Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Characterization of the

Performance of Bin Blenders


Part 2 of 3: Free-Flowing Mixtures
Albert Alexander, Osama Sudah, Paulo Arratia, Chris Goodridge, Laman Alani, and Fernando Muzzio*

P
In part two of a series of three articles, art one in this series of articles discussed the general ap-
mixing rates and mechanisms are examined proaches to determining the mixing performance of bin
using rectangular bin blenders and two free- blenders in terms of sampling methods and tools, vari-
able analysis, and expected effects of common experi-
flowing mixtures.
mental settings (fill percentage, rotation rate, loading method),
followed by an example of a case study and a brief discussion
about the effects of discharge (1). In the second part of this se-
ries, mixing rates and mechanisms are examined using rectan-
gular bin blenders and two free-flowing mixtures. Specifically,
300L the effects of loading conditions, fill level, baffle implementa-
56L tion, rotation rate, and scale-up are addressed for binary mix-
tures of colored sand and a mixture of granulated aceta-
minophen with common pharmaceutical excipients.

14L Blenders
All of the experiments in this article have been performed in
rectangular bin blenders of various sizes. The three blenders,
of 14-, 56-, and 300-L total capacity, are shown in Figure 1. The
two larger blenders are manufactured by GEA Gallay (Birm-
ingham, UK) and the smallest blender is a custom-made ves-
Figure 1: The three blenders discussed in this article include a sel that is run under stepper motor control. The geometry of
14-L transparent, a 56-L stainless steel, and a 300-L stainless these blenders is formed by joining together a pyramidal hop-
steel tote-blender (Gallay, Birmingham, UK). per section and a rectangular bin section. The blenders are sym-
metrical in design (in a front-to-back and left-to-right sense)
Albert Alexander and Chris
but the rotation axis is offset 30% to the long axis of the blender,
Goodridge are post-doctoral researchers, which produces asymmetric flow patterns in the blender (see
and Fernando Muzzio, PhD, is a Figure 2a). Each of the major dimensions for this blender geom-
professor, all in the Department of Chemical etry is identified in Figure 2b and the measured lengths and an-
and Biochemical Engineering at Rutgers gles are shown in Table I.
University, 98 Brett Road, Piscataway, NJ
08854, tel. 732.445.3357, fax 732.445.6758,
In the two larger blenders, a circular opening is cut into the
muzzio@sol.rutgers.edu. Paulo Arratia is a top of the blender to enable access to the material within it. In
post-doctoral researcher at Haverford the 56-L blender, the opening is 11.75 in. in diameter, which en-
College (Haverford, PA), Osama Sudah is ables direct access to 41% of the blender surface area, leaving 59%
an engineering associate at Merck Research of the mixture surface partially accessible (i.e., sampling must be
Laboratories (Rahway, NJ), and Laman
Alani is the executive director at Merck
done on an angle or near the surface). In the 300-L blender, the
Research Laboratories (West Point, PA). opening is 18 in. in diameter, which opens 35% of the surface
Fernando Muzzio, PhD, is also a member of area to direct sampling. In the smallest vessel, the entire top is re-
Pharmaceutical Technology’s editorial movable, enabling access to the entire mixture surface.
advisory board. A diamond-shaped baffle is located in the center of each
*To whom all correspondence should be addressed. blender perpendicular to the axis of rotation (identified in Fig-
ure 2). The purpose of the baffle is to aid axial mixing by in-
56 Pharmaceutical Technology JULY 2004 www.phar mtech.com
Sampling tools. For the sand experiments, core samplers were
used to gather samples from the mixture. Core samplers are hol-
low tubes with one end filed to create a sharp edge to minimize
bed disturbance when thrust into the mixture. Upon retrieval of
the sampler, the contents are extruded in a last-in–first-out man-
ner (1, 2). During initial experimentation, the sand flowed out
of the core samplers when they were removed from the granu-
lar bed. Coating the insides of the tubes with a concentrated soap
solution to increase friction between the material and the wall
effectively retained the material within the tube upon the removal
Figure 2: The rotational axis of Gallay tote-blenders are offset 30% to of the sampler from the mixture. Each core was extruded into
the long axis of the blender (a). The various dimensions for this blender 5–15 samples of 8 g.
geometry are sketched and labelled in section (b). The acetaminophen mixture was sampled using the side-
sampling groove thief because the material did not stay in the
Table I: Bin blender dimensions (in.) core sampler when the sampler was removed from the powder
H W D V  bed. Coating the samplers with soap solution was not workable
14-L 12 11 9 3.5 35 because the quantification method (near-infrared [NIR] spec-
56-L 22 18 15 6.0 35 troscopy) is sensitive to the addition of impurities in the sam-
300-L 40 30 24 8.0 35 ples, and adding fluid could dissolve some of the sample. The
groove sampler used in these experiments consisted of an outer
H denotes height, W denotes width, D denotes depth, V denotes hollow sleeve (1 in. in diameter) with an opening running the
valve opening,  denotes hopper angle.
length of the pipe (66 in.), surrounding a rotating inner pipe
(1, 3). The sampler was inserted into the powder bed with an
Table II: Baffle dimensions (in.) open cavity. The inner tube is rotated to trap a core of mater-
ial. The sample-bearing thief is placed on a frame and opened,
Relative
discharging the material into a series of adjacent trays. The sam-
Baffle Baffle Blender baffle
ple size can be roughly varied by using trays of different widths.
w. h. W. W.
In these experiments, the trays were 0.75 in. in width, which re-
14-L 1.25 2.00 11 11.4%
sulted in 4–8 samples of 1.6–2.1 g.
56-L 2.00 3.25 18 11.1%
300-L 6.00 10.50 30 20.0%
Analysis
Baffle w denotes baffle width, Baffle h denotes baffle height, The sand mixtures were characterized using a gray-scale image
Blender W denotes blender width, Relative baffle W denotes
analysis technique developed by Wightman et al. to quantify
relative baffle width.
the component concentrations in binary mixtures of contrast-
creasing axial flow as the mixture tumbles in the blender. The ing colors (4). For the mixture of black and white sand, each
baffle is removable in the two smaller blenders but is perma- collected sample was spread in a circular dish 2.25-in. in di-
nently attached in the largest blender. The relative size of the ameter. Under controlled lighting conditions, a picture of the
baffle, which can have a significant effect on mixing perfor- sample was taken using a digital camera. The resulting image
mance, is identified by comparing the width of the baffle to the was then processed using Scion Image Software (Scion Corpo-
width of the blender, as shown in Table II. ration, Frederick, MD, and the National Institutes of Health,
The relative size of the baffle is greater in the 300-L blender, Bethesda, MD) to give the mean gray-scale value of the sam-
than in the other two, and as a result the effect of the baffle on ple. Using a calibration curve of known component concen-
axial mixing rates may also be greater in the 300-L blender. Fur- tration standards, the mean gray-scale value of each sample was
thermore, the lack of geometric similarity in baffle size with converted to concentration data.
increasing blender size may affect mixing performance when Acetaminophen concentration in the powder experiments
scaling-up. was quantified using NIR spectroscopy. Two primary advan-
Materials. The following two free-flowing mixtures will be tages of NIR spectroscopy are its nondestructive, nonintrusive
discussed: nature, which enables the acquisition of a spectra of solid sam-
● Sand mixture: 50/50 w/w % mixture of black and white sand ples with minimal or no pretreatment, and its relatively fast rate
particles of a similar size (400 m) of analysis, scanning each sample in 30 s. Both calibration
● Acetaminophen mixture: 25% by weight granulated aceta- and experimental samples were placed in scintillation vials
minophen (100 m), blended into a matrix of microcrys- (Fisher brand, 20-mL in volume, 1.5-cm in diameter, and 6-cm
talline cellulose (Avicel PH102), mannitol, aspartame, and in height) that were suitable for NIR analysis. The NIR spec-
magnesium stearate (MgSt). trometer produces radiation that penetrates 4 mm into a pow-
The sand mixture is very free-flowing. The acetaminophen der sample before becoming too attenuated to produce a use-
mixture investigates mixing of a free-flowing active with a mod- ful signal. Consequently, the size of a sample was selected to
erately free-flowing excipient. produce a 4-mm height, resulting in a sample size of 2 g.
58 Pharmaceutical Technology JULY 2004 www.phar mtech.com
Figure 3: The two extreme initial condition possibilities are sketched.
Figure 3 (a) shows top-to-bottom and left-to-right loadings, which
emphasize radial and axial mixing, respectively (rotation is into the
page). The evolution of RSD is compared for these extreme loading
conditions in (b), showing that radial mixing is an order of magnitude
faster than axial mixing.

A rapid content analyzer manufactured by FOSS NIRSys-


tems, Inc. (Silver Spring, MD) collected the NIR data, and the
company’s software (VISION 2.10) was used to evaluate the
data using a variety of mathematical treatment options. De-
scriptions of selected sampling of the use of NIR to identify
pharmaceutical ingredients are available (5, 6). Typically, the
spectrum of an unknown sample is analyzed with a model equa-
tion derived from a set of calibration samples of known con-
centrations. In this study, both the calibration and experimen-
tal NIR spectra were conditioned using the second derivative
and partial least square treatments over the entire spectrum
(1140–1830 nm). Good agreement was found between actual and
predicted values (3% error) and NIR was deemed suitable for
quantifying acetaminophen concentrations for this system.

Results
Loading conditions. Probably the most important factor affect-
ing mixing rates in tumbling blenders is the method by which
materials are loaded into the blender. Without a preblending
step, loading the blender results in an initial unmixed condi-
tion of the various components. For practical purposes, there
are only two extreme outcomes from blender loading: top-to-
bottom or left-to-right initial distributions. The two extreme
examples of completely segregated initial configurations within
a bin blender are sketched in Figure 3a. The difference in mix-
ing rates that arose from these two initial conditions was quan-
tified by comparing the evolution of mixture relative standard
60 Pharmaceutical Technology JULY 2004 www.phar mtech.com
than axial mixing. The mix- ure 4a. For top-to-bottom loading, there
ing constant, k, was defined was a notable difference in the early (20
in Part 1 of this series of ar- revolutions) variance decay but all exper-
ticles as the slope of the iments reached an asymptotic variance
lined formed by plotting ln value near 32 revolutions, beyond which
(RSD) versus revolutions there was little difference in mixture vari-
(1). For these experiments, ance for the three fill levels. The mixing
k was 0.253 for top-to-bot- rate was much slower at 80% fill but the
tom loading and 0.011 for difference in mixture quality at different
left-to-right loading, which fill levels was erased within 30 revolutions.
quantifies the radial mixing Furthermore, it is apparent that dead
in a 56-L vessel as more than zones did not form for the 80% case be-
20 times faster than axial cause RSD reached comparable levels at
mixing for this free-flowing 128 revolutions. Also, the slower mixing
mixture. Even though radial at 80% fill was still fast compared to typ-
mixing was very fast, there ical blender mixing times (order 100–200
will always be some degree revolutions) such that the difference in
of axial variance (especially mixing rates does not have severe conse-
because the initial condi- quences on blender performance.
tions are not perfectly radi- Similar results are noted for the sand
ally symmetrical) and axial mixture when mixed in the 56-L blender
mixing will always control at 20%, 40%, 60%, and 80% fill with the
asymptotic homogenization baffle (Figure 4b). For all fill levels below
(1). The order of magnitude 80%, RSD dropped to a value near 5% in
difference in the mixing rate 16 revolutions and did not vary significantly
between the above cases un- at 32 revolutions. In agreement with the
derscores the importance of data from the acetaminophen mixture, the
loading conditions. Acci- mixing rate was considerably slower at 80%
dentally loading the blender fill but it appears that a similar RSD would
with left-to-right asymme- be achieved near 50 revolutions. Hence,
try can have significant con- it appears that the change in mixing rates
sequences on mixture qual- at 80% fill will not be important when com-
ity if the mixing protocol pared to typical mixing times for either of
assumed top-to-bottom these free-flowing mixtures (nor do dead
Figure 4: The effects of changing the fill percent in the
loading had occurred. zones appear to form at 80% fill for either
blender are shown for a 56-L blender loaded from top-to-
Fill level. In previous stud- mixture). For fill levels below 80%, the RSD
bottom with the acetaminophen mixture (a) and the sand
ies, the fill level of the at 32 revolutions showed a gradation: the
mixture (b). Results from left-to-right loading with the sand
blender has been shown to 60% fill case is less homogeneous than at
mixture are shown in (c).
have a measurable effect on 40% fill which is less homogeneous than a
the mixing rates of free- system with 20% fill. These variations in
deviation, or RSD (RSD  /M, in which flowing materials (7, 8). Increasing the fill RSD likely arose because the two compo-
 is standard deviation and M is the av- level decreases the relative amounts of ma- nents in this sand mixture contained slightly
erage values of the samples used to gen- terial in the flowing region (the volume of different size distributions and, hence,
erate ). Changing the loading condition the flowing region does not significantly showed a slight tendency to segregate rather
has the effect of accentuating either radial change with variations in fill level). All than mix perfectly.
mixing (perpendicular to the axis of ro- mixing action must take place in the flow- The mixing rate was faster at 40% than
tation) for top-to-bottom loading, or axial ing region; therefore, decreasing the rela- at 60%, which is in agreement with the ac-
mixing (parallel to the axis of rotation) tive size of the flowing region decreases etaminophen data; however, the 20% re-
for left-to-right loading. mixing rates. Also, at higher fill levels, each sults were slower than the 40% results,
The results from blending the sand mix- particle spends less time in the cascading which is a curious result. In principle, the
ture for top-to-bottom and left-to-right region during each blender revolution and mixing rate at 20% should be faster than
loading conditions in a 56-L bin blender dead zones can form. For the 56-L bin at 40%, however, the sand mixture
at 60% fill without a baffle are shown in blender, with the baffle in place and using slumped (i.e., there was slip at the vessel–
Figure 3b. RSD decreases much faster in the free-flowing mixture interface) when loaded at 20%.
the top-to-bottom loaded experiment than acetaminophen mixture, the evolution of Slumping affects the amount of time it
in the left-to-right loaded experiment, in- top-to-bottom loaded experiments at takes for the mixture to “turn over” and
dicating that radial mixing is much faster 40%, 60%, and 80% fill is shown in Fig- interferes with the basic striation-forming
62 Pharmaceutical Technology JULY 2004 www.phar mtech.com
Figure 6: Sketch of the flow patterns
within the tote-blender. The mixture
travels down the cascade and then hits on
Figure 5: (a) Probability density function (pdf) of axial velocities from DEM simulations when the blender is in a slanted wall, which imparts axial flow to
the upright position; a vertical red line is shown at 0. The orientation of the blender is shown from the side in (b). the left in this configuration.

mechanism that determines radial mixing rates (9). Mixture time scale for radial mixing is mixture turnover, not blender
turnover is defined as the time it takes for an individual parti- revolution.
cle to return to near its original position. Changes in fill level For the left-to-right loaded case, there is no difference in the
have significant effects on turnover time; at low fill levels, the observed mixing rates at 40% and 60% fill for the sand mixture
mixture can undergo more turnovers/blender revolution than (see Figure 4c). This result is peculiar because if the major mech-
at higher fill levels. These differences in turnovers/revolution in- anism for axial mixing is dispersion, then the fill level should
dicate that variations in mixing performance with changes in have an effect on mixing rates. Higher fill levels cause the mix-
fill level do not necessarily reflect an increase in mixing efficiency ture to turn over fewer times than lower fill levels do for the
but may simply be an increase in mixing time. The true mixing same number of revolutions and obviously involve larger

Pharmaceutical Technology JULY 2004 63


amounts of material, requiring more particle movement to reach during a blender revolution. If dispersion is the dominant mode
an equivalently mixed state. Because 40% and 60% mix at nearly of axial displacement, then the average axial velocity should be
identical rates, the implication is that dispersion is not the only 0. A probability density function (pdf) of axial velocities when
mechanism at work. the blender is in the upright position (as in Figure 1) is shown
Previous studies have demonstrated that axial mixing rates in Figure 5a. At this orientation, the average axial velocity is
were affected by changes in fill percentage for experiments done 0.0083 cm/s in the positive direction (to the right), which is
in V-blenders and double cones (7, 8). Those blenders are sym- 2.7% of the average downstream velocity in the cascade (0.3
metrical with respect to the plane bisecting the vessel perpen- cm/s). This small but significant axial flow indicates that blender
dicular to the axis of rotation. This bin blender is rotated on an geometry is inducing axial flow in a specific direction. The av-
axis 30 offset to the long axis of the blender, which induces an erage axial velocity is shown in Table III for subsequent blender
asymmetrical flow topology during each complete blender rev- positions, using 0 as the upright state.
olution. The mixture within the blender flows into angled walls The velocities in Table III indicate that axial flow in the bin
(i.e., the walls are not perpendicular to the dominant flow di- blender changes direction from right-to-left and back as the
rection) that impart axial velocity gradients to the mixture. Fur- blender undergoes one complete revolution. Axial flows develop
thermore, the blender also is asymmetrical in a top-to-bottom because the mixture in the blender does not flow in a geomet-
fashion and the flow patterns that evolve as the mixture flows rically symmetrical environment. In the upright position, the
into and out of the hopper and the bin section are markedly mixture flows towards a wall that is sloped to the right, which
different, which has an effect on axial mixing. Axial mixing rates causes the entire mixture to flow to the right. Similarly, in the
are not likely to be affected by fill level in these bin blenders be- inverted position, flow is toward a leftward slanting wall, caus-
cause the presence of geometrically induced axial flow effectively ing a flow to the left (see Figure 6).
causes convective axial mixing that exceeds dispersive mixing (10). More-diverse flow patterns occur when the mixture is flow-
Computer simulations using the discrete elemental method ing into or out of the hopper section. As the mixture flows from
(DEM) allow for the study of axial flow gradients (1). The ef- the hopper section into the bin section, particles on either side
fects of blender geometry on particle velocities are investigated of the mixture diverge as the bin section is filled. Simulations
by extracting the average axial velocity (i.e., parallel to the axis capture this flow. At 1⁄4 turn, average axial velocities in the left
of rotation) of all free-flowing particles at specific instances half of the blender are 0.0028 cm/s while in the right half av-

64 Pharmaceutical Technology JULY 2004 www.phar mtech.com


decrease in RSD that results from top-to- the mixing rate resulting from the presence
Table III: Average axial velocity at bottom (Figure 7a) and left-to-right loaded of the baffle: the mixing constant increases
specific blender orientations. (Figure 7b) experiments with the sand mix- by a factor of 1.6 (0.0106 to 0.0165). The
Blender Average axial velocity ture when the 56-L blender is run with and modest increase in mixing rate with baffle
orientation (cm/s) without a baffle. For top-to-bottom load- addition is far less than the order of mag-
0 0.0083 ing, there is virtually no effect coming from nitude improvements seen in work using
1
⁄8 0.0144 the presence of the baffle because axial mix- 1-gal double-cone blenders and a different
1
⁄4 0.0051 ing is not a mixing-limiting step for that type of baffle (8). However, the effects of
3
⁄8 0.007 initial condition. However, when loaded baffle use must be examined by comparing
1
⁄2 0.006 left-to-right, there is a modest increase in baffle geometry, size, and location in the ex-
5
⁄8 0.011
3
⁄4 0.0055
7
⁄8 0.0085
1 0.0083

erage axial velocities are 0.019 cm/s. Par-


ticles in the left half are moving leftward
and in the right half are moving rightward,
and thus particles are diverging. The con-
verse situation arises when the mixture
flows from the bin section into the hop-
per section. At 13⁄16 turn, average axial flow
in left half is 0.0053 cm/s and in the right
half is 0.014 cm/s indicating that parti-
cles are converging. Similar flow patterns
have been shown to lead to the formation
of segregation patterns in double-cone
blenders (11) and similar segregation pat-
terns form in bin blenders (1). The ag-
gregate effect of these leftward, rightward,
converging, and diverging flows is to in-
duce axial flow in many directions as the
mixture tumbles in the blender. Because
the mixture does not turn over at the same
frequency as the blender rotates, different
portions of the mixture come into con-
tact with all of these diverse flows. Thus,
mixing efficiency is maximized because
different portions of the mixture come
into contact with these different velocity
fields.

Baffle use
In many tumbling blenders, a baffle can
be installed in the vessel as a means of im-
proving blending performance. The effi-
cacy of these inserts has not been quanti-
fied but it is commonly believed that any
disruption of normal, regular flow pat-
terns should have beneficial effects on
mixing rates. In this rectangular bin
blender, the baffle is diamond-shaped and
located in the center of the vessel, perpen-
dicular to the axis of rotation (shown in Fig-
ure 2). The intended effect is to improve
axial mixing by increasing the amount of
axial displacement. Figure 7 compares the
Pharmaceutical Technology JULY 2004 65
angle to the axis of rotation does not affect mixing rates for top-to-
in the middle of the blender. bottom loading in double cones or rotat-
In these double-cone exper- ing cylinders (7, 8) and has been shown
iments (with free-flowing to hold true for bin blenders (13) because
glass beads), the axial mixing the mechanism for radial mixing is the
rate was improved by a fac- same (iterative striation formation, 9). For
tor of 20 when the baffle was left-to-right loading, there was also no
placed in the blender, com- RPM effect noted for double cones or
pared to the nonbaffle con- V-blenders (7, 8). However, because axial
dition. However, the blender mixing in bin blenders involves convec-
was only 24.5 cm in diame- tion (axial mixing in the other blenders is
ter, which means that the baf- driven by dispersion), the effects of RPM
fle was almost half the width on left-to-right loading for free-flowing
of the blender. Also, the baf- mixtures may not be similar to other tum-
fle was not attached to the bling blenders. Therefore, only the effects
blender shell and did not ro- of rotation rate on axial mixing rates are
tate, hence the baffle was reported here for the bin blender. The re-
nearly perpendicular to the sults of experiments run at 6, 10, and 14
flowing region at all times. rpm are shown for experiments using the
Clearly, this large nonrotat- sand mixture in the 56-L bin blender
ing insert in the middle of the loaded from left-to-right in Figure 8. Once
blender will have an enor- again, there is no difference in the mixing
mous impact on flow pat- rates. This result indicates that particle ve-
Figure 7: The effects of a diamond-shaped baffle located in terns and the resultant mix- locities do not affect axial mixing rates,
the center of the blender are shown for top-to-bottom (a) and ing rates. In comparison, the which signifies that convective and dis-
left-to-right (b) loadings in the 56-L blender. baffles in the bin blenders persive flow is not affected in a mixing
were much narrower relative sense for free-flowing mixtures by rotation
to blender width (11–20%, rates or particle velocities.
see Table II) and would be
expected to have a lesser im- Scale-up
pact on flow patterns and The design of new batch manufacturing
mixing rates. processes often involves increasing the size
of the process equipment on progression
Rotation rate from lab- to pilot- to full-scale produc-
Another important factor tion equipment. The scale-up of mixing
regarding the performance rates in tumbling blenders remains de-
of tumbling blenders is the pendent on heuristics and experience
effect of rotation rate on rather than solid fundamental reasoning
mixing rates. Clearly, very backed by mechanistic explanations and
slow (0.1 rpm) or very fast experimental testing. The results from ex-
Figure 8: The rotation rate of the blender has no effect on the (
75 rpm) rotation rates periments in 14-, 56-, and 300-L bin
evolution of RSD for left-to-right loaded experiments at 60% have large effects on mixing blenders using the sand mixture loaded
fill in the 56-L blender. because the mode of granu- from both top-to-bottom and left-to-right
lar flow changes from being at 60% fill are shown in Figure 9. For top-
periments because the effects of baffle use characterized by discrete avalanches to to-bottom loading, there is virtually no
are closely tied to these variables. Simply cataracting flow (i.e., as portions of the difference in the mixing rate among the
adding a “baffle” of unspecified design and mixture become airborne) (12). However, three blenders (covering a
20-fold in-
location is no guarantee of improved the operating range of commercial crease in blender capacity). These results
blender performance. blenders lies in an interval far from both indicate that blender performance for free-
Comparing baffle effects in this bin of these two extreme flow conditions and flowing mixtures and top-to-bottom load-
blender to those observed in a 1-gal capac- within a range of rotation rates for which ing conditions are nearly impervious to
ity double-cone blender illustrates how re- flow characteristics are qualitatively very changes in blender size, making the radial
sults from baffle use must be viewed in the similar (i.e., the rolling regime). This raises mixing process scale-independent. While
context of baffle geometry and operating the question of whether mixing is affected the rate of mixing is nearly the same, the
principle. The baffle in the double-cone by changes in rotation rate in the rolling asymptotic RSD increases with increasing
blender was a rectangular plexiglass sheet, regime. blender size (similar trends were seen with
12.5 cm 10 cm, that was placed at a 45 Previous work has indicated that RPM increasing fill %, although not as stark).
66 Pharmaceutical Technology JULY 2004 www.phar mtech.com
spectively. Although a direct istics. The fill level had no effect on axial
relationship between blender mixing rates (up through 60% fill) in the
size and axial mixing rate asymmetrical bin blender because axial
does not exist, there is prob- mixing rates are controlled by convective
ably a strong, although cur- flow, not dispersive flow. The axial veloc-
rently unknown, connection ities imposed by rotating the blender on
between the two. Another a skewed axis increased the amount of
factor may be the relative axial mixing by imposing axial flow gra-
volume of the flowing mass dients on the mixture. This result is note-
to the total volume of the worthy because fill level does affect axial
mixture. In this system the mixing rates in symmetrical blenders that
surface area of the mixture rely on dispersion for axial mixing.
when the blender is upright The addition of relatively small baffle
is 99 in.2 and 270 in.2 re- (10–20% of the total blender width) had
spectively for the 14-L and no effect on radial mixing and a marginal
56-L blenders. If we assume effect on axial mixing, with mixing rates
that the flowing layer is con- increased by a factor of 1.6. We remind
stant in depth for both sys- the reader that this baffle is not optimally
tems (which may or may not designed or located to aid in increasing
be reasonable), then the ratio axial mixing rates. Similar to results seen
of the flowing mass to the in previous work on other tumbling
total volume is 0.116 and blenders, the rotation rate had no effect
0.079 for the smaller and on mixing efficiency in this study. When
Figure 9: The effect of changing the size of the blender is
larger blenders, respectively. free-flowing materials are mixed in
shown for top-to-bottom (a) and left-to-right (b) loading
Thus, there may be relatively blenders of increasing size, the loading
conditions.
more mass in motion in the method plays a decisive role in determin-
smaller blender, which could ing the scale-up approach. For top-to-
This gradation in asymptotic RSD could also be a factor affecting the axial mixing bottom loading, there is little or no ap-
be caused by segregation of the mixture rate. These results indicate that scaling rules parent difference in the mixing rates, but
(the particle size range of the two batches are very dependent on the loading condi- for left-to-right loading, the width of the
of colored sand were slightly different) or tions in the blender and that ensuring con- blender plays an important role in deter-
by dispersive limitation. The differences sistency in the loading method and initial mining the overall mixing rate. Using typ-
in asymptotic RSD will subsist with fur- conditions may be much more important ical scaling criteria such as tangential ve-
ther revolutions if segregation is the cause to the overall scale-up of the process than locity or the Froude number is irrelevant
but will disappear if dispersive limitation changes in vessel particulars such as fill as the rotation rate has no effect on mix-
is the cause. level, blender size, or rotation rate. ing rate.
Although radial mixing rates are inde- These results are compared and con-
pendent of scale, the axial mixing rate, as Conclusion trasted with similar experiments using co-
seen by examining left-to-right loaded ex- Generally, free-flowing materials are well- hesive materials in part three of this series
periments, does depend on blender size behaved and mixing results can be read- of articles. The effects of various experi-
(Figure 9b). For a change from 14 L to 56 ily explained and quantified. Radial mix- mental parameters on agglomerate com-
L, there is a decrease in the axial mixing rate ing (which is stressed by top-to-bottom minution and lubrication processes are
by a factor of 2 (the mixing constant de- loading) is more than an order of magni- also discussed.
creased from 0.0203 to 0.0106). Although tude faster than axial mixing (emphasized
axial mixing is not completely dependent by left-to-right loading). Adding more ma- References
on dispersion, axial mixing rates are still terial into a blender (i.e., increasing the 1. A. Alexander et al., “Characterization of the
very slow in comparison to radial mixing fill level) decreased radial mixing rates, Performance of Bin Blenders, Part 1 of 3:
Methodology,” Pharm. Technol. 13 (5), 70–86
rates. Because a single particle does not de- but as long as dead zones did not form (2004).
viate axially by more than a few particle di- (which is typical for fill levels 60% and 2. F.J. Muzzio et. al.,“An Improved Powder Sam-
ameters during a single pass down the cas- apparently true up through 80% fill for pling Tool,” Pharm. Technol. 23 (4), 92–110
cade, the width of the blender plays a large the mixtures tested here), the change in (1999).
role in the axial mixing rate in the blender. mixing rate would not have catastrophic 3. F.J. Muzzio et. al., “Sampling and Character-
ization of Pharmaceutical Powder and Gran-
A direct relationship between blender width effects on mixture quality for typical mix- ular Blends,” Inter. J. Pharm. 250 (1), 51–64
and axial mixing rates does not exist. The ing times. The highest permissible fill level (2003).
blenders are 18 in. and 11 in. in width, re- that does not cause dead zones to form is
spectively, resulting in blender-width:mix- unclear and is likely dependent on both
ing-constant) ratios of 890 and 1040, re- blender geometry and mixture character-
Pharmaceutical Technology JULY 2004 67
4. C. Wightman, F.J. Muzzio, and J. Wilder, “A Quantitative Image Analy-
sis Method for Characterizing Mixtures of Granular Materials,” Pow-
der Technol. 89, 165–176 (1996).
5. M. Blanco et al., “Near-Infrared Spectroscopy in the Pharmaceutical
Industry,” Analyst 123 (8), 135R–150R (1998).
6. S.P. Jacobsson et. al, “Quantitative Determination of Sulfasalazine by
Near-Infrared Spectroscopy and Multivariate-Analysis in Reflectance
Mode with a Fiberoptic Probe,” J. Phar. Biomed. Anal. 13 (4–5), 415
(1995).
7. D. Brone, A. Alexander, and F.J. Muzzio, “Quantitative Characteriza-
tion of Mixing of Dry Powders in V-Blenders,” AIChE J. 44 (2), 271–278
(1998).
8. D. Brone and F. Muzzio, “Enhanced Mixing in Double-Cone Blenders,”
Powder Technol. 110 (3), 179–189 (2000).
9. T. Shinbrot, A. Alexander, and F. Muzzio, “Spontaneous Chaotic Gran-
ular Mixing,” Nature 397, 675–678 (1999).
10. O. Sudah et. al., “Simulation and Experiments of Mixing and Segre-
gation in a Tote-Blender,” AIChE J. (currently in press).
11. A. Alexander, T. Shinbrot, and F.J. Muzzio, “Granular Segregation in
the Double-Cone Blender: Transitions and Mechanisms,” Phys. Flu-
ids 13 (3), 578–587 (2001).
12. H. Henein, J.K. Brimacombe, and A.P. Watkinson, “Experimental Study
of Traverse Bed Motion in Rotary Kilns,” Metall. Trans. 14B, 191–205
(1983).
13. O. Sudah, D. Coffin-Beach, and F.J. Muzzio, “Effects of Blender Rota-
tional Speed and Discharge on the Homogeneity of Cohesive and Free-
Flowing Mixtures,” Inter. J. Pharm. 247 (1, 2), 57–68 (2002).PT

Please rate this article.


On the Reader Service Card, circle a number:
345 Very useful and informative
346 Somewhat useful and informative
347 Not useful or informative
Your feedback is important to us.

68 Pharmaceutical Technology JULY 2004 www.phar mtech.com

You might also like