Chapter 19

Mechanical Phase Separations

§19.0 INSTRUCTIONAL OBJECTIVES

After completing this chapter, you should be able to:
Have a broad understanding of the entities that are responsible for air pollution.
Understand the spectrum of particle sizes and identify the devices that could be used to separate them from the
fluids in which they are suspended.
Understand the force balances on which settling equations are based.
Know the sizes of the particles whose behavior is described by Newton’s law, Stokes’ law, or Brownian motion.
Use settling equations modified by hindered settling parameters and adjusted for centrifugal forces to design particle-
fluid or microorganism-fluid separators.
Describe a knock-out drum, coalescer, vane filter, cartridge filter, demister, baghouse, electrostatic precipitator,
settling pond, plate-and-frame filter, cyclone, vacuum drum filter, and settling tank.
Analyze filtration data to ascertain if the operation is at constant or variable rate, or constant pressure, and if the
filter cake is compressible.
Use mathematical models based on Darcy’s law or flow through packed beds to predict required filter areas for
drum and plate-and-frame filters.
Understand the nature of filter aids and the utility of washing and pressing cycles.
Know how centrifuges are designed and what they are used for.
Describe how extracellular and intracellular bioproducts from fermentors are recovered (prior to purification).
Understand precipitation, flocculation, and agglomeration processes, and the need for cell disruption in
bioprocesses.

This chapter does not deal with separations where one or
more chemicals are removed from a feed mixture; instead, it
describes mechanical devices used to separate one bulk phase
from another. Mundane, household examples of such devices
include air conditioning and heat-pump filters to prevent dust
and solid particles from clogging heat-exchange surfaces,
paper filters in drip coffee makers to prevent coffee grounds
from entering the brew, water filters attached to home water-
supply units in locations where water quality is suspect, and
electric precipitators used in homes where an occupant has
serious pollen and dust allergies.
The word ‘‘filter’’ was derived from the Latin ‘‘filtrum,’’
which in turn may be traced to the word ‘‘feltrum,’’ which
describes felt or compressed wool, and which in turn is fur-
ther related to the Greek word for wool or hair. An Egyp-
tian papyrus dating from the third century A.D. and known
as the ‘‘Stockholm Papyrus’’ describes the process of pro-
ducing caustic soda and the use of a filter for clarifying it,
including the application of clay as a filter aid. The first
patent for a filtering device was granted to Joseph Amy in
1789 by the French government. Thereafter, for the next 50
years, most filter patents pertained to treating water or sew-
age. The modern rotary-drum vacuum filter and pressure
leaf filter were developed in the late 1800s by mining engi-
neers, including W. J. Hart, E. Sweetland, E. L. Oliver, and
J. V. N. Dorr for use in the cyanide process for recovering
gold [1].
To be described in this chapter are means of:
Removing airborne liquids, solid particles, microorgan-
isms, and vapors from air streams when a clean, sterile-
air supply is required to prevent contamination or infec-
tion of a product, process stream, or the environment.
Separating entrained liquids from vapor streams as in a
flash distillation chamber, or partial condenser.
Designing an optimal air-purification system comprised
of multiple particle-capture devices.
Condensing vapors from air streams when downstream
conditions favor an undesirable condensation.
Eliminating pollutant particles, mists, and fogs from
gases that are vented to the atmosphere from manufactur-
ing plants.

778

Removing droplets of one liquid suspended in another as
in hydrocarbon-water decanters.
Recovering, as a cake, solid particles suspended in
liquids, by means of plate-and-frame, drum, leaf, and
other filters; and determining cake wash cycles.
Operating filters at constant pressure and variable rates,
using pump characteristic curves.
Designing and analyzing cyclones and centrifuges.
Applying mechanical separations to bioprocesses: cell
disruption, precipitation, and flocculation (preceded by
coagulation).
This is only a sample of a plethora of applications. In a sense,
the membrane processes described in Chapter 14 could also
be classified as mechanical separations and included in this
chapter. However, there is a major difference between the
two, insofar as the design methods for most membrane
devices typically involve molecular diffusion. Pressure drops
are high and mass transfer is slow. The devices described in
this chapter are bulk-flow units operating mostly at relatively
low pressure drops, and the design equations are based on
hydrodynamics involving settling velocities of macropar-
ticles rather than molecular diffusion of individual species.
Though both are ‘‘mechanical separations,’’ there are major
differences between design methods for calculating the mem-
brane area required for a diffusing molecular flux, and the
demister area needed to retain dust particles or droplets, or
filtration areas for pressure- or vacuum-driven devices for
solid–liquid filtration. In solid–liquid filtration the screen is
not the filter; the particles form a ‘‘cake’’ and this, rather
than the screen or fabric, is the filter. This situation is quite
different from that when a membrane is the filter, as in
Chapter 14.
An important aspect of filtration is that particles sus-
pended in the liquid or gas are retained, as in a ‘‘drip grind’’
coffee maker, where the grounds are retained on the filter
medium. Periodic ‘‘blowback,’’ scraping, or other particle-
removal methods are required; otherwise, both the filter and
the retained particles must be disposed of or processed. If the
concentration of particulate matter in the gas or solution is
high, large amounts of solids must be removed. For large
quantities of inexpensive industrial liquids and wastes, use of
filters is prohibitively expensive, so the liquids are placed in
retention (holding) tanks or ponds, where the particles are al-
lowed to settle, often with the help of coagulants, settling
agents, and mild, directed stirring to speed the settling. If the
particle is an industrial product, it is not processed in a reten-
tion pond or settler. What is used instead is a filter press that
can handle a high concentration of particles in the 10–50
micron range inexpensively and in large volumes. Here, pres-
sure or vacuum drive a solution through fabrics or screens,
frequently precoated with ‘‘filter aids.’’ As the retained parti-
cles accumulate on the screen, they form a ‘‘cake,’’ which
then becomes the filter.
Mechanical Phase Separations 779
This chapter has its own vocabulary. To be encountered in
the upcoming pages are settlers, decanters, coalescers, vanes,
centrifuges, demisters, knock-out drums, electrostatic precip-
itators, mesh pads, cyclones, impingement separators, bag fil-
ters, and drum, plate-and-frame, and vacuum filters. Design
methods and applications for these devices are unique in the
sense that each is designed for a specific purpose and a spe-
cific range of particle sizes. A device whose design was
based on inertial impingement would be inadequate for an
application involving particles whose hydrodynamic behav-
ior is characterized by Brownian motion. As a rule, prelimi-
nary selection of a specific mechanical separation device is
based on particle size and phase. After the device is selected,
laboratory and/or pilot-plant data are analyzed to establish
values of empirical constants in the design equations used to
size the plant unit. The design variables include particle size
and density; fluid velocity, density, and viscosity; the external
force field; and device parameters. Except for vacuum and
drum filters and centrifugal units, the design equations for
the devices are largely based on settling velocities predicted
by Newton’s gravitational law or Stokes’ law.
Industrial Example
The problem of producing enormous quantities of sterile air
for aerobic fermentation exists only in biochemical engineer-
ing. In the 1960s, cotton plugs were satisfactory for use in
test tubes or shake flasks. For pilot-plant fermentors, small
fibrous filters were deemed satisfactory. For plant applica-
tions, cotton fibers and activated carbon were standard, but
glass fibers were recognized as being a better filter medium,
and 6-foot-deep beds containing glass fibers 5–19 microns
(mm) in diameter came to be in use. The sterilization process
was aided and abetted by inefficient air compressors, which
heated the air to about 150 C (which is not close enough to
the 220 C at 5 minutes required to kill bacterial spores). The
pressure drop through the various filters, plus the spargers
and 20-or-more-feet-high fermentors, is well over 1 atm, so
today, expensive compressors or blowers, rather than fans,
are used. These must be protected from damage by solid par-
ticles and vapor droplets in the inlet air by suitable means,
which include knock-out drums and coalescent filters. A seri-
ous source of contamination is the oil mist and oil/water
emulsions emitted by compressors.
As discussed by Shuler and Kargi [2], one 50,000-liter
aerobic fermentor, during a 5-day fermentation, requires
about 2 108 liters of absolutely sterile air. Because banks
of 10 or more 100,000-liter fermentors may be housed in one
building, sterilization processes such as UV radiation, steam,
ozone, or scrubbers are not economical. Minimization of
pressure drop is critical, as is dependable protection of the
high-value product in the fermentors. Not only must the air
entering be sterile; if the fermentation involves pathogens or
recombinant DNA, the concentration of microorganisms in
the exiting gas, which is far higher than that in the inlet gas,
must also be reduced to zero. Catalytic combustion has been
780 Chapter 19 Mechanical Phase Separations
used to accomplish this, but membrane products, which are
much less expensive, have made serious inroads into the
market.
Particulate concentrations in air streams vary widely. Popu-
lations of microorganisms, which vary from 0.5 to 1 mm in
size, have been measured in all parts of the world and range
from 1 to 10 per liter, which is 103–104/m3. Because of their
small size, this translates into only about 10 8 mg/m3, which is
relatively insignificant compared to the dust, vapor, and oil
loadings. Reasonable numbers for concentrations of particles in
city air are 35 mg/m3, which is about 500 106 particles/m3,
80% of which cannot be seen by the human eye [3]. The haze
that one sees when flying over any large city in the world is
not usually visible from the street. Additional problems arise
from the concentration of water vapor in the air, which is
highly variable, as are the fumes emitted by automobiles.
Air compressors are housed in a building, but provisions
must be made for ambient fog, rain, snow, and possible air-
borne construction-site debris. The sterile supply of air is
critical, and although air-purification units upstream of the
compressors can be serviced, no downstream maintenance is
feasible because of the possibility of contamination. Air
cleanup after compression is required because of the oil mist
and water/oil emulsions emitted by compressors, which can
be as much a source of contamination as the entering air
stream. Air-filtering devices downstream of the compressors
should be cleanable by blowback, and there must be a certain
amount of redundancy. An ever-present danger is wetting of
the filters, which increases pressure drop and provides paths
for pollutant short circuits.
Historically, depth filters with glass–wool fibers akin to
building insulation were used, but these have been almost
totally replaced with membrane or cartridge surface filters of
the type described in Chapter 14 and later in this chapter.
Possible mechanisms for capture of bacteria, about 1 mm in
size, are direct interception, electrostatic effects, and inertial
effects. In depth filters, as the gas flows in streamlines around
the fibers, particles with sufficient mass will, because of an
inertial effect, maintain a straight-line trajectory and be cap-
tured by the fiber. Brownian motion is unimportant for bacte-
ria, but for viruses, which are smaller than bacteria, it is
important. In the case of surface filters used for sterilizing
air, sieving effects play a major role.
Figure 19.1 shows a cyclone and compressor air-intake fil-
ters. The cyclone removes particles above 10 mm with high
efficiency and will prolong the life of the suction-line filter,
which may be a simple panel or cartridge filter but more
likely is a more modern coalescing filter, possibly preceded
by a vane impingement device. Most viruses, bacteria,
vapors, odors, and submicron particles pass these filters.
Water vapor will generally pass, and will not condense in the
compressor because of the adiabatic temperature rise, but it
will condense when the compressed gas is cooled.
Downstream of the compressor is an aftercooler, a prefilter
to drain condensate, an adsorption dryer, and a high-efficiency
HEPA filter to reduce microorganisms to a level of 100 parti-
cles/m3, which is standard for a sterile work area. The prefilter
will generally be a two-stage, self-draining, coalescing glass-
fiber device specifically designed to reduce oil carry over to
0.001 mg/m3. The dryer can be activated alumina or zeolite,
and if oil vapor is a potential problem, it may be backed up
with an activated carbon adsorber to remove hydrocarbon
gases. All of these filtration units are subject to governmental
and industry standards and performance tests.
§19.1 SEPARATION-DEVICE SELECTION
Separation-device selection is based largely on the size of the
particles carried by the fluid. Other considerations such as
density, viscosity, particle concentration, and flow rate also
enter into the selection process, as do particle and fluid dollar
value and the device particle-capture efficacy and cost, but
they are secondary. Furthermore, as will be seen shortly,
mathematical models for many filtration devices are based
on particle settling velocities, and these are based on hydro-
dynamic equations in which particle size is a key variable.
The lists in Tables 19.1 and 19.2 were compiled from manu-
facturers’ product bulletins and various other sources, many
of which differ considerably, and must be viewed as a prelim-
inary guide to the selection process.
Helpful guidelines to the interpretation of the particle-size
data in Table 19.1 are the entries regarding the limit of visi-
bility and the size of a human hair. It is often surprising to see
the amount of condensate dripping out of a coalescer when
the air entering the coalescer seemed clear of mist. The range
of particle sizes given in Table 19.1 is indicative not only of
the fact that the sizes of the particles in atmospheric fog vary
from day to day and location to location, but also that within
a given fog, the particles have different sizes. This is impor-
tant because the particle-capture efficiency for the devices
listed is a function of particle size. American Filter Co. Inc.,
for example, distributes product literature that states that
their high-efficiency cyclone has a 50% efficiency for
Figure 19.1 Air-
purification system.
 §19.2 Industrial Particle-Separator Devices 781

Table 19.1 Typical Particle Sizes Table 19.2 Particle-Size Ranges for Particle-
Capture Devices
Particle Size, mm
Particle-Capture Device Size Range, mm
Large molecules 0.001–0.004
Smoke 0.005–1 Membranes 0.00001–0.0001
Fume 0.01–0.1 Ultracentrifuges 0.001–1
Tobacco smoke 0.01–0.12 Electrical precipitators 0.002–20
Smog 0.01–1 Centrifuge 0.05–5
Virus 0.03–0.1 Cloth collectors 0.05–500
Mist 0.1–10 Fiber panels and candles 0.10–10,000
Fog 0.1–30 Elutriation 1–100
Spores 0.5–1.80 Air filters 2–50
Bacteria 0.5–10 Centrifugal separators 2–1,000
Prokaryotic cells 1–10 Impingement separators 5–2,000
Dust 1–100 Vane arrays 5–10,000
Limit of visibility 10–40 Cyclones (high efficiency) 6–35
Filter presses 10–50
Liquid slurries 10–50 Cyclones (low efficiency) 15–250
Eukaryotic cells 10–100 Cloth and fibers 20–1000
Drizzle 10–400 Gravity sedimentation 45–10,000
Spray 10–1000 Screens and strainers 50–1,000
Pollen 20–80 Sieving screens 50–20,000
Mist 50–100
Human hair 50–200
electric-field gradient; (6) agglomeration by particle–particle
Rain 100–1,400 collisions; and (7) sieving, where the flow pathway is smaller
Heavy industrial dust 100–5,000 than the particle. Mechanisms 2–4 are depicted in Figure
19.2. Note that in interception the particle follows the stream-
line, while in impaction it follows a direct path. Generally,
capturing a 6-mm particle, and a 99% efficiency for capturing devices that operate by a combination of mechanisms 2 and 3
a 25-mm particle. combine impaction and interception in one empirical design
The nomenclature in this field is far from standardized. equation. In many devices, synergistic mechanisms are used.
The term aerosol, for example, is used to describe suspended In cyclones, for example, gravity settling is abetted by centrif-
liquid or solid particles that are slow to settle, be they sub- ugal force. A generic consideration in collection devices is the
micron or 50 mm in size. Mists are generally described as problem of re-entrainment. The inertial forces that deposit a
particles upward of 0.1 mm in size that arise because of vapor particle on a fiber can also blow the particle off the fiber. Cy-
condensation. Sprays are the result of intentional or clones, for example, are more efficient for liquid droplets than
unintentional atomization processes. for solid particles because droplets are more likely to coalesce
In developing a flowsheet for a particle-collection system, and agglomerate at the bottom than are solid particles.
it is well to remember the strongest of the process design
heuristics: ‘‘Cheapest first.’’ In terms of the devices listed in §19.2.1 Gravity Settlers
Table 19.2, this means removing large particles by
inexpensive settling chambers, vane arrays, or impingement If the velocity of the carrier fluid is sufficiently low, all parti-
devices, and then removing the small amount of remaining cles whose density is above that of the carrier will eventually
particles with the higher-capital-cost units like membranes, settle. Terminal velocities of droplets and solid particles are
centrifuges, or electric precipitators. such that the required size of the settling chamber usually

§19.2 INDUSTRIAL PARTICLE-SEPARATOR

DEVICES
The operative mechanisms for the particle separators to be Interception
described are: (1) gravity settling, where the force field is ele- Filter
vation; (2) inertial (including centrifugal) impaction, where Diffusion fiber
the force field is a velocity gradient; (3) flow-line (direct)
interception or impingement, where the particle is assumed to Impaction
have size, but no mass, and follows a streamline; (4) diffu-
sional (Brownian) deposition, where the force field is a con-
centration gradient; (5) electrostatic attraction due to an Figure 19.2 Particle-collection mechanisms.
782 Chapter 19 Mechanical Phase Separations
Gas in
Larger particles
Dust-collecting hoppers
Figure 19.3 Horizontal settling chamber.
becomes excessive for droplets smaller than 50 mm and for
solid dusts smaller than 40 mm. For solid particles, air veloc-
ities greater than 10 ft/s lead to re-entrainment of all but the
heaviest particles.
In the horizontal settling chamber of Figure 19.3, the gas ve-
locity, upon entering the chamber, is greatly reduced. The key
design variable, the particle-residence time, computed as the
length of the chamber divided by the gas velocity, determines
whether or not the chamber is long enough to allow the particle
to fall to the bottom. The width of the chamber must be such
that the gas velocity is below the ‘‘pick-up’’ velocity that will
cause re-entrainment. For low-density materials such as starch,
this is 5.8 ft/s. For gas–solid systems, settling chambers have
advantages of minimal cost and maintenance, rapid and simple
construction, low pressure drop, and dry disposal of solids.
A crude classification of solids takes place in the sense
that the first of the dust-collecting hoppers contains larger
particles than the ones that follow, but little use is made of
that because particle sizes overlap. Many variations of the
simple enclosure in Figure 19.3 exist. The height a particle
has to fall can be decreased by banks of trays set within the
chamber, as in the Howard multitray settling chamber. Baf-
fles can be used to direct the gas flow downward to add a
momentum effect to the gravitational force. Baffles and tortu-
ous paths also aid particle capture by inertial mechanisms,
but the cost in terms of pressure drop is high.
For solid–liquid systems, devices based on gravity are called
sedimenting separators, clarifiers, thickeners, flocculators, and
coagulators. Coagulation is the precipitation of colloids, by floc
formation, caused by addition of simple electrolytic salts, which
modify electrostatic forces between the particles and fluid. The
Drive
Skimmer
Influent Well
Influent
Effluent Collector Arm
Sludge Concentrator
term flocculation is generally used to describe the action of
Gas out water-soluble, organic, polymeric molecules that may or may
not carry a charge, such as polyacrylamide, which promotes set-
tling. Figure 19.4 depicts a liquid-settling device of the type
widely used for wastewater treatment, which is equipped with a
Smaller slowly moving rake that revolves at about 2 rph and moves the
particles
sludge downward to promote particle agglomeration. The vol-
ume of clear liquid produced depends primarily on the cross-
sectional area and is almost independent of the tank depth.
Liquid–liquid gravity separators are important in the oil
industry, where mixtures of water and oil are commonplace,
and in the chemical industry, where extractive distillations
and liquid–liquid extractions are carried out extensively. In
liquid–liquid separators, called decanters, there is often a con-
tinuous phase with a discontinuous phase of dispersed drop-
lets. The two phases must be held for a sufficient time for the
droplets to settle if heavy, or rise if light, so that the two
phases disengage cleanly. A completely clean disengagement
is a rarity because, unless the liquids are unusually pure, dirt
and impurities concentrate at the interface to form a scum, or
worse yet, an emulsion that must be drained off. Figure 19.5
shows a continuous-flow gravity decanter designed to separate
an oil layer from a water layer that contains oil droplets. It
does not show the perforated underflow and interface baffles,
outlet nozzles, or inlet flow distributors.
The unit does not run full, and the design involves balanc-
ing the liquid heights due to the density difference of the
phases, and determining the settling velocities of droplets
moving up or down from the dispersed to the continuous
phase. Needless to say, rules of thumb and years of experi-
ence are required to design units that work well. Some design
methods are based on the time it takes particles to move
through a semihypothetical interface between the heavy and
light fluids. Example 19.6 shows how the dimensions for a
continuous-flow decanter are obtained. Methods for design-
ing a vertical decanter are given in Exercise 19.8.
§19.2.2 Impaction and Interception Separators
Inertial impaction and interception mechanisms, shown
above in Figure 19.2, consist of a particle colliding with a
Scum Trough
Scum
Draw-Off
Figure 19.4 Liquid sedimentation and
Sludge Draw-Off flocculation.
 §19.2 Industrial Particle-Separator Devices 783

Feed

Light phase overflow

Top of light phase

Drain interface Light phase

for emulsion Heavy phase out

Heavy phase

Heavy phase out Light phase out

Interface
Figure 19.5 Gravity-flow decanter.

target that can be anything from a screen, a bed of fibers,
staggered channels, or louvers. Inertial forces accelerate
large particles less than small particles, and this, coupled
with re-entrainment and variable drag coefficients due to
shape, make theoretical prediction of capture efficiency and
velocity distributions within a cloth or mesh filter virtually
impossible. Instead, impingement separators are designed on
the basis of system-specific constants provided by device
manufacturers and used in conjunction with the Souders–
Brown equation, (6-40), developed in §6.6.1 to describe
droplet behavior in distillation columns [3]. Also provided
by the manufacturer are recommendations on allowable gas
or liquid velocities and pressure drops. For particle-capture
devices, performance parameters cannot be calculated from
physical properties; if the velocity is lower than what is rec-
ommended, impingement of small particles may not take
place, and if it is too high, re-entrainment will occur. In addi-
tion, use is frequently made of generalized or device-specific
information regarding collection efficiency as a function of
Reynolds number or particle size. When impingement
devices are used to capture liquid droplets, they coalesce and
the liquid must be drained from the collector device. Often,
modern coalescence devices combine vane and channel
impingements with waffled filters.
An endless array of governmental and industry standards
and regulations apply to products manufactured for the pur-
pose of removing particles and contaminants from air
streams. Not only do public health laws, with respect to the
quality of the air emitted, exist, but there are also industry
standards for how devices that impact the environment are to
be tested. Based on these tests, products are graded and cate-
gorized. This is typical for industrial products intended for a
specific use such as filtering air for hospital operating rooms
or removing oil mists generated by air compressors. The
Eurovent standards for flat-panel ventilation filters shown in
Table 19.3 were set by the quasi-governmental agency the
European Committee of Air Handling and Refrigeration
Equipment Manufacturers, and apply to both glass-fiber me-
dia and synthetic organic fibers. Parallel specifications have
been set by American manufacturers and trade organizations

Table 19.3 Cen/Eurovent Filter Classification

Type Class Eurovent Designation Efficiency, % Measured by

Coarse dust filter EU1 <65 Synthetic dust

EU2 65–80
EU3 80–90
EU4 >90

Fine dust filter EU5 40–60 Atmospheric

EU6 60–80 Dust spot
EU7 80–90 Efficiency
EU8 90–95
EU9 >95

High-efficiency EU10 85 Sodium chloride

particulate air EU11 95 or liquid
filter (HEPA) EU12 99.5 aerosol
EU13 99.95
EU14 99.995

Ultra low EU15 99.9995 Liquid aerosol

penetration air EU16 99.99995
filter (ULPA) EU17 99.999995
784 Chapter 19 Mechanical Phase Separations

Table 19.4 Coalescing-Filter Media Grades filter element, held between stainless steel support sleeves, is
shown in Figure 19.6. Because of the coalescing filter
Pressure
Color medium, the condensate is drained, and the elements are self-
(bar)
Grade Efficiency Coalescing Maximum regenerative as far as removal of liquid is concerned. How-
Code (%) Carryover Oil Dry Wet ever, it is advisable that prefilters capable of removing parti-
cles down to 5 mm or less be placed in the line ahead of
2 green 99.999+ 0.001 mg/m3 0.1 0.34
the coalescing filter or it will quickly be plugged. In this table
4 yellow 99.995 0.004 mg/m3 0.085 0.24
the coalescing efficiency was measured using 0.30–0.6 mm
6 white 99.97 0.01 mg/m 3 0.068 0.17
particles based on 50 ppm maximum inlet concentration.
8 blue 98.5 0.25 mg/m 3 0.034 0.19
A well-designed filtration system, as shown in Figure
10 orange 95 1.0 mg/m3 0.034 0.05
19.7, will have elements such as an inexpensive coarse-
particle pre-filter collector like a screen filter or cyclone, fol-
such as the American Petroleum Institute (API). Not shown lowed by an extended-surface filter that is effective down to
in this table are specifications regarding particle size, but the micron level, and then a submicron filter where the veloc-
they do exist [4]. ity is lower and the particle capture is principally by
Table 19.4 shows the internationally accepted grading sys- Brownian motion and/or sieving.
tem for coalescing filter media used to capture liquid oil, oil/
water emulsions, and oil aerosols emitted by oil-lubricated
§19.2.3 Fabric Collectors
compressors. These are glass microfibers in the 0.5–0.75 mm
range, which will trap up to 99.99999% of oil/water aerosols A very common industrial filtration device is a fabric dust col-
and dirt particles in compressed air, down to a size of 0.01 lector. In industry, multiple collectors are housed in enclosures
mm. The mechanical sandwich construction of the two-stage called baghouses. These are relatively inexpensive installations

Figure 19.6 Brink fiber-bed mist

collector. (Courtesy of MECS, Inc.)

HEPA filter
media velocity 0.02 m/s

Prefilter Extended surface filter

media velocity media velocity 0.11 m/s
2.5 m/s

1.3 m/s 1.3 m/s

Face 2.5 m/s
velocity
2.5 m/s

(Viscous (Interception
impingement) + diffusion)

Figure 19.7 Multistage filter

(Diffusion) system.

D 0.5 DC
F
E
A Dust-Laden Air Inlet
B Dust Hopper
G C Filter Bag (TYP)
D Clean Air Plenum
E Clean Air Outlet
F Compressed Air Source
C DC
G Bag Support Cage
A B
Figure 19.8 Tubular bag filter with pulse jet cleaning.
capable of capturing particles down to 0.05 mm. As shown in
Figure 19.8, particles are collected on the outside of a fabric-
encased, porous, cylindrical candle. The device has a vibratory
or compressed-air blowback system to remove the particles
trapped on the outside of the filter element. Liquids as well as
solids are processed in units of this type. For both liquids and
gases, as the particles on the cloth build up, they form a cake
that acts as a filter, and often is a more effective filter than the
fabric or screen. This makes screen and fabric collectors sys-
tem specific; there is no way to predict performance other than
to take laboratory data because the filtering action of the cake
cannot be predicted analytically.
§19.2.4 Vanes and Louvers
Another device that falls in the aerodynamic-impingement
category is the vane or louvered particle collector. Here,
the carrier fluid is forced through a maze, changing
direction frequently. This type of device is most effective
for collecting droplets or mists and fogs that coalesce
and can then be drained from the system. Most often, if
pressure drop allows, vane units are used as prefilters for
mesh filters, particularly for very small droplets that coa-
lesce upon impingement.
§19.2.5 Cyclones and Centrifuges
For a centrifuge or cyclone, centrifugal acceleration is substi-
tuted for gravitational acceleration in the appropriate fluid-
dynamics equations. The complicating factors are that centrifu-
gal force depends on the distance from the axis of rotation,
which depends on the complex geometry and flow patterns in
the device, and that the concentration of particles may be so
high that hindered (by neighboring particles)-settling equations
are necessary. A typical design method, applied to a Podbiel-
niak centrifugal extractor, was demonstrated in Example 8.11.
This design strategy consists of finding the optimal conditions
for the centrifuge from test runs using a small laboratory unit,
and then using a set of scientifically deduced, semi-empirical
rules for scale-up to a large industrial unit. This methodology,
as will be seen, is also used to design cyclones.
Because cyclones are inexpensive and durable, with a
decent collection efficiency for particles larger than about
§19.2 Industrial Particle-Separator Devices 785
0.5 DC × 0.2 DC
0.5 DC
1.5 DC
2.5 DC
Collecting
hopper
diameter DC
0.375 DC
Figure 19.9 Standard high-efficiency cyclone dimensions.
5 mm, they are the most widely used device for industrial dust
collection. If the efficiency is not high enough, multiple units
can be placed in series. The dust-laden stream enters the top
section of the cylindrical device tangentially, which imparts a
spinning motion. Centrifugal force sends the particles to the
wall, where they agglomerate and fall to the bottom. The spin-
ning gas also travels toward the wall, but it reverses direction
and leaves the device from a sleeve at the top, whose bottom
extends to below the inlet, as shown in Figure 19.9, which
includes standard-dimension relations. The path is usually
axial, there being an inner up-flow vortex inside the downward
vortex. In liquid cyclones (hydroclones), the upward flow is
separated from the downward flow by an outer jacket wherein
the liquid flows up. Separation depends on settling velocities,
particle properties, and geometry of the device. By directing
the inlet flow tangent to the top of the cyclone, centrifugal
force can be utilized to greatly enhance particle collection.
Well-designed cyclones can separate liquid droplets as small as
10 mm from an air stream. Small cyclones are more efficient
than large ones and can generate forces 2,500 times that of
gravity. For solids, re-entrainment problems can be reduced by
water sprays and vortex baffles at the outlet.
§19.2.6 Electrostatic Precipitators
Electrostatic precipitators are best suited for the collection of
fine mists and submicron particles. The first practical appli-
cation was fashioned by Cottrell in 1907 for abating sulfuric-
acid mists. A particle suspended in an ionized gas stream
within an electrostatic field will become charged and migrate
to a collecting surface. Care must be taken that the particles
do not re-entrain, but are removed from the device. Two
types of devices are available, one in which ionization and
786 Chapter 19 Mechanical Phase Separations

Discharge system High voltage cable

support insulator

Precipitator plate cover

D. C. output

Collecting
(positive)
plates Clean gas
outlet

Figure 19.11 Collection efficiency for cyclone.

Discharge
(negative)
electrons Solution

Feed Performance of Parallel Cyclones

Direction of
of gas flow Particle % of Total % % of % of % of
Size, Particles Collection Particles Particles Not Total Particles
mm in Feed Efficiency Collected Collected Not Collected
Transformer
rectifier set 104–150 3 100.0 3.0 — —

Collecting A.C. input 75–104 7 99.1 6.9 0.1 0.6

(positive) plates 60–75 10 98.5 9.9 0.1 0.6
40–60 15 97.3 14.6 0.4 2.5
30–40 10 96.0 9.6 0.4 2.5
Figure 19.10 Electrostatic precipitator.
20–30 10 94.3 9.4 0.6 3.8
15–20 7 92.0 6.4 0.6 3.8
collection are combined, and one in which they are separated.
10–15 8 89.3 7.1 0.9 5.7
In Figure 19.10, the chambers are combined. To obtain ion-
7.5–10 4 84.2 3.4 0.6 3.8
ization, the voltage must be high enough to initiate a corona 5–7.5 6 76.7 4.6 1.4 8.8
discharge, but not so high as to cause sparking.
2.5–5 8 64.5 5.2 2.8 17.6
Recent innovations to electrostatic precipitators include
0–2.5 12 33.5 4.0 8.0 50.3
adding water sprays and two-stage ionizing wet scrubbers. 100% 84.1% 15.9% 100%
Of course, any device that adds water to a dry powder com-
plicates the waste-disposal problem and increases power con- The numbers in Column 3 (C3) were obtained from Figure
sumption. Waste disposal of particle-laden water streams is a 19.11; numbers in C4 are computed from (C2 C3=100); numbers
general problem with such devices as spray chambers, wet in C5 are from (C2 – C4); and numbers in C6 are from (C5 100=
scrubbers, packed absorption columns, and plate scrubbers, sum of C5). Total collection efficiency for the cyclones is 84.1%,
which are not elaborated on in this brief introduction to but particle capture for particle sizes below 5 mm is low, so the
mechanical separations.
Example 19.1, adapted from Nonhebel [6], illustrates the
role of electrostatic precipitators in an industrial environ-
ment, and the importance of particle-size distribution in
assessing the effectiveness of a pollution-control system.

EXAMPLE 19.1 Cyclones in Series with an

Electrostatic Precipitator.
The first two columns in the table below give the dust content of a
feed to a bank of precleaning, parallel cyclones in series with an
electrostatic precipitator having a contact time of 3 s. Given the effi-
ciency versus particle-size performance data for each of the two
devices in Figures 19.11 and 19.12, obtain emission-collection effi-
ciencies for each of the devices and for the overall system if the gas
flow rate is 240,000 m3/hr, gas density is 10 g/m3, and particle load-
ing is 10/m3, all at the same standard conditions. Figure 19.12 Collection efficiency for electrostatic precipitator.
 §19.2 Industrial Particle-Separator Devices 787

effluent from the cyclones is sent to an electrostatic precipitator. The Wash

feed to the precipitator has a particle loading of (10)(100 84.10)= distributors
Dewatering
100 ¼ 1.59 g/m3.
The performance of the electrostatic precipitator, set forth in the
table below, is obtained in exactly the same manner as for the

Ro
cyclones, using Figure 19.12 for the collection efficiency of the pre-

ta
cipitator. From the results, it operates at an efficiency of 85.4%, with

tion
an exit dust loading of 1.59(100 85.4)=100 ¼ 0.23 g/m3. Final
Cake washing dewatering
(max allowable)
Initial
Feed from Cyclones Performance of Electrostatic Precipitator dewatering

Particle % of Total % % of % of % of
Size, Particles Collection Particles Particles Total Particles Discharge
mm in Feed Efficiency Collected not Collected not Collected Slurry level

75–104 0.6 99.2 0.6 — trace

60–75 0.6 98.7 0.6 — trace Filtering Discharged
filter cake
40–60 2.5 97.7 2.4 0.1 0.7
30–40 2.5 96.8 2.4 0.1 0.7
Figure 19.13 Rotary-drum vacuum filter.
20–30 3.8 96.5 3.7 0.1 0.7
15–20 3.8 96.0 3.7 0.1 0.7
10–15 5.7 95.5 5.4 0.3 2.1
rotating drum over which a fabric sleeve is stretched. The
7.5–10 3.8 94.7 3.6 0.2 1.4
volume inside the drum is divided into zones. One zone of
5–7.5 8.8 94.0 8.3 0.5 3.4
the drum, which rotates at from 0.1 to 10 rpm, is in contact
2.5–5 17.6 90.5 16.0 1.6 11.0
with an agitated trough containing the slurry, which is drawn
0–2.5 50.3 77.0 38.7 11.6 79.3
100% 85.4% 14.6% 100%
onto the filter cloth by a vacuum of about 500 torr inside the
drum. As the drum, along with the newly formed cake,
rotates out of the trough, the cake enters a washing zone
The overall collection efficiency for the system of cyclones and where water-soluble impurities are washed out of the cake.
precipitator is 97.7%. It is interesting to note that although these The wash water may be added to the filtrate. In the next
engineering calculations, made in the 1960s, are the same as those zone, the cake may be dewatered by vacuum, mechanical
made today, the allowable air-pollution standards have been pressing, and/or an inflatable diaphragm. After that, the cake
tightened.
is removed from the cloth by blowback pressure for high ro-
tation rates, a knife blade or scraper for low rotation rates, or
by other means. Following this, there may be a brush to clean
the filter cloth; sometimes a precoat of diatomaceous earth
§19.2.7 Filter-Cake Filtration Devices (silica) or perlite fiber is applied. There are many possible
In the above description of bag filters, it was pointed out that drum variations, including pressurized drums.
particles collected on the outside of the cloth form a cake, In recent years, vacuum belt filters have taken a share of
which also acts as a particle collector. For solid–liquid sys- the market for large, continuous industrial filters. A sche-
tems, there is a class of equipment where the slurry is matic of a vacuum belt filter is shown in Figure 19.14. It is
pumped or vacuum driven through a cloth filter, with the sectioned in the same way as a rotary-drum vacuum filter.
cake acting as a filter medium. Not all slurries can be treated Since the early days of the chemical industry, the ven-
this way, but if a cake-based filter is an option, the first step is erable plate-and-frame filter press has been an industry
to make laboratory tests to ascertain under what conditions warhorse. Many variations of this press, which is suitable
the solid will form a suitable cake. Usually, as discussed in only for batch operations, are in use. The design permits
§14.8, the proper concentration and type of filter aid that
needs to be added to the slurry must be researched, and, be- Feed Wash
cause the pressure may not be constant through the entire fil-
ter cycle, the effect of pressure on cake permeability must be
studied. Generally, the slurry should contain less than 35 vol
% solids and the particle size should be above a few microns.
Pc-SELECT, an expert system to guide the laboratory study
and device selection, has been formulated by Wakeman and
Tarleton [7]. The use of a laboratory test leaf filter, usually of Mother Wash Filter
0.1 ft2 filter area, is highly recommended and is described in liquor liquor medle
detail in [11]. Cake
Shown in Figure 19.13 is a rotary-drum vacuum filter, Filter belt Support belt
which operates continuously and consists of a hollow, Figure 19.14 Belt filter.
788 Chapter 19 Mechanical Phase Separations
Inlet
Plate Outlet
delivery of the slurry to the filter cloth, which is backed
by a metal plate; discharge of the filtrate and retention of
the cake; and addition of wash water, with, in some mod-
els, the impurities leaving through a different port. The
device can have from two to four separate ports, and
some presses embody features such as inflatable dia-
phragms that enable cake dewatering by compression, a
process called expression. After the cycle, the press is dis-
assembled, and the cakes are collected manually.
Figures 19.15 and 19.16 show the most common, sim-
plest, two-port configuration, which consists of alternate
plates and frames hung on a rack and pressed together with a
closing (and opening) screw device. The filter cloths, which
have holes to align with the inlet and outlet ports, are hung
over the plates and act as gaskets when the press is closed. A
very large plate-and-frame filter press may have as many as
100 plates and frames, and up to 300 square meters of filter
area. Slurry feed enters from the bottom, and feeds the cavi-
ties in parallel. The filtrate flows through the cloth, channels
in the plate, and out the top, while the cake builds up in the
frame. The frame is full when the cakes, which build up on
both sides of the frame, meet. Other versions of the plate-
Fixed head Solids collect
in frames
Plate Frame
Clear-filtrate
outlet
Material enters
under pressure
Filter cloth
Figure 19.16 Plate-and-frame filter press.
Frame Figure 19.15 Plate-and-frame pair.
and-frame filter press have three- and four-port systems,
which facilitate washing, when required, because if the slurry
fills the frame, the wash water may be blocked if it enters
through the slurry feed lines.
A type of filter press that competes with plate-and-frame
devices in batch-production processes is the pressure leaf fil-
ter, which has the advantage of not having to be disas-
sembled completely after each cycle. Most leaf filters
resemble the baghouse device shown in Figure 19.8. Hori-
zontal and vertical versions of pressure leaf and plate-and-
frame filters are available. Choice of filter equipment is gov-
erned mostly by economic factors, which include relative
cost of labor, capital, energy, and product loss, but attention
must be paid [8] to: (1) fluid viscosity, density, and chemical
reactivity; (2) solid particle size, size distribution, shape,
flocculation tendency, and compressibility; (3) feed slurry
concentration; (4) throughput; (5) value of the product;
(6) waste-disposal costs and environmental problems; and
(7) completeness of separation and material yields. Experi-
mental data are required to establish these parameters, and
pilot-plant testing is a necessity. Proper choice and concen-
tration of filter aid, and the choice and pretreatment of the
Movable head
Closing device
Side raits

filter cloth, are also critical. As in all engineering ventures,
careful attention to details is a necessity.
§19.3 DESIGN OF PARTICLE SEPARATORS
§19.3.1 Empirical Design Equations
What is called an ‘‘empirical design equation’’ here is an
equation that, although it may be scientifically based, has
constants that are device- and substance-specific. Engineer-
ing handbooks, for example, describe generalized graphs and
equations that allow the design of piping systems for all com-
mon fluids, and are valid for all sizes and configurations; no
experimental data are required. This is not the case for
particle-collection devices; their performance has to be cali-
brated. Its collection efficiency for talc particles may be
entirely different from its efficiency for collecting gypsum
particles even though the particle-size distributions are the
same. Particle compressibility, electrical charge, aerody-
namic shape, and agglomeration tendency are some of the
variables for which no general science-based equation exists.
Design equations are empirical, with constants that are use-
specific. Frequently, the equations are not dimensionless, but
must be used only with specified dimensions.
§19.3.2 Mesh Filters
The most frequently encountered equation in the specification
of mesh and vane filters is the Souders–Brown equation (6-40),
which was developed in 1924 to calculate settling velocities for
the purpose of modeling entrainment in distillation columns. In
the version used here for mist-eliminator design, the empirical
constant KL is known as a system load factor, or simply load
factor. Its value is such that the velocity it predicts, the bulk
impingement velocity, u, is normally considerably higher than
the particle settling velocity widely used in mechanical separa-
tions and calculated by the following equation, which looks
exactly like (6-40), but with KL in place of C, which was
obtained in §6.6.1 from a rigorous force balance.
!1=2
rp rf
u ¼ KL (19-1)
rf
General industrial practice is for a manufacturer to pro-
vide KL values that are appropriate for its devices and a spe-
cific use. Amistco, in Alvin, Texas, a manufacturer of
particle-removal equipment, has the following entries in its
product literature [9]:
Table 19.5 Recommended Design Values for
KL in (19-1)
Typical Wire-Mesh Pad, (No-co-knit yarn)
Vertical Flow . . . . . . . . . . . KL ¼ 0.35 ft/s
Horizontal Flow . . . . . . . . . KL ¼ 0.42 ft/s
Typical Vane Unit
Vertical Flow . . . . . . . . . . . KL ¼ 0.50 ft/s
Horizontal Flow . . . . . . . . . KL ¼ 0.65 ft/s
§19.3 Design of Particle Separators 789
KL values for horizontal units are higher than those for
vertical units because the filter is designed to remove liquid
droplets from gas streams, where the droplets are coalesced
and drain from the unit. Drainage of liquid is facilitated so
that the gas velocity can be higher, and re-entrainment
minimized.
For mesh pads and vanes, as with all particle-fluid separa-
tors, pressure drop is a significant operating-cost factor. As
will be seen in the next example, this information can also be
correlated using KL factors.
EXAMPLE 19.2 Removal of Droplets with a Mesh
Filter.
Determine the diameter, D, of a vertical cylindrical vessel, wherein
a 6-inch-thick TM-1109 Amistco pad is used to separate water drop-
lets from air at 70 F. Also, obtain the pressure drop through the
mesh. Pertinent data, which apply to water droplets in air, are: Q,
volumetric vapor flow ¼ 200 ft3/s; rp, droplet (water) density ¼ 62.3
ft3/lb; rf, fluid (air) density ¼ 0.0749 ft3/lb; and manufacturer’s sug-
gested KL ¼ 0.35 ft/s.
Solution
Substituting the above values into (19-1),
1=2
62:3 0:0749
u ¼ 0:35 ¼ 10:09 ft/s
0:0749
Flow area ¼ (200)=(10.09) ¼ 19.8 ft2; vessel cross-sectional area ¼
(3.14)(D)2=4 ¼ 19.8. Solving, vessel diameter ¼ D ¼ 5 ft.
Knowing the vessel diameter is not enough to complete
the detailed design. The particle-size distribution in the
incoming gas feed and the collection efficiency of the mesh
for each particle size, as in Example 19.1, are needed. Also,
it must be ascertained that the calculated velocity is high
enough for inertial-capture mechanisms, but not so high as to
initiate re-entrainment. With respect to the pressure drop, this
can be obtained from the manufacturer. A typical plot is
Figure 19.17, where the pressure drop is shown as a function
of superficial pressure drop, with liquid loading as a parame-
ter. The pressure drop for Example 19.2 is from approxi-
mately 0.2 inch of water at this low liquid loading.
Another example of an empirical approach is the determi-
nation of the depth of mesh filters required to reduce particle
concentrations to a desired level. Here it is assumed that
every differential layer of mesh removes the same fraction of
particles. If N is the number of particles per unit volume and
x the filter depth, it follows that
dN=dx ¼ KN
Integration from N ¼ No, x ¼ 0 to N ¼ N, x ¼ x gives
lnðN o =N Þ ¼ Kx (19-2)
The constant K is evaluated experimentally using a mesh
whose depth is known. This equation must be used carefully
because it is valid only for a very narrow particle-size distri-
bution. An alternative is to use a different K for each size
range in the distribution, if data are available.
790 Chapter 19 Mechanical Phase Separations

Figure 19.17 Pressure drop

for Amistco filter.
[From Amistco Separation
Products, Inc., Alvin, TX,
with permission.]

EXAMPLE 19.3 Removal of Microorganisms with a standard or a prototype cyclone, as in Figure

Filter.
A 15-cm-thick mesh filter is being used to filter uniformly sized
microorganisms from a 10 m3/minute air stream, effectively reduc-
ing the concentration from 200 microorganisms/m3 to 10/m3 over a
100-hour fermentation period. New regulations mandate a reduction
to 1/m3. If the air velocity and ambient conditions do not change,
what depth of the same type of filter is required?
Solution
First, K is evaluated for the current 15-cm filter using (19-2) and the
performance data. The original contamination of 200 microorgan-
isms/m3 was reduced to 10. Thus, ln[(200=10)] ¼ K(15). Solving,
K ¼ 0.2 cm 1. For the new mesh, x is obtained from ln[(200=1)] ¼
(0.2)x. Solving, x ¼ 26.5 cm. The filter required to meet the new reg-
ulations will cost almost twice as much as the old filter. However,
these specifications are not attainable by any present-day filters.
19.18.
2. Get an estimate of the particle-size distribution in the
feed stream to be treated.
3. Estimate the number of cyclones in parallel required.
4. Calculate the cyclone diameter for an inlet velocity
of 15 m/s. Scale the other cyclone dimensions from
Figure 19.9.
5. Calculate d2 using d1, the mean-particle diameter, from
Figure 19.18 and
h i1=2
d 2 ¼ d 1 ðDc2 =Dc1 Þ3 ðQ1 =Q2 ÞðDr1 Þ=ðDr2 Þðm2 =m1 Þ
(19-3)
where d2 ¼ mean diameter of the particles separated in
the proposed design, at the same separation efficiency;
Dc1 ¼ diameter of the standard cyclone ¼ 8 in. (203
mm); Dc2 ¼ diameter of the proposed cyclone, mm;

§19.3.3 Cyclone Design 100

For cyclones, the effect of feed and device parameters is 90
complex, and interdependencies are to be expected. Larger 80
particles go to the wall quickly, but the smaller ones are sepa-
70
Grade efficiency, %

rated from the gas near the bottom vortex where the gas
reverses direction. Design methods, first developed by Stair- 60
mand [10], are based on obtaining particle-collection effi- 50
ciency data for a cyclone of diameter D and establishing 40
geometric ratios that permit scaling up or down. Design 30
methods for solid–liquid cyclone separators are similar to
20
those for solid–gas or liquid–gas units. Stairmand’s design
procedure, as presented by Towler and Sinott [5], who show 10
detailed calculations, is as follows:
0 10 20 30 40 50 60 70
1. Obtain a collection-efficiency versus particle-size Particle size, µm
curve for a feed mixture from the literature for the Figure 19.18 High-efficiency cyclone performance curve.

Q1 ¼ standard fluid flow rate, 223 m3/h; Q2 ¼ proposed
fluid flow rate, m3/h; Dr1 ¼ particle-fluid density dif-
ference, standard cyclone, 2,000 kg/m3; Dr2 ¼ den-
sity difference, proposed design, kg/m3; m1 ¼ standard
fluid viscosity (air at STP) ¼ 0.018 mN-s/m2; and m2 ¼
viscosity, proposed fluid, mN-s/m2.
6. Calculate the cyclone performance and recovery of
particles (efficiency). If the results are unsatisfactory,
try a smaller diameter.
7. Calculate the pressure drop using (19-4) and select a
blower
DP ¼ rf =203 u21 þ u21 2w2 ½ð2rt =re Þ 1 þ 2u22
where DP ¼ cyclone pressure drop, mbar; rf ¼ fluid
density, kg/m3; u1 ¼ inlet duct velocity, m/s; u2 ¼ exit
velocity, m/s; rt ¼ radius of circle to which the center-
line of the inlet is tangential, m; re ¼ radius of exit
pipe, m; w ¼ factor given in [5]; c ¼ parameter in [5]
given by c ¼ fc(As=At); fc ¼ friction factor, 0.005 for
gases; A1 ¼ area of inlet duct, m2; and As ¼ surface
area of the cyclone exposed to the spinning fluid, where
length equals total height times cross-sectional area of
a cylinder with the same diameter, m2.
§19.3.4 Hydrodynamic-Based Equations
Mathematical models used to describe the behavior of parti-
cles that separate from fluids, primarily because of gravita-
tional forces, invariably are based on the terminal velocity of
the particle, ut, which is defined as the fluid velocity that ren-
ders a particle, subject to gravitational force, motionless
when suspended unhindered in an upward-flowing fluid
stream. At that condition, the drag force on the particle plus
the buoyant force balance the force of gravity. The terminal-
velocity concept was previously encountered in §6.6.1, where
100,000
10,000
Drag coefficient, CD = 2Fd /Apρf u 2
Spheres
1000
Disks
100
Cylinders
10
1.0
0.1
0.0001 0.001 0.01 0.1 1.0 10 100
Reynolds number, NRe = dpuρf /µ
§19.3 Design of Particle Separators 791
it was used to derive (6-40) to model distillation-column and
reflux-drum diameters. This equation, a combination of
(6-40) and (6-41), is
2 31=2
4d p g rp rf
ut ¼ 4 5 (19-5)
3C D rf
From (6-39), the drag coefficient CD in (19-5) is related to
the drag force Fd on the projected area, Ap , of a spherical parti-
cle by
!
pd 2p u2
F d ¼ CD r (19-6)
4 2 f
(19-4) where: dp is particle diameter; rp is particle density; rf is fluid
density; ut is terminal velocity (or settling velocity in a quies-
cent fluid); g is acceleration due to gravity; and CD is the
dimensionless drag coefficient. If AE units are used in (19-6),
the denominator must include gc (e.g., 32.174 lbm-ft/lbf-s2), to
convert mass to force.
§19.3.5 Drag Coefficient
Essential to the use of Equation (19-5) are numerical values
for the drag coefficient. Fortunately, measurements of drag
coefficients and their theoretical interpretations have been
the subject of extensive research, and correlations such as
Figure 19.19, which is a plot of CD versus particle Rey-
nolds number, NRe ¼ dpur=m, are available. Use of this plot
in conjunction with (19-5) frequently leads to trial-and-error
calculations because the Reynolds number, which contains
particle velocity as a variable, must be known before CD
values can be obtained. Drag coefficients may also be a
function of variables not displayed in the plot, which leads
to additional correlations and equations. These include: (1)
particle velocity history, (2) particle shape, (3) the effect of
walls and collisions with other particles, and (4) random
Brownian movement, if the particles are very small.
Figure 19.19 Effect of particle
Reynolds number on drag coefficients.
1000 10,000 100,000 1,000,000
[From Lapple and Shepherd, Ind.
Eng. Chem., 32, 605 (1930).]
792 Chapter 19 Mechanical Phase Separations
A particle’s history, which includes movement at any
velocity other than the terminal velocity, is universally
neglected in evaluating CD, although methods of correcting
for past accelerations have been researched. The particle-
shape factor poses more of a problem, as can be inferred
from the fact that there are separate curves for disks, cylin-
ders, and spheres in Figure 19.19. Falling objects may rotate
as they fall, which makes the calculation difficult, with the
net effect of increasing drag on the particle; and circulation
can occur in droplets. The drag coefficient is discussed in
depth in advanced fluid-mechanics textbooks. Other impedi-
ments to falling are collisions with walls and other particles.
This leads to the development of equations for hindered set-
tling (in contrast to free settling). The last of the factors to be
considered here, Brownian motion, is random particle motion
occasioned by the collision of very small particles with sur-
rounding gas molecules or atoms. This will be treated as a
separate subject but applied, if necessary, as a correction fac-
tor to the settling velocity equations.
§19.3.6 Settling Equations
To avoid having to make trial-and-error calculations for CD,
and to facilitate calculations for settling velocities, it is con-
venient to divide Figure 19.1 for spherical particles into low,
high, and intermediate regions, wherein equations can be
written to relate u to CD.
Low Reynolds Number Region, Stokes’ Law
Stokes’ law, Fd ¼ 3pmudp, which applies at NRe < 2, gives
CD ¼ 24=NRe. Substitution into (19-5) gives the settling
velocity for a spherical particle of diameter dp as
gd 2p rp rf
ut ¼ (19-7)
18m
This equation can be used for Reynolds numbers from
0.001 to 2, with an error for CD, at the highest NRe, of
about 10%. This translates into a usually negligible 5%
error in particle velocity. Note that (19-7) and the ensuing
equations in this section are limited to spheres falling in a
gas or liquid of low viscosity. For a listing of recent liter-
ature on particles falling through Bingham plastics and
other non-Newtonian fluids, and corrections for nonspheri-
cal particles, see [11]. Small variations in the numerical
constants in the above equations as well as in other equa-
tions in this section appear in the literature. The values
here stem from [12].
High Reynolds Number Region, Newton’s Law
For Reynolds numbers between 500 and 200,000, the drag co-
efficient is almost independent of the Reynolds number, and
the corresponding settling velocity and drag force for a spheri-
cal particle are, respectively, C D ffi 0:44 and F d ¼ ð0:055Þ
pd 2p u2 rf , resulting in
2 31=2
d p g rp rf
ut ¼ 1:744 5 (19-8)
rf
Intermediate Reynolds Number Region
Between the Stokes and the Reynolds regions, where NRe lies
between 2 and 500, C D ffi 18:5N Re0:6 , resulting in
0:71
ut ¼ 0:153g0:71 d 1:14
p rp rf r0:29
f m0:43 (19-9)
Cunningham Correction to Stokes’ Law
A correction to Stokes’ law is important for particles under
3 mm in diameter for settling in gases and under 0.01 mm for
settling in liquids. In gases, small particles can slip between
the gas molecules with less drag, resulting in a terminal
velocity higher than that predicted by Stokes’ law, (19-7).
This occurs when the mean free path of the gas, l (0.0065
mm for ambient air), is comparable to the particle diameter.
The increase in terminal velocity can be predicted with the
Cunningham slip correction factor, Km, which is a multiplier
to the Stokes settling velocity, ut, given by
l 0:656d p
Km ¼ 1 þ 1:644 þ 0:552 exp
dp l
(19-10)
For a 0.01-mm particle falling in ambient air, Km ¼ 2.2. Thus,
the particle falls more than twice as fast as predicted by
Stokes’ law.
Brownian Motion
Oscillatory, zigzag motion of particles whose size falls in the
0.1–0.001-mm range was first observed in 1826 by the British
botanist Robert Brown. It was the first visual confirmation of
the correctness of the kinetic theory of matter, which pre-
dicted that this motion is due to unbalanced impacts of mole-
cules or atoms on particles. Einstein [13] was the first to
obtain the following theoretical expression for the average
distance Dx moved through by a particle of radius r in a liq-
uid of viscosity m during time t, where NA is Avogadro’s
number and T is absolute temperature. There is a correspond-
ing equation for the rotary movement.
ðDxÞ2 ¼ RTt=3pmN A r (19-11)
Particles larger than 2–3 mm are collected by inertial im-
paction and direct interception, but for smaller particles,
Brownian movement becomes important. The displace-
ments due to Brownian movement for water droplets in
air have been measured by Brink [14] and are given in
Table 19.6. Because of this motion, submicron particles,
given enough time, will coagulate. In general, in fiber and
other types of fine-particle mist collectors, where the gas
 §19.3 Design of Particle Separators 793

Table 19.6 Brownian Displacement where the constants are:

Particle Brownian
Diameter, mm Displacement, mm/s Law b n

0.1 29.4 Stokes 24 1.0

0.25 14.2 Intermediate 18.5 0.6
0.5 8.92 Newton 0.44 0
1.0 5.91
2.5 3.58
5.0 2.49 Hindered Settling
10.0 1.75
The settling-velocity equations apply to single particles and
predict higher settling velocities than are observed when the
concentrations of particles are high enough that settling is
velocity is less than 0.075–0.20 m/s, Brownian movement hindered by particle–particle collisions. Various approaches
is the controlling mechanism for particle collection; how- for correcting terminal velocities for hindered settling appear
ever, the design techniques combine all hydrodynamic in the literature; the one used here is due to Carpenter [16].
mechanisms into some type of an empirical correlation For a spherical particle of uniform size,
based on experimental data. a
Brownian motion is superimposed on the particle velocity, ush ¼ u t 1 fparticles (19-14)
and it lowers the efficiency of capture devices that are based
on collisions of particles with fibers because small particles, where ush ¼ hindered settling velocity, fparticles ¼ the volume
at very low velocities, will follow the streamlines around the fraction of particles, and a ¼ an exponent whose value is
fiber collectors. given in Table 19.8.

Criteria for Settling Equations

Table 19.8 Hindered Settling
Although the range of applicability of the above settling
NRe a
equations has been stated, a commonly used concept is to use
a single criterion based on the highest Reynolds number for 0.5 4.65
0.0875
which the equation applies, which for Stokes’ law is 2. The 0.5 NRe 1300 4.374(NRe)
criteria is based on a general procedure used by McCabe NRe 1300 2.33
et al. [15] and Carpenter [16] that eliminates the terminal-
velocity factor, ut, from the Reynolds number by substituting
one of the settling equations. This results in an empirical
equation, EXAMPLE 19.4 Settling of Particles.
h i1=3
For a particle 0.01 inch in diameter, determine (a) the proper equa-
K c ¼ 34:81d p rf rp rf =m2f (19-12)
tion to use for the settling velocity, (b) the terminal (unhindered)
velocity, (c) the hindered settling velocity, and (d) the velocity in a
where dp is in inches and mf is in cP, with densities in lb/ft3. centrifugal separator where the acceleration is 30 g. The pertinent
The constant Kc, which is listed in Table 19.7 along with the data are: rf ¼ 0.08 lbm/ft3, mf ¼ 13.44 10 6 lbm/ft-s ¼ 0.02 cP, rp
range of applicability of the settling equations, can be used to 500 lbm/ft3, and fparticles ¼ 0.1.
determine if the equation is suitable for the particle size in
question. Having obtained Kc, it is then convenient to calcu- Solution
late the settling velocity from a general settling-velocity
(a) and (b) Solving (19-12),
equation, h i1=3
h i K c ¼ 34:81ð0:01Þ 0:08ð500 0:08Þ=ð0:02Þ2 ¼ 16:16
ð1þnÞ ð1 nÞ 1=ð2 nÞ
ut ¼ 4g d p rp rf =3bmnf rf (19-13)

Table 19.7 Ranges of Settling Equations

Newton’s Law Intermediate Law Stokes’ Law Stokes–Cunningham Law Brownian Movement

dp 100,000–1,500 1,500–100 100–3 3–0.1 0.1–0.001

NRe 200,000–500 500–2 2–0.0001
Kc 43.6 3.3 Kc 43.6 3.3

with dp in microns
794 Chapter 19 Mechanical Phase Separations

For Kc ¼ 16.16 in Table 19.7, the settling velocity is in the Examination of (19-17) shows that the separation can be
intermediate range, so b ¼ 18.5 and n ¼ 0.6. Using (19-13) for sharpened by choosing a fluid that has a density close to that
the settling velocity, of one of the particles. A liquid-phase density can be altered
ut ¼ ½4ð32:2Þð0:01=12Þ1:6 ð500 0:08Þ= by adding a thickener or very fine particles that do not settle.
0:6
3ð18:5Þ 13:44 10 6 ð0:08Þ0:4 1=1:4 ¼ 11:78 ft/s Density adjustment is also the basis for separating enzymes
and other biological systems by aqueous two-phase extrac-
It is possible to check if the selection of the intermediate range tion. For this application, centrifuges are used [17].
is correct by calculating the particle Reynolds number. NRe ¼
dputrf/mf ¼ [(0.01/12)(11.78)(0.08)/(13.44 10 6)]¼ 58.4.
Checking Table 19.7, it is seen that the correct choice of region
EXAMPLE 19.5 Separation of Particles by
was made.
Classification.
(c) For NRe ¼ 58.4, by Table 19.5, a ¼ 4.374(NRe) 0.0875 ¼ 3.064.
Thus, for a 0.1 volume fraction of solids, the hindered settling A mixture of particles A and B is to be separated by classification
velocity from (19-14) is ush ¼ 11.78(1 0.1)3.064 ¼ 8.53 ft/s. using water. The size range for both A and B is between 7 and 70
This is a 28% reduction in velocity. mm, with rA ¼ 8 g/cm3 and rB ¼ 2.75 g/cm3. Assume unhindered
(d) To find the velocity at 30 g acceleration, it is necessary to multiply settling and a water viscosity of 1.0 cP (0.01 poise). (a) What veloc-
g by 30 in (19-13), so ut in the centrifugal field is the velocity ity will give a pure A product? (b) What is the size of the largest A
from part (c) multiplied by 30 to the 1/(2 n) power, where particle swept out with the B particles?
n ¼ 0.6. Thus, ut ¼ 8.53(30)0.714 ¼ 96.7 ft/s.
Solution
(a) Since A will settle faster than B, the water velocity must be
§19.3.7 Particle Classification larger than the settling velocity of the largest B particle. Assum-
ing that ut of the largest B particle is in the Stokes’ settling
Separation of particles in accordance with their size is called domain, (19-7), with CGS units, gives
classification. If two particles have different terminal veloc- ut ¼ 980:7ð0:007Þ2 ð2:75 1Þ=18ð0:01Þ ¼ 0:468 cm/s
ities in air, it is possible to adjust the air velocity so that one
particle remains suspended while the particle having the The Reynolds number, NRe ¼ (0.007)(0.468)(1)=0.01 ¼ 0.33,
higher terminal velocity falls. Likewise, as is illustrated in which is in the Stokes’ law region, as assumed.
Figure 19.20, if a group of particles is injected into a moving (b) It is necessary to calculate the size of the heavier A particle that
body of water, the particle with the lowest terminal velocity settles at 0.468 cm/s. From a rearrangement of (19-7),
2 31=2
will be found farthest downstream. For two groups of 1=2
18ut m 5 ¼ 18ð0:468Þð0:01Þ
different-density particles, 1 and 2, with a range of sizes, and dA ¼ 4 ¼ 0:0035 cm
g r r 980:7ð8 1Þ
with 1 denser than 2, complete separation is unlikely because A f
the size range overlaps. This overlap occurs when particles in ¼ 35 mm
the two groups have equal terminal velocities. If NRe < 2
for all particles, from (19-7), Thus, any A particle smaller than 35 mm will be swept out
along with all the B particles.
ut ¼ gðd 1 Þ2 r1 rf =18m ¼ gðd 2 Þ2 r2 rf =18m
(19-15)
Dividing the two RHS equalities,
§19.3.8 Gravity Decanter
ðd 1 Þ2 =ðd 2 Þ2 ¼ r2 rf = r1 rf (19-16)
The design method suggested here for a liquid-liquid separa-
For all NRe, the general result from (19-5) is tion is due to Schweitzer [18], as used by Coker [12].
n
ðd 1 Þ=ðd 2 Þ > r2 rf = r1 rf (19-17)
1
where n ¼ /2 for laminar flow; n ¼ 1 for turbulent flow; EXAMPLE 19.6 Separation of Oil from Water by
n ¼ 0.625 for intermediate flow. Settling.
A decanter to separate oil from water is needed. The oil flow is
Fluid in, Fluid out, 8,500 lb/hr, and the water rate is 42,000 lb/hr. It is anticipated that
wide range some fine there will be oil droplets in the water layer. Obtain the dimensions
of particle particles
sizes
of the horizontal decanter. The tank will have an L/D (length-to-
diameter) ratio of 5. Ignore the hindered settling effect. The follow-
ing nomenclature and data apply: ut ¼ oil-droplet terminal velocity,
ft/s; g ¼ acceleration of gravity, 32.2 ft/s2; dp ¼ oil-droplet diameter,
Coarse Intermediate Fine
particles particles particles assumed uniform at 150 mm (0.00049 ft); rp ¼ oil density, 56 lb/ft3;
rf ¼ fluid (water) density, 62.4 lb/ft 3; mf ¼ fluid (water) viscosity,
0.71 cP or 4.77 10 4 lbm/ft-s; moil ¼ oil viscosity, 9.5 cP or
63.84 10 4 lbm/ft-s; Qoil ¼ 8,500/(56)(3,600) ¼ 0.0422 ft3/s; and
Qwater ¼ 42,000/(62.3)(3,600) ¼ 0.1873 ft3/s
Figure 19.20 Classification by gravity settling.
 §19.4 Design of Solid–Liquid Cake-Filtration Devices Based on Pressure Gradients 795

Solution Slurry Feed

Although Qoil Qwater, it is best to make sure that oil is the dis- z P
persed phase. The criteria suggested by Schweitzer [18] are based
on a parameter, u, in terms of light, l, and heavy, h, phases:
dz Filter Cake
0:3
Q ¼ ðQl =Qh Þ½rl mh =rh ml (1)
where Pi
Filter
Po Medium
Q RESULT z=0
Filtrate Flow
<0.3 Light phase always dispersed
0.3–0.5 Light phase probably dispersed
Figure 19.21 Filtration profile.
0.5–0.2 Phase inversion probable
2.0–3.3 Heavy phase probably dispersed
>3.3 Heavy phase dispersed experiments he found the flow rate to be proportional to the
pressure drop, indicating laminar flow, which is also the case
Applying (1), for cake filtration when Reynolds numbers are below one and
4 4 0:3 inertial effects are negligible. As has been pointed out by
Q ¼ ð0:0422=0:1873Þ ð56Þ 4:77 10 =ð62:3Þ 63:84 10
Wakeman and others [7], Darcy did not include viscosity in
¼ 0:10 his original equation, which, with viscosity for application to
Clearly, oil will be the dispersed phase. Next it is necessary to filtration, is
decide which settling law applies, so Kc is calculated using (19-12), k dP
with AE units, and substituting the absolute density difference for J¼ (19-18)
m dz
(rp rf):
h i1=3 where dP is the pressure drop through thickness dz of a
K c ¼ 34:8ð0:00049Þ ð62:4Þð6:4Þ=ð0:71Þ2 ¼ 0:128 medium of permeability k and J ¼ u is the fluid velocity (vol-
ume flow/unit filter area). For cake filtration, it is customary
With reference to Table 19.7, it is seen that Kc is in the Stokes’ law
to replace the permeability, k, with a, the specific cake resist-
range.
Since turbulence in a gravity settler is undesirable, it is also nec-
ance, and dP=dz with dP=dW, where W is the mass of the dry
essary to check the Reynolds number of the fluid after the vessel filter cake per unit filter area and dW and dz are related by
dimensions are established. By the criteria in Table 19.7, NRe should dW ¼ rc ð1 eÞdz (19-19)
be < 2.
where e is the fraction of voids in the filter cake and, thus,
By Stokes’ law, (19-7), the settling velocity (in this case, the rise
velocity of the oil droplets) is is a measure of the volume of flow paths through the
medium. Unfortunately, for filter cakes, the complexity of
ut ¼ 32:2ð0:00049Þ2 ð56 62:4Þ= 18 4:77 10 4
¼ 0:0058 ft/s their structure, and their dependence on pressure, preclude
The negative value arises because the oil droplets rise rather than direct calculation by either e or a from scanning-electron
settle. It is now possible to obtain the vessel dimensions. Assuming micrographs or other means. Nevertheless, Kozeny, in
that the length of the vessel is five times the diameter, L=D ¼ 5, and 1927, and Carman, in 1938, developed a theoretical flow
that the width of the interface, which is not at the top, is 0.8D, the model based on pores being replaced by a bundle of capil-
phase-interface area, A, is (0.8D)(5D) ¼ 4D2, and since Qwater=A lary tubes whose orientation is at 45 to the surface. Their
ut, D 0.5(0.1873=0.0058)1/2 ¼ 2.84 ft, and L ¼ (5)(2.84) ¼ 14.2 ft. equation is based on the Poiseuille equation for laminar
For additional design calculations to establish Reynolds numbers flow through a straight capillary, but with the straight cap-
and specifications for this separator, see [12].
illary being replaced by more complex geometric con-
structs. §14.3 and 6.8 contain expositions on this subject
as it applies to flows of liquids through membranes and
packed beds. The Kozeny–Carman equations have limited
§19.4 DESIGN OF SOLID–LIQUID CAKE- applicability to cake filtration because e and a must be
FILTRATION DEVICES BASED ON PRESSURE evaluated experimentally. A more empirical and widely
GRADIENTS used approach to finding a relationship between process
variables in filtration is to consider the two pressure drops
Pressure-filtration devices consist of a cloth or mesh barrier in Figure 19.21, one through the medium and the other
(the medium) that retains suspended solids (the cake), while through the cake. Denoting the medium resistance to flow
allowing the fluid in which the solids are suspended (the fil- by Rm and the cake resistance to flow as Rc, Darcy’s law,
trate) to pass through. As shown in Figure 19.21, it is custom- (19-18), can be applied to each resistance,
ary to treat the pressure drops through the cake and the
medium as separate entities. Filtration models in this chapter u ¼ ðP Pi Þ=mRc ¼ ðPi Po Þ=mRm (19-20)
and (14-81) are based on Darcy’s law, developed in 1855 to Letting the total pressure across the cake and medium
describe the flow of water through sand beds. In his DP ¼ (P Po), V ¼ volume of filtrate, Ac ¼ cake area, and
796 Chapter 19 Mechanical Phase Separations
u ¼ (dV=dt)(1=Ac).Combining with (19-20) gives
dV=dt ¼ DPAc =mðRm þ Rc Þ (19-21)
which states that dV=dt is directly proportional to Ac and DP,
and inversely proportional to the filtrate viscosity, m, and the
sum of the resistances of the cake and medium. Usually, Rm
is much smaller than Rc after the cake begins to build up.
When pressure is in AE units, the RHS must be multiplied
by the gravitational constant, gc; Rc and Rm have dimensions
of reciprocal length.
As filtration progresses, cake thickness, filtrate volume, and
resistance to flow increase, but Rm is assumed to remain con-
stant. W, the weight of the dry cake, is related to V, filtrate
volume, and cF, the dry cake mass per unit volume of filtrate,
by W ¼ cFV. However, care must be exercised when applying
this formula because the cake is wet, and then dried, so cF will
have different values depending on whether wet or dry cake
masses are used. It now becomes useful to replace Rc by a, the
specific cake resistance, which replaces k in the original Darcy
equation, (19-18), where now Rc ¼ aW=Ac ¼ acFV=Ac. Thus, if
length is in ft, Rc has units of 1/ft and a has units of hr2/ft
because the compressibility has been multiplied by gc for di-
mensional consistency [15, 19]. For SI units, a has dimensions
of m/kg. Substituting the definition of a into (19-21),
d ðV =Ac Þ DP
¼ (19-22)
dt m½Rm þ acF ðV=Ac Þ
It is common to consider the application of (19-22) to two
regimes of filtration: constant-pressure and constant-flow rate.
§19.4.1 Constant-Pressure Filtration
Assuming the filtration area is constant and the pressure drop
is constant, (19-22) can be integrated from V ¼ 0 to V ¼ V for
t ¼ 0 to t ¼ t to yield the constant-pressure form of what was
termed the Ruth equation in §14.3.1, and which was devel-
oped in 1933 [20]:
V 2 þ 2VV o ¼ K t or t=V ¼ ðV þ 2V o Þ=K (19-23)
where V o ¼ Rm Ac =acF and K ¼ 2A2c DP=acF m
It is well to consider what kind of a pump can deliver a
slurry to a filtration unit at constant pressure. Such a pump is
certainly not a positive displacement pump, because it oper-
ates at constant flow rate. Centrifugal pumps are sold with
performance charts (characteristic curves) that display their
flow rate as a function of pump pressure or head, so it is pos-
sible to devise a pump control system that forces a centrifu-
gal pump, up to a point, to maintain a constant pressure even
though the flow diminishes. If the pump is not controlled, it
will maintain an output flow and pressure that follow the
characteristic curve supplied by the manufacturer.
In practice, regardless of what type of filtration device is
used, the cake will be wet, and if it is the final product, it
must be dried before being sold. Often before drying, which
is energy-intensive, the cake is subjected to expression to
wring out excess moisture. Predrying devices include
machines that are vice-like presses, centrifugal separators
like a laundry dryer without a heater, or inflatable
diaphragms inserted between frames of a plate-and-frame fil-
ter. Frequently, the filtrate imbibed in the filter cake contains
water-soluble impurities, which must be washed out of the
cake prior to expression and final drying. The optimal eco-
nomics is to use the filtration apparatus to also conduct the
water wash and the expression so that, if necessary, a ‘‘wash
cycle’’ and time for expression are appended to the filtration
cycle, and the throughput rate for a filtration system is
obtained by dividing the total throughput by the sum of the
required wash, expression, and filtration times. Once the Vo
and K constants are obtained, the Ruth equation can be used
to obtain either wash or filtration rates and filter areas.
When the constants Vo and K are evaluated from constant-
pressure laboratory or pilot-plant data, (19-22) can be used to
model large-scale units operating with the same feed, con-
centration of filter aids (if any), and pressure drops. For
example, (19-23) can be modified to model a continuously
rotating drum filter operating at constant speed of n rpm,
where u is the fraction of the drum immersed in the slurry
tank, V 0 /n represents the filtrate volume filtered in one revo-
lution, and A0 is the drum area. The modified equation is
2
ðV 0 =nÞ þ 2ðV 0 =nÞðV o Þ ¼ K ðu=nÞ (19-24)
If the resistance of the medium is negligible in compari-
son to the resistance of the cake, Vo ¼ 0 and (19-1) becomes
2unDP 1=2
Volume of filtrate per unit time ¼ V 0 n ¼ A0
acF m
(19-25)
As discussed in §14.8, in many applications, a, the spe-
cific resistance of the cake, is a function of the pressure drop
across the cake because the filter cake is compressible. This
is particularly true for bacteria and other ‘‘soft’’ cakes, where
an increase in pressure does not, as predicted by (19-22), pro-
duce a directly proportional increase in the volume of filtrate
and cake. In that case, an adjustment to all of the above equa-
tions that contain a should be made by relating the pressure
difference to the cake compressibility by an empirical equa-
tion. Table 19.9 [21, 22] lists cake compressibility factors for
several inorganic filter cakes for the equation
a ¼ a0 ðDPÞs (19-26)
0
where DP is in psi. Note that in the table, a varies by a factor
of 10,000 and s varies from 0.2664 to 1.01.
The constant s, which is zero for an incompressible cake,
must be evaluated from experiments in which the filtration
pressure is varied. All filtration equations should be modified
Table 19.9 Filter-Cake Compressibility
10
Substance a0 10 , m/kg Exponent s
Calcium carbonate 1.604 0.2664
Kaolin, Hong Kong 101 0.33
Solkofloc 0.0024 1.01
Talc 8.66 0.51
Titanium dioxide 32 0.32
Zinc sulfide 14 0.69
 §19.4 Design of Solid–Liquid Cake-Filtration Devices Based on Pressure Gradients 797

by substituting (19-26) for a when the filter cake is com-

pressible. Modified equations and examples can be found in
[21]. The wide range in values for s and a 0 underscores the
need for experimental data prior to undertaking a new design.
Reliable predictive equations are not available.
Tables of cake compressibility in the literature also often
contain values for e, the cake void fraction. If sample calcula-
tions are not provided, it is best to consult the original article
to determine how the a and e were calculated. If separately,
then a can be used with (19-23) and similar equations that do
not require void fractions. The values of e can then be used to
obtain cake properties such as particle surface area and to
calculate an accurate cake thickness, as will be shown in
Example 19.8. Sometimes, however, the e and a are calcu-
lated simultaneously by curve fits, so they are interrelated,
and care must be taken in applying the values. In either case,
it is always worthwhile to consult the original journal refer-
ences unless sample calculations are offered. Another consid- for each run. (b) Use the data to obtain the cake compressibility factor,
eration is that these ‘‘constants’’ are pressure dependent and a 0 , and s in (19-1). (c) The data will be used to size a production unit
represent average values over a range of pressure, so unless that will process 300 ft3 of filtrate, with a filtration time of one hour
this range is reported, nothing is known with certainty for each cycle. If a plate-and-frame filter press is used, what filter area
because extrapolations should not be made. will be required if the anticipated pressure drop is 5.5 psi? (d) A
In Examples 19.7 and 19.8, it will be seen that a and e are rotary-drum vacuum filter is available at the plant. The fraction of the
both functions of pressure. According to the laws of Darcy and drum area submerged is 0.30, and the rotation speed is 10 rph. The
Kozeny, the hydraulic pressure gradient should be linear, and drum is 6 ft in diameter and 10 ft wide, and the system is expected to
when it is not, it is because of substantial variations in a and e run at DP ¼ 5.5 psi. Is this device suitable for the application?
through the cake. The more compressible the cake, the larger
the changes. It has been inferred from experiments that, for Solution
cubic packed, hard spheres, there is no change in void fraction Inspection of the data reveals that, as expected for constant-pressure
through the cake, but for highly compressible latex slurries, the runs, the rate of filtration decreases with time because the cake
flow area is reduced from 5% to 50% from the top of the cake thickness increases. That the rate is linear can be deduced from the
to the surface of the medium. In that case, the solid actually fact that the data for each of the three test runs, after a short time,
flows through the medium as it replaces the fluid in the voids can be plotted as straight lines, as shown in Figure 19.22. This veri-
at a velocity that can be 19–50% of the liquid velocity. fies the mathematical model, (19-23), which has a slope of 1=K and
an intercept of 2Vo=K on a plot of t=V versus V. The data points
below 1 L were neglected and the intercept was obtained by extrap-
EXAMPLE 19.7 Design of a Filtration System. olation because often there is a brief, higher constant-rate period
before the cake starts to build up. If the specific cake resistance is a
A process transmittal from R&D to the engineering department function of pressure, this can be ascertained by calculating a for
requests that a filtration system be designed based on information each of the three runs, which were at different pressures.
from three laboratory filtrations conducted at constant pressure as Once Vo and K are known, Rm and a are obtained from their defin-
follows: ing equations below (19-23). Another way of solving the problem is to
use any two of the data points from a given run, substitute them into
Test 1, Test 2, Test 3, (19-23), and solve the two equations simultaneously for Vo and K.
DP ¼ 5.5 psi DP ¼ 18 psi DP ¼ 53 psi However, this method is not as reliable as making a plot and fitting
Filtrate
the best line through all five data points because the two points chosen
Volume (L) t=V, s=L t, s t=V, s=L t, s t=V, s=L t, s
may be unrepresentative. Using AE units, calculations are illustrated
0 (extrapolated) 27 — 20 — 5 — for Test 1, using the best line through the data in Figure 19.22.
0.5 38 19 25 12.5 8 4 (a) For Test 1: K ¼ 1/slope ¼ (4 1)/(70 38) ¼ 0.0938 L2/s ¼
1 38 38 28 28 9 9 1.17 10 4 ft6/s
2 48 96 36 72 14 28 Intercept ¼ 27 s/L ¼ 764 s/ft3
3 59 177 42 126 17 51 Using the equations just below (19-23) to obtain Vo and Rm,
4 70 280 51 204 21 84 Intercept ¼ 2Vo=K. Therefore,
5 — — 58 370 26 130 4
V o ¼ ð764Þ 1:17 10 =2 ¼ 0:045 ft3
6 — — 66 396 30 180
Rm ¼ ðADP=mÞðInterceptÞ ¼ ð0:75Þ½ð5:5Þð144Þ ð764Þð32:2Þ=
4
6 10 ¼ 2:44 1010 =ft
It is reported that the filter area was 0.75 ft2, m ¼ 6 10 4 lbm/ft-
s, cake density is 200 lb/ft3, and cF ¼ 1.5 lb/ft3. (a) Obtain values for a ¼ Rm A=V o cF ¼ 2:44 1010 ð0:75Þ=½ð0:045Þð1:5Þ ¼ 2:71
constants Vo and K in (19-23) and specific cake resistance a, and Rm 1011 ft/lbm
798 Chapter 19 Mechanical Phase Separations

At 10 rph, the drum filter can process (10)(5.8) ¼ 58 ft3 of fil-

trate/h, which is much less than 300 ft 3/h. So the existing
rotary-drum filter is inadequate.

EXAMPLE 19.8 Selection of a Filter from Lab Data.

The characteristics of an aqueous slurry at 68 F are being
investigated in a laboratory apparatus to determine what class of
filter equipment would be suitable. The mass fraction of solids in
the slurry, xs, is 0.01, and their specific gravity is 2.67. Labora-
tory runs were made at constant pressure drops of 10, 20, 35, and
50 psi, with the data at 10 psi shown in the table below, and data
for the other runs plotted in Figure 19.24, where V/A is the fil-
trate volume flow rate divided by the filter area. The filter area is
0.01 ft 2, and xc, the weight fractions of moisture in the cakes,
were 0.403, 0.431, 0.455, and 0.470, respectively, as measured
Figure 19.23 Cake compressibility factors for Example 19.7.
after each run. Determine cake thicknesses at 30 minutes and
average cake porosities.

Similarly, the values for Test 2 and Test 3 were calculated and
are listed in the following summary:
Experimental Data for 10 psi Run
A ¼ 0.01 ft2, xc ¼ 0.403
Test 1 Test 2 Test 3 t, min V, ml V=A, ft3=ft2 t=(V=A) 10 3, s/ft

Vo, ft3 0.045 0.048 0.021 1 18 0.063 0.94

K, ft6/s 1.17 10 4 1.62 10 4 4.61 10 4 3 42.5 0.150 1.20
a, ft/lbm 2.71 1011 6.45 1011 10.35 1011 5 59.0 0.208 1.44
10 96.0 0.338 1.78
Rm, 1/ft 2.44 1010 5.90 1010 4.35 1010
15 120.0 0.424 2.12
20 143.0 0.505 2.37
25 165.0 0.583 2.58
(b) All a values listed above are plotted against pressure drop
in Figure 19.23. A least-squares fit of the data using Equation 30 181.0 0.639 2.82
(19-1), a ¼ a 0 (DP)s, gives a 0 ¼ 1.04 1011 and s ¼ 0.59.
(c) From part (a), for DP ¼ 5.5 psi, Rm ¼ 2.44 1010 ft 1, and a ¼
2.71 1011 ft/lbm. Solution
Using (19-23), (300)(300) þ (2)(300)Vo ¼ Kt.
The experimental data above and the plot in Figure 19.24 were, as
V o ¼ Rm Ac =acF ¼ 2:44 1010 Ac = 2:71 1011 ð1=5Þ will be shown, used to calculate the values in the table below, where
¼ 0:0600 Ac ft3 with Ac in ft2 cF is the lb dry cake/ft 3 of filtrate, a is the cake compressibility, and
K ¼ 2A2c DPgc =acF m ¼ 2A2c ð5:5Þð144Þ32:2= Rm is the medium resistance.

2:71 1011 ð1:5Þ 6 10 4

¼ 2:1 10 4 A2c ft3 /s 3
2
with A in ft 30 min

Time, t, ¼ 1 hour ¼ 3,600 s

Therefore, (19-23) becomes 90,000 þ 600(0.0600)Ac ¼ 2.1
i
ps

10 4A2c (3,600).
10

2
Solving for the positive root, Ac ¼ 370 ft2.
t/, (V/A) × 10-3, s/ft

(d) Assume the same type of filter cloth is used on the large filter. 20 min
i
ps
20

2
Area of drum ¼ ðperimeter ÞðwidthÞ ¼ ð3:14Þð6Þð10Þ ¼ 188:4 ft
p si
Available area for filtration per revolution ¼ ð0:3Þð188:4Þ 10 min 30
¼ 56:5 ft 2 1 si
p
50
Time per rotation ¼ 1=10 ¼ 0:1 h ¼ 360 s
Time for filtration per rotation ¼ 0:3ð360Þ ¼ 108 s
By interpolation of the test at DP ¼ 5.5 psi for 108 s, get 2.18 L
of filtrate for 0.75 ft 2. 0
0 0.5 1.0 1.5
2
Therefore, for 56.5 ft of the drum, get 2.18(56.5/0.75) ¼ 164 L V/A, ft 3/ft2
of filtrate/rotation, or 164(0.0353) ¼ 5.8 ft3 of filtrate/revolution.
Figure 19.24 Experimental filtration data for Example 19.8.
 §19.4 Design of Solid–Liquid Cake-Filtration Devices Based on Pressure Gradients 799

Calculated Values for Constant-Pressure Filtration Tests of Figure 19.24

DP, cF , e, void L, cake Slope 10 3 Intercept 10 3

a 10 11, Rm 10 10
,
psi lb/ft3 fraction thickness, inch Fig. 19.24 Fig. 19.24 ft/lb ft 1

10 0.640 0.798 0.144 3.252 0.72 6.99 4.97

20 0.639 0.779 0.180 1.946 0.370 8.60 5.10
35 0.638 0.762 0.210 1.314 0.200 10.17 4.82
50 0.638 0.750 0.231 1.025 0.140 11.32 4.83

The slope and intercept were used to obtain values for a and Rm
as in Example 19.7. To obtain e, the void fraction (porosity) of the
cake and the thickness of the cake, the following equations apply,
where W=A ¼ mass of dry cake/unit filter area, xc ¼ mass-fraction
solids in the cake, V=A ¼ volume filtrate/unit filter area, rf ¼ filtrate
density, rs ¼ the true density of the solids in the cake, L ¼ cake
thickness, and cF ¼ mass of dry cake/volume of filtrate.
ðW=AÞ ¼ cF ðV=AÞ (1)
ðW=AÞ ¼ rs 1 eavg L (2)
For a unit volume of cake, xc is given by
h i
xc ¼ rs 1 eavg = rs 1 eavg þ rf eavg (3)
up, the ability of the pump to develop pressure becomes the
limiting factor and the process continues at constant pressure
and a falling rate. For constant dV=dt, (19-23) becomes
V=Ac DPftg
¼ (19-27)
t m½Rm þ acF ðV=Ac Þ
Rearranging (19-27) after substituting u ¼ V=t, the super-
ficial velocity of the filtrate through the cake, the variation of
the pressure drop with time is
DP ¼ au2 t þ bu (19-28)

Equation (3) can be solved for the porosity, using xc ¼ 0.403, where a ¼ acF m=A2c (19-29)
rs ¼ 2.67(62.4) lb/ft3, and rf ¼ 62.4 lb/ft3:
b ¼ Rm m=Ac (19-30)
0:403 ¼ ð62:4Þð2:67Þ 1 eavg = ð62:4Þð2:67Þ 1 eavg þ 62:4eavg

which yields eavg ¼ 0.798. For a filtering time of 30 min, V=A ¼ Since for a constant rate of filtration, u must be constant,
0.639 ft. From (1), the corresponding W=A ¼ 0.640(0.639) ¼ 0.409 (19-28) defines a straight line on a plot of DP versus t, as was
lb=ft2. Solving (2) gives shown in Example 14.8.
W=A 0:409
L¼ ¼ ¼ 0:012 ft ¼ 0:146 in:
rs 1 eavg 62:4ð2:67Þð1 0:798Þ
§19.4.3 Variable-Rate Filtration
From the above table of calculated values for all runs, there is a
fairly strong variation of a with pressure, and some dependence of e. The most realistic, and in many respects the simplest,
Using the method of Example 19.7 with (19-26), a ¼ 3.50 1011 filtration scenario is when the flow and pressure both
(DP)0.3 and, similarly, (1 e) ¼ (0.15)(DP)0.13. vary, and the filtration rate varies in accordance with
There is no numerical or theoretical relationship that links e to a the pump characteristic curve provided by the pump
for compressible cakes. For incompressible cakes, where the manufacturer. Figure 19.25 shows such a curve. The fol-
Kozeny–Carman formulation is valid and when the cake particles lowing example, adapted from Svarovsky [23], demon-
are spherical of diameter dp:
strates the procedure.
ð1 eÞ
a ¼ 150
rs d 2p e3

The calculated cake thicknesses are small for a filtration time

of 30 minutes. For a rotary-drum vacuum filter with 30% sub-
2
mergence, the rate of rotation would be only 0.01 rpm, which is
too low. Either centrifugation or pressure filtration is required.
Another point to note is that the analytical equations for com- Pump outlet
pressibility and void fractions as a function of pressure should gage
not be extrapolated. pressure, 1
barg

§19.4.2 Constant-Rate Filtration

0 10 20 30 40 50
In a plate-and-frame filter or a pressure leaf filter, where cen-
3
trifugal pumps are used, the early stages of filtration are fre- Capacity, m /h
quently at a reasonably constant rate; then, as the cake builds Figure 19.25 Pump characteristic curve for Example 19.9.
800 Chapter 19 Mechanical Phase Separations

EXAMPLE 19.9 Filtration Time in a Plate-and-Frame

Filter. Q, m3/h DP 10 5, N/m2 V, m3 1/Q, s/m3
Tests conducted on a laboratory plate-and-frame filter produced the 45 0.2 0.72 80
data given below. Determine the time required to process 50 m3 of 40 0.75 12.71 90
the same filtrate in a filter press with an area of 50 m2, using the 35 1.15 24.53 103
same cloth and filter aid as the laboratory unit. 30 1.4 36.10 120
Data: rc ¼ 2,710 kg/m3; medium resistance, Rm ¼ 6.462 1010/m;
25 1.6 50.62 144
filtrate viscosity, m ¼ 2.78 10 7 N-h/m2 ; c F ¼10.037 kg/m3 ;
20 1.75 70.31 180
a ¼ 1.069 1011 m/kg.
15 1.8 97.55 240
Solution
Graphical integration gives 1.5 h.
The pump characteristic curve, Figure 19.25, shows the pump dis-
charge pressure, as a function of volumetric flow rate through the
pump, Q. Assume the flow rate is the filtrate, where Q ¼ dV=dt. The
time required for a volume of filtrate, V, is
Z §19.5 CENTRIFUGE DEVICES FOR
V
dV SOLID–LIQUID SEPARATIONS
t¼ (1)
0 Q
Centrifuge devices can greatly increase the rate of sedimenta-
which can be numerically integrated as follows. Assume the DP tion or filtration, particularly when particles are very small
across the filter medium and cake in bar ¼ the discharge pressure of (<10 mm), the liquid is very viscous, and/or the density dif-
the pump in barg. Rearrange the Darcy equation, (19-22), so that V
ference between particles and liquid is very small. They are
is a function of Q:
also favored over cake-filtration devices when the liquid is
Q ¼ Ac DP=m½Rm þ amcF ðV=Ac Þ (2) the main product. Two general types are: (1) sedimentation
centrifuges and (2) filtering centrifuges. Three common pro-
Solving for V,
duction centrifuges, shown schematically in Figure 19.27, are
Ac Ac DP (a) the tubular-bowl centrifuge, (b) the disc-stack centrifuge
V¼ mRm (3) with or without a nozzle discharge, and (c) the basket-
macF Q
filtering centrifuge.
Thus, using SI units, The tubular-bowl centrifuge, whose bowl is suspended
50 from an upper bearing and motor drive through a flexible-
V¼ 7 drive spindle, is widely used in the biochemical, food, and
2:78 10 1:069 1011 ð10:037Þ
pharmaceutical industries, particularly for the separation of
50ðDPÞ 7 (4)
2:78 10 6:462 1010 cells and viruses from broths. The slurry feed is introduced
Q
from the bottom through an orifice, followed by a distributor
¼ 0:00838½DP=Q 359
and baffle to accelerate the feed to the very high circumferen-
Using the pump characteristic curve of Figure 19.25, tabulate Q, tial speeds. The clarified liquid (centrate) exits from the top
DP, V from (4), and 1/Q starting from Q ¼ 45 m3/h and marching by overflowing a ring weir. Solids sediment moves to the
down in increments of 5 m3/h. Then plot 1/Q versus V until it just bowl wall. When its buildup begins to reduce the clarity of
exceeds 50 m3/h. From (1), by graphical integration, the area under the centrate, the process is stopped and solids are removed.
the curve from V ¼ 0 to 50 m3/h, as shown in Figure 19.26, is equal Bowl diameters, D, range from 1.75–5 inches. Corresponding
to the filtration time. rotation rates, v=2p, vary from 50,000–15,000 rpm, resulting
in centrifugal forces of up to 62,100 times the gravitational
force, as computed from v2D=2g. Typically, the solids are
less than 1% of the feed, and the centrifuge can handle no
300
more than 10 pounds of solids. Liquid throughputs are as low
as 0.05 gpm and as high as 20 gpm.
The disc-stack centrifuge is also vertically mounted. The
0
1/Q, feed enters at the top near the vertical axis, flows to the bot-
s/m3 tom, and is then accelerated by a radial vane assembly. It then
100
passes through a stack of 50–150 closely spaced (0.4–3 mm)
conical disks at an angle with the vertical of 35–50 . Solids
settle against the underside of the disks, from where they
move to the bowl wall. Liquid flows upward and exits
0 20 40 60 80 through overflow ports. If the fraction of solids in the feed is
V, m3 small, they remain in the centrifuge until the process is
Figure 19.26 Graphical integration for Example 19.9. stopped so that they can be manually removed. For larger
feed concentrations, the centrifuge can be fitted with an outer
 §19.5 Centrifuge Devices for Solid–Liquid Separations 801

Figure 19.27 Common solid–liquid

centrifuges: (a) tubular-bowl, (b) disc-stack,
(c) basket-filtering.

housing for periodic solids removal or with nozzles for con- distance from the tube center to the inner wall of the tube.
tinuous removal. For the latter type, bowl diameters range The feed enters at the bottom, passing up a length, L, and
from 10–30 inches with corresponding liquid throughputs of exiting at the top. Large-scale centrifuges of this type are
10–400 gpm and bowl speeds of 10,000–3,300 rpm. Corre- sized from laboratory data using a sigma scale-up theory,
sponding solids throughputs are 0.1–11 tons/h. developed next.
When the solids are the main product and high product For unhindered settling by centrifugal force that obeys
purity, low cake moisture content, and/or high solids recov- Stokes’ law, the terminal radial velocity is given by a modifi-
ery is desired, the basket-filtering centrifuge, operating cation of (19-7), where gravitational acceleration, g, is
batchwise in a cyclic operation that can include washing, replaced by centrifugal acceleration, Rv2 :
may be preferred. These devices have a large solids-holding
2 2
capacity, but are not suitable for soft biological solids. As dR Rv d p rp rf
discussed in great detail in [11], many designs are available, ut ¼ ¼ (19-31)
dt 18m
including both vertical (top or bottom unloading) and hori-
zontal units. The velocity of the liquid in the upward direction, z, is
Centrifugal separators are widely used in bioseparations
where density differences between particles and fluids are so dz Q Q
¼ ¼ (19-32)
small that gravity settlers are not effective. A typical applica- dt A p R20 R21
tion is separation of microorganisms from fermentation
broths. A common device is the tubular-bowl centrifuge, Dividing (19-31) by (19-32) gives an equation for the par-
where the solids move to the wall and form a sludge. Key ticle trajectory, dR=dz. Integrating it between R ¼ R1 to R0
variables for this centrifuge, shown in Figure 19.28, are rota- and z ¼ 0 to L, and multiplying and dividing by g, gives
tional speed, v in radians per unit time; R1, the distance from 2 3
2 3
2
the tube center to the surface of the sludge; and R2, the gd p rp rf 6pLv2 R2 R2 7 X
Q¼4 56
4
0 1 7
5 ¼ ð u t Þ g
18m R0
g ln
R1
(19-33)
wherePðut Þg is given by (19-7), Stokes’ law for gravity settling,
and A with units of length2 is a centrifuge sigma factor for
centrifuge A. For scale-up from a small laboratory centrifuge
(A) to a large production centrifuge (B), with different dimen-
sions and for operation at a different rotation rate, assuming
that ðut Þg remains the same, application of (19-33) gives
P
QB ¼ QA PB (19-34)
A

The assumptions in (19-34) and sigma theory are: (1) The

Figure 19.28 Tubular-bowl centrifuge.
particles are evenly distributed in the continuous liquid and
the concentrations are low, so settling is not hindered.
(2) Streamline flow at a Reynolds number below 0.2, with
the liquid rotating at the same velocity as the bowl. (3) No
re-entrainment, displacement of the flow pattern by the
deposited material, or nonuniform liquid feed.
802 Chapter 19 Mechanical Phase Separations

Despite these limitations, the theory has been shown to usually on the solutes. The symbol So will represent the ini-
work well for scale-up calculations for centrifuges of the tial concentration of solubles in the cake before washing.
same type [1, 20]. When wash liquid enters a homogeneous cake, it first dis-
Use of the sigma scale-up theory for a tubular-bowl centri- places solubles and liquid having the same composition as
fuge is illustrated in the following example, where the capac- the initial filtrate retained in the cake after filtration. If the
ity of a laboratory centrifuge is estimated and compared to wash cycle is ‘‘ideal,’’ the minimum volume of wash water
the measured value. In Exercise 19.19, a plant centrifuge is required is equal to the exact volume of liquid lodged in the
selected. Application of the sigma scale-up theory to a disc- voids of the cake. The wash liquid simply replaces a volume
stack centrifuge is considered in Exercise 19.20. equal to its own volume. Unfortunately, ideal, plug-flow
wash cycles do not exist because some washable material dif-
fuses out of the cake very slowly and, more importantly, the
EXAMPLE 19.10 Feed Capacity of a Tubular-Bowl wash liquid may carve channels in the cake, allowing pure
Centrifuge. wash liquid to break through the cake without displacing ini-
A laboratory tubular-bowl centrifuge has the following dimensions,
tial filtrate. At that point, the instantaneous concentration of
with respect to Figure 19.28, and operating conditions: bowl speed exiting wash is different from the average concentration
800 rps, R0 ¼ 0.875 inch, R1 ¼ 0.65 in., and bowl length ¼ L ¼ 4.5 remaining in the cake. Thus, three wash phases exist [1]:
inches. When used to remove E. coli cells from the following fer- (1) Liquid is displaced in the pores by wash liquid. Normally,
mentation broth, a satisfactory volumetric feed capacity of the cen- at least 50% of the material is removed. (2) A mixture of
trifuge, Q, of 0.11 gpm is achieved. breakthrough wash liquid appears along with the displaced
liquid. This phase ends when roughly 70–95% of the wash-
Broth: rf ¼ 1.01 g/cm3 and m ¼ 1.02 10 3
kg/m-s ables have been removed from the cake, at which point the
E. coli: smallest diameter ¼ dpmin ¼ 0.7 mm and rp ¼ 1.04 g/cm3 wash effluent equals about twice the cake void volume. (3)
Assuming the applicability of Stokes’ law, estimate the feed capac- The displacement has ended; solutes (or liquor) can be
ity of the centrifuge. removed only by diffusion. At this point, it is probably ad-
vantageous to repulp the cake and wash in additional coun-
Solution tercurrent or concurrent cascades, as described in §16.2.
To establish a wash cycle, it is preferable to conduct labo-
Compute the sigma factor for the laboratory centrifuge from (19-33)
ratory experiments to measure the instantaneous concentra-
using the given dimensions and operating conditions. The rotation
rate, v, in radians/s ¼ 2(3.14)(800) ¼ 5,030 s 1. Using AE units, tion of solutes in the wash effluent in terms of S=So, as a
2 3 function of the wash ratio W, defined as the volume of wash
X 6pLv2 R20 R21 7 liquid used/initial volume of filtrate in the cake. A typical
¼6
4
7
5 plot is shown as Curve 2 in Figure 19.29. Included is Curve 1
R0
g ln for the ideal, plug-flow wash cycle case. Note that S=So refers
R1
2 !3 to washable entities, which must be carefully defined and are
4:5 2 0:875 2 0:65 2
usually, but not always, just the solubles.
63:14 12 5030 12 12 7
6 7 It is useful to convert the experimental instantaneous
¼6
6
7 ¼ 7;400 ft2
7
4 0:875 5 effluent concentration plot to a plot of R, the fraction of ini-
32:2 ln
0:65 tial solutes still retained in the cake, as a function of W. By a
4
solute material balance, the total amount of solutes minus the
From (19-7), using CGS units with dp ¼ 0.7 10 cm and m ¼
1.02 10 2 g/cm-s,
2
981 0:7 10 4 ð1:04 1:01Þ 7
ut ¼ ¼ 7:85 10 cm/s
18 1:02 10 2
P
From (19-33), using SI units with ¼ 690 m2 and ut ¼ 7:85
10 9
m/s, Q ¼ 7:85 10 (690) ¼ 5.42 10 6 m3/s ¼ 0.0860
9

gpm, which is 78% of the measured rate.

§19.6 WASH CYCLES

Many filtrations and centrifugations are followed by a wash
cycle whose purpose is to recover solutes, filter aids and addi-
tives, and otherwise cleanse the cake, if the cake is the prod-
uct. Or the purpose may be to wash out the retained liquid in
the cake and add it to the filtrate, if the filtrate is the product.
The symbol S will be used to denote the instantaneous con- Figure 19.29 Instantaneous effluent concentration as a function of
centration in the effluent washed out of the cake, based wash ratio.
 §19.6 Wash Cycles 803

Z W
S
dW
W S=So 0 So R

0.00 1.000 0.000 1.000

0.10 1.000 0.100 0.902
0.20 1.000 0.200 0.803
0.30 1.000 0.300 0.705
0.40 1.000 0.400 0.606
0.50 1.000 0.500 0.508
0.60 0.998 0.600 0.409
0.70 0.979 0.699 0.312
0.80 0.897 0.793 0.220
0.90 0.725 0.874 0.140
1.00 0.500 0.935 0.080
Figure 19.30 Retention of solutes as a function of wash ratio. 1.10 0.295 0.975 0.040
1.20 0.151 0.997 0.018
1.30 0.068 1.008 0.008
amount of solutes in the effluent equals the solutes still 1.40 0.028 1.013 0.003
retained in the cake. Thus, 1.50 0.010 1.015 0.001
Z 1 Z W Z 1 1.60 0.004 1.015 0.000
S S S 1.70 0.001 1.015 0.000
dW dW ¼ R dW
0 S o 0 So 0 So 1.80 0.000 1.015 0.000
(19-35) 1.90 0.000 1.015 0.000
Rearranging (19-35), R as a function of W can be com- 2.00 0.000 1.015 0.000
puted by integrating, e.g., Curve 2 in Figure 19.29. The result
of that integration for Curves 1 and 2 is shown in Figure
19.30. What fractions of broth are recovered for wash ratios of 0.5, 1.0, 1.5,
Z W and 2.0? (c) What mass of wash water is required to recover 98% of
S the broth?
dW
S
RfW g ¼ 1 Z 0 1 o (19-36)
S
dW Solution
0 So
There are two approaches to designing wash cycles. The (a) One thousand kg of wet cake consists of 550 kg broth and 450 kg
dry biomass. Broth volume ¼ 550 kg/m3/1,050 kg/m3 ¼ 0.524 m3
first is to establish, by experiment, tables of data or plots like
and biomass volume ¼ 450 kg/1,150 kg/m3 ¼ 0.391 m3. Thus, the
Curve 2 in Figure 19.29, and then manipulate the data to find total volume is 0.915 m3 and volume fractions are biomass
the wash cycle that satisfies design objectives, which might 0.427 and broth ¼ 0.573. The wash water replaces the broth, so
be reducing the concentration of dissolved solutes to a certain the void volume eavg ¼ 0.573. The volume of water for W ¼ 1
level, displacing a certain amount of mother liquor with wash (0.573)(0.915 m3) ¼ 0.524 m3, or a wash water mass of 0.524
liquid, or, if the wash takes place on a rotary vacuum filter, (1,000) ¼ 524 kg.
determining the amount of wash liquid required to remove a (b) From (19-36), values of R can be computed by graphical inte-
stipulated amount of material in a given amount of time. gration for values of the wash ratio, W. Because the experimen-
Example 19.11 illustrates typical data manipulations. An tal data are closely spaced, reasonable accuracy is achieved
alternative is to devise general mathematical models for using the trapezoidal rule with a spreadsheet. For each interval
washing in much the same way as leaching and drying opera- in W, the arithmetic-average value of S=So is determined and
tions are modeled [7]. multiplied by DW. For example, in the interval of W from 0.80
to 0.90, the average value of S=So is (0.897 þ 0.725)=2 ¼ 0.811,
which, multiplied by DW ¼ 0.1, is 0.0811. In the above table,
EXAMPLE 19.11 Washing a Biomass Cake. the change in the integral for this interval in W is (0.874
0.793) ¼ 0.081. Values of the integral and R for a sequence of
Following fermentation, biomass is separated from the broth, often values of W are included in the above table. Equation (19-36)
by centrifugation. Then the wet biomass is washed to recover the requires the value of the integral for W ¼ 1. In the above table,
occluded broth, if the desired product is in the broth, e.g., if the bio- it is seen that by W ¼ 1.60, the integral is no longer changing; so
product is extracellular. A 1,000-kg biomass sample has been freed that value of 1.015 can be used.
of broth and found to weigh 450 kg. (a) What is the average porosity
(c) For recovery of 98% of the broth, the above table shows that a
of the biomass if the density of the broth is 1,050 kg/m3, the biomass
wash ratio of 1.2 is sufficient, or a mass of wash water ¼ 1.2
density is 1,150 kg/m3, and the wash liquid is water at 1,000 kg/m3?
(524) ¼ 629 kg. In this example, washing is very efficient.
(b) The following experimental wash data were obtained for S=So as
Often, it is not.
a function of W:
804 Chapter 19 Mechanical Phase Separations

§19.7 MECHANICAL SEPARATIONS cells can be disrupted differently. Gram-positive bacteria

IN BIOTECHNOLOGY have a cell wall about 0.3 mm thick composed of peptido-
glycan, teichoic acid, and polysaccharides, which is followed
Figure 19.31 is a schematic of the processing steps necessary
by a fragile membrane made of proteins and phospholipids.
to separate bioproducts obtained from plants and fermenta- The cell wall of gram-positive bacteria is susceptible to lysis
tion of bacteria, molds, and fungi from mammalian cells or
by the enzyme lysozyme, which degrades peptidoglycan.
by recombinant methods, which include insertion of DNA
Gram-negative bacteria are enveloped by multilayer mem-
into appropriate hosts. An introduction to these methods was
branes significantly thinner than the walls of gram-positive
given in §1.9. When the bioproduct is produced extracellu-
bacteria, and cannot be lysed. Osmotic shock (simply
larly, the biomass is separated from the broth by vacuum or
immersing a cell in distilled water) can be used to recover
pressure filtration, centrifugation, or by membranes (micro-
periplasmic proteins if the cell wall is breached or nonexist-
filtration or ultrafiltration). Expression, the de-liquoring of
ing. Yeast and mold cells have walls 0.1–0.2 mm thick, but
the biomass by compression, may be done if it is economi- mammalian cells do not have walls and are relatively fragile.
cally viable. The filtrate is then subject to an initial purifica-
In general, the fragile plasma membranes are readily destabi-
tion, which will include precipitation from solution or
lized by acids, alkali, detergents, or solvents.
methods described in previous chapters of this book. The
Cell debris are removed by centrifugation, microfiltration,
subsequent candidate purification and concentration opera-
or filtration under vacuum or pressure. The broth, which
tions are all described in previous chapters.
characteristically contains very low concentrations of the tar-
If the product resides intracellularly, the cells must first be
get species, then undergoes an initial purification to increase
harvested (separated from the broth). Then they are subject
the product concentration, to reduce the cost of subsequent
to cell disruption, a homogenization process wherein the cell
purification steps, and to prevent fouling of ion exchangers,
walls are breached so the product can be extracted. Intra-
adsorbents, chromatography columns, etc. Precipitation or
cellular products include recombinant insulin and growth
extraction are possible venues. Both the range of products
factors. A number of recombinant products form relatively
and the media in which they are produced are enormous, so
insoluble inclusion bodies; others, such as porcine insulin,
generalizations are difficult. Special attention must be paid to
need to be removed from pig pancreas. Different types of

Figure 19.31 Sequencing of bioseparations.


maintaining delicate protein structures. Damage by shear,
temperature, pH changes, contamination, and deactivation by
endogenous proteases are primary considerations, particularly
for products (and byproducts) that are biologically active.
Of the technologies listed in Figure 19.31, only cell dis-
ruption has not, as yet, been described in this book, and pre-
cipitation has received only brief mention. Interestingly
enough, of the six most commonly employed bioproduct sep-
aration operations, these two are the most commonly
employed. They appear on 80% and 40%, respectively, of
bioprocess flowsheets [11].
§19.7.1 Precipitation
The objective of the initial purification is to recover and con-
centrate the product, which can be in solution or in the pre-
cipitate. Precipitation can be induced by temperature, pH
adjustment, or addition of salts, solvents, polymers, or bio-
specific agents. The precipitates are generally amorphous
because of occlusions consisting of salts, solvents, or im-
purities, so accurate phase diagrams of the type seen in Chap-
ter 17 for crystals are not obtainable. Crystallization is a
special type of precipitation in which the product is crystal-
line and is produced slowly under very controlled conditions.
Nevertheless, in precipitation, the physical chemistry princi-
ples described in §17.11 and Chapter 1 apply.
Temperature
Temperature change, one way of precipitating crystals, is not
useful, by itself, for bioproducts. However, cooling, used in
conjunction with solvents and salts to maintain system stability,
is widespread. Temperatures below 0 C are not uncommon.
pH
The solubility of proteins depends on pH, with the minimum
generally at the isoelectric point. In theory, this can be the
basis of a protein-separation process; however, the differ-
ences in the isoelectric points of proteins are not large
enough for commercialization.
Salts
The effectiveness of salts in precipitating proteins follows the
Hofmeister series with ammonium sulfate, which is antichao-
tropic, being the most commonly used. The protein–salt aggre-
gates formed are shear-sensitive, so diaphragm pumps or gravity
feed are required. Downstream complications include salt
removal and disposal. The salting-out mechanism is complex: it
occurs partly because the salt removes water by associating with
water molecules, leaving fewer for the proteins, and partly by
shielding the electrostatic protein charges that account for pro-
tein–protein repulsion. An equation due to Cohn can be used to
predict the results of salt-induced protein separations [30]:
lnðSÞ ¼ B K s Cs (19-37)
where B ¼ the natural log of the solubility of the protein in
water, which depends on pH and temperature; Ks ¼ salting-out
§19.7 Mechanical Separations in Biotechnology 805
constant in m3/kmol; Cs ¼ salt concentration in kmol/m3; and
S ¼ protein remaining in solution in kmol/m3. Exercise 19.23
demonstrates use of this equation.
Organic Solvents
The most commonly used organic solvents, acetone and etha-
nol, can denature protein products, so low temperatures are
employed to mitigate protein denaturation. Precautions need
to be taken in mixing the solvents with aqueous solutions to
prevent regions of high, local solvent concentrations. The
precipitates formed with solvent addition are frequently very
fine powders, which are difficult to grow by aging, compared
to powders formed by salting-out and polyelectrolytes.
Nevertheless, organic solvents are widely used for RNA,
DNA, and plasma–protein precipitation. They lower the di-
electric constant of the solution, thus disrupting the electro-
lyte stability. The general equation that models solvent-based
precipitation of proteins [29] is:
lnðS=Sw Þ ¼ ðA=RTÞ½ð1=ew Þ ð1=eÞ (19-38)
where S ¼ solubility of protein in the medium (kmol/m3), Sw
solubility of protein in water (kmol/m3), A ¼ a constant, e ¼
dielectric constant of the medium, and ew ¼ dielectric con-
stant of water.
EXAMPLE 19.12 Precipitation of Ovalbumin.
The solubility of ovalbumin (OA) in water at 20 C is 42 mg=mL.
Thirty percent of the ovalbumin precipitates when 25 mL of ethanol
is added to 100 mL of a 40 mg=mL aqueous solution of ovalbumin.
How much ethanol would have to be added to the original aqueous
solution to precipitate 90% of the ovalbumin at the same tempera-
ture? Assume no volume of mixing, and that the dielectric constant
is linear with volumetric composition. At 25 C, the dielectric con-
stants for water and ethanol are 78.4 and 24.4, respectively.
Solution
The amount of OA in the aqueous solution ¼ 42(100) ¼ 4,200 mg.
For the addition of 25 mL of ethanol, 30% of the OA is precipitated,
leaving 70%, or (0.70)(4,200) ¼ 2,940 mg of OA left in a solution of
25 þ 100 ¼ 125 mg. Thus, the solubility of OA in this solution is
2,940=125 ¼ 23.5 mg/mL. The dielectric constant of this solution of
25=125, or 20%, ethanol is 24.4 þ (1 0.20)(78.4 24.4) ¼ 67.6.
Substituting these values into (19-38) gives
lnð23:5=42Þ ¼ ðA=RT Þ½ð1=78:4Þ ð1=67:6Þ
Solving, A=RT ¼ 285. Let x ¼ the volume in mL of added ethanol
needed to precipitate 90% of the OA. The amount of OA that would
be left in the solution ¼ (1 0.90)(4,200) ¼ 420 mg. The solubility
of OA in the resulting solution would be 420=(100 þ x). The di-
electric constant of the resulting solution would be 24.4 þ [x=
(100 þ x)](78.4 24.4). Substituting these values into (19-38) gives
420
ln 42 ¼
ð100 þ xÞ
285½ð1=78:4Þ ð1=f24:4 þ ½1 x=ð100 þ xÞ ð78:4 24:4ÞgÞ
Solving, the volume of ethanol required ¼ x ¼ 257 mL. This is a
very large amount, and thus this method may not be practical to pre-
cipitate the OA.
806 Chapter 19 Mechanical Phase Separations
Water-soluble powders of the type used in flocculation,
which were introduced in §19.2, can be used to precipitate
proteins. In the research stage are affinity precipitants, where
a conformal ligand attached to the polymer can couple with a
target protein to further enhance aggregation. Here, as with
the other precipitation processes, pH is important since pro-
teins exhibit their lowest solubility at the isoelectric point.
§19.7.2 Coagulation, Flocculation, Clarification,
and Sedimentation
A precise lexicographic definition of these processes is not
possible because they may be proceeding simultaneously and
be viewed functionally. Sedimentation, in Perry’s Chemical
Engineers’ Handbook [11], is defined as ‘‘the partial separa-
tion or concentration of suspended solid particles from a liq-
uid by gravity settling. This process may be divided into the
functional operations of thickening and clarification. The
purpose of thickening is to increase the concentration of sus-
pended solids while that of clarification is to produce a clear
effluent.’’ In all aspects but one, clarifiers and thickeners are
identical. The one difference is that clarifiers are usually ligh-
ter in construction because the average density and viscosity
are lower, because the suspended solid concentration is
lower. This makes the definition function specific.
Small particles dispersed in a suspension are stabilized by
forces due to the surface charges of the particles, which is
why they do not agglomerate spontaneously due to Brownian
motion. Bacterial cells and most solids suspended in water
possess negative charges at neutral pH. The source of the sur-
face charges is the surface groups, which are capable of ion-
ization. A second source of surface charge is the preferential
adsorption of ions in the solution.
The physical process of sedimentation is enhanced by
coagulation and flocculation, which may occur sequentially,
as in Figure 19.32, but often occur simultaneously, as do pre-
cipitation and agglomeration of proteins if a polyelectrolyte
is present when the temperature of a saturated solution of
proteins is lowered. Flocculation is thus defined as the further
agglomeration of the small, slowly settling floc formed dur-
ing coagulation to form a larger aggregated floc particle. The
relative sizes of suspended particles encountered in biologi-
cal systems are shown in Table 19.10.
Table 19.10 Relative Sizes of Suspended
Particles
Class Diameter, mm
Colloids 0.0000001–0.001
Dispersed 0.001–0.1
Coagulated 0.1–1.0
Flocculated 1.0–10
Organic particles below the size visible to the human eye,
approximately 0.04 mm (40 microns), generally have settling
times that are unreasonably long, and thus coagulation and
flocculation, as well as mild agitation, are required to achieve
economically sized equipment.
Table 19.11 provides a list of inorganic and organic coag-
ulants as well as some coagulant/flocculant aids, which are
used, in part, because they shorten settling times by increas-
ing the density of the suspended microorganisms. These are
hydrophyllic and associated with both internal and surface-
bound water, so their density is very close to that of the broth.
Inorganic coagulants are water-soluble inorganic acids,
bases, or salts that, when dissolved, produce cations or
hydrolyzed cations. Increasing the concentration of salt com-
presses the electrical double layer surrounding a suspended
particle and decreases the repulsive interaction between par-
ticles, thus destabilizing them. In flocculation there is further
agglomeration by an organic polyelectrolyte. One end of a
flocculant molecule attaches itself to the surface of one parti-
cle at one or more adsorption sites, and the other extended,
unadsorbed end of the same molecule bridges and adsorbs to
one or more additional particles, thus forming a larger aggre-
gate of floc particles.
The coagulant/flocculant aids in Table 19.11 are insoluble
particulates generally used to enhance solid–liquid separa-
tions, where slime and glue-like interactions are troublesome.
It is known, for example, that broth cultures of actinomy-
cetes, such as Streptomyces greisius, are difficult to filter or
settle and require the addition of about 2–3% diatomaceous
filter aid to form a satisfactory cake. Usually, large quantities
of these filter aids are required, and this raises the need of
recovery or waste-disposal processes. In general, filtration of
biosystems is difficult and centrifugation is preferred.
Figure 19.32 Coagulation,
flocculation, sedimentation sequence.
 §19.7 Mechanical Separations in Biotechnology 807

Table 19.11 Coagulants/Flocculants

Polyelectrolyte Coagulants
Inorganic Coagulants Type Charge Examples Coagulant/Flocculant Aids

Acids
Hydrochloric Anionic Negative Polyacrylamide Activated carbon
Sulfuric Polyacrylic acid Activated silica
Bases Polyacrylate Bentonite
Calcium hydroxide Polystyrene sulfonate Clay
Sodium hydroxide Cationic Positive Polyalkylene polyamine Metal oxides
Salts Polyepichlorohydrin Paper pulp
Aluminum chloride Polyethylenimine
Aluminum sulfate Polyaminoethyl
Ammonium sulfate polyacrylamide
Calcium chloride Polyvinylbenzyl trimethyl
Ferrous chloride ammonium chloride
Ferric chloride Polydimethyl diallyl
Ferric sulfate ammonium chloride
Ferrous sulfate Nonionic Neutral Polyacrylamide (1–5%) hydrolyzed
Polyethyleneoxide
Miscellaneous Albinic acid, dextran, guar gum, starch

Destabilization of suspended biological particles is rela-
tively difficult compared to the problems seen in chemical
plants and wastewater-treatment facilities. Very frequently,
pretreatment is required prior to centrifugation or filtra-
tion. The salt content, pH, and temperature of the system
affect the surface charges of the suspended solids. The
signs, magnitudes, and distribution of the surface charges
influence the type and quantity of coagulant/flocculant
used. Negatively charged moieties are flocculated by cat-
ionic flocculants. Negatively charged solids may also be
coagulated by inorganic ions and flocculated by anionic
flocculants. ‘‘Whatever it takes’’ is an appropriate cliche.
Natural systems are complex and laboratory studies are a
necessity.
§19.7.3 Cell Disruption
If the target protein resides intracellularly, cell walls need
to be breached, and the microorganism homogenized prior
to extracting the product. Bacteria, molds, and yeast are
only about 1 mm in diameter. Proteins may be present as
soluble or folded, or as insoluble, misfolded inclusion
products, where all cysteine amino acid residues are fully
reduced, making homogenization a considerable challenge.
Inclusion aggregates need to be recovered (usually by cen-
trifugation because they are denser than the cell debris),
dissolved, and then re-suspended, and the proteins folded
to restore biological activity. The substrate is shear-, tem-
perature-, and pH-sensitive, and mild conditions and buff-
ering are mandated.
The two prevalent approaches to cell disruption are me-
chanical and chemical. The chemical methods are costly,
and removal of additives is always an issue. Enzymes
such as lysozyme, which is effective for lysing gram-
positive bacteria, is less effective for gram-negative bacte-
ria. Other options are detergents like TritonTM X-100, the
chaotropes urea and guanidine hydrochloride, or the che-
lating agent ethylenediaminetetraacetic acid (EDTA),
which extracts divalent ions from the cell wall. All of
these chemical methods are expensive, and are accompa-
nied by protein denaturation. Other laboratory methods
include heat-shocking, freeze-drying, and changes in
osmotic pressure induced by adding salts or immersing
cells in distilled water. Osmotic shock is effective for
mammalian cells, or gram-negative bacterial or fungal
cells, whose walls have been weakened by methods such
as immersion in an isotonic fluid.
Mechanical means are more attractive from an industrial
standpoint. Ultrasound generators, producing pressure waves
at about 20 kcycles/s by piezoelectric, titanium transducers,
are effective in small vessels. In larger vessels, field
inhomogeneity, heat dissipation, cell fragmentation, and
enzyme denaturation make the method less suitable for
large-scale application.
Industrial-sized equipment includes ball mills wherein
cells are mixed with as many as 80%, by volume, of 20–50
mesh beads and then passed through discs rotating at high
speed. The cell walls are broken by shear and impact. A unit
of this type, called the Dyno-Mill, is capable of processing
more than 4,000 lb/h. Even higher capacities, up to 40,000
lb/h, are achieved by pressure-based homogenizers, where
cells suspended in an aqueous medium are forced to flow
through narrow, adjustable gaps at high speed and pressures
up to 50,000 psig. Rotor-stator devices are also on the
market.
Large amounts of heat and energy are involved in cell
disruption, which is analogous to particle-size reduction, a
widely practiced art in chemical manufacture. Prediction
808 Chapter 19 Mechanical Phase Separations
of power consumption for these so-called crushing and
grinding operations is correlated by Kick’s law, which is
based on particle-radius-size reduction, or Rittinger’s law
of grinding, where the energy is correlated to the change
in surface area. Ghosh [29] suggests that both of these
laws are suited for tissue grinding, but are not useful for
modeling cell disruption, and instead suggests an equation
of the form
n
C=Cmax ¼ ½1 expð t=uÞ (19-39)
3
where C ¼ concentration of released product (kg/m ), Cmax
maximum concentration of released material (kg/m3), t ¼
time (s), u ¼ time constant (s), and n ¼ number of passes
through the device.
EXAMPLE 19.13 Ultrasonification of a Cell
Suspension.
A cell suspension has an intracellular antibiotic concentration of 10
mg/mL. If, at the end of 5 minutes of ultrasonification, 20% of the
antibiotic is released, how long will it take to release 95% of the
antibiotic?
SUMMARY
1. Particle size can be used as a criterion for an initial
appraisal of what type of mechanical separation device
is suitable.
2. Except for devices employing force fields other than
gravity, equations for particle settling velocities are the
basis for many of the mathematical models used to
design industrial separators.
3. The major settling laws, including Newton’s and
Stokes’, can be modified to include centrifugal forces
and hindered particle settling. Equations used to design
centrifuges, cyclones, and filtering devices generally
include particle terminal velocity as a variable.
4. Particle-capture devices can be designed on the basis of
efficiency and pressure-drop data provided by manufac-
turers. Many standards and test procedures are regulated
by trade organizations and government agencies.
5. Pressure filtration using plate-and-frame, leaf, or rotary
vacuum drum filtration devices is suitable for separating
solids approximately 10–50 mm in size and in concentra-
tions of about 1–35 vol% from liquids. At much higher
concentrations and particle sizes, settling devices may
be more economical; at lower concentrations, particle-
size sieving should be considered.
6. The model used for design of plate-and-frame or vacuum
solid–liquid filtration devices assumes the filtration rate
is proportional to the pressure gradient and inversely
proportional to filtrate viscosity. The model can be used
for constant-pressure and variable- or constant-rate fil-
tration, depending on pump characteristic curves.
Solution
First compute the time constant using the given data. Using (19-39)
with n ¼ 1,
0:20ð10Þ
¼ ½1 expð 300=uÞ 1
10
Solving, u ¼ 1,345 s. Then, for 95% release,
0:95ð10Þ
¼ ½1 expð t=1345Þ 1
10
Solving, t ¼ 4,028 s ¼ 67 minutes.
Product quality and yield are key considerations for all
cell-disruption schemes. If there are six processing steps in
the product-recovery and purification scenario, and the yields
in each step are 90%, then 47% of the valuable product is
lost. In cell disruption, losses from the release of proteases
can lead to enzymatic degradation, and loss of product from
heat or oxygen degradation needs to be avoided. Therefore,
reducing and chelating agents as well as buffers are often
used, particularly if residence times are long. Some labora-
tory devices disrupt the cells by freezing, or apply mechani-
cal shear to partially frozen paste to avoid loss of biological
activity.
7. An alternative model for filtration is based on the
Hagen–Poiseuille formulation for pressure drop through
a packed bed. Here, void fractions and cake flow paths
must be identified.
8. The Ruth equation, used to model solid–liquid filtration,
has three constants and predicts a straight-line relation-
ship between filtrate volume and time, divided by filtrate
volume for a constant-pressure filtration.
9. Two of the three constants in the Ruth equation can be
eliminated if the filter cake is incompressible and if the
filter medium pressure drop is insignificant compared to
the cake pressure drop.
10. The diverse and unpredictable nature of filter-cake com-
pressibility has hindered development of generalized
correlations and mathematical models.
11. Centrifuge devices can greatly increase sedimentation
and filtering rates. A commonly used design scale-up
method is the sigma method.
12. Squeezing liquid out of a filter cake is called expression.
13. Washing is done to remove undesirable solutes or to
recover occluded product.
14. Precipitation can be accomplished by addition of salts,
solvents, polymers, and surface-active agents, aided by
temperature and pH changes.
15. When the bioproduct of a fermentation resides intra-
cellularly, the cell must be disrupted so that the bioprod-
uct can be recovered.

REFERENCES
1. Tiller, F.M., Theory and Practice of Solid-Liquid Separation, Chemical
Engineering Department, University of Houston (1978).
2. Shuler, M.L., and F. Kargi, Bioprocess Engineering, Prentice Hall PTR,
Upper Saddle River, NJ (2002).
3. Souders, M., and G.G. Brown, Ind. Eng. Chem., 26(1), 96 (1934).
4. Sutherland, K., Filters and Filtration Handbook, 5th ed. Buttersworth-
Heinemann, Burlington, MA (2008).
5. Towler, G., and R. Sinott, Chemical Engineering Design, Elsevier, Bur-
lington, MA (2008).
6. Nonhebel, G., Processes for Air Pollution Control, Butterworth & Co.,
Cleveland, OH (1972).
7. Wakeman, R.J., and E.S. Tarleton, Filtration, Elsevier Science, New
York (1999).
8. Foust, A.S., L.A. Wenzel, C.W. Clump, L. Maus, and L.B. Anderson,
Principles of Unit Operations, J. Wiley & Sons, New York (1960).
9. Amistco Corporation, Alvin, Texas.
10. Stairmand, C.J., Trans. Inst. Chem. Eng., 29, 356 (1951).
11. Perry’s Chemical Engineers’ Handbook, 8th ed., D.W. Green and R.H.
Perry, Eds., McGraw-Hill, New York (2008)
12. Coker, A.K., Chapter 6, ‘‘Mechanical Separations,’’ in Ludwig’s Ap-
plied Process Design for Chemical and Petroleum Plants, 4th ed. Vol. 1,
Elsevier Publishing, New York (2007).
13. Einstein, A., Ann Physik, 17(4), 549 (1905).
14. Brink, J., Can. J. Chem. Eng., 41, 134 (1963).
15. McCabe, W.L., J.C. Smith, and P. Harriott, Unit Operations of Chemi-
cal Engineering, 4th ed. McGraw-Hill Book Co., New York (1985).
STUDY QUESTIONS
19.1. Why is particle size the main parameter used in selecting a
mechanical phase-separation device?
19.2. At the particle settling velocity, what force balances the
drag force plus the buoyant force?
19.3. Into what four regions are settling equations for particles
divided?
19.4. What form of the Souders–Brown equation is used to cor-
relate empirical settling data?
19.5. What criteria have been developed for deciding which set-
tling equation is applicable for a given particle diameter?
19.6. How are settling velocity equations modified to take into
account particle–particle collisions and particle-shape differences?
19.7. What empirical equations, with constants obtained from
experimental data, are frequently used to design many particle-fluid
separation devices?
19.8. Why do governmental regulatory agencies and trade orga-
nizations set many design and performance specifications for parti-
cle emissions?
19.9. Why is centrifugal force frequently applied to speed up
and facilitate particle-fluid separation?
19.10. Why have theoretical analyses that treat voids in filter
cakes as flow channels not been applied industrially?
19.11. In a filtration cycle, why does constant-pressure filtration
usually occur near the end of the cycle and constant-rate filtration at
the beginning?
Study Questions 809
16. Carpenter, C.R., Chem. Eng., 90(23) 227–231 (1983).
17. Kula, M.R., K.H. Kroner, and H. Hustedt, Advances in Biochemical
Engineering, 24, 73 (1984).
18. Schweitzer, P.A., Handbook of Separation Techniques for Chemical
Engineers, McGraw-Hill Book Co., New York (1979).
19. Peters, S.M., K.D. Timmerhaus, and R.E. West, Plant Design and Eco-
nomics for Chemical Engineers, 5th ed. McGraw-Hill, New York (2003).
20. Ruth, B.F., G.H. Montillion, and R.E. Montonna, Ind. Eng. Chem., 25,
76, 153 (1933).
21. Tiller, F.M., Chem. Eng., 73, (13) 151 (1966).
22. Silla, H., Chemical Process Engineering, Marcel Dekker Inc., New
York (2003).
23. Svarovsky, L., Solid-Liquid Filtration, 3rd ed. Butterworths, London
(1990).
24. Chopey, N., Handbook of Chemical Engineering Calculations, 3rd ed.
McGraw-Hill Book Co., New York (2003).
25. Aiba, S., A.E. Humphrey, and N.F. Mills, Biochemical Engineering,
Academic Press, New York (1965).
26. Blasewitz, A.G., and B.F. Judson, ‘‘Filtration of Radioactive Aerosols
by Glass Fibers,’’ Chem. Eng. Progress, 51(1), 6 (1955).
27. Stairmand, C.J., Trans. Inst. of Chem. Engrs., 28, 131 (1950).
28. Walas, S.M., Chemical Process Equipment, Butterworths, Boston
(1988).
29. Ghosh, R., Principles of Bioseparations Engineering, World Scientific
Publishing Co., Hackensack, NJ (2006).
19.12. For what particle-size and particle-concentration ranges
are vacuum rotary-drum, leaf, and plate-and-frame filters generally
used?
19.13. For what assumptions do filtration data plot as a straight
line for V versus t=V coordinates?
19.14. Why are precoat and filter aids generally used in solid–
liquid plate-and-frame or vacuum rotary-drum filtrations?
19.15. Why are wash periods followed by expression often part of
the filtering cycle?
19.16. What are the assumptions in the Ruth equation for
filtration?
19.17. How are empirical constants in filtration models
determined?
19.18. Why are pump characteristic curves important in pressure
filtration?
19.19. What is the sigma theory and how is it applied?
19.20. How do processes for separating extracellular and intra-
cellular bioproducts differ?
19.21. What steps can be taken to speed coagulation of particu-
lates from bioreactors?
19.22. How are washing cycles determined?
19.23. Name five methods for cell disruption.
810 Chapter 19 Mechanical Phase Separations

EXERCISES
Section 10.3
19.1. Particle settling velocity.
For a solid, spherical particle of 0.8 mm in diameter and a den- Inlet dust
sity of 2,600 kg/m3 that is immersed in a fluid of density 1,200 kg/ particle-size dp, average Grains/ Weight %
m3 and a viscosity of 1.0 cP, calculate (a) the unhindered terminal range, mm diameter, mm std ft3 of particles
velocity in m/s, and (b) the hindered terminal velocity if the volume
fraction of such particles is 0.05. 0–20 10 0.0062 2.7
19.2. Particle velocity prior to terminal velocity. 20–30 25 0.0159 6.9
Consider a spherical particle 4 microns in diameter with a 30–40 35 0.0216 9.4
density of 3,000 kg/cm3 falling through water of 0.001 N-s/m 2 40–50 45 0.0242 10.5
viscosity. At 10 5 s, what is the particle velocity? Confirm that 50–60 55 0.0242 10.5
the particle is in the Stokes’ law range. Assume the particle is
60–70 65 0.0218 9.5
at its terminal velocity. Hint: An inertial term must be added
to (6-40). 70–80 75 0.0161 7.0
19.3. Settling velocity of microorganisms. 80–94 87 0.0218 9.5
For separations by settling and centrifugation of bacteria, yeast, 94þ 94+ 0.0782 34.0
fungi, and mixed-culture activated sludge from fermentations and
sewage systems, values of cell density, equivalent diameter, settling Adapted from [24].
velocity, and volume fraction of cells in suspension must be esti-
mated. The data below and procedures used to obtain these values
are described in [23].

Volume ut, cm/s, rp, rf, m, g/cm-s, gZ, dp,

Species Fraction 10 4 g/cm3 g/cm3 10 3 cm/s2 microns

Yeast 0.041 0.83 1.09 1.00 11.3 981 5.5

3
Bacteria 0.31 10.13 1.03 1.00 8.3 981 10 1.0
Fungi 0.353 1.92 1.003 1.002 9.7 981 143
Sludge 0.279 4.09 1.013 1.00 12.5 981 (71)

In the above table, gZ is the gravitational constant times the cen-
trifugal field strength. The bacteria-settling study was conducted in
centrifuges, while the others were done in simple gravity settlers.
The particle size was observed microscopically, except for the
sludge, which was calculated from the settling velocity using
Stokes’ law. The authors claim good correlations between micro-
organism diameters calculated from settling velocities and those
determined microscopically. Verify their claim by calculating the
diameters from the settling velocities, including a verification of
their calculated settling velocity for the sludge. Discuss your results
in terms of possible hindered settling, aggregation, and particle-
shape properties.
19.4. Efficiency of a settling chamber.
A settling chamber at a coal-burning installation has the fol-
lowing chamber dimensions, operating conditions, and inlet parti-
cle-size distribution, where 1 pound ¼ 7,000 grains. Determine
the collection efficiency for each particle-size range, and the
overall particle-collection efficiency. Assume the settling veloc-
ity, ut, is one-half of that computed by Stokes’ law.
Chamber width ¼ 3.29 m, chamber height ¼ 0.75 m, and cham-
ber length ¼ 4.57 m.
Volumetric gas feed rate ¼ 70.6 ft3/s at std. conditions of 32 F
and 1 atm.
Actual gas temperature and pressure is 446 F and 1 atm.
Actual gas viscosity ¼ 2.60 10 5 N-s/m2 , and gas density
is negligible compared to particle density, which is
2.65 g/cm3.
19.5. Diameter of demister pad.
A 6-inch demister pad is to be used to separate liquid entrain-
ment from a gas in a horizontal flash drum. Calculate the demister
diameter using the following data:
Vapor flow rate ¼ 465 cfm at 110 F and 50 psia; vapor density
0.30 lb/ft3; and liquid density ¼ 33 lb/ft3, K ¼ 0.35.
19.6. Aerosol filtration by glass fiber mats.
An ambient air stream at a superficial velocity of 9.84 ft/min, con-
taining an aerosol of 1-mm-diameter particles at an estimated concen-
tration of 104 particles/m3, is to be filtered through a bed of glass
fibers, 19 mm in diameter, with a bed void fraction, eb, of 0.0033. An
empirical equation for the particle-collection efficiency as a function
of bed thickness for glass fibers of this type is given by [26]:
0:075L0:9 rb us 0:4
Fractional efficiency ¼ 1 10 (1)
where L ¼ bed depth in inches, rb ¼ bed bulk density ¼ 5.15 lb/ft3,
and us ¼ superficial air velocity in ft/min. The pressure drop across
the filter per unit bed depth is given approximately in terms of a
modified drag coefficient as
DP 2ru2s
¼ CD (2)
L pgc d p
where r ¼ air density ¼ 1.2 10 3 g/cm3, dp ¼ glass fiber diameter
19 10 4 cm, and the drag coefficient is given by CD ¼ 50=N Re,
where NRe ¼ dprus=m and m ¼ air viscosity ¼ 1.8 10 4 g/cm-s.
Calculate: (a) bed depths in cm for a series of particle-collection
efficiencies over a range of 90 to 99.99%; and (b) pressure drop per
unit bed depth in kg/m2-m. Adapted from [25].
 Exercises 811

100
90
80
70
ES
60 RV
CU
50 NT
ME
40 N GE
PI
30 IM
TARGET EFFICIENCY%

7µ 5µ 4µ 3µ 2.5µ 2µ 1.5µ 1µ 0.5µ

20

10
9
8
7
6
5 0. 5
µ
4 1µ

3 2µ
3
4µµ
2 5µ
7µ

Figure 19.33 Diffusion and impingement target

1
.01 0.1 1.0 10.0 efficiencies for Exercise 19.9. Adapted from Stairmand
GAS VELOCITY (FT./SEC) [27]. In the graph, (TS, symbol for Greek mu), is microns.

19.7. Particle settling. and dust density of this exercise. If each particle collision results
Aerosol particles 0.001 inch in diameter with specific grav- in capture, the overall % efficiency, ho, for each particle size is
ity 0.8 are to be settled from an air stream. If the settling in terms of the fractional single-layer target efficiency of Figure
chamber is 2 ft deep, what should the residence time be? At 19.33, given by
the process conditions, rf ¼ 0.08 lb/ft 3 , rp ¼ (0.8)62.4 ¼ 49.92
lb/ft 3 , and m (air) ¼ 0.02 cP. ho ¼ 100½1 expð ht So Þ
19.8. Design a vertical decanter. where So is the number of layers or stages of filter. Assume the
Use the design method suggested by Towler and Sinott [5] to 1-mm-thick filter cloth has 50 stages.
design a vertical decanter to separate oil droplets suspended in
19.10. Cyclone rating.
water (the continuous phase). The design method suggests that
A cyclone of diameter Dc of 2 ft, whose dimension ratios are as
an oil-droplet particle size, dp, of 150 mm be assumed; that the
given in Figure 19.9, is being considered to remove dust from a
height of the decanter be increased by 10% to allow for a disper-
cement kiln. The gas feed, at inlet velocity ui of 20 ft3/s, has
sion band at the oil/water interface (where the feed enters); that
0.5 grain/ft3 of particles of average diameter dp of 7 mm and density
the velocity of the continuous phase, uc, be less than the settling
rp of 175 lb/ft3, where 1 lb ¼ 7,000 grains. At the operating condi-
velocity of the oil droplets, us; and that the decanter length be
tions, air viscosity is 1.21 10 5 lb/ft-s, and the air density is negli-
twice the diameter. Check the dimensions of your decanter to
gible compared to the particle density. The local air-pollution
make certain that the residence time is reasonable (2–5 minutes),
authority requires an effluent of less than 0.1 grain/ft3. Can this be
and for the maximum size of water particles that may be carried
achieved with the present cyclone? This can be determined by com-
out by the oil phase.
puting the fractional collection efficiency for the 7 mm particles from
For the oil, flow rate ¼ 1,000 kg/h, density ¼ 900 kg/m3, and h i
2 1
viscosity ¼ 3 cP. E ¼ 1 þ Dpcrit =d p (1)
For the water, flow rate ¼ 5,000 kg/h, density ¼ 1,000 kg/m3, and
viscosity ¼ 1 cP. where Dpcrit ¼ the diameter of the smallest particle that is theoreti-
cally separated from the gas stream at a collection efficiency of
19.9. Collection efficiency of a bag filter. 50%. From [28],
Calculate, using single-fiber collection efficiencies, the over- 2 312
all collection efficiency of a bag filter comprised of glass fibers
9mD c
of mean diameter 10 mm in the form of a cloth 1-mm thick, Dpcrit ¼ 4 5 (2)
taking into account both diffusion (Brownian motion) and 4pN t ui rp rf
impingement. The gas contains 1 g/m3 of dust of specific grav-
ity 2 g/cm 3 and has a velocity of 0.1 ft/s. The particle-size where Nt, the number of rotations made by the gas stream in the
distribution is as follows: cyclone, is given by
6
N t ¼ ui 0:1079 0:00077ui þ 1:924 10 u2i (3)
Dust particle size, mm 3–5 2–3 1–2 <1
with ui in ft/s.
Mass fraction of total dust 0.45 0.20 0.20 0.15
19.11. Scale-up of test leaf filter data.
A test leaf filter is used to determine the filtration rate for an
Under these conditions, impaction does not occur and re- unclarified broth from a fermentor. At a pressure drop of 50 kPa,
entrainment can be neglected. Figure 19.33 gives the single-layer 150 mL of filtrate are collected in 30 minutes. If the resistance of
target collection efficiencies, ht, as a function of gas velocity for the filter cloth is negligible, will 300 mL of filtrate be collected in
diffusion (straight lines) and impingement (curved lines) of dust 30 minutes if the pressure drop is doubled? If not, how many mL of
particles ranging in size from 0.5 to 7 mm for the fiber diameter filtrate will be collected?
812 Chapter 19 Mechanical Phase Separations

Section 19.4 1.5 4,043 0.32 17,850

19.12. Constant-pressure filtration. 1.5 4,398 0.34 17,800
A slurry is being filtered in a plate-and-frame filter press at a 1.5 4,793 0.36 18,450
constant pressure drop of 10 psi under conditions where the resist- 1.5 5,190 0.38 18,800
ance of the medium is negligible. The fluid viscosity is 2.0 lb/hr-ft; 5,652 19,660
1.5 0.40
5 lb of dry cake are collected for every ft3 of filtrate; and the cake
1.5 6,117 0.42 20,260
is incompressible, with a specific cake resistance, a, of 8.34 1010
ft/lbm. What filter area is required to process 50 ft3 in one hour? 1.5 6,610 0.44 20,890
19.13. Scale-up for a rotary vacuum filter. 1.5 7,100 0.46 21,340
A pilot-plant rotary-drum vacuum filter, with 10 ft2 of filter area 1.5 7,608 0.48 21,790
in tests involving a new product, delivers 9 ft3/minute (cfm) of fil- 1.5 8,136 0.50 22,250
trate at a total pressure differential of 15 psi. The drum rotates at 2.5 1.5 8,680 0.52 22,700
rpm and 25% of the drum is submerged in the feed reservoir. The
1.5 9,256 0.54 23,210
pilot-plant data are to be used to design a large plant unit, made by
the same manufacturer, to deliver 100 cfm of filtrate. The new unit is
also expected to run at a DP of 15 psi, but with an rpm of 2.0 and a
submergence of 30%. The cake compressibility, filtrate viscosity, as The filter pump was controlled manually until the pressure
well as the medium and cake resistance in the pilot and plant units became constant at 150,000 N/m 2, which occurred at approximately
are expected to be very much alike. (a) What should be the drum V ¼ 0.3 m 3, t ¼ 3,686 s. Make a plot of the data, t/V versus V, and
area for the plant unit? (b) Your boss has asked you to recommend determine the specific cake resistance, a, and the medium resist-
desirable and future actions to be taken if production rates have to ance, Rm.
be increased by 50–400%. 19.16. Filtration at constant rate, followed by constant
19.14. Scale-up of filtration data. pressure.
A filtration system for a fermentation effluent is being designed A slurry is passed through a filter cloth 0.02 m2 in area at a con-
on the basis of the data taken using a pilot-plant plate-and-frame stant rate, with 4 10 5 m3/s of filtrate being collected. After 100 s,
filter press. The filter area is 0.3 m 2, cF ¼ 2,000 kg/m3, DP ¼ 2.5 the pressure is 4 104 N/m2; after 500 s, it is 1.2 105 N/m2.
10 4 N/m2, and filtrate m ¼ 3 10 3 kg/m-s. The liters of filtrate, The same filter-cloth material will be used in a plate-and-frame
V, collected during 120 minutes were: filter press having an area of 0.5 m2, which can hold a cake of thick-
ness, L, of 0.04 m, and which will process the same slurry. Assume
the filtration will be at a constant rate, and then at a constant pres-
V,L 120 360 700 840 1,100
sure after the pressure reaches 8 104 N/m2. If the volume of cake
t , min 5 20 45 75 120 per volume of filtrate is 0.02, calculate the time required to fill the
filter frame.
(a) Obtain values for Vo and K in (19-23). (b) Calculate a, the 19.17. Area of a rotary-drum vacuum filter.
specific cake resistance, in m/kg. (c) If the anticipated plant plate- Determine the surface area of a rotary-drum filter that processes
and-frame filter unit operates at the same pressure, what area is 20 m3/hr of a calcium carbonate slurry at a pressure drop of 0.679
required to handle 5,000 L/h of feed? (d) A rotary-drum vacuum bar with 37.5% submergence, and a drum rotation rate of 0.2 rpm.
filter is being considered for the same application. The fraction of Pertinent data are: T ¼ 20 C, mass-fraction carbonate in the feed
the drum area submerged in the slurry is 0.35, the pressure drop 0.15, mass fraction of water in the filter cake ¼ 0.40, filtrate density
maintained by the vacuum pump is the same as for the plate-and- 998.3 kg/m3, filtrate viscosity ¼ 1 cP, and dry carbonate density
frame filter press, and n ¼ 1 rpm. What drum area is needed? 2,709 kg/m 3. Assume the media resistance is negligible. Use Table
19.15. Filter resistances from test data. 19.8 for cake compressibility.
Filtration tests were carried out with a plate-and-frame filter press 19.18. Constant-rate filtration of compressible talc.
at 20 C under the following conditions: rs ¼ 2,710 kg/m3, filter area Talc is to be filtered at a constant filtration rate under conditions
0.37 m2, filtrate m ¼ 0.001 N-s/m2, and cF ¼ 10.037 kg/m3. listed below. The pressure drop is expected to rise from 2.0 psi to
100 psi. Generate a table of pressure versus time at 10, 20, 40, 60,
80, and 100 psi. Assume Rm in (19-27) is negligible, but justify your
Filtration data from Svarovsky [23] assumption. Obtain aavg by integrating (19-26), which gives aavg ¼
(1 s)a 0 (DP)s [23]. The conditions are: percent solids in slurry
DP 10 5, N/m2 t, s V, m3 (t ts)=(V Vs), s/m3 1%, filtrate density ¼ 62.4 lb/ft 3, viscosity ¼ 1.49 cP, talc density
167 lb/ft3, filtration rate ¼ 0.000362 ft3/s-ft 2, cake compressibility
0.4 447 0.04 factor ¼ a ¼ 8.66 1010(DP, psi)0.506 m/kg, and cake porosity ¼
0.5 851 0.07 e ¼ 0.86(DP, psi) 0.045.
0.7 1,262 0.10
0.8 1,516 0.13 Section 19.5
1.1 1,886 0.16 19.19. Sigma factor for a centrifuge.
1.3 2,167 0.19 In Example 19.10, a laboratory test was conducted with a small,
2,552 tubular-bowl centrifuge on a fermentation broth. At the operating
1.3 0.22
conditions, a sigma value of 7,400 ft2 was computed, with a meas-
1.3 2,909 0.25 ured volumetric flow rate of 0.11 gpm. For the commercial plant
1.5 3,381 0.28 that will process the same broth, the largest tubular-bowl centrifuge
1.5 3,686 0.30 available has the following characteristics: bowl speed ¼ 15,000
 Exercises 813

Section 19.7
19.22. Cell disruption.
The release of protein from a suspension of Saccharomyces
crevisiae by disruption in an industrial homogenizer was inves-
tigated by Hetherington. [Trans. Instn. Chem. Engrs., 49, 142
(1971)]. The homogenizer pressure was varied, and it was oper-
ated on a recycle basis. The results were correlated by the
equation
log½Rm =ðRm RÞ ¼ KNPc
where R is the amount of protein released (mg/g yeast), Rm is the
maximum amount of protein available for release, K is a tempera-
ture-dependent dimensional constant, P is pressure in kg f/cm 2, N
is the number of passes through the homogenizer, and c is a con-
stant. Make a plot of log[Rm=(Rm R)] versus N with P as a
parameter, using the data below, and determine the value of the
constants K and c.

Figure 19.34 Disc-stack centrifuge.

log[Rm=
(Rm R)] P(N)
rpm, R0 ¼ 5 cm, R1 ¼ 2 cm, and bowl length ¼ L ¼ 10 cm. Using the 1 500(4.7) 400(7.2)
sigma scale-up theory, calculate how many gpm this unit can
0.8 500(3.7) 400(5.7)
process.
0.6 500(2.8) 400(4.2) 360(8.3)
19.20. Disc-stack, centrifuge.
A dilute aqueous slurry of viscosity 0.001 N-s/m2 and density 0.4 500(1.9) 400(2.9) 360(5.5) 270(9.3)
1,000 kg/m3 is to be separated from the solid particles, which have a 0.2 500(0.9) 400(1.5) 360(3.0) 270(4.8)
density of 3,000 kg/m 3, by centrifugal sedimentation. Pilot-plant 0.1 500(0.4) 400(0.7) 360(1.5) 270(2.8) 200(7)
runs at a feed rate of 10 10 6 m3/s using a disc-stack centrifuge, 0 500(0) 400(0) 360(0) 270(0) 200(0)
shown in Figure 19.34, with n ¼ 20 disks; a disk angle, u, of 45
from the vertical; an outer radius R0 ¼ 0.075 m and an internal
radius R1 of 0.025 m; and operating at a bowl speed of 20,000 rpm, 19.23. Precipitation of monoclonal antibody with ammonium
give satisfactory results. The same separation is to be carried out in a sulfate.
plant using a disc centrifuge with R0 ¼ 0.25 m and R1 ¼ 0.075 m, In a laboratory experiment using 10 L of cell culture, ammonium
100 disks, a disk angle of 50 , and rotating at 14,000 rpm. The sulfate powder is added to the solution to precipitate a monoclonal
sigma scale-up theory used in Example 19.10 with (19-33) applies, antibody. Eighty percent of the antibody, whose initial concentration
except that for a disc centrifuge [24], was 0.8 mg/mL, with MW of 150,000, precipitates in the 1.0-M
X 2pnv2 R30 R31 cotu solution. If the solubility of the antibody in pure water is 350 kg/m3,
¼ (1) what concentration of ammonium sulfate is needed to precipitate
3g
98% of it?
where v is the rotation rate in radians/unit time. What is the 19.24. Rotor-stator mechanical cell disrupter.
expected production rate in m3/s of liquid? A rotor-stator cell disruptor consists of a tapered cavity (sta-
tor) and a cone-shaped rotating rotor. The cell suspension is
Section 19.6 pumped through the very small gap between the rotor and stator,
19.21. Filter-cake washing. thus being disrupted by the high shear rate. It was found that at a
The experimental data given below are the instantaneous concen- residence time of 10 s, an intracellular antibiotic release was
tration of solute S, which is to be washed out of a filter cake as a 2.65 mg/mL, and at a residence time of 30 s, it was 4.77 mg/
function of W, the wash ratio (volume of wash liquid/volume of liq- mL. If the residence time is increased to 60 s (by slowing the
uid in the cake), and So, the initial solute concentration prior to feed rate), what concentration of antibiotic is predicted? What is
washing. What is the fractional recovery of solute for W ¼ 1 and 2? the percent extraction?

S/So 1.0 1.0 0.9 0.80 0.7 0.55 0.43 0.23 0.15 0.1 0.04 0.02 0.1 0.04 0.02
W 0 0.2 0.4 0.6 0.8 0.9 1.0 1.2 1.4 1.6 1.8 2.0 1.6 1.8 2.0