You are on page 1of 20

Characteristics of Batch Rotor–Stator Mixer

Performance Elucidated by Shaft Torque and


Angle Resolved Piv Measurements
Hans Henrik Mortensen,1 * Richard V. Calabrese,2 Fredrik Innings3
and Lasse Rosendahl4
1. Tetra Pak Scanima, Gugvej 152B, 9210 Aalborg East, Denmark
2. Department of Chemical & Biomolecular Engineering, University of Maryland, College Park, Maryland 20742-2111
3. Tetra Pak Processing Systems, Research & Technology Bryggaregatan 23, 22186 Lund, Sweden
4. Institute of Energy Technology, Aalborg University, Pontoppidanstraede 101, 9220 Aalborg East, Denmark

Characteristics of batch rotor–stator mixer performance are elucidated by shaft torque and angle resolved 2D PIV measurements obtained in a
full-scale, custom build, bottom-mounted, rotor–stator mixer unit operating in the turbulent regime with water as working fluid. Measurements
have been acquired at various rotor speeds corresponding to impeller based Reynolds numbers between 2.0 × 105 and 8.5 × 105 . The use of a
transparent Plexiglas stator facilitated PIV measurements inside and outside the stator as well as into the stator slots themselves. The governing
mechanisms controlling the complex flow structures, flow rates, power dissipation, velocity fields, strain rate fields and turbulence intensity fields
are explained, highlighting the influence on rotor–stator mixer performance. The results indicate that scale-up of mixing processes that depend
on macro-scale phenomena should be based on constant rotor tip speed.

Keywords: rotor–stator mixer, PIV, cavitation, mean velocity, strain rate, turbulence

INTRODUCTION In the past decade, several researchers have tried to apply


known theory and mechanistic models from conventional stirrers

R
otor–stator mixers are commonly employed in the pro-
to rotor–stator mixers despite the fact that only one geometri-
cess industries to carry out liquid–liquid homogenisation,
cal parameter (the rotor diameter) is included. Padron (2001)
dispersion and emulsification as well as solid–liquid dis-
measured the power draw of three different lab-scale rotor–stator
persion, dissolving and grinding. A variety of different designs
batch mixers, namely a Ross ME100LC, a Silverson L4R and a
exist but their operating principle is basically similar. Stator ele-
Greerco XLR Homogeniser. Using conventional impeller defini-
ments surround a high-speed rotor causing a complex flow pattern
tions of Reynolds and power number, the power number was
with high velocity gradients and turbulence. The local shear and
found to be inversely proportional to the Reynolds number in
energy dissipation rates are typically several orders of magni-
the laminar regime and approximately constant in the turbulent
tude higher than in conventional stirred vessels (Meyers et al.,
regime; similar to what is seen for conventional stirrers. However,
1999; Atiemo-Obeng and Calabrese, 2004). Despite their com-
he also found that the power number changed significantly with
mon application, there are only few archival publications about
these devices and the user is left with little basis to theoretically
predict or experimentally assess their performance. Design and ∗ Author to whom correspondence may be addressed.
scale-up of a given process is further complicated by the var- E-mail address: hanshenrik.mortensen@tetrapak.com
ious geometrical parameters introduced by the presence of the Can. J. Chem. Eng. 89:1076–1095, 2011
stator, such as geometry, size and number of stator openings, sta- © 2011 Canadian Society for Chemical Engineering
DOI 10.1002/cjce.20587
tor thickness, rotor–stator clearance and number of concentric Published online 21 June 2011 in Wiley Online Library
stages. (wileyonlinelibrary.com).

| 1076 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 89, OCTOBER 2011 |
mixer and stator type and in some degree with rotor–stator clear- scaled with the cube of rotor speed. The results showed good
ance gap. For a specific stator type having slotted openings, the agreement with comparative LDA measurements in the stator jets
turbulent power number increased with the number of stator slots and bulk flow. Later, Utomo et al. (2009) extended this study to
while the power number per stator slot remained almost constant. include the effect of three different stator geometries. For narrow
It was therefore concluded that energy dissipation is controlled by stator openings, the simulation revealed a bulk flow field rotating
fluid impingement on stator slot surfaces or by turbulence in the opposite to the impeller. Calculated power numbers were found
emanating stator jets. Power draw was also measured by Doucet to be proportional to the flow rate while flow rate correlated with
et al. (2005) who carried out experiments on a VMI lab-scale the total stator open area. The energy dissipated in the jet and
rotor–stator batch mixer operating in both viscous Newtonian rotor swept volume depended linearly of the flow rate while the
and shear-thinning fluids (laminar regime). Again using conven- energy dissipated in the stator openings correlated with the total
tional Reynolds and power number definitions, the overall shape surface area of the leading and trailing stator opening edges.
of the power number curves resembled the findings of Padron Given the complex nature of the flow field, the numerous pos-
(2001). Several approaches were applied to establish a fluid inde- sible geometrical parameter combinations, the variety of different
pendent power curve and, although the classical Metzner and Otto product applications and the complicated dispersion mechanisms
approach seemed applicable, it was not possible to determine a involved, a general characterisation of rotor–stator mixer perfor-
unique value of the Metzner-Otto constant, limiting the practical mance by simple mechanistic models does not seem realistic. On
application of the approach. In the same study, an attempt was the other hand, even relatively simple numerical simulations have
made to predict the cavern size (well mixed region surrounded proven to provide reliable results. Development of robust and vali-
by stagnant fluid) but correlations from the literature, used dated CFD-based models therefore seems to be a trustworthy path
for conventional impellers, were unable to fit the experimental forward, especially in the light of rapid advancements in com-
data. puter speed and numerical methods. Regarding the evaluation
Investigation of dispersion mechanisms and prediction of equi- and prediction of emulsification performance, single-phase simu-
librium drop size distributions in dilute liquid–liquid rotor–stator lations will enable utilisation of local strain and energy dissipation
emulsification processes has been the objective of several studies. rates, quantities that theoretically relate to drop break-up. More
The theory and mechanistic models known for conventional stir- sophisticated multiphase simulations could even predict coales-
rers were applied. Francis (1999) obtained experimental drop size cence or break-up phenomena. Another perspective using CFD
data for inviscid dispersions using the same Ross and Silverson is the possibility to improve the rotor–stator equipment itself
lab-scale mixers as Padron (2001). The mixers were fitted with in a rational way. Validated models could serve as a numerical
similar stators and the data were well correlated by the mech- laboratory for the design engineer trying to develop new and
anistic Weber number model originally proposed by Chen and more efficient equipment. Although computationally expensive,
Middleman (1967). The study was continued by Phongikaroon the prospect of automated shape optimisation is realistic, Lund et
(2001) to elucidate the effect of drop viscosity and stator geome- al. (2003), and fluid–structure interaction, dynamic meshes and
try. The data were best fit by a modified Weber number model that embedded optimisations algorithms are in fact starting to appear
accounts for drop viscosity, first derived by Wang and Calabrese in commercial software packages. Motivated by this perspective,
(1986). The influence of different stator geometry was clearly seen the primary objective of this study is to provide the insight and val-
in the data. An additional experiment was carried out by Cal- idation base for subsequent CFD modelling of a Tetra Pak Scanima
abrese et al. (2002) using the previously mentioned Ross mixer. (TPS) batch rotor–stator mixer. Shaft torque and high resolution
Two almost identical stators were tested, one of them machined to 2D angularly resolved PIV measurements have been obtained in a
double the rotor–stator clearance. Surprisingly it was found that full-scale, custom built mixer unit with water as the working fluid.
the larger clearance gap provided equal or smaller drop sizes. By use of a Plexiglas stator, PIV measurements were obtained
Numerical calculations has been carried out by Calabrese et inside (rotor region) and outside the stator as well as in the stator
al. (2002) who presented a 2D transient CFD simulation of a slots. Opposed to other known batch rotor–stator mixer studies,
prototype IKA rotor–stator in-line mixer with prescribed water this work has been carried out on a full-scale unit eliminating
throughput. The RANS equations were solved using a stan- the uncertainty involved due to improper scaling; an important
dard k-ε turbulence model and the sliding mesh method was detail, since even geometric similarity is often violated in scaled
used to accommodate the rotor–stator interface. The simulation down rotor–stator laboratory equipment.
revealed an extremely complex flow pattern with circulation cells
in each stator slot and re-entrainment of volute fluid back into
the shear gap. For validation purposes, 2D LDA measurements METHODS AND MATERIALS
were acquired in the same device and although the measured
velocity field showed reasonable qualitative agreement with the Mixer System
simulation, the data contained significant quantitative differences. In its simplest form, a TPS batch mixer is comprised of a baffled
Barailler et al. (2006) carried out a laminar, full 3D transient mixing vessel with an integrated bottom-mounted, centre-located,
CFD simulation of the same VMI mixer used by Doucet et al. rotor–stator mixing unit. Unfortunately, bearings and drive com-
(2005). The calculated shaft torque showed good agreement with ponents located beneath the vessel prevent space and optical
experimental measurements and the presence of caverns, exper- access for lasers and cameras. Therefore, a custom built mixer
imentally determined by Doucet et al. (2005), were also seen in has been developed for use in this and subsequent experiments.
the simulations. Recently, Utomo et al. (2008) studied the flow A detailed description and mechanical considerations are beyond
pattern and energy dissipation rate in a lab-scale Silverson L4RT the scope of this article, so only a brief presentation of relevant
batch rotor–stator mixer using a full 3D sliding mesh RANS sim- features will be given. The prototype mixer, as well as the camera
ulation and a standard k-ε turbulence model. The calculated jet and laser set-up, is shown in Figure 1.
velocities and flow rates through the stator openings were found A top entry drive configuration plus an internal rotor hub-
to be proportional to the rotor speed while energy dissipation rate bearing system are used to maximise the free space beneath the

| VOLUME 89, OCTOBER 2011 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 1077 |
vessel. The optical access for the camera is secured by a glass imise reflections, all internal metal surfaces are coated with flat
window (pos. 1) in the flat base plate while another glass win- black paint.
dow, mounted in a horizontal nozzle (pos. 2) allows the laser The shaft torque is supplied by a top mounted, 22 kW, fre-
sheet to enter the vessel. A transparent Plexiglas stator (pos. 3) quency controlled e-motor (pos. 5) and transmitted to the rotor
permits illumination into the interior rotor swept volume. To min- via a torque transmitter (pos. 6), an intermediate shaft (pos. 7)

Figure 1. The prototype mixer used in the experiments. The PIV laser and camera are also shown.

| 1078 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 89, OCTOBER 2011 |
and three flexible bellows couplings (pos. 8). The measurement
range of the torque transmitter (type HBM TWN20) is 0–100 Nm
with an accuracy of 0.2 Nm. Precise rotor speed measurements
are facilitated by a drive shaft encoder (pos. 9) giving 360 pulses
per revolution. A secondary encoder (pos. 10) giving 1 pulse per
revolution is used in combination with a timer to reference the
rotor position and to trigger the laser (pos. 16) and camera (pos.
18), allowing images to be recorded at any desired rotor angle.
The 220-L vessel (pos. 11) is filled by gravity from a deionised
water storage tank (not shown). Properly located ventilation
valves (pos. 12) prevent entrapment of air that otherwise might
cause reflections and noise in the PIV recordings. The vessel
design and a small interconnected buffer tank with a regulated
compressed air supply (not shown) allows the system to be pres-
surised up to 1.5 bar, a feature that can be utilised to prevent
reflections from dissolved air expanding in low pressure regions,
and to filter out or investigate the influence of cavitation. The exte-
rior vessel surface and the four baffle plates are fitted with serial
connected cooling jackets (pos. 13) in order to control the ves-
sel’s water temperature. The cooling water flow is regulated by a
Figure 2. Fields of view and coordinate systems employed in the PIV
modulating valve (not shown) based on input from a temperature experiments.
sensor (pos 14) located inside the vessel. A syringe injection port
(pos. 15) allows easy injection of the seeding particle suspension.
The design allows different rotor–stator sizes and types to be Fields of View
mounted in the same base-plate. Important dimensions of the Two different fields of view are employed in these experiments.
rotor (pos. 4) and stator (pos. 3) used in these experiments are The size and locations are shown on Figure 2. The larger field of
shown in Figure 1. The diameter of the vessel is 650 mm and the view (FOV2) is used to capture the overall flow features within-
width of the baffles is 80 mm. and outside the stator while the smaller field of view (FOV1) is
used to enhance the spatial resolution in the high shear intensity
stator slot and nearby area. The applied magnification factor, M,
PIV System is given by the ratio WCCD /WFOV , where WCCD is the width of the
Again, refer to Figure 1. The light sheet is generated by a New CCD chip and WFOV is the width of the field of view. By recording
Wave Gemini PIV-15 laser (pos. 16) containing two separate Q- images of a high precision ruler, placed in the centre of the laser
switched Nd:YAG cavities, some beam combining optics, a second sheet, the magnification factors corresponding to FOV1 and FOV2
harmonic generator and an optical attenuator. The dual cavity were found to be 0.606 ± 0.002 and 0.1935 ± 0.0002 respectively.
system provides maximum user control over the pulse separation
time while Q-switching allows the light energy to be emitted in
ultra short pulses facilitating the recording of un-blurred frozen Coordinate Systems and Measurement Plane
images. The pulse duration is 10 ns and the maximum repeti-
tion rate for successive pulse pairs is 15 Hz. By use of the optical To facilitate description and data analysis, two different coordinate
attenuator the laser pulse energy can be adjusted between 0 and systems are defined. A global cylindrical coordinate system (, r)
120 mJ. For eye safety, a second harmonic generator is utilised with origin placed in the centre of the base plate and a local Carte-
to produce 532-nm visible green light by doubling the frequency sian coordinate system (x, y) with origin placed in the centre of the
of the primary infrared beams. The emitted beam is transformed viewed stator slot—equal to (30◦ , 94.65 mm) in the global coor-
into a light sheet by a frontal lens system (pos. 17) that allows dinate system. The former is used to reference the angular rotor
adjustment of the laser sheet thickness. The complete assembly is position, while the latter is used for data analysis. Both coordi-
mounted on a vertical and horizontal traverse system (not shown) nate systems are shown in Figure 2. The horizontal measurement
providing easy adjustment of the laser sheet position. plane is located 18 mm above the flat base plate corresponding to
The image recording is done by a Dantec 80C60 HiSense cross midway between top and bottom surfaces of the stator slots.
correlation camera (pos. 18) mounted with a 60 mm Nikkor lens
(pos. 19) with adjustable aperture (f-number between 2.8 and 32). Seeding
The image sensor consists of a Peltier cooled 1280 × 1024 pixel The fluid motion in a rotor–stator mixer is characterised by strong
CCD chip having a 6.7 micron pixel pitch and 12 bit greyscale res- and alternating acceleration fields. To ensure proper flow track-
olution. The frame straddling technique is used to exploit the short ing, the use of small diameter tracer particles with fluid-matched
pulse separation time allowed by the dual cavity laser. The camera density is of utmost importance. For that reason and due to their
and lens are rigidly mounted beneath the vessel with the optical acceptable light scattering properties, polyamide particles with an
axis perpendicular to the light sheet. average diameter of 5 ␮m and a specific gravity of 1.03 are used. A
Data acquisition is controlled via Dantec FlowManager soft- particle/ethanol suspension was injected into the water filled ves-
ware v.5.4.71 and a Dantec FlowMap 2500 processer unit (not sel while the mixer was running to ensure a fast and homogenous
shown). The former provides the user interface for system set-up, distribution. The amount of seeding was adjusted in an iterative
database handling and analysis control, while the latter contains procedure, together with laser pulse energy, laser pulse separation
dedicated correlator units, frame grabbers and buffers, as well as time and interrogation volume, in order to obtain a high vector val-
synchronisation systems for the laser, camera and trigger signals. idation rate. The final seeding concentration resulted in more than

| VOLUME 89, OCTOBER 2011 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 1079 |
five and usually about 10 particle images per interrogation area, Measurement Accuracy and Spatial Resolution
consistent with the recommendation by Keane and Adrian (1990). Estimation of the accuracy of a given PIV experiment is far from
trivial. The list of errors is long and their influence is often com-
Particle Image Formation plex. Monte–Carlo simulations, based on synthetic PIV images
Using a three-point Gaussian peak estimator for sub-pixel interpo- with known displacement fields, have often been used to quantify
lation, particle images above 2 pixels were found to be essential these errors (e.g. Westerweel, 1993). However, today it is widely
to avoid peak locking effects. Unfortunately, selection of lens f- accepted that the particle image position can be determined within
numbers above 5.6 was unfeasible due to the limited laser power 0.1 pixel in properly set-up experiments (Raffel et al., 2007).
available; so it was therefore necessary to slightly defocus the Design rules for set-up, mainly formulated by Keane and Adrian
image. To avoid touching the duly positioned camera, the laser (1990, 1992) provide requirements for particle image density,
sheet was moved instead; about 1.5 mm upwards. Using this tech- displacement and diameter, as well as particle image displace-
nique together with an f-number of 5.6, particle images of 2 and ment gradients. In this experiment, the set-up is close to optimal.
3 pixels were obtained; this being in compliance with the opti- Therefore, using 16 × 16 pixel interrogation areas and enforcing
mal particle image diameter recommended by several authors, the one-quarter-rule (particle image displacement should be less
for example Adrian (1991); and Westerweel (2000). The required than or equal to one-quarter of the interrogation area width), a
laser attenuation for this set-up was about 85% of maximum, dynamic velocity range (ratio of maximum to minimum resolv-
corresponding to approximately 100 mJ per pulse. The laser sheet able velocity or displacement) of about 40 and an accuracy of
thickness was about 1 mm. about 2.5% of full scale should be expected. However, for our
case, the use of an adaptive correlation algorithm allowed captur-
Cross-Correlation and Validation ing displacements of approximately 7.5 pixels, thus providing a
The laser pulse separation time was adjusted to obtain displace- dynamic velocity range of about 75 and accuracy of about 1.4%
ments of 4 pixels for the rotor tip at each speed setting. Using an of full scale. It must be emphasised though, that additional uncer-
adaptive correlation or often named multigrid algorithm, that is tainties concerning rotor speed, fluid temperature and geometrical
Scarano and Riethmuller (1999), allowed satisfactory capture of dimensions must be taken into account when comparing these
higher stator jet velocities. The final interrogation area was set to results with other experiments or CFD results.
16 × 16 pixels, with 50% overlap between each area, resulting in Based on the number of non-overlapping interrogation areas,
more than 95% validated vectors in the high-speed regions and the obtained dynamic spatial range is 80, corresponding to a spa-
almost 100% validation in other regions (except in certain areas tial resolution of 0.18 mm in FOV1 and 0.55 mm in FOV2 (the data
influenced by shadows or reflections; see Results and Discussion resolution is actually twice as high due to the use of 50% overlap
section). Further increase in the laser pulse separation time caused between interrogation areas). In contrast, the Kolmogorov length
rapid escalation in the number of outliers. Deviation from the scale is estimated to range between 5 and 15 ␮m assuming that
local mean was used as the validation criterion and invalid vec- all power is dissipated within twice the rotor swept volume, an
tors were replaced by linear interpolation. All cross-correlation assumption consistent with the findings of Utomo et al. (2008).
and validation were carried out using Dantec FlowManager soft-
ware, while further processing was done in MATLAB® . Note that Power Draw
no data filtering has been employed. Calculation of power draw is facilitated by simultaneous measure-
ments of speed and torque using the sensors described in Mixer
Measurement Matrix System section. Data acquisition is controlled via LabViewTM 8.5
For each field of view, measurements were obtained at 12 different software in conjunction with a NI USB-6215 data acquisition
rotor speed settings ranging from 333 to 1433 rpm (corresponding board. The sampling rate for each signal is 50 kHz and power draw
to Reynolds numbers, NR from 2.0 × 105 and 8.5 × 105 ), separated
by 100 rpm increments. The upper limit was enforced by incip-
ient vibrations of the Plexiglas stator while the lower limit was
assumed to border the region of practical interest. For each speed
setting, angularly resolved measurements were acquired at 12 dif-
ferent rotor positions with 5◦ increments. Given the 60◦ symmetry
of the rotor, this maps a full rotor revolution. For the small field
of view, FOV1, the measurements were additionally resolved in 1◦
increments within the 10◦ region surrounding the viewed stator
slot. It should be noted that the recordings showed some variance
in the rotor positions due to inherent inaccuracy within the syn-
chronisation system. The accuracy of synchronisation, in terms
of the recorded angular rotor position, is estimated to be ±0.3◦ .
For each specific rotor position, speed setting and field of view,
500 image pairs were acquired. During most measurements the
temperature was controlled to 21.0 ± 0.5◦ C and the vessel pressure
was maintained at 1.00 ± 0.05 bar. However, for speed settings
ranging from 1033 to 1433 rpm in FOV1, a low-pressure vortex
core caused severe reflections from cavitation bubbles. To cir-
cumvent this, the temperature was reduced to 15 ± 0.5◦ C and the Figure 3. Example of a raw PIV image obtained in FOV2. Notice the
pressure was increased to 1.45 ± 0.05 bar, sufficient to suppress shadow stripes inside the stator slot and the reflection zones near solid
the aforementioned phenomenon. surface contours.

| 1080 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 89, OCTOBER 2011 |
is calculated from the average value over a 2-s period. About 200 stator material and surrounding fluid. The reduced illumination
periods are used for estimation of power draw statistics. To com- directly influences the noise in these stripes, something that is
pensate for losses in couplings, bearings and seals, measurements clearly seen in the data. Reflections from rotor and stator surfaces
were also obtained using an empty but pressurised vessel. are another disrupting factor, and the resulting pixel saturation
made it necessary to discard (zero-out) data in these areas. Dis-
carded data are masked with black colour in all subsequent vector
RESULTS AND DISCUSSION and scalar plots. In these plots, the rotor blade is dark grey and
Before presenting and interpreting results it is necessary to high- the solid portion of the stator is light grey. The direction of blade
light some characteristic features of the raw image data. For that rotation is clockwise. When referring to the left hand wetted
purpose a typical image, obtained in FOV2, is shown in Figure 3. stator slot surface, we will use the terms upstream surface and
The shadow stripes inside the stator are caused by the inclined upstream edge(s), to conform with the direction of blade rota-
stator slot edges and the difference in refractive index between tion. When referring to the right-hand-wetted stator slot surface,

Figure 4. Mean velocity magnitude normalised by rotor tip speed for rotor positions  = 25◦ to  = 30◦ in FOV1. Measurements obtained at 1033 rpm.

| VOLUME 89, OCTOBER 2011 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 1081 |
we will use the terms downstream surface and downstream  = 25◦ to  = 40◦ . The measurements were obtained at 1033 rpm
edge(s). and the rotor tip speed is used to normalise the data. In addition,
selected streamlines are included to enhance visualisation of the
various flow structures. Note that in certain circulation zones,
Flow Topology some streamlines are so close together that they appear as a single
The measured flow fields are complicated and inherently unsteady thick black line rather than as thinner separate lines.
due to the transient passage of rotor blades. Several characteristic As the rotor blade approaches the stator slot ( = 25◦ ), a stagna-
flow structures are revealed by the angularly resolved velocity tion zone is formed in front of the inner edge of the downstream
measurements. Most of these structures appear at all rotor speed stator slot surface where the fluid entering the slot is separated
settings in both the instantaneous and mean fields. Figures 4 and from the fluid remaining inside the stator. For the part impinging
5 show the spatial variation of the measured mean 2D velocity on the downstream stator slot surface practically all the tangential
magnitude in FOV1 at 12 different rotor positions ranging from momentum is converted into radial momentum. The fluid then

Figure 5. Mean velocity magnitude normalised by rotor tip speed for rotor positions  = 31◦ to  = 40◦ in FOV1. Measurements obtained at 1033 rpm.

| 1082 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 89, OCTOBER 2011 |
accelerates and emanates from the stator opening entraining the wards. The path and strength of the vortex is illustrated in Figure 7
surrounding slow moving bulk fluid. Only part of the slot near the where the Q-criterion, originally proposed by Hunt et al. (1988), is
downstream surface is occupied by this jet stream. Most of the slot used to identify the vortex core and to quantify the vortex strength
is occupied by a circulation loop where bulk fluid is re-entrained at various rotor positions. An expression for the two-dimensional
back into the stator slot due to the low pressure caused by the Q criterion, Qxy , is given in Equation (1). Clearly, Qxy is increas-
accelerating jet, a phenomenon also reported by Calabrese et al. ingly positive in areas where the vorticity, ωz , dominates the rate
(2002); and Utomo et al. (2008). The re-entrainment flow clearly of strain, Sxy .
separates at the outer edge of the upstream stator slot surface gen-
erating one or more slowly moving circulation zones. Notice that 1 2
Qxy = (ω −2Sxy ) (1)
the velocity of the emanating jet is about 1.3 times the rotor tip 4 z
speed while the velocity of the fluid downstream the stator slot is
<0.8 times the rotor tip speed. As the vortex strength reduces the core pressure increases and
As the rotor blade begins to block the stator slot (from  = 26◦ ) the cavitation bubbles implode. The resulting pressure waves and
the speed of the emanating jet reduces. The flow enters the slot velocity gradients would most likely influence the emulsifica-
in two separate streams. Fluid behind the rotor blade enters tan- tion and dispersion performance of the mixer. In this study, the
gentially via the rotor–stator clearance gap and fluid downstream cavitation phenomenon begins at approximately 1000 rpm at a
the rotor blade enters radially via the unblocked portion of the fluid temperature of 21◦ C and a vessel pressure 1 bar. However,
stator slot opening. The two streams merge near the leading edge the onset is likely influenced by several other parameters such
of the rotor tip causing a sudden radial diversion of the tangential as the geometry and size of stator openings, stator thickness,
stream. A strong vortex is generated in the corner of the resulting rotor–stator clearance gap and rotor blade width.
bi-directional shear layer ( = 29◦ ). The strength of this vortex is The magnitude of the 2D mean velocity field in FOV2 is shown
illustrated by the fact that expanding cavitation bubbles are pro- on Figures 8 and 9 for 12 different rotor positions separated by
duced in the low-pressure core. As mentioned in Measurement 5◦ increments. Given the 60◦ symmetry of the rotor this ade-
Martrix, these bubbles caused severe reflections in the PIV record- quately maps a full rotor revolution. Again, the rotor speed is
ings. Examples of this are given in Figure 6. The left image shows 1033 rpm and the rotor tip speed is used to normalise the data
a cavitation bubble located outside the light sheet, while the and selected streamlines are included to enhance visualisation of
right image shows the reflections from a cavitation bubble located the various flow structures. Note that in certain circulation zones,
within the light sheet. Both images were obtained at 1033 rpm some streamlines are so close together that they appear as a single
without the use of seeding. Note that the presence of seeding thick black line rather than as thinner separate lines.
amplifies the reflections and strongly increases the number of The flow in the rotor region is predominantly tangential except
saturated pixels. near stator slots, where the fluid starts to move in the radial
The most severe reflections were observed in the situation direction driven by centrifugal forces and pressure differences
where the rotor blade completely blocks the stator opening between the interior and exterior stator volume. Note the differ-
( = 30◦ ). In this case, a narrow high-speed jet enters the slot ence between the high velocity field trailing and the low velocity
tangentially via the rotor–stator clearance gap. The jet abruptly field in front of the rotor blade. The velocity of the former is about
re-directs in the corner between the rotor tip and the downstream 1.3 times the rotor tip speed while the velocity of the latter is
stator slot surface, resulting in very high velocities and steep veloc- about 0.7 times the rotor tip speed. The variation in velocity can be
ity gradients around the aforementioned vortex core. A complex explained by spatial differences in pressure caused by the impeller
flow pattern is also seen in the clearance gap near the leading rotation and varying displacement through stator openings.
edge of the rotor. Low-speed fluid from this gap moves counter The plots also reveal extremely complex flow patterns in the
clockwise into the stator slot, despite the clockwise moving rotor bulk flow outside the stator. The emanating jets intermingle and
blade, generating one or more eddies in the high shear zone near create various circulation loops depending on the rotor position.
the blade tip (not shown). As the rotor blade progresses (from The interaction between two adjacent jets also differs, likely influ-
 = 31◦ ) the strong vortex moves outwards within the strong radi- enced by their position relative to the surrounding baffle plates. At
ally directed velocity field. Note that the velocity here is up to 1.9 certain rotor positions the re-entrainment of bulk fluid back into
times the rotor tip speed but reduces as the vortex migrates out- the stator slot emanates from the slot’s own jet (e.g. at  = 10◦ ),

Figure 6. Left image shows a cavitation bubble located outside the light sheet. Right image shows reflections from a cavitation bubble located in the
light sheet. Both images are obtained at 1033 rpm and without the presence of seeding.

| VOLUME 89, OCTOBER 2011 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 1083 |
Figure 7. The two dimensional Q criterion, Qxy , calculated from the mean velocity field, plotted for selected rotor positions in FOV1. Measurements
obtained at 1033 rpm.

while in other positions (e.g. at  = 35◦ ) it emanates from the adja- located outside the stator. More information about the out-of-plane
cent slot’s jet. In all rotor positions, the jet is directed somewhat flow (unmeasured velocity component) can be extracted from the
counter to the direction of impeller rotation primarily, induced by divergence of the mean velocity field, based on the two measured
aforementioned re-entrainment flow. Utomo et al. (2009) reported components. For an incompressible fluid, ∇U = 0 so calculated
similar results for narrow stator openings in a non-baffled mix- non-zero divergence values indicate the presence of out-of-plane
ing vessel. In fact, the bulk flow was found to rotate opposite the extensional strain, and therefore out-of plane flow. Figure 10
impeller direction. shows the divergence of the mean velocity field for rotor posi-
While the flow interior to and inside the stator slots is strongly tion  = 30◦ in FOV2. Apparently, the flow is three-dimensional
confined by solid surfaces, the opposite is the case outside, in the in the emanating jets but even more so in the rotor–stator clear-
remainder of the tank. The bulk flow is more three-dimensional, ance gap. The former is likely due to jet entrainment while the
enhanced by the presence of baffles. In addition, out-of-plane latter could be due to the vertical leakage flow between the hor-
motion is indicated by the various source and sink structures izontal rotor plate and the inside stator surface. Note that the

| 1084 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 89, OCTOBER 2011 |
Figure 8. Mean velocity magnitude normalised by rotor tip speed for rotor positions  = 0◦ to  = 25◦ in FOV2. Measurements obtained at 1033 rpm.

horizontal clearance gap area between the rotor and stator cor- least for applications where constant slot and jet mean flow char-
responds to the area of almost two stator slots (out of 26 slots in acteristics are preferred. Of course, geometric similarity is also
total). A smaller rotor–stator clearance gap, a labyrinth seal or required to maintain stator slot and jet flow characteristics.
similar could reduce this leakage flow and thereby increase the
efficiency of the mixer. Flow Rate
Normalisation of the absolute velocity by the rotor tip speed The volume flow rate per unit depth, V̇xy , across a given contour,
has been used throughout this section. This is not without merit. C, can be expressed as the line integral along said contour of the
Figure 11 shows the normalised mean velocity field at two dif- dot product between the local mean velocity vector, U and the path
ferent rotor speeds for both fields of view. In all cases, the rotor normal vector, n. The term dS denote the length of an infinitesimal
position  = 30◦ . The almost identical plots reveal that velocity segment of the contour, C.
magnitude scales with the rotor tip speed while the flow topology 
remains self similar. This indicates that scale-up of rotor–stator
mixer applications should be based on constant rotor tip speed, at V̇xy = (Un) dS (2)
C

| VOLUME 89, OCTOBER 2011 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 1085 |
Figure 9. Mean velocity magnitude normalised by rotor tip speed for rotor positions  = 30◦ to  = 55◦ in FOV2. Measurements obtained at 1033 rpm.

Figure 12 shows the path of integration (red contour) used in stator slot. The incipient decrease in outflow is associated with
each field of view to estimate the volume outflow per unit depth the directional change of the velocity field leading the rotor blade.
of the centered stator slot, V̇xy, slot . Simpsons rule is applied to As velocities become more tangential to the stator slot entrance,
evaluate the integral from the discretised data. The values on C are an increasing amount of fluid is able to by-pass the slot opening
determined by linear interpolation from surrounding grid values. and remain in the rotor region. When the rotor blade begins to
Obviously errors are introduced due to the masked out regions block the stator slot, a rapid decrease in outflow occurs. This sce-
(light reflections), especially in FOV2. nario continues as the rotor blade passes the stator slot but with
The estimated slot outflow, for each rotor position, is given in a slightly lower rate as the stator slot begins to open ( ≥ 30◦ ).
Figure 13. Data for both fields of view are included. The shown The outflow starts to increase again between  = 30◦ and  = 35◦ .
cycle repeats due to the 60◦ symmetry of the rotor. The outflow The flow rate from  = 35◦ to  = 55◦ is somewhat lower than that
increases linearly as the rotor blade approaches the stator slot from  = 0◦ to  = 20◦ , due to the difference in pressure in the fluid
( = 0◦ –20◦ ). The maximum value is reached between  = 20◦ and leading and trailing the rotor blade. There is a small discrepancy
 = 25◦ ; that is slightly before the rotor blade begins to block the between FOV1 and FOV2 data. The former is likely more accurate

| 1086 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 89, OCTOBER 2011 |
V̇stator can be obtained by multiplying the stator outflow per unit
depth by the height of the stator slots. The flow number, NF given
by Equation (3) is then estimated to be 0.11 using the data from
FOV1.

V̇stator
NF = (3)
N · D3

Strain Rate
The nominal shear rate (Utip /ı) in the rotor–stator clearance gap is
often used by vendors to characterise the emulsification and dis-
persion ability of a given device. The value is easy calculated and
does provide a simple and intuitive explanation of drop or parti-
cle breakup. However, results by Calabrese et al. (2002) strongly
indicate that breakup is controlled by forces induced in the stator
Figure 10. The divergence field for rotor position  = 30◦ in FOV2. openings or in the emanating jets rather than by forces induced in
Measurements obtained at 1033 rpm. the rotor–stator clearance gap. Our data allows direct estimation
of the four in-plane strain rate components present in these areas.
due to the higher spatial resolution and the relatively limited From these, the 2D mean strain rate, Sxy , commonly defined as
size/influence of the masked out region. the magnitude of the 2D strain rate tensor eigenvalues, can be
The total stator outflow per unit depth V̇xy, stator can be estimated calculated. Expressed in terms of derivatives of mean velocity, Sxy
as the stator slot outflow per unit depth averaged over all rotor is given by:
positions and multiplied by the number of stator slots. Figure 14
shows V̇xy, stator as a function of the rotor speed, N. Clearly, V̇xy, stator  2  2
depends linearly of the rotor speed as also described by Utomo et dux duy ((dux /dy) + (duy /dx))2
Sxy = + + (4)
al. (2008). A rough estimate of the total flow rate out of the stator dx dy 2

Figure 11. Mean velocity magnitude normalised by rotor tip speed for two different rotor speed settings in the two different fields of view.

| VOLUME 89, OCTOBER 2011 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 1087 |
Figure 12. Mean velocity vector plot with path of integration (red contour) in FOV1 and FOV2 used for calculation of V̇xy,slot .

where ux and uy are the measured mean velocities. Note that, in accelerates in this region. Similar mechanisms are responsible for
some literature,
√ the expression in Equation (4) is multiplied by a the latter area. Note that the mean strain rate is up to 2.3 times the
factor of 2 due to a different definition of the various strain rate nominal shear rate in the clearance gap. There also is a rhombus
components. pattern present in the stagnation zone adjacent to of the down-
Figure 15 shows the mean 2D strain rate field normalised by the stream stator slot surface (the rhombus pattern is encircled by a
nominal shear rate in the rotor–stator clearance gap at six selected dashed line). It is unclear whether this arises from noise or is in
rotor positions in FOV1. The data are obtained at 1033 rpm and the fact a real phenomenon. However, thorough examination of the
nominal shear rate is approximately 20 000 s−1 . Note the elevated raw PIV images excludes non-uniform illumination (reflection or
noise level in these plots caused by numerical differentiation of shadow stripes) as the cause. Interestingly, similar patterns are
the mean velocities to obtain the derivative terms entering the reported by Cimbala et al. (1988); and Scarano and Poelma (2009),
definition of Sxy . The elevated noise level is clearly visible in the both describing rhombus shaped vortex cells in cylinder wakes.
shadow stripes behind the stator slot edges. However, whether there is a connection to present phenomenon
As the rotor blade approach the stator slot ( = 25◦ ) two dis- is uncertain.
tinct areas of high strain rate are observed; one near the inner Another distinct area of high strain rate is seen near the leading
edge of the upstream stator slot surface and one near the outer edge of the rotor tip as the rotor blade begins to block the stator
edge of the downstream stator slot surface. The former is a result slot ( = 29◦ ). The zone is formed by two sources, namely the
of the shear strain generated between the high-speed fluid enter- shear strain associated with the abrupt re-direction of the incom-
ing the slot and the low-speed fluid re-entrained from the bulk ing tangential jet and the elongational strain associated with the
flow, combined with the elongational strain generated as the fluid acceleration as it turns (de-acceleration prior to and acceleration

Figure 13. Estimated slot outflow, V̇xy,slot , as a function of rotor position. Figure 14. Estimated stator outflow, V̇xy,stator , as a function of rotor
Measurements obtained at 1033 rpm. speed, N.

| 1088 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 89, OCTOBER 2011 |
Figure 15. 2D mean strain rate, Sxy , normalised by nominal shear rate in the rotor–stator clearance gap at six selected rotor positions in FOV1.
Measurements obtained at 1033 rpm.

after the turn). Note that the absolute strain rate is about 2.6 strain in the jet shear layer. It varies with rotor position but is
times the nominal gap shear rate. As the rotor blade completely significantly lower than the nominal shear rate.
blocks the stator slot ( = 30◦ ), the zone of high strain rate follows It is clear that a large portion of the flow by-passes the high
the high-speed flow adjacent to the strong vortex. This scenario strain rate zones. Although this ratio likely depends on vari-
continues as the rotor blade moves forward and the vortex struc- ous geometrical parameters, the results indicate that single row
ture moves outward. The maximum strain rate is obtained at rotor–stator mixers are unsuitable as high intensity reactors (e.g.
 = 30◦ , and is about 3.2 times the nominal shear rate. in applications involving competing reactions). It also explains
Figure 16 shows similar plots for two selected rotor positions in why single-stage rotor–stator mixers are seldom used in single
FOV2. Again the data are obtained at 1033 rpm and the nominal pass inline processes, but more commonly used in batch or semi-
shear rate used for normalisation is approximately 20.000 s−1 . The continuous processes using recirculation to insure multiple passes
absolute strain rate in the jets is primarily induced by the shear through the rotor–stator unit.

| VOLUME 89, OCTOBER 2011 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 1089 |
Figure 16. 2D mean strain rate, Sxy , normalised by nominal shear rate in the rotor–stator clearance gap at two selected rotor positions in FOV2.
Measurements obtained at 1033 rpm.

The strain rate plots presented above were normalised by the velocity instead of Utip . However, normalisation by the local mean
nominal shear rate in the rotor–stator clearance gap which for leads to excessively large values in low-speed regions. Hence, cau-
constant gap size is proportional to the rotor tip speed. As illus- tion should be taken when comparing present observations with
trated in Flow Typology section, the mean velocity scales with the results from the literature.
rotor tip speed while the flow topology remains constant, inferring Figures 18 and 19 shows the turbulent intensity, Tu, at selected
that spatial velocity derivatives and therefore the mean strain rate rotor positions in FOV1 and FOV2 for the data obtained at
also scales with the rotor tip speed. This is indeed what is seen in 1033 rpm. It is clear that regions of high turbulence coincide with
the present data. However, whether or not the strain rate scales the regions of high mean strain rate. This can be expected from the
with the nominal shear rate remains unanswered, since the gap transport equations for the turbulent kinetic energy. Using Ein-
width was not varied. It is likely though, that geometric param- stein notation, the term responsible for production of turbulent
eters do influence the strain rate, for example the geometry and kinetic energy is given by:
size of stator openings and the width of the rotor–stator clearance
gap.     ∂ui
− ui uj (7)
∂xj
Turbulent Fluctuations
Turbulence is central to mixing and the random velocity fluctua- Thus, regions of high turbulence are expected where the mean
 
tions are responsible for break-up mechanisms in emulsification velocity gradients ∂ui /∂xj and/or the Reynolds stresses −ui uj 
processes. The turbulent velocity fluctuations are illustrated in are high and that is indeed what is seen. The highest values of
Figure 17 which shows the deviation between a single instanta-
neous velocity field sample and the ensemble mean.
It should be mentioned that the mean field structures, described
in Flow Typology section also appear in most of the instantaneous
fields. However, their spatial position and shape varies signifi-
cantly from sample to sample. In addition, various small eddies
are always present, mainly in shear layers but frequently in other
regions. The varying structure positions are inherently due to
the stochastic nature of the velocity fluctuations. A quantitative
measurement of these random unsteady velocity fluctuations is
given by the measured portion kxy of the turbulent kinetic energy,
defined as the mean deviation of instantaneous data from the
ensemble mean results:

1  2   2 
kxy = ux + uy (5)
2

where   denotes a mean quantity. The turbulence intensity Tu


may be calculated from kxy using the rotor tip speed as reference:

kxy
Tu = (6)
Utip
Figure 17. Deviation between a single instantaneous velocity field
An alternative definition of the turbulence intensity, frequently sample and the ensemble mean for  = 25◦ in FOV2. Measurements
appearing in the literature, uses the magnitude of the local mean obtained at 1033 rpm.

| 1090 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 89, OCTOBER 2011 |
Figure 18. Turbulence intensity, Tu, at six selected rotor positions in FOV1 for data obtained at 1033 rpm.

turbulence intensity (Tu about 70%) are observed in the stator macro-scale dependent rotor–stator mixing processes should be
slot near and inside the stator slot vortex. This is partially due based on constant rotor tip speed.
to the high strain rate field adjacent to the vortex core but also
to the Reynolds stresses caused by the expanding and imploding
cavitation bubbles. The large region of high turbulence intensity Power Draw
(Tu from 20% to 40%) occurring in the emanating jets indicates The power dissipated in the fluid P and the impeller power num-
that the majority of energy may be dissipated in that region. ber NP [calculated from Equation (8)] are plotted in Figure 21
Figure 20 shows the turbulent intensity fields at two different versus rotor speed, N.
rotor speeds in both fields of view. All measurements are obtained
at rotor position  = 30◦ . The employed definition of the turbu-
lence intensity, using the rotor tip speed as a reference, seems to P
NP = (8)
result in self similar Tu fields. Again this implies that scale-up of N 3 D5

| VOLUME 89, OCTOBER 2011 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 1091 |
Figure 19. Turbulence intensity, Tu, at six selected rotor positions in FOV2 for data obtained at 1033 rpm.

The rotor speed range corresponds to impeller Reynolds num- examining comparable rotor–stator mixers. Padron (2001); and
bers NR , [calculated from Equation (9)], between 2.0 × 105 and Doucet et al. (2005) report power number values between 1.5
8.5 × 105 , implying the flow is within the turbulent regime. and 3, suggesting emulsification performance superior to present
mixer configuration. A more detailed assessment of emulsifica-
ND2 tion performance requires comparison of the local maximum in
NR = (9)
␮ the spatially dependent energy dissipation rate.
The power dissipated in the fluid scales with the rotor speed
The nearly constant power number curve is typical for both agi- with an exponent of 2.97; very close to the theoretical value of
tators and rotor–stator mixers operating in the turbulent regime 3 for conventional agitators operating in the turbulent regime.
(Atiemo-Obeng and Calabrese, 2004). However, the constant This indicates that power draw in is primarily controlled by the
value of about 1.2 is somewhat lower than seen in other studies turbulent fluctuations in the emanating jets.

| 1092 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 89, OCTOBER 2011 |
Figure 20. Turbulence intensity fields, Tu, at two different rotor speeds (333 and 1033 rpm). Data are shown for both fields of view and rotor position
 = 30◦ .

with water as working fluid. Measurements have been acquired


at various rotor speeds corresponding to conventional impeller
based Reynolds numbers between 2.0 × 105 and 8.5 × 105 . The
use of a transparent Plexiglas stator facilitated PIV measurements
inside and outside the stator as well as into the stator slots them-
selves, something that has not been reported before.
The measured flow fields are complex and inherently unsteady
due to the transient passage of rotor blades past stator slots. Sev-
eral characteristic structures are present in both the instantaneous
and mean fields. These include high speed jets emanating from
the stator slots, circulation loops that entrain bulk fluid back into
the stator slot and rotor–stator clearance gap, and strong stator
slot vortices caused by the passing rotor blades. The mean veloc-
ities show large spatial variations with the highest values present
in the emanating stator jets and adjacent to aforementioned slot
vortices. The mean velocities in the former and latter regions were
found to be about 1.3 and 1.9 times the rotor tip speed respectively.
The mean velocities were found to scale with the rotor tip speed,
while the flow topology remains self similar.
Figure 21. Power dissipated in the fluid, P, and power number, NP , Expanding cavitation bubbles were produced in the low-
versus rotor speed, N. pressure core of the strong stator slot vortices. As the vortex is
carried out of the stator slot by the radial jet, its strength reduces
and the core pressure increases causing the cavitation bubbles to
CONCLUSIONS implode. The resulting pressure waves and velocity gradients most
Characteristics of batch rotor–stator mixer performance are likely influence the emulsification and dispersion performance of
elucidated by shaft torque and angularly resolved 2D PIV mea- the mixer.
surements obtained in a full-scale, custom built, bottom-mounted, Based on divergence plots, the flow seems to be highly
TPS rotor–stator mixer unit operating in the turbulent regime, three-dimensional in the emanating jets and in the rotor–stator

| VOLUME 89, OCTOBER 2011 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 1093 |
clearance gap. The former is likely due to jet turbulence while WCCD width of the camera CCD chip (m)
the latter could be due to the vertical leakage flow between the WFOV width of a given field of view (m)
horizontal rotor plate and the inside stator surface. The flow rate
out of a specific stator slot shows a sinusoidal character with a
period corresponding to the time between the passage of two suc- Greek Symbols
ceeding rotor blades. The total outflow scales linearly with the ı width of rotor-stator clearance gap (m)
rotor speed, corresponding to a constant impeller flow number of  rotor position given in the global cylindrical coordinate sys-
0.11. tem (◦ )
The highest values of the mean strain rate fields are observed in ␮ dynamic viscosity (Pa s)
the shear layers of the stator slot jet and near the stator slot vor-  density (kg/m3 )
tex. The values are up to 3.2 times the nominal shear rate in the ωz vorticity component in z direction (s−1 )
rotor–stator clearance gap. Large portions of the flow by-passes
the high strain rate zones. The mean strain rate field is found to
scale with rotor tip speed. As expected, regions of high turbulence
ACKNOWLEDGEMENTS
are observed where the mean velocity gradients are the largest. The financial support from Tetra Pak Scanima (DK), Department
The highest values of turbulence intensity (Tu about 70%) are of Energy Technology at Aalborg University (DK) and the Dan-
found near and inside the stator slot vortex. This is partially due ish Department of Research and Innovation (DK) are gratefully
the high strain rate field adjacent to the vortex core but also due to acknowledged.
expanding and imploding cavitation bubbles. The large region of
high turbulence (Tu between 20% and 40%) occurring in the ema-
nating jets indicate that the majority of energy is dissipated in this REFERENCES
region. Defining the turbulence intensity by normalisation with Adrian, A. J., “Particle Imaging Techniques for Experimental
the rotor tip speed results in self similar Tu fields, independent of Fluid Mechanics,” Annu. Rev. Fluid Mech. 23, 261–304
rotor speed. (1991).
The power draw is found to closely follow the theoretical Atiemo-Obeng, V. A. and R. V. Calabrese, “Rotor-Stator Mixing
curve for conventional stirrers operating in the turbulent regime, Devices,” in E. L. Paul, V. A. Atiemo-Obeng and S. M. Kresta,
enabling the use of conventional definitions of power and flow Eds., “Handbook of Industrial Mixing: Science and Practice,”
numbers. The constant turbulent power number is found to be John Wiley & Sons., Inc, Hoboken, NJ, USA (2004), pp.
1.2. 479–505.
The results indicate that scale-up of batch rotor–stator mixing Barailler, F., M. Heniche and P. A. Tanguy, “CFD Analysis of a
processes that depend on macro-scale quantities should be based Rotor-Stator Mixer With Viscous Fluids,” Chem. Eng. Sci. 61,
on constant rotor tip speed, since this leads to self similar flow 2888–2894 (2006).
topology, mean velocity, strain rate and turbulence intensity fields. Calabrese, R. V., M. K. Francis, K. R. Kevala, V. P. Mishra, G. A.
A prerequisite for this scale-up rule is that geometric similarity is Padron and S. Phongikaroon, “Fluid Dynamics and
maintained. Emulsification in High Shear Mixers,” in “Proc. 3rd World
Congress on Emulsions,” Lyon, France (2002).
Chen, H. T. and S. Middleman, “Drop Size Distribution in
NOMENCLATURE Agitated Liquid–Liquid Systems,” AIChe J. 13(5), 989–995
D rotor diameter (m) (1967).
Ds inside stator diameter (m) Cimbala, J. M., H. M. Nagib and A. Roshko, “Large Structure in
kxy two-dimensional turbulent kinetic energy (m/s) the Far Wakes of Two-Dimensional Bluff Bodies,” J. Fluid
M magnification factor Mech. 190, 265–298 (1988).
n normal vector Doucet, L., G. Ascanio and P. A. Tanguy, “Hydrodynamics
N rotor speed (rps) or (rpm) Characterisation of Rotor-Stator Mixer With Viscous Fluids,”
NF impeller flow number Chem. Eng. Res. Des. 83(A10), 1186–1195 (2005).
NP impeller power number Francis, M. K., “The Development of a Novel Probe for the In
NR impeller Reynolds number Situ Measurement of Particle Size Distributions and
P power dissipated in the fluid (W) or (kW) Application to the Measurement of Drop Size in Rotor-Stator
Qxy two-dimensional Q-criterion (s−1 ) Mixers,” Ph.D. thesis, University of Maryland, College Park,
Sxy two-dimensional strain rate (s−1 ) MD, USA (1999).
Tu turbulence intensity Hunt, J. C. R., A. A. Wray and P. Moin, “Eddies, Stream and
U two-dimensional velocity vector Convergence Zones in Turbulent Flows. Report CTR-S88,”
Utip rotor tip speed (m/s) Center for Turbulence Research. NASA-Ames Research Center,
ux mean velocity component in x direction (m/s) Stanford University, California, USA (1988).
uy mean velocity component in y direction (m/s) Keane, R. D. and R. J. Adrian, “Optimisation of Particle Image

ux fluctuating velocity component in x direction (m/s) Velocimeters. Part 1. Double Pulsed Systems,” Meas. Sci.

uy fluctuating velocity component in y direction (m/s) Technol. 1, 1202–1215 (1990).
V̇stator volume flow rate through all stator slots (m3 /s) Keane, R. D. and R. J. Adrian, “Theory of Cross-Correlation
V̇xy volume flow rate per unit depth (m2 /s) Analysis of PIV Images,” Appl. Sci. Res. 49, 191–215 (1992).
V̇xy, slot volume flow rate per unit depth through a single stator Lund, E., H. Moeller and L. A. Jakobsen, “Shape Design
slot (m2 /s) Optimisation of Stationary Fluid-Structure Interaction
V̇xy, stator volume flow rate per unit depth through all stator slots Problems With Large Displacement and Turbulence,” Struct.
(m2 /s) Multidisc. Optim. 25, 383–392 (2003).

| 1094 | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | | VOLUME 89, OCTOBER 2011 |
Meyers, K. J., M. F. Reeder, D. Ryan and G. Daly, “Get a Fix on
High-Shear Mixing,” Chem. Eng. Prog. 95, 33–42 (1999).
Padron, G. A., “Measurement and Comparison of Power Draw in
Batch Rotor-Stator Mixers,” M.Sc. Thesis, University of
Maryland, College Park, MD, USA (2001).
Phongikaroon, S., “Effect of Dispersed Phase Viscosity and
Interfacial Tension on Drop Size Distribution in a Batch
Rotor-Stator Mixer,” Ph.D. thesis, University of Maryland,
College Park, MD, USA (2001).
Raffel, M., C. Willert, S. Wereley and J. Kompenhans, “Particle
Image Velocimetry—A Practical Guide,” 2nd ed., Springer,
Berlin, Heidelberg, New York (2007).
Scarano, F. and C. Poelma, “Three-Dimensional Vorticity
Patterns of Cylinder Wakes,” Exp. Fluids 47, 69–83 (2009).
Scarano, F. and M. L. Riethmuller, “Iterative Multigrid Approach
in PIV Image Processing With Discrete Window Offset,” Exp.
Fluids 26, 513–523 (1999).
Utomo, A. T., M. Baker and A. W. Pacek, “Flow Pattern,
Periodicity and Energy Dissipation in a Batch Rotor-Stator
Mixer,” Chem. Eng. Res. Des. 86, 1397–1409 (2008).
Utomo, A., M. Baker and A. W. Pacek, “The Effect of Stator
Geometry on the Flow Pattern and Energy Dissipation Rate in
a Rotor-Stator Mixer,” Chem. Eng. Res. Des. 87, 533–542
(2009).
Wang, C. Y. and R. V. Calabrese, “Drop Breakup in Turbulent
Stirred-Tank Contactors: II. Relative Influence of Viscosity
and Interfacial Tension,” AIChE J. 32(4), 667–676 (1986).
Westerweel, J., “Digital Particle Image Velocimetry—Theory and
Application,” PhD Dissertation, Delft University Press, Delft
(1993).
Westerweel, J., “Theoretical Analysis of the Measurement
Precision in Particle Image Velocimetry,” Exp. Fluids Suppl.
29, S3–S12 (2000).

Manuscript received September 2, 2010; revised manuscript


received March 26, 2011; accepted for publication March 28, 2011.

| VOLUME 89, OCTOBER 2011 | | THE CANADIAN JOURNAL OF CHEMICAL ENGINEERING | 1095 |

You might also like