometric inline torque transducer (described below) which was in turn
characteristics of the system, such as impeller type, impeller connected to a variable-speed motor (Model CG-2033-11; Chem-
clearance, impeller diameter-to-tank diameter ratio, and others glass Life Sciences, Vineland, NJ). The motor was equipped with a
(Rushton et al., 1950; Yianneskis et al., 1987; Dyster et al., 1993; controller to adjust the agitation speed in the range 0–500 rpm
Lane and Koh, 1997; Patwardhan and Joshi, 1998; Murthy and and with an rpm-meter, measuring the rotational speed within
Joshi, 2008; Yang and Takahashi, 2010). This topic has been 71 rpm. The agitation speed was also periodically checked using a
the subject of significant work. For example, Mahmoudi and digital tachometer (Model HT-4100, Distek, Inc., New Brunswick,
Yianneskis (1992) investigated the effect of the impeller position NJ). A traversing system supporting the tank was used to set the
on mixing time when the process is operated under different flow vertical position of the tank with respect to the impeller, Cb,
patterns. Rutherford et al. (1996) and Chapple et al. (2002)
studied the effect of the impeller dimension on mixing character-
istics such as Power Number and Pumping Number. Zhao et al.
(2011) studied the flow characteristics for different impeller types
and impeller spacing.
In addition, other less easily quantifiable phenomena are also
critical for the success of mixing operations, such as the effect of
the intensity and the characteristics of the flow pattern in the
lower portion of the vessel on the attainment of the complete
suspension of solids off the tank bottom. Suspension of solid
particles is a complex phenomenon caused by strong turbulent
disturbances near the tank bottom, which can lift the particles
upwards, combined with a sufficiently high average velocity field
‘‘sweeping’’ the tank bottom and distributing the particles in
other regions of the vessel (Baldi et al., 1978). Therefore, the
vessel configurations and impeller types and geometries that
generate stronger turbulence near the bottom of the vessel are
typically capable of suspending solids at lower agitation speeds,
often with lower energy consumption.
The effect of the intensity and the characteristics of the
flow pattern near air–liquid interface are also important for a
number of processes such as for the case of surface air entrain-
ment (Bhattacharya et al., 2007; Mali and Patwardhan, 2009).
Impeller type, geometric and operating variables that produce an Fig. 1. Schematic of experimental set-up of agitation system.
234 S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250

measured from the bottom of the impeller to the bottom of the approach has been used by several investigators to quantify
tank. Cb in this work was always equal to 0.082 m (Cb/T¼ 1/3). the flow characteristics of mixing vessels (Sheng et al., 2000;
The liquid medium was always distilled water. The vessel was Montante et al., 2001a, 2001b; Brown et al., 2004; Li et al., 2011;
filled with water at the desired liquid level, H. As a result, the Zhao et al., 2011). In all experiments performed here, the water in
impeller submergence, Sb, i.e., the distance between the bottom of the vessel was seeded with neutrally-buoyant 10 mm silver coated
the impeller (bottom of the blades) and the air–liquid interface particles (Dantec Measurement Technology USA, Mahwah, NJ,
under no-agitation condition, could also be varied. USA) that could follow the fluid flow pattern very closely and
scatter the incoming laser light. Two different but similarly
3.2. Experimental determination of the impeller power dissipation operating sets of PIV apparatuses, labeled ‘‘PIV1’’ and ‘‘PIV2’’,
and Power Number were utilized in this work. PIV1 was a two-dimensional TSI PIV
apparatus (TSI Incorporated, Shoreview, Minnesota, USA)
The motor was mounted on a steel framework separate from equipped with two individually water-cooled laser heads, while
the tank (Fig. 1) to minimize vibrations that could have affect PIV2 was a Dantec FlowMap 2D PIV apparatus (Dantec Dynamics
torque measurement. Appropriate couplings were selected to A/S, Skovlunde, Denmark) equipped with one laser head. Both PIV
mount the torque transducer on the shaft between the motor systems used a double pulsed 120 mJ Nd-YAG laser (New Wave
and the impeller to minimize wobbling of the impeller shaft and Research model Gemini PIV 15, Fremont, CA, USA) consisting of
prevent the introduction of friction sources in the torque mea- two infrared laser heads combined in a single package with a
surement, as reported in other systems (Chapple et al., 2002). second harmonic generator and two discrete power supplies
The torque, G, applied to the impeller by the motor was emitting light at a 532 nm wavelength. The laser produced two
experimentally measured with a strain gage-based rotary torque pulsed infrared laser beams that passed through an optical
transducer (Model, T6-5-Dual Range, Interface, Inc. Scottsdale, AZ) arrangement of lenses to generate a laser light sheet acting as
connected to an external multi-channel signal conditioner, dis- the photographic flash for the single CCD digital camera
play and controller (Model 9850). The transducer operated by (PIV1: PIVCAM 10–30, TSI model 630046; PIV2: Dantec HiSense
producing an output voltage proportional to the applied torque, PIV/PLIF). The cameras had a spatial resolution of about
which was generated by the change in electric resistance within 1000  1016 pixels (PIV1) and 1280  1020 pixels (PIV2), respec-
the strain gages bonded to the torque sensor structure within tively. The laser and the digital camera were connected to a
the transducer (Model DCVA). A controller was used to feed the synchronizer (LASERPULSE Synchronizer, TSI model 610034),
torque transducer with a supply DC voltage, which was converted which was in turn connected to a computer (DELL Precision
to AC in the transducer stator, transferred inductively to the WorkStation 530). All components were controlled by dedicated
transducer rotor electronics, rectified, stabilized, and fed to the software (PIV1: Insight PIV Software; PIV2: FlowManager
strain gage bridge also contained within the rotor. The output Software) on the computer. Guided by the settings from the
from the bridge was conditioned in an amplifier, converted to a software, the synchronizer accepted the trigger dictated by the
digital signal, and transferred back to the stator by a rotating encoder and proceeded to synchronize the laser pulses with
transformer. The transducer could measure the torque applied the capture of appropriate images using the digital camera.
to the rotating shaft in two different ranges, i.e., 0–0.5 Nm and The software collected pairs of digitized images from the CCD
0–5 Nm. However, only the first range was used in this work. The camera (with the two images in each pair being collected at a
torque measurement error was 70.1% of full scale. The same small but known time interval, typically 400 ms), which were
instrument could also measure the agitation speed, N, and then subdivided into small interrogation areas, typically 64  64 pixels.
internally calculate and display the instantaneous power deliv- Each pair of frames for a given interrogation areas was then
ered, P, by the shaft, according to Brown et al. (2004): analyzed using cross-correlation to determine the spatial x-dis-
placement and y-displacement that maximize the cross-correla-
P ¼ 2pN G ð3Þ
tion function for that interrogation area. The resulting
The controller utilized its own software (M700) to interface displacement vector obtained by dividing the x- and y-displace-
with a computer, which was used for data acquisition, storage, ments by the time interval was taken as the fluid velocity in that
and post processing. The data acquisition rate was 100 Hz. Before interrogation area.
collecting experimental torque data, the torque transducer was
statically calibrated by blocking the motor and then mounting on
the shaft, below the transducer, an arm extending perpendicu- 3.4. Experimental determination of impeller pumping flow rate and
larly to the shaft, which pushed on a dynamometer (Shimpo impeller Pumping Number
Digital Force Gauge, Model FGV-0-5XY, Cole-Parmer, IL), placed
perpendicular to the arm. The torque calculated by multiplying An impeller rotating in an incompressible fluid produces a
the dynamometer force display by the length of the arm was pumping action that can be measured as the flow rate discharged
compared with the torque measured by the torque transducer. by the impeller, Qout, which must be balanced by an incoming
The agreement was always within 73%. flow rate toward the impeller, Qin. If the impeller pumping flow
The experimental power dissipation was measured for differ- rate is defined as Q, then Q¼Qin ¼Qout. Thus, the experimental (or
ent liquid levels and agitation speeds. In all experiments (con- computational) determination of Q requires only the determina-
ducted at least in duplicate), the system was allowed to stabilize tion of either Qout or Qin, which can be obtained by constructing
for 2 min before collecting power data in 20 s time intervals. an appropriate control volume around the impeller and measur-
Eq. (1) was used to obtain the experimental non-dimensional ing (or computing) the outgoing or incoming flow rate across the
Power Number, Po from P. surfaces defining such control volume. Ideally, a cylinder with
diameter and height exactly equal to the impeller diameter and
3.3. Experimental determination of the local liquid velocity by blades height should be used. In practice, this cylinder needs to
particle image velocimetry (PIV) be slightly bigger so that average velocity data can be measured
by PIV. For radial impellers such as disk turbines the discharge
Particle image velocimetry (PIV) was used to determine flow is nearly exclusively radial, which implies that the flow
experimentally the local fluid velocity in the mixing tank. This rate discharged radially across the side surface of the control
S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250 235

volume, Qr, is equal to Qout and hence to Q. Then, Qr can be pre-CFD mesh generator (Gambit 2.4.6) coupled with a CFD
obtained from: software package (Fluent 6.3.26). Simulations were conducted
Z wCV =2 on only half of the tank (1801-tank geometry) by applying
Qr ¼ U r outr ¼ DCV =2 pDCV dz ð4Þ periodic boundary conditions since this approach was less com-
wCV =2
putationally intensive. The exact geometry of each component of
In this equation, DCV and wCV denote, respectively, the diameter the system (vessel, impeller, shaft, baffles, etc.) was utilized to
and height of the cylindrical control volume surrounding the reproduce the exact shape of the impeller and vessel in which the
impeller where the velocities are obtained. numerical CFD simulations were conducted. An unstructured
In this work, the control volume was a cylinder, 4.2 cm in radius tetrahedral cell grid was generated around the impeller. A
and 2 cm in height, centered on the impeller. Radial velocities were structured hexahedral-cell type mesh was used in the rest of
measured by PIV1 along the side of this cylinder, i.e., on a vertical the domain. About 75% of the cells had an Equisize Skew value
segment (rake) 3.8 mm away from the impeller blades in the radial less than 0.1. In only 0.3% of the cells, mainly around the impeller,
direction, corresponding to a non-dimensional radial distance of this value was between 0.7 and 0.8. Four different mesh densities
r/R¼0.34. Velocity measurements were made at seven equally and qualities (i.e., meshes with 441,364 cells, 575,616 cells,
spaced locations along this rake, which extended 2.8 mm above 768,852 cells and 2,159,742 cells), were tested here in order to
and 2.8 mm below the impeller blade (corresponding to vertical assess the quality of the meshes and also to determine the
locations of z1 ¼0.072 m and z2 ¼0.093 m, respectively). The fol- smallest mesh size that produced mesh-independent results
lowing equation was used to obtain the experimental value of Qr: (Deglon and Meyer, 2006; Bai et al., 2007; Coroneo et al., 2011;
iX
¼6
Bai et al., 2011). The typical mesh used in most simulations
Qr ¼ U ri ðr,zi ÞDzi 2pr ð5Þ contained 768,852 cells for the H¼T case. Results were obtained
i¼1 using both the standard k–e and the realizable k–e models for
turbulence. For each model, first-order discretization and second-
where U ri ðr,zi Þ is the average of the two radial velocities measured
order discretization were used. However, only the results with
by PIV at zi (U ri ðr,zi Þ) and at zi þ 1 (U rði þ 1Þ ðr,zi þ 1 Þ). The velocity at
second-order discretization are presented in this paper because of
each location was the average of 500 PIV radial velocities measure-
their better agreement with the experimental results.
ments at any given r and zi. Since the PIV data were taken at
random position of the impeller relative to the baffles, the average
4.1. Multiple reference frame (MRF) method
velocity obtained via PIV could be taken as the average velocity at
any point along a circle with radius r ¼ DCV/2 and at an axial
The Multiple Reference Frame (MRF) method, a steady state
position zi (circumferential average). After computing the appro-
CFD modeling approach, was used to model the velocity distribu-
priate average radial velocity, this velocity value was multiplied by
tion (Montante et al., 2001a, 2001b; Aubin et al., 2004; Deglon
the elemental surface area (Dz2pr) of the cylindrical surface area
and Meyer, 2006; Coroneo et al., 2011). Accordingly, the total tank
where the measurements were taken (Dz¼3.5 mm).
volume was divided in two volumes: the first volume was a
Q values were obtained for different liquid fill ratios and were
cylinder with a radius of 0.068 m and a height of 0.08 m enclosing
used to determine the corresponding non-dimensional radial
the impeller region and centered on the impeller, while the
Pumping Number, NQ, from Eq. (2). Since Qr ¼Q for disk turbines
second volume included the remainder of the vessel including
it follows that that the radial Pumping Number and the impeller
the baffles. A flat air–water interface was assumed in all cases in
Pumping Number were identical in this case (NQ ¼ NQr).
which the MRF approach was used. Fluent solved the mass
momentum balance equations inside the rotating frame enclosing
3.5. Experimental determination of impeller discharge angle
impeller while those outside the rotating frame were solved in
the stationary frame. A steady transfer of information was made
The impeller discharge angle is defined as the angle between
at the MRF interface as the solution progresses. In general, if the
the direction of average vector of the radially discharge stream
interaction between impeller and baffles is weak, the effect of
produced by a radial impeller and the horizontal plane (Montante
relative orientation of impeller and baffles can be neglected and
et al., 2001a, 2001b). The impeller discharge angle was experi-
the MRF model can be successfully used (Marshall and Bakker,
mentally obtained here by taking the average angle of the velocity
2004). This was the case for this work as well as for similar
vectors in the r–z plane on the previously mentioned vertical rake
studies that have appeared in the literature (Montante et al.,
at r/R¼ 0.34. However, only the velocities at the five points on this
2001a, 2001b; Aubin et al., 2004; Coroneo et al., 2011). Since the
rake with elevations spanning over the blade height (i.e., between
MRF simulation is a steady state simulation of an intrinsically
(z1 ¼0.075 m and z2 ¼0.0904 m) were included in the calculation.
unsteady-state system, the average velocity field along a circum-
Positive angles indicate an upward flowing stream.
ference at any radial location was taken as the average velocity at
that location and compared with the experimental PIV data.
3.6. Experimental determination of the minimum agitation speed for
surface aeration 4.2. Volume of fluid (VOF) method

NE can be defined as the minimum agitation speed that results The VOF method can model the location of two immiscible fluids
at which the surface entrainment of air bubbles into the vessel (or phases) by solving a single set of momentum equations and
commences (Bhattacharya et al., 2007; Mali and Patwardhan, tracking the volume fraction of each of the fluids throughout the
2009). This speed was determined visually in this work, as domain (Hirt and Nichols, 1981). This method is especially useful to
previously reported (Joshi et al., 1981; Tanaka and Izumi, 1987; track the location of the interface between the two fluids (Gueyffier
Veljkovic et al., 1991; Bhattacharya et al., 2007). et al., 1999; Pilliod and Puckett, 2004). In previous stirred tank
applications, the VOF model was used to determine the shape and
4. Numerical simulation method location of the air–liquid interface under different operating condi-
tions (Serra et al., 2001; Haque et al., 2006; Mahmud et al., 2009).
Numerical simulations of the velocity distribution and turbu- In this work, the VOF model was used to model the air–water
lence levels in the tank were conducted using a commercial interface for different liquid levels. Starting from an initial setting
236 S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250

in which the top portion of the tank domain was computationally 5. Results and discussion
filled with air (corresponding to a volume fraction of water, ai,
equal to zero in those cells) and the lower portion with water 5.1. Results for the standard configuration (H/T¼1; Sb/D ¼2.23)
(corresponding ai ¼1 in those cells) the momentum balances in all
cells in the domain were solved until convergence was achieved. The most common configuration for stirred vessels is the one in
Those cell for which 0o ai o1 were at the air water interface. which H/T¼1, and C¼T/3 (Sb/D¼2.23). Therefore, this configuration
Whenever this occurred, Fluent applied a Geo-Reconstruct model, was used in this work to validate the appropriateness of the
i.e., an explicit time marching approach that captured the most experimental technique and the computational approach used here.
accurate shape of interface in this cell and the in neighboring cells. Results were obtained for N¼300 rpm, corresponding to an impeller
(http://www.fluentusers.com/support/ugm04/auto/multiphase.pdf, tip speed of 1.2 m/s and an impeller Reynolds Number, Re, of 29,250,
accessed October, 2010). where Re¼ rND2/m. Since this value falls within the turbulent
regime, a turbulence model was incorporated in the simulation to
account for turbulence effects. The MRF approach was used in
4.3. Computational prediction of impeller power dissipation and combination with both the standard k–e model and the realizable
impeller Power Number k–e model, denoted, respectively as ‘‘FS-SKE CFD Simulations’’ and
‘‘FS-RKE CFD Simulations’’ (where ‘‘FS’’ stands for ‘‘flat air–liquid
Once the flow field in the tank was computed via CFD, the surface’’). The experimental results were obtained as averages of 500
predicted values of the pressure on the front and back side of each PIV measurements by PIV1. This eliminated any periodic effects
impeller blade (p1 and p2, respectively) and the shear stress on the caused by the impeller position. Fig. 2 presents the dimensionless
blade width areas, disk, hub and shaft were obtained. The torque, mean axial and radial velocities obtained by CFD and PIV at three
G, applied to the impeller was then calculated from: different axial levels, i.e., below the impeller (Z/T¼0.14), above the
X  X impeller (Z/T¼0.38) and in the impeller region (Z/T¼ 0.33). This
G¼ p1 p2 i  r i dAi þ tj  r j dAj ð6Þ figure show that there was, in general, good agreement between the
i j
experimental data and the predicted results for the velocity dis-
where the calculations were performed over all control cells. The tribution, and that the velocity profiles predicted using both
subscript i refers to the cells on each blade and the subscript j to turbulence models (SKE and RKE) were similar to each other,
all the cells on the blade width areas, disk and hub. In this although those obtained with the RKE model were in slightly better
equation, r is the force arm, dA is the cell surface area, and (p1–p2) agreement with the experimental measurements.
is the pressure difference between the front and back of the blade Another approach used here to validate the simulation results
for each cell at any radial distance. Once the torque was predicted, was by calculating and experimentally measuring parameters
the same equations reported in the experimental section such as the Power Number, Po, and the impeller Pumping
(Eq. (1) and (3)) were used to calculate the predicted power Number, NQ. A ‘‘double loop’’ recirculation flow (DL Regime)
dissipation and hence the predicted Power Number, Po. was visually observed in the experiments, in which the radial
jet emanating from the impeller impacted the vessel wall produ-
cing a recirculation loop above the impeller and another one
4.4. Computational prediction of impeller pumping flow and below it. For this kind of flow pattern, Qout around impeller can be
Pumping Number expected to be almost identical to radial flow out, Qr. Therefore,
the impeller Pumping Number, NQ and the impeller radial Pump-
The same approach described above in the experimental ing Number, NQr were expected to be the same. Tables 1 and 2
section was used here computationally to calculate Qr (Eq. (4)) present a summary of the results for Po and NQ. Good agreement
and NQr (Eq. (2)) using the computationally predicted velocities was found between the experimental data and the computational
instead of the experimental results. Two approaches were used to results obtained here for NQ and Po. The RKE-based results for Po
calculate the radial discharge flow rate through the side surface of were found to be in slightly closer agreement with the experi-
cylindrical control volume, i.e., (a) using the velocity distribution mental measurements. The Pumping Number predicted by both
calculated via CFD and calculating Qr based on Eq. (4); and (b) turbulence models were the nearly identical. The results obtained
using a subroutine in Fluent to calculate the net flow rate through in this work are in agreement with those available from the
a selected surface (side surface of the cylinder). Both approaches literature (Wu and Patterson, 1989; Dyster et al., 1993; Lane and
gave identical results. The impeller radial pumping rate and the Koh, 1997; Murthy and Joshi, 2008; Yang and Takahashi, 2010).
radial Pumping Number (NQr) were obtained for different liquid fill Based on the computational results achieved in the basic
ratios. In all computations, the flow rates in and out of the control configuration (H/T ¼1), the RKE model was used as the turbulent
volume around the impeller were always in close agreement, thus model for the remaining of the simulations conducted in this
validating the mass balance. The largest error (¼(Qin–Qout)/Qin) work. The flow pattern and the velocity field in the same stirred
was found to be 1.12%. vessel at different fill ratios (H/T), i.e., different impeller submer-
gence ratios (Sb/D), were then examined.

4.5. Computational prediction of impeller discharge angle 5.2. Results for different impeller submergence ratios
(H/To1; Sb/D o2.23)
The impeller discharge angle was also predicted using CFD
simulation data and compared to that obtained experimentally 5.2.1. Effect of impeller submergence ratio on flow pattern
(Montante et al., 2001a, 2001b). The radial and axial velocity The velocity vectors on a vertical cross sectional area in the
components predicted by CFD on the side surface of the cylind- impeller region, obtained through PIV2 measurements, are pre-
rical control volume with r/R¼0.34 and height equals to the blade sented in the center panels in Fig. 3 for N¼300 rpm. These results
height (z1 ¼7.5 cm and z2 ¼9.04 cm) were used to calculate were obtained for different impeller submergence ratios, i.e., for
impeller discharge angle at each point on this surface. Then the Sb/D equal to 2.23 (H/T¼1), 1.24 (H/T¼0.69), 0.98 (H/T¼0.61), 0.77
final impeller discharge angle was reported as the average of all (H/T¼0.54), and 0.59 (H/T¼0.49). This figure also presents the
those data (some 2000 points). corresponding values predicted by CFD simulations (right-hand
S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250 237

Fig. 2. Comparison between PIV experimental results and CFD predictions of radial velocities (left-hand side panels) and axial velocities (right-hand side panels) using
both standard k–e (SKE) and realizable k–e (RKE) on three different horizontal iso-surfaces at different vertical locations for Cb/T¼ 0.30, D/T¼ 0.31 and N ¼300 rpm.

Table 1 predictions and PIV measurements of the velocities were, in


Power Numbers for H¼ T, D/T¼ 0.31 and Cb/T¼0.30. general, in substantial agreement with each other for all impeller
submergence levels examined here, although some discrepancy
Source Po
between the experimental data and the computational prediction
Torque measurements (this work) 4.89
could be observed for for Sb/D¼0.59, which was the most difficult
FS-SKE CFD simulation (this work) 4.66 case to examine experimentally and to properly simulate compu-
FS-RKE CFD simulation (this work) 4.74 tationally. However, and more importantly, it can be observed from
Literature dataab 4.4–5.5 Figs. 3 and 4 that the flow pattern for Sb/D¼ 0.59 (H/T¼0.49) is
a different from the flow patterns for all other cases in which the
Experimental values of Wu and Patterson (1989),
Dyster et al. (1993), Murthy and Joshi (2008), and Yang submergence level is larger, both computationally and experimen-
and Takahashi, 2010. tally. It appears that when the submergence level Sb/D reaches a
b
Simulations results of Lane and Koh (1997). critical value between 0.59 and 0.77 the upper recirculation flow
above the impeller is suppressed and the flow regime transitions
from a ‘‘double-loop’’ recirculation flow (DL Regime), where the
Table 2 fluid flow forms two recirculation loops one above and the other
Pumping Numbers for H¼ T, D/T¼ 0.31 and Cb/T¼0.30. below the impeller, to a ‘‘single-loop up-pumping’’ recirculation
flow (SL-up Regime), in which impeller pumps the liquid radially
Source NQin NQout or NQr
and upwards forming a single recirculation loop and creating a
PIV measurements (this work) 0.808
partially segregated core region just above the impeller, as evident
FS-SKE CFD simulation (this work) 0.790 0.790 from Figs. 3 and 4, and as more clearly and schematically shown in
FS-RKE CFD simulation (this work) 0.790 0.790 Fig. 5. The submergence level at which this flow reversal occurs is
Literature dataa,b 0.70–0.85 defined here as a critical submergence level, Sbcr. For the stirred
a vessel studied in this work the critical submergence level was
Experimental values of Wu and Patterson (1989) and Dyster et al. (1993).
b
Simulations results of Lane and Koh (1997). conservatively taken as the higher of the two Sb/D values between
which the transition occurred, i.e., Sbcr/D¼0.77 (H/T¼0.54).
side panels), in which a flat surface for the air–water interface was
assumed (the flow pattern predicted by CFD simulations are
presented for a larger cross sectional area than the PIV measure- 5.2.2. Effect of impeller submergence ratio on impeller discharge
ments for better visualization). The corresponding velocity contour angle, Pumping Number, and Power Number
plots, also obtained both experimentally and computationally, are The flow regime transition (from DL Regime to SL-up Regime)
shown in Fig. 4. The results in this figure indicate that the CFD shown in Figs. 3 and 4 was also found to be associated with
238 S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250

Fig. 3. Velocity vectors on a vertical cross sectional area for the different Sb/D ratios shown in the left-hand side panels (Cb/T¼ 0.30, D/T¼ 0.31 and N ¼ 300 rpm). Central
panels: PIV2 experimental measurements. Right-hand side panels: FS-CFD predictions. The rectangular area in each of the CFD simulation panels shows the areas of the
tank that was investigated by PIV and reported in the central panels.

changes in the impeller discharge angle, Pumping Number and and Po were nearly unaffected by submergence. However, when
Power Number, as shown in Fig. 6. the impeller submergence ratio decreased below the critical level
For relatively large impeller submergences (Sb/D 40.98 (H/T4 (Sbcr/D ¼0.77) and reached Sb/D ¼0.59, i.e., when the flow transi-
0.61)) the average impeller discharge angle was slightly positive tion occurred, these two dimensionless numbers decreased dras-
(41–71), i.e., the radial jet emanating from the impeller was tically. The FS-CFD predictions for Po and NQr appear to be in very
slightly pointed upwards. However, when submergence was good agreement with the experimental data for the liquid levels
closer to the critical value (Sbcr/D ¼0.77; H/T¼0.54) this angle equal to, and higher than, the critical submergence level (Sbcr/D Z
began to increase and it suddenly and significantly increased to 0.77 (H/TZ0.54)), i.e., before flow pattern transition occurred.
60.11 after the flow reversal had occurred (Sb/D¼ 0.59; H/T¼0.49). When Sb/D ¼0.59 the agreement was still substantial, although
This rapid transition was experimentally observed and computa- the FS-CFD simulations slightly overpredicted both Po and NQr.
tionally predicted (Fig. 6a). This figure additionally shows that a
substantial agreement between predictions and results exists at 5.2.3. Effect of impeller submergence ratio on velocity distribution
all submergence levels: the simulations underpredicted the dis- A quantitative comparison of the FS-CFD simulation results for
charge angles by 6–12%, i.e., by less than one degree for first two the radial and axial velocities for different impeller submergence
cases and by 5 degrees for Sb/D ¼0.59 (H/T¼0.49). ratios, i.e., for Sb/D ¼2.23 (H/T¼1), Sb/D ¼1.24 (H/T¼0.69),
Fig. 6b and c show the computational predictions and experi- Sb/D¼0.77 (H/T¼0.54), and Sb/D¼0.59 (H/T¼0.49), at five different
mental measurements of the Power Numbers and the radial axial (vertical) locations in the liquid is presented in Fig. 7. This
Pumping Numbers, respectively, at impeller submergence. When figure shows that the axial and radial velocity profiles were not
the impeller submergence ratio Sb/D was higher than 0.59, NQr strongly affected by the submergence ratio when this ratio was
S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250 239

Fig. 4. Velocity contours on a vertical cross sectional area for different Sb/D ratios (Cb/T¼ 0.30, D/T¼0.31 and N ¼ 300 rpm). Left-hand side panels: PIV2 experimental
measurements. Right-hand side panels: FS-CFD predictions.

Fig. 5. Schematic representation of flow regimes in a stirred tank equipped with a disk turbine at different submergence ratios: (a) Double-loop recirculation flow (DL
Regime) for impeller submergences larger than the critical submergence value; (b) Single-loop up-pumping recirculation flow (SL-up Regime) for impeller submergences
smaller than the critical submergence value.
240 S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250

submergence systems and especially solid suspension, which


requires higher average and fluctuating velocities close to tank
bottom to avoid particle settling.
Another less pronounced change in the flow field can be seen
in Fig. 7 as the impeller submergence was lowered from 1.24 to
0.77. The radial velocity in upper portion of the liquid (Z/T¼0.45)
for Sb/D ¼0.77 was mostly negative, implying that the flow in that
region was directed inward, i.e., toward the impeller. However,
for Sb/D Z1.24 this velocity was small but positive. This observa-
tion further confirms the smaller changes in flow pattern pre-
sented in Fig. 3.

5.2.4. Effect of agitation speed on flow pattern, surface air


entrainment, and impeller flooding
The flow patterns shown in Fig. 3 for different impeller
submergence ratios (Sb/D) were obtained at the same agitation
speed (300 rpm). These results clearly show that flow transition
occurs as a result of changes in the liquid level (and hence
impeller submergence), but they still leave open the question of
the effect of agitation speed on flow pattern, flow pattern
transition, and other important phenomena such as surface air
entrainment, which is typically observed for low submergence
ratios. In another words, without additional information it is still
unclear if the flow regime transition can only be attributed to
geometric variables (such as Sb/D) or if dynamic effects (such as
agitation speed) are also important. To address this question, the
flow pattern at low submergence was determined for different
agitation speeds using PIV2. The results were obtained for
Sb/D ¼0.77 (H/T¼0.54), i.e., for Sb/D ZSbcr/D, and Sb/D ¼0.59
(H/T¼0.49), i.e., for Sb/D oSbcr/D, for five different agitation
speeds, i.e., 50 rpm, 200 rpm, 300 rpm, 400 rpm and 500 rpm.
The results are presented in Fig. 8 for Sb/D ¼0.77 and in Fig. 9 for
Sb/D ¼0.59. The right-hand side panels in these figures show the
actual CCD images for each case, in order to evidence possible air
entrainment effects. Fig. 8 shows that for Sb/D ¼0.77 the ‘‘double-
loop’’ recirculation flow (DL Regime) was not affected by N even
when the agitation speed was very intense (N ¼500 rpm). The
double-loop flow pattern was observed for this impeller submer-
gence, even when the agitation speed was as high as 600 rpm
(results not shown for this case). A closer look at the flow patterns
in the left-hand side panels in Fig. 8 reveals some difference in the
upper zone around shaft at different agitation speeds. At higher
agitation speeds, the upper recirculation pattern was able to
produce a stronger flow that penetrated the upper area more
deeply, reaching the impeller shaft. However, at lower agitation
Fig. 6. Experimental and predicted values of (a) impeller discharge angle, speeds the upper recirculation was weaker and the loop ‘‘closed’’
(b) radial Pumping Number, NQr, and (c) Power Number, Po for different Sb/D more closely to the impeller blades than to the shaft. Despite
ratios (Cb/T¼0.30, D/T¼0.31 and N ¼ 300 rpm). these differences, the double-loop flow pattern was unaffected for
this Sb/D ratio.
sufficiently high, i.e., above the critical level (Sbcr/D Z0.77), However, some small surface air entrainment effects could be
although this was less so for velocities at the same Z/T location visually observed as the agitation speed was increased, even in
but for different submergence, as for the case of Sb/D ¼0.77. the presence of the DL Regime. NE, the minimum agitation speed
However, for Sb/D ¼0.59, i.e., after the flow transition to the at which the onset of surface air entrainment could be observed,
SL-up Regime had occurred (i.e., below Sbcr/D), the velocities were was visually determined using the criteria previously reported to
significantly different from all other cases in both intensity, and, describe this phenomenon (Bhattacharya et al., 2007; Mali and
in some cases, direction. Fig. 7 also shows the experimental Patwardhan, 2009) and was found to be approximately equal to
velocity measurements for Sb/D ¼0.59, which were found to be 450 rpm for this system (Sb/D ¼0.77), as very small bubbles first
in substantial agreement with the CFD predictions, thus again appeared to be entrained in the liquid and then become dispersed
validating the change in the flow pattern and the weakening of even in lower portion of the tank because of the double-loop flow.
the overall circulation flow that occurred when the submergence For N ¼500 rpm, small bubbles could be observed even in the CCD
ratio was lower than the critical value. image, as shown in the top right-hand side panel in Fig. 8. As the
The decrease in the radial and axial velocities that accompa- agitation was further increased, the number of small air bubbles
nied the flow transition was especially pronounced in the lower in the liquid increased. However, the presence of these small
region of the tank, i.e., for Z/Tr0.07. This phenomenon is bubbles did not alter in any way the double-loop flow generated
important since it is likely to affect the mixing efficiency of low by the impeller for this submergence ratio. A similar pattern was
S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250 241

Fig. 7. CFD predictions of radial velocities (left-hand side panels) and axial velocities (right-hand side panels) on five different horizontal iso-surfaces at different vertical
Z/T locations for different Sb/D ratios (Cb/T¼0.30, D/T¼0.31 and N ¼ 300 rpm). PIV1 experimental results for Sb/D¼ 0.59 (H/T¼ 0.49) are also presented.

observed also for Sb/D 40.77. For example for Sb/D ¼2.23, NE was pattern was present at all agitation speeds, including extremely
found to be equal to  600 rpm. low agitation speeds (N ¼50 rpm). This flow pattern resulted in
A very different pattern emerged when the same experiment the formation of a partially segregated core region just above the
was repeated at Sb/D ¼0.59 (H/T¼0.49), i.e., for a submergence impeller at any agitation speed for this submergence level. When
level below the critical submergence level at which flow reversal the agitation speed was low, the upward and radially outward jet
occurs (Sb/D oSbcr/D). Fig. 9 shows the PIV flow patterns and CCD emanating from the impeller was not strong enough to ‘‘pull’’
images for this submergence at the same agitation speeds used in fluid away from the core region just above the impeller and no air
the previous case (Fig. 8). Fig. 9 shows that the dominating flow entrainment occurred. However, when the agitation speed was
pattern at this submergence level was the SL-up Regime, as increased, the centrifugal force associated with this radial jet was
already mentioned above. Even more relevant it that this flow able to draw away more liquid from the upper central core region
242 S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250

Fig. 8. Velocity profiles measured by PIV2 (left-hand side panels) and CCD images of the PIV investigation area (right-hand side panels) at different agitations speeds for
Sb/D ¼0.77 (H/T¼ 0.54) (Cb/T¼ 0.30 and D/T¼ 0.31).
S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250 243

Fig. 9. Velocity profiles measured by PIV2 (left-hand side panels) and CCD images of the PIV investigation area (right-hand side panels) at different agitations speeds for
Sb/D ¼0.59 (H/T¼ 0.49) (Cb/T ¼0.30 and D/T ¼0.31). PIV data in the shaded were ignored due to vortex formation and surface air entrainment, thus making the velocity
profiles measurements in the liquid in those zones unreliable.
244 S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250

Fig. 10. Transition to impeller flooding condition: (a) Gas-sparged gas–liquid mixing system: impeller becomes flooded when more gas is sparged or the agitation speed is
decreased (Middleton and Smith, 2005; Middleton et al., 2004); (b) Liquid mixing system with reduced impeller submergence: impeller becomes flooded from surface
aeration when the impeller submergence is reduced below the critical Sbcr/D level and the agitation speed is sufficiently high (N ¼ 300 rpm; Cb/T¼0.30 and D/T¼0.31);
(c) same as (b) but at N ¼ 400.

resulting in a vortex which, depending on the agitation speed, However, this type of air entrainment was very different from the
could be more or less strong and possibly result in surface air limited air entrainment of single bubble observed at higher sub-
entrainment. mergence levels when the DL Regime dominated. For example, for
More specifically, the experimental results presented in Fig. 9 Sb/D¼0.77 (DL Regime), the few and small bubbles that were
along with visual observations showed that for moderate agitation entrained were distributed in the entire mixing vessel including
speeds (starting at 200 rpm) the air–water interface started to the bottom portion below the impeller (Fig. 8; N¼500 rpm), which
become deformed and a vortex began to appear. By further was a common observation even for higher submergence ratios.
increasing the agitation speed, the vortex depth increased. At about However, when Sb/D¼0.59, i.e., below the level that produced
N¼300 rpm the vortex reached the impeller disk and this greatly a change in flow pattern to result in the SL-up Regime, massive
facilitated surface air entrainment (NE for this case was 300 rpm). air entrainment occurred, large bubbles appeared, and the they
S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250 245

remained mostly close to the air–liquid interface (Fig. 9; N¼300 rpm, observed. For relatively low agitation speeds (N ¼100 rpm and
N¼400 rpm and N¼500 rpm), with no bubbles observed in the lower 200 rpm corresponding to Fr¼0.022 and 0.087, respectively), the
portion of the vessel below the impeller. As a result, the PIV velocity Power Number decreased from about 5 (for Sb/D ZSbcr/D,) to
vectors in the shaded areas in Fig. 9 should be disregarded because about 4 (for Sb/D oSbcr/D,). This drop can be attributed exclusively
the air–liquid interface is no longer flat (for NZ200 rpm), and to the change in flow regime (from DL to SL-up). For example for
because of the presence of many large air bubbles (especially for N ¼200 rpm, Po decreases from 4.7 to 3.9 by decreasing impeller
NZ300 rpm). submergence ratio from 0.77 to 0.59, i.e., as the flow transition
When surface air entrainment occurred in the SL-up Regime, occurred. Similar changes in power number and related mixing
‘‘flooding’’ of the impeller was also observed. Impeller flooding is parameters (such as the minimum agitation speed required for
a term commonly used in gas–liquid mixing systems (where a gas just solid suspension, Njs) have been reported before in the
is sparged below the impeller) when excessive bypassing of gas literature for the case in which a radial impeller was placed so
bubbles occurs around the impeller and the agitation speed is not low in the vessel (typically below Cb/Tffi 0.15) that the lower
high enough to disperse the bubbles throughout the vessel. As a recirculation in the bottom portion of the impeller was sup-
consequence, the bubbles swarm will go through the impeller pressed (Montante et al., 2001a, 2001b). For example, the Power
region and remain closer to the shaft (Fig. 10a). When impeller Number for disk turbines has been shown to drop from  5 for
flooding occurs in gas–liquid systems, the impeller power con- Cb/TZ0.17 to  4 for Cb/Tr0.13 (Armenante and Uehara Nagamine,
sumption is reduced drastically, the system loses its ability to 1998). It is interesting to notice that this drop is exactly the same
provide adequate gas hold-up, liquid pumping by the impeller is numerically as that which occurred when the top recirculation was
greatly reduced, and poor process results often result (Middleton altered as a result of lowering the impeller submergence, as found
and Smith, 2004; Middleton et al., 2005). In the system studied here when the DL to SL-up Regime transition occurred.
here, impeller flooding was observed only when Sb/D was below For higher agitation speeds i.e., N4200 rpm (corresponding to
the critical level (i.e., in the SL-up Regime) and for sufficiently Fr40.087), Fig. 11 shows that Po for impellers with submergence
high agitation. However, in this case flooding occurred as a result ratios below Sbcr/D (i.e., at Sb/D¼0.59 and Sb/D¼0.46) was much
of surface air entrainment, not gas sparging, and the large bubbles smaller not only of the Power Number for impellers with higher
responsible for it stayed mostly in the upper portion of the vessel submergence ratios but also than the Poffi 4 value found at lower
and especially around the impeller (Figs. 10b and c). agitation speeds with the same low submergence. For example, for
In order to describe the surface air entrainment process in N¼300 rpm (Fr¼0.195), Po decreased from 4.7 to 2.8 when Sb/D
greater detail, the Power Number was experimentally obtained was lowered from 0.77 to 0.59, and to 2 when Sb/D was lowered to
at different agitation speeds (i.e., 100 rpm, 200 rpm, 300 rpm, 0.46. Even bigger drops in Po could be observed as the agitation
400 rpm and 500 rpm) and for different submergence ratios speed was further increased (Fig. 11). The reduction in Po at low
(Sb/D equal to 0.46, 0.59, 0.77 and 0.98). Vortex phenomena in submergence ratios was associated with significant surface air
stirred vessels are usually interpreted based on the Froude entrainment, but this only happened because of the existence of
Number defined as: the partially segregated zone above the impeller observable only
when the SL-up Regime was in place, which became depleted of
N2 D
Fr ¼ ð7Þ fluid when the agitation speed was increased, thus resulting in
g
a vortex. It is interesting to notice that such a drastic reduction
Therefore, the Po data are reported here as a function of the Fr at in Po with Fr at low Sb/D ratios is similar to that observed in
the corresponding N (0.022, 0.087, 0.195, 0.347 and 0.542, submerged gas-sparged systems when the impeller flooding condi-
respectively), and are presented in Fig. 11. This figure shows that tion is approached as the gas flow rate is increased at constant
for impeller submergence ratios at, or higher than, the critical impeller speed (Middleton and Smith, 2004; Middleton et al., 2005).
level (i.e., for Sb/D ¼0.77 and 0.98; DL Regime), Po remained This implies that when the impeller is located below the
constant when the agitation speed, and hence Fr, increased. critical submergence level (Sbcr/D ¼0.77) and, at the same time,
However, for lower impeller submergence ratios associated with the Froude Number is above a critical value, Frcr, (taken con-
SL-up Regime (Sb/D ¼0.59 and 0.46), two operating regions servatively here as Frcr ¼0.087), the impeller undergoes flooding
corresponding to different levels of power reduction could be through surface air entrainment.

Fig. 11. Power Number as a function of Froude Number for different Sb/D ratios (Cb/T¼ 0.30 and D/T ¼0.31).
246 S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250

Fig. 12. Air–liquid interfaces predicted by VOF-CFD simulations at N ¼ 200 rpm (top panels) and 300 rpm (bottom panels) (Cb/T¼0.30 and D/T¼0.31). Left-hand side
panels: Sb/D ¼0.77 (H/T¼ 0.54; DL Regime); right-hand side panels: Sb/D ¼ 0.59 (H/T¼0.49; SL-up Regime).

In summary, the transition from the DL Regime to the SL-up mergence ratios, i.e., Sb/D ¼0.77 (H/T¼0.54) and Sb/D¼0.59 (H/
Regime resulting from lowering the submergence level is a T¼0.49), at 200 rpm and 300 rpm (based on experimental obser-
necessary but not sufficient requirement for surface air entrain- vations, modifications to the air–liquid interface for higher
ment. Air entrainment will result only if the agitation speed, and impeller submergence ratios were negligible). The contours of
hence the Froude Number, are high enough to generate a the air–liquid interface predicted by the VOF-CFD simulations are
sufficiently deep vortex that draws air into the liquid. If the shown in Fig. 12. For Sb/D ¼0.77 the simulations predicted a
agitation speed is further increased and the vortex reaches the nearly flat interface irrespective of the agitation speed, as experi-
impeller, impeller flooding will occur. In all cases in which the mentally observed. This is also consistent with the presence of the
flow transitions from DL Regime to SL-up Regime Po will drop. DL Regime. In such cases, the flow patterns predicted by the VOF-
However, this drop will be much more significant if surface CFD simulations were essentially identical to those predicted
aeration is additionally present. based on the flat surface assumption by FS-CFD in Fig. 3 (results
not shown). However, for Sb/D ¼0.59, i.e., when the SL-up Regime
5.3. Volume of fluid (VOF) simulation for critical impeller was controlling, the interface was no longer predicted to be flat,
submergence Ratios as shown in the right-hand side panels Fig. 12. Surface deforma-
tion was clearly present at N ¼200 rpm. For N ¼300 rpm the
The air–water interface is typically flat in baffled tanks if vortex at the air liquid interface had reached the impeller. The
H/T¼1 and, in many cases, even if H/To1. Therefore, the simulation was not able to predict actual bubbles, although a few
assumption of a flat interface was acceptable for most of the cells with significant presence of air are clearly visible below the
CFD-simulations conducted here. However, when the impeller surface in the lower, right-hand side panel in Fig. 12. However,
submergence ratio was decreased to a value below the critical this approach clearly predicted the correct vortex depth and
value, then the air–liquid interface was observed to form a vortex, shape. Visual observations confirmed that flooding was well
which modified the liquid surface and could result in surface air underway at this and higher speeds, as displayed in the (b) and
entrainment. In order to attempt to capture this phenomenon, the (c) right hand side panels in Fig. 10.
multiphase VOF simulation approach was additionally used here. Having validated the key feature of the SL-up Regime, the VOF
By generating the appropriate mesh at the interface between two approach was then used to predict additional hydrodynamic
phases, the VOF model could, in principle, be used to predict air characteristics of the flow at low impeller submergences. In
entrainment. However, such simulations are computationally Fig. 13, the radial and axial velocity CFD predictions obtained
intensive since in order to capture air bubbles in the liquid phase using VOF are compared with those obtained using FS, as well as
the mesh size for that zone must be some 10–20 times smaller with the corresponding PIV measurements at Sb/D ¼0.59 shown
than the real bubble size, thus requiring long time for convergence above in Fig. 7. In general, the VOF-CFD predictions were found to
if the real phenomena is to be captured. In this work, a mesh size be in better agreement with the experimental data especially in
of 1 mm was used around the air–liquid interface to ensure that the lower portion of the tank (Z/T¼0.04, 0.19 and 0.27), i.e., even
the mesh quality was within a range acceptable for VOF simula- in regions far away from the upper region where the vortex
tions of the shape of the interface, although clearly too large to formed. Impeller discharge angles, Pumping Numbers and Power
capture the presence of individual bubbles. Simulating the hydro- Numbers were also calculated using VOF-CFD simulations for
dynamics in stirred tank with Sb/D¼0.59, N¼ 300 rpm and using Sb/D ¼0.77 and 0.59 at 300 rpm and are reported in Fig. 14. This
the above-mentioned mesh size and 650,000 cells required at least figure shows that the VOF-CFD and FS-CFD predictions were in
400,000 iterations to converge, especially in those cases in which good agreement with each other and with the experimental
the interface was significantly deformed. values for Sb/D ¼0.77, when the DL Regime was dominant and
VOF simulations incorporating the MRF approach (VOF-CFD no significant vortex could be observed. However, for Sb/D¼ 0.59,
simulations) were conducted for two different impeller sub- i.e., when the SL-up Regime was in place, the VOF-CFD predictions
S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250 247

Fig. 13. Comparison between radial velocities (left-hand side panels) and axial velocities (right-hand side panels) predicted by VOF-CFD and FS-CFD simulation methods
on five different horizontal iso-surfaces at different vertical Z/T locations for Sb/D ¼ 0.59 (H/T¼ 0.49; SL-up Regime) (Cb/T¼ 0.30, D/T ¼0.31 and N ¼ 300 rpm). PIV
experimental results for are also presented.

were in better agreement with the experimental data compared 3. A critical impeller submergence ratio, Sbcr, was identified for
to the FS-CFD simulations. As expected, the VOF method was able the system under investigation (Cb/T¼0.30, D/T¼ 0.31). It was
to predict the flow transition more accurately. conservatively assumed that Sbcr/D ¼0.77. The following flow
pattern transitions were then observed:
a. for Sb/D ZSbcr/D, the disc turbine produced a standard
6. Conclusions ‘‘double-loop’’ recirculation regime (DL Regime) in which
the radial jet emanating from the impeller impacted the
The following conclusions can be drawn from this work: vessel wall resulted in a recirculation loop above the
impeller and another one below it;
1. The CFD predictions of the velocity components, Power Num- b. for Sb/DoSbcr/D, the disc turbine produced a ‘‘single-loop-up-
bers, and Pumping Numbers obtained in this work for a pumping’’ recirculation flow (SL-up Regime), creating not
conventional stirred vessel (H/T¼1 and C¼ T/3), were in sig- only a single recirculation flow but also a partially segregated
nificant agreement with experimental measurements obtained core region just above the impeller. This is the first time that
here and with data available in the literature (Poffi5; NQ ffi0.8). this regime has been reported and characterized.
2. The impeller submergence ratio Sb/D, which is a linear func- 4. Whenever Sb/D oSbcr/D, the flow regime transitioned from DL
tion of the H/T ratio for a fixed impeller off-bottom clearance, to SL-up and the following phenomena occurred:
can play a critical role on the flow regimes produced in stirred a. the Power Number and the radial Pumping Number dropped
vessels when H/To1. significantly in correspondence of the flow transition;
248 S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250

b. For Fr4Frcr, a vortex began to appear above the impeller.


The depth of this vortex rapidly increased with Fr and
resulted in significant surface air entrainment that flooded
the impeller. Whenever this happened Po dropped to values
between 1 and 3, depending on the agitation speed, and NQ
dropped to  0.2.
7. It can be concluded that lowering the submergence level
below Sbcr is a necessary but not sufficient requirement for
significant surface air entrainment and impeller flooding.
Impeller flooding can occur only if:
a. Sb/D oSbcr/D, i.e., the impeller operates under the SL-up
regime; and
b. Fr 4Frcr, i.e., the agitation speed, is high enough to generate
a sufficiently deep vortex that draws air into the liquid.
8. The velocity profiles contours, Power Number and Pumping
Number at different impeller submergence ratios predicted by
FS-CFD simulations were in a good agreement with those
measured experimentally through PIV. VOF-CFD simulations
were able to predict values that were in even closer agreement
with the experimental data for lower Sb/D ratios including
when surface deformation occurred (Sb/D ¼0.59).

The results of this work indicate that the impeller may lose its
effectiveness as a mixing device even when it is still fully
submerged, i.e., long before it becomes exposed to air. This may
be very relevant to industrial processes in which the liquid level is
dropped as a result of operations involving partially emptying the
vessel. This knowledge can help practitioners avoid operating
their equipment in regions where the desired mixing effects are
not achievable.

Nomenclature

C distance from bottom of the impeller disk and tank


bottom, (m)
Cb distance from bottom of the impeller blades and tank
bottom, (m)
Cb/T impeller off-bottom clearance ratio, dimensionless
CV cylindrical control volume around impeller
D impeller diameter, (m)
DCV diameter of the cylindrical control volume around
impeller, (m)
Fr Froude Number, dimensionless
Fig. 14. Comparison between predictions of VOF-CFD and FS-CFD simulations Frcr Critical Froude Number, dimensionless
and experimental results for Sb/D ¼ 0.77 (H/T¼ 0.54; DL Regime); and Sb/D¼ 0.59
(H/T¼ 0.49; SL-up Regime) (Cb/T¼ 0.30 and D/T¼ 0.31 and N ¼300 rpm):
g gravitational acceleration, (m s  2)
(a) Impeller discharge angle; (b) radial Pumping Number, NQr; and (c) Power H maximum liquid level, (m)
Number, Po. H/T fill ratio, dimensionless
Hcr critical maximum liquid level, (m)
b. the impeller discharge angle, defined as the angle between N agitation speed, (rpm)
the direction of average vector of the radially discharge NE agitation speed at onset of air entrainment, (rpm)
stream produced by a radial impeller and the horizontal NQ Pumping Number or Flow Number, dimensionless
plane, suddenly increased from some 41–71 (DL Regime) to NQr radial Pumping Number, dimensionless
about 601 (SL-up Regime). p pressure, (N m  2)
5. The transition from DL Regime to SL-up Regime was found to P instantaneous power dissipated by the impeller, (J s  1)
be only a function of the Sb/D ratio (and possibly other Po Power Number, dimensionless
geometric ratios), but not of agitation intensity. Q total discharge flow rate from impeller, (m3 s  1)
6. For Sb/D oSbcr/D (SL-up Regime), the Froude Number, Fr, and Qr radial discharge flow rate from impeller, (m3 s  1)
hence the agitation speed, were found to be important in the r radial position, (m)
description of the mixing performance of the impeller. A R tank radius, (m)
critical value of Fr, i.e., Frcr (where Frcr ¼0.087) was identified, Re Reynolds Number, dimensionless
defining two sub-operating ranges: Sb impeller submergence (distance from bottom of the
a. For FrrFrcr, the Power Number dropped to  4 (from 5 in impeller blades to the air–liquid interface), (m)
the DL Regime), but no significant surface air entrainment Sb/D impeller submergence ratio, dimensionless
was observed; Sbcr critical impeller submergence, (m)
S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250 249

T tank diameter, (m) Chapple, D., Kresta, S.M., Wall, A., Afacan, A., 2002. The effect of impeller and tank
U velocity, (m s  1) geometry on power number for a pitched blade turbine. Chem. Eng. Res. Des.
80 (4), 364–372.
Urz velocity magnitude on r–z surface, (m s  1) Coroneo, M., Montante, G., Paglianti, A., Magelli, F., 2011. CFD prediction of fluid
Utip velocity at the tip of the impeller, (m s  1) flow and mixing in stirred tanks: numerical issues about the RANS simula-
wCV height of the cylindrical control volume around tions. Comput. Chem. Eng. 35 (10), 1959–1968.
Deglon, D.A., Meyer, C.J., 2006. CFD modeling of stirred tanks: numerical con-
impeller, (m) siderations. Miner. Eng. 19 (10), 1059–1068.
z axial position, (m) Dyster, K., Koutsakos, E., Jaworski, Z., Nienow, A., 1993. A LDA study of the radial
discharge velocities generated by a Rushton turbine: Newtonian fluids, ReZ 5.
Chem. Eng. Res. Des. 71, 11–23.
Gueyffier, D., Li, J., Nadim, A., Scardovelli, R., Zaleski, S., 1999. Volume-of-fluid
Greek letters interface tracking with smoothed surface stress methods for three-dimen-
sional flows. J. Comput. Phys. 152 (2), 423–456.
Haque, J., Mahmud, T., Roberts, K., 2006. Modeling turbulent flows with free
a volume fraction of a given phase in each cell, dimensionless surface in unbaffled agitated vessels. Ind. Eng. Chem. Res. 45 (8), 2881–2891.
G torque, (Nm) Horowitz, A., 2010. Is the reign of batch processing coming to an end? Life Sci. Leader.
Hirt, C.W., Nichols, B.D., 1981. Volume of fluid (VOF) method for the dynamics of
m viscosity of water, (Pa s) free boundaries. J. Comput. Phys. 39 (1), 201–225.
r liquid density, (kg m  3) Ibrahim, S., Nienow, A.W., 1996. Particle suspension in the turbulent regime: the
t shear stress on impeller surfaces, (N/m2) effect of impeller type and impeller/vessel configuration. Chem. Eng. Res. Des.
74 (6), 679–688.
Ibrahim, S., Nienow, A.W., 1999. Comparing impeller performance for solid-
Subscripts suspension in the transitional flow regime with Newtonian fluids. Chem.
Eng. Res. Des. 77 (8), 721–727.
Joshi, J.B., Pandit, A.B., Sharma, M.M., 1981. Mechanically agitated gas–liquid
A air reactors. Chem. Eng. Sci. 37 (6), 813–844.
i counter in summation Li, Z., Bao, Y., Gao, Z., 2011. PIV experiments and large eddy simulations of single-
loop flow fields in Rushton turbine stirred tanks. Chem. Eng. Sci. 66 (6),
j counter in summation 1219–1231.
r radial direction Lane, G.L., Koh, P.T.L., 1997. CFD simulation of a Rushton turbine in a baffled tank.
z axial direction Int. Conf. CFD Min. Met. Process. Power Gener., 377–385.
Mahmoudi, S.M., Yianneskis, M., 1992. The variation of flow pattern and mixing
time with impeller spacing in stirred vessels with two Rushton impellers. In:
King, R. (Ed.), Fluid Mechanics of Mixing. Kluwer Academic Publishers,
Acknowledgement Dordrecht, The Netherlands, pp. 11–18.
Mahmud, T., Haque, J., Roberts, K., Rhodes, D., Wilkinson, D., 2009. Measurements
and modeling of free-surface turbulent flows induced by a magnetic stirrer in
The authors would like to thank Merck & Co. for donating one an unbaffled stirred tank reactor. Chem. Eng. Sci. 64 (20), 4197–4209.
of the two PIV systems used in this work, and in particular Scott Mali, R.G., Patwardhan, A.W., 2009. Characterization of onset of entrainment in
Reynolds, Marc Steinman, and Stephen Pafiakis. One of us (SM) stirred tanks. Chem. Eng. Res. Des. 87 (7), 951–961.
Marshall, E.M., Bakker, A., 2004. Computational fluid mixing. In: Paul, E.L., Atiemo-
wishes to express her appreciation for the financial support Obeng, V.A., Kresta, S.M. (Eds.), Handbook of Industrial Mixing. Wiley,
provided by the Otto H. York Department of Chemical, Biological Hoboken, New Jersey, pp. 292–294.
and Pharmaceutical Engineering at the New Jersey Institute of Middleton, J.C., Smith, J.M., 2004. Gas–liquid mixing in turbulent systems. In: Paul,
E.L., Atiemo-Obeng, V.A., Kresta, S.M. (Eds.), Handbook of Industrial Mixing.
Technology in order to complete her Ph.D. dissertation. Wiley, Hoboken, NJ, pp. 599–601.
Middleton, J.C., Smith, J.M., Armenante, P.M., 2005. Gas–liquid mixing in agitated
reactors. In: Sunggyu Lee (Ed.), Encyclopedia of Chemical Processing. Taylor &
References Francis, Boca Ratton, FL, pp. 1131–1141.
Montante, G., Lee, K.C., Brucato, A., Yianneskis, M., 2001a. Experiments and
Akiti, O., Armenante, P.M., 2004. Experimentally-validated micromixing-based predictions of the transition of the flow pattern with impeller clearance in
CFD model for field-batch stirred-tank reactors. AlChE J. 50 (3), 566–577. stirred tanks. Comput. Chem. Eng. 25 (4–6), 729–735.
Armenante, P.M., Huang, Y.-T., Li, T., 1992. Determination of the minimum Montante, G., Lee, K.C., Brucato, A., Yianneskis, M., 2001b. Numerical simulations
agitation speed to attain the just dispersed state in solid–liquid and liquid– of the dependency of flow pattern on impeller clearance in stirred vessels.
liquid reactors provided with multiple impellers. Chem. Eng. Sci. 47 (9–11), Chem. Eng. Sci. 56 (12), 3751–3770.
2865–2870. Murthy, B.N., Joshi, J.B., 2008. Assessment of standard k–e, RSM and LES
Armenante, P.M., Chou, C.C., 1996. Velocity profiles in a baffled vessel with single turbulence models in a baffled stirred vessel agitated by various impeller
or double pitched-blade turbines. AlChE J. 42 (1), 42–54. designs. Chem. Eng. Sci. 63 (22), 5468–5495.
Armenante, P.M., Luo, C., Chou, C.C., Fort, I., Medek, J., 1997. Velocity profiles in a Nienow, A.W., 1968. Suspension of solid particles in turbine agitated baffled
closed, unbaffled vessel: comparison between experimental LDV data and vessels. Chem. Eng. Sci. 23 (12), 1453–1459.
numerical CFD predictions. Chem. Eng. Sci. 52 (20), 3483–3492. Ochieng, A., Onyango, M.S., Kumar, A., Kiriamiti, K., Musonge, P., 2008. Mixing in a
Armenante, P.M., Uehara Nagamine, E., 1998. Effect of low off-bottom impeller tank stirred by a Rushton turbine at a low clearance. Chem. Eng. Process. 47,
clearance on the minimum agitation speed for complete suspension of solids 842–851.
in stirred tanks. Chem. Eng. Sci. 53 (9), 1757–1775. Patwardhan, A.W., Joshi, J.B., 1998. Design of stirred vessels with gas entrained
Armenante, P.M., Uehara Nagamine, E., Susanto, J., 1998. Determination of from free liquid surface. Can. J. Chem. Eng. 76 (3), 339–365.
correlations to predict the minimum agitation speed for complete solid Paul, E.L., Midller, M., Sun, Y., 2004. Mixing in the fine chemicals and pharmaceu-
suspension in agitated vessels. Can. J. Chem. Eng. 76 (3), 413–419. tical industries. In: Paul, E.L., Atiemo-Obeng, V.A., Kresta, S.M. (Eds.), Hand-
Aubin, J., Fletcher, D.F., Xuereb, C., 2004. Modeling turbulent flow in stirred tanks book of Industrial Mixing. Wiley, Hoboken, NJ, pp. 1027–1064.
with CFD: the influence of the modeling approach, turbulence model and Pilliod, J.E., Puckett, E.G., 2004. Second-order accurate volume-of-fluid algorithms
numerical scheme. Exp. Therm. Fluid Sci. 28 (5), 431–445. for tracking material interfaces. J. Comput. Phys. 199 (2), 465–502.
Baldi, G., Conti, R., Alaria, E., 1978. Complete suspension of particles in mechani- Rushton, J.H., Costich, E.W., Everett, H.J., 1950. Power characteristics of mixing
cally agitated vessels. Chem. Eng. Sci. 33 (1), 21–25. impellers. Chem. Eng. Prog. 46 (8), 395–476.
Bai, G., Armenante, P.M., Plank, R.V., Gentzler, M., Ford, K., Harmon, P., 2007. Rutherford, K., Mahmoudi, S.M., Lee, K.C., Yianneskis, M., 1996. The influence of
Hydrodynamic investigation of USP dissolution test apparatus II. J. Pharm. Sci. Rushton impeller blade and disk thickness on the mixing characteristics of
96 (9), 2327–2349. stirred vessels. Trans IChemE 74 (3), 369–378.
Bai, G., Wang, Y., Armenante, P.M., 2011. Velocity profiles and shear strain rate Serra, A., Campolo, M., Soldati, A., 2001. Time-dependent finite-volume
variability in the usp dissolution testing apparatus 2 at different impeller simulation of the turbulent flow in a free-surface CSTR. Chem. Eng. Sci. 56
agitation speeds. Int. J. Pharm. 403, 1–14. (8), 2715–2720.
Bhattacharya, S., Hebert, D., Kresta, S.M., 2007. Air entrainment in baffled stirred Sheng, J., Meng, H., Fox, R.O., 2000. A large eddy PIV method for turbulence
tanks. Chem. Eng. Res. Des. 85 (5A), 654–664. dissipation rate estimation. Chem. Eng. Sci. 55 (20), 4423–4434.
Brown, D.A.R., Jones, P.N., Middleton, J.C., Papadopoulos, G., Arik, E.B., 2004. Experi- Tanaka, M., Izumi, T., 1987. Gas entrainment in stirred-tank reactors. Chem. Eng.
mental methods. In: Paul, E.L., Atiemo-Obeng, V.A., Kresta, S.M. (Eds.), Handbook Res. Des. 65 (2), 195–198.
of Industrial Mixing. Wiley, Hoboken, New Jersey, pp. 161 and 237–249. Veljkovic, V.B., Bicok, K.M., Simonovic, D.M., 1991. Mechanism, onset and intensity
Campolo, M., Paglianti, A., Soldati, A., 2002. Fluid dynamic efficiency and scale-up of surface aeration in geometrically-similar, sparged, agitated vessels. Can.
of a retreated blade impeller CSTR. Ind. Eng. Chem. Res. 41 (2), 164–172. J. Chem. Eng. 69 (4), 916–926.
250 S. Motamedvaziri, P.M. Armenante / Chemical Engineering Science 81 (2012) 231–250

Wu, H., Patterson, G.K., 1989. Lasser-Doppler measurements of turbulent-flow Yianneskis, M., Popiolek, Z., Whitelaw, J.H., 1987. An experimental-study of the
parameters in a stirred mixer. Chem. Eng. Sci. 44 (10), 2207–2221. steady and unsteady-flow characteristics of stirred reactors. J. Fluid Mech. 175,
Yang, T., Takahashi, K., 2010. Effect of impeller blades angle on Power Number and 537–555.
flow pattern in horizontal stirred vessel. J. Chem. Eng. Jpn. 43 (8), 635–640. Zhao, J., Gao, Z., Bao, Y., 2011. Particle image velocimetry study of flow patterns
Yeoh, S.L., Papadakis, G., Yianneskis, M., 2004. Numerical simulation of turbulent and mixing characteristics in multiple impeller stirred tank. J. Chem. Eng. Jpn.
flow characteristics in a stirred vessel using the LES and RANS approaches with 44 (6), 389–397.
the sliding/deforming mesh methodology. Chem. Eng. Res. Des. 82 (7),
834–848.

Anda mungkin juga menyukai