RECIPES FOR FLUID MIXING

Understanding the process objective is essential to equipment sizing and

selection
David S. Dickey American Reactor Corp. Ramesh R. Hemrajani Exxon Research and
Engineering Co.
Often misconstrued as a unit operation, mixing is perceived by some to be a mature
technology. More accurately, however, mixing is the mechanics of contacting the fluid
phases that demand special attention during design and operation of various process
systems in the chemical process industries (CPI). Also, any description of the
mechanisms of mixing requires some details of the chaotic fluid motion. Such
information is only now being gathered, using laser Doppler anemometry and
computational fluid-dynamic models of flow processes.
Of course, engineers cannot afford to wait until all the details of the mechanisms of
mixing are fully understood. They have to make the most of the empirical -- as well as
fundamental -- knowledge available today. This is because mixing is at the heart of such
processes as heat transfer, mass transfer, and physical or chemical change. Together with
such parameters as processing time, temperature, pressure and catalyst activity, mixing
plays an important role in the control of yield and product quality.
Knowing the variables that determine the rate and degree of mixing is essential in
achieving the most effective process design. Too often, failure to do so results in
inefficient heat and mass transfer, slow reaction rates, unacceptable selectivity in product
formation, plugging of catalysts and reaction vessels and so on. Clearly, it is important to
seek design methods that are flexible, predictive and reliable.
A mixer may be designed for just mixing. But sometimes the indirect effects of mixing,
such as the reaction rates and heat- or mass-transfer efficiency, are the desired outcomes.
In these situations, mixing may be only a means to achieve a desired process objective. In
this sense, a mixer is unlike any other piece of process equipment that usually has certain
performance requirements of its own.
For example, a pump is required to deliver a certain height of fluid head at a given
flowrate. A heat exchanger is called upon to transfer a specified amount of thermal
energy from a hot to a cold medium. Similarly, a distillation column is designed to meet
the specification of the user in terms of the number of transfer units (NTU). A mixing
equipment, on the other hand, may be required to achieve some hard-to-quantify process
results.
A mixer wears many `hats'
Not surprisingly, most mixers are designed to accomplish several tasks, often within the
same application. For example, in many batch processes the mixer must keep up with
agitation objectives that change with the progress of a chemical reaction.
If a significant amount of liquid is added or removed during the process, the mixer must
be able to operate at different liquid levels. Also, each of the mixer objectives may
require a different intensity of mixing. What's more, this mixing intensity, which is
related to fluid velocity, tends to drop off quite rapidly with the distance of the fluid
particles from the source of agitation. Therefore, the ideal mixer for each situation may
be quite different, depending on the peculiarity of the objective.
The best design is likely to be a compromise among some competing demands of the
process. This is because the design goal is to come up with a mixer that can handle the
most difficult portion of the process, without causing major problems during execution of
the other steps.
Unless the liquid is highly viscous, or a very short mixing time is required, most
processes involving liquid blending are easy to handle with standard mixing equipment.
More difficult is to design mixers for chemical reaction processes. In addition to
changing chemical composition, reactions may change viscosity, create or eliminate
suspended particles, alter batch volume, and affect many other fluid properties related to
mixer performance. (Some of the specific issues are addressed in Part 2, p. 92.)
Heat transfer is another difficult area as regards to mixer design. It is difficult because
mixing can improve heat-transfer efficiency only on the process-side of the vessel, by
maintaining the uniformity of the batch temperature. And this may have a very limited
effect on the overall heat-transfer coefficient when mixing low-viscosity fluids. In
addition, even large changes in mixing intensity have only modest effects on heat transfer
through the stagnant fluid ``film'' near the wall.
Stirring has its limits
In cooling of high-viscosity materials, mixing may be counterproductive. This is because
the increased power input to the agitator is dissipated as heat, raising fluid temperatures
more rapidly than the enhanced cooling due to improved heat transfer. In this situation,
increased heat-transfer area and temperature difference are likely to have a greater effect
on overall heat-transfer rates than will mixing.
Solids suspension, and liquid and gas dispersion also require mixing. Solids suspension is
better understood than some of the other multiphase phenomena, such as gas-liquid
flows, because particle size, density difference and the resultant settling velocities help
quantify the mixing problem.
Often a part of storage operations, solids suspension maintains a slurry of uniformly
suspended particles. Other situations involve dissolution or formation of particles,
normally by physical, rather than chemical, means.
It is also possible that some liquid dispersions become suspended as solid particles during
such processing steps as suspension polymerization. In suspension of catalyst particles,
the catalyst remains unchanged, of course. But mixing ensures enhanced mass transfer
and chemical reaction.
Operations involving single-phase liquids also encompass many applications for mixing.
Where differences in composition and temperature cause problems, mixing can be used
as a means of improving uniformity. If the uniform conditions are suitable for a reaction
to take place, the products are likely to be consistent from batch to batch in terms of yield
and quality.
Differences in fluid viscosity and density can also be eliminated by single-phase
blending, although these differences can often be attributed to variations in fluid
composition and temperature. Similar problems involving two-phase liquids may also be
solved by simple blending.
Basic parameters, such as drop size distribution, may be important in liquid dispersion.
But a knowledge of surface tension and phase viscosity may be necessary to tailor a
mixer to meet the given process objective. The same applies to gas dispersion, where the
objective is usually interfacial mass transfer.
Most often, the transfer is from the gas to the liquid, but in some stripping applications,
the transfer of mass is the other way around. Because gas bubbles are usually small and
well mixed, the mass transfer is limited mostly by the resistance on the liquid side.
Intense agitation on the liquid side not only promotes mass transfer, it also reduces the
bubble size while increasing the surface area. As a result, gas dispersion is one mixing
application where more agitation almost always improves mass-transfer rate and
productivity.
However, the initial gains are met with diminishing returns as mixing intensity is
increased. Consequently, in such applications as waste treatment where efficient use of
power is critical, mixer designs strive for gas dispersion levels that lie just above the
required minimum. On the other hand, when productivity is important, as is the case in
process reactors, units are designed to deliver the maximum intensity of mixing.
Knowing the process is a must
Prior to selection of a mixer, the most critical step is to identify the problem. This step
enumerates the requirements of the physical or chemical process under consideration. In
other words, to meet the process objectives, the mixer must ensure:
--Reaction rate that is high enough to complete the process in a given time
--Process conditions that are suitable to achieve the prescribed percent conversion to the
product
--Production of a high percentage of the desired product relative to the less-desired
byproducts
--Sufficient heat-transfer rates
--Uniformity of temperature or concentration within specified limits
--Particle-size distribution (for gas-liquid or liquid-liquid systems) within prescribed
upper and lower limits
The above are the so-called positive requirements, which relate to the desired process
objectives. There are also a few negative requirements, which must be avoided to meet
the overall process objective. These include:
--Eliminating hot or cold spots at the heat transfer surface
--Preventing material from sticking to the wall
--Avoiding accumulation of stagnant material on the liquid surface
--Allowing no solids to settle or accumulate at the bottom of the vessel
While these requirements represent typical process objectives, they do not translate into
criteria that can be used to assess a mixing task. In fact, mixing is best evaluated in terms
of three principal parameters: intensity, difficulty and size of mixing.
How to gauge a mixing job
The intensity of mixing is related to fluid velocity. In other words, intensity increases
with a rise in the induced fluid motion. Because all a mixer does is produce fluid motion,
the consequences of that fluid motion must be considered relative to process
performance.
Fluid velocity, which accounts for both magnitude and direction of fluid motion, is an
essential parameter for quantifying mixer performance. Because the fluid velocity can be
used to relate the performance of a mixer with its other basic operating parameters, there
exists a sound basis for estimating mixing intensity.
In blending applications, mixing difficulty is characterized by fluid viscosity. In solids
suspension, such parameters as particle settling velocity and the percentage of solids
content contribute to mixing difficulty. Extreme values of two other parameters, namely
particle size distribution and particle shape, can also contribute to the difficulty in some
subtle ways. This is because very large particles or large size distributions may behave
very differently from uniform, small particles. Also, particles of flat or rough shapes may
be more difficult to suspend than spherical particles. (Factors that affect solids suspension
are discussed in more detail in Part 2 beginning p. 92.)
In gas dispersion, the quantity of gas compared to that of liquid has a direct influence on
mixing difficulty. Superficial gas velocity (that is, volume flowrate divided by vessel
cross-sectional area) usually provides a good general description for the quantity of gas
present in an agitated tank. Although the initial dispersion of a large quantity of gas is
somewhat difficult, such larger quantities of the gas generally provide greater surface
area for increased overall mass transfer. This makes it easier to meet certain process
requirements.
Much of the effect of the size of mixing relates to the quantity of material that must be
mixed. Whether the fluids are measured on a mass or volume basis is determined by the
design variables that need to be evaluated. For example, impeller size and speed can be
related to fluid volume while torque and power need to be related to mass quantities.
A secondary, yet important aspect of mixing size, is what may be called mixing shape.
This relates to the shape of the mixing vessel. For example, a quantity of fluid contained
in a cylindrical vessel, where the height of the liquid level is more or less equal to the
diameter, should be the easiest to agitate with a turbine impeller.
Unusually tall and short vessels with the same quantity of liquid may require multiple
impellers or some special designs. Often, mechanical or physical considerations, such as
high pressure or heat-transfer efficiency, make tall or slender tanks more desirable. For
similar reasons, large, field-erected storage tanks tend to be short relative to their
diameter.
Fluid properties, especially liquid viscosities, are critical to the correct assessment of a
mixing job. Low-viscosity, Newtonian liquids may be so easily moved that a small
propeller-type mixer will provide all the mixing needed to prevent liquid stagnation, even
in a large tank.
High-viscosity, non-Newtonian fluids may require the use of special impellers that act
directly on a large proportion of the material. Such large, close-clearance impellers do not
rely on induced liquid motion to bring about mixing action. As a rule, small impellers
may be adequate for small tanks, but much larger units are required for large tanks.
The quantity that better defines mixing characteristics and fluid motion as a function of
fluid properties is Reynolds number:
NRe = 0.17rN delta2/m (1)
where delta is the impeller diameter; N, the rotational speed; r, the fluid density; and m,
the fluid viscosity. The ``conversion factor'' 0.171 is explicitly indicated to ensure
consistency of the units (nomenclature box).
The significance of NRe is the same as that for pipe flow, representing a ratio of inertial
to viscous forces in the fluid. To reflect the mixer performance, impeller diameter and
rotational speed are substituted for pipe diameter and fluid velocity.
Viscous conditions exist at low Reynolds number (NRe less than 10), while turbulent
conditions prevail at high values of NRe (NRe greater than 20,000). The wide range in
the transition region (10 less than NRe less than 20,000) suggests that, even at high
viscosities, turbulent conditions may exist near the impeller while viscous conditions may
prevail near the wall.
Mixing and pumping action
An agitator made up of one or more radial-flow turbines, a driver and a gearbox promotes
fast mixing by producing centrifugal pumping action. However, axial flow domiantes if
the agitation is produced by axial-flow turbines. In this sense, a mechanical mixer acts as
a centrifugal or axial pump by producing a circulating capacity and a velocity head.
Unlike a pump, however, the internal circulation in a mixer is not as confined and
directed as in a pipe. The pressure head generated by the pump is equivalent to the
mixer's velocity head, which can also be considered a degradation of kinetic energy. H is
proportional to shear in mixing because the head from kinetic energy generates shear
through the jet or pulsating motion of the fluid.
Total flow in a mixer includes the primary flow (induced directly by the impeller) and the
secondary flow (entrained by the liquid jet). The entrainment depends on the mixer
geometry, particularly that of the impeller. As a rule, small impellers enhance the
contribution of entrainment to the total flow more effectively than do their large
counterparts.
However, because of shifts in flow patterns it is difficult to formulate simple rules that
are generally applicable. Also, because of recirculation and intermingling of fluids in a
mixer, it is not possible to define and measure simple flow quantities.
The primary flow in a mixing tank can be measured by invasive tools, such as pitot tubes
and velocity probes. However, until recently it has been difficult to measure total flow
because of the limitations of these techniques in measuring the randomly directed
velocities of the entrained fluid. Of late, developments in laser Doppler anemometry have
greatly improved the means to measure velocities and identify the flow patterns. Now the
challenge is to devise ways of using this information, and somehow relate it to the
process objectives.
Depending on the process needs, one can change the combination of pumping action and
fluid shear by changing the impeller diameter and agitator speed. As shown in Figure 1, a
large-diameter impeller at slow speed would provide the high pumping action necessary
for such applications as blending and solids suspension. On the other hand, an impeller
with a small diameter and high speed is needed for high-shear systems where mass
transfer is important.
Shear stress and mixing
Whenever there is relative motion between liquid layers, shearing forces (also known as
shear stress, which is related to the fluid velocity gradients via the coefficient of
viscosity) produce fluid intermixing. This is how mixing is carried out by dispersing gas
bubbles, and stretching or breaking liquid drops.
Shear rate -- defined by the velocity gradients -- is what is delivered by the mixer. These
velocity gradients can be produced either by entraining the liquid with a propeller jet, or
by creating velocity profiles with rotating impellers.
Velocity gradients on the propeller jets are proportional to the fluid velocity at the blades,
while the velocity profiles induced by an impeller are related to its tip speed, pN delta .
Of course, the shear rate in different parts of the mixing tank will be different. As a result,
there are several definitions of shear rate:
--The average shear rate in the impeller region is proportional to the rotational speed. The
proportionality constant lies between 8 and 14 for all impeller types
--The maximum shear rate occurring near the blade tip is proportional to the impeller tip
speed, pN delta
--The average shear rate in the entire tank is an order of magnitude less than the average
shear rate in the impeller region
--The minimum shear rate is about 25% of the average shear rate in the tank
Twists in impeller types
There are many different types of impellers used for agitation. Some, such as turbine
impellers, are built for the economy of construction. Others such as the close-clearance
impellers to handle high viscosity, are designed for special applications.
A few are built the way they are because of special fabrication requirements, such as the
rounded corners and smooth curves of glass-coated impellers. Some impellers are
designed to satisfy process efficiency without incurring excessive manufacturing costs.
And, of course, there are some that seem to defy any reasonable explanation, other than
someone's speculation that the design should work.
Turbine impellers have the widest use in low- and medium-viscosity applications. They
are also used, almost exclusively, for solids suspension and gas dispersion.
Most of today's turbine-style impellers have evolved from either a simple two-blade
paddle or a marine propeller. Those derived from a paddle have vertical blades to
produce radial flow. The ones derived from the marine propeller have angled blades, and
they produce axial flow.
One of the most commonly used impellers is the disk-style turbine (see top of p. 83).
Although rarely applied in liquid blending, this radial-flow device is used quite
extensively for gas dispersion. During general processing, the disk maintains a stable
discharge pattern by separating return flows from above and below the impeller. In gas-
dispersion applications, the disk tends to ensure that the rising gas bubbles move through
the blade tip region at the first pass.
In addition, the disk provides a convenient means for attaching six blades to a central hub
and shaft. Equipped with multiple bolt holes at different radial positions, it also offers
some field adjustability.
Axial-flow turbines are used for solids suspension, blending of liquid, and a diversity of
other situations. One of the most common and versatile designs is that of the pitched-
blade turbine (see p. 83). Because impellers of different size can be made by using plates
of different sizes placed on the same hub, this device is simple to build.
The discharge flow is primarily axial, although it contains some radial and tangential
components. Local turbulence is generated through a series of trailing vortex structures
behind each blade.
Over the last decade or so, several impeller designs have been introduced, offering
performance advantages over the conventional pitched-blade turbine. The advantages
include lower capital and operating costs, and often improved performance. For lack of a
better term, these devices are called ``high efficiency'' impellers (see p. 83).
The primary difference between the high-efficiency impellers and the pitched-blade
turbines is the shallow angle at the tip of the blade. Many of the high-efficiency units
have three, instead of four, blades that may have a curved cross-section with a taper or
twist from hub to tip.
All these features tend to increase the percentage of axial flow (relative to the total flow
in the mixer), and reduce the drag around the blade. The other performance features of
these turbine impellers relate to flow patterns.
Importance of mixer geometry
The recirculating fluid motion is instrumental to the flow patterns in a mixer. The vessel
geometry dictates the possible distribution of the developed flow patterns. Typically, the
impeller diameter, delta , lies between 10% and 70% of the tank diameter, T.
Very large impellers are often used in mixing of viscous fluids, where the impellers are
required to create additional fluid motion in the outer annular area.In most low-viscosity
applications, tanks are equipped with baffles to limit swirl. Baffles with widths, WB,
ranging from one-twelfth to one-tenth the tank diameter are usually mounted a short
distance from the wall. Such geometric arrangements ensure that most of the annular,
induced-recirculation area of the tank is between the baffles.
A rotating impeller creates liquid motion primarily by creating a differential pressure
across the blade. While related to the blade angle, the direction of the liquid motion
cannot be quite perpendicular because fluid does not pass directly through the blade. The
result: Fluid tends to move away from the direction of the blade faces, while it acquires
motion parallel to the face of the blade. The fluid motion in the immediate vicinity of the
impeller blade is very complicated. In many ways, it is similar to the flow around marine
propellers on ships.
Impeller-induced flow patterns
The discharge characteristics of an impeller determine the recirculation patterns of the
fluid. As a result, very different flow patterns are generated by radial- and axial-flow
impellers. For example, radial-flow impellers create two recirculation loops, while axial-
flow units produce various types of single loops. Typical flow patterns for a radial-flow
turbine, pitched-blade turbine and a high-efficiency impeller are shown in Figure 2.
Indeed, it is the diversity of recirculation patterns that makes an agitated tank a versatile
piece of process equipment. The same fluid can be processed over and over again, with
different portions of the fluid being mixed during each pass through the impeller region.
Axial-flow impellers can handle a wide variety of mixing applications, making an
understanding of their basic flow patterns -- and some of the subtleties -- critical to the
success of a physical or chemical process.
Bear in mind that fluid motion accelerates as the fluid is drawn into an impeller, ensuring
high fluid velocities to be present in the discharge. This discharge flow pattern leaving
the impeller and the resulting recirculation patterns subsequently influence the fluid
velocity and flow pattern entering the impeller region. As inputs to the impeller, these
entering patterns and velocities in turn influence the discharge flow.
The important aspect of the close connection between the entering and discharging flows
is that recirculation is an integral part of impeller flow, even around the blades. As a
result, identical impellers in different tanks may have entirely different flow patterns.
An easy way to consider the effects of recirculation is to look at the simplified flow
patterns in a tank for different diameters of the same-style impeller (Figure 3). The
changes in the ratio of the diameter of the impeller to that of the tank is called `` delta /T
effects.''
At small delta /Ts, the impeller operates almost independently of the tank. The discharge
pattern in the immediate region of the impeller would be much the same in any larger
tank.
Flow from a pitched-blade turbine is downward with an outward and rotational
component (Figure 3). The downward pattern changes direction at the bottom of the tank,
goes outward toward and up the wall, and then back into the return flow to the impeller.
At the tank wall, the baffles eliminate most of the rotational component of flow. The
result: a general pattern that is axial rather than rotational.
The flow pattern at small delta /T generates intense agitation in the impeller region and a
large, less-intense recirculation pattern. The power required to achieve similar process
objectives is usually highest for small delta /Ts. The flow patterns for medium and large
impellers in Figure 3 show a more-desirable distribution of strong recirculation flow
generated by the impeller.
A very large impeller approaches the limits of practical recirculation. Here the influence
of the impeller extends beyond the blade tip. As a result, the swept area of the large
impeller approaches half the cross-sectional area of the tank, and recirculation is
diminished. The large turbine also establishes separate patterns above and below the
impeller, and secondary recirculation near the bottom.
One important design consequence of these flow patterns for different delta /T ratios is
that mixer performance may change unevenly with impeller diameter. For example,
solids suspension or surface motion may be very different as a function of delta /T.
Recent developments indicate that axial-flow impellers that generate greater axial flow
with less of a radial and a rotational component are very efficient in delivering mixing
action for the same power or torque. To further improve efficiency, impeller blades with
a curved cross-section -- designed to imitate an airfoil shape -- can be used.
The more-axial, less-radial flow profile of the high-efficiency impeller is likely to do a
better job of creating top-to-bottom recirculation. Compared with a pitched-blade turbine,
some of the turbulence and vortex structure near the blade tips is also reduced. This loss
of the effect of turbulent mixing seems to have little adverse effects on the mixer's
blending performance.
Another factor that can influence impeller performance is the liquid level. If the liquid
level is lowered during a process and the mixer happens to be equipped with a short shaft,
the impeller may find itself rotating ``up in the air'' above the liquid level. The other
extreme is when the liquid level becomes significantly greater than (more than 1.2 times)
the tank diameter. In this case, additional impellers may be required.
With appropriate impeller spacings, multiple impellers tend to reinforce a single
recirculating-flow-pattern. Widely spaced pitched-blade turbines or radial-flow turbines
tend to create multiple recirculating cells. Under some circumstances -- where stages with
slightly different conditions may improve selectivity or yield, for example -- these cells
may offer a processing advantage.
Viscosity produces one of the more-subtle effects on flow patterns. With increases in
viscosity, the discharge from the impeller is less effective at inducing recirculating
motion. As the strength of the return flow to the impeller diminishes, axial flow is more
difficult to achieve. Also, blade angle and shape become less important because the
blades tend to act like bars attached to the shaft.
The result of increased viscosity is an eventual radial-flow-pattern with almost any style
of an impeller. The transition from axial to radial flow for a pitched-blade turbine is
shown in Figure 4. Complicating the picture is the fact that many viscous fluids also have
non-Newtonian characteristics, making the exact Reynolds number or transition range
difficult to predict.
Evaluating mixer performance
Given the complexity of the mixing process, it is impossible to predict all the flow effects
associated with different impellers operated under various circumstances. Yet a
systematic approach to assess mixer performance has been developed in a series of
articles [1]. In another article [2], the author defines methods for quantifying mixing, and
shows how agitation intensity relates to mixer design.
A convenient way to simplify the performance evaluation of a mixer is to assign a scale
of agitation. Ranging from 1 to 10 for the minimum to maximum levels of agitation, the
scale covers, by definition, the entire spectrum of mixing. To justify this one-to-ten scale
that is based on fluid velocity, it has been argued [3] that fluid velocity is more relevant
in evaluating mixer performance than any other simply defined quantity.
In fact, the details of the velocity profile are of secondary importance. Most important are
the magnitudes of velocities that tend to produce characteristics associated with similar,
although not identical, mixing objectives, especially in equipment of different sizes.
Mixing experts differ on how to create the similar intensities of agitation, when applied to
the same vessel, however.
For example, two authors [2, 3] use constant torque, N2D5, for equal motion, while
another [4] shows that equal Reynolds number produces an ``equal process result.''
Fortunately, the differences in the calculations resulting from these assumptions are small
in single tanks.
Assuming constant Reynolds number in a given tank, one can obtain the agitation
intensity, which is conveniently defined on a scale from one to ten, from the following
equation:
IA = 6.9 10-4Fm[N delta2/V1/3] (2)
where Fm is a performance characteristic accounting for viscosity and impeller type. In
the turbulent range, Fm has values of 1.0 and 0.65 for a pitched-blade turbine and high-
efficiency impeller, respectively (Figure 5). Before entering the graphs in Figure 5, use
Equation 1 to determine NRe.
Keep in mind that these curves do not portray exact relationships for all similar impellers,
or for all types of applications. They are meant to provide good comparative estimates for
typical impellers. The accuracy of these relationships are especially limited for the high-
efficiency impeller because there are several designs provided by different manufacturers
for different applications.
Specifying design parameters
Clearly, whenever it is possible to evaluate mixer performance in some rational fashion,
one must be able to use that information for design purposes as well. As indicated in the
literature [1], the complete design procedure for mixing equipment requires evaluation of
both the process objectives and the mechanical capabilities of the unit. For blending
applications, a preliminary design estimate for the agitator speed can be obtained from
the following equation:
N = 1450 [IAV1/3/(Fm delta2)] (3)
When used in design calculations, Equation 3 above -- valid for impeller-to-tank-
diameter ratios in the range: 0.2 less than delta /T less than 0.7 -- does require some
initial estimates for impeller diameter and viscosity factor. Typical delta /T values may
range from 0.25 to 0.5. Small impellers are commonly used for low-intensity agitation,
while larger devices are preferred for high-intensity applications.
If one picks the value of the viscosity factor corresponding to the turbulent range, an
initial estimate of agitation speed can be computed. Following this initial estimate, one
can compute the Reynolds number and have the value of the viscosity factor rechecked.
A few such iterations between agitator speed, Reynolds number and the revised viscosity
factor should provide a reasonable estimate for design. Note that the design speeds are
usually adjusted to speeds obtainable with fixed gear reductions from standard motor
speeds.
An appropriate agitator speed may be based on certain standard features of the
equipment, but an impeller diameter can be chosen to meet the user requirement more
closely. This is because an impeller can be specified to be of any diameter; there are
virtually no requirements for fixed increments. Therefore, an alternative design approach
would be to specify the agitator speed and then compute the impeller diameter:
delta = 38 [IAV1/3/(FmN)]1/2 (4)
Before proceeding further, the authors would like to offer this piece of advice: Do not get
too precise with your design estimate based on any of the correlations. These equations
are designed to provide only estimates. Besides, differences between agitation levels of 6
and 7 may be almost impossible to distinguish.
The next step in the design procedure is to size the motor used to turn the impeller. Once
impeller diameter and shaft speed have been determined, power requirements are well
correlated. The correlation is based on the power number, NP, for the impeller:
NP = 9.5 1014[P/(rN3delta5)] (5)
The power number is analogous to the friction factor for pipe flow. It remains constant
for turbulent conditions. One can adjust the value of NP for turbulent conditions by
exploiting certain factors of impeller geometry [5]. Typical power numbers for the two
different impellers shown in p. 83 are: NP = 1.37 for pitched-blade turbines; NP = 0.33
for high-efficiency impellers.
However, when flow shifts into the transition and viscous ranges, the power number
increases with decreasing Reynolds number. Then, NP values for an impeller depend on
such details as blade width and so on.
The effects of viscosity on NP are similar for similar-style impellers. Therefore, instead
of providing a series of curves for different values of NP, it is more appropriate to
estimate a power correction factor, FP (Figure 6), which is analogous to the viscosity
factor used above to characterize mixer performance. Both factors have a value of 1.0 in
the turbulent range.
The final step in using NP and FP is to compute impeller power. Given the fluid density,
one can calculate the shaft speed and impeller diameter from the equations presented
above. The impeller power is given by
P = 10-15rNPFPN3delta5(6)
Given the calculated impeller power, the actual motor power is usually chosen to be a
value corresponding to the next larger, standard motor size. The calculated power
required by the impeller may be 85% or 90% of the actual motor power. This partial
loading of the motor may be more than just a safety factor. The extra power accounts for
the power losses through the gear reducer, shaft seals, and so on.
To sum up, before specifying a mixer the user must understand and clearly define the
process objectives. These objectives must be converted to operations that can be
accomplished by fluid motion. An understanding of velocity patterns helps solve the
mixing aspects of the design problem.
The defined scale for agitation intensity provides a convenient criterion that can be used
in the sizing of equipment to meet process objectives. Last, but not least, is the caveat
that the diversity of mixer applications allows exceptions to all the rules presented here.
In those cases, experience-based engineering judgment is the best recourse to achieving
the desired process results.
NOMENCLATURE
delta Impeller diameter, in.
FPImpeller power factor
FmViscosity performance factor
HVelocity head, in.
IAAgitation intensity number
NShaft speed, rpm
NPPower number
NReReynolds number
PImpeller power, hp
QCirculating capacity, gal/min
TTank diameter, in.
VVolume, gal
WBBaffle width, in.
mViscosity, cP
rDensity, lb/ft3
Edited by Gulam Samdani
[Photograph]

Photograph: When scaled up, a mixer may perform quite differently from its laboratory-
scale counterpart
[Photograph]

Photograph: The high-efficiency impeller offers greater axial flow (relative to the radial
and tangential components) and significantly reduces the drag around the blades. The
simple disk-style turbine is most often used for gas dispersion. The pitched-blade turbine
combines axial, radial and tangential motions to bring about thorough mixing
Lightnin
 [Illustration]

Figure 1: For the same power dissipation (vertical axis), generation of shear or local
eddies prevails over pumping or bulk fluid motion in a mixing tank, for relatively small
diameter and high speed of the impeller. The opposite is true for large diameter and low
speed of the impeller
CARLA MAGAZINO
[Illustration]

Figure 2: The disk-style turbine creates the least amount of axial flow, forming
recirculation patterns above and below the rotating device. The pitched-blade turbine
boosts the axial flow, but not as much as the high-efficiency impeller
[Photograph]

Photograph: Proper installation of a production-scale mixing equipment is critical to its

efficient operation and maintenance
Lightnin
[Illustration]

Figure 3: Increasing the ratio of impeller to tank diameter ( delta /T) results in gradual
changes in the flow patterns above and below the pitched-blade turbine. While the
recirculation patterns at the bottom get bigger, there is more-thorough mixing with
increasing delta /T
[Illustration]

Figure 4: Fluid viscosity can change the predominantly axial-flow pattern to a more-or-
less radial one, with major recirculation patterns above and beneath the rotating blades.
Highly viscous fluids often behave in a non-Newtonian fashion, which makes the
corresponding Reynolds number for the flow transition difficult to predict
[Illustration]

Figure 5: The difference in the viscosity factor for a pitched-blade turbine and a high-
efficiency impeller gives rise to different agitation intensities (via Equation 2)
[Illustration]

Figure 6: Given the Reynolds number, one can obtain the power factor from the curve for
the appropriate type of impeller, and use it in Equation 6 to determine the power
requirements. The power factor levels off to a value of 1.0 in the turbulent-flow regime

[Biography]

David S. Dickey is the technical director for American Reactor Corp. (1700 Dalton Drive,
New Carlisle, OH 45344-2307; tel. (513) 849-6264). He has more than 15 years of
experience in the design and application of chemical process equipment, with special
emphasis on mixing and agitation technology employed for spearations and heat transfer.
He has been the past chairman of the American Institute of Chemical Engineers (AIChE)
Equipment Testing Procedures Committee, and has contributed to the procedure for
mixing equipment. A member of AIChE, American Soc. of Mechanical Engineers and
American Chemical Soc., he has a B.S. from the University of Illinois, and a master's
degree and Ph.D. from Purdue University, all in chemical engineering.

Ramesh R. Hemrajani is an engineering associate in chemical engineering technology at

Exxon Research and Engineering Co. (P.O. Box 101, Florham Park, NJ 07932-0101; tel.
(201) 765-6478). A recognized expert in mixing technology, he specializes in difficult
mixing problems related to chemical manufacturing, synthetic fuels development and
crude oil production. He received his Ph.D. in chemical engineering from Northwestern
University and a post-doctoral fellow at McMaster University. Prior to joining Exxon, he
worked for Aluminum Company of America.

Made a mess of mixing? Try some proven answers

Good design and proper testing can eliminate some puzzling problems, as
shown in this series of operational examples
While the mixing business has seen a number of substantial innovations in recent years --
e.g., the emergence of high-efficiency mixers; the use of lightweight, advanced
composite materials -- designed to improve mixing speed and process performance, good
operational results often owe more to proper system design than to new mixing
technology. ``Everything is specific to a particular process, and when you consider all the
multiphase effects, the problems are often extremely subtle,'' says Gary Tatterson, a
professor at Texas A&M University's Dept. of Mechanical Engineering (College Station).
He adds that ``mixing is often taken for granted in the overall process, but if it doesn't
work it becomes a major problem.'' Some interesting case histories of mixing mishaps
were discussed in a special symposium aimed at consultants, equipment manufacturers
and research and plant engineers, and entitled ``War Stories in Industrial Mixing,'' that
Tatterson chaired at AIChE's 1987 Spring Annual Meeting, held in Houston, Mar. 29 -
Apr. 2. The examples included solids suspension in large tanks, reactor mixing, erosion
of rubber-lined impeller blades in an abrasive application, and mixing problems
experienced in liquid-liquid extraction. TRY IT BEFOREHAND -- A basic theme of the
symposium was empiricism: the importance of adequately testing a mixing-system design
before building the equipment and installing it in the operating unit. ``I would say that the
overall principles of mixing are reasonably well understood, but there's a lot of detail you
have to work out for each application,'' comments James Y. Oldshue, vice-president of
Mixing Equipment Co. (Rochester, NY), one of the speakers at the meeting. The
empirical approach is probably justified, especially because new applications account for
a substantial portion of mixer sales, say manufacturers. This is true for Otto H. York Co.
(Parsippany, N.J.), says Roger W. Cusack, the company's manager of process equipment,
who adds that ``most applications are in areas whose characteristics are not well
understood, such as specialty chemicals, new pharmaceutical products and wastewater
cleanup.'' Mixing Equipment's Oldshue described a case that involved the suspension and
leaching of uranium solids (about 78 wt %) in ten 1,400-m tanks that measured
approximately 12 m in dia. by 12 m deep. The mixers had been designed to use 200 shaft
kW*, and the impellers were 45-degree axial-flow turbines measuring 178 in. in dia
(including a protective rubber covering). After startup, operators found that the mixers
were drawing only 140 shaft kW, and that the viscosity of the slurry was below
specifications. The reason, Oldshue told the meeting, was that the user had added several
inches of rubber thickness to the impellers, which changed both their profile and
efficiency. The problem was solved by lengthening the impeller blades to obtain mixer
shaft-power of 200 kW (the alternative would have been to restore the profile of the
blades to that in the original design). GAS FOULS IT UP -- Hoechst Celanese Corp.
experienced a variety of interrelated problems with a new batch reactor designed to make
multiple products by catalyzed direct esterifications. Michael Lakin, a staff engineer with
the company's technical center (Corpus Christi, Tex.), explained the situation at the
meeting. In the process, solvent, alcohol (liquid or solid), catalyst and an organic acid are
placed into the vessel. The charged materials are then agitated, and heated with internal
steam heaters, while gas is sparged from the bottom of the reactor to help maintain the
activity of polymerization inhibitors. There were problems from the start, says Lakin: low
raw-material efficiencies (confirmed by residual solids), excessive acid carryover,
unwanted polymerization, and long reaction times. Vendor suggestions for improvements
centered on greater mixer horsepower, and on changing the pitched-blade, axial-flow
impeller for similar or high-efficiency impellers. Celanese and the vendor tested new
agitator designs at 1/10th the scale and discovered that the presence of gas (sparged gas
plus vapors generated in reaction) prevented the pitched-blade turbine from providing
adequate fluid circulation and solids suspension. Inadequate solids suspension was a
serious problem that correlated with the low material efficiencies and long reaction times.
The answer proved to be a dual-impeller system. A radial-flow impeller (flat-blade disc
turbine) was placed just below the liquid surface, and an upward pumping, pitched-blade
axial-flow impeller was installed near the bottom of the vessel, below the heater bundles.
(Originally, the pitched-blade unit was higher -- between the heater bundles.) According
to Lakin, the tests showed that the original agitator was designed simply for blending and
solids suspension, but because gas was present solids suspension was not the controlling
design parameter. ``A critical piece of information had not been properly identified and
communicated, or had been neglected in the initial equipment selection and design
process,'' Lakin told the meeting. He observed that chemical companies are often
reluctant to reveal sensitive process information, but noted that this barrier can be
overcome by the performing of in-house studies or by signing confidentiality agreements
with vendors, or both. A general approach to the design of three-phase systems is to base
it on the requirements for solids suspension, then add an overdesign factor for the gas,
says David Short, a mixing consultant in the engineering department of Du Pont Co.
(Wilmington, Del.). ``The gas is going to be fairly well dispersed anyway,'' he explains,
``but it will interfere with solids mixing. So after you have designed for solids, you may
have to increase the rpm of the mixer by up to 70% to keep the solids in suspension.''
Along with fellow Du Pont consultant Arthur Etchells III, Short manages a Du Pont in-
house mixing laboratory for troubleshooting and testing process designs that are not well
understood. The company has a 0.9-m-dia. by 1.8-m-high vessel with a 4-kW variable-
speed mixer, and a 1.5-kW variable-speed drive that is used in several smaller tanks.
Short says that although mixer manufacturers make their testing facilities available to
customers, an in-house laboratory is much more convenient, ``especially if you are in a
hurry.'' However, he agrees that a company must be large to justify such a facility, and
notes that Du Pont's laboratory also does a lot of work for the company's customers.
While energy saving still remains an important objective, the firm has learned that if
better mixing achieves a 1% yield improvement, the mixer can be overdesigned by
10,000% before the extra energy cost negates the chemical saving. On the other hand,
Short says that an additional, unnecessary impeller on a shaft adds a cost of about
$500/(yr)(hp). SOME OBSERVATIONS -- Many common mixing processes, such as
paint mixing, are not generally understood because information specific to those
techniques is not in the literature, says Short. Du Pont, a major supplier of titanium
dioxide to paint companies, went through extensive testing to develop what it feels to be
the best mixer for its customers' TiO slurry tanks: a four-bladed, radial turbine mounted
5-10 cm off the bottom, and a four-bladed, pitch-bladed turbine at the tank's midpoint.
``Mixing is an area in which people are learning a lot, but not inventing a lot,'' says
Etchells. Tatterson, of Texas A&M, holds to the same opinion. ``Mixing used to be a
black art,'' he notes, ``but now a lot of research is being done by institutions on the
fundamentals and mechanisms, particularly in Europe and Japan.'' Sometimes when a
mixer does not appear to be working well, the problem may lie elsewhere, such as the
tank inlet or outlet pipes. This is especially true for slurries, says Short. ``We had one
case where a pipe had not been sized correctly it was too big, and the flow velocity was
inadequate, so the solids settled and blocked the pipe, restricting the flow even more. The
operator thought it was a mixing problem because there were solids in the bottom of the
mixer, but once we got the feed velocity right everything worked beautifully.'' Blade
erosion is a major problem in slurry mixing, says Julian Fasano, technical director of
Chemineer Inc. (Dayton, Ohio). At the symposium, he told of an ore-slurry application in
which the substitution of four high-efficiency blades for three pitched blades on an
impeller doubled or tripled the life of the rubber covering. Originally, the covering had
lasted little more than six months. The high-efficiency blades have a contoured shape, in
contrast with the single flat plane of pitched blades. This minimizes microflow, the main
cause of erosion in this case, according to Fasano. Also, impellers fitted with these blades
typically generate about 70% more flow than a standard impeller, thus reducing the
power requirement. ``In this particular case, the operator saved about 2 million kWh/yr in
10 tanks,'' he says. MORE CASE HISTORIES -- Several mixing problems and answers
that involve agitated liquid-liquid extraction columns were described by York's Cusack.
One concerned a 3-ft-dia. by 40-ft-high reciprocating-plate column used in a process for
making an agricultural chemical. The process worked satisfactorily when piloted at the 1-
in.-dia. by 8-ft scale, but the production column recovered only 99.6%, against a
guaranteed 99.8%. This small difference was significant in monetary terms, Cusack told
the meeting. The system had been designed for the light phase (fed from the bottom) to
be continuous, while the heavy phase was dispersed from the top. But tests showed that
the continuous phase's specific gravity was 0.2 higher at the top of the column than at the
bottom, and this caused large top-to-bottom (axial) circulation of the continuous phase,
resulting in poor performance. The answer: The heavy phase was made continuous, and
the light phase dispersed; this improved product recovery to 99.8-99.9%. Cusack noted
that the problem was not a big factor in the original pilot column, but became significant
as the diameter increased. In another case, extraction of a pharmaceutical product from a
whole fermentation broth was impeded by the broth's tendency to form emulsions, so the
manufacturer used batch extraction, followed by three passes through a centrifugal
extractor. The processing of 9,000 gal of whole broth took about 20 h and required
30,000 gal of chloroform (the solvent), said Cusack. York substituted a reciprocating-
plate extraction column, and reduced solvent usage per batch to 13,500 gal, with
significant savings in maintenance and operating costs. Cusack stresses that the
reciprocating plate unit works because of its uniform shear characteristics, but that other
types of agitated extraction-columns may not be suitable for systems that tend to
emulsify. He also referred to a sudden emulsion problem in an extraction column that had
operated successfully for several years in a chemical plant. It was suspected that the
cause was grease from the stuffing boxes of new solvent-feed pumps that had recently
been installed (the previous pumps had mechanical seals). The pumps were replaced by
new ones with mechanical seals, and the problem disappeared. In a pilot-plant test of a
process involving an aqueous solution of a food product, the feed would coalesce on
stainless-steel internals, preventing good dispersion in the column. Cusack said that this
was solved by substituting Teflon for the steel. At the mixing symposium, a method for
designing new agitators or mixers to match process modifications in existing equipment
was described by David Barclay, a senior engineering scientist with Vista Chemical Co.
(Ponca City, Okla.). Barclay plots a line of constant process effect for one parameter
(e.g., power),. This is combined with data from plant tests and/or lab simulations to
develop agitator dimensions.
*Power delivered by the impeller to the liquids. Source: Hoechst Celanese Corp.
Gerald Parkinson
[Photograph]

Dual-impeller system (right) provides better circulation than original design (left)
[Photograph]

Vessel is used in various multiphase studies at an inhouse mixing lab Du Pont Co.

UNRAVELING THE MYSTERIES OF SHEAR-

SENSITIVE MIXING SYSTEMS
This article represents an important contribution to the understanding of
the fluid mechanics involved in shear-sensitive mixing. It offers a fresh
interpreta- tion, original equations, and the first design proce- dure.It
enables the engineer to easily calcu- late shear rates, flowrates, and
agitation intensity.
Richard L. Bowen, Jr., Tensco
Certain applications of open-impeller mixing require only fluid motion in the agitator
vessel. These flow-sensitive applications have until recently been satisfied in the
chemical process industries by axial-flow impellers -- propellers for small applications,
and 45-deg. 4-blade pitched-blade turbines for large ones 1. Approximately 80% of all
mixing applications -- including the common ones of blending and solids suspension --
involve flow-sensitive mixing 2. In the other 20%, shear rate, with the corresponding
shear stress, is the controlling variable. These shear-sensitive applications are handled by
the radial-flow turbine. Types of radial-flow turbines The radial-flow turbines considered
here are shown in Fig. 1. One of the earliest is the flat-blade disk turbine, also known as
the "Rushton" turbine (Fig. 1a) 3. It was used almost exclusively with all top-entering
agitators until mixing parameters became better understood, then the axial-flow turbine
was instead applied to flow-sensitive regimes. The flat-blade disk turbine is still used in
applications requiring high shear or turbulence, as well as for gas-liquid contacting,
because the flat disk forces the gas, introduced from below, along a path into the
discharge jets (in contrast to the vertical path through an impeller without a disk).
Originally, the number of blades varied from four to eight, or even ten, but the norm
today is six blades. Dimensions have been standardized: disk diameter at 2D/3, blade
width at D/5, and blade length at D/4. More studies have probably been made with this
type of impeller than with all the others put together. The flat-blade turbine is sometimes
supplied for applications requiring relatively high shear (Fig. 1b). This turbine originally
had six or eight blades; but now, four or six blades having a W/D ratio of between 0.15
and 0.20 are common. Flows generated by impellers The axial-flow turbine generates
flow downward to the tank bottom, then up the sides and back down the center to the
turbine suction. The radial-flow turbine, on the other hand, produces flow radially from
the turbine blades toward the sides of the vessel, where the flow splits, one part going
upward along the sides and back to the suction, the other going downward along the sides
and bottom, then back to the suction. Any type of impeller produces two different flows.
The first is the discharge flow, which is confined to the diameter of an axial-flow
impeller and the blade width of a radial-flow impeller. The discharge flow entrains fluid
at the boundaries to produce the total, or circulation, flow. In the case of the axial-flow
impeller, the discharge flow is usually of no concern to the engineer, because it is the
circulation flow that turns over the contents of the vessel to achieve the desired results.
The discharge flow from a radial turbine is represented by:
(See June 9, 1986 issue of Chemical Engineering for equation)
Here, Nq is a constant known as the pumping, or discharge-flow, number; it is dependent
on the type of impeller, system geometry, and Reynolds number. In determining values of
Nq for various impellers, early researchers always produced a unique value that appeared
to be independent of system geometry. However, all the values were of the same order of
magnitude. Shear produced by radial-flow turbines Although the engineer would not be
interested in the discharge flow from axial-flow impellers in flow-sensitive applications,
such is not the case with radial-flow turbines in shear-sensitive applications. In the case
of radial turbines, it is the discharge flow that produces shear, and the circulation flow
that returns the fluid to the zone of shear. In a rotational viscometer, shear rate is
proportional to speed for Newtonian fluids, with shear stress indicated on a torque scale.
Viscosity = shear stress shear rate. In mixing, the calculation is reversed. Shear rate is
calculated from the velocity profiles at the discharge of the turbine blade. Shear stress is
then determined by multiplying this rate by the viscosity. It is shear stress that ultimately
reduces particle size and disperses droplets in emulsions and suspensions. Although
viscosity may have little effect on power consumption in the turbulent region, the shear
stress produced by a particular impeller and speed is proportional to viscosity. A fluid
having a viscosity of 100 cP produces 100 times more shear stress than a 1-cP one.
Determining velocity profiles Virtually no shear differential develops in the stream
leaving the axial-flow impeller. Flow from downwardly discharging marine propellers
and hydrofoil impellers produces essentially uniform velocities from centerline to tip,
because of the design; therefore, shear does not develop in the stream. The radial-flow
turbine is diametrically different. Each of its blades produces jets. As the impeller rotates,
centrifugal force moves the fluid in front of the blades radially, to produce the discharge
flow. Fluid discharging from the front of the blades has a tangential velocity at the
impeller tip equal to the peripheral velocity of the turbine. The tangential- and radial-
velocity components produce the resultant flow from the turbine in the form of jets.
These are shown by studies of the gas-filled vortices leaving the downstream side of the
blades of flat-blade disk turbines (Fig. 2)4. The mainstream of the jet flows off the
leading edge of the blades. The vortices leaving the trailing edge, which represent a low-
pressure area, give a good indication of the direction of the main high-pressur

Scaleup and Design of Industrial Mixing Processes

Joan Schweikart
Scaleup and Design of Industrial Mixing Processes. By Gary B. Tatterson. McGraw-Hill,
Inc., 11 West 19th. St., New York, NY 10011. 1993. 312 pages. $56.00.
Reviewed by Stanley, Grossel, Process Safety & Design, Inc., Clifton, NJ
This book complements the author's previous book, ``Fluid Mixing and Gas Dispersion in
Agitated Tanks,'' and uses technical information in a practical manner to solve industrial
problems. The book's five chapters cover the following topics: basic processing concepts,
power and flow, processes and process objective, process translation, and pitfalls in
process translations and use of analogies.
Chapter 1 discusses basic mixing concepts including process objectives, different
mechanisms of fluid motion and mixing, types of mixing equipment, mixing design
geometries, commonality of mixers, properties and characterization of flow fields,
optimum designs and optimum conditions, mixing tests, and other basic considerations.
Chapter 2 presents power calculations for the purpose of judging the controlling mixing
and contacting mechanisms.
Chapter 3 is a general review of different mixing processes, covering such topics as
process objectives, different processes in mixing and contacting (typical correlations for
mechanical mixing, jet mixing, and mixing for continuous-flow and fed-batch systems),
multiphase applications, and computer modeling.
Chapter 4 discusses the need for pilot plant studies and scaling, and the basis for scaling
of a number of processes.
Chapter 5 reviews pitfalls in process translations, and dimensionless groups and
correlations with respect to their relevance and accuracy in scaling. A discussion of
analogies is also presented.
Many technical problems are presented and are worked out to completion or to a
reasonable conclusion. The information in this book can be applied to such areas as
chemical reactor design, pilot plant studies, design and scaling, speciality chemical
production, biotechnology and fermentation, pharmaceuticals production, and general
single and multiphase processing. This book is very useful to chemical engineers faced
with design and scaleup of mixing processes.

CONTROLLING THE MIX

Customization and control are driving mixer designs
Mixing, one of the most fundamental of all plant operations, is also one of the most
complex. An array of high-tech methods based on such concepts as chaotic flow, and
using computational fluid dynamic models and Doppler anemometry, is bringing new
insights into a process step once taken for granted.
As they respond to the latest discoveries in mixing science, equipment makers are also
trying to keep up with the changing -- and increasingly specialized -- needs of the
chemical process industries (CPI). ``Customers want more control over what is going on
in the mixing tank, and want different designs for different tasks,'' notes Jack Cassetta,
manager of marketing services at Lightnin (Rochester, N.Y.)
Driving today's designs are strong pharmaceutical and biotechnology industries, and
environmental markets such as wastewater treatment. New devices incorporate such
features as low-shear and custom shaped impellers, to handle specific mixing problems.
Meanwhile, fugitive emissions regulations are influencing sealed mixer designs. Here,
one new twist has been magnetic drives, explains Lightnin divisional sales manager Jeff
Caulfield. For cleanroom-level purities, his company has developed the Mag Mixer, a
sealed, magnetically driven mixer, available for tank volumes from 25 to 6,000 L in
pharmaceutical and food processing and hazardous materials handling.
For mixing suspensions with high solids content, or large volumes of low- viscosity
fluids, jet mixing is making a name for itself. Unlike most liquid-liquid and liquid-solid
mixing techniques, this type of mixing relies on hydraulic, rather than mechanical, force.
Instead of shearing fluid and propelling it around the
Mixing in the Process Industries
Joan Schweikart
Mixing in the Process Industries. Second edition. Edited by N. Harnby, M.F. Edwards
and A.W. Nienow, Butterworth-Heinemann, 80 Montvale Ave., Stoneham, MA 02180.
1992. 414 pages. $59.95.
Reviewed by Stanley S. Grossel, Senior Engineering Fellow Hoffman-La Roche, Inc.,
Nutley, NJ
The second edition of this book is a rewritten, expanded, and updated revision of the
volume first published in 1985. It is essentially the lecture notes from a post-experience
course produced by the University of Bradford and the Institution of Chemical Engineers
in the U.K. Experts from both academia and industry have contributed to this book,
providing both a theoretical and practical perspective.
There are seventeen chapters, written by thirteen authors. Some of the topics include:
characterization of powder mixtures; mixing in fluidized beds; the dispersion of fine
particles in liquid media; mixing liquids in stirred tanks; jet mixing; mixing in single-
phase chemical reactors; laminar flow and distributive mixing; the suspension of solid
particles; and the mixer as a reactor.
The book covers a wide variety of topics in detail, but there are some shortcomings, in
my opinion. There are several topics which could have been covered in more detail, e.g.,
a more fuller discussion of newer types of impellers such as hydrofoils, equations for
blending time for liquids, to name just two. However, the book does present a fairly
comprehensive treatment of liquid, solid, and multiphase mixing processes and
equipment.

MIXING
Mixing is one of the most common and most complex operations performed in chemical
process plants. Accounting partly for both its universality and complexity is that it
encompasses substances in all the physical states and in all combinations of them.
Also contributing to the complexity is that the design of mixing equipment depends on an
unusually large number of variables, including impeller type, size and speed, tank size
and shape, and the physical characteristics of the ingredients, each of which can range
enormously (viscosity, for instance, from less than 1 to well over 1 million centipoise).
Further complicating mixing operations and equipment design are the many, varied
objectives, which include blend ing (from low-viscosity to nonNewtonian fluids), solids
suspension, liquid-liquid emulsification, gas dispersion in a liquid, and mass and heat
transfer between liquid and solids, liquid and liquid, gas and liquid, and even gas and
solids.
The complexity of, and the wide-ranging applications for, mixing have long challenged
researchers attempting to understand the phenomena and engineers designing equipment
for this process operation. By and large, the engineers have answered the challenge with a
large, immensely diverse array of equipment that serves as a testament to their ingenuity.
This creativity can be seen in the following examples. For more information on each,
circle the Reader Service number that follows.

Different types of agitation combine for fast mixing

The Cyclomix 50L is a high-shear mixer for the processing of powders. Featuring fast,
intensive mixing and ergonomic design, the unit has a capacity of 50 L. Typical mixing
times range from 30 s to 5 min. A high-speed, rotating shaft has paddle-shaped mixing
elements that rotate close to the mixer vessel's inner wall, causing intense agitation.
Booth 4 F-14-G13. -- Hosokawa Micron B.V., Doetinchem, Netherlands

Scaleup and Design of Industrial Mixing Processes

Joan Schweikart
Scaleup and Design of Industrial Mixing Processes. By Gary B. Tatterson. McGraw-Hill,
Inc., 11 West 19th St., New York, NY 10011. 1993. 312 pages. $56.00.
AGITATION INTENSITY: KEY TO SCALING UP FLOW-SENSITIVE
LIQUID SYSTEMS

Only two variables are involved in scaling up flow-sensitive operations. For geometrically similar
systems, the relationship of impeller diameter and speed is well known. This article shows how to
scale up (or down) when the geometry is not similar.

Richard L. Bowen, Jr., Tensco

Of all the unit operations, liquid agitation undoubtedly occurs in more different processes than any other.
The most common
application of agitators in the chemical process industries (CPI) is for systems requiring equal fluid
velocities when the controlling
variable is viscosity -- such as the blending of miscible fluids. Certain storage tanks require minimum fluid
velocities to keep the
solids from settling. And there are some solids suspensions, as of finely divided solids, that behave like
single-phase fluids. Fluid
motion may be used to increase the effective rate of heat transfer in an agitated vessel. In general, heat
transfer occurs as part of
processing operations -- e.g., chemical reactions in miscible fluids. In many cases, the fluid motion required
by the process is
sufficient to provide adequate heat transfer. However, there are cases where fluid velocities greater than
those necessary for the
process may be required to improve convective-heat-transfer coefficients. The agitation of flow-sensitive
systems is most
efficiently and economically handled by axial-flow impellers, either propellers or pitched-blade turbines.
As there is a limit to the
useful size of cast metal three-blade propellers (17 in. dia.), the use of these on portable agitators in top-
entering vessels is
limited to smaller applications (up to several thousand gallons and 3 hp).* The axial-flow turbine having
four blades set at 45
deg is universally used in the CPI for flow-sensitive applications not handled by propellers. Circulation
capacity of impellers
Although agitation is still largely an empirical science, some very simple mathematics will enable us to
understand and visualize
the mechanics of the agitation of flow-sensitive systems. Mechanically operated agitators are caseless
pumps used to place the
entire fluid contents of vessels in motion. The gross effective pumping, or circulation, capacity of an
impeller operating in a vessel
may be given by:

(See March 18, 1985 issue of Chemical Engineering for equation)

The pumping number (NQ) is dependent upon the type and geometric design of the impeller, the D/T ratio,
and the Reynolds
number. Values of NQ for a 45-deg, 4-blade, pitched-blade, axial-flow turbine having a projected blade
height of W = 0.14D
have been published (Fig. 1) 1. Intensity of agitation A certain minimum dynamic response is required to
satisfactorily solve an
agitation problem. This response for flow-sensitive systems, specifically, is represented by the magnitude
of the fluid velocities in
the agitator. Over twenty years ago, agitation intensity had been broken into three categories: low, medium,
and very high.
Although this breakdown continues to be used, there was originally no accepted quantification of the
various categories so far as
any variable was concerned -- whether fluid velocity, power per unit volume, impeller tip speed, or tank
turnover rate. Holland's
recommendation that agitation intensity be related to impeller tip speed -- 500 ft/min for low agitation and
1,100 ft/min for very
high agitation 2 -- was not widely adopted. Weber suggested that agitation intensity be characterized as
``mild,'' `` medium'' and
``violent'' 3. These terms are still used today. No quantification of intensity was offered, other than by a
nomograph that
indicated that four times the power per unit volume was required to increase agitation from mild to violent.
Tank turnover rate
has been used as a criterion for agitation intensity: the faster the contents of a tank are turned over, the more
violent the
agitation. If one assumes agitation intensity to be proportional to the rate at which a specific tank's contents
are turned, intensity
would vary linearly with the fluid velocities in the tank. Such a criterion, recently proposed, suggested that
the range from mild to
violent intensity be defined by average bulk fluid velocities (vb) ranging from 6 to 60 ft/min 4. A 1-to-10
scale was established
to cover this range. The scale values may be defined as agitation intensity numbers (NI):

(See March 18, 1985 issue of Chemical Engineering for equation)

Now, for the first time, agitation intensity was quantitatively defined as varying linearly with tank bulk-
fluid velocities. This
definition fixed its relationship to tank turnover rate. Average bulk fluid velocities The flow produced by an
impeller as in Eq. (1)
may be divided by the cross-sectional area of the agitator vessel to obtain what may be defined as the
average bulk fluid
velocity 5:

(See March 18, 1985 issue of Chemical Engineering for equation)

Obviously, flow in a batch agitator vessel cannot be uni- directional over the vessel's entire cross-section,
as this definition of
bulk velocity requires. Flow is normally downward in a circular area around the impeller shaft, and upward
in the annular area
around the central downflow. Localized velocities vary greatly in an agitated vessel, reaching a maximum
just after leaving the
impeller, and approaching zero in the center of the torus present with all types of impellers. The bulk
velocity obtained by Eq.
(3) tends to be conservative, possibly half that determined by dividing the maximum velocity by two.
Nevertheless, this is a
useful means of correlating results between different-sized vessels, because localized velocities are
proportional to impeller tip
speed and the geometric location in geometrically similar agitators. Similarity concepts require fixed ratios
of quantities. With
geometric similarity, all dimensions must have the same ratio as the two tank diameters. Kinematic
similarity in agitation is
normally defined in terms of ratios of correspondingly localized velocities. It occurs with geometric
similarity in turbulent
agitation. However, the definition of the bulk velocity is concerned with the actual magnitude of the
velocity. In order to
duplicate a velocity magnitude within a kinematically similar system, one known velocity, such as the tip
speed of the impeller,
must be held constant. Eq. (1) may be substituted into Eq. (3) to get:

(See March 18, 1985 issue of Chemical Engineering for equation)

For a constant value of D/T, Eq. (4) reduces to vb = kND. A comparison with Eq. (5) shows that vb is
proportional to impeller
tip speed, as would be expected since kinematic similarity exists. Process results at varying agitation
intensities Eq. (3) may be
substituted into Eq. (2) to get:

(See March 18, 1985 issue of Chemical Engineering for equation)

Process results obtained by varying agitation intensity as defined by Eq. (6) may be generally correlated
against NI. The results
of several such basic correlations are shown in Table I 1. Two of these variables (specific gravity and
viscosity) may be
represented graphically (Fig. 2). This shows how progressively difficult process results must be met by
increased agitation
intensity. The values shown in Fig. 2 are not intended to serve as design points. Rather, they should be
considered the limiting
cases, with the required agitation intensity being somewhat greater than that shown. After the process
results that a certain
agitation intensity will produce are known, a value of NI may be selected to obtain the results desired in a
new application. Tank
turnover rate and agitation intensity By the Eq. (3) and (6) definitions, the flow required in a tank to
produce a certain agitation
intensity is the product of the bulk fluid velocity (or 6 NI) and the horizontal cross-sectional area of the
tank. This flow can be
calculated in gpm by means of Eq. (7):

(See March 18, 1985 issue of Chemical Engineering for equation)

The volume, V, of the tank in gallons equals AZ(7.48), with Z being the liquid height. Multiplying the
numerator and
denominator by T yields:
(See March 18, 1985 issue of Chemical Engineering for equation)

Tank turnover rate, in turns/min, is simply Q/V. Eq. (7) may be divided by Eq. (8) to get:

(See March 18, 1985 issue of Chemical Engineering for equation)

This means that the tank turnover rate varies directly with the scale of agitation and inversely with the tank
diameter and the Z/T
ratio (or inversely with the magnitude of Z ). Eq. (8) may be rearranged as follows:

(See March 18, 1985 issue of Chemical Engineering for equation)

Now the turnover rate varies inversely as the one-third power of volume, and the two-thirds power of the
Z/T ratio. For
``square'' batches (Z/T= 1), we can calculate the turnover rate for values of NI and V from Eq. (11); these
are plotted in Fig. 3.
Turnover rate varies as 1/(Z/T )2/3. A correction scale may be added to the right of Fig. 3 so that the
turnover rate for any Z/T
may be found. To use the correction scale, one finds the value of Q/V for a square batch (e.g., Q/V = 3.0
for a 10,000-gal tank
at NI = 6), runs this value horizontally to a Z/T of 1, moves up or down a line parallel to the scale lines to
the actual Z/T value
(e.g., Z/T = 1.5), then goes horizontally back to the Q/V axis to read the actual turnover rate (e.g., 2.3). At
any value of NI, the
rate is less for tall tanks and greater for shallow ones. Impeller tip speed and agitation intensity Eq. (4)
shows that the bulk
velocity (vb) is proportional to impeller tip speed (vp). Substituting Eq. (1) into Eq. (6), and solving for
ND, results in:

(See March 18, 1985 issue of Chemical Engineering for equation)

If NQ values of the D/T parameters at high Reynolds numbers in Fig. (1) are plotted against D/T, a line
having a slope of 1/2
results on logarithmic paper. The line from such a plot of NQ vs. D/T may be represented by Eq. (13):

(See March 18, 1985 issue of Chemical Engineering for equation)

Eq. (13) may be substituted into Eq. (12) to obtain Eq. (14):

(See March 18, 1985 issue of Chemical Engineering for equation)

Multiplying both sides of Eq. (14) by yields:

(See March 18, 1985 issue of Chemical Engineering for equation)

Eq. (15) indicates that impeller tip speed varies directly with NI but inversely with (D/T )l.5. For
geometrically similar systems,
(D/T )1.5 is constant, so vp varies only with NI. For D/T = 1/3, vp = 180 NI ; for NI = 10, tip speed is 1,800
ft/min. Thus,
lines of constant NI in Fig. 3 are also lines of constant impeller tip speed for geometrically similar systems
in the turbulent region.
However, when D/T varies, so does the ratio vp/NI. For example, for D/T = 0.20, vp = 386 NI, and the
impeller tip speed is
3,860 ft/min at NI = 10. Therefore, tip speed is not an indicator of agitation intensity with varying D/T
ratios. However, with
geometrically similar systems, one can scale to similar agitation intensities by using a constant tip speed, or
simply a constant
ND. This is, indeed, one procedure that has been recommended for finding the impeller speed in scaling to
larger geometrically
similar vessels in flow-sensitive systems: N2 = N1 (D1/D2)6 . But one cannot scale to a constant NI value
in geometrically
different systems with a constant ND. The new value of ND for a different D/T ratio is given by Eq. (14).
Holland's proposed
quantification of agitation intensities may be examined in light of these observations. The D/T ratio was 1/3
and the impellers
were 6-blade radial-disk turbines, for which probably NQ = 2.8 at D/T = 1/3 7. With a radial-flow turbine,
flow is split, with
half going to the top part of the vessel and the other half to the bottom part. For this reason, the flow
produced by the impeller
as in Eq. (1) must be divided by two to obtain the flow going to each half. Substituting this equation into
Eq. (6), and multiplying
both sides by p, yields:

(See March 18, 1985 issue of Chemical Engineering for equation)

Putting NQ = 2.8 and D/T = 1/3 into Eq. (16) yields NI = 0.0105 vp. Values of NI for Holland's tip speeds
are:

(See March 18, 1985 issue of Chemical Engineering for equation)

The most violent agitation has an NI value of 11.6, well in line with the maximum of 10 used here.
However, the mildest
agitation has an NI of 5.3, which is really medium agitation. For NI = 10, the tip speed of a radial turbine
would only have to be
952 ft/min, just about half that of the axial-flow turbine at the same D/T ratio. Torque/unit volume in
scaleup It is evident that
impeller tip speed can be used for scaling up if systems are geometrically similar, but not if they are
geometrically different.
Another suggested criterion for scaleup is constant torque per unit volume 8. Torque and power are related
by N because
PfQN, and both are found in the dimensionless power number:

(See March 18, 1985 issue of Chemical Engineering for equation)

For the same fluid at high Reynolds numbers, torque is:

(See March 18, 1985 issue of Chemical Engineering for equation)

For a scaleup in which the larger tank is geometrically similar and has a diameter of T = nT, with D = nD,
then V = nV, and DN
= DN = nDN. Therefore, N = N/n, and TQ = k(N/n)(nD) = knND. If we divide this by V = nV, we get:

(See March 18, 1985 issue of Chemical Engineering for equation)

Thus, for geometrically similar systems, torque per unit volume is a constant parameter, if N1D1 = N2D2.
It also turns out that
torque per unit volume remains constant with changing D/T ratios at constant values of NI. Dividing Eq.
(18) by the tank
volume, Eq. (8), yields:

(See March 18, 1985 issue of Chemical Engineering for equation)

Eq. (14) may be substituted into Eq. (20) for (ND):

(See March 18, 1985 issue of Chemical Engineering for equation)

Eq. (21) shows that, for a particular impeller, a fixed Z/T ratio, and a specific fluid, TQ/V is independent of
all other variables
and is simply proportional to the square of the agitation intensity number, NI. It is remarkable that the D/T
terms all cancel out.
It should be emphasized that this is fortuitous, and happens only because the exponent on the D/T term in
Eq. (13) is 1/2. If it
were anything else, there would be a D/T term in Eq. (21). Eq. (21) indicates that one can scale to any D/T
ratio at a constant
NI and still maintain constant the torque per volume. With changing D/T ratios, impeller speed is
calculated by means of Eq.
(14). A result somewhat similar to Eq. (21) and based on Fig. 1 has been published graphically as a
parameter consisting of
P/qNV vs. Reynolds number 9. The parameter is simply TQ/V divided by q:

(See March 18, 1985 issue of Chemical Engineering for equation)

The value of the parameter in Eq. (22) varies by about a hundredfold in going from NI = 1 to 10 in the
foregoing source,
showing that it adheres to Eq. (21). The power required to turn an agitator is simply kTQ N. If TQ/V is
constant, P/V is
proportional to N for geometrically similar agitators. With geometrically similar systems, an increase in
tank size proportionately
increases impeller diameter. At equal values of NI, ND remains constant, so that N varies as 1/D.
Therefore, power varies as
1/D. For example, in a fivefold scaleup (as from a 5- to a 625-gal vessel), the diameter would be five times
as large but speed
only one-fifth as large; therefore, P/V for the larger vessel would only be one-fifth that of the smaller one.
This is far different
from the constant P/V used for scaling such applications twenty-five years ago. Tank turnover time and
agitation intensity
Another criterion for agitation intensity used by some is turnover time -- the time required to turn over the
contents of a tank
once. This is expressed as V/Q, simply the reciprocal of the turnover rate in Eq. (9):

(See March 18, 1985 issue of Chemical Engineering for equation)

Values of hT in minutes are plotted against V in Fig. 4 (Z/T = 1), with a Z/T correction scale at the right.
Turnover time has
recently been used to quantify agitation intensity: mild -- 360 min; medium -- 1/23 min; violent -- 1030 s
10. For any scale of
agitation, these values are said to be independent of tank size, and are, thus, constant. However, Fig. 4 and
Eq. (23) show that,
for any NI scale, turnover time increases with the one-third power of the tank volume. Very mild agitation
intensities In certain
agitation applications, NI values of less than 1 are encountered. Most of these are found with large storage
tanks in the
petroleum industry, and a majority of these have side-entering propellers of relatively small size. Weber
gives an example of a
420,000-gal kerosene blending tank whose contents are turned completely every 30 min by a side-entering
propeller 3. If the
batch were square (i.e., Z/T = 1), vb = 1.4 ft/min, based on the horizontal cross-sectional area, and NI =
0.23. Recently,
Bathija suggested a turnover time of 60 min for a 460,000-gal tank (Z/T = 0.5) agitated with a jet 10. This
requires that vb =0.5 ft/min, or NI = 0.08. Other examples of tanks holding up to 4.5 million gal have
equally slow turnover times 11. Lines for
NI = 1/3 are given in Figs. 35 for very mild agitation. Consideration of blend time Neither turnover rate
(Q/V ) nor turnover
time (hT ) is required in design procedures based solely on agitation intensity (that is, without consideration
of blend time),
inasmuch as the required flowrate (Q) is easily calculated directly from Eq. (7). However, Figs. 3 and 4 are
useful in that they
lucidly show the relationships between Z/T ratios, turnover rates, and turnover times for a large range of
tank sizes and the
complete span of agitation intensities. The foregoing considerations do not take into account blend time
(hB) -- the time required
to attain satisfactory blending. Tank turnover rate (Q/V ) multiplied by hB yields a dimensionless group
(hBQ/V ), which has
been used to correlate blend-time data 12. The group represents the total number of tank turnovers required
to attain blending.
It has been stated that under some (unspecified) conditions hBQ/V may be considered constant 13. The
dimensionless group
hBN, which is more commonly used in blend-time correlations, represents the number of impeller
revolutions that accomplishes
the required blending. This group is related to hBQ/V, as might be expected. The flow generated by an
impeller is given by Eq.
(1). Multiplying the numerator and denominator of Eq. (8) by D yields:

(See March 18, 1985 issue of Chemical Engineering for equation)

In the turbulent region, hBN is usually constant for geometrically similar conditions 14. This means that
hBQ/V would also be
constant under similar conditions. Therefore, Fig. 4 can be used for a rough scaleup of blend times for
similar fluid systems and
geometries in the turbulent region. For example, if we have a 100-gal tank (Z/T = 1), the turnover rate is 7
turns/min at NI = 3.
If complete blending takes 10 min, a 1,000-gal tank, also at NI = 3, would get 3.25 turns/min and be
blended in (7)(10)/3.25 =
21.5 min. Both sides of Eq. (11) may be multiplied by hB to arrive at:

(See March 18, 1985 issue of Chemical Engineering for equation)

Because the turns required to accomplish satisfactory blending remain constant:

(See March 18, 1985 issue of Chemical Engineering for equation)

This shows that blend time, hB, varies directly with V 1/3 and (Z/T )2/3, and inversely with agitation
intensity. When scaleup is
to larger tank sizes, blend time increases with the one-third power of vessel size at a constant value of NI,
as shown in the
previous example. To reduce blend time, the scale of agitation must be increased. Increasing the scale of
agitation in the
previous example from NI = 3 to NI = 6 for the larger vessel cuts blend time in half. However, this is
accomplished at a penalty
in higher initial equipment cost that can be severe. Eq. (21) shows that the torque increases as the square of
NI ; thus, if blend
time is cut in half by doubling the intensity of agitation, torque is increased four times, and power is
increased eight times. An
economic balance may be reached between lower operating and higher equipment costs. Or, a comparison
may be made
between a single large agitator and a number of smaller ones. Under the heading ``Tank turnover time and
agitation intensity,''
we noted that it has been suggested that, at any scale of agitation, turnover time remains constant with
increasing tank size. This
is equivalent to achieving a constant blend time 15. Because hBQ/V is a constant, if V/Q or hT is held
constant, Q/V is also
constant; therefore, hB must likewise remain constant. If hB in Eq. (27) is constant and Z/T = 1, then:

(See March 18, 1985 issue of Chemical Engineering for equation)

This means that the scale of agitation must be increased as the one-third power of the tank size. If the
scaleup is from a 100-gal
to a 10,000-gal vessel, the scale of agitation must be increased 4.6 times from the pilot size, and torque
increases 21 times. If
Eq. (28) is substituted into Eq. (14):

(See March 18, 1985 issue of Chemical Engineering for equation)

Substituting Eq. (24) into Eq. (29) results in:

(See March 18, 1985 issue of Chemical Engineering for equation)

At constant Z/T and D/T ratios, N is constant for increasing tank sizes. To hold the speed constant, this
requires that equipment
be excessively large in large-scale applications. For this reason, constant blend time is rarely used 16. It is
unreasonable to
expect large tanks to blend as fast as small ones; moreover, larger tanks take longer to fill and empty. Eq.
(27) shows that if the
tank size is increased 10 times the blend time increases 2.15 times; but if it is increased 100 times, blend
time is increased only
4.6 times. This does not seem unreasonable. Demanding faster blend times for larger vessels unnecessarily
increases the
equipment size and cost. Flow required in designing flow-sensitive systems The first step in designing a
fluid motion system, if
one does not have prior data, is to estimate the agitation intensity required. Then, the flowrate in the tank to
satisfy this intensity
must be calculated 17. Eq. (7) is related to tank diameter. Substituting Eq. (10) results in:

(See March 18, 1985 issue of Chemical Engineering for equation)

Tank fluid flow, Q, is plotted against V for various values of NI at Z/T = 1 in Fig. 5, with a Z/T correction
scale on the right.
From this plot, one can determine the flow required in the design of axial-flow turbines 17 or jet mixers*
10 in flow-sensitive
systems for various values of NI. Examples of flow-sensitive systems A number of examples of flow-
sensitive systems
encountered in the chemical process industries are shown in Table II, taken from a broader summary that
includes
solids-suspension and gas-dispersion processes 18. Note that a number of solids suspensions are included.
Most of these
examples involve the storage of suspensions already made up. In some instances, however, the behavior of
finely divided solids
suspended in a liquid may be essentially that of a single-phase system. Examples are the suspension of
slaked lime in water
treatment, and finely divided clay in ceramics manufacture. In some cases, Table II shows a single value of
NI, because the
process conditions are so well defined. In others, a range is specified because no single NI value is
universally accepted. A
minimum value may indicate acceptable performance, and a maximum value exceptionally efficient
operation (but higher
equipment costs). Economic factors usually determine the value to be used. Note that, at lower values of
NI, an increase of one
scale level has a much greater effect than at higher values. For example, an increase in NI from 1 to 2
boosts agitation intensity
100%, whereas an increase from 9 to 10 only hikes intensity 11%. Although a change to an adjacent scale
level will produce
distinct differences in dynamic response, it may not be possible to distinguish process differences in the
upper levels (e.g., 710).
Even a hike in intensity from 7 to 8 represents only a 14% change. For this reason, lines for NI values of 4,
6, 8 and 9 are left
off the graphs. Determining existing agitation intensity In many process plants, agitation equipment is
operated economically,
accomplishing satisfactory blending, even though the process engineer may have only a rough idea of the
intensity of agitation.
Actually, intensity in the turbulent region may be easily determined for any tank agitated with an axial-flow
turbine from only the
knowledge of impeller and tank diameters, and impeller speed. Eq. (14) may be rearranged to give:

(See March 18, 1985 issue of Chemical Engineering for equation)

Taking D/T to be 1/3, values of NI, N and D may be calculated, as shown in Fig. 6. As NI varies with
(D/T )1.5, a correction
for D/T is given to the right in Fig. 6. To use the graph, select a speed (e.g., 45 rpm), move vertically to the
impeller diameter
(e.g., 42 in.), that indicates an NI value of 2.77 at a D/T of 1/3 (for a 10.5-ft-dia. tank). For other values of
D/T (e.g., 0.5 for a
7-ft-dia. tank), move horizontally to the right-hand correction scale (e.g., D/T = 1/3), go up or down a line
parallel to the
correction lines to the actual D/T (e.g., 0.5), then horizontally to the left to the NI scale to get the actual
intensity of agitation
(e.g., 5.1). Circulation capacity of propellers All of the equations (14, 21 and 32) involving the effective
flow produced by an
impeller are based on Fig. 1, which is for a 4-blade pitched-blade turbine. Many smaller blending
applications are handled by
top-entering propellers, often ``portable'' types. Because the propeller is also an axial-flow impeller, the
same principles as
those already developed for the axial-flow turbine apply. However, as common as the propeller is, no data
similar to those in
Fig. 1 exist for it. There are some data available that will enable us to calculate the gross flow produced by
the three-blade
square-pitched marine propeller. Porcelli and Marr collected data on the circulation times of propellers
centrally mounted in a
baffled vessel (D/T = 0.22 0.43), and presumably were able to separate the direct flow produced by the
propeller from the
flow entrained by the direct flow 19. They found that the direct flow was independent of the D/T ratio, with
flow represented by
q = 0.55ND. The entrained flow was proportional to the cross-sectional area between the propeller blades
and the tank walls,
divided by the area enclosed by the propeller diameter, which took the form of (T D)/D, or (T/D) 1. The
total flow produced
by a propeller is given by:

(See March 18, 1985 issue of Chemical Engineering for equation)

Dividing both sides of Eq. (33) by ND gives NQ:

(See March 18, 1985 issue of Chemical Engineering for equation)

Nagata included Eq. (34) (with the 0.55 replaced by a general term) after a table summarizing the discharge
performance of a
variety of impellers, mostly radial-flow types 20. It is implied that, via this equation, circulation rates could
be calculated from
the discharge rates given in the table. Specifically developed from data on axial-flow propellers, Eq. (33)
cannot be applied to
radial-flow turbines. Eq. (34) plotted as NQ vs. D/T forms a gentle curve (Fig. 7); a straight line through it
may be defined by:

(See March 18, 1985 issue of Chemical Engineering for equation)

Eq. (35) deviates only a few percent from Eq. (34) between D/T values of 0.2 and 0.5. More useful than Eq.
(34), it is similar
in form to Eq. (13), which is also shown in Fig. 7. Eqs. (34) and (13) are good for NRe = 10 to 10. Fluid
mechanics for
propellers Eq. (35) may be substituted into Eq. (12) to get:

(See March 18, 1985 issue of Chemical Engineering for equation)

Multiplying both sides by p will yield the relationship between intensity of agitation and impeller tip speed:

(See March 18, 1985 issue of Chemical Engineering for equation)

Although the constant is about 38% greater, the effect of the D/T ratio is much less than in Eq. (15). For the
two examples
examined under Eq. (15) for D/T = 1/3 and 1/5, Eq. (37) yields vp = 103 NI and 147 NI compared with 180
NI and 386 NI,
respectively, for the pitched-blade turbine. Substituting Eq. (36) into Eq. (20) results in:

(See March 18, 1985 issue of Chemical Engineering for equation)

Eq. (38) differs radically from Eq. (21) in that the TQ/V ratio varies with (D/T )1.6, whereas the
relationship for pitched-blade
turbines showed that the TQ/V ratio was independent of the D/T ratio. Eq. (36) may be rearranged to give:

(See March 18, 1985 issue of Chemical Engineering for equation)

Eq. (39) may be used to calculate the agitation intensity of existing propeller agitators mounted centrally in
baffled vessels
(NRe>10, m<about 60 cP). Application of propeller agitators Portable propeller agitators are run at
relatively high speeds
(considering that pitched-blade turbines run from about 30 to 230 rpm). Some are directly driven with
1,750-rpm motors, with
propellers ranging from 3 to 7 in. dia. Others are directly driven with 1,150-rpm motors with slightly larger
propellers (some
manufacturers have discontinued this speed). The majority are gear-driven at about 400 rpm, with
propellers ranging from 7 to
17 in. dia. (up to 3 hp). Propeller agitators are sized by horsepower according to the volume and viscosity
of the batch.
Maximum tank size ranges up to 5,000 gal for low viscosities (<100 cP), and maximum viscosities extend
to 5,000 cP (batches
<100 gal). At any horsepower, the maximum-diameter propeller is used with the lowest-viscosity fluid; as
viscosity increases,
propeller diameter is decreased. Because a larger propeller may be used at the lower speed (400 rpm) to
produce the same
flow at a much lower horsepower, the majority of propeller agitators are gear-driven. The 1,750-rpm
agitators are limited to
small batches of moderate viscosities at low D/T ratios (about 1/10, or less). Higher D/T ratios (about 1/6)
are used at 400
rpm. A plot of NI vs. speed for various propeller sizes at D/T = 1/6, according to Eq. (39), is shown in Fig.
8. Intersects of the
propeller sizes used with the 400-, 1,150-, and 1,750-rpm agitators are indicated. For the 400-rpm
propellers, NI values vary
from 4 to 10, which is medium to violent agitation. For D/T = 0.10, the NI values shown in Fig. 8 should be
multiplied by about
0.7, which would give NI values for the 1,750-rpm propellers of from 6 to 13, also medium-to-violent
agitation. However, all
manufacturers state that the agitation produced in low-viscosity fluids by these propellers and speeds is
mild-to-medium. The
discrepancy results from the installation of the portable propeller agitator, which is used without baffles and
is mounted on the
edge of the tank, with the shaft tilted and turned to one side. This is done to eliminate the swirl (tangential
flow with little
top-to-bottom flow) and vortex that occur when any agitator is centrally mounted without baffles.
Eliminating the swirl and
vortex drastically suppresses flow below that predicted by Eq. (33), which is based on baffled vessels. If
the agitation intensity
for propeller agitators set off-center without baffles is assumed to be NI = 2 for mild agitation, the flow in
such cases is reduced
to about one-third of what could be obtained with baffles. Virtually every equation (except that for the flow
produced by
impellers) that has been discussed is based on the simple assumption that process results are determined by
the magnitude of the
scale of agitation intensity (NI) or the bulk velocity (vb), which are defined by the simple equations NI =
vb/6 = q/6A and vb =
q/A, with q being the effective flow produced by an impeller. Everything has been developed from this
very simple relationship.

*Propellers of up to 3 ft dia. are used on short shafts of side-entering agitators (up to 75 hp) on large
storage tanks.

*Jet mixers may be used for the simplest liquid blending operations, with viscosities not exceeding 30 cP.

J. Matley, Editor
[Illustration]
Figure 2: Agitation intensities required in blending applications for various viscosity ratios and specific-
gravity differences
[Illustration]
Figure 1: Impeller pumping numbers for pitched-blade axial-flow turbines at various Reynolds numbers
and D/T ratios
[Illustration]
 Table I: Intensity of agitation needed for blending miscible fluids, as determined from basic correlations 1
[Illustration]
Figure 3: Tank-turnover rates vs. tank volume at various intensities of agitation, with correction for (Z/T)
ratio
[Illustration]
Figure 4: Tank turnover time (hT) vs. tank volume at various intensities of agitation, with correction for
(Z/T) ratios
[Illustration]
Figure 5: Tank fluid flow vs. tank volume for various values of NI gives required design flow
[Illustration]
Figure 6: Agitation intensity vs. impeller speed for various diameters of pitched-blade turbine, with the
L/T correction at the
right
[Illustration]
Table II: These examples of flow-sensitive systems in the chemical process industries include a number of
solids suspensions
18
[Illustration]
Figure 7: NQ vs. D/T for square-pitched marine propellers and pitched-blade turbines in turbulent region
[Illustration]
Figure 8: Intensity of agitation vs. propeller speed, with D/T = 1/6, for various propeller sizes

[References]
Hicks, R. W., Morton, J. R., and Fenic, J. G., Chem. Eng., Apr. 26, 1976, pp. 103-104; Dickey, D. S.,
Chem. Eng., Jan. 9,
1984, p. 78.
Holland, F. A., Chem. Eng., Sept. 17, 1962, pp. 179-184.
Weber, A. P., Chem. Eng., Sept. 2, 1963, p. 91.
Hicks, R. W., et al., op. cit., p. 102.
Connolly, J. R., and Winter, R. L., Chem. Eng. Prog., August 1969, p. 77.; Hicks, R. W., et al., op. cit., p.
103; Corpstein,
R. R., and Dove, R. A., Chem. Eng. Prog., February 1979, p. 72
Rautzen, R. R., Corpstein, R. R., and Dickey, D. S., Chem. Eng., Oct. 25, 1976, pp. 121-123.
Uhl, V. W., and Gray, J. B., ed., ``Mixing Theory and Practice,'' Vol. 1, Academic Press, New York, 1966,
p. 206.
Connolly, J. R., and Winter, R. L., op. cit., p. 76.
Corpstein, R. R., and Dove, R. A., op. cit., p. 67.
Bathija, P. R., Chem. Eng., Dec. 13, 1982, p. 93.
Uhl, V. W., and Gray, J. B., op. cit., p. 238.
Ibid., pp. 209-210.
Ibid., p. 247.
Ibid., p. 213.
Nienow, A. W., Chem. Eng. Sci., Vol. 29, 1043, 1974.
Rautzen, R. R., et al., op. cit., p. 123.
Hicks, R. W., et al., op. cit., p. 102.
Gates, L. E., Hicks, R. W., and Dickey, D. S., Chem. Eng., Dec. 6, 1976, p. 165.
Porcelli, J. V., and Marr, G. R., Ind. Eng. Chem. Fund., Vol. 1, No. 3, 1962, pp. 172-179.
Nagata, S., ``Mixing Principles and Applications,'' Halsted Press, Wiley, New York, 1975, p. 138.

[Biography]
Richard L. Bowen Jr. is president of Tensco (P.O. Box 207, Barrington, RI 02806), a manufacturer of
mixers. He wrote the
``Drives'' section of the ``Encyclopedia of Chemical Process Equipment'' (Reinhold, 1964), and a series on
``Non-Newtonian
Fluid Flow'' (Chem. Eng., 1964). Holder of 19 U.S. and British patents, he has an A.B. in chemistry from
Princeton
University, an M.S. and Sc.D. in chemical engineering from Massachusetts Inst. of Technology, and is a
member of AIChE.,
ASME, ISA and Sigma Xi