SOLIDS HANDLING

BLENDING, SAMPLING AND

SEGREGATION
By Thomas G. Troxel, Jenike & Johanson, Inc. | October 1, 2012

Blending and segregation are two opposite and competing processes in

solids handling that ideally fit the expression “two sides of the same
coin.” But, if you think about it, coins have three sides and sampling
completes the idiomatic expression perfectly. Blending is a necessary and
essential process operation widely employed in many industries and
materials, from nano-scale powders to run-of-mine ores. Segregation is
almost always an unwanted consequence of handling operations that
reverses the blending process or creates a need for blending where it
might not otherwise be necessary if it were not for the effects of
segregation. Sampling is the vitally important tool for measuring and
quantifying a blend, understanding the sources and effects of segregation
and for troubleshooting problems.

BLENDING
Blending is the process of combining two or more materials to achieve a
combined product. The mixture may be a combination of dissimilar
materials such as cement, sand and aggregate to make concrete; or cereal
flakes, raisins, nuts and marshmallows to make breakfast cereal. A blend
may also be a combination of chemically similar particles blended to
create a uniform mixture of particle sizes, or another property such as
color, flavor, melt index and so on. Blending bulk solids can be achieved
in many ways including batch and continuous methods and with a wide
variety of equipment, covering the range from extremely low intensity
blending with gravity recirculation or layering components onto a belt, to
high intensity blending with high-speed, high specific-horsepower
blenders.
All of the various equipment and processes for blending solids rely on
three principal mechanisms for achieving a blend: convection, diffusion
and shear.

Convection.Convection is the transfer of a collection of particles from

one location to another. This can occur as a result of material cascading in
a tumble blender, material moving against the blade of a ribbon or paddle
blender or as a result of gas pressure pulses in a pneumatic blender.

Diffusion.Diffusion is the random redistribution of particles that occurs

as a result of increased particle mobility. Increased mobility typically
occurs when the bulk density of the material is decreased sufficiently to
allow individual particles to move relative to one another. Fluidization in
fluidized bed reactors or granulators results in diffusion. Mechanical
blenders move collections of particles by convection, but when the speed
of the agitator is sufficient to locally fluidize material, diffusion occurs.
Tumble blenders, continuous screw blenders and agitated blenders also
provide opportunities for diffusion where material is cascading or in free-
fall.

Shear. The definition of shear as it applies to solids blending is

sometimes misunderstood. Shear occurs in a flowing granular solid as a
result of a velocity gradient, and can develop as either a discontinuous
shear (for instance, a shear or slip plane) or as a continuous gradient of
velocity. In either case there can be some overlap in what could be called
shear and what might be characterized as convection. The important
difference, at least as it applies to solids blending, is in the intensity.
Shear planes that develop in gravity blending or in a tumble blender or
screw blender will achieve blending predominantly by a convective
process. Shear in a high-speed mixer will have a different end result in
that it will be much more effective in breaking up agglomerates of fine
powders and distributing small-particle-size material with high surface
activity.

Figure 1 is a cross section view through a bin showing the displacement

of material as it flows. The pattern shows how the faster velocity in the
center displaces material relative to the slower material at the walls. This
displacement pattern can be described as shear deformation, but the
blending that occurs is caused by convection of collections of material
from the central part of the bin that reach the outlet sooner than material
near the wall. If material withdrawn from the bin is circulated back to the
top, the discreet fill pattern that was placed into the bin will be smeared
and homogenized, and some degree of blending will be achieved after
several recirculations [ 1].

Figure 1. This pattern shows how the faster velocity in the center
displaces material relative to the slower material at the walls

In a high-speed agitated blender, shear occurs with significantly higher

intensity near the agitator but affects a relatively small volume of
material. The high-intensity shear can break agglomerates, plate fine
particles onto the surface of larger particles and redistribute particles with
much higher intensity, causing better diffusion of small particles that may
adhere to themselves or other particles. Ideally, the convective
component of blending in this type of blender brings all of the material in
the blender into the high shear zones.

BLEND QUALITY
In casual conversation, terms such as “uniform blend”, “homogenous
mixture”, and “well blended” are often used without quantifying what
they mean. In the heavily scrutinized pharmaceutical industry, regulations
and guidelines dictate specific and often rigorous methodology for
quantifying the uniformity of a mixture. Food and other consumer
products must also meet trade regulations for delivering the stated
quantity of product and composition of ingredients. Quantifying the
structure of a blend requires defining not only the composition (for
instance, the proportion of each component and allowable variation) but
also the quantity of material that is to be measured. The compositional
requirement of a blend is usually well known because it is either the
entire objective of the process to produce a particular mixture, such as a
pharmaceutical tablet or box of cereal, or it is an intermediate mixture
that has well defined requirements. Determining the amount of material
to measure is sometimes less clear, and may be different depending on
who is answering the question. In pharmaceutical applications the answer
is clear in that it is almost always based on the unit dose of the blend that
makes up the final tablet.

In other applications the sample size subjected to scrutiny is not as clear

cut. Take, for example, a box of cereal. In one specific example a
manufacturer produced a cereal with a cereal particle and a marshmallow
candy particle. The packaging line filled boxes from a series of conveyors
that fed a mixture of the two components into a gravimetric packaging
machine that dispensed the required weight of the blend into each
package. The two components of the mixture were of similar size and
density and the only blending that occurred was in the handling system
along several conveyors and in a surge hopper. Each component was
metered onto the conveyor system at the correct ratio, but there was not a
discreet blending process. The manufacturing plants fine tuned their
systems to get an acceptable mixture into each box.

In this case, each box was always within weight tolerance for the stated
package quantity, but there was no direct control of the ratio of the two
components in any box. In an effort to improve quality, a new packaging
line was installed that handled each component individually up to the
packaging head where each component was weighed in a series of weigh
hoppers, and then each box was filled from several hoppers of each
component, selected by the packaging system controller, to control both
the proportion of each component and the total weight in a box. This
would seem an ideal solution since it would guarantee both accurate
package weight and composition. From the producers point of view, they
had improved quality by insuring that every package met their quality
standard.
However, to everyone’s surprise, consumer complaints from packages
produced at this plant went up. What the plant had gained in
compositional accuracy they had sacrificed in uniformity. Consumers
found that individual bowls of cereal poured from the new packages had
much wider variations of the two components than boxes filled from the
old filling line. The new filling line had given up control of blending,
even though the blending in the old system was very “low tech”. The new
system allowed only about one second from the time the ingredients were
separate until they were combined in the package. There was very little
opportunity to do any blending, and the result was that some boxes were
well blended while others were segregated. From the manufacturers point
of view, they had “improved quality” and were taking responsibility for
what they could control to the highest degree of precision, but they were
not controlling the sample size of interest to consumers.

Random blend. The simplified case of blending two different ingredients

that do not have significant bonding or other interactions is often referred
to as a random blend. In a random blend, all individual particles are free
to move relative to each other, and particle bonding is very weak or does
not occur. Particles that form random blends can be easy to blend,
because the particles move easily relative to one another, but that also
means that they can readily separate from each other and collect in zones
of similar particles when forces such as gravity, airflow, or vibration act
on the blend. This process is referred to as segregation. Random blends
are never perfect, and even in the best-blended state there will be random
variations in concentrations of components.

Ideal random blends are uncommon in most industrial applications,

although the cereal example described above comes close. In most dry
blending applications, particles have some tendency to interact with one
another by way of chemical, molecular, physical, or other means such
that individual particles can agglomerate, coat or bond to one another.
When particle interaction occurs, the blend is referred to as an ordered,
structured, or interactive blend. In an ideal world, this type of blending
process could create new particles, each consisting of the combined
individual ingredients. A perfect particle created in this blender would
consist of the desired ratio of each ingredient. This concept of combining
ingredients to create new, chemically homogeneous particles is one of the
primary objectives of wet and dry granulation processes. In most
industrial blending processes, the reality is somewhere between a random
blend and an ordered blend. Some particles of the blend may have very
little tendency to interact while other blend components may have
significant interaction.
 Figure 2. Sifting segregation occurs when a mixture
containing a range of particle sizes is allowed to form a pile.
Smaller particles sift between larger particles as each layer
of material slides along the pile surface

Figure 3. In the example described in the box on p. 44,

cylindrical pellets with a nominal size of about 3 mm were
being blended with nearly spherical beads with a size range
from about 1 mm down to 150 microns

Figure 4. Implementation of a more-rigorous sampling

method in the example on p. 44, revealed inadequate
blending. Area between dotted lines represents desired
blend ratio
Figure 5. Based on the improved sampling results, the
example blender was modified to improve the blending
performance and minimize segregation mechanisms

Figure 6. The sampling results for the new blender design

show the improved performance. Area between dotted lines
represents desired blend ratio

SEGREGATION
Segregation is the separation of particles that have identifiable
differences. By definition, identical particles cannot segregate.
Segregation can be driven by many forces including gravity, electrostatic
forces, fluid drag forces and elastic forces.

Sifting. The most prevalent type of segregation is sifting, which results in

separation by particle size. Sifting segregation occurs to some degree in
virtually every industry that handles bulk solids. Sifting will occur in
mixtures of different size particles when particles are sufficiently large so
that surface forces are weak relative to gravity (usually larger than 100
microns), when particles have mobility relative to one another (in other
words, they do not adhere to each other) and when there is some
mechanism to allow particles to move relative to one another. Collections
of different size particles will not spontaneously segregate when at rest,
but will often readily segregate when allowed to move in many common,
industrial handling processes.

Figure 2 shows an example of sifting segregation that occurs when a

mixture containing a range of particle sizes is allowed to form a pile.
Smaller particles sift between larger particles as each layer of material
slides along the pile surface. As the pile builds up, the finer fraction
deposits preferentially near the center and coarser fraction near the edges.

A blend of perfect structured particles of identical size will not segregate

after discharge from the blender. However, if there are size differences
between the particles, then segregation by size may occur. Even if all the
particles are chemically homogeneous, this segregation may cause
problems because it may affect other blend properties, such as solubility
or bulk density. There are many examples where this type of segregation
is important, such as with spray-dried laundry detergent and food
products, and other natural materials such as clay cat litter, coal and other
mineral materials. In these examples the particles vary in size, but are
virtually identical from a chemical standpoint.

In many other industrial-blending applications, particles not only differ in

size, but also in chemical composition. And in most cases, particles of
different size also have different chemical composition, so sifting
segregation not only results in physical property differences associated
with size distribution, such as bulk density and fluidization behavior, but
also differences in chemical composition.

Fluidization segregation and particle entrainment. Other common

mechanisms of segregation that act primarily on particle size are
fluidization segregation and particle entrainment or dusting segregation.
Fluidization segregation occurs where gas movement through a bed of
solids causes finer, lighter particles to rise to the top surface of a fluidized
blend of powder, while the larger, heavier particles concentrate at the
bottom of the bed. Particle entrainment or dusting segregation occurs
when fine particles in a blend are carried by air currents (such as during
transfer of a blend into a container) and then settle preferentially at the
container walls.

SAMPLING
Sampling is essential in determining the state of the blend in a blender, in
downstream equipment or in the final delivered product. Samples are
analyzed to measure the variables of interest to the application, such as
particle size, chemical composition, pH, dissolution rate, color and so on.
The overall average of the sample results represents the average
composition of the blend. Variations between samples provide a measure
of blend uniformity. Variability can be expressed by statistical measures,
such as standard deviation, coefficient of variation, or relative standard
deviation, as well as other measures [ 2].

The single most important rule of sampling is to collect a sample that

accurately represents the state of the blend at the point where it is
sampled. In many industrial processes this is a significant challenge due
to either the difficulty of getting a sample at the desired location or
because of the potential for sampling error that is introduced by the
sampling method. The two “golden rules” of sampling are to always
collect a full stream sample, and always collect a sample while it is
moving [ 3].

In the case of the cereal example given above, sampling is easy because
the sample is either a whole box of cereal or individual servings poured
from the box. In either case the sample is easy to get and it is possible to
analyze the entire sample. For other processes, significant challenges can
prevent the use of good sampling techniques. In an example involving
limestone at a circulating fluidized-bed boiler, the size distribution of the
limestone used for fluegas desulfurization came into question. This
became a critically important question because the plant could not meet
emission regulations above about 50% output, and the blame was falling
on the size distribution of the limestone produced by the onsite
preparation plant. The plant produced about 50 ton/h of limestone, and to
get a meaningful sample was not a trivial task. To satisfy all parties
involved, the plant had to modify its chute systems to put in large diverter
gates and position several operators to coordinate the collection of
simultaneous full-stream samples in 55-gal drum quantities.

Samples of this size pose another challenge in analysis. Usually it is not

practical or convenient to analyze such large sample quantities, so
samples must be reduced or split. Simply grabbing a small sample from a
large sample container can introduce significant sampling error,
especially if sifting segregation can occur. The preferred method of
splitting samples is to use a spin riffler, which can divide a sample into a
number of smaller splits, usually from 6 to 16 divisions. Spin rifflers are
capable of providing a high degree of uniformity between the individual
sub-samples.

Another challenging task is sampling from a static bed in a blender.

Clearly these samples cannot be collected while the material is moving,
so some means of getting a meaningful sample must be used. Simply
scooping material from the top surface is insufficient for many reasons,
including particle segregation, which could make the top surface
significantly different from material deeper in the bed. A sampling thief is
often used to collect samples from a stationary bed.

AN EXAMPLE
In an example involving plastic compounding, engineers for the project
had experience using a V-type rotating blender for similar materials at
other plants. Batch ingredients would be fed into a scale hopper and then
discharged into the blender. The blended batch would then be discharged
into a surge bin feeding an extruder. While this process had worked
successfully at other compounding plants, this plant would blend
ingredients with a wider size range. The most extreme case involved
blending the two materials shown in Figure 3, which consisted of
cylindrical pellets with a nominal size of about 3 mm and nearly spherical
beads with a size range from about 1mm down to 150 microns.

Engineers for the project ran tests in a small-scale V-blender and

evaluated blend uniformity by taking samples from the blender. The
overall blend results showed a lot of scatter. Some samples were
acceptable, while others were significantly outside the desired blend ratio.
Engineers also noticed that when the batch was discharged after a test, the
material in the receiving container was severely segregated. This
prompted development of a more rigorous sampling method that included
collecting samples from the discharge of the blender from beginning to
end. When this was done the results, shown in Figure 4, clearly showed
inadequate blending. Based on this realization, further testing was done to
determine if a tumbling type of blender could be used. After a series of
trials, a cylindrical blend vessel with an conical hopper and insert was
found to produce a blend within specification, and discharge a blended
stream to the downstream vessel.

This example illustrates several key points about blending, segregation

and sampling. First, the blender must provide the correct blending action
for the solids. These two materials are very free flowing and hence blend
easily. Blend time in the tumbling type blender is very short, on the order
of tens of revolutions. However, the other side of the coin is equally
important because the material also segregates readily, so maintaining the
material in a blended state is difficult. This required two modifications to
the geometry of the blend container. First, the dead zone near the center
of the blender had to be reduced because the blender geometry did not
allow enough diffusion to mix material in the center of the blender.
Convection dominated the action in the blender, and diffusion occurred
on the cascading surfaces, but some material remained poorly blended in
the center. In addition to the blending performance, the flow pattern that
developed during discharge had to be changed from funnel flow, where
some of the blend remained stagnant during discharge, to mass flow,
where all of the blend flowed as it discharged. In funnel flow, the highly
segregating material could de-mix along shear surfaces during discharge.
A mass flow pattern was achieved by designing a sufficiently steep
hopper-shaped blend container with an internal bullet to displace material
in the center of the container. The final configuration of the blender,
shown in Figure 5, produced the results shown in Figure 6.

The other important point illustrated in this example is sampling. Without

sampling the blender discharge stream from start to finish, it would be
impossible to know if the blend delivered to the extruder would be
uniform and within specification.

A blending process shouldn’t be considered as an independent step in a

block diagram, but really must be treated as a synergy of the mixing
activity, segregation and sampling to produce the desired end result.â•‘

A sampling thief is a probe that can be inserted into a bed of material to

collect a sample from below the surface. Many designs exist, but in their
basic form, they consist of a rod, frequently with a pointed tip. The rod
has cavities cut into it along its length which are covered by a tube that
fits closely over the rod and has openings that correspond to the cavities
in the rod. The tube can be rotated to cover the cavities or open them.
When the rod is inserted into the bed of material, and the tube rotated to
exposed the cavities, blended material will fill the cavities, which can
then be covered again and withdrawn. Thief sampling has been used
extensively and studied thoroughly. While it remains the only practical
way to collect samples from all positions in a static bed, it has been
shown to have biases and limitations that can cause sampling error. These
can come from properties of the blend and from operator technique. Some
typical sources of error include local segregation as material fills the
cavities, smearing as the thief is inserted into the bed, and preferential
adhesion or repulsion of some components of the blend.

Sampling will always introduce some degree of error, which can affect
the sampling results in three ways. Perhaps the most common effect is
that sampling increases the variability of samples, which produces data
showing that the uniformity of the blend is worse than it actually is. In
some cases, this presumed “fact” is used to suggest that a blend is
actually better than what the sampling data shows.

Sampling error can also produce bias, or an overall shift in the results,
indicating that the measured average from the collected samples is higher
or lower than the known blend composition. This type of error can be
confounding because people often jump to the conclusion that there is
“missing” or “extra” material, somehow. It is more likely that one
component adheres to, or is repelled by, the thief, or preferential flow of
one component into the thief shifts the sample composition. Another
possibility is that a sample is collected from an inappropriate location,
such as a known dead zone in the blender that may be holding the
“missing” material. When this type of error is observed it usually requires
additional detective work to determine the cause of the bias.

Sampling devices can also produce samples that show lower variability
than are actually present due to smearing of the sample. This type of error
is often referred to as “counterfeiting.” Counterfeiting is hard to detect
because it may produce results that show blending is good.

The purpose of sampling is to accurately quantify the state of a blend or

uncover blend quality issues, not to mask or overlook suspected
problems. Samples should be intentionally collected from suspected
regions where the blend may be less uniform, not just from the middle of
the blender, or during the “best” part of a production run. Fringe regions
may include the surface of the material, regions where the material may
have been stagnant during blending and at the beginning and end of
discharge from a batch bin.

Edited by Rebekkah Marshall

REFERENCES
1. Carson, J.W., and Royal, T.A., Techniques of in-bin blending, IMechE,
C418/056, 1991.

2. Fan, L.T., Chen, S.J., and Watson, C.A. Solids mixing. Industrial Eng.
Chem. 62, 53–69, July 1970.

3. Allan, T. “Particle Size Measurement”, 2nd Ed., Chapman & Hall Ltd.,
London, 1975.

AUTHOR

Thomas G. Troxel is a vice president at Jenike & Johanson, Inc. (3485 Empresa Dr., San Luis Obispo,
CA 93401; Phone: 805-541-0901; Email: tgtroxel@jenike.com). He holds a B.S. in Engineering
Science (ES) from California Polytechnic State University (Cal Poly) in San Luis Obispo. After
graduating as the Outstanding Senior in the ES program in 1981, he worked at General Dynamics Co.
He then helped to open Jenike & Johanson’s San Luis Obispo, California facility in 1982. Troxel has
been intimately involved in many aspects of the firm’s consulting and research activities on a wide
range of projects including flow properties testing, modeling, blending, pneumatic conveying and
fluidization. He has been a major force behind the firm’s expansion in mechanical design engineering
and custom-built equipment.

Click here for full pdf version of this article – includes all graphs, charts, tables, and author
information