MIXING OF POWDERS 585

27. W. J. B. van der Bergh, B. Scarlett, and Z. I. a Radioactive Tracer Technique," Powder Technol
Kollar, "Computer Simulation Model of a Nauta® 4:345-350 (1970-71).
Mixer," Powder Technol 77:19-30 (1993). 46. KOMAX Systems, Inc., 1947 E. 223rd Street, Long
28. Prater Industries Inc., 1515 South 55 Court, Beach, CA 90810.
Chicago, IL 60650. 47. Charles Ross & Son Company, 710 Old Willets
29. V. Mixers are available from Patterson-Kelley Co., Path, Hauppauge, NY 11787.
Division of Harsco Corp., East Stroudsberg, PA 48. Toray Industries Inc., 3 to 3 Nakanoshima Kita-ku,
18301. Osaka 530, Japan.
30. V Mixers and Double Cone Mixers are available 49. Lightning Mixer Equipment Co. Inc., 128 Mount
from the General Machine Company of New Road Blvd., Rochester, NY 14603.
Jersey, Inc. (GEMCO), 55 Evergreen Avenue, 50. R. H. Nielsen, N. Harnby, and T. D. Wheelock,
Newark, NJ 07114. "Mixing and Circulation in Fluidized Beds of
31. V Mixers and Ribbon mixers available from O'Hara Flour," Powder Technol 32:71-86 (1982) describes
Manufacturing Ltd., 65 Skagway Avenue, Toronto, the use of Cabosil added to the flour to facilitate
Canada, M1M 3T9. fluidization and minimum fluidization velocity.
32. Ribbon and other mixer systems available from 51. Statitec Mixing Systems, EMI Inc., P.O. Box 912,
Munsun Machine Company Inc., 210 Seward Clinton, CT 06413. The passive mixer available
Avenue, Utica, NY 13503. from Statitec is known as the Statiflo mixer.
33. Mulling equipment is available from National 52. D. A. Pattison, "Motionless Inline Mixers Stir Up
Engineering Company, 20 North Wacker Drive, Broad Interest, Chem. Eng. 11:94 (1969).
Chicago, IL 60606. 53. D. J. Cassidy, B. G. Scribens, and E. E. Michaelides,
34. The Turbula® system was developed by Willy A. "An Experimental Study of the Blending of Granu-
Bachofen A.G., Maschin en fabrik, C.H. 4005 Basel, lar Materials," Powder Technol 72:177-182 (1992).
Utengasse 15, Switzerland. Available in North 54. J. R. Johanson, "In Bin Blending," Chem. Eng.
America from Glen Mills Inc., 395 Allwood Prog. 66(6):50-55 (1970).
Avenue Road, Clifton, NJ 07012. 55. M. R. Stein, "Gravity Blenders: Storing and Blend-
35. The AeroKaye® mixer is manufactured by Amherst ing in One Step," Powder Bulk Eng., Vol 4, No. 1,
Process Instruments Inc., Mountain Farms, Tech- pp. 32-36 (1990).
nology Park, Hadley, MA 01035-9547. 56. A. W. Roberts, "Storage and Discharge of Bulk
36. L. T. Fan and Y-M. Chen, "Recent Developments Solids from Silos with Special Reference to the Use
in Solid Mixing," Powder Technol 67:255-287 of Inserts," POSTEC-Research Report, May 1990.
(1990). 57. A. W. Roberts, "Design of Bins and Feeders for
37. N. Harnby, M. F. Edwards, and A. W. Anti-segregation and Blending," in Proceedings of
Nienow, Mixing in the Process Industries, 2nd edit. the Institute of Mechanical Engineers, Bulk
Butterworth, London (1992). Materials Handling—Towards the Year 2000,
38. Air mixers are available from Andritz Sprout-Bauer London 1991.
Inc., Muncy, PA 17756. This equipment is manu- 58. H. T. Young, Apparatus for Gravity Blending
factured in the United States under license from of Particulate Solids, U.S. Patent No. 4,353,652,
Gebruder Grunkg Lissberg, Germany. October 12, 1982.
39. Dynamic Air Conveying Systems, 1125 Walters 59. C. E. Roth, Blending System for Dry Solids, U.S.
Blvd., St. Paul, MN 55110. Patent 4,358,207, November 9, 1982.
40. V. A. Fauver and A. E. Hodel, "Pulsed Air Blender 60. I. A. S. A. Peschl, "Universal Blender—A Blending
Produces Uniform 15 Ton Lots in 20 Minutes," and Mixing for Cohesive and Free Flowing
Chem. Proc. (1986). Powders," Bulk Solids Hand. 6(3) (1986).
41. Blendicon is available in Canada from Ward Iron 61. H. Wilms, "Blending Silos. An Overview," Powder
Works, Ltd., 1223 Victoria Street, P.O. Box 511, Hand. Proc. 4(3) (1992).
Welland, Ontario, L3B 5R3. 62. J. W. Carson and T. A. Royal, 1991. "Techniques
42. Fuller-Kovako Corporation, 3225 Schoeperville of In-Bin Blending," in International Conference
Road, P.O. Box 805, Bethlehem, PA 18016-0805. on Bulk Materials Handling—Towards 2000, 1
43. J. M. Ottino, "The Mixing of Fluids," Sci. Am. Mech. E., London.
56-67, Vol. 260, No. 1 (1989). 63. K. S. Manjunath, S. R. de Silva, A. W. Roberts, and
44. Chemineer Inc. manufactures a passive mixer S. Ballestad, "Determination of the Performance
known by the trade name Kenics® Static Mixer. of Gravity Blenders with Emphasis on Plane Sym-
125 Flagship Drive, North Andover, MA 01845. metric Designs. POSTEC-Research Report
45. L. T. Fan, S. J. Chen, N. D. Eckhoff, and C. A. 921600-2, June 1992. Available from POSTEC Re-
Watson, "Evaluation of a Motionless Mixer Using search A / S , Kjolues Ring, Porsgrunn, Norway.
 12
Size Reduction of Solids
Crushing and Grinding
Equipment
L. G. Austin and O. Trass

CONTENTS

12.1 INTRODUCTION 586

12.2 A BRIEF REVIEW OF FRACTURE MECHANICS 587
12.3 SIZE REDUCTION MACHINES 598
12.4 THE ANALYSIS OF SIZE REDUCTION PROCESSES 605
12.5 NEW MILLS 623
12.6 FUTURE WORK 631
REFERENCES 631

12.1 INTRODUCTION mental aspects are emphasized rather than

mechanical or process engineering aspects, to
The unit operation of the size reduction or form a background for intelligent decision-
comminution of solids by crushers and mills is making in the choice and analysis of size re-
a very important industrial operation involving duction systems.
many aspects of powder technology. It is esti- In many operations, a material must be
mated that mechanical size reduction of rocks, reduced from lumps of up to a meter in size to
ores, coals, cement, plastics, grains, etc. in- a fine powder, sometimes a powder essentially
volves at least a billion tons of material per less than 100 jjum in size. It is clear that size
year in the United States alone. The operation reduction over many orders of magnitude in
ranges in scale, for a single device, from a few size cannot be efficiently achieved in a single
kilograms per hour for speciality products to machine and a sequence of different types of
hundreds of tons per hour for metallurgical machine is used, each machine designed for
extractive purposes. In this chapter, the funda- efficient operation on a particular feed size.
586
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 587

Machines for breakage of large lumps are gories, elastic and ductile, with the corre-
called crushers and machines for smaller sizes sponding failure under stress termed brittle or
are called mills, with a range of overlap where nonbrittle fracture, respectively. Consider a
either a fine crusher or a coarse mill can be simple tensile stress, as illustrated in Figure
used. The operation of crushing normally does 12.1. Stress is defined as a = F/A, and Figure
not give problems because the energy con- 12.2 shows the characteristics of elastic and
sumption and capital cost per ton per hour is ductile materials. An elastic material can be
not high. The principal requirement for crush- stressed, producing elongation, and the mate-
ers is a mechanical requirement—they must rial returns to its original shape when the
be very robust because of the high stress re- stress is removed. However, if the solid is
quired to crush a large lump. On the other stretched too far, catastrophic failure occurs
hand, fine grinding consumes a great deal of and the solid fractures at a stress termed the
energy and may lead to high abrasive wear, so
tensile strength. Ductile materials undergo a
the major scientific and technical problems are
partially irreversible stretching before failure
concerned with fine grinding and most current
occurs.
research is focused on these problems.
Elastic materials fail at small strain so or«
Before discussing the various types of com-
0-o and the strain-stress relation up to where
minution equipment in detail, it is invaluable
failure occurs is the empirical Hooke's law:
to have a clear idea of the fundamental physi-
cal laws involved in size reduction. These in-
volve the areas of fracture mechanics, particle- = Ye = Y- (12.1)
fluid dynamics, agglomerative forces (dry and
wet), and powder flow. The last four topics are
covered elsewhere in this book and are men- where Y is Young's modulus, s is strain. For a
tioned here only as they arise. Fracture me- perfect crystal Y depends on the orientation of
chanics are discussed in some detail. Since the the stress, but most brittle solids are polycrys-
objective of size reduction is to obtain a suit- talline with a random arrangement of crystal-
able product size, the accurate measurement lites, so Y is an effective isotropic elastic con-
of powder size distributions is a basic feature
stant. The work done on the solid to go from
of the process; this is also covered in detail
zero external stress to a stressed state by slowly
elsewhere. However, the prediction of size dis-
tributions and how they change with mill oper-
ation is dealt with in depth.

12.2 A BRIEF REVIEW OF

FRACTURE MECHANICS

12.2.1 Stress, Strain, and Energy

To produce size reduction the lumps of solids Cross-Section A
must be fractured, and they must be stressed
to produce fracture. Quantitative theoretical
analysis is possible only for relatively simple
states of stress, but the concepts that emerge
are qualitatively useful for the complex stress-
ing conditions of industrial crushers and mills. \
Materials are divided into two broad cate- Figure 12.1. Simple tensile stress.
588 HANDBOOK OF POWDER SCIENCE

at the point can be resolved into two compo-

elastic . nents, the normal component perpendicular to
the plane and the shear component in the
plane. The normal stress tends to pull A away
failure
from B (tension) or force A into B (compres-
sion), whereas the shear stress tends to make
Strain e
A slip sideways with respect to B. From a
Figure 12.2. Illustration of stress-strain curves for sim- molecular aspect, a solid consists of an array
ple tensile tests; cr0 = force/original cross-section; of atoms, molecules, or ions at rest (although
strain e = x/L0. vibrating) with respect to one another, so that
the attractive and repulsive forces between
them are exactly balanced. Viewing these
increasing F up to a final stress of cr is
forces as acting like springs, Figure 12.5 illus-
f[o Fdx and using Hooke's law:
trates the three stress states. Obviously, un-
Work per unit volume = Ye1/! = a2/2Y even compression or tension across a solid
(12.2) must produce shear stress.
Drawing an arbitrary set of axes through the
This reversible strain energy is stored in the point that defines x, y, z directions (see Fig-
solid. If the solid is immediately loaded to a, ure 12.3), the shear stress can be resolved into
the work done is crAeL, which is cr2/Y per the components rxy, rxz. The sign convention
unit volume. Half of this is strain energy and is that material —x is dragging material at +x
the other half will accelerate the solid and in the y direction with a force per unit area of
cause it to oscillate until frictional damping Txy, when the sign convention for normal stress
converts the kinetic energy to heat. Similarly, is positive for compression, negative for ten-
if a solid suddenly expands at a constant a, sion. Taking moments about a point it is read-
the work done per unit volume is a2/Y and ily shown that rxy = ryx, ryz = Tzy, TZX = TXZ
again only half is reversible strain energy. (see Fig. 12.4).
More generally, consider a stressed solid at
equilibrium. At a differential plane at any 12.2.2 Directions of Normal
point in the solid there is no net force (since and Shear Stress
there is no movement of one part of the solid To describe the process of fracture it is neces-
with respect to another), as illustrated in Fig- sary to know the normal and shear stresses
ure 12.3, and the force of material A acting on and their directions in the solid. The relations
material B must equal the force of B acting on between stress and direction can be readily
A. The force per unit area of A acting on B is developed for a planar solid (two-dimensional)
called stress, and equals B on A, so stress is a as follows. Consider an arbitrary direction de-
force transmission through the solid. The stress

T xy dy dxdz
s
Tyzdzdxdy

sman plane in
the solid

Figure 12.3. Illustration of stress through a point in a Figure 12.4. Moments about a point in the zy plane:
stressed solid at equilibrium. material outside square acts on material inside.
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 589

Compression -^—ar From Eq. (12.4), a particular value of a,

^/W\/\OW\AA/\AO^AA/W\AO^/WW that is a, can be obtained that makes r = 0.
Tension —*»o"
V\AAAO\AAAAAAO\AAAAAAO\AAAA/ T
xy/(°x ~ °"y) =
I1 (12.5)

New x, y axes are defined along this direction,

as shown in Figure 12.6c; then rxp = Tyx = 0,
and Eqs. (12.3) and (12.4) become
Figure 12.5. Illustration of states of stress on a molecu-
lar basis. a = ax cos2 p + av sin2 fi (12.6)

T = (12.7)
fined by a in Figure 12.6a, and imagine the
shaded differential element of solid at equilib-
rium to be acted on by forces from the outside where /3 is now a general direction variable
material, as shown. Because the element is a (angle) measured from the new axes and a, r
differential element, the forces are uniform are the stresses at angle /3 (at angle a = p +
over the small lengths of side and represent a; see Figure 12.6d). These axes are called the
the forces at a point in the solid. The relative directions of principal stress and ax, or^ are
lengths along x,y, and the hypotenuse are the principal stresses.
cos a: sin a: 1, and since rxy = ryx a force bal- Eliminating p between Eqs. (12.6) and
ance gives (12.7) gives the equation of a circle, so the
relation between T and a at any angle p can
a = ax cos2 a + cry sin2 a + 2rxy cos a sin a
be represented by the Mohr circle as shown in
(12.3) Figure 12.7. The maximum shear stress occurs
in a direction of p = 45° ( = 135°) and
- a;, sin 2 a + r cos 2 a (12.4)
T == — xv

The force balance will apply for any other a 4- T 2

(see Fig. 12.6b), with a different a and r, of
> - «rl/2 -
I 2 ) " •

(12.8)
course, but the same ax, ay, rxy.
o- at TT = (or, + oj)/2 (12.9)

The maximum normal stress is clearly the

larger of the principal stress values. Also, it is
readily shown that the principal stresses are

(c) (d)
Figure 12.6. Equilibrium stress conditions in a planar 135°
element. Figure 12.7. A Mohr stress circle for a planar system.
590 HANDBOOK OF POWDER SCIENCE

related to the normal stresses in the original where X, Y are the body forces in the x and y
coordinates by directions at the point.
crv + a;, The differential strains at point x,y are
(12.10) defined by ex = du/dx, ey = dv/dy for the
linear strains, where u is the change in x
Thus, knowing crx, cry, rxy at any point in the dimension from the nonstressed state at point
solid, the direction and magnitudes of the x, y; v is change in y dimension. The differ-
maximum shear stress, tensile stress, and com- ential planar shear strain yxy is illustrated in
pressive stress are readily calculated. Figure 12.9 and is defined by yxy = angular
A similar treatment 1 in three dimensions, deformation 6X + 62. Clearly 6X =
considering the six stress components, leads to (du/dy) dy/dy and yxy = yyx = du/dy +
Mohr circles for the three planes of principal dv/ dx. The empirical physical laws relating
stress as illustrated in Figure 12.8, where stress and strain are Hooke's law, ex = crx/Y,
(T3,a2, ai are principal stresses ranked in or- and the fact that a strain in the x direction
der of magnitude. It is concluded that the causes a proportional dimensional change in
maximum tensile stress has the magnitude and the y direction (stretching in x gives a con-
direction of the largest negative value of the traction in y, compression an expansion). Thus
three principal stresses and the maximum ey due to ex equals —vex, where v is Poisson's
shear stress occurs at 45° between the av a3 ratio ( « 0.25). For small elastic planar defor-
directions, with a magnitude given by Eq. mations the total strains are:
(12.8).
€
12.2.3 Differential Stress-Strain x =Y + \ p f ) (m3a)
Equations
e + (12 13b)
The second step is to find the values of y= Y (~P~Y) *
ax,ay, rxy at all points in a solid, since these Defining a modulus of rigidity G = Y/2(l +
can be converted to maximum stresses and v), it can be shown from Hooke's law that:
directions. For planar stress, a differential
force balance of a rectangular differential ele- 2(1 + v)
ment at position x, y in the solid gives Jxy = rxy/G = rxy (2.14)

0 = •yx
(12.11) Using the definitions of strain
dx dy
d\ d\ d2yxv
day, dr,xy
0 = —- + + Y (12.12) dx2 dxdy
dy dx

<r. ,cn_ plane

(j| ,ov> plane

dy stressed

v
non- stressed

Figure 12.8. Mohr principal stress circles for a three- Figure 12.9. Illustration of differential strains at a point
dimensional solid. x, y in a planar solid.
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 591

and from Eqs. (12.13) and (12.14) dimensional tension. The tension stretches the
bonds between the molecules, as illustrated in
+ dx2 Figure 12.10, where the arrows indicate inter-
dx2
molecular attractive-repulsive forces. In the
2(1 + v) stretched state, any molecule still has a bal-
(12 15
Y ,dxdy
, - > ance of forces on it but, as Figure 12.10b
If the body forces are known, Eqs. (12.11), shows, the movement away from the non-
(12.12), and (12.15) are three simultaneous stressed equilibrium against attractive forces
differential equations in the unknowns requires addition of energy (integral of
ax,ary, rxy. They are solved using the stress force X distance) and the solid reaches a new
and/or strain boundary conditions, that is, the equilibrium at a higher energy state (stored
stress-strains imposed on the solid from exter- strain energy). The maximum attractive force
nal action. For negligible X, Y the solution that the solid can exert on the surface layer is
procedure is to define the Airy stress function the inflection point of the potential energy
F(x,y) such that ax=d2F/dy2 and ay = curve since force = d(energy)/d(separation
82F/dx2, for then rxy = - d2F/dxdy and distance), and an external tension that exceeds
from Eqs. (12.13) and (12.15) (d4F/dy4) + this maximum causes an unbalance of forces
(d4F/dx4) + 2{d4F/dx2dy2) = 0. Solving this and acceleration of one plane of molecules
equation with the transformed boundary con- away from another. The solid would catas-
ditions gives F(x,y) and ax, cry,Txy follow by trophically disintegrate at all planes in the
double differentiation. Equivalent but more solid. Assuming Hooke's law to apply up to the
complex equations exist for three dimensions. inflection point, the strain energy per unit
The strain energy above the nonstressed state volume of solid is, from Eq. (12.2), a2/2Y.
is calculated from The area produced per unit volume is 2N
where TV is the number of planes per unit
crzez Txyyxy length; N equals 1/d where d is the interpla-
nar spacing. Thus,
(12.16)

12.2.4 Ideal Strength, Stress

Concentration, and the Griffith °"ideal failure ~ V ~ T (12.17)
Crack Theory
The concept of ideal strength can be illustrated where y is the surface energy defined as the
by considering an ideal solid made up of work required to create a unit area of surface
planes of molecules, subjected to simple one- from the unstressed solid. Equation (12.17)

(Repulsion

Cohesive Plane of atoms

forces Attraction
between
atoms Surface plane

i i i
Applied external tension
Separation in direction
of tension
(a) (b)
Figure 12.10. Illustration of forces between molecules in a solid, (a) Cohesive forces; (b) energy of position.
592 HANDBOOK OF POWDER SCIENCE

must underestimate the ideal strength since where a is the ellipse axis in the y direction, b
Hooke's law underestimates the force re- in the x direction. For an elliptical hole with
quired to reach the inflection point. Since y is its long axis perpendicular to the stress direc-
known for simple solids, it is readily shown tion, a is greater than b, and stress concentra-
that the tensile force for real fracture is orders tion can be very high if a » b.
of magnitude less than ideal. Griffith2'3 argued that real solids contain
The concept of stress concentration or stress many minute flaws corresponding to the
intensity factor can be illustrated by consider- three-dimensional equivalent of the elliptical
ing a planar solid containing a small hole, holes discussed above and that these points of
under a uniform externally applied tensile weakness initiate cracks at stress levels much
stress of S in the x direction and zero in the y below ideal. He made four basic assumptions:
direction. Without the hole, the solution is (1) that stress concentration occurs at the tip
intuitively obvious as ax = S, ay = 0, rxy = 0 of the flaw, (2) that the solid is stressed to
for all values of x and y. With a small hole of where the intermolecular bonds at the tip are
radius a present (see Fig. 12.11), the added stretched to breaking point, (3) that the stress
boundary condition is state is reproduced at the tip for an infinites-
imal expansion of the flaw and, (4) that energy
for expanding the flaw as a propagating crack
is available because the solid cannot immedi-
since there is no external stress inside the ately relax from its externally applied stressed
hole, and the solution is: or strained state. The solution of the
stress-strain equations for a long ellipse gives
the extra strain energy due to the presence of
the ellipse as Az TTC2<J2/Y where c is the long
(12.18) half-axis, that is, half the crack length, and
Az is the crack width. Thus, dwx/dc =
which gives a maximum stress of 35 in the x Az2irc(r2/Y. A sudden irreversible change
direction at 6 = 90° and 270°. Since a crack from c to c + dc at the instant of fracture is
will open up under tension it is reasonable to like a loaded solid suddenly expanding dc at
expect that the solid will fail by cracks starting constant load, so that the work done is twice
at the top and bottom of the hole and pro- the (reversible) strain energy, dw3/dc =
gressing in the ±y direction. The solution for 2Az 2TTC(T2/Y. The energy necessary to break
a small elliptical hole is more complex but bonds is Aye Az for a crack of half-length c,
gives a maximum stress of so dw2/dc = 4yAz. Using the principle of
virtual work, dw3 = dwx + dw2 at crack initia-
2a tion and the critical tensile strength is
OL^/S =1+ — (12.19)
(12.20)

rx (r,e) Comparing Eq. (12.20) with Eq. (12.17), val-

ues of d are no more than a few Angstroms,
so a flaw with a half length of hundreds of
Angstroms can give orders of magnitude re-
duction in tensile strength To from the ideal
(a) (b) strength. As the crack progresses after initia-
Figure 12.11. Illustration of stress concentration in a
tion, dw3/dc > (dw1/dc) + (dw2/dc) and ex-
plane due to a circular or elliptical hole; s = applied tra energy is available to accelerate the crack
tensile stress. tip. The system is unstable and the crack
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 593

rapidly expands, accelerating to high veloci- may run into a region of compression that
ties. The strength is lower than ideal because prevents further crack growth. Also, solutions
the bulk stress does not have to be sufficient of the stress-strain equations for simple com-
to break all the bonding forces at once, since pression of discs, cylinders in the "Brazilian"
only the bonds around the crack tip are break- radial mode of testing, and spheres, show that
ing at any instance of time. In addition, Eq. tensile stresses are present, with maximum
(12.20) is valid for a single flaw whereas the values along the loaded axis. Even for cubes
presence of many flaws close together will give and cylinders loaded along the axis, friction
further reductions in strength. between the loading platen and the sample
Obviously, pure compressive stress does not leads to nonuniform compressive stress and
cause the flaw to open and will not cause regions of tensile stress. Thus compressive
crack propagation, so tensile stress is neces- loading of irregularly shaped lumps or parti-
sary for brittle failure. It might be thought cles will certainly produce local regions of
that tensile stress will not exist under condi- tensile stress and, hence, brittle fracture.
tions of simple one-dimensional compression. Ductile materials, on the other hand, un-
However, a more detailed analysis considering dergo plastic deformation due to sliding of
all possible orientations of the flaws shows planes of solid over one another, with the
that tensile stresses are produced at the tip of fundamental mechanism being that of move-
an ellipse at a suitable orientation even under ment of dislocations under stress gradients. In
conditions of bulk compression. The result for this type of movement, the bonding forces
a planar system with bulk normal stresses orl between planes are not broken all at once, but
and cr2 and flaws of a size that would give a only enough bonds are broken to allow the
tensile strength of TQ under one-dimensional dislocation to move to the next position, the
tension (with the crack axis perpendicular to bonds reform behind the dislocation, and so
the stress) is shown in Figure 12.12. The com- on, thus leading to slip of one plane over
pressive strength under one-dimensional com- another by a series of low-energy steps. We
pression is ST0, that is, compressive strengths have already seen that the maximum shear
of brittle materials, are about an order of force occurs at 45° to the direction of principal
magnitude higher than tensile strengths. stress, so plasticity and failure by shear will
Under combined stressed conditions the appear as illustrated in Figure 12.13. The slip
crack will propagate in a direction perpendicu- process appears as the region of yielding in
lar to the local tensile stress conditions and Figure 12.1, and is quite unlike the unstable
initiation of brittle failure. Slip may initiate
from a suitably oriented flaw that gives stress
concentration, but there is no opening of a
locus of fail-safe stress
combinations o~|£ »°~2C crack comparable to that under tensile stress.
However, other factors come into play once
plastic yield has commenced. The plastic slip
may cause part of the solid to act as a wedge,
thus creating tensile forces that then propa-
gate brittle fracture, as illustrated in Figure
Oj? Compression 12.13. Also, the movement of dislocations can
pile up dislocations at a grain boundary, thus
-| + < T 2 > O leading to a small hole that can nucleate a
Griffith crack. Highly ductile materials under
simple one-dimensional tensile loading will
Figure 12.12. Illustration of effect of combined stress
on failure from Griffith Flaws with simple tensile neck down, giving increased stress at the neck
strength of To: equations are equations of locus. and, eventually, complete slip failure with pos-
594 HANDBOOK OF POWDER SCIENCE

Load sion and induced tension because a large strain

is required to make the solid reach a highly
stressed state, and because dw4 is large. Rub-
ber materials have this property because of
the shape and flexibility of the long molecules,
which can coil and uncoil, bend, and straighten.
A high degree of crosslinking bonds will re-
duce the flexibility, so these materials are
Slip planes at approximately 4 5 ° weakly crosslinked, which means they are weak
to shear stress. Thus, the best conditions for
(a)
failure are tensile strain which straightens out
the tangled and coiled molecules into a paral-
Load
lel array like a crystal, with superimposed shear
that breaks the few crosslinking bonds.

12.2.5 Qualitative Applications of

Fracture Theory: Grinding Energy
Rocks, ores, and coals being broken in size
reduction machines will normally undergo
Plastic slip brittle fracture via preexisting Griffith flaws.
The strength or grindability of these materials
(b)
will correlate only roughly with the hardness
Figure 12.13. Illustration of failure by shear: slip leads or chemical bond strength, because the num-
to brittle fracture. ber, size, and orientations of the flaws are
additional variables. The materials are stronger
sible cleavage along crystallographic planes of in compression than tension. To calculate the
weakness. strength of a lump or particle being subjected
The Griffith treatment is extended4 to allow to stress, from an a priori theory of fracture
for plasticity by including a term, dw4, for the mechanics, it would be necessary to: (1) solve
energy required for plastic deformation caused the stress-strain equations for the geometry
by the moving stress field around the crack tip. and conditions of applied stress; (2) convert
Then the initiation condition is dw3 — dw1 = the results to the local magnitude and direc-
dwt > dw2 + dw4. The value of dw4 depends tion of the principal stresses at all points in
on the size and density of dislocations in the the solid; (3) consider the density (number per
solid and dominates over bond energy dw2 for unit volume) distribution of sizes, and orienta-
ductile materials. Thus, ductile materials are tion (possibly random) of flaws in the solid;
stronger than purely brittle materials. Once and (4) determine the places where local ten-
fracture commences, however, the term for sile stress can activate the flaws to the point of
plastic energy may decrease because the crack fracture initiation, with failure commencing at
moves at high velocity in relation to the time the weakest location. Such a calculation is
scale for movement of the dislocations that clearly impossibly complicated for most real
give plasticity. conditions in a mill, and can be attempted
Some polymeric materials have the ability only for idealized solids and simple stress con-
to deform to high strain without fracture, for ditions [see Equation (12.23) for an example].
example, rubber, and the description of their In addition, most grinding machines have
failure can be considered as a separate prob- some degree of impactive stress that propa-
lem. They are difficult to break by compres- gates stress waves through the solid, activating
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 595

flaws to tensile fracture in the process. The to give yield, and vice versa. The slip surface is
size distribution of the suite of fragments pro- now along the direction of r - /JLCT > T 0 . The
duced on fracture is as important as the frac- value of /n is normally small so that the tensile
ture itself (see later), and there exists no known strength is fairly close to the compressive
theory for its prediction. Theory predicts, and strength, and slip surfaces tend to lie fairly
experiment confirms, that a fracture propagat- close to 45° to the principal stress directions.
ing under local tensile stress rapidly reaches From Eqs. (12.6) and (12.7) it is readily shown
high velocity (unless it reaches a zone of local that
compressive stress), of the order of the veloc-
ity of sound in the solid. This leads to a stress IJL = ° ~ ? (12.22)
wave that propagates from the crack tip and
this stress wave in turn initiates more fracture
at flaws in the path of the crack. This leads to where Co, To are the magnitudes of simple
bifurcation of the crack, with bifurcation of one-dimensional compressive and tensile
each of the new arms, and so on, to give a stresses required to give yield. It will be re-
"tree" of cracks through the solid (see Fig. membered that the maximum shear stress for
12.14). The energy associated with the rapidly principal stresses of cr-, ^ in two dimensions
moving stress wave is normally sufficient to is |or, — orx\2, so slip is aided by a combination
pass the crack through grain boundaries and of compressive and tensile stresses.
through regions of bulk compressive stress. A comparison between the failure of brittle
Ductile materials fail by initial shear, and it and ductile materials shows the following
is again necessary to find the magnitude and major features:
direction of shear at all points through the
solid. The Mohr-Coulomb criterion is that 1. Pure brittle failure is almost independent
failure occurs when shear stress reaches the of temperature, but as temperature in-
yield point given by creases to where dislocations are more
mobile, the failure may change to slip, and,
= T0 + (12.21) hence, lower strengths. Pure ductile failure
gives decrease of strength with increase of
where T 0 is the yield shear stress under condi- temperature owing to greater mobility of
tions of zero tensile or compressive stress per- dislocations. For brittle failure with a sig-
pendicular to the shear stress plane and /JL is nificant plastic energy term, strength in-
called the coefficient of internal friction. Equa- creases with temperature owing to the in-
tion (12.21) states that a high compressive crease of the plastic zone around the tip,
stress perpendicular to the shear plane will then decreases as failure changes to slip.
tend to prevent slip, thus requiring a higher r 2. For failure from Griffith cracks, a smaller
particle has a smaller probability of con-
taining a large flaw and will be relatively
stronger. Put another way, as brittle materi-
als break, the remaining fragments are
stronger because the larger flaws have bro-
ken out. On the other hand, failure by yield
is not very size-sensitive because the dislo-
cations are very small compared to lumps
or particle sizes.
3. The rate of stress application is more im-
portant with ductile materials than with
Figure 12.14. Tree of cracks in brittle failure. purely brittle materials, because a high rate
596 HANDBOOK OF POWDER SCIENCE

of stress application may give brittle failure to cool to very low temperatures, using liquid
whereas the same stress reached by slow nitrogen (77 K).
steps would give time for ductile behavior. There has been a great deal of misconcep-
4. Ductile materials demonstrate work hard- tion in the grinding literature concerning
ening, that is, initial deformation produces grinding energy. The previous discussions show
movement and pile-up of dislocations and that a strong solid must be raised to a higher
further deformation is more difficult. They state of stress for fracture to proceed, espe-
also demonstrate stress fatigue, again owing cially from applied compressive forces. Once
to the gradual accumulation of dislocations the fracture has initiated, only a fraction of
on repeated cycles of stress. the local stored strain energy around the
5. Loading of brittle materials with uniform propagating cracks is used to break bonds (the
triaxial compressive stress, hydrostatically y term). The fragments of solid are removed
for example, leads to greatly increased from external stress when the solid disinte-
strength by reducing local tensile forces grates, and the rest of the strain energy stored
and preventing cracks from opening. in the solid is converted to heat and sound.
Experiments on mills show that the fraction of
In the case of tough, rubbery materials, the the electric power input to the mill that is
best stress application for size reduction is the used directly to break bonding forces is very
scissors type of action, that is, a cutting action. small ( < 1%), usually less than the errors
This has three main features: (1) a large com- involved in the measurement of the energy
ponent of shear stress, (2) a high strain and balance. Rittinger's law,5 that the "energy of
stress caused by two forces applied in opposite size reduction is proportional to the new sur-
directions by the blades (or stator and rotor), face produced," has no correct theoretical
and (3) the creation of a surface flaw by the base.
very high local stress of a sharp blade pene- To make size reduction more energy effi-
trating the material. These features are illus- cient it is necessary to: (1) match the machine
trated in Figure 12.15. For rubbery polymers to the particles being broken, so that mill
with a substantial degree of crosslinking, which energy is efficiently transferred to stressing the
gives high shear strength, cooling the material particle; (2) get nonuniform stress conditions
to a low temperature can convert it to a brittle in the particles, because nonuniform stress
material, which can then be broken like other
generates local tensile stress to activate flaws
brittle materials. The action of the cooling is
to the point where fracture can initiate; and
to reduce the flexibility (ability to rotate and
(3) generate the right type of stress to match
bend) of the bonds joining the groups making
the failure characteristics of the material. The
up the polymer chains; it is normally necessary
specific energy consumption per unit of area
produced, for example, Joules/m 2 , can be used
as a comparative guide to efficiency, because a
higher value is certainly an index of more size
reduction per unit of energy input. It will not
necessarily be constant for a given machine
and material because it may increase or de-
crease with a greater degree of size reduction.
On the other hand, in many cases, the produc-
tion of extra fine material is undesirable, and
then the specific surface area of the product is
Figure 12.15. Illustration of shear-cutting actions. obviously not a good guide to mill efficiency,
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 597

because the specific surface area is con- edly the high reactivity of freshly fractured
tributed largely by the extra fine sizes. clean surfaces.

12.2.6 Property Changes and Reactions 12.2.7 Abrasion

It is known that prolonged treatment of mate- Abrasion is a special type of fracture—the
rials with repeated stress, by batch ball milling tearing out of small pieces of material from
for long periods, for example, can cause mas- the surfaces of the components used to apply
sive changes in the properties of the materials. stress to the material being fractured. It is
Rose6 showed that quartz underwent phase obvious, of course, that these grinding compo-
change from one form to another during ball nents must be strong enough to stress the
milling and the topic has been reviewed,7 giv- material being comminuted without bulk frac-
ing many examples. It has been suggested that ture themselves, but this is no guarantee that
shear stress will cause nucleation and growth their surface will necessarily be abrasion resis-
of one phase from crystallites of another in a tant. The fracture mechanics of abrasion is not
particle. In ball milling, tough organic poly- well developed theoretically, but it certainly
mers can undergo a delay period in which they involves high local surface stresses owing to
hardly break at all, followed by breakage. Pre- asperities in the rock and in the grinding sur-
sumably the pounding by the balls makes the face, plus local surface microflaw structure,
material weaker by causing some degree of ductility, friction, and possibly high local tem-
crystallization (molecular alignment). It is peratures. High rates of surface stressing
known that repeated light taps on a friable caused by high relative speed between stress-
coal create or extend cracks in the coals, so ing and stressed agents undoubtedly assist
that it eventually fails. Coal is a brittle poly- abrasive fracture.
mer with planes of weakness caused by the The chemical environment at the surface
geological process of laying down the material, can play a significant role, by two mechanisms.
but presumably other materials could show The first and most obvious is that an environ-
the same effect. ment that attacks the grinding surfaces will
Benjamin8 has discussed the formation of cause surface flaws and weakness and acceler-
solid solutions of ductile metals by prolonged ate abrasion. This effect is well recognized in
ball milling of a mixture of powders of the wet grinding. Second, there is some
components, and the similar creation of a fine evidence9'10 that an environment can change
dispersion of a brittle material in a ductile the bond strength and ductility of material
matrix. The mechanism appears to be cold- close to the surface, by strong chemical ad-
welding of clean surfaces produced by fracture sorption onto the surface. Such an effect will
or flattening, so that size reduction and size not change the bulk strength (unless condi-
growth occur simultaneously. In this case, the tions are such that fracture commences at
mill action must be such as to force particles surface flaws) but it can change the abrasive
together as well as fracture particles. It is comminution.
known that organometallic compounds can be The terms "hard" and "soft" are often in-
formed by ball milling chromium and nickel in discriminately used to characterize both the
organic liquids, accompanied by rearrange- bulk strength or resistance to comminution of
ment of the organic molecule to other organic a material and its ability to penetrate or wear
molecules with some H 2 , CH 4 , and CO 2 evo- another material. It would be better to use the
lution. Similarly, reactions such as Cr(s) + terms "strong" and "weak" for bulk comminu-
3TiCl4(l) -> CrCl3(s) + 3TiCl3(l) occur in an- tion properties, and reserve "hard" or "soft"
hydrous liquids. Again, the cause is undoubt- for the characteristics measured by one of the
598 HANDBOOK OF POWDER SCIENCE

usual hardness tests such as the Rockwell or Feed

Vickers tests. For example, coal can be consid-
ered to be a weak rock, but certain coals are
abrasive due to inclusions of quartz. Again, a
tough plastic such as Teflon is not hard or
abrasive but it can be very difficult to com-
minute.

12.3 SIZE REDUCTION MACHINES

12.3.1 Crushers
Product
There are many different types of machines Figure 12.17. Toothed single-roll crusher.
for size reduction, and almost every method of
breaking lumps that one might think of has
been incorporated into a crusher or grinder.
Figures 12.16, 12.17, and 12.18 show common material moving down until all of it falls
types of industrial crushers that are available through the gap. The crusher "capacity," that
in a wide range of sizes. The reciprocating is, the kg/s passed, is determined by the area
action of the movable jaw in a jaw crusher available for this mass flow. Feed or product
strains lumps of feed to the point of fracture, fragments less than the gap setting pass out of
as does the nonsymmetric movement of the the breakage zone and cannot be overground,
rotating mantle in a cone or gyratory crusher so these devices can be referred to as once-
(the nonsymmetric movement is produced by through machines.
the bottom end of the mantle shaft being set The machines are applying nonuniform
in an off-center, eccentric bearing). The size compressive stress and the mill power must be
reduction ratio, defined approximately as the sufficient to compress all the large pieces of
largest feed size divided by the largest product rock to the fracture point when the crusher is
size, is of the order of 10 and is varied by the full of large lumps. The stronger the rock, the
adjustable gap setting. The basic action is that larger the power required. Solution of the
entering brittle material is crushed, the broken equilibrium stress-strain equations for ideal
products fall under gravity into a narrower diametral loading of spheres of brittle material
space, and bigger fragments are crushed as the
metal-lump-metal space closes again, with

Feed
-l
- Bearing
Feed

Non-symmetric
Mantle

Product
Figure 12.16. One type of jaw crusher. Figure 12.18. Gyratory crusher.
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 599

gives the relation between force, diameter,

Feed Inlet
and maximum tensile stress a as
21
(28 4- 20P)
,(12.23)

where P is the force, r the sphere radius, v is

Poisson's ratio, and tensile stress is negative,
or

<rc = - — (12.23a)

where P* is the force that causes fracture and

ac is called the compressive resistance. Thus,
Figure 12.19. Heavy duty hammer crusher (Jeffrey Coal
the force to produce the tensile stress for Buster).
fracture is roughly proportional to r2. The
crushing surfaces of jaw and roll crushers are
often ribbed or toothed (1) to help prevent suited for nonabrasive materials, although it is
slippage of the rock as it is compressed, thus sometimes used for fairly abrasive rock be-
ensuring additional shear stress and (2) to give cause of its low capital cost: the user must
higher local stress at the surface of the mate- then resurface the hammers at frequent inter-
rial, thus activating or even producing local vals. Figure 12.20 shows a Cage-Pactor
flaws. An important point to remember is that crusher, in which solid is crushed as it passes
the same mass flow is occurring through an
ever-decreasing area, so the broken material
consolidates to a bed of lower porosity. It is
known that a highly consolidated bed of low
porosity is difficult to compress further (it has
a high Young's modulus) so that high stresses
can be produced on the metal surfaces squeez-
ing the bed near the gap. This is generally
avoided by controlling the rate of feed to a
crusher to prevent excessive consolidation at
the gap. As discussed later, compression of a
bed of particles (bed compression) can pro-
duce fine material, which is often undesirable
in a crusher because it may lead to excessively
dusty conditions in the work area.
Figure 12.19 illustrates a hammer crusher.
Material is broken by direct impact of the
hammers, by being thrown against the case or
breaker bars, and by compression and shear
when nipped between the hammers and the
case. The hammers are mounted on a heavy
rotor and/or the shaft is attached to a heavy Product
flywheel, to give a high moment of inertia of Figure 12.20. Cage mill (T. G. Gundlach Machine
the rotating mass. This type of crusher is best Company Cage-Pactor).
600 HANDBOOK OF POWDER SCIENCE

between rows of rotating bars. Again, this is Porticle nipped into

best suited for relatively weak nonabrasive Crushing Action
material such as coal.

12.3.2 Roll and Rod Mills

Figures 12.21, 12.22, and 12.23 show machines
suitable for intermediate size reduction. The
medium-duty hammer mill is especially suit-
able for sticky or tough materials that cannot
Figure 12.22. Smooth roll crusher.
be efficiently broken in rolls or rod mills,
because of its ability to shear in addition to
compress. The smooth roll crusher is widely surface, reenter the bed, and get carried up
used for size reduction in the laboratory, but it again. The rods rolling over one another act
also has industrial application in preparing like sets of rolls, stressing particles in a similar
material with a top size of, say, 12 mesh (1.70 manner. The power to the mill is used to lift
mm) and a minimum of fines (see Section the rods against gravity; the resulting potential
12.4). Because of abrasive wear the rolls have energy of position is converted to kinetic en-
to be resurfaced at frequent intervals if the ergy as the rods fall, which in turn is converted
crushed material is strong and abrasive. Again, to strain energy and, finally, to heat and sound.
it is important to control the feed rate into the However, there are two major differences
rolls to prevent the damaging forces arising between smooth roll crushers and rod mills.
from bed compression in place of steel- First, the rod mill is a retention device because
lump-steel crushing. The rod mill acts some- fine fragments have to pass along the mill to
what like a multiple set of rolls as the cylinder overflow at the exit and can be rebroken again
rotates; the bed of rods is carried up until it and again, so the mill is acting on a reservoir
lies at an angle to the horizontal. It is then of powder. In this type of device, the residence
unstable and rods start to roll down the bed time distribution (see below) of material in the
mill is of importance, and more fines are pro-
duced. Second, there is obviously no con-

Feed

Figure 12.21. Medium-duty hammer mill (Jacobson

Crusher Co.). Figure 12.23. Illustration of a tumbling rod mill at rest.
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 601

trolled gap setting or controlled power to the discharge for continuous wet grinding, while
turning rod, so it is not always possible to discharge through slots or grates that retain
break large, strong lumps, which can then leave the balls is often used for continuous dry
in the overflow. The force available for frac- grinding. For grinding coal, the mill is swept
ture is increased by making the steel rods with hot air to dry the coal and the fine coal
heavier (larger diameter) and the mill diame- removed in the exit air stream. Ball mills can
ter larger, but this is limited by excessive dam- be used for very fine dry grinding by air sweep-
age to the mill lining by the falling rods. Thus, ing, with return of oversize particles to the
the feed to a rod mill is normally less than mill feed from a high-efficiency (rotary) size
about 25 mm in top size, depending on mate- classifier cutting at a small size to give a high
rial strength, It is normally used for wet grind- circulating load.
ing. Abrasive wear on the rods means that
worn-down rods must be removed and re-
12.3.4 Autogenous and
placed with fresh rods at suitable intervals.
Semi-Autogenous Mills
12.3.3 Tumbling Ball Mills Autogenous tumbling mills are similar in prin-
ciple to the tumbling ball mill, but use the
Figure 12.24 shows the tumbling ball mill, also material being broken as the breakage media.
a retention mill, which is very widely used for There are four major types. The first is essen-
dry and wet grinding to relatively fine sizes. tially identical in construction to a ball mill,
The principle is identical to that of the rod but the feed consists of two streams, a narrow
mill, but the maximum force available to break size range of lumps of rock (e.g., 75 mm X 150
large, strong lumps is even less, so the feed to mm) and the normal fine crushed feed. The
the mill is rarely larger than 10 mm for strong large rocks wear to round pebbles (hence, the
rock. Because of its great industrial impor- name pebble mills) on tumbling and then act
tance this type of mill has been widely investi- like steel balls on the rest of the feed. The
gated, and is discussed in detail below. Abra- feed rate of large rock is adjusted to keep a
sive wear is easily handled by topping up the suitable load of pebbles in the mill. The sec-
charge with fresh balls at frequent intervals ond type has a large diameter-to-length ratio
and it is not necessary to stop the mill to add (typically 2:1) and takes a natural crushed
the balls. The mill shown has an overflow feed containing rock typically up to 200 to 300
mm, with discharge through slots of typically
20 mm width. Since the feed rate has to be in
balance with the rate at which the large lumps
break themselves to less than 20 mm by their
own tumbling action, it is not possible to vary
the product size distribution over a wide range.
In fact, the third type, semi-autogeneous mills,
are identical but add some charge of large (4
in. = 100 mm) steel balls, typically a few per-
cent of the mill volume, to increase output
capacity. The Scandinavian countries and
South Africa use a variant of this type with a
smaller diameter-to-length ratio (typically 0.5
Grate Discharge
Ball Mill to 1), which behave like semi-autogeneous
pebble mills.
Although very similar to tumbling ball mills,
Figure 12.24. Illustration of a tumbling ball mill at rest. autogeneous and semi-autogeneous tumbling
602 HANDBOOK OF POWDER SCIENCE

mills have some distinct features in their into a shaking mechanism is similar in princi-
breakage action. Since rock has a lower den- ple and very useful for preparing laboratory
sity than steel, the power input per unit of mill samples of fine powders. The planetary or
volume is lower than in ball mills, so the centrifugal mill15'16 contains two or more ro-
equivalent ball milling action is reduced. How- tating cylinders partially filled with balls,
ever, a gradual decrease of the size of large mounted at the periphery (parallel to the axis)
lumps of rock is not a typical disintegrative of a bigger cylinder or frame that is also
breakage but has a major component of a rotated. The respective speeds of rotation are
chipping action in which pieces are broken off set by gears to use the centrifugal force of the
irregular feed shapes to give rounded material. outer rotation to throw the balls across their
The rounded lumps then abrade until the size cylinders as they rotate, thus replacing gravita-
is small enough to be broken by a larger lump. tional fall with much higher centrifugal force
Both chipping and abrasion give small product and also greatly increasing the number of balls
fragments, so the mills give suitable qualities moved per unit volume and time. A fairly
of finely ground material even when the prod- recently developed mill16 accomplishes the
uct contains substantial amounts of very coarse same purpose with a single horizontal mill
particles.11"14 Autogenous mills have lower ca- shell mounted on an eccentric (with counter-
pacity for a given mill volume than a ball mill balance weights), with the radius of gyration
and, hence, higher capital cost per unit of chosen to produce the effect of a centrifugal
output, but they do not have the continuing field moving around the mill with each gyra-
cost of replacement steel balls. The use of tion. This gives a high-force tumbling action of
semi-autogeneous mills allows the best eco- the ball charge but avoids the high force on
the drive produced by the vibrating ball mill
nomic balance to be reached between capital
and is much simpler mechanically than plane-
cost and cost of replacement steel.
tary mills. The power input and capacity per
The fourth type of autogeneous mill, the
unit volume of the mill is very high and it is
rotary breaker, is specific for coarse size re- suitable for underground treatment of ores in
duction of coal. It has the added feature that mining tunnels, thus saving millhouse con-
the cylinder case is lightweight and contains struction costs. Abrasive wear is high and the
many holes (typically 50 to 300 mm), so that mill is designed for rapid replacement of a
material broken less than the desired top size removable lining in the mill.
falls through and forms the product. Coal is
light enough and friable, enough that self- 12.3.6 Roller-Race Mills
breakage by tumbling gives high output with-
Figure 12.25 gives an example of the class of
out requiring a heavy shell to withstand
mills known as vertical spindle mills or roller-
pounding and abrasion.
race mills. The rotating table throws material
through the roller-race and the pulverized ma-
12.3.5 Vibrating / Planetary / Centrifugal terial passes over the rim and is swept up by
Ball Mills an air stream flowing through the annulus
between the rim and the case. The stream
There are two other variants of the ball mill. passes to a classifier that returns oversize to
In the vibrating ball mill the cylinder is not the table, so that the rollers are acting upon a
rotated to cause tumbling but is packed almost fairly thick bed of material. The basic action is
full with balls and mounted on an eccentric that the rotation of the race pulls material
that jerks it around the cylinder axis, thus under the roller, the roller is driven by this
causing the balls to vibrate in the cylinder. material, and the bed of material passing un-
The mechanical stresses on the drive are high der the roller is nipped and crushed as it
and the mill is not conveniently scaled to high passes through the gap between the roller and
continuous capacity. A small ball mill fitted the race. The rollers are loaded with massive
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 603

Feed to ever, with a proper size range of feed matched

Classifier Coal Feed Coarse Particles to roll diameter they are efficient mills with a
From Classifier
lower energy consumption per ton of product
than many other mill types. The compression
of beds of powder by pulling material between
rolls is used in three recent developments:
the high-pressure grinding rolls machine (the
Schonert mill), the Szego mill, and the
Horomill. These are discussed in detail in
the section on new developments (see later).
Sweeping
Air
12.3.7 Hammer Mills
High-speed hammer mills, similar to the mill
shown in Figure 12.21, are also used for rela-
tively fine dry grinding of many nonabrasive
Heavy
Drive Housing Rejects materials, with much of the grinding action
Figure 12.25. Illustration of roller-race mill being by shear between the hammer tips and
(Krupp-Polysius Co.). the case. The mills are air swept with built-in
rotary size classifiers to retain coarser material
in the mill.
springs to give a high compressive force and
the gap automatically adjusts to a height such 12.3.8 Disc Mills
that the mass flow of material pulled in equals Figure 12.26 illustrates the disc mill, which
the mass flow of compressed material passing also consists of surfaces rotating at high speed
through the gap. The bed compression pro- with respect to one another, but with the gap
duces breakage just like putting a bed of parti-
cles into a cylinder and applying high pressure.
Fragments broken from the particles fit within
spaces between the feed particles and the Feed Inlet
porosity of the bed decreases to a minimum at
the gap where the pressure is highest (see
later).
These mills are widely used for coal grind-
ing, again with hot air drying and conveying,
and for raw material grinding for cement man-
ufacture. They give a produce size distribution
of about 80% < 75 /im, from feeds with a top
size of about 25 to 75 mm depending on the Feed
diameter of the rollers. The mills work well Entrance
only for brittle feeds of a natural size distribu- Between
Grinding Plates
tion: if the feed consists entirely of large lumps, Plate
Grinding
the rollers nip lumps rather than a feed bed. "Plate
As a result, the rollers ride up, then fall, the
mill runs roughly, and gives high abrasion of
both rollers and bed plates. If the feed con-
tains excessive fine material, mill operation is
again unstable with gouging, and slip of the Grinding Disks 7
rollers over the bed of powder occurs because Product Outlet
the bed develops fluid-like properties. How- Figure 12.26. Disc mill.
604 HANDBOOK OF POWDER SCIENCE

between the discs readily adjustable during then obtained, but at a cost of wall and media
operation. The force application is by shear wear.
and compression as particles move into the
narrower portions of the gap. There are sev- 12.3.10 Fluid Energy Mills
eral machines similar in principle but with
different plate geometry. Figures 12.28 and 12.29 shows types of fluid
energy mill, in which small particles are sus-
pended in high-velocity streams of air or steam
12.3.9 Stirred Media Mills
obtained by expansion through nozzles with
Figure 12.27 shows a sand mill or Attritor, inlet pressures of 5 to 10 atmospheres. In the
which consists of paddles turning in a bed of device illustrated in Figure 12.28, the tangen-
water and sand or small steel or ceramic balls. tial entry of high-velocity fluid creates a
The large number of grinding particles give doughnut of swirling particles and fluid in the
many breakage actions per unit time but the grinding chamber, which retains coarser parti-
breakage action is mild, and the mill is most cles by centrifugal action. The microturbu-
often used for comminution or deagglomera- lence of the gas stream causes high-speed
tion of small, relatively weak particles or ag- impact of particle-on-particle, and the cen-
glomerates, such as dyestuffs, pigments, clays, trifugal size classification allows only fine sizes
etc. A similar principle is used in the high- to leave the breakage zone. In Figure 12.29,
energy ball mill, with larger balls and high the opposed jets cause high-speed collision of
paddle speeds which give much higher forces the particles, and a size classifier and fan
and a high power input per unit of mill vol-
ume. These are used on a relatively small scale
for preparing mechanical alloys by dry grind- Hypothetical
Feed Injector Tangent Circle
ing of ductile metals. Larger versions are used
for fine grinding of limestone and other fairly
weak materials. In shear mills, slurry is flowing TOP
in a narrow annulus between a rotating drum VIEW

and a stationary cylinder, with breakage caused

by the high fluid shear forces across the annu-
lus. They are generally suitable only for small Fluid Inlet Jet Axes
weak particles or weak agglomerates. In some I 2"Drilled Orfices
mills, a wider annulus is filled by small media. Feed
Pressure Manifold Feed Manifold
More intensive and uniform grinding action is
Outlet
8 Feed
Inlets

Product Grinding Chamber

in Liquid
Concentric
Collector

Feed
Particles ag or Product
in Liquid Bin
Figure 12.27. Stirred ball-particle mill: Attritor. Figure 12.28. Fluid energy mill: Sturtevant Micronizer.
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 605

Finished
Product

•Screen

Rotary
Knife

Fan Air Stationary

Knife

Figure 12.30. Illustration of rotary knife cutter mill.

Feed
12.4 THE ANALYSIS OF SIZE
REDUCTION PROCESSES
Compressed
Air,Steamfor Gas Opposed Jets 12.4.1 General Concepts
Figure 12.29. Fluid energy mill: Majac Jet Pulverizer. It is clear from the previous section that the
multiplicity of mill types and breakage actions
make it virtually impossible to formulate a
system in the device returns larger sizes into
general theory of the unit operation of size
the jet stream. The mills are designed to give
reduction. In most cases good mill design
fluid boundary layers on the containing sur-
has evolved by trial-and-error starting from
faces, to reduce particle impact on the sur-
common-sense applications of the concepts of
faces and the consequent abrasion. The spe-
fracture. However, for devices that reduce
cific energy consumption calculated from the
large tonnages of material, using substantial
energy required for air compression or steam-
electrical energy, there is considerable impe-
raising is high compared to mechanical
tus for accurate process design rules and for
grinders, but the mills are capable of produc-
techniques for optimization of the system. As
ing very fine material (e.g., - 5 /mm) and are
in other unit operations, it is invaluable to
used primarily for specialty grinding of high-
construct mathematical models of the opera-
value materials or where cheap waste steam is
tion to aid in its understanding and optimiza-
available.
tion. In the last decades, considerable ad-
vances have been made in this respect using
12.3.11 Shredders and Cutters concepts very similar to those of chemical
Figure 12.30 illustrates a whole class of mills reactor theory.17'18 The mill is considered
designed specifically for size reduction of tough equivalent to a reactor that accepts feed com-
but nonabrasive materials such as polyvinyl ponents (the set of feed sizes) and converts
chloride, Teflon, rubber, wood, etc. They rely them to products (the set of product sizes),
on the cutting action, like scissors, between and a size-breakage rate (population) balance
rotating and static sharp edges with narrow is performed on the reactor.
clearance. The efficiency of this type of mill is The rate at which a material breaks in a
highly dependent on maintaining sharp cutting mill depends on its particle size as well as its
edges. Shredders, for example, for waste strength characteristics. Normally, for any
paper, and hogs for waste wood and bark fit given mechanical action there will be particle
into this category. A number of mechanical sizes that are too big for efficient breakage
arrangements are used. because the action is not powerful enough,
606 HANDBOOK OF POWDER SCIENCE

and particles that are too small for efficient

breakage because the statistics of applying the
action are not favorable. It is also apparent
that the specific energy (kWh/ton) used for
size reduction increases for breakage of finer
and finer sizes because (1) it becomes more
and more difficult to apply stress efficiently to
millions of tiny particles and (2) the basic
strength of brittle particles increases because
large flaws (which can be stress-activated to
fracture at low stress) become broken out as
grinding proceeds to finer sizes. It is necessary,
then, to analyze the breakage of each size
range. It has been found convenient to use a
yfl screen sequence to define the size ranges o.oi
0.031 0.063 0.125 0.25 1.0
(e.g., "size" is defined as 16 X 20 mesh, 20 X RELATIVE SIZE Xj/x:
30 mesh, etc.) because material in one of these Figure 12.31. Typical cumulative primary progeny frag-
size intervals appears to behave like a uniform ment distribution: ball milling of 20 X 30 mesh quartz.
material, to a sufficient approximation. Since a ( • ) dry; (O) wet.
geometric progression never reaches zero it is
necessary to define a "sink" interval contain-
similar to those of Figure 12.31 for many brit-
ing all material less than the smallest size
tle materials and machine types: it is not dif-
measured. Thus, a feed size range can be split
ficult to see that this form is compatible with
into n intervals, numbered 1 for the top size
the tree of cracks illustrated in Figure 12.14.
interval to n for the sink interval.
The slope y of the finer end of the B plot is
Using this basis, the size distribution from characteristic of the material and appears to
breaking a given "size" in one pass through be the same for all breaking sizes. In many
the device is called the progeny fragment distri- cases, the B values are size-normalizable, that
bution, and is conveniently represented in the is, the curves of Figure 12.31 fall on top of one
cumulative form "Dtj = weight fraction less another for different breaking sizes. Thus, the
than size xt from breakage of larger size ;," "weight fraction less than a given fraction of
where xt is the top size of interval /. Obvi- the breaking size" is constant and
ously, Djj = 1 and 1 - Dj+1} is the fraction of
size j that remains of size ; after passing; the
fraction of size / transferred to size / is dtj =
Dtj - Di+lj. The set of numbers di} is called For retention mills, the concept of specific
the transfer number matrix. For a once-through rate of breakage St is applicable. Consider a
machine such as a roll crusher these values mass W of powder in the mill, of which a
can be determined experimentally by crushing weight fraction w; is of size /. The specific rate
each size independently. of breakage S, for example, for size interval ;',
For a retention machine such as a ball mill, Sj is defined by:
it is extremely valuable to define a primary
progeny fragment distribution, Bij9 again cumu- Rate of breakage of size ; to smaller sizes
lative, which is the mean set of product frag- (12.24)
ments produced from one breakage action,
with the products then mixed back into the It has units of time" 1 and is comparable to a
mill contents to wait to be selected for a first-order rate constant in chemical kinetics.
second breakage, and so on. It has been found A batch grinding test on a feed of size / is
that the form of the primary B values is comparable to an homogeneous first-order
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 607

"rate-of-reaction" experiment, and if S is con- lumps is stressed and fractured independently

stant, Eq. (12.24) goes to of the other sizes, as it reaches the region in
the rolls where it is nipped and compressed.
d(wjW)/dt= -Sj For a given gap setting, xg say, breakage of
each size will produce a mean set of progeny
and
fragments denoted by DtJ. Then, for a feed
(12.25) consisting of weight fraction fl of size 1, f2 of
size 2, etc., the fraction pL of product in size /
Figure 12.32 shows a typical result. This is:
first-order relation is observed so frequently
that it can be called "normal" breakage,
whereas non-first-order kinetics indicate some or
"abnormal" feature. Methods of estimating S
and B values from experimental tests have (12.26)
been described,18 but they are primarily useful
for laboratory or pilot-scale test data, and it is n >i >1
at present frequently necessary to infer values where Pi is the cumulative fraction of material
for large devices by extrapolation from smaller less than xt in the crusher product. The logi-
scale results. cal analysis of how the D values vary with
conditions is given later in this section as
12.4.2 Mill Models an example of the analysis of once-through
devices.
The function of a mill model is to describe the
Retention grinders such as the tumbling
product size distribution. The model can then
ball mill are very important industrially, and
be used to assist in the analysis of the influ-
the mill model applicable to these is devel-
ence of design and operating variables on mill
oped as follows. First, consider the simplest
performance. For example, consider a simple
system of batch operation, with the powder
once-through device such as the smooth roll
getting finer and finer for longer and longer
crusher (see Fig. 12.22). It can be assumed as a
grinding times. Using the concept of primary
first approximation that each size of the feed
progeny fragment distribution joined with Eq.
(12.24) the "net rate of production of size i
material is the sum of its production from all
larger sizes minus its rate of breakage," or

W W
dt

n>i>j>l (12.27)
where W is the mass of powder in the mill and
bitj is the primary progeny fragment distribu-
tion in the interval form, bi}, = Bi}•, - Bi+lj.
o S 16x20 = 0.606 Min.-' This set of n equations is known as the batch
A S4Ox5O=O.29O Min" 1
S|40x200 s 0-088Min.-' grinding equation. If btj and St do not vary
with time, it has the solution:18'19
10 15 20

GRINDING TIME (MINUTES)

Figure 12.32. Typical first-order plot of batch grinding

data, various sizes of cement clinker.
(12.27a)
608 HANDBOOK OF POWDER SCIENCE

where the set of transfer numbers dt j is com- mill was plug flow, Eq. (12.27) would still apply
puted from the algorithms with a grind time r of r = W/F, F being the
mass flow rate through the mill. However,
(0 retention mills will generally have a residence
e-stt time distribution (RTD) defined by <£(0 dt =
=j
weight fraction of feed in at time 0 which
a
2^ i,kaj,k leaves between time t and t + dt. This is due
k=j to mixing in the mill which brings some feed
quickly to the discharge, while other material
is back-mixed to the feed end and leaves later.
• E "i,kaj,k
k=i Figure 12.34 gives an example determined by
using a pulse of radiotraced powder in the mill
l 'i feed and counting at the mill exit.21 Then the
^TT E $kbi,kak,j steady product size distribution will be made
up of material ground for all times over the
RTD range, in a weighted sum:18
The equations are programmed20 for compu-
tation on a PC, and the solution starts with
/ = 1, then / == 2, etc., using the feed size (12.28)
distribution w((0). Figure 12.33 shows the com-
puted solution compared to the smoothed ex-
where wt(t) is the solution of Eq. (12.27) for
perimental points for grinding of a narrow
the mill feed. For a fully mixed mill the mass-
feed size, using experimentally determined
rate balance is "the rate of flow size / out =
values for S and B.
rate of flow size / in plus rate of production of
Second, consider a retention grinding ma-
size / by breakage of all larger sizes minus rate
chine where the powder flows uniformly, is
of breakage of size /." Thus,
ground, and is then fine enough to exit through
an overflow or grate without preferential re- i i
tention of larger sizes. If the flow through the (12.29)

Rogers/Gardner
— Semi-infinite

COMPUTED
O EXPERIMENTAL

100
1 2
SIZE /im Dimensionless time, t*
Figure 12.33. Comparison of computed to experimen- Figure 12.34. Residence time distribution for a 4.57 m
tal size distributions for batch grinding. diameter X 9.2 m long wet overflow discharge ball mill.
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 609

However, pj = Wj for a fully mixed system bled (k = 2), the required residence time is
with no size classification at the mill exit. halved. Thus, there can be similitude between
Using T = W/F a small mill and a large mill, with only a
/-i difference in time scale. The same result is
ft + r
£ hjsjPj obtained for batch or plug flow grinding, and
for Eq. (12.28) providing the RTD is normaliz-
able with respect to T, that is, 4>(t/r) is the
1+ same from one mill to another.
(12.29a) The use of these models is illustrated below.
This set of equations is readily computed se- Experimental measurement of the variation of
quentially starting at / = 1. the values of Sj with mill conditions is the
The variable used in the computations is the most explicit and logical means for describing
mean residence time T, and any model can be mill operation and mill efficiency.
computed for a range of r values. Since T = It is useful to have an approximate mill
W/F, the value of r that gives the desired model that is simple enough for quick-hand
product size also specifies the mass W neces- calculations. The results of Figure 12.33 allow
sary to get a desired production rate F. Then the deduction that Bond's "law"22 applies to a
the mill size needed to contain W is calcu- reasonable approximation,
lated. Of course, it is also necessary to have lOOjum 100 ^
equations that give mill power, in order to = mpt/W=El\
v
determine the specific energy of grinding. 80F
An important general conclusion can be (12.30)
reached by considering Eqs. (12.27) or (12.29)
applied to a comparison of two milling systems where m p is the shaft mill power, JC80P is
operating on the same feed. Suppose that the the size in micrometers at which 80% passes
B values are the same between the two sys- that size in the product, x 80F is the 80%-
tems, but that S values are different by a passing size of the feed, and the energy index
constant factor, S- = kSt. Using Eq. (12.29a) EY is determined from the data. E is the
as an example, applied to both mills, specific energy of grinding (kWh/ton) re-
quired to go from a specified feed of x 80F to a
desired product of JC80P. This empirical equa-
KisjPj tion enables rapid estimation of the grinding
time or specific energy to go from any feed to
(Mill 1)
any product, assuming that Ex is a constant. It
does not give any information on the size
p-= distribution of the product nor does it take into
account the size distribution of the feed. As
(Mill 2) might be expected, Ex is not closely constant
Substituting for S't in the second equation, from one mill to another, or for different mill
conditions. As used in practice, Ex is deter-
mined for a given material from an experi-
Pi ment under standard conditions23 using an
empirical correlating equation that converts it
(Mill 2)
to the value expected for an 8-ft diameter wet
Obviously, p\ = pt when krr = T, that is, an overflow ball mill operating in closed circuit.
identical set of size distributions is produced El is then known as the Bond Work Index WY,
in mill 2 as in mill 1 but with residence times which has the physical meaning of the hypo-
decreased by the factor k. If S values are dou- thetical kWh/ton necessary to go from a very
610 HANDBOOK OF POWDER SCIENCE

large feed to 80% passing 100 ^m, in the 8-ft These are used in conjunction with the appro-
diameter mill. Empirical correction factors priate mill model to predict the circuit product
based on prior experience are used to allow size distribution from a mill circuit
for different conditions and mill diameter.22 simulation.18
Figure 12.37 shows one interesting result
12.4.3 Mill Circuits: Classification from a simulation of a tumbling ball mill. If a
mill circuit is designed to produce a size distri-
In industrial practice, mills are frequently used
bution passing through a control point (if/%
in closed circuit, where the mill product is
passing size x*) from a given mill, then this
passed through a size classifier that gives two
specification can be met by a suitable feed rate
exit streams, a coarser stream returned to the
through a classifier with set st values, or by a
mill feed and a finer stream, which is the final
different feed rate with the classifier adjusted
product. The operation of the classifier is best
to cut at smaller sizes (and, hence, give more
described by the set of classifier selectivity
recycle and a larger C value). It is seen that
numbers, st, defined as the weight fraction of
there is a permitted band of size distributions
size / presented to the classifier that is sent to
through the control point, from C = 0 to C =
the coarse stream. These are readily calcu-
oo. Austin and Perez24 have shown that the
lated from experimentally measured size dis-
limiting (steepest) size distribution obtained at
tributions of the three streams.18 Figure 12.35
high circulating load depends only on the pri-
gives a typical example. It can be seen that a
mary progeny fragment distribution. Thus, it is
typical classifier is not ideal. It sends some
a material characteristic and it is not possible
coarse material to the product and returns
for a customer to specify a steeper distribu-
some fine material back to the mill. The
tion. The higher circulating load also gives a
smaller the value of d50, the bigger the overall
higher circuit output rate Q tph (tons/h). The
fraction of the classifier feed that is directed
physical reason for these effects is that a high
into the recycle stream. The relation between
flow rate through the mill, F = (1 + C)Q,
the circuit feed and product and the mill feed
brings fine material rapidly to the classifier
and product is shown in Figure 12.36: defining
and removes it before it is overground. Thus,
the circulation ratio by C = T/Q, then
the mill contents contain on the average less
0= fines and more coarser material, and coarser
material breaks faster than fine material. The
and
general reason for closed circuit operation is to
remove particles that are already fine enough, to
prevent energy being wasted on grinding them
even finer.
100
The return of fine material back to the mill
feed, due to the apparent bypass of the classi-
fier as shown in Figure 12.35, decreases effi-
' / ^ Measured;
Measured selectivity ciency by leading to overgrinding. In principle,
*^ curve s(Xj)
this can be compensated by higher circulation,
^ Ideal Classification but in practice (1) it may not be possible to
(S.I. = 1.0)
pass enough mass through the mill to ap-
-S.I. = 0.6 proach this limit without overfilling the mill
leading to poor breakage action and (2) in-
creased mass flow through a classifier may also
increase the bypass fraction, thus defeating
100 500 1000
Size (microns) the action. For these reasons it is advanta-
Figure 12.35. Illustration of selectivity values of a size geous for a classifier to approach as closely as
classifier: a is an apparent bypass. possible the ideal classification shown in Fig-
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 611

CLASSIFIER

Figure 12.36. Normal closed circuit.

ure 12.35. The function of efficient classification small size in retention devices such as ball
is to reduce the proportion of fine material by mills and roller-race mills. As fine material
avoiding overgrinding of fines. The concept of builds up in the bed of powder, the breakage
indirect inefficiency is that although a mill may of all sizes slows down. This appears to be
be operating efficiently in transferring input partly due to coating of the grinding surfaces
energy to breakage it can be inefficient if that but principally due to a cushioning action. In
energy is used to break material that already dry grinding, it is argued25 that the agglomera-
meets specifications. tive forces between fine particles impart a
fluid-like nature to the bed that can absorb
12.4.4 Non-First-Order Grinding and impact without giving high stress to particles
Slowing of Grinding Rate directly under the stressing surfaces. This can
be likened to trying to grind particles sus-
It can be reasoned from fracture mechanics
pended in a sponge; the energy of a falling ball
and the difficulty of efficiently stressing unit
or passing roller is spread over a large elastic
mass of very small particles that the specific
mass instead of being concentrated on a small
rates of breakage are smaller for small parti-
mass of solid. In addition, air trapped in such
cles than for larger ones. This has been con-
a bed cannot rapidly flow out of the bed in the
firmed for every type of mill investigated to
path of the stressing surface because of the
date. However, there is an additional effect of
high drag forces, so it moves away carrying
particles with it, much like a liquid parting to
100p let a solid ball fall through.
It is sometimes possible to predict the cor-
C =D
rect product size distribution even in the pres-
ence of slowing-down effects, by performing
the simulation with a false residence time 6
that is less than the real residence time t. A
10 slowing-down factor K can be defined by K =
0/t, which then also represents the ratio of
the actual mean value of 5,- from time 0 to t
Q.
to the first-order value St. Figure 12.38 shows
values of K for four different materials, plot-
0 1 2 3 4 5 6 7 8 ted against the fraction of fine material less
C
, I. , , , l , , , , , , , , . I , . i than 10 fim in size. It is apparent that differ-
10 50 100 500 1000 ent materials develop the slowing-down pro-
Size fjjn
cess at different amounts of fines. The magni-
Figure 12.37. Permitted band of size distributions pass-
ing through a desired point, with varying circulating tude of the effect can be seen from Figure
load. 12.39, where it takes 20 min to reach a size
612 HANDBOOK OF POWDER SCIENCE

similar phenomenon is observed with wet

grinding of slurries at high solids content. In
very fine ball milling there may be changes in
• 0.8
the primary progeny distribution as well.26'27

• 0.6 12.4.5 Analysis of Smooth Roll Crushers

The unit operation of crushing does not usu-
0.4 ally give problems for brittle materials such as
rocks and ores, so it has not yet received the
a Cement Clinker (typeA) amount of theoretical analysis given to fine
£0.2 o
v
Cement Clinker (type B )
Lower Kittanning Coal
grinding. However, the general concepts of the
a West Kentucky No. 9 Coal process engineering analysis of crushers can
be illustrated by using smooth roll crushers as
O.I 0.5 I.O 5.0 I0.0 50.0
an example. There are five facets to the analy-
PERCENT LESS THAN
sis: (1) the correct feed size, (2) the maximum
Figure 12.38. Representation of slowing down of rate force required, (3) the capacity in tons per
of breakage with build-up of fines, K values, for dry
grinding in a batch ball mill.
hour, (4) the product size distribution, and (5)
the maximum power required.
Gaudin28 gives the relation between angle
of nip ©, coefficient of friction rj, particle size
distribution which would have been obtained x, gap x g , and roll diameter d as
in 7 min if grinding had stayed first-order.
Since the rate of energy input to the ball mill tan(®/2) V
is almost constant, the slowing-down process d (12.31)
produces greatly reduced grinding efficiency cos(®/2) =
d+x
and leads to high specific grinding energy. A
See Figure 12.40. If x is too big the particle
will not be nipped since © will give
100
tan(©/2) > rj. Austin et al.29 have pointed
out that there is little published information
on the effective values of 17 between the
crushed materials and the rolls as a function
of material, roll speed, surface roughness, etc.
When the gap is small compared to roll diam-
eter, Eq. (12.31) gives
X
- < (1 + 7j 2 ) 1/2 - 1 (12.32)
d
to assure nipping of feed size x. For a value of
7] = 0.5 this states that the maximum lump
size should be less than 1/10 of the roller
diameter.
The maximum force tending to separate the
500 1000
SIZE >im
rollers can be estimated by assuming the worst
possible case, that is, simultaneous compres-
Figure 12.39. Comparison between computed and ex-
sion to failure of the maximum lumps of size
perimental size distributions of 20 X 30 mesh Lower
Kittanning coal ground for different times in the ball xm at all places along the rolls. Assuming that
mill. 6 = first-order time, t = real time. the lumps are small compared to the roll di-
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 613

at that location. Since the geometry of the

system requires that xg < xc it is clear that the
material is consolidated as it passes towards
the gap. To avoid reaching a highly consoli-
dated bed that acts like a noncompressible
solid, the feed rate is controlled (nonchoke
feeding) to give a high porosity of the feed so
that compression to the gap size gives porosi-
ties greater than about 0.3, thus <2max ~
0.7pweg. In practice, Q is even lower, as deter-
mined by experience. It depends on the forces
required to consolidate the bed and typical
values for coals30 are 0.2 to 0.9 of <2max, with
lower values for strong materials and smaller
Compression
rolls.
If rolls are operated at different speeds, the
arithmetic mean is used for u. Different roll
speeds are sometimes used to give an extra
component of shear in addition to compres-
sion. If the feed contains lumps too big to be
fully nipped, such lumps will build up in the
Figure 12.40. Illustration of nip angle in rolls and
crushing-shear forces.
mill inlet and will eventually abrade to a size
which allows them to be nipped and pulled in
completely. This, of course, reduces the feed
ameter, unit length of rolls contains l/xm large rate. In circumstances where it is desirable to
lumps each requiring the force P* given by use too large lumps (to reduce the number of
Eq. (12.23a). Thus, stages of crushing, for example), ribbed or
toothed rolls are used to increase the ability of
Maximum force = x m cr c L/4 (12.33)
the rolls to pull in larger lumps, especially for
where <xc is the compressive resistance defined relatively weak and friable materials.
by crc = P*/(xm/2)2, and L is the length of A method of predicting the product size
the rolls. distribution obtained from any feed has been
28
Gaudin has described the concept of ca- proposed by Austin et al.29 They showed that
pacity calculation. Assuming that the feed is in the values of the transfer numbers dtj depend
sizes less than xm the maximum capacity (at on the ratio of the particle size x to the gap
choke feeding) per unit length of rolls is given size x g , providing the roll diameter is much
by the ribbon of solid which can be pulled bigger than x and xg. Thus, values of dtj
through the rolls: measured experimentally for one gap setting
can be converted to any other gap setting.
Qmax = uP(i - eg)xg = uP(i - ec)xc Then Eq. (12.26) is used to calculate the prod-
(12.34) uct size distribution. However, this method
requires the experimental measurement of the
where 1 - 0g is the volume fraction of solids total dtj matrix, which is very time consuming.
at the gap, p the true solid density, and u the To reduce the description of the dtj matrix
circumferential velocity of the rolls. Obviously to a few parameters, they assumed that the
(1 - 6g)xg = (1 - 6c)xc, where 6C is the breakage processes could be described as
porosity of the feed picked up by the rolls at shown in Figure 12.41. First, feed particles of
the critical angle of nip and xc the value of x size j have a probability Sj of breaking and a
614 HANDBOOK OF POWDER SCIENCE

so it is expected that s't will be greater than or

equal to st.
Product
Considering the repeated fracture of size 1
material, 1 - st falls through the rolls to
size 1 product, and s1 breaks. The material
resulting from the breakage follows two
Figure 12.41. Equivalent circuit for a once-through roll
routes—material that passes the gap to give
crusher with multiple fracture actions.
product and material retained to fracture
again. Let atj = btj(\ - sp, which has the
probability 1 - Sj of falling through the rolls physical meaning "when size j breaks, atj- is
without fracture: clearly, the fraction of larger the fraction sent to product size /." Let ctj =
sizes that fall through the roll gap without btjs'i which is "when size ; breaks, ctj is the
breakage is zero, Sj, = 1, and sizes much fraction sent to size i for a further breakage."
smaller than the gap do not break, Sj = 0. Then, the broken quantity st distributes itself
Experimental values are shown in Figure 12.42
as a2l m a31
and were found to fit the empirical relation
1
in size 4, and so on; a41 is the product from
6.6 (12.35)
breakage of size 1 to size 4, c21a42 is 1
1 + breaking to 2 breaking to 4, c31a43 is 1 break-
ing to 3 breaking to 4, c2Xc32a43 is 1 break-
where d50/xg is characteristic of the material. ing to 2 breaking to 3 breaking to 4, and so on.
Second, it is assumed that all sizes break into Thus,
a normalized primary progeny fragment distri- dhl = 1 - ^
bution b;_j, where bx is the weight fraction of
breakage products of one size that appears in
the next lower size, b2 appears in the size d3X = c21a32) (12.36)
below that, etc. Third, it is assumed that a d41 =
fragment of size i produced by fracture in the
rolls has in turn a probability s't of being
etc., until c values become zero. The equation
rebroken or 1 — s't of passing through the roll
is readily converted to dtj replacing 1 with ;
gap. Since this material results from fracture it
and 2 with ; + 1, etc. Then the total size
is already in a favorable position to be nipped,
distribution from a feed of ft is obtained from
Eq. (12.26).
i.o Austin et al. 29 ' 31 treated the above problem
o Primary bypass somewhat differently by developing the mass
A Secondary by-pass balance equations for the equivalent circuit of
Figure 12.41 as if st and s'( were due to exter-
nal classifiers and they developed a method
0.5 o for calculating btj values from the test data.
They found that the values of bi_j in the
cumulative form fitted the empirical function
(see Fig. 12.31):

0.71
RELATIVE SIZE
28
x;/x;
1.0
^-•(ir) +«-«(ir
(12.37)
Figure 12.42. Measured primary bypass (fraction un-
broken) and estimated secondary bypass for feeds of y2 where 4>, y, /3 are characteristic parameters
screen intervals of Lower Freeport coal. for the material, as shown in Table 12.1.
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 615

Table 12.1. Characteristic Breakage Parameters Determined from Smooth Roll Crusher Tests31

MATERIAL 7 P d5o/xg
Rhyolite 0.29 0.83 3.6' 1.45
Diabase 0.40 0.84 4.0 1.40
Coals
Shamokin anthracite, PA 0.30 1.05 5.0 1.70
Illinois # 6 0.36 0.81 3.0 1.66
Ohio # 9 0.33 0.95 4.2 1.93
Western Kentucky # 9 0.47 1.05 4.0 1.81
Belle Ayre, Wyoming 0.49 1.17 4.0 1.70
Pittsburgh E. Seam, PA. 0.32 0.81 3.0 1.66
Upper Freeport, PA. 0.39 0.96 4.0 1.56
Lower Freeport, PA. 0.50 1.05 4.5 1.54

By a trial-and-error matching of computed They also simulated the effect of passing

size distributions with experimental values they the product through a screen and recycling
determined that s't could be estimated from s( above-size material to the roll feed. The mini-
values by mum production of fines was obtained when
the gap and the screen were of the same size
i < ig - 1 (see Figure 12.44) even though the circulating
s't = j (sig_1 + sig_2)/2 i = ig - 1 (12.38) load was relatively small. Larger gap settings
and the associated high circulating load pro-
duced very little change in the final product.
where ig is the interval number corresponding This is because fine material is not acted on by
to the gap setting. Thus, a simulation model is the crusher as it passes through, so a high
constructed for smooth roll crushers which has circulating load is no advantage.
the experimentally determined material char- The process of fracturing unit volume of
acteristics of <l>, y, /3, and d50/xg. Figure feed to less than the gap setting requires
12.43 shows a typical match of computed stressing the original broken volume (5X) of
versus experimental results.

FEED

0.25 0.5 1.0 2.50 5.0

SIZE mm
Figure 12.44. Simulated circuit product size distribu-
Figure 12.43. Crusher product size distribution from tion for 3 X 30 mesh Illinois # 6 coal as a function of
3 X 12 mesh feed. gap setting at ideal screening of 12 mesh.
616 HANDBOOK OF POWDER SCIENCE

particle size 1, stressing again the fraction of later, fracture by bed compression in place of
this volume that undergoes a second fracture, steel-particle-steel nipping fracture tends to
stressing again the fragments of these frag- produce size distributions with proportionally
ments that undergo a third fracture, and so more fines than expected; also additional en-
on. The total stressed volume is readily calcu- ergy is used in the bed compression.
lated as s1 plus the sum of all c terms, that is, The capacity and product size distributions
of other crushers can be analyzed in a similar
fashion.33"36 For example, a jaw crusher acts
+C
2,1 C 3,2 + C
2,1 C 4,2 on a maximum solid volume rate of A{\ -
+ C 2,1 C 3,2 C 4,3 )• 6c)u, where A is the throat area, 6C is the feed
" t " C 3,l C 4,3
porosity, and the velocity of flow u is deter-
If it is assumed that the strain energy per unit mined by the fall of solid under gravity as the
stressed volume required to produce fracture jaw opens. There is repeated breakage and fall
is a constant, which is known32 as Kick's "law," as the material moves down the crusher until
the total stressed volume is proportional to the it passes the gap which is a mean of the open
ideal specific energy required to grind size 1 to and closed side settings. The analysis is similar
less than the gap setting. Defining a reduction for gyratory crushers, although the rotational
ratio by xx/xv Figure 12.45 shows the relation motion can aid the rate of material moving
of the volume of repeated crushing to reduc- down.
tion ratio. In practice, it is usually found that a
larger reduction ratio requires a bigger in- 12.4.6 Analysis of Tumbling Ball Milling
crease of specific energy than that predicted
by Figure 12.45 because smaller lumps become 12.4.6.1 Influence of Mill Conditions
relatively stronger (require higher stress to The tumbling ball mill is the most widely used
cause breakage). device for fine grinding of brittle materials on
If the crusher is run nearer to choke feed- an industrial scale. Because of its simplicity, it
ing then breakage owing to bed compression is mechanically reliable, which is very impor-
becomes an additional factor. As we will see tant in continuous process streams, and it is
available in sizes ranging from small labora-
tory mills to industrial mills of 5 m diameter
by 10 m long, or even larger. It is a retention
device, where a bed of powder is acted upon
by the tumbling balls and the mean residence
time of solid in the bed is typically a few
minutes to 30 min depending on the desired
degree of size reduction. It has certain disad-
vantages. First, the mill power is almost inde-
pendent of the level of filling by the powder,
so a mill operated at lower than design capac-
ity is inefficient because (1) if the powder level
is held at a normal level, a low solid feed rate
gives a long residence time (r = W/F\ and
the energy is used to grind finer than neces-
2.O 5.O 2O.O sary and (2) if the level is dropped to keep r
Reduction Ratio, x,
constant, the energy is used to tumble balls on
Figure 12.45. The total crushed volume per unit feed balls without enough powder between them,
volume for roll crushing of a coal (Upper Freeport)
also giving excess ball wear. Second, the cost
through a smooth roll crusher, as a function of the
particle size to gap size ratio. of replacing steel balls as they wear is substan-
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 617

tial, and the steel or rubber lining of the mill O.9

has also to be replaced every 2 or 3 years.

Third, the "slowing-down" process comes into
play for very fine dry grinding or fine wet
grinding of viscous pulps. Thus, grinding can
become inefficient and consume high energy.
The major fraction of the direct power re-
quired to turn the mill (excluding motor and
drive losses) is used in the act of raising the
balls. On the other hand, the more balls raised
per unit time, the higher the rates of breakage
of powder in the mill because the tumbling of
the raised balls gives the breakage action.
Thus, the variation of power input with mill
conditions is likely to be a direct index of the
best breakage conditions. Figure 12.46 shows
typical power variation with rotational speed
and with ball filling. The critical rotational
speed is defined as the speed where balls on O.2 O.4 0.6 O.8 I.O 1.2
Fraction of Critical Speed
the case would start to centrifuge and is read-
ily shown to be Figure 12.46. Variation of net mill motor power with
critical speed as a function of ball loading: 2-ft diame-
76.6 ter laboratory mill.
critical speed =
]/D - d, in feet
42.2 however, are normally run at / < 0.4 to pre-
rpm vent balls blocking the overflow or the feed
/ D — d, in meters
entry.
(12.39) The specific rates of breakage can be deter-
where D is mill diameter, and d is ball diame- mined in laboratory or pilot-scale mills by
ter. Figure 12.46 shows that the power passes batch tests with controlled powder and ball
through a maximum at about 80% of critical filling, and controlled pulp density if wet.
speed. This varies somewhat with mill diame- Figure 12.47 shows a typical result for the
ter and ball load because the force of a heavy variation of St with particle size. The rates of
ball charge acting on the case tends to prevent breakage are low for sizes that are relatively
slip between the balls and the case (thus aid- large with respect to the ball diameter because
ing the raising of balls) for larger mills. For (1) the particles are so big that the force
large mills with steel balls the rotational speed required to break them is achieved only by
is usually in the range 65% to 75% of critical relatively few of the tumbles and (2) the parti-
speed to avoid cataracting of balls onto the cles are too big to be nipped by a ball-ball
mill case, which can damage the mill lining. collision [see Eq. (12.31)]. Small sizes also
The figure also shows that the maximum power break slowly because (1) their basic strength is
is obtained at about 45% filling of the mill higher due to removal of large flaws and (2)
volume by the ball bed at rest (calculated the mass of particles captured in a ball-ball
assuming the ball bed has a porosity of 0.4), collision becomes smaller and smaller as parti-
/ = (M/p b F)(l/0.6), where V is mill volume, cle size decreases with respect to ball size.
M mass of balls, p b true density of ball mate- Large ball diameters are better for breaking
rial, and / is the fraction of mill volume filled large particles but small balls are better for
by the bed of balls. Continuous overflow mills, breaking small particles because there are
618 HANDBOOK OF POWDER SCIENCE

many more ball-ball collisions for a given the ball density (the balls must have a hard
mass of small balls than for the same mass of surface)
large balls. This means that there is an opti- the load of powder or suspended solid in
mum mixture of ball sizes in the mill to go the mill
from any feed size distribution to any ball mill the rotational speed of the mill, as a frac-
product. tion of critical speed, and the lifting action
The slope a shown in Fig. 12.47 is charac- of mill lifters built into the mill lining
teristic of the material. It is also found that the slurry density and viscosity in wet milling
the primary progeny distributions in the the dispersing action of chemicals used as
first-order breakage region, which occurs to grinding additives
the left of the maxima in the curves, can be
fitted by Eq. (12.37), and the values of <1>, y, plus, of course, the diameter and length of the
and /3 are also characteristic of the material. mill. In addition, the degree of recycle and the
Examples are given in Ref. 18. Especially, a efficiency of size classification or air (gas)
material with a small value of y will produce sweeping to remove fines are also important
proportionately more fines on grinding. factors to prevent overgrinding or the develop-
ment of slowing-down effects. For example,
12.4.6.2 Major Variables tests show that a ball mill that is underfilled
The major variables involved in ball milling, in with solid is inefficient because the breakage
addition to these material characteristics are: zones where balls collide with balls or the case
are not filled and energy is wasted by steel-
• the ball loading in the mill on-steel collisions. On the other hand, over-
• the distribution of ball sizes in the mill and filling by powder or slurry is also found to be

1 1 1 1 1 1 I I 1 1 1 1 1—1 11

BALL DIAMETER IN mm _
/

UJ
Q / \
o
21.1 . >J
<
Ul / /
cr -
/
m /

o
UJ
0.5-
-
/yOL y
/
V
i-
/
/
O 1 9 \ -
/ \
U. j
/
O / / / /
Ul 25.A 31.8 38.1
Q. \
0.1 i i i 1 I i 11 1 1 1 1 1 >i I I 1 ^
0.1 1.0 10
PARTICLE SIZE Xj.mm
Figure 12.47. Predicted variation of Sz values with particle size for different ball diameters: copper ore (\/2~
intervals).
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 619

inefficient because it appears to cushion the expressing it as a formal "grinding pressure" P

breakage action. A general rule-of-thumb is defined by
that the solid should just occupy the interstices <P
of the ball bed calculated with the bed at rest. P=— (12.40)
Inefficiencies of this type, or the use of a La
mismatch of ball sizes to the particle size, or where L is the length of the roller and d its
the use of too dilute or too concentrated slurry, diameter. Figure 12.48 illustrates the geome-
etc., are examples of direct inefficiency as dis- tries involved, which shows that the system is
tinct from the indirect inefficiency of over- like a choke-fed single roll crusher operating
grinding. The reader is referred to Ref. 18 for against a flat plate. It can be treated in a
more detailed discussions of the effects of the somewhat similar fashion to the double roll
major variables in ball milling. Since this infor- crusher system, except that the gap xg is not
mation is fairly up-to-date, it will not be re- set but is a natural consequence of the mate-
peated here. More recent work includes ex- rial pulled under the roller, that is, it is a
tended treatments of the optimization of the floating roller.
distribution of ball sizes in the mill,37 the Immediately, the mass flow under the roller
influence of slurry density in wet ball is given by Eq. (12.34) modified to include
milling,38'39 the mass transport of slurry roller length
through a ball mill40'41 and predictive equa- Q = puLiX - 6) = puUl - 6c)xc (12.41)
tions of mill power.42'43
Models for autogeneous and semiautoge-
neous grinding mills are not so well developed
although the basic principles are very similar
to those for other tumbling media mills. Re-
cent work on constructing these models44"46
includes the kinetics of chipping of large rock
to form smaller pebbles, self-fracture of rock
by its own tumbling action, mass transport
through discharge grates, and mill power
equations.
mm
12.4.7 Analysis of Roller-Race Mills
TABLE
t t
ii ••
-y —

The type of mill exemplified in Figure 12.25 is (a)

the second most important type of mill (after
tumbling media mills) from the aspect of the
tonnage of material ground annually. A recent
analysis47 has given a detailed account of the
powder technology associated with this type of
mill and the analysis is summarized here. The
rotation of the table (race) brings material
under the rollers, which ride up and rotate as
the material passes underneath. Since the
rollers are heavy and are loaded by massive
springs, there is a vertical force acting down
on the roller that generally depends on how (b)
high the roller is forced against the springs.
Figure 12.48. Illustration of roller geometry and nota-
Let this force per roller be denoted by <P. The tion for the analysis of roller-race mills (race is called
force is scaled with respect to mill size by table).
620 HANDBOOK OF POWDER SCIENCE

where 6C and xc are defined where slip ceases is compressed to zero porosity, emax = 6C from
and material is pulled in without further slip Eq. (12.44). Then Eq. (12.43) becomes
and moves at the horizontal table velocity u. If
)de
there are no large lumps in the feed (to avoid P=\-rL (12.47)
chatter of the floating roller), the material is
pulled in as a bed and crushed by compression
using Eq. (12.45) and its differentiation ($ c in
of the bed. Until the bed is nipped for crush-
radians). Thus, the strain at the gap under a
ing there is very little work done on the mate-
grinding pressure P is determined by the func-
rial. The vertical compression pressure is es-
tion P(e) and the critical angle of nip,
sentially zero at the critical angle of nip (f)c,
but it increases as the material moves toward
the gap and reaches a maximum at the gap P= | — |/,(€„) (12.47a)
where the degree of compression is highest,
where Ix is the integral of Eq. (12.47), which
0 = 0. Let the resolved vertical pressure at <f>
increases as eg increases.
be denoted by P(<f>). Since the critical angle of
nip for bed crushing is less than 12°, sin $ ~ (/> Now consider the work done as the column
and cos 0 « 1.0 and the total vertical force is of powder is compressed. By integrating force
times the distance the force moves, from <j>c to
ILd\ r(i)e (f> = 0, it is readily shown that
* = i — U P(<l>)d4> (12.42)
uLd \ r(k

From the definition of formal grinding pres- }f*s() (12.48)

sure where m p is the net mill power per roller.
With the same substitutions as before,
(12.43)
g
Consider a thin vertical column of powder mp = C P(e)de (12.49)
nipped at </>c and moving at velocity u toward
the gap. Define a linear strain e by the frac- or
tional change in vertical dimension e = (12.49a)
(xc — x)/xc. The relation between strain e mp =
and porosity 6C between </>c and <fi = 0, with Thus, mill power is proportional to the velocity
e = 0 at <f)c, is of the roller, its length and diameter, and the
number of rollers. The compression character-
g
(12.44) istics of the material being ground enter via
i-eg
the critical angle of bed nip, the feed porosity,
A simple geometric construction gives the re- and the function of Eq. (12.46). The effect of
lation between e and <f> as grinding pressure also depends on the bed
compression characteristics, as can be seen by
substituting Eq. (12.47a) into Eq. (12.49a),
(12.45)
I2(€g)/I1(€g) (12.50)
Let the relation between the vertical pressure It is convenient to put Eq. (12.50) in the
P((f>) and the linear strain e be the (unknown) form
function "stress = function of strain," that is,
mp=$uLdP (12.51)
P(c/>)=P(e) (12.46)
where <j) is a dimensionless factor called the
where e is the strain at </>. A hypothetical specific power factor (per roller). Since a bed
maximum strain emax is defined when the bed becomes more difficult to compress further
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 621

once it is partially compressed, the value of $ approximately proportional to P over a lim-

generally decreases as grinding pressure is in- ited range of P.
creased. The factor is constant only at suffi- It must be understood that Q is the rate of
ciently low grinding pressures where the rela- material being crushed per roller, not the flow
tion of Eq. (12.46) is linear, P(e) = Ce, since rate in and out of the mill. The centrifugal
then the integrals become Ix = (f)Ce g and action of the table is constantly throwing ma-
h = (l^Ceg a n d <£ = (|)^c* For a typical terial out of the race, where it is swept up in a
angle <£c of 12° = 0.21 radians, <£ = 0.079 (per high-velocity air stream.. Larger particles fall
roller). back into the race as the gas velocity decreases
It must be realized that the formal grinding above the annulus, and larger particles (and
pressure defined by P = force/Ld is much some fines) are returned to the race from the
smaller than the actual maximum pressure at built-on classifier at the top of the mill. Thus
the gap. The average pressure over the region the mill can be considered as a fully mixed
(£c to 0 is 2P/(f)c, but this is the integral of a retention mill where there is breakage action
sharply rising stress-strain curve, so much under the rollers and a reservoir of powder
higher stress exists at the gap. Bed compaction not under the rollers. Let pt be the product
involves fracture of particles, with small prod- size distribution out of this reservoir of weight
uct fragments fitting into the interstices of W, ft the size distribution of feed into the
larger particles. reservoir, and wt the size distribution within
The flow rate under the roller will also the reservoir. A mass breakage rate balance
generally decrease as grinding pressure is in- on material entering and leaving the breakage
creased because xc becomes smaller as xg zones and the reservoir gives
becomes smaller. Again, it is convenient to
express the flow equation, Eq. (12.41), in the
form:
Q = puLd(l - 0g)(xg/d) = mpuLd (12.52) (12.55)
where m is a dimensionless factor called the
where F is the feed rate in and out of the race
specific capacity factor. It is readily shown from
and FN is the rate in and out of N rollers in
the definition of strain and Eq. (12.44) that
the race.
1 - cos <bc \ For the fully mixed assumption wt = pt, and
m=\ || 1 - | (12.53)
rearranging Eq. (12.55) using r = W/F gives
Since the strain at the gap increases as grind-
ing pressure increases via the relation of Eq.
(12.47), the value of m decreases. It should be Pi
noted that the roller-race cannot be operated r(FN/W)(l - du)
at very low pressure because there would be (12.56)
insufficient downward force to prevent slip where the apparent primary breakage distribu-
between the roller and the powder bed, and tion "bt j is defined by the breakage products in
the table would not transfer rotational velocity one pass under the roller,
to the roller.
The specific grinding energy E for material h (12.56a)
flowing under the roller is mp/Q and
Comparing with the usual equation, Eq.
E= (12.54) (12.29), it is seen that the specific rate of
mp
breakage is given by
Since both </> and m decrease as grinding
pressure is increased, the value of E can be S. = (l -diJ)(FN/W) (12.57)
622 HANDBOOK OF POWDER SCIENCE

and classification and recycle in the usual manner18

can then be solved for any value of F, and the
value of F adjusted to give the desired final
7=1
Pi = (12.58) product size. If the total circulation ratio is C,
1+
the actual k g / s of final product, Q say, is
where FN is given by NQ, Q being the flow given in the usual way by:
rate under each roller at choke feeding, Eqs.
(12.52) and (12.53). Thus, it is seen that the Q = F/(l + C) (12.60)
roller-race mill is one mill where it is possible where C is defined as the ratio of mass flow of
to describe S( and bitJ- values in terms of a final product to material returned to the race
precisely known breakage zone. from internal or external size classification.
The values of St and ~bt j have been deter- Tests on a limited number of U.S.A. coals
mined in a laboratory scale roller-race mill gave the following empirical relations for AiT
and the equations enable these data to be and ~bt j as a function of the Hardgrove Grind-
scaled for pilot-scale and full-scale simula- ability Index and the grinding pressure:
tions, as follows. To start, because it is not HGI
/ \
easy to determine the mass of the reservoir of
A0T = 0.17211 + 1.08 -jQJj-k (12.61)
powder in an operating mill, it is convenient to
replace S( values with the absolute rate of where P is expressed in MPa and A0T is the
breakage At defined by At = StW. At has the absolute rate of breakage of 18 X 25 mesh
dimension mass/time, for example, kg/s, and (1 X 0.841 mm) coal in kg/min:
is physically the instantaneous rate of break-
age of size i (under specified conditions) if all (12.62)
of W were of size i. Equation (12.58) then
becomes where x is the standard size of 1 mm and a
0
is the material characteristic given by
Pi = (12.58a)
1+At/F a = 0.58 - (2A)(10-3)HGI (12.63)
Then, from Eqs. (12.52), (12.53), and (12.57) and the characteristic breakage distribution
A i = NQ(1 - dit) (12.59) parameters of Eq. (12.37) are given by
or 0 =5 |

( 1 - e max \ I 1 - cos cf)c y = 1.23 - (2.32)(10~ 3 )//G/ \ (12.64)

0 = 0.58 + (2.6)(10" 3 )//G/ j
xd-dtjXpuLd) (12.59a) The values were determined for test condi-
At the same grinding pressure in the labora- tions of 7VT = 2, uT = 0.0565 m/s, dT = 0.060
tory mill as in the full-scale mill it can be m, L T = 0.016 m, and a sufficient depth of
assumed that the bed compression eg, the bed to ensure choke-feeding to the two rollers.
hypothetical maximum strain emax, the critical Some comments can be made. First, the
angle of nip (f>c and the degree of breakage in specific rates of breakage for coal ground in
one pass under the roller, 1 - dit, are the the laboratory mill (dT = 0.060 m) are shown
same since these values depend on grinding in Figure 12.49. It is seen that the simple
pressure, not the size of the roller. Thus At power function of Equation (12.62) does not
values are scaled by apply to larger sizes, where xt/d is greater
than about 1/25. The increased breakage rates
uLdN .
lT (12.59b) above this size are due to the greater ability of
\uTLTdTNT/ a roller to nip single particles than to nip a
where the suffix T refers to the laboratory test bed of fine feed. The decrease at even larger
conditions. Equation (12.58a) combined with sizes is due to the inability of the rollers to nip
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 623

the rise of the roller forces against the spring

: " i ' ' ' i i i 1 L

and the grinding pressure increases

P = P0 + kxg + Mg (12.65)
where k is the spring constant of the precom-
pressed spring and g is the gravitational con-
O.I stant. For example, an industrial mill with
%
steel rollers of 1.22 m (48 in.) diameter and
•s -_
^^•^SLOPE « d
K
0.43 m (17 in.) length subjected to a preload
Q ~
per roller of 1.8 X 105 Newtons (40,000 lbf)

-PREDICT!
will have a minimum grinding pressure of
about 0.42 MPa. However, such a roller is
0.01 , , , | , , ', i i i expected to rise about 38 mm ( « 1.5 in.),
io3
and with a spring constant of 0.72 X 104
SIEVE SIZE, |im Newtons/mm (40,000 lbf/in.), this will give an
Figure 12.49. Absolute breakage rate of 18 X 25 mesh extra grinding pressure of about 0.48 MPa,
Elkhorn coal as a function of particle size. that is, the total grinding pressure per roller is
about 0.9 MPa. Equation (12.58a) shows that a
lower mill capacity F gives a finer product size
distribution. However, the equation is valid
only with almost constant At values as long as
particles that are larger still. However, feeds
the reservoir W in the mill is sufficient to
containing particles too large in reference to
choke-feed the rollers. If the feed rate is made
the roller diameter are avoided in practice
too small, the value of W will fall below this
because they give rise to chattering of the
level as the rotating race throws material out,
rollers. Second, the linear increase in specific
Q in Eq. (12.59) will change to a lower value
breakage rates with increasing grinding pres-
and F and At each change by the same factor.
sure cannot be extrapolated to high grinding Then the product size distribution will not get
pressures because the coals (especially soft finer and, in fact, the smaller raise of the
coals) will cake onto the rollers and cause slip, rollers will reduce the grinding pressure, cause
which leads to loss of energy as frictional heat less breakage and the product size distribution
instead of causing breakage. Third, the frac- may get coarser, as demonstrated by Austin et
tion of particles of a given size that do not a| 48-50 The m m power w m fall a s t h e r0Hers
break in one pass under the rollers are rein-
are underfed and, to get fine product, it is
corporated into a new bed fed into the next
necessary to have a race designed to retain
pass and can break at the same specific break-
powder, plus efficient classification to give a
age rate, thus preserving the first-order nature high rate of recycle to the bed. Finally, the
of the breakage kinetics. Every reapplication empirical equations for A0T, a, /3, y, and 4>
of grinding pressure will cause further break- are based on limited data and it is advisable
age. A typical result is that the feed to the mill for values to be determined directly for any
is rolled over about 10 times before it leaves coal or other material under study.
the classifier as final product.
Fourth, in practice the rollers in an indus-
trial mill are generally loaded with massive 12.5 NEW MILLS
springs initially compressed to a preload, and
any material passing under the roller is sub- 12.5.1 High-Pressure Grinding Rolls Mill
jected to this minimum grinding pressure, Po New designs of mills are constantly being
say, plus the weight of the roller, M say. patented and constructed in small-scale ver-
However, as the bed is pulled under the roller sions, but most are variants on existing mill
624 HANDBOOK OF POWDER SCIENCE

designs and operate with the same fundamen-

tal principles. However, there are several new
mill designs that result in large part from the
investigations of Professor Klaus Schonert in
Germany. By studying the breakage of powder
beds by compression in a piston-cylinder sys-
tem, he showed51 that the specific energy of
size reduction was significantly less (20% to
30%) than that for tumbling ball mills, that is,
the use of energy to cause breakage was more
efficient in this type of system. This is because
the particles are all subject to the stressing
action and less energy is wasted (1) by impacts
that are not sufficient to cause fracture, (2) by
steel-steel collisions that do not trap particles
for breakage, and (3) due to frictional losses
from powder and media movement in uncon- Figure 12.50. Principle of the high pressure grinding
fined systems. To apply this principle in prac- rolls (HPGR) mill (KHD-Humboldt Wedag).
tice, he invented a mill that is essentially a
double roll crusher that has one roll free to
move against a large applied force and that is while compacted softer materials such as coal
choke-fed from a hopper above the gap. This tend to stick to the rolls and have to be
type of mill, called a high-pressure grinding rolls removed with scalping blades. The mill has
(HPGR) mill, has been developed commer- also been used for grinding diamond-bearing
cially by the Krupp-Polysius Company and rock to liberate the diamonds since there is
(under license from Schonert and Krupp- less breakage of the strong diamonds and more
Polysius) by the KHD-Humboldt Wedag preferential fracture along a diamond-rock
Company, both of Germany, and by others interface.53
(also under license). Figure 12.50 shows the Although very different in appearance, the
principle:52 both rolls are driven by electric basic action is very similar to that of the
motors connected by special couplings that roller-race mill discussed previously. The feed
permit the free roll to move in its containing material pulled in the rolls is nipped with a
tracks and the force is applied by an hydraulic critical angle of nip, compressed (which causes
pressure system which allows for easy control breakage) to a maximum high pressure at the
of the grinding pressure. The very high stresses gap, and the gap automatically adjusts to pass
at the gap require that the rolls be of strong the compressed cake. As we have already
and hard material to avoid surface cracking noted, the roller-race mill is more efficient
and reduce abrasion and the rolls must be than many other types of mill, and so is the
thick enough to withstand the strain. HPGR mill: confined compression of beds of
The mill has been very successful in the particles is generally more efficient than other
cement industry as a pregrinder to conven- grinding methods. The major differences be-
tional long (tube) tumbling ball mills. When tween the roller-race mill and the HPGR mill
grinding cement clinker at formal grinding are (1) that the critical angle of nip for the
pressures of 2 to 6 MPa, the resulting com- HPGR mill (two rolls of equal diameter) is
pressed powder passes through the gap as a half that for the roller-race (flat surface) mill,
coherent strip that can then be deagglomer- (2) the grinding pressures used in the HPGR
ated with a hammer mill or in a following ball mill are several times higher, and (3) the
mill. Much harder materials do not briquette, HPGR mill always has two rolls. Austin47 has
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 625

shown that allowing for these differences gives

descriptive equations identical to those for the
roller-race mill and that it is possible to take
data from a laboratory HPGR mill and predict
the mill power and capacity of roller-race mills.
Usually a roller-race mill performs many re-
peated compressions, each at relatively low
pressure, while the HPGR mill performs one
compression at high pressure. The specific en-
ergy of grinding for a given duty is probably
very similar whichever method is used. This is
especially true when the HPGR mill is oper-
ated in closed circuit with a classifier.
Nearly all the comments made about
roller-race mills also apply to HPGR mills. For
example, both mills will chatter if feed sizes
are too large for the roller or rolls diameter. Figure 12.51. Studded roller surface for autogenous
wear protection (KHD-Humboldt Wedag).
Both machines give flat product size distribu-
tions with a relatively high proportion of fines
because fine material produced in the initial gap, owing to compression of the air contained
compression is further broken as the material in the bed which is escaping upward. It has
is pulled into the gap. Overgrinding of fines also been suggested47 that the fines impact a
will be less in the roller-race mill if a high fluid-like property to the dry bed so that in-
recirculation from an efficient classification is stead of moving into the rolls (or under a
used, because of the lower pressures. The sim- roller) as a locked bed, the bed can shear and
pler mechanical design of the HPGR mill collapse. Under these circumstances the pres-
makes it easier to scale to high capacities and ence of some moisture may be beneficial by
KHD offers sizes up to 1.7 m roll diameter providing capillary forces between particles,
and 2 m roll length with capacities up to 500 but the water content must be low enough to
tons/h. 53 There are two major disadvantages allow free bed compression.
to the mills. Unlike tumbling media mills where
wear of steel media is readily compensated by
frequent addition of fresh media without stop- 12.5.2 The Horizontal Roller Mill
ping the mill, wear on rollers, races and rolls (Horomill®)
has to be corrected by dismantling the equip- Figure 12.52 shows another form of bed com-
ment when wear has progressed too far. The pression mill, recently introduced by the FCB
Krupp-Polysius Company designs mills where company of France. The mill is specifically
this disassembly and roller replacement or designed for dry grinding of cement clinker
resurfacing with segment sections can be done and consists of a horizontal stationary roller
rapidly. KHD-Humboldt Wedag54 have rollers that rotates on its axis, inside of a horizontal
fitted with hardened studs that hold com- mill cylinder that is driven to rotate on the
pressed cake on the rolls, thus giving autoge- mill axis. The grinding portion of the roller
neous surface protection (see Figure 12.51). presses with a controlled force against a grind-
Another disadvantage is that feeds containing ing track on this inside of the mill cylinder.55
a high proportion of fines give rise to erratic Like roller-race mills, the mill is a retention
mill operation and, hence, machine vibration. mill where the level of material in the mill is
This has been suggested to be caused by flu- controlled to give choke feeding of the gap
idization of the bed as it passes toward the between the roller and the track; the roller
626 HANDBOOK OF POWDER SCIENCE

Figure 12.52. The Horomill® (FCB Groupe Fives-Lille).

will float to pass the material pulled into the an advantage for roller-race mills when used
gap, and the product passing under the roller on softer materials such as coals which tend to
will mix into the reservoir of material and be form strong compacts under high pressure.
reground by repeated passes under the roller.
Dry powder flows out of the mill and is lifted 12.5.3 The Szego Mill
in a bucket elevator to a high-efficiency air The original concept is due to the late L. L.
classifier, with return of coarse material to the Szego and the mill has been developed in
mill feed. Toronto, Ontario by General Comminution,
The grinding pressure is quoted as "mod- Inc., in close collaboration with University of
erate" and the mill is not air-swept like a Toronto researchers in the Department of
conventional roller-race mill. The comments Chemical Engineering. As a result, while in-
made on roller-race mills and high-pressure dustrial utilization of the mills is still modest,
grinding rolls apply also to this mill and the there is a great deal of published material
mills will probably give similar specific grind- available. The mill is a planetary ring-roller
ing energies, although the power used for clas- mill, consisting principally of a stationary
sification is probably higher for air-swept grinding cylinder inside which a number of
roller-race mills. It is easier to ensure choke- helically grooved rollers rotate, being flexibly
feeding in the Horomill® and in the HPGR suspended between flanges connected to a
mill as compared to roller-race mills where the central drive shaft (see Fig. 12.53).
rotating table both drives the rollers and The material is fed by gravity, or pumped
throws material into the air stream, but the into a top feed cylinder if wet, and is dis-
deagglomeration and rapid removal of fines is charged continuously at the bottom of the
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 627

the central axis in the mill cylinder, with a

higher velocity giving a higher force per unit
Particles to be ground
mass of roller.
An important feature is the ability of the
Helically grooved roller
roller grooves to aid the transport of material
rotates clockwise, driven through the mill, thus providing a means to
by friction on grinding shell
control the residence time, the number of
times material is rolled over, and, hence, mill
Stationary grinding shell of capacity and product size distribution. This
hardened steel
transporting action is particularly important
with materials that do not readily flow by
gravity, such as pastes and sticky materials.
Material to be ground fed
into mill The mill has several design variables that can
Bearing assembly be utilized to meet specific product require-
ments. The important variables are the num-
Roller shaft ber of rollers, their mass, diameter and length,
and the shape, size, and number of starts
Roller of hardened steel
radially mobile
of the helical grooves on the rollers. Increase
in the number of starts gives a steeper angle
for the helical grooves. As the number of
rollers is increased, the product becomes finer.
Heavier rollers and higher rotational speeds
generate the greater crushing forces which
Ground product
leaving mill may be needed for strong materials. The
ridge/groove size ratio can be changed to
increase or decrease the effective pressure act-
Figure 12.53. The Szego Mill (General Comminution
Inc.).
ing on the particles. The common groove
shapes are rectangular and tapered; the latter
will decrease the chance of particles getting
mill. The feed particles are repeatedly crushed stuck in the grooves.
between the rollers and the stationary grinding If several passes through the mill are
surface. The crushing force is created mainly required to get a sufficiently fine product,
by the radial acceleration of the rollers; shear- multiple-stage mills can be used that have
ing action is induced by the high velocity gra- several sets of rollers fixed onto the same
dients generated in the mill. Hence, the pri- rotor. This allows various design combinations
mary forces acting on the particles are the of different roller sizes and ridge/groove size
crushing and shearing forces produced by the ratios in different stages for optimal mill per-
circumferential motion of the rollers. The formance. The operating variables for the mill
basic action of the mill, is somewhat similar to are the material feed rate, its consistency (if
that of the roller race mill and the HPGR mill. wet), and the rotational speed of the rotor.
The rollers rotate about their own axes, pull Typically, the rotational speed is between 400
material under the rollers with a critical angle and 1200 rpm, depending on equipment size,
of nip, and pass out compressed broken mate- which translates to roller velocities of 6 to 10
rial. The rollers will float away from the sta- m/s.
tionary grinding cylinder to a gap that depends Most work with the Szego Mill has been
on the centrifugal force and the compression done on the grinding of coal in oil or water 56
properties of the bed. The force on the rollers for the preparation of coal-slurry fuels. Lime-
is controlled by the speed of rotation around stone, mica, talc, and other filled materials
628 HANDBOOK OF POWDER SCIENCE

have been tested,57'58 as have various waste Other combined processes tested involve
materials, for example, hog fuel,59 sawdust,60 grinding and extraction, applied to oil extrac-
and waste paper,61 the latter for use as a tion from rapeseed (canola);68 and simultane-
reinforcing filler in cellulose-plastic compos- ous grinding and reaction, in a coal liquefac-
ites. Wet grinding of grains, as a preprocessing tion study.69 When a thick slurry is being
step for hydrolysis and fermentation to alco- ground and a very fine product is required, a
hol,62 is another interesting application. The continuous recycle system without classifica-
mill is characterized by high capacity per unit tion is used since classification is very difficult
volume and modest power consumption. It is at high slurry or paste viscosity. The mill is
very versatile; in wet grinding it can also han- then run long enough to give the product the
dle highly viscous materials such as thick desired fineness. Metals have been ground that
pastes, that is, high solids concentrations, way down to submicron flake thicknesses.70
without extreme loss of efficiency.63 Within A significant effort has been expended on
reason, not only particle size distribution but mill modeling. This includes performance
also particle shape can be controlled, for ex- modeling using the population balance ap-
ample, from granular to flaky.64 proach,71'72 with breakage functions and
Another group of applications involve grind- grinding kinetics for single and multipass
ing combined with other operations or pro- grinding for both wet and dry operation. A
cessing. The simultaneous grinding and ag- dynamic model73 of fluid flow between a roller
ridge and the stationary grinding cylinder has
glomeration (SGA) process, 65 ' 66 as an
been made for wet grinding. The centrifugal
example, combines grinding and selective oil
forces are balanced by pressure development
agglomeration of coal with oil in water for coal
in the squeezed film of paste; the model al-
beneficiation. In the conventional process, de-
lows, currently for a Newtonian fluid, com-
veloped at the National Research Council of
putation of the total dynamic force field,
Canada, oil or a hydrocarbon solvent is added
velocities, shear stresses, etc., as well as the
to finely ground coal in water. Intense mixing
clearance between the roller-ridge and the
breaks the oil into fine droplets and allows the
grinding surface. Integration of these events,
hydrophobic coal particles to collect onto the in combination with a confirmed mechanism
droplets, leaving the hydrophilic ash (noncom- of material transport through the mill, allows
bustable mineral matter) behind in the water.67 prediction of the residence time distribution
A period of milder stirring allows the coal-oil and an upper limit to the product particle size
particles to grow into larger spherical agglom- distribution.73
erates for separation from the aqueous phase
Szego Mills are available in laboratory and
by screening or other means. The combined
pilot sizes as well as in small industrial sizes
SGA process uses the Szego Mill to replace with throughputs of 1 to 10 tons/h. Compared
the grinding and high-shear mixing steps, with to a ball mill, throughput per unit volume in
considerable equipment simplification and en- the Szego Mill is some 30 times higher and the
ergy savings,66 with results comparable to the specific power consumption due to the high
conventional process. Other grinding mills power density is typically 30% lower, as is
such as ball or agitated media mills are not characteristic of bed compression mills. While
suitable, as the sticky agglomerates would coat the Szego Mill is a compact and efficient
the balls and either reduce the grinding effi- grinder for many applications, very hard and
ciency greatly or block the mill, whereas the abrasive materials excluded, its special niche is
Szego Mill will operate owing to the positive grinding wet at high solids loading; a tooth-
transporting action of the roller grooves. The paste-like consistency appears to be the best.
objective of those studies was to make benefi- Special mills have been built for operation at
ciated coal-oil-water slurry fuels as an oil high temperatures and pressures, further en-
replacement in industrial or utility boilers. hancing the range of applications of this mill.
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 629

12.5.4 The DESI Mill ter of the rotors passes through the working
zone within a few hundredths of a second. The
This mill is another example of a mill that uses
particles are disintegrated by collision with the
a principle similar to that of an existing type of
mill but that incorporates changes allowing multiple rows of grinding elements and by
it to embrace also new applications. It has particle-particle attrition in the air stream.
been developed in Estonia by the company The grinding elements serve as targets for the
Desintegraator and is in use in various parts of colliding material and as accelerators for the
the former Soviet Union, with applications next collision (see Fig. 12.54). The material
ranging from industrial minerals to fuels to typically undergoes two to eight collisions with
biological materials. A great deal of work on the grinding elements.
the mill has also been done at the Tallinn Whereas many mills, including the HPGR
Technical University, but there are relatively mill, break particles by internal tension pro-
few publications, and most of these are in duced by compressive forces applied relatively
Russian. During privatization in the early slowly, in high-speed impact mills, the DESI
1990s, the original company was broken into included, breakage occurs by a different pro-
smaller entities and information is available cess of producing tension. The particles expe-
from the Desintegraator Association or from rience free, unrestricted impact at high veloc-
DESI-E Ltd., both in Tallinn, Estonia. ity, typically in the 30 to 200 m / s range in the
Invented by the late Dr. J. Hint some 40 DESI. (It has been shown by Vervoon and
years ago, the DESI mill was first used with Austin75 that pellets moving at 30 m / s reach a
the development of silicalcite, a strong build- maximum impact force within a few microsec-
ing material made of sand and lime ground onds after impact when they strike a rigid
together. Mechanical activation imparted to target containing a force transducer). An in-
the materials by the mill accounts for its high tensive compression wave starts from the area
strength; the development of both silicalcite of contact and surges through the particle at
and the mill is described in a 600 page mono- high velocity, with the stresses exceeding the
graph by Hint.74 The DESI is an impact mill normal compressive strength of the particle.
comprising of two rotors moving at high speed When the compression wave reaches the op-
in opposite directions. Thus the mill has the posite side of the particle, it is reflected as a
same principle as the Cage-Pactor mill shown tension wave of the same intensity. The parti-
in Figure 12.20 but it is specifically designed cle then starts to break up. The multiple prop-
for fine grinding. The material fed to the cen- agation of waves in the particle and its

51

Figure 12.54. Operating principle of the DESI impact-roller mill (DESI-E Ltd.).
630 HANDBOOK OF POWDER SCIENCE

fragmentation are believed to activate the ma- amount of work has been done on wear, with
terial chemically.76 Hence, mechanochemical many combinations of both target and abra-
activation of the material occurs which may sive particle materials as well as velocity, parti-
have beneficial effects on downstream process- cle size, impact angle, etc.77
ing, or even for simultaneous grinding and The main unique feature of this type of mill
reaction. Such activation effects have been is the ability to mechanically activate many
observed with chemical catalysts, building ma- materials.78'79 Such a claim is supported by
terial (e.g., silicalcite), fertilizers, and in vari- extensive research; a more recent presenta-
ous biological systems. The DESI mill can be tion80 has summarized some of this work, in-
used for selective grinding of weaker compo- cluding mechanical activation of polymers and
nents in a heterogeneous material by judicious biological systems in the disintegrator. Mill
selection of the speed of rotation to give im- design and operating conditions were related
pact forces between those required to break to the resultant activation. Again, custom de-
the respective materials.76 Besides effective sign is essential, for the desired objectives and
grinding, the fast rotation of the grinding ele- the particular materials, in situ reactions or
ments in opposite directions allows excellent enhanced downstream processing. Of course,
micromixing of solids or solids and liquids. the same comments can be made about high-
The mill can also be used to treat sticky mate- speed hammer mills, which operate at similar
rials since the powerful centrifugal forces dis- impact velocities.
courage adhesion.
For fine, and especially ultrafine grinding,
12.5.5 The Nutating Mill
the DESI mill is used with a built-in aerody-
namic classifier, which recycles coarse material This mill is being developed by the Warmley
for regrinding. The fine product enters a col- company in Australia,81'82 specifically for dry
lector and de-dusting system. DESI mills are or wet grinding at high power density of brittle
available in a wide capacity range, from small materials such as metalliferous ores. It has
laboratory units with capacities of 5 to 10 several similarities to the planetary and cen-
kg/h through to industrial units with capaci- trifugal mills16 described previously since it is
ties up to 100 t/h, the latter for limestone a mill that uses grinding balls at high g forces,
grinding in a DESI 31 M-8 mill. The total but these forces are produced in a different
assembly weighs 14 t, with gross dimensions, way. The mill shell is in the form of an in-
m, of 4.5 length, 2.6 width, and 2.4 height, verted cone, with feed from above into the
including motors, and a power rating of 500 to narrow end of the cone. The shell is rotated
1200 kW. about the center line of the cone, which is at
There are many DESI mills in industrial use an angle to the vertical. This axis is mechani-
covering a number of applications, with a range cally forced to rotate at the same time to form
of quoted product particle sizes varying from the surface of a narrow cone with the tip of
90 wt% < 5 /im to 90 wt% < 3 mm. Many the cone at a fixed point on the vertical (just
more materials have been ground in labora- like the earth rotating on its own axis but also
tory settings down to the micrometer size. moving in orbit with its axis not perpendicular
Apparently, most units are custom-designed, to the plane containing the orbit path). This
with the number of rows as well as size and wobbling planetary action produces high g
inclination of the grinding elements being im- forces and rapid movement around and across
portant variables in addition to the rotor di- the mill of the balls inside. The mill grinds
ameter. The mill rotors are self-balancing and very rapidly because of the high forces and the
the grinding elements are reinforced with high power density and the feed discharges at
wear-resistant ceramics: chamber walls are also the large end of the cone. The mill is capable
reinforced where required. An extensive of very fine grinding by adjusting the feed and
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 631

discharge rate to give a long mean residence new types of mill is proceeding, of course, but
time while maintaining an appropriate hold-up until this research produces industrially impor-
of powder or slurry to avoid steel-on-steel tant results it falls outside of the scope of this
collisions. As with all high power density mills chapter.
using grinding media, the wear rate of media The methodology of characterizing a size
and shell liners is high and the energy effi- reduction operation by examining the specific
ciency is not going to be better than that of a rates of breakage and the primary progeny
more conventional tumbling media mill, but fragment distributions has proved very infor-
the mills are small for a high capacity. High mative. Again, however, there are no precise
power density machines are especially suited descriptions of why the values of S, and Btj
for very fine grinding, to avoid having to use a vary in the ways observed. The variations are
large machine to give a small amount of suit- often sensible from simple physical reasoning,
able product. The application of the concepts but the quantitative relations involved are still
of mill modeling to the nutating mill is well essentially empirical.
advanced and it is possible to predict optimum The choice of a certain crusher-mill combi-
conditions, capacities, and product size distri- nation for a given job is generally made intu-
butions from tests on a new material in a itively at present; the choice is not the logical
laboratory-scale mill. result of a precise set of rules or calculations.
Programming of the calculations for computa-
tion with current desktop computers and avail-
12.6 FUTURE WORK able software is not the problem: it is inade-
quate systemic, quantitative descriptions of
It is still true that much work remains to be how machines and materials behave that pre-
done to raise the technical understanding of vent full use of the techniques of mill and mill
the unit operation of size reduction to that of circuit simulation.
the other (perhaps fundamentally simpler) unit The mechanisms of the slowing down of size
operations such as heat transfer, distillation, reduction that is observed as fines accumulate
absorption, etc. The mechanical stressing con- remain to be investigated in detail, and this
ditions inside mills are complex, and the frac- branch of investigation will undoubtedly in-
ture and disintegration of natural materials is volve the nature of the cohesive interaction
a complex phenomenon. It must be empha- between particles, dry and in dense slurries,
sized that for size reduction we are concerned and the effect of grinding additives on these
not only with the conditions at which fracture forces.
occurs but also the size distribution of the set The better utilization of many ores, fuels,
of fragments resulting from the fracture. and other materials in the future may involve
The conversion of electrical energy via me- requirements of mechanical reduction to ul-
chanical action to surface energy of fracture is trafine sizes. This represents a branch of inves-
thermodynamically very inefficient. However, tigation that has come to the fore but that
based on the industrial requirements of cost, poses many problems in theory, experimental
throughput, wear, and reliability of operation, technique, and engineering design.
it is difficult to see how to improve existing
devices substantially or how to invent new
ones with much greater efficiency. The mate- REFERENCES
rial in this chapter has been limited to the
powder technology relevant to crushers and 1. A. Nadai, Theory of Flow and Fracture of Solids,
McGraw Hill, New York, p. 89 (1950). See also
mills that are in commercial operation with Developments in Fracture Mechanics, Vol. 1, edited
proven benefits for particular applications. Re- by G. G. Shell, Applied Science Publishers, London
search on different methods of breakage and (1979).
632 HANDBOOK OF POWDER SCIENCE

2. A. A. Griffith, "Phenomena of Rupture and Flow New Grinding System and its Applications." Proc.
in Solids," Philos. Trans. R. Soc. Lond. 221A:163 14th IMPC, Toronto (1982).
(1920). 17. L. G. Austin, "A Review Introduction to the De-
3. A. A. Griffith, "The Theory of Rupture," Proc. scription of Grinding as a Rate Process," Powder
First Int. Conf. for Applied Mechanics, Delft (1924). Technol. 5:1-17 (1971/72).
4. G. R. Irwin, Fracture Dynamics: Fracturing of 18. L. G. Austin, R. R. Klimpel, and P. T. Luckie, The
Metals, American Society of Metals (1948); Orowan, Process Engineering of Size Reduction: Ball Milling,
E., "Fracture and Strength of Solids," Reports of AIME, New York, 561 p (1984).
Progress in Physics, Physical Society, London, 72:185 19. K. Reid, "A Solution to the Batch Grinding Equa-
(1949). tion," Chem. Eng. Set 20:953 (1965).
5. R. P. von Rittinger, Lehrbuch der Aufbere- 20. T. Trimarchi and L. G. Austin, "A Ball Mill Circuit
itungskunde, Ernst v. Korn., Berlin (1857), quoted Simulator in Object-Oriented Programming," avail-
in many surveys of grinding theory. able from the Mineral Processing Section, Depart-
6. H. E. Rose, private communication (1964). ment of Mineral Engineering, The Pennsylvania
7. I. J. Lin and S. Nadir, "Review of the Phase Trans- State University, University Park, PA 16802.
formations and Synthesis of Inorganic Solids by 21. R. S. C. Rogers and R. P. Gardner, "Use of a
Mechanical Treatment," Mat. Set Eng. 39:193-209 Finite-stage Transport Concept for Analyzing
(1979). Residence Time Distributions of Continuous Pro-
8. J. S. Benjamin, "Mechanical Alloying," Set Am. cesses," AIChE J. 25:229 (1979).
234:41-48 (May 1976). See also: C. Suryanarajan, 22. F. C. Bond, "Crushing and Grinding Calculations,"
Bibliography on Mechanical Alloying and Milling, Brit. Chem. Eng. 6:378 (1965).
Cambridge Interscience Publ., 380 pp. (1995). 23. C. A. Rowland, Jr. and M. M. Kjos, "Rod and Ball
9. N. H. Macmillan, "Chemisorption Induced Varia- Mills," in Mineral Processing Plant Design, edited by
tions in the Plasticity and Fracture of Non-metals," A. L. Mular and R. B. Bhappu, AIME, New York,
in Surface Effects in Crystal Plasticity, Nordhoff, pp. 239-278 (1978).
Leyden, p. 629 (1977). 24. L. G. Austin and J. W. Perez, "A Note on Limiting
10. A. R. C. Westwood and J. J. Mills, "Application of Size Distributions from Closed Circuit Mills,"
Chemo-mechanical Effects to Fracture-dependent Powder Technol. 76:291-293 (1977).
Industrial Processes," ibid., p. 835. 25. L. G. Austin and P. Bagga, "An Analysis of Fine
11. L. G. Austin, C. A. Barahona, and J. M. Menacho, Dry Grinding in Ball Mills," Powder Technol
"Fast and Slow Chipping Fracture and Abrasion 25:83-90(1981).
in Autogenous Grinding," Powder Technol. 26. L. G. Austin, M. Yekeler, and R. Hogg, "The
46(l):81-87 (1986). Kinetics of Ultrafine Dry Grinding in a Laboratory
12. L. G. Austin, N. P. Weymont, C. A. Barahona, and Tumbling Ball Mill," Proceedings of Second World
K. Suryanarayana, "An Improved Simulation Model Congress Particle Technology, Kyoto, Japan, p
for Semi-Autogenous Grinding," Powder Technol. 405-413 (September 1990).
47(3):265-283 (1986). 27. L. G. Austin, M. Yekeler, T. F. Dumm, and R.
13. L. G. Austin, C. A. Barahona, and J. M. Menacho, Hogg, "Kinetics and Shape Factors of Ultrafine
"Investigations of Autogenous and Semi-Autoge- Grinding in a Laboratory Tumbling Ball Mill,"
nous Grinding in Tumbling Mills," preprinted for Particle Particle Syst. Character. 7:242-247 (1990).
World Congress Particle Technology, Nuremburg, 28. A. M. Gaudin, Principles of Mineral Dressing,
Federal Republic of Germany, April 1986; Powder McGraw-Hill, New York, p 41-43 (1939).
Technol 57:283-294 (1987). 29. L. G. Austin, D. R. Van Orden, and J. W. Perez,
14. L. G. Austin and S. Tangsriponkul, "A More Gen- "A Preliminary Analysis of Smooth Roll Crushers,"
eral Treatment of Abrasion-Chipping Processes Int. J. Miner. Proc. 6:321-336 (1980).
Applicable to FAG/SAG Milling," Particle Particle 30. L. G. Austin and J. D. McClung, "Size Reduction
Syst. Character. 77:345-350 (1994). of Coal," in AIME Handbook, Coal Preparation,
15. A. A. Bradley, P. S. Lloyd, D. A. White, and P. W. Harvey Mudd Series, edited by J. Leonard, p
Willows, "High-Speed Centrifugal Milling and Its 189-219 (1991).
Potential in the Milling Industry," S. Afr. Mechan. 31. L. G. Austin, K. Shoji, D. R. Van Orden, B.
Eng. 22:129-134 (1972). McWilliams, and J. W. Perez, "Breakage Parame-
16. A. L. Hinde and F. B. Verardi, Studies on Design ters of Some Materials in Smooth Roll Crushers,"
of Centrifugal Mill Grinding Circuits." Proc. 3rd Powder Technol. 25:245-251 (1981).
IF AC Symposium, Automation in Mining, Mineral 32. F. Kick, Dinger Polytech. J. 247:1 (1883); 250:141
and Metal Processing, Montreal, Canada, p 283-294 (1883).
(Aug., 1980). See also: L. P. Kitschen and P. J. 33. W. J. Whiten, "Simulation of Crushing Plants with
Lloyd, "The Centrifugal Mill: Experience with a Models Developed Using Multiple Spline Regres-
 SIZE REDUCTION OF SOLIDS CRUSHING AND GRINDING EQUIPMENT 633

sion," /. S. Afr. Inst. Mining Metal 72:257-264 nia State University, University Park, PA 16802,
(1972). submitted for publication.
34. W. J. Whiten, "Application of Computer Methods 48. L. G. Austin, J. Shah, J. Wang, E. Gallagher, and
in Mineral Industries," Proc. 10th Intl. Mining Pro- P. T. Luckie, "An Analysis of Ball-and-Race
cessing Congress; ibid. 73:317-323 (1973). Milling: Part I, The Hardgrove Mill," Powder
35. A. Kumar, "An Investigation of a General Mathe- Technol. 29:263-275 (1981).
matical Model for Predicting the Product Distribu- 49. L. G. Austin, P. T. Luckie, and K. Shoji, "An
tion from a Roll Crusher and a Cone Crusher." M. Analysis of Ball-and-Race Milling: Part II, The
S. Thesis in Mineral Processing. The Pennsylvania Babcock E-17 Mill," Powder Technol. 33:113-125
State University, University Park, PA 16802 (1986). (1982).
36. V. Singhal, "An Investigation of the Applicability 50. L. G. Austin, P. T. Luckie, and K. Shoji, "An
of a Crusher Model to Jaw Crushing" M.S. Thesis Analysis of Ball-and-Race Milling: Part III, Scale-
in Mineral Processing, The Pennsylvania State Uni- up to Industrial Mills," Powder Technol. 33:127-134
versity, University Park, PA 16802 (1985). (1982).
37. F. Concha, R. Santelices, and L. G. Austin, "Opti- 51. K. Schonert, "Energetische Aspekte des Zerklein-
mization of the Ball Charge in a Tumbling Mill," erns sproder Stoffe," Zement-Kalk-Gips, 32(0:1-9
XVI International Mining Processing Congress, (1979).
Stockholm (June 1988). 52. F. Fischer-Helwig, "Current State of Roller Press
38. C. Tangsathitkulchai and L. G. Austin, "The Effect Design," KHD Symposium '92 "Modern Roller
of Slurry Density on Breakage Parameters of Press Technology," KHD Humboldt-Wedag AG,
Quartz, Coal and Copper Ore in a Laboratory Ball Cologne, p 73-79 (1992).
Mill," Powder Technol 42:281-296 (1985). 53. H. Kellerwessel, "High-Pressure Particle-Bed
39. C. Tangsathitkulchai and L. G. Austin, "Slurry Comminution: Principles, Application, Testing and
Scale-up, Details of Equipment Design," KHD
Density Effects on Ball Milling in a Laboratory
Humboldt-Wedag AG Paper, Cologne, 51 p (1993).
Ball Mill," Powder Technol. 59(4):285-293 (1989).
54. S. Strasser, "Current State of Roller Press Tech-
40. R. C. Klimpel, L. G. Austin, and R. Hogg, "The
nology," KHD Symposium '92 "Modern Roller
Mass Transport of Slurry and Solid in a Laboratory
Press Technology," KHD Humboldt-Wedag AG,
Overflow Ball Mill," Miner. Metal. Proc. (5:73-78
Cologne, p 11-21 (1992).
(1989).
55. The Horomill, Objectif 93/9 A2B2, FCB, Division
41. R. C. Klimpel and L. G. Austin, "An Investigation
Cimenterie, Groupe Fives Lille, Lille, France.
of Wet Grinding in a Laboratory Overflow Ball
56. E. A. J. Gandolfi, G. Papachristodoulou, and O.
Mill," Miner. Metal. Proc. 6(0:7-14 (1988).
Trass, "Preparation of Coal-Slurry Fuels with the
42. L. G. Austin, W. Hilton, and B. Hall, "Mill Power Szego Mill," Powder Technol. 40:269-282 (1984).
for Conical (Hardinge) Type Ball Mills," Miner.
57. E. A. J. Gandolfi, V. R. Koka, and O. Trass, "Fine
Eng. 5(2):183-192 (1992).
Grinding Applications with the Szego Mill," in Proc.
43. J. J. Cilliers, L. G. Austin, P. Leger, and A. Deneys, 12th Powder & Bulk Solids Conference / Exhibition,
"A Method of Investigating Rod Motion in a Labo- Rosemount, IL, p 448-457 (1987).
ratory Rod Mill," Miner. Eng. 7:533-549 (1994).
58. O. Trass and E. A. J. Gandolfi, "Fine Grinding of
44. L. G. Austin, J. M. Menacho, and F. Pearcy, "A Mica in the Szego Mill," Powder Technol.
General Model for Semi-Autogenous and Autoge- <50(3):273-279 (1990).
nous Milling," Proc. 20th Int. Symp. on the Applica- 59. O. Trass and R. Gravelsins, "Fine Grinding of
tion of Mathematics and Computers in the Mineral Wood Chips and Wood Wastes with the Szego
Industries, edited by R. P. King and I. J. Barker, Mill," in Proc. 6th Bioenergy Seminar, Vancouver,
Mintek, Johannesburg, South Africa, 2:107-126 B.C., February 1987, p 198-204 (1988).
(October 1987). 60. R. Gravelsins and O. Trass, "Wet Grinding of
45. L. G. Austin, "State of the Art in Modeling and Wood with the Szego Mill," in Proc. 7th Cdn.
Design of Autogenous and SAG Mills," in Chal- Bioenergy R & D Seminar, edited by E. N. Hogan,
lenges in Mineral Processing, edited by K. V. S. Ottawa, Ontario, p 281-286 (April 1989).
Sastry and M. C. Fuerstenau, Society of Mining 61. T. Molder and O. Trass, "Grinding of Waste Paper
Engineering, Inc., Littleton, CO, p 173-193 (1989). and Rice Hulls with the Szego Mill for Use as
46. L. G. Austin, "A Mill Power Equation for SAG Plastics Fillers," Int. J. Miner. Proc. (in press).
Mills," Miner. Metal. Proc. 7(0:57-62 (1990). 62. O. Trass, E. A. J. Gandolfi, and E. Daugulis,
47. L. G. Austin, "The Theory of Roller-Race Mills," "Development of an Integrated Fine-Grinding,
available from the Mineral Processing Section, De- Hydrolysis, Ethanol Fermentation Process," in Pro-
partment of Mineral Engineering, The Pennsylva- ceedings, "Energy from Biomass and Wastes XIV"
634 HANDBOOK OF POWDER SCIENCE

Conference, Lake Buena Vista, Florida, 16 p Parameters in Grinding Operations using a Direct
(Jan./Feb. 1990). Search Method," Int. J. Miner. Proc. 23:137-150
63. O. Trass, E. Edusei, and E. A. J. Gandolfi, "Wet (1988).
Grinding of Coal and Limestone with the Szego 73. O. Trass and G. L. Papachristodoulou, "Dynamic
Mill at High Solids Concentrations," in 14th Intl. Modelling of Wet Grinding in the Szego Mill," i n
Conf. on Coal Slurry Technology, Clearwater, FL, Proceedings, 2nd World Congress Particle Technol-
April 24-27, 1989; also Proc. 15th Conf., p ogy, Kyoto, Japan, Vol. II, p 471-179 (1990). See
A115-128 (1990). also: G. L. Papachristodoulou, "The Dynamic
64. V. R. Koka, G. Papachristodoulou, and O. Trass, Modelling of the Szego Mill in Wet Grinding Oper-
"Particle Shapes Produced by Comminution in the ations," Ph.D. Thesis, University of Toronto (1982).
Szego Mill," Particle Particle Syst. Character. 74. J. Hint, "Fundamentals of the Manufacture of Sili-
22:158-165 (1995). calcite Products," Gosstroiizdat, Leningrad, 601 p
65. O. Trass and O. Bajor, "Modified Oil Agglomera- (in Russian) (1962).
tion Process for Coal Beneficiation. II. Simultane- 75. P. M. M. Vervoorn and L. G. Austin, "The Analy-
ous Grinding and Oil Agglomeration," Can. J. sis of Repeated Breakage Events as an Equivalent
Chem. Eng. 66:286-290 (1988). Rate Process," Powder Technol. 63:141-147 (1990).
66. O. Trass, P. D. Campbell, V. R. Koka, and E. R. 76. A. Tymanok, "Grinding by Collision. Disintegrator
Vasquez, "Modified Oil Agglomeration Process for and its Use in Technology: Review of Principles
Coal Beneficiation. IV. Pilot Plant Demonstration and Recent Results," Internal Report, Tallinn
of the Simultaneous Grinding-Agglomeration Pro- Technical University, Estonia, 8 p (1993).
cess," Can. J. Chem. Eng. 72:113-118 (1994).
77. H. Uuemois, H. Kangur, and I. Veerus, "Wear
67. C. E. Capes and R. G. Germain, "Selective Oil
in the High-Speed Impact Mills," in Proc. 8th
Agglomeration in Fine Coal Beneficiation," in
European Symposium on Comminution, Stockholm,
"Physical Cleaning of Coal, Present and Developing
Sweden, p 513-524 (May 1994).
Methods," edited by Y. A Lin, Marcell-Dekker,
New York, p 293-359 (1982). 78. J. Hint, "Uber der Wirkungsgrad der Mechanis-
chen Aktivierung. Eininge Ergebnisse der Ak-
68. L. L. Diosady, L. J. Rubin, and O. Trass, "Solvent
tivierung von Feststoffen mittels grosser Mechanis-
Grinding and Extraction of Rapeseed," Proc. 6th
cher Energien," Aufbereitungstechnik (1971).
World Rapeseed Congress, Paris, France, p
1460-1465 (May 1983). 79. J. Hint, "About the Fourth Component of Technol-
69. O. Trass and E. R. Vasquez, "Liquifaction of Coal ogy," Valgus, Tallinn, Estonia, p 66-72 (in
with Simultaneous Grinding," in Proc. 15th Intl. Russian) (1979).
Conf. on Coal Slurry Technology, Clearwater, FL, p 80. B. Kipnis and L. Vanaselja, "Uber die Anvendung
337-349 (1990). von Desintegratoren in Technologie der Mechano-
70. O. Trass and T. Lustvee, "Preparation of Alu- aktivierung und Mechanochemie," Intl. Fachtagung
minum Pastes with the Szego Mill," Pacific Region "Forstchritte in Theorie und Praxis der Aufbereitung-
Meeting, Fine Particle Society, Honolulu, Hawaii stechnik," Freiberg, Germany, p 155-160 (1989).
(August 1983). 81. J. M. Boyes, "High-Intensity Centrifugal
71. V. R. Koka and O. Trass, "Determination of Milling—A Practical Solution," Int. I. Miner. Proc.
Breakage Parameters and Modelling of Coal 22:413-430 (1988).
Breakage in the Szego Mill," Powder Technol. 82. D. I. Hoyer and J. M. Boyes, "The High-Intensity
57(2):201-214 (1987). Nutating Mill—A Batch Ball Milling Simulator,"
72. V. R. Koka and O. Trass, "Estimation of Breakage Miner. Eng. 3:35-51.
 13
Sedimentation
Wu Chen and Keith J. Scott1

CONTENTS

13.1 INTRODUCTION 635

13.2 THEORY OF SEDIMENTATION 639
13.3 THICKENING 657
13.4 CLARIFICATION 666
13.5 NONCONVENTIONAL SEDIMENTATION PROCESSES AND EQUIPMENT 672
LIST OF SYMBOLS 676
REFERENCES 678

13.1 INTRODUCTION this chapter concentrates on sedimentation

in liquids.
Gravity sedimentation is a widely used method Suspensions of solids normally settle natu-
of separating solids/liquid mixtures and in- rally, as long as there is a difference in density
cludes diverse applications such as clarifica- between solid and liquid. Given time a suspen-
tion of waste water, thickening of milled gold sion separates into a clear liquid layer above, a
ore pulps, flotation of suspended sewage solids, supernatant, and a sediment below, which re-
and countercurrent washing of soluble metal mains "saturated" with liquid. Such batch sed-
from acid-leached suspensions. These opera- imentation can be carried out on a large scale
tions typically are performed in relatively large in tanks, ponds, or lagoons.
single-compartment tanks such as shown in To achieve continuous operation, it is nec-
Figures 13.1, 13.2, and 13.3.The discussion in essary merely to supply a steady stream of
fresh suspension, the feed, to the center (or
end) of the sedimentation vessel and to re-
Deceased. move continuously the lighter liquid phase,
635
636 HANDBOOK OF POWDER SCIENCE

SUPERSTRUCTURE
y

PLAN
Figure 13.1. Cross-sectional view and plan of a thickener. The tank may be constructed of steel or concrete. The
rake lifts vertically if it encounters an unusual resistance.

termed the overflow. Solids removal is nor- while usually achieving a high degree of clarity
mally achieved by continuous raking of the in the overflow.
thickened sediment toward the center (or op- Other solids/liquid separation techniques,
posite end) of the tank, from where it is however, need to be considered as an alterna-
pumped out as the underflow stream. tive, or addition to, gravity sedimentation if:
The relative simplicity of both the process
and the mechanical equipment involved makes 1. The solids stream must have a low moisture
gravity sedimentation the least costly of the content.
available solids/liquid separation tech- 2. The loss of 10% to 15% of the liquid in the
niques.2'3 feed to the underflow is not acceptable.
The process has the capability of treating high 3. The cost of the required floor space is
water flow rates with relatively little hardware1 excessive or space is not available.
 SEDIMENTATION 637

operating clarifler. The larger tank, the secondary clarifier, represents one of the final
drinking water from purified sewage in the Standard Water Reclamation Plant, a pilot
. (Courtesy of National Institute for Water Research, SCIR.)

an empty thickener at a Transvaal gold mine. Such tanks handle up to 15,000 tons/day of
e times as much water.
638 HANDBOOK OF POWDER SCIENCE

4. The process must be carried out under high proportion of liquid in the underflow can
pressure. be tolerated. On the other hand, in thickening,
the minimum quantity of liquid in the under-
Table 13.1 shows the advantages of various flow is the main objective and the presence of
solids/liquid separation methods. up to a few percent of suspended solids in the
Combinations of techniques may be used to overflow (often harmlessly recirculated) is of
improve the effectiveness of separation, such secondary concern. The distinction is there-
as a vacuum filter immediately following a fore in the end result rather than the process;
thickener to dewater or wash the thickener in thickening, the solids concentration in the
underflow. In selecting a separation process it feed stream is increased by sedimentation
is essential, therefore, to consider wider as- while in clarification the solids are removed by
pects than just the pros and cons of individual this process.
techniques. Some guidelines are available in Each of the two functions can be optimized
the literature for selecting equipment2'5"7 but and controlled separately. The turbidity of a
these should be supplemented by sufficient clarification tank overflow is related to slowly
knowledge in this field. Discussions with settling fine solids which may be flocculated to
specialists or equipment suppliers can help form larger faster settling units. The control of
in formulating likely solutions for a given overflow clarity is therefore affected by the
problem. selection of flocculant, its dosage, and by
control of the volumetric feed rate.
Underflow density of a thickener depends
13.1.1 Objectives in Gravity
on the height of sediment in the tank, the
Sedimentation
degree of flocculation in the suspension (floc-
Sedimentation is distinguished into two pri- culated material tends to incorporate more
mary functions. The first is clarification, in liquid than dispersed particles), and on the
which absence of solids in the liquid overflow underflow pumping rate. Because flocculation
is the essential requirement and a relatively and feed rate affect both the overflow clarity

Table 13.1. A Qualitative Comparative Guide to the Particular Advantages of Various Solids/Liquid
Separation Techniques.

RATIO OF
SOLIDS / LIQUID THROUGHPUT CLARITY MOISTURE EASE OF OVERALL COST
SEPARATION TO FLOOR OF CONTENT OF WASHING CAPITAL PLUS
TECHNIQUE AREA LIQUID SOLID STREAM SOLIDS OPERATING
Sedimentation
Gravity Low Good High Require repeat Low"
operations
Centrifugal High Good to Medium Possible High
excellent
Cyclone Very high Very poor High Require repeat Low
operations
Filtration High Good Low Easy High*
Screening High Very Poor Medium Easy Medium
Drying Medium — Extremely low — High
a
The operation cost of vacuum filters in the S.A. gold mining industry ( ~ 75 million tons/yr) is six to eight times
the cost of gravity thickening.4
 SEDIMENTATION 639

and the underflow density it is seldom possible ever. The complicating factors that arise in
to optimize both clarification and thickening real situations, dealt with more fully in subse-
simultaneously.8 quent sections, are:

13.1.2 Applications of Gravity 1. Nonspherical and irregularly shaped

Sedimentation particles
2. The simultaneous presence of large number
Sedimentation processes are used extensively
particles
throughout the world in many industries, wa-
3. The presence of mutual particle attraction
ter purification, and waste water treatment.
in which individual particles lose their iden-
These large-tonnage operations are often car-
tity and are grouped into agglomerates
ried out in remote locations or nonurban
(floes) by chemical-physical forces
areas where land is available and relatively
4. The method of measuring of settling rate
inexpensive, and hence the use of large tanks
for gravity settler design
is not a serious disadvantage.
5. Wall effects.
Sedimentation is also practiced on a smaller
scale in a variety of processes. Increasing at-
tention is being given to development of higher 13.2.1 Sedimentation of a Sphere in an
capacity thickeners, that is, those of reduced Infinite Fluid
area per unit throughput or thickeners that When a single spherical particles is suspended
can produce thick underflows equivalent to at rest in a liquid it experiences two opposing
filter cakes. This interest is minimizing the forces, BF and G F , as shown in Figure 13.4.
disadvantages of sedimentation, rather than Provided the densities of the solid and the
selecting alternative types of equipment, is an liquid are not equal, there will be an unbal-
indication of the desirability of the positive anced force, the difference between G F , the
features of gravity sedimentation as a means downward gravitational attractive force, and
of separating solids/liquid mixtures. BF, the Archimedean upward thrust or buoy-
ancy force. This unbalanced force, (G F - BF\
equal to V( ps - p L )g, causes the particle to
13.2 THEORY OF SEDIMENTATION accelerate (downward if positive) and attain a
velocity relative to the liquid. Skin friction,
This section covers the fundamental aspects of that is, the resistance offered by a fluid to the
the sedimentation of particles, whether as sin- motion of a solid, then results in the develop-
gle spheres in an unbounded liquid or in a ment of a drag force, FD, which opposes the
mixture of many other particles in a finite motion and increases with increasing particle
suspension. As much of the published verifi-
cation of the theory was carried out in labora-
tory measuring cylinders, this section also dis-
BUOYANCY FORCE » V./D Q
cusses the theory of batch settling tests. How
the settling behavior of suspensions observed
in batch tests is used in the design of sedimen-
tation equipment is covered separately (p.
1002), while the nomenclature used is given at F o » DRAG FORCE GRAVITATIONAL ATTRACTION » V./J .Q

the end of the chapter. V = VOLUME OF PARTICLE

The settling of a single sphere in an un-

bounded fluid represents the simplest case of
solids/liquid sedimentation. This ideal condi- Figure 13.4. Forces acting on a spherical particle in a
tion is seldom encountered in practice,1 how- liquid.
640 HANDBOOK OF POWDER SCIENCE

velocity. The drag reduces the acceleration, significant, the flow is called transitional. These
and finally the value of the drag force becomes limiting values for the particle Re are orders
equal to the original driving force (G F — BF) of magnitude lower than for flow in pipes in
and there are no further unopposed forces which the fluid streamlines are constrained by
acting on the particle, it continues to travel at the boundary walls.
a constant rate called its terminal settling ve-
locity, u^. We may then write: 13.2.1.2 Laminar Flow

^ D = FXPs - PO§ 13
( -D The analytical solution for the magnitude of
the drag on a single sphere, settling under
This equation evaluates the magnitude of the
streamlined flow conditions in an unbounded
drag force for any size particle but does not
liquid, is given by Stokes9 as:
relate it to its unknown settling velocity.
This relationship has been formulated for a FD = (13.3)
sphere in an infinite fluid9 for slow flows but
its general solution depends on the type and where u^ = terminal velocity of the sphere in
magnitude of flow around the particle as char- an infinite fluid in streamlined flow.
acterized by the dimensionless entity known as Even for this simplified condition, however,
the Reynolds number. Eq. (13.3) is only a close approximation and,
for greater accuracy, additional terms have
13.2.1.1 Fluid Flow Around a Particle and the been found to be necessary. Proudman and
Reynolds Number Pearson10 for example, advocate the equation:

When the particle velocity is low, drag is due

largely to the viscosity of the liquid and this
flow is called viscous, laminar, or streamlined.
At high velocities, the fluid streamlines do not (13.4)
continue completely around the particle but
break up into vortices with the result that
It has become common practice11 to express
turbulent eddies and inertial forces also con-
the forces exerted on moving bodies by the
tribute to the drag. This finally develops into
fluid in terms of a dimensionless drag coeffi-
fully turbulent flow.
cient C D , obtained by dividing the drag force
The criterion for distinguishing between flow FD by pLMsr/2 and by the area of the body
conditions is the dimensionless particle projected onto the plane normal to wsr. For a
Reynolds number: sphere, this area is ird^/A\ hence the coeffi-
d
p'Us * PL cient is:
Re (13.2)
(13.5)
where
dp = particle diameter
wsr = relative velocity between particle and which, together with Eqs. (13.2) and (13.4) and
liquid setting uST = u^, becomes
p L = liquid density
IJL = liquid viscosity. 24 27 Re

For particles,12b streamlined flow occurs be- (13.6)

low Re ~ 0.3, turbulent flow above Re ~
2 X 105, whereas in the intermediate region, The practical significance of terms following
in which inertial forces become increasingly 24/Re can be tested by considering the largest
 SEDIMENTATION 641

Reynolds number likely to be encountered in a spherical quartz particle in water at 20°C,

real situations. The maximum overflow rate of the maximum permissible diameter is dp ~ 85
operating sedimentation equipment quoted by jLtm if the error in u^ using Eq. (13.8) is not to
Perry and Green12a is usr ~ 0.8 mm/s; the exceed 5%.
95% upper limiting diameter in their particu-
late slurries is estimated as dp ~ 0.2 mm, and, 13.2.1.3 Transitional and Turbulent Flow
if we accept the most common sedimentation
Grit chambers, a special type of sedimentation
liquid as being water at 20°C, p L = 1 X 103
basin used in sewage treatment, are designed
kg/m 3 , jit = 1 X 10~3 kg/ms, and therefore
for removal of coarse sands larger than 200
Re should lie largely in the range 0 to 0.16.
jLtm. Such large particles settle in transitional
The accurate value of C D for Re = 0.16
or turbulent flow but for the determination of
according to Eq. (13.6) is 153.95, while using
their terminal velocities, no alternative simple
the first term only, C D = 150.0, representing a
expression similar to Eq. (13.8) exists. To cal-
difference of 2.6%. This maximum error is
culate their terminal velocity, u^v the particle
quite acceptable in settler design as variations
Reynolds number must first be known, but this
in ambient temperature of only 1 to 2°C can,
cannot be known until the value of u^ is
by changing the liquid viscosity, result in
determined. A trial-and-error solution is one
greater variations in sedimentation rate. The
means of arriving at its value.13
second and third terms in Eq. (13.6) may
The following more direct and accurate al-
therefore be safely neglected and the equation
gorithm has, however, been found useful in
simplified to give:
avoiding both this repetitive procedure and
24 the inaccurate graphical or cumbersome inter-
CD = — (13.7)
polation of the Re vs. C D Re 2 values given in
which is an alternative expression of Eq. (13.3). Perry and Green.12b The algorithm includes
As the particle is a sphere, V in Eq. (13.1) is also the laminar flow region discussed previ-
ITdl/6 and from Eqs. (13.1) and (13.3) we can ously.
now write
Algorithm for Calculating u^ from d
Ps ~ p.
(13.8) 2
1. Calculate the entity / = C D Re which does
This equation provides a means of calculat- not contain w^t:
ing the terminal settling velocity of a single / = C n Re 2
D
sphere of diameter dp in an unbounded fluid,
in streamlined flow, as determined by the 4- (p s ~
•g (13.9)
physical properties of the sphere and fluid, 3"
that is, their densities and the fluid viscosity. 2. From / calculate the required values of a
Alternatively, it permits the estimation of the and b according to the data given in Table
diameter of a particle by observing its settling 13.2 using the appropriate range of /
velocity under these prescribed conditions. For values.

Table 13.2. Values of a and b for Calculating Re from J = C D Re 2 in Any Flow Regime.

FLOW REGIME ~dp T = CDRe2 a b -Re

Laminar (Stokes' law) 1-75 fim 0-10 24 1 0-0.4

Transition region I 75-350 fim 10-10 3 33.7/- °-19 1.05/-°-05 0.4-20
Transition region II 0.35-2 mm 10 3 -2 X 10 5 63.6/" 0 2 9 L 0 / - 0.046 20-600
Transition region III 2-15 mm 2 X 10 5 -l X 10 8 91.7/" 0 - 30 0.96/-° 0 3 9 600-1.5 X 104
Newton's Law 1.5 X 10 cm 1 X 10 8 -1.7 x 1010 0.67 0.5 1.5 X 10 4 -2 X 105
642 HANDBOOK OF POWDER SCIENCE

3. Calculate initial spin, it will not rotate to a different

b position but fall vertically in its original orien-
J
Re = — (13.10) tation. Most real particles are not symmetrical,
a however, and they experience not only drag
4. Finally, determine forces parallel to the stream velocity, but also
lateral (lift) forces at right angles to the stream.
Re This may cause drift to one side during
(13.11)
settling, rotation to a position of maximum re-
sistance, steady rotation, or even a wobbling
The values calculated in this way are accu- motion.
rate to within 5% to 6%. A subroutine for the Even when such nontranslational motions
trial-and-error solution mentioned earlier13 is are neglected, the calculation of the terminal
less accurate; for example, at Re ~ 70 the settling velocity of nonspherical particles using
error is 18%. The algorithm covers a wide an equation similar to Eq. (13.8), that is wns =
range of sphere diameters from 1 fim up to 10 / ( p s , p L , /jL, dp) requires first that the shape is
cm and embraces the laminar, transitional, known or can be determined; second, that a
and Newton's flow regimes. When, however, representative "diameter" dp can be assigned
the flow is known beforehand to be laminar to this shape; and third, that a drag equation
(Re < 0.3), Eq. (13.8) provides exactly the be available similar to Eq. (13.3) for a sphere.15
same answer in fewer steps. The problem is thus complicated, and cer-
A theoretically derived120 equation: tainly no analytical solution exists for the
irregular shaped particles encountered in
Re = 20.52[(l + 0.0921/ 05 ) 0 " 5 - i f (13.12) practice.
However, empirical methods for dealing with
gives results that are within 7% of the experi-
shape are available. These are presented
mental values for Re up to 7000.
briefly with reference to some of the effects of
A free-settling equation, valid not only for
increasing departure from spherical form
all particle sizes but also covering a wide range
shown in Table 13.3. This table compares par-
of naturally occurring shapes (see following
ticles of various shapes on the common basis
section), is presented by Swanson;14 his calcu-
of having the same volume (arbitrarily 1 mm3),
lated values are within 20% of the measured
that is, of each possessing an equal gravita-
values in most cases.
tional attractive force. Column 3 shows that as
the shape departs increasingly from spherical,
13.2.2 Nonspherical Particles in an
the surface area increases from 4.84 for a
Infinite Fluid
sphere to 8.35 for a cylindrical needle of the
Spherical particles rarely occur in solids-liquid same volume, that is, an increase of 73%. The
separation practice, some more common skin friction and hence drag for nonspherical
shapes being angular and flaky particles de- particles will thus be greater than their termi-
rived from crushing or weathering, amorphous nal settling velocities correspondingly lower
fluffy precipitates, randomly formed agglomer- than for a sphere of the same volume.
ates, and regular polyhedral or needle-shaped Skin friction is related to surface area. If, in
crystals. the drag Eq. (13.3), dp is taken as the diame-
The drag on a nonspherical particle de- ter of a sphere having the same surface area
pends on its shape and its orientation with as a given shaped particle,15 from Eqs. (13.1)
respect to the direction of motion.12b If a body and (13.3) it can be shown that:
possesses spherical isotropy (such as a cube or
octahedron) and is placed initially in any ori- ^ = fp=K Q3.13)
entation in a liquid and allowed to fall without
Table 13.3. Various "Diameters" of Particles of Equal Volume (1 mm3) but Differing in Their Shape.

DIAMETER OF A SPHERE (mm)

SHOWING THE SAME PASSING THE SAME
HAVING THE PROJECTED AREA MINIMUM SQUARE
SAME SURFACE AS PARTICLE APERTURE
CALCULATED AREA AS (SIZING BY (SCREEN
SURFACE AREA PARTICLE MICROSCOPE) ANALYSIS) RATIO
SHAPE DESCRIPTION (mm2) dA dM ^scr ^M/^scr

Sphere — 4.84 1.24 = dp 1.24 1.24 1.00

Icosahedron 20 Equilateral 5.15 1.28 1.32 1.23 1.07
triangles
Dodecahedron 12 Pentagons 5.31 1.30 1.35 1.27 1.06
Cube-octahedron 6 Octagons and 8 5.69 1.35 1.14 — —
equilateral triangles
Octahedron 8 Equilateral 5.72 1.35 1.36 1.24 1.10
triangles
Hexahedron 6 Squares 6.00 1.38 1.13 1.00 1.13
(cube)
Tetrahedron 4 Equilateral 7.20 1.51 1.51 1.44 1.05
triangles
Plate (5:5:1) — 8.19 1.61 1.93 1.45 1.33
Needle (10:1) Right circular 8.35 1.63 1.80 0.50 3.60
cylinder
644 HANDBOOK OF POWDER SCIENCE

where apertures in the most favorable position for

passing, dSCY. Thus, for an elongated or platy
u^t = settling rate of sphere of diameter dp particle, the two larger dimensions are
having the same volume as the non- recorded by microscopy while for sieving the
spherical particle two smaller dimensions are recorded.
dA = diameter of sphere having same surface Columns 5 to 7 in Table 13.3 compare the
area as the nonspherical particle mean diameters of various shaped particles as
K = shape factor. measured by these two methods. It can be
seen that dM is, as expected, consistently larger
Kp is normally smaller than 1, indicating lower than dSCT, except for a sphere where they are
settling rate owing to the larger surface area equal. The ratio dM/dSCT, therefore, also pro-
in nonspherical particles. This treatment is not vides a basis for formulating shape correction
rigorous because Eq. (13.3) applies strictly to factors but it is less consistent than the ratio
spheres. However, values of Kp calculated as dp/dA. The latter can, however, be used only
above using the data in Table 13.3 compare if dp and the shape are known.
reasonably well with the experimentally deter- Much work has been done on characterizing
mined values16 as shown in Table 13.4. The the shapes of microscopic images alone,16"18
agreement, although not exact, serves to illus- but the weakness in applying such results to
trate the principle of using shape correction sedimentation19 is that only two dimensions
factors, based on more than one characteristic are considered. Thus, a cube and a square
diameter of a particle, to estimate its settling plate both of side d settle at quite different
rate. The particle must, however, be of known rates but are indistinguishable under the light
shape. microscope while a right circular cylinder of
The more usual situation is that neither the equal height and diameter may appear spheri-
shape nor the size is known, that is, dp and dA cal to one observer and cubic to another de-
both need to be determined experimentally. pending on its position on the microscope
Two commonly used methods of measuring slide. Different shape factors would therefore
particle size are microscopy and sieve analysis. by applied to the same particle.
For spheres both techniques give identical re- Both the experimental procedures for mea-
sults but for nonspherical particles, the two suring shape factors as well as their interpreta-
methods, because they measure different tion are therefore involved but the main pur-
properties, give different characteristic diame- pose is merely to apply a correction factor to
ters for the same particle. Stokes' law so as to be able to calculate the
Thus, in microscopy the diameter of a parti- settling velocity of any nonspherical particle.
cle, dM, is the diameter of a circle with the The necessity for this, however, is completely
same projected area as the particle which nor- eliminated if for the measurement of dp, we
mally lies "flat," while in sieving the particle use an alternative, very commonly used method
has a probability of being presented to the of particle sizing—the sedimentation method.

Table 13.4. Comparison of Calculated and Observed Shape Factors Kp.

SHAPE FACTOR Kp
SHAPE CALCULATED, dp/dA EXPERIMENTAL RESULT16
Sphere 1.0 1.0
Cube-octahedron 0.92 0.96-0.98
Octahedron 0.92 0.93-0.95
Cube 0.90 0.92-0.94
Tetrahedron 0.82 0.82-0.86
 SEDIMENTATION 645

This technique supplies information on the the data are scattered but average out at about
settling velocity, uns, of any particle of un- 0.8; for Re > 2000 the shape factor is steady
known shape that when inserted into Eq. (13.8) at 0.47, while in the transition region the fac-
gives the exact equivalent Stokes' diameter, tor shows a steady decrease with increasing
dns. This is the diameter of a sphere of settling Re.
velocity identical to that of the nonspherical The shape factor for a given nonspherical
particle. It combines both the "true diameter" particle is therefore not even a constant for
of the particle and its shape correction factor that particle but dependent also on the pre-
into a single term. For the transitional and vailing conditions.
turbulent flow regimes such direct observation Measurements of settling velocity of nonspher-
is the only means of determining the settling ical particles are therefore simpler and of more
velocities of nonspherical particles.15 Investi- use than prediction of these velocities from inde-
gations of settling of such particles have there- pendent size and shape determinations.
fore been mainly experimental.16'20'21 Most
particles encountered in industrial practice are
far from spherical to such an extent that not 13.2.3 Settling in the Presence
only the overall shape plays a role but also the of Other Particles
microsurface topography. Many correlations have been presented to
The experimental observations of Richards describe the effect of higher solids concentra-
and Locke21 on the terminal settling velocities tions on the settling rate of uniformly dis-
of various sized quartz particles wns, obtained persed particles.23 Two effects have been ob-
by screening, may be used to determine the served. One is that some particles may loosely
shape factor for this irregular material. The associate into a group, separated from each
results are plotted in Figure 13.5. other by several diameters, and act as an en-
It can be seen that the shape factor depends tity descending at a higher rate than that
on the size dp. Relating the shape factor to corresponding to the expected terminal veloc-
the corresponding Reynolds number Re < 1, ity of the individual particles. Such "clusters"
are often transient and their occurrence has
been observed24"26 predominantly at low par-
ticle concentrations. Other particles in the
PARTICLE REYNOLDS NUMBER
0,01 0,1 I 10 tOO 1000 10 000 same suspension remain single and may even
show negative settling rates when being car-
ried upward by the return flow from the rapidly
descending clusters. Tory and Pickard27 pre-
sent a stochastic model that accounts for these
wide variations in settling rate. They noted
that in spite of variations of settling velocity
between particles, the mean settling velocity as
shown by their overall rate of descent re-
mained remarkably constant.
The second effect is that as concentration
increases each particle is subjected to in-
creased drag owing to the higher volume of
0,01 OpZ 0,05 0,1 0,2 0,5 I 2 5 10 20 return flow fluid displaced by the sedimenting
d - PARTICLE SIZE, mm particles. Alternatively, the ideal fluid flow
Figure 13.5. Variation of shape factor of quartz with around each particle is disturbed by the pres-
particle size and Re. ence of its neighbors.
646 HANDBOOK OF POWDER SCIENCE

13.2.3.1 Suspension of Uniform Particles For streamlined flow, which has been shown
to be the most common in thickening practice
With closely sized particles uniformly dis-
and hindering settling, n becomes indepen-
tributed in a settling cylinder, a visible inter-
dent of Re and was determined experimen-
face between suspension and liquid forms at
tally28 to be
commencement of settling even at low concen-
trations because of the constant means de-
n = 4.65 + 19.5dp/D (13.15)
scent rate of each particle. In effect this inter-
face is the one that exists between suspension
and hence a second constant for a given sys-
and air before commencement of sedimenta-
tem. An exact value of n in the range 4.65 to 5
tion. It is usually hazy, however, because parti-
is, however, seldom critical and, as shown later,
cles are never exactly identical. At higher con-
a value of n = 4.7 is found to give a satisfac-
centrations this interface becomes increasingly
tory correlation for a large number of real
well defined and sharp, forming even for parti-
suspensions even where dp is unknown, that
cles with a considerable range of sizes (see
is, when an exact value of n cannot be calcu-
later). As such suspensions separate when di-
lated from Eq. (13.15). A similar compromise
lute, because of the wide variation of settling
value of n = 4.7 (n between 4.65 and 4.78)
velocities present, the formation of a distinct
was arrived at also by Watanabe.29
interface at higher concentrations indicates
Although Eq. (13.14) was derived empiri-
that interference between particles is such that
cally, various authors 15 ' 30 ' 31 have shown its
particles of all sizes descend jointly, that is,
general validity on theoretical grounds. The
they are in hindered settling.
hindrance factor term is a simple one15 that
permits its modification to deal with sedimen-
Richardson and Zaki Equation. The settling tation of irregularly shaped particles and parti-
velocity of the interface, us, was noted by cle aggregates which will be discussed later.
Richardson and Zaki 28 to be related to the Alternative hindering settling equations are
velocity of a single particle w^, and the con- much more complicated.23 In some the con-
centration (f)s, by: centration term appears in various forms up to
five times in one equation rather than once as
= ujl - (13.14) in Eq. (13.14).
where
Experimental Verification of the Richardson
us =mean settling rate of particles and Zaki Equation. As (1 - <£s) is a fraction
(particle-supernatant interface) in a and n a positive number, Eq. (13.14) indicates
container in the presence of many a decrease in particle settling rate with in-
others creasing volume fraction of solids.
Moo = terminal velocity of a single repre- Figure 13.6 compares the experimentally ob-
sentative particle, that is, ws when served values of us for two suspensions of
4>s = 0, under otherwise similar con- glass spheres with predicted values based on
ditions. It is a constant for a given Eq. (13.14). The value of u^ in this equation
solid-liquid system and equivalent to was calculated from Eq. (13.8) and n calcu-
uw in Eq. (13.31) lated from Eq. (13.15). The agreement can be
(f>s = volume fraction of particles (dimen- seen to be good. For an uncharacterized sus-
sionless) = C/ps pension for which dp and hence the constants
C =mass concentration of particles, for MM and n are not known, it should be possible
example, kg/m 3 to estimate them from Eq. (13.14) by means of
n = a constant = f[dp/D, Re] a plot of In ws versus ln(l - </>s) and using the
u^ =the hindrance factor = (1 - </>s)n. intercept and slope of the best fitting straight
 SEDIMENTATION 647

0,8
ent materials with various shapes and densi-
u0 (mm/t) ties. Although such particles tend to segregate
GLASS d
n
BEAOS CALCULATED BEST FIT when dilute, at normal thickener feed concen-
FROM d TO DATA
tration, mutual retardation of the particles in
X 63,6 4,70 3,53(«) 3,54
a batch test causes hindered of "zone-
0 26,2 4,66 0,S0(«) 0,49
settling"34 with a uniform particle settling
velocity regardless of size. It is therefore less
reliable to calculate u^ from Eq. (13.8) and,
for reasons that are discussed in the next
section, use is made of an alternative form of
Eq. (13.14):
X DATA OF SHANNON •« Ol (REF 3 2 )
0 DE JAGER J.PJ. (REF 4 )
A/n ul/n _ l/n .(
=
(13.16)

EQUATION ( I I ) where n may be taken as equal to 4.7, that is,

l/n = 0.213.
A plot of w°213 versus <£s should therefore
give a straight line from which either the slope
or the intercept can now be used to determine
Woo. Although, because of differing dp values,
the values of ux vary between different sus-
pensions, the hindrance factor of any suspen-
0,5 1.0 sion (1 - </>s)4J, should be constant at a given
4>, - SOLIDS VOLUME FRACTION
concentration. The relative velocity, uju^,
Figure 13.6. Settling velocity of glass bead suspension should therefore be the same for each suspen-
as a function of solids concentration.
sion at the same concentration and plots of
(ws/Moo)0'213 versus <f>s should yield a single
line. The intercept is ux and the slope is n. straight line of slope = - 1 because from Eq.
This is one basis for using hindered settling (13.16):
rates, ws, to determine the mean particle size
of a suspension of particles. 0.213
Two independent values of u^ estimated in — | = (1 - (13.17)
this way are compared to Figure 13.6 and the
excellent agreement found between them has
The data of Figure 13.6 are replotted on
been shown33 to apply also to a variety of
this basis in Figure 13.7 and they can be seen
solids and liquids (dp = 13 to 1740 fim; p L =
to fall closely on this theoretical line with
890 to 1070 kg/m 3 ; and /x = 1 to 7 cp). The
intercepts at +1 and a negative slope of 45°.
reliability of the Richardson and Zaki equa-
In addition, summarized data on spheres from
tion, together with their value of n, is there-
an extensive survey of the literature23 are in-
fore well established for relatively closely sized
cluded and the agreement with Eq. (13.17) is
spherical particles up to the maximum attain-
within 6%. Bearing in mind that no allowance
able free-settling concentrations.
was made in this work for fairly wide varia-
tions in dp/D and its effect upon n and ux,
13.2.3.2 Suspensions Consisting of a Range
this agreement is reasonable.
of Particle Sizes, Shapes, and Densities
When, however, the sedimentation rates of
Suspensions consisting of uniform particles are angular quartz particles rather than spheres
rarely met within practice. Real slurries con- are compared on the same basis, Figure 13.7
tain a range of particle sizes and often differ- shows that the settling rate now decreases
648 HANDBOOK OF POWDER SCIENCE

X
fixed water. Thus, if a quartz suspension at
AS FOR PREVIOUS FIG
SPHERES 0
MEAN VALUE OBTAINED FROM
<£s = 0.2 settles according to Eq. (13.17) as if it
379 DATA POINTS EXTRACTED
FROM 17 PUBLISHED PAPERS experienced the same drag as a suspension of
\ AND THEIR 9 5 % CONFIDENCE
LIMITS-REF.23 spheres at <£s = 0.5, its effective total solids
ANGULAR 0
QUARTZ
PARTICLES A
d :26^m
HATTON.R. REF4 volume must in fact be equal to 0.5, with the
d =2O M m
extra volume being made up of stagnant water
than moves with the particle.
Because the stagnant water behaves as if
solid, the lines for quartz in Figure 13.7 re-
main straight but because of the unknown
quantity of water, they are of unknown slope.
If we call the slope &v, Eq. (13.17) can be
rewritten
0.213
(13.18)

0,4 0,6
or
-SOLIDS VOLUME FRACTION 4.7
= w j l - ky(/)s) (13.19)
Figure 13.7. Settling velocity of spherical and angular
particles plotted according to Eq. (13.17). Comparing with Eq. (13.14) it can be seen
that the original concentration term <£s is re-
placed by kv(f)s, which now represents the
more rapidly with increasing concentration
effective solids volume fraction. For spheres,
than predicted by Eq. (13.17). The spheres
where there is no stagnant liquid, ky = 1, while
reach a hindrance factor us/ux = 0.0385 at
for the two quartz suspensions kv-~ 2.5 with
4>s = 0.5 while this same retardation is experi-
the finer sample carrying relatively more
enced by quartz particles at a concentration as
water, that is, having a slightly higher value
low as 0.2. As the retardation in settling veloc-
of Jfcv.
ity of a particle in hindered settling is due to
Correlation of the settling data for quartz
the interference offered to its ideal return
according to Eq. (13.14), that is, ky = 1 and n
fluid flow pattern by the presence of its neigh-
variable, gives values of n ranging from 11.8 to
bors, it must be concluded from hydraulic sim-
14.6 depending on $ s . The value of n is there-
ilarity considerations that quartz particles pre-
fore not only much higher than expected
sent a greater effective blockage to the return
from theory30 (n lies between 1 and 8) but
flow than can be expected from their volume.
more seriously is not a constant for a given
A unit volume of quartz must in fact have the
system. Many such correlations have been at-
retardation effect of 0.5/0.2 = 2.5 volumes of
tempted,35"40 with values of the exponent as
an equivalent sphere. The plausible inference
high as 466.7, but as shown by Capes,41'42
is that such angular particles carry with them
these are reduced to expected levels if due
attached water because of their roughness14'15
allowance is made for the fixed water associ-
and this stagnant water behaves as if the vol-
ated with particle agglomeration or irregular
ume of the particle were effectively increased.
shape.
The net solids concentration is therefore
greater than the volume of dry solids present.
By assuming the effective solids fraction for 13.2.4 Aggregated Suspensions
any degree of retardation to be similar to the Natural aggregation is frequently present in
volume fraction of spheres at the same retar- particle suspensions,1'43 especially at higher
dation, we can calculate the proportion of concentrations such as in thickener feeds,
 SEDIMENTATION 649

where the mutual proximity of the particles Dilute Suspensions. When the suspension is
causes them to adhere and settle together as dilute the aggregates (floes) are formed inde-
clumps rather than as single particles. This pendent of each other—they are widely spaced
increase in "particle" size results in faster in the intervening liquid and descend through
settling, and in thickening the effect is often it as individual entities. After agitation ceases
exploited. In clarification, the solids concen- (t = 0) the floe formation time is fast46 com-
tration in the liquid is much lower, a natural pared to the time over which sedimentation is
aggregation is largely absent. Artificial observed in a batch test. For instance, in the
flocculation is therefore always required. presence of a coagulant, the silky appearance
It can be brought about by reducing the mutu- of dispersed micaceous clays noted during stir-
ally repellent charges on the particles by means ring disappears within seconds after agitation
of electrolytes (coagulation) or by bridging is stopped. Dilute suspensions of floes there-
particles by the simultaneous adsorption of fore show a constant interface descent rate
polymers. from zero time (curve A1 in Figure 13.8).
In all cases, aggregates are produced, each
consisting of a large number of varying size
primary solid particles, associated together into Intermediate Concentrations. At higher
a single relatively large sedimentation unit or concentrations, the particles have a better
floe. Such a floe includes not only this loosely chance of forming larger floes. At the start of
held solids structure but also the interstitial a batch settling test, that is, after the cessation
stagnant water.15 Floes have a density lower of agitation, the suspension appears to have an
than the solid particles, due to this water, but "induction" period (curve Bv Figure 13.8)
have a greatly increased diameter so that their during which the relatively low initial sedimen-
settling rates are several orders of magnitude tation rate u{ increases with time either grad-
higher than those of the original individual ually,44 or in discrete steps45"51 and subse-
particles. No distinction is made here between quently reaches a higher constant rate, us.
the terms floes and aggregates or in their The maximum steady value is accepted as the
method of production, as only the sedimenta- settling rate in a static batch test.43'52"54 Two
tion behavior of the final aggregates is of phases in this acceleration process are shown
concern at this stage. in Figure 13.9A to C.
This rate is higher than would be expected
13.2.4.1 Types of Settling Behavior from extrapolation of sedimentation data in
in Aggregated Suspensions the dilute rate (Fig. 13.10), indicating that the
Previous sections dealt with discrete individual mode of sedimentation is now different.58 The
particles as the primary sedimentation units. higher settling rates are attained not only by
Resuspension of these unaggregated pulps in a the formation of larger floes but also by the
batch test to prepare a uniform suspension reduction of resistance to relative movement
usually does not alter the size, shape, or set- of floes and liquid. An anisotropic structure in
tling characteristics of these units from test to the suspension with liquid channels of low
test. In flocculated suspensions, however, the flow resistance in an upward direction is
sedimentation units (aggregates) are freshly formed during the induction period. A similar
formed only after agitation ceases. The shape argument was used to explain the accelerated
of the resultant sedimentation curve (height of settling of intermediate concentration suspen-
interface H versus time t) depends on solids sions in the presence of particles of density
concentration,44'45 that is, on the number of lower than,59 equal to,60 or greater than61 the
primary particles present and their mutual density of the fluid.
proximity when agitation is stopped and the At intermediate concentrations the floes
shearing force is removed. must be closer together than in dilute suspen-
650 HANDBOOK OF POWDER SCIENCE

WITH SLOW STIRRING

QUIESCENT CONDITIONS INTERMEDIATE-B CONCENTRATED-C

QUIESCENT

STIRRING
^STIRRING
LESLIE MINE ORE
4>s = 0 , 2 8
' LESLIE MINE ORE * ^
PYROPHYLUTE <f>s =0,015 I ^ s =0,15 QUIESCENT [STANDER,J.W.-REF^
kv4>s = O,2 ! /cv4>s-O,55
[REF 47]
I [STANDER.J.W.-REF4]

20 0
i
t,
. 100 200
t-TIME, min t - TIME, min t-TIME, min

Figure 13.8. Mode of settling in aggregated suspensions depends on solids concentration and presence of mild
agitation.

sions. Their probability of touching or bridging the primary particles.64 Some channeling may
by particle growth is therefore much higher occur and often a few very large channel open-
and this three-dimensional interaction be- ings (volcanoes) are observed at the interface.
tween solids is likely to be involved in the Settling curves such as those shown in
formation of the channel structure with flow Figure 13.8 A1 to Cx were obtained66 also for
channels being developed between the floes.44 sedimentation of the coal particles in oil, that
It is therefore not surprising that the induc- is, a nonaqueous system, indicating their
tion period increases with concentration.121 In general nature.
the intermediate concentration range the max-
imum steady settling rates us decrease with Slow Agitation in Aggregated Suspensions.
concentration, but to a lesser extent than for The effect of very mild stirring (0.1 to 2 rpm)
pulps in the dilute range (Fig. 13.10). The depends on the concentration regime present.
decrease is not unexpected being at higher floe In dilute suspensions the formation and the
concentrations less voidage between them is subsequent sedimentation of the floes is nei-
available for channel flow and fewer and ther aided nor hindered and settling rates are
somewhat narrower channels may be formed. therefore little affected (curve Ax and A2,
Fig. 13.8). In intermediate suspensions, hori-
Concentrated Suspensions. The solid parti- zontal shear hinders the formation of short-
cles are in a compression zone. The suspen- circuit flow channels and materially decreases
sion does not attain any degree of "mobility" the maximum settling rate attained47'62 (curve
but subsides at a sluggish and ever decreasing B2 rather than Bx).
rate. When agitation is stopped the particles In concentrated slurries, mild mechanical
are closer together and are able to form a disturbance promotes the shearing of the
three-dimensional structure like a packed bed. particle-particle links. Under quiescent condi-
The lower layers can be further compacted by tions the three-dimensional structure that
the weight of solids from above. The particles forms after cessation of agitation tends to
collapse inwardly toward each other and con- resist collapse because of friction at the points
sequently liquid is expressed from these layers. of particle contact and the support from the
This liquid moves upward through the bed and base and the walls. The mass of solid above
because of the tight packing, mainly between may not be sufficient to overcome the strength
 SEDIMENTATION 651

MAXIMUM SETTLING
\ RATE ATTAINED AT
115 min

2
or
T 20 min 75 min
x
<

\ U = 0,036 IMI/I

o
u
x \
= 0,0485
X
kv<£s = 0,45

i i 1 1 1

100 200
t-TIME, min
(a)

(c)
47
Figure 13.9. (A) Intermediate setting in a desanded mine pulp showing the height of the pulp interface when
photographs in B and C were taken. (B) Commencement of the break-up of the initially gelled mass; t = 20 min.
(C) Agglomerated structure beginning to appear; t = 75 min.
652 HANDBOOK OF POWDER SCIENCE

-VOLUMETRIC SOLIDS CONCENTRATION , therefore replace these corresponding terms

0,05 0,10
for spheres in Eqs. (13.8) and (13.14).
SYMBOL MATERIAL
CONCfi*
SCALE
Observation of the interface setting rate u
0
X
FERRIC HYDROXIOE - REF 55
ACTIVATEO SLUDGE - REF 5€
_
of such suspensions over a range of concentra-
BELOW
[USING ONLY RESULTS TOR TALL
CYLINDERS ( H o ^ 1800mm)]
• tions <j)s can be expected to be correlated
A COAGULATEO PYROPHYLLITE
CLAY-REF 59 (AVERAGED DATA)
ABOVE - according to Eq. (13.19), which may be rewrit-

\V-
ten as:
213
= u° (l - (13.20)
where
uF = terminal settling velocity of one repre-
sentative floe
kv<f)s =volume fraction of floes
kv = volume of floe per unit volume of con-

: \
-
tained solid.

Where, as in the case of hydroxide precipi-

tates or organic suspensions, the volume frac-
C-SOLIDS CONCENTRATION, 9 / 1

tion of the actual solids is not readily deter-

Figure 13.10. Settling velocities at low concentrations
according to Eq. (13.20) of three widely different types mined, ky can be equally well expressed as
of suspensions. k' = volume of floe per unit mass of contained
solid and concentration C = the mass of solids
per unit volume of suspensions. As k' = ky/ps
of the particle structure and hence slow shear- and C = 0S p s , k'C becomes identical to ky</)s,
ing of the suspension results in bonds being that is, it is not necessary to know p s .
broken, a realignment of the particles, release As for dispersed particles in hindering set-
of water, and hence promotion of the subsi- tling, a plot of the experimental values w°213
dence of the interface, curve C 2 in Figure and 0S should give a straight line of intercept
13.8. / = Wp213, and of slope / • ky over the concen-
tration range pertaining to dilute suspensions.
This has been found to apply to a large variety
13.2.4.2 Calculating Settling Velocity of suspensions and plots for a typical metal
of Aggregated Suspensions hydroxide precipitate, clay-mineral and acti-
vated sludge, are shown in Figure 13.10.
The Richardson and Zaki equation [Eq.
A similar settling pattern is demonstrated
(13.19)] had been applied to predict the set-
by each of the suspensions shown in Figure
tling velocities of aggregated suspensions. It
13.10 despite the variation in their nature and
was quite successful in the dilute concentra-
composition. There is the expected linear plot
tion range. Deviations from predictions were
at low concentrations followed by settling rates
encountered for the intermediate suspensions.
that deviate positively at higher values.
The data of Figure 13.10 are replotted, re-
Dilute Suspensions. With water in the voids placing the abscissa ^>s by the floe volume
between the particles of a single aggregate the concentration kcf)s, and normalizing the inter-
volume occupied by the floes, defined as the face velocity with respect to uF, in Figure
total volume of the envelopes surrounding 13.11. It is now the equivalent of Eq. (13.17)
each, is therefore greater than the volume with (assumed) spherical floes in the place of
fraction of actual solids present. The diameter, solid spheres and should therefore appear as
density, and concentration of the floes should the 45° line in Figure 13.7.
 SEDIMENTATION 653

1,0 V^ 1 I 1 1 1
MATERIAL SYMBOL AUTHOR(S) REFERENCE

ACTIVATED SLUDGE + KEINATH et al 67

X COLE 56

0,8 -
• v « FERRIC HYDROXIDE
PYROPHYLLITE
Y
0
A
RUDOLFS ft LACY
BRETTON
SCOTT
68
55
57
FLOCCULATED GLASS
SPHERES NO. 1 0 STEINOUR 69

>^ KAOLIN

CALCIUM CARBONATE
t

A
GAUDIN a
FUERSTENAU
TORY
70
71

DILUTE PULPS ^ ^ ^ ^ - ^ ^ L
0,6
-A
INTERMEDIATE PULPS

• ^

0,4 S ^ ^ ^ ACTIVATED SLUDGE

0,2

EQUATION ( I 4 r \ .
kc = 0,55

i i i i i 1 i i i NI i i i
0,2 0,4 0,6 0,8 1,0 1,2 1,4
/CV<£S-FLOC VOLUME CONCENTRATION

Figure 13.11. Dimensionless maximum interface sedimentation rates; data of Figure 13.10 and other sources
plotted on normalized coordinates.

To demonstrate the general applicability of dilute range settles faster than expected from
this plot, other aggregated suspensions have a suspension of floes assumed to consist of
been included. For all suspensions in the di- individual rigid spheres. This onset of interme-
lute range, where the floes behave as separate diate settling behavior is observed at ky(f)s
entities, there is a straight line of slope = - 1 values from about 0.35 onward (Fig. 13.11).
as expected, up to a characteristic value of The onset would be expected to be related to
&v(/>s, followed by increasing positive deviation the size and shape of the floes and the inter-
from Eq. (13.20). This marks the onset of particle and interfloc forces, which all play a
channel flow and an increased permeability of role in the formation of a structure in the
the suspension as a whole. flocculated suspension. The settling rate is
For a specific slurry the point of departure better determined by direct measurement.
from the straight line indicates the upper lim-
iting concentration of its dilute range. Similar 13.2.5 Measuring Settling Rates
materials, for example, activated sludges (Fig. As discussed earlier, the Richardson and Zaki
13.11) or red mud,58 show similar limiting con- type equations correlate experimental data
centrations, that is, kw<f)s at the point of depar- fairly well to a certain extent. A real world
ture from Eq. (13.20) does not differ signifi- slurry contains a variety of particles with dif-
cantly for the same material even though it ferent shapes, sizes, and densities. The uF in
may be derived from different sources. the Richardson and Zaki equation cannot be
calculated directly but must be determined by
Intermediate Concentrations. The slurry in- actual settling tests (as in Fig. 13.10). As dis-
terface at concentrations higher than for the cussed earlier the settling rate deviates from
654 HANDBOOK OF POWDER SCIENCE

the equation as the concentration becomes than the observed velocity. It is given by
thicker and thicker. It is obvious that an actual
settling test is still essential in studying sedi- = U
s - W
L =
mentation phenomena.
Sedimentation phenomena are usually stud- (13.23)
ied by observing the behavior of suspensions The relative velocity is larger than the abso-
placed in cylinders (frequently with a volume lute velocity (ws, the velocity relative to the
of 1 to 2 liters). Sedimentation in a cylinder wall of settling chamber, that is, the velocity
consists of descent of particles and rise of measured during the settling test). As c£s was
sediment from the bottom. Two typical sedi- essentially zero for the sedimentation of a
mentation curves are shown in Figure 13.12 single particle, the Stoke's velocity requires no
for a slow settling clay (attapulgite) and a fast correction.
settling microbarite (principally BaSO4 used In Figure 13.13 the various stages involved
for weighting purposes in drilling muds). The in a batch sedimentation of dilute to moder-
height of the descending interfaces along AB ately concentrated suspension in a cylinder are
and AC and the rise of the sediment along OB illustrated. Along AB, the rate of sedimenta-
and OD are shown. The slopes of the lines tion is constant and this rate is taken as the
AB and AC yield the settling velocities rela- settling velocity at this initial concentration.
tive to the container walls. For microbarite, From B to C, known as the first falling rate
ws = (420 - 70)/500 = 0.7 mm/s and for period, the slope decreases, indicating that the
attapulgite, line AC yields us = 0.032mm/s. concentration is increasing. Simultaneously,
As the solids settle, liquid is displaced up- the sediment is rising from the bottom as
ward. The downward flux ws(/>s equals the shown by the L versus t curve. When the
upward liquid flux (1 - </>s)«L and upper descending boundary meets the ascend-
ing sediment at the compression point C, the
ws<£s + (1 - 0s)wL - 0 (13.21)
compression period (also called the second
The velocity of the liquid is given by falling rate period) begins. Further decrease in
height is effected solely by flow of liquid out of
= - - 7 - us = •M, (13.22) the compaction zone because of the weight of
the solid particles. When the final structure
The velocity wsr of the solids relative to the carries the entire weight of the sediment,
liquid is the most significant quantity rather liquid flow ceases.
For thickener design (to be discussed later),
the relationship between solid settling flux and
concentration is required. A series of batch

, A CONSTANT
f RATE PERIOD N vs 1 CLEAR

VB .FIRST FALLING
C O N S T J \ f RATE PERIOD M vs 1
CONSTANT
.SECOND FALLING
RATE PERIOD """
MICROBARITE <fe = 0.0305
VARIABLE

—L
MICROBARITE200J 400 \ 600 ^ 8 0 0 1000
ATTAPULGITE 5000 10000 15000 20000
0 TIME, s TIME

Figure 13.12. The sedimentation of microbarite and Figure 13.13. The various stages of sedimentation are
attapulgite. The initial slurry heights are 420 mm for illustrated. Conditions in the cylinder at a time corre-
microbarite and 405 mm for attapulgite. sponding to height H are shown.
 SEDIMENTATION 655

settling tests at different concentrations lead zone mode. Fine, dispersed particles with di-
to a relationship between usr and </>s as shown ameters less than 0.1 micron will diffuse out of
in Figure 13.14. Particles tend to settle inde- the descending slurry-liquid interface into the
pendently in dilute slurries, and consequently, supernatant region.
there is no unique settling velocity for such In Figure 13.14A, the relative sedimentation
slurries. As concentration of the slurry 4>s in- curve is shown as terminating at a value of
creases and settling of large particles is im- ^s = eso> where the particles enter into physi-
peded by the presence of small particles, a cal contact and form a cake. The velocity of
point is reached were all particles presumably the cake surface is no longer a unique func-
have identical velocities and settle as a "zone." tion of the slurry concentration. It depends on
Ultimately, as the concentration continues to the rate at which liquid is squeezed out of the
increase, a point is reached where the solids cake by the weight of the cake. Nevertheless,
form a cake capable of transmitting stresses many investigators have mathematically
through points of contact. The solids then treated the compression zone in the same
enter into the matrix. As a crude approxima- manner as the first falling-rate period. Ulti-
tion the null stress solid concentration es0 mately as shown in Figure 13.12 by the point
marks the beginning of the cake zone. marked t = °o, the sediment reaches a point at
When uniform particles settle, a distinct which there is no more compaction. At that
interface is present even for dilute slurries; point, the solid velocities are everywhere zero.
and the distinction between zone and dilute Very few reliable data involving sedimentation
settling disappears. The extrapolated velocity velocities at concentrations near the cake re-
corresponds to the Stokes velocity. Although gion have appeared in the literature. Data are
there is not theoretical Stokes velocity when different to obtain and difficult to interpret.
slurries with particles having a range of sizes
are involved, an extrapolation to point A as
13.2.5.1 Kynch Theory
shown in Figure 13.14A is employed to pro-
duce a pseudo-Stoke's velocity that can be Kynch132 (1952) made an important improve-
used in empirical correlations. ment in the sedimentation theory. Instead of
In the zone settling region, it is generally performing a series of batch tests to obtain the
assumed that the relative settling velocity is a flux-concentration relation, Kynch developed
unique function of concentration. If the size a means to achieve that by a single batch
range does not include large, dense particles sedimentation test.
or submicron particles with high diffusion co- The first falling rate period as shown in
efficients, settling will be predominantly in the Figure 13.13 is the result of action that takes

A B
Figure 13.14. Relative settling velocities and relative flux as a function of concentration.
656 HANDBOOK OF POWDER SCIENCE

place at the bottom of the cylinder. When

particles reach the bottom and start to form a
cake, the liquid is squeezed out and flows
upward. The upflowing liquid retards the set-
tling of the particles above, resulting in a more
concentrated slurry. The retarding effect prop-
agates upward through the settling particles
and can be treated as a signal carried by a
characteristic wave of constant concentration.
The constant rate period comes to an end
when the first characteristic reaches the
supernatant-slurry interface. Successive sig-
nals originating at the surface of the sediment
lead to a decreasing sedimentation velocity
throughout the suspension.
The equation of continuity in a settling col-
umn is given by
dd), d(u-d)J
—- + —?- = 0 (13.24) Figure 13.15. Kynch construction for a batch settling
dt dx curve.
where
x = distance up from bottom of the settling
moment the particles pass the characteristic,
column
they have the same concentration and same
t = time.
settling rate. A material balance over the char-
acteristic takes the form
If the settling rate is a function of concentra-
tion only, tR<t>sR{v + (-us)} = cf>s0H0 (13-27)
where
d(us4>s)/dx = d{us4>s)/d4>s • d<f>s/dx
<£s0 = initial slurry concentration
c/>sR = slurry concentration at point R
Equation (13.24) can be rewritten as: v = characteristic velocity.
dfa dfc From Figure 13.15, it can be seen v = HR/tR,
—
dt
+*;(&)— = 0 (13.25)
dt ddx substituting into Eq. (13.27) gives
The solution of Eq. (13.25) is
= <f>soHo/iHR + ( - W S ) ' R ) (13.28)
</>SR
x = v((/)s)t + constant (13.26) The settling velocity us is the slope of the
settling curve at R, (dH/dt)t=tR. Also the
Equation (13.26) represents a straight line for
intercept of the tangent at R is
a characteristic of constant concentration.
The method by which Kynch obtained the Hz = HR + (-us)tR (13.29)
flux curve from a single settling plot is illus- Substituting Eq. (13.29) into Eq. (13.28) yields
trated by Figure 13.15. If a characteristic with
slope v emanating from the origin travels up- &R = - T 7 - ^ (13.30)
ward, it meets the supernatant-slurry inter-
face at R. The total solids in the column will Therefore, a relation between the solid con-
pass this characteristic during this period (t = centration and sedimentation rate can be ob-
0 to tR). Because the settling velocity is con- tained along the first falling rate period of a
sidered as function of concentration only, the settling curve.
 SEDIMENTATION 657

Kynch ignored the sediment at the bottom Although wall effects in full-scale equip-
of the settling chamber. Therefore, he argued ment may be safely ignored, design of such
that the constant in Eq. (13.26) is zero and all equipment is often based on settling rates
the characteristics emanate from the origin of observed in small laboratory glassware, for
Figure 13.15. Tiller124 took into account the which corrections may therefore sometimes be
effect of the sediment rising from the bottom, required. In practice this correction is, how-
revised Kynch's argument, and suggested that ever, usually neglected as the consequent
the characteristics come from the surface of error is both small and conservative.
the sediment. Fitch125 considered the charac-
teristic as a kind of concentration discontinu-
ity that emanates either from the origin or 13.3 THICKENING
from the cake surface depending on the initial
concentration of the suspension and the shape Gravity thickening provides a means for eco-
of the flux curve. He states that the surface of nomically removing a large fraction of the
the sediment was also a concentration discon- liquid in a slurry. The process is shown
tinuity propagating upward. At the moment schematically in Figure 13.16 and the equip-
the characteristic leaves the cake surface, these ment used in Figures 13.1 and 13.3. In thick-
two discontinuities should have the same ve- ener technology, a slurry, sludge, pulp, or mud
locity. Therefore, a characteristic should rise all describe a suspension of solid particles in a
tangentially from the cake surface. liquid. Schematically clarifiers and thickeners
appear to be identical, and there is no sharp
13.2.6 The Effect of Container Walls line between the two. In general, clarification
When a particle sediments in a closed column involves suspensions in the dilute ppm
rather than in an infinite liquid, it displaces its (mg/liter) range whereas thickening tends to
volume of liquid from a lower to higher level treat more concentrated slurries in the 1%
and the wall interferes with the ideal liquid and above range. However, it needs to be
flow pattern. This results in an additional drag noted that the meaning of "dilute" or "con-
on the particle and a reduction in the free centrated" varies from industry to industry.
settling velocity ux by a factor W, which de- For instance, the feed to a thickener used in
creases with increasing ratio dp/D, where D mineral industries could be 5% by volume
is the diameter of the container. The retarda- which could be equal or higher than the con-
tion effect may be expressed as follows: centration of a cake produced in municipal
waste water applications. Care must be exer-
uw = u00-W (13.31) cised in interpretation of concentration limits
and many expressions for W have been
proposed.1
According to Francis22
EXCESS

1 ~0A75(d/D)]~4 OILUTE
FEED -
UQUIO

W= (13.32) SLURRY

1 - (d/D) \
for streamlined flow, while Garside and Al-
Dinbouni23 give a simpler equation: W =
[1 + 235(d/D)]~\ applicable for Reynolds
numbers between 3 and 1200.
For a particle of dp = 200 ^m, settling in a
one-liter graduate cylinder (D ~ 60 mm) or in THICKENED SLURRY
UNDERFLOW PUMP
a 25 m diameter tank, the calculated settling VOLUMETRIC RATE 0 m S /h
CONCENTRATION C u k g / m *

velocities, ww, are 99.3% and 99.998% respec- Figure 13.16. Schematic view of continuous thickening
tively of its velocity in an unbounded fluid. process.
658 HANDBOOK OF POWDER SCIENCE

suggested as typical for feed and outputs of the symbols used in this text is listed in Table
solids-liquid separation processes. 13.5. For the most part in the industry, con-
Recovery and further processing of solids centrations are given in mass/unit volume
generally following thickening. Consequently, (kg/m 3 or lb/ft 3 ). They are represented by
the concentration of the underflow is critical the letter C, which is generally employed by
to subsequent operations. The density of the authors writing on free-settling theory. The
underflow from a clarifler is of less impor- same letter C has been used for both the
tance. As less solids are involved, the mechani- suspension and the sediment. The new trend
cal equipment for clarifiers is light compared in the solids-liquid separation field is to use
to that for thickeners in which large volumes volume fractions that provide true concentra-
of dense materials generally require heavy tion comparisons among different processes. It
raking systems. was also found advantageous to use different
In idealized free settling theory, the settling symbols for the free-settling (</>s) and compres-
rate is considered to be a unique function of sion zones (e s ).
slurry concentration. Design methods based
on this principle apply to cases when no sedi- 13.3.2 Thickening in the Free Settling
ments are present, and the underflow is simply Region
a suspension with a higher concentration than Most of the existing design methods for con-
the feed. When higher underflow concentra- tinuous gravity thickeners fall into this cate-
tions are desired, sediments subject to com- gory. The methodology provides a means of
pressive effects due to the unbuoyed weight of determining the area requirements of thicken-
the solids are required. In the sediment, parti- ers. The settling velocities and fluxes are
cles enter into contact, and the solid velocity is required.
no longer a unique function of concentration.
The liquid and solid fluxes are determined 13.3.2.1 Design Procedures
through the use of the Darcy-Shirato126 equa-
tions relating the relative velocity of the solids The Coe and Clevenger Method. Coe and
to the liquid, pressure gradient, and the Clevenger52 were the first authors to establish
permeability. a rational method for the sizing of thickeners.
These two distinct mechanisms of thicken- They studied the settling of metallurgical pulps
ing process are discussed separately in this and correlated batch sedimentation phenom-
text. ena with the design of continuous thickeners.
In a continuous thickener, the settling flux
13.3.1 Nomenclature is taken relative to the bulk flow of the slurry.
Different nomenclatures used in this field have The slurry as a whole is also moving downward
been a source of confusion. A comparison of owing to continuous volumetric draw-off at

Table 13.5. Symbols for Thickening.

VOLUME CONCENTRATION
MASS CONCENTRATION:
SUSPENSION OR SEDIMENT SUSPENSION SEDIMENT
Variable concentration
Feed
Underflow
Critical concentration C crit
Settling velocity
Solid flux (Cws) G
Solid flux at critical
concentration
 SEDIMENTATION 659

the base. If the underflow pumping rate is Q from the one representing the feed to the
m 3 /h and the area of cylindrical section of the thickened underflow. All intermediate concen-
thickener is A m2 then the bulk slurry velocity trations will therefore exist even if only as
is Q/A m / h and the solids flux due to under- transients. The maximum solids throughput of
flow pumping alone, called the underflow flux, a thickener is governed by the concentration
is (Q/A)C. The total flux G T , the solids flux layer that has the lowest solids flux. The mini-
relative to the walls, is the sum of the settling mum value Gmin is then selected for designing
flux and the underflow flux, that is, the cross-sectional area for the thickener.

GT = usC + (Q/A)C = (us Q/A)C Example 13.1. An aqueous slurry of a min-

(13.33) eral is to be thickened from 10 lb/ft 3 to a
concentrated underflow of 60 lb/ft 3 . The
The concentration of the underflow under amount of solids to be recovered is 350
steady-state conditions must meet the materi- tons (dry basis). The density of the solids is
als balance: rate at which solids pass through 200 lb/ft 3 .
the thickener equals the rate at which solids
are discharged in the underflow, that is, a. Calculate the overflow and underflow rates.
The volumetric concentrations of the feed
«,.£. (13.34) (</>sF) and underflow (</>su) are
<£sF = 10/200 = 0.05
Eliminating Q/A between Eqs. (13.33) and
cf>m = 60/200 = 0.30
(13.34) yields
The volumetric rate of the feed and the
GT = (13.35) underflow are
j
QF = 350 • 2000/10 = 70,000 ft 3 /day
Qu = 350 • 2000/60 = 11,667 ft 3 /day
Equation (13.35) is the design equation. To
The overflow rate is
apply Coe and Clevenger's method, the batch
settling rates us at concentrations ranging from Cover = e F - G u = 70,000-11,667
that of the feed to the thickest "free-settling" = 58,333 ft 3 /day.
slurries need to be measured. By inserting the
corresponding values of us and C in Eq.
b. Determine the unit area and the total area.
(13.35), G T can be calculated from the feed
C F to the underflow C u . Thickening is a pro- Batch settling tests need to be conducted at
cess of reducing the mean interparticle dis- different slurry concentrations to obtain the
tance. Therefore, sometime during their pas- settling rate and flux information. The fol-
sage through the thickener, feed particles have lowing batch sedimentation velocities were
to traverse the range of interparticle distances determined at 80°F.

C, lb/ft 3 10 15 20 25 30 40 50 60
us, ft/h 6.13 4.13 2.66 1.65 1.10 0.65 0.40 0.26
2
G T , lb/ft • h 73.6 82.6 79.8 70.7 66.0 78.0 120.0 00
660 HANDBOOK OF POWDER SCIENCE

The solid flux G T is calculated according to both summer (80°F) and winter (40°F) oper-
Eq. (13.35): ation.
The up flow velocity of water is
GT = ^ - £ ) = 73.6 lb/ft 2 • h 58,333/1000/24/3600 = 0.000675 ft/s =
0.000206 m/s. Using Stoke's settling law
The minimum value of G T is 66 lb/ft 2 • h [Eq. (13.8)] us = d2(ps - p L )g/18^, for
at C = 30 lb/ft 3 . This represents the choke summer
points where the thickener is expected to
operate. The total area required can then 0.5
be calculated

area = 350 • 2000/24/66 = 411 ft2 18-0.00086-0.0002061 05

(3210 - 1000) • 9.8 J
Applying a safety factor of 25% leads to
550 ft2, corresponding to a diameter of = 12
27 ft.
For convenience, the variables were con-
c. During winter, the average water tempera- verted to SI units during the calculation.
ture is 40°F. How should this factor affect The result shows a particle larger than 12
the thickener design? fim would not be carried over in the over-
Without an actual settling test of the slurry flow. For winter operation
at 40°F (which is probably the most reliable
way), it can be assumed that the settling 0.5
velocities are inversely proportional to the d=
viscosity of water. Viscosities are
18 -0.00155 -0.000206 "T5
80°F 0.86 cp
(3210 - 1000) -9.8 J
40°F 1.55 cp
= 16
With the velocity reduced by the ratio
0.86/1.55 = 0.55, the minimum flux be- Larger particles can flow out. The thick-
comes 0.55 • 66 = 36.3 lb/ft 2 • h. The area ener has poorer performance in the winter.
required is Bear in mind that this is an oversimplified
calculation as the flow patterns in a thick-
ener involve circulation owing to the intro-
area = 350 • 2000/24/36.3 = 804 ft2
duction of the feed and are complex.
If a 25% safety factor were applied, a 1000
ft2 (36 ft in diameter) unit should be de- The Hassett 65 Method. This is very similar
signed. Clearly, varying seasonal effects are to the Coe and Clevenger method. The graphi-
significant and must be considered. cal representation of the total flux curve [Eq.
(13.33)] is shown in Figure 13.17. Disregarding
d. Assuming that the distance from the sus- the very dilute concentrations that normally
pension surface to the overflow is 5 ft, apply only in clarification, the total flux curve
determine the largest sized particle that can be seen to show a minimum value at M.
would be carried over in the overflow for The concentration of the minimum flux zone
 SEDIMENTATION 661

TYPICAL RANGE OF UNDERFLOW

rYPICAL RANGE OF CONCENTRATIONS C u
THICKENER FEED
CONCENTRATIONS C p
y
^/^ TOTAL FLUX
/

\
TYWCAL RANGE OF
MnLpUK.^C
CLARIFIER FEED

I CONCENTRATIONS

^ ^ | SETTUNG FLUX « « C ^

1
1
1
c
crtt Cu
C -CONCENTRATION

Figure 13.17. Underflow and settling fluxes may be summed graphically to give the total flux (Hassett method).

is (by definition) Ccrit and its total flux (the The Yoshioka Method. In 1957, Prof. N.
minimum value of G T ) is Yoshioka of Kyoto University developed a
procedure that has been looked on with favor
Gmin = ( l ^ + Q/A)Ccrk (13.36) by authors writing on thickening. The Hassett
method suffers from the need to plot a sepa-
where L7crit = settling velocity at Ccrit. Material
rate curve for each underflow rate. In the
balance [Eq. (13.34)] gives
Yoshioka procedure, only one graph is needed.
For illustration, Figure 13.17 is replotted in
r =® (13.37) 13.18. A line is drawn through point P (C u on
the abscissa) at an angle of which the tangent
The underflow flux line (Q/A)G reaches the is —Q/A. Congruency considerations dictate
value Gmin at the concentration C = Cu (Fig. that this line intercepts the ordinate at the
13.17). value of Gmin. The equation for this line is
Determination of Gmin according to Has-
sett's construction requires the drawing of a
total flux curve for each value selected for the (13.38)
pumping rate Q and gives Cu only after Gmin y4
is determined. Although it has the advantage
of clearly illustrating the minimum value of which at Ccrit attains the value Gmin —
G T , it is cumbersome as Cu is normally the QCCTit/A. At this concentration the equation
primary thickening objective. for the settling flux, which is usC = G T -
662 HANDBOOK OF POWDER SCIENCE

QC/A [Eq. (13.33)], also attains the value usC curve. Thus Gmin is obtained from the
Gmin - QCcrit/A. Since G T reaches its mini- intercept with the ordinate and the corre-
mum at M where C = C"ci-it
n sponding pumping rate Q, found from the
slope of the tangent line.
dGT d(usC) Q
03.39)
aC aC Example 13.2. Rework Example 13.1 with
Thus, d(usC)/dt = -Q/A at C = Ccrit. The the Yoshioka method.
line drawn from Gmin to P therefore coincides
with the settling flux curve at Ccrit and forms a b. Determine the unit area and the total area.
tangent to this curve at N.
Yoshioka et al.72 proposed determining Gmin The same batch settling data in Example
more directly than via the total flux curve by 13.2 are used to calculate batch settling flux
starting from Cu and drawing a tangent to the and plotted in Figure 13.19.

C, lb/ft 3 10 15 20 25 30 40 50 60
£/, ft/h 6.13 4.13 2.66 1.65 1.10 0.65 0.40 0.26
2
UC, lb/ft • h 61.3 62.0 53.2 41.3 33.0 26.0 20.0 15.6

The Gmin determined graphically is 66 tion correlates the initial slurry height and
lb/ft 2 • h at C = 30 lb/ft 3 , which is the concentration in a batch test to the continuous
same as obtained by the Coe and Clevenger thickener underflow concentration Cu and its
method. equivalent batch test solids height is

The Talamage and Firch Method. As only Hu = COHO/CU (13.41)

one concentration limits thickener throughput
for a selected underflow concentration (or un-
derflow pumping rate) it is an advantage if this To apply this method:
value could be arrived at directly, thereby
eliminating the need to measure the settling 1. An attainable and acceptable C u is se-
rates of many irrelevant concentration values. lected.
Talmage and Fitch63 developed a procedure 2. Hu is calculated from Eq. (13.41).
for obtaining the required minimum flux from 3. An "underflow line" is drawn parallel to
a single batch settling test on the proposed the time axis at H = Hu. If it intersects the
thickener feed by observing not only the initial settling curve above the compression point,
steady settling rate but also the complete then tu is read directly from the settling
H — t settling curve. curve (Fig. 13.20A). If the underflow line
The Kynch model discussed earlier was ex- intersects the settling curve below the com-
tended by Talmage and Fitch to design a con- pression point, tu is obtained as the inter-
tinuous thickener. The solids flux that can be section of the underflow line with a tangent
passed through the thickener is drawn to the settling curve at the compres-
sion point (Fig. 13.20B). Once the value of
G = C0H0/tu (13.40) tu is obtained, the solids flux which corre-
where tu is the time required for settling solids sponds to the required thickener area is
to reach underflow concentration. The equa- calculated from Eq. (13.40).
 SEDIMENTATION 663

FLUX 6

Cent
C -CONCENTRATION

Figure 13.18. Yoshioka construction on a settling flux curve.

13.3.3 Thickening in the concentration. Such an assumption works rea-

Compression Region sonably well until particles enter into contact
and form a sediment. In the sediment, Darcy's
All the above design procedures employ flux
law applies, and the solid flux is no longer a
theory which is based on the rate of sedimen-
unique function of the solids concentration.
tation being uniquely dependent on the solid
The solid flux is constant and independent of
depth in the steady-state flow. However, con-
centration changes with depth and is a func-
tion of underflow rate and compressibility
parameters.
Basic variables involved in gravity thicken-
ing are underflow concentration (e su ), solid
flux (gs), and the height (L) of the compres-
sion zone. Other important quantities are per-
meability (K\ local solidosity (volume fraction
of solids, es), liquid viscosity (/x), densities of
solids (p s ) and liquid (p L ), and gravity (g).
Flux theory omits most of these parameters
and depends on the solution of the continuity
0 10 20 30 40 50 60 70
equation [Eq. (13.24)]. Design of thickeners
Concentration, C Ib/cubic ft with a compression zone depends on the inte-
Figure 13.19. Yoshioka construction on Example 13.2. gration of the Darcy equation.127"130
664 HANDBOOK OF POWDER SCIENCE

u
TIME
TIME

Ca ) Hu > Hc % -b ,) ..
Hu, <. H
.. c

Figure 13.20. Locating underflow time in the Talmage and Fitch method.

13.3.3.1 Model for Continuous Thickener uid is squeezed upward. At the surface of the
sediment, all the solids are accepted. A por-
The model is based on an idealized flat-
tion of the liquid is rejected and flows upward
bottomed, steady-state thickener. In Figure
and out from the overflow. A material balance
13.21, the various regions are listed as over-
of the flow in the sediment compression
flow, feed-transition, thickening, and dis-
zone is
charge. The increasing volume fraction of
solids from es0 to esu is shown. Both solid and <?F<£SF = 4s = 4su = €sUs = *sBWsB (13.42)
liquid fluxes are constant at any point x above It needs to be noted that qs is equivalent to
the bottom of a thickener. Both solid and G T used in Eq. (13.33) for the Coe and
liquid have downward (positive) velocities, with Clevenger method. The flux of solids (qs) is
the solids flowing down more rapidly. No liq- also called the "superficial solid velocity." The
first term in Eq. (13.42) gives the solids flux
from the feed. The second and third terms
simply reaffirm that the flux qs at an arbitrary
OVERFLOW point equals the flux in the underflow of the
thickener. The last two items provide the same
FEED information with respect to the product of
TRANSITION volume fraction of solids and the local true
80 average velocity (ws,wsB). The pressure drop
LIQUID in a continuous thickener can be expressed by
the Darcy-Shirato equation as:

DISCHARGE U

Figure 13.21. Idealized thickener.

 SEDIMENTATION 665

where the relative velocities of solids and liq- dPs

uid are used. For compressible sediments, the
structure parameters such as permeability (K),
(13.48)
porosity (e), and solidosity (e s ) are considered
to be functions of solids compressive pressure The solids flux qs becomes an input param-
(ps). The liquid pressure pL in Eq. (13.43) eter as well as the underflow concentration
must be replaced by ps before integration is esu. It is necessary to have constitutive equa-
possible. Assuming point contact among parti- tions relating K and es to ps. It is best to
cles, a force balance over a distance dx (Fig. obtain these constitutive relations experimen-
13.22) leads to tally. The following expressions, which have
been used for cake filtration, are adopted here:

i t + ^ +g(Ps£s+PL€)= ° (13*44) ec = e,:

s0 1 + = * „ ! +—
-8

where the effective pressure (ps) represents

the net vertical stress divided by the cross- (13.49)
sectional area. Combining Eqs. (13.43) and
where es0 and Ko are solidosity and perme-
(13.44), the particulate structure equation with
ability under null stress. The degree of com-
ps results:
pressibility is related to the parameters /3, 5,
and pa. All these parameters need to be deter-
mined experimentally.

(13.45) 13,3,3.2 Solution Methodology

The underflow concentration is given by €su = For a continuous thickener operated under
steady-state conditions, the solidosity at the
tfsuAtfsu + 4u> = <!*(<!* + iX s o <1 c a n b e ex-
bottom of the thickener is assumed to be the
pressed as:
same as the underflow solids concentration.
This also implies that the velocity of the solids
(13.46) equals the velocity of the liquid at the bottom.
To start the solution, selected values of esu
and qs are substituted into Eqs. (13.47) and
Substituting Eq. (13.46) into Eqs. (13.43) and
(13.48). At the sediment surface ps = 0. At the
(13.45) and eliminating q yields
bottom ps must equal a value that corre-
sponds to esu. Equations (13.47) and (13.48)
can be solved numerically as long as constitu-
tive relations such as Eq. (13.49) are available.

P + dP dP 13.3.3.3 Thickener Behavior

s s L

X
1 i 1
Equations (13.47) and (13.48) were solved for
kaolin flat D (a type of clay). Figure 13.23
shows that at a given underflow concentration
dx esu, increasing qs results in higher values of L.
t 1 It can also be noted that the plots are charac-
terized by two asymptotes. The horizontal
asymptote corresponds to a long detention
time in which the Darcy term in Eq. (13.48) is
Figure 13.22. Force balance. negligible. A thickener operating far into this
666 HANDBOOK OF POWDER SCIENCE

region would be oversized. The vertical asymp- verted 10 traditional multideck thickeners to
tote represents the maximum flow rate that is deep thickeners.
possible with the required underflow concen-
tration. Physically, the vertical asymptote cor-
responds to a condition in which the unbuoyed 13.4 CLARIFICATION
weight of the sediment is balanced by the
Darcian drag. Therefore, no compressive pres- As implied in the name, the purpose of clari-
sure is available to thicken the solids and an fication is to remove turbidity or suspended
infinite height of sediment is required to ob- solids from a murky liquid and render it crys-
tain the given underflow concentration. Oper- tal clear. It is used in a wide variety of indus-
ation should lie in a range in which flux varies tries, applied to raw materials, intermediates,
from 25% to 75% of the limiting value of the and products and, increasingly in recent years,
flux. to waste streams.
The treatment of raw water supplied; clari-
13.3.3.4 Application of Deep Thickener fication of solutions in the sugar, metal, and
inorganic chemical industries; removal of fine
Deep thickener technology was developed and catalyst particles from petroleum intermedi-
exploited in Alcan alumina plants.131 The sig- ates; polishing of beer in racking tanks after
nificantly higher underflow concentration is addition of finings; and the disposal of indus-
the most significant advantage over conven- trial waste water are but a few examples of the
tional thickeners. It was reported that they can use of sedimentation for the commercial-scale
achieve underflow concentration from 90% to clarification of liquids.
95% of that obtained with rotary vacuum fil-
Sedimentation is, however, not the only
ters. In addition, it had the advantages of
means of achieving this, 5 ' 6 ' 74 and for a fuller
lower capital and maintenance costs, lower
coverage of clarification, the chapters dealing
area requirement, increased recovery of valu-
with filtration methods should also be con-
able chemicals, and even the production ca-
sulted.
pacity. Table 13.6 shows the performance com-
Clarification by gravity sedimentation is car-
parisons of a deep cone thickener with the
ried out in circular tanks similar to that shown
conventional thickeners. As a result, Alcan
in Figure 13.2, but of lighter construction than
had installed 20 new deep thickeners and con-
those shown in Figures 13.1 and 13.3, and also
in rectangular tanks.75 In the treatment of
10 potable water, long rectangular basins are con-
sidered to be hydraulically more stable, with
' TIME. DAYS less short-circuiting between feed and over-
flow points, especially in larger plants.76 In
1.0 flocculated sewage treatment in the Toronto
LU area,77 a long, rectangular horizontal-flow set-
X tling tank was stated to be "much better" than
z a circular tank even when based on the same
LU 0.1 overflow rate and detention time.
5 In the food industry, the relatively pro-
LU
CO longed residence time in normal gravity set-
0.01 tling tanks sometimes leads to fermentation
-8 -6
10 10 10 and deterioration. The processing time and
SOLID UNDERFLOW RATE, the liquid inventory may both be reduced,
q su , m3/m2/s however, by the use of centrifugal clarifiers. In
Figure 13.23. Height versus solid underflow rate at these the gravitational force, g, is increased to
constant underflow concentration for kaolin flat D. 1000 to 10,000 times with a corresponding
 SEDIMENTATION 667

Table 13.6. Comparisons of Conventional with Deep Thickener.11

CONVENTIONAL
THICKENER DEEP THICKENER
Diameter 120 ft 40ft
Height 15-20 ft 40-60 ft
Underflow concentration 30-35 wt% 45-55 wt%
Overflow clarity < 200 mg/liter < 100 mg/liter
Flocculant dosage 20-40 g/ton 50-80 g/ton
Upward velocity 0.5 m/h 3 m/h
Solids loading 1 to 2 mt solids/m2 day 10-15 mt solids/m2-day
Capital costs very high 2-4 times lower
From Ref. 131

increase in the settling velocity of the solids. As a consequence, the incoming feed, which
Examples are the treatment of olive oil to may differ both in density and temperature
avoid rancidity and the separation of yeast from the contents of the tank, can readily
cells from beer to cause a rapid termination of upset the ideal flow pattern. The mode of feed
their growth not possible in normal gravity entry and overflow removal is therefore more
sedimentation.120 Detailed discussion of cen- critical than in thickeners76'78 and model stud-
trifugal sedimentation5 is, however, beyond the ies should be used to investigate the hydraulic
scope of this chapter. effects of novel designs79 including the effects
Clarification of gold-bearing solutions in of baffles, weirs, and distributor plates.80
Southern Africa is traditionally done by pre- In thickening, the solids settle by hindering
coat filtration. A full-scale test133 has shown, settling with an interface between suspension
however, that a prior gravity sedimentation and liquid so that size segregation of fines is
step can reduce the overall cost of clarification minimal. Because of the lower solids concen-
to 60% that of filtration alone. This specific tration in clarifiers the floes descend indepen-
example confirms that the relative lower cost dently, with the larger particles or floes reach-
of gravity sedimentation in general, indicated ing the sludge level faster than the slower
in Table 13.1, applies also to clarification. settling fine material. If it were not for this
range of settling rates, the normal settling flux
13.4.1 Comparison of Clarifiers curve (Figs. 13.17 and 13.18) could be used in
and Thickeners design. Instead, a different flux curve would be
required for each species of particle size and
The concentration of solids in feed to clarifiers particle concentration present. However, as
ranges from 0.1 to 10 kg/m 3 which is between the maximum solids throughput is not of
20 and 1000 times more dilute than for thick- primary importance in clarifiers, a different
eners. As the main purpose of clarification is approach is used in design.
the removal of the solid matter from the liq- More details of the conditions necessary for
uid, the concentration at which these solids the separation of various sizes and densities of
(usually waste) are rejected from the clarifier particles in a mixed suspension are given by
is of reduced importance. The solids are usu- Masliyah.123
ally finer than in thickener feeds and require
flocculation for efficient settling. In a fully
loaded thickener, the deep layers of sediment- 13.4.2 Pretreatment for Sedimentation
ing and compacting solids are a prominent Effective clarification often depends on floc-
feature of its depth-concentration profile culating agents for success. Even when sedi-
whereas in clarifiers the major volume of the mentation is feasible without such an aid,1
tank is occupied by relatively quiescent liquid. pretreatment can result in both a reduction of
668 HANDBOOK OF POWDER SCIENCE

the size of tanks required and an increased ent mechanisms, coagulation is a reversible
clarity of the product. Pretreatment also ap- process whereas flocculation is not.
plied to thickener feed when throughputs have The combined use, first of an electrolyte to
to be increased but this is considered to be a reduce repulsive charge, followed by a reduced
more expensive remedy than installing addi- quantity of the relatively more expensive poly-
tional tanks for the long term.73'79 mer, often leads to a less costly pretreatment
Flocculants are additives that cause sus- process than the use of either alone. There are
pended solids to agglomerate into floes which of course restrictions in selecting flocculants
act like single large particles and therefore for potable water and foodstuffs. Less efficient
settle more rapidly than their smaller compo- but edible natural products such as starches
nents. Floe formation is brought about by co- and gums find a useful application here.
agulation, by capture in hydrous precipitate, or The quantity of flocculant normally re-
by the formation of polymer bridges between quired to cause efficient flocculation is only a
particles. small fraction of that which can be adsorbed
Coagulation occurs when the mutual elec- on the relatively large solid surface available,
trostatic repulsive forces between particles are and polymer flocculant molecules are usually
sufficiently reduced, by the addition of ions of quickly and completely removed from solution.
opposite sign, to permit the London-van der Contamination of the clarified liquid with
Waals attractive forces to cause aggregation of residual free flocculants is therefore usually
the particles. This requires either a pH change absent but not impossible and could lead to
or the addition of preferably polyvalent ions or problems at another point in the circuit1 or in
a combination of these actions. Lime and alum the application of the product. A bigger prob-
are common coagulation additives. lem at the flocculant addition stage is to en-
At a suitable pH, alum addition will lead to sure that the limited quantity of flocculant is
the formation of a hydrous aluminum hydrox- equally distributed between all the particles in
ide precipitate in which particles may be cap- the suspension.
tured. Although the principles of flocculation are
Polymer flocculation, whether by natural or reasonably well understood, selection and ap-
synthetic neutral polymers or polyelectrolytes, plication of the best flocculant for a particular
may be considered82 to take place in two suspension is still an art. 1 ' 5 ' 83 ' 84 Thus nega-
stages: tively charged solids may be flocculated by
cationic flocculants as expected but it is also
possible that better results may be obtained
1. adsorption of the polymer onto a particle
with an anionic polymer after addition of a
surface, attributed to hydrogen bonding or
divalent cation.82 Determination of the best
ion adsorption, and
conditions and selecting the best product from
2. flocculation of the particles either as a a range of similar type flocculants is therefore
direct result of the London-van der Waals based on results of laboratory batch tests on
attractive forces or due to physical polymer the liquid to be clarified.87 Such tests, if prop-
bridges formed between the particles. erly carried out, can indicate not only the
chemicals to add, the required amounts and
These bridges may be formed by the two ends the order of their addition, but also the degree
of one polymer molecule being attracted to of stirring, method of application, and the wait
two different particles,83 or by the "loops" of period required either before the next addi-
polymer chains on one particle being attracted tion or the commencement of sedimentation,
to the loops of another.82 Polymer flocculation that is, the point in time at which the floccu-
is extremely sensitive to the molecular weight lated suspension should be admitted to the
of the polymer used.84 Because of the differ- clarifier.
 SEDIMENTATION 669

A typical sequence is: claimed that well-formed floes, produced in a

central (feed well) zone, are smoothly trans-
1. addition of lime solution to the feed under ferred to the clarification zone with minimal
conditions of rapid mixing to ensure good floe disruption. This leads to a simpler process
dispersion and adsorption onto the solids, and reduced reagent consumption.
2. addition of polymer solution, as dilute as In repeated settling of the same flocculated
possible (normally a few ppm) and at more solids such as in countercurrent decanting, floe
than one point of application, also under shear in the transfer pumps is unavoidable.
conditions of rapidly mixing to ensure good Although this can be minimized,91 it is nor-
distribution in the liquid, and finally mally necessary to add a further quantity of
3. slower mixing or gentle shear of the clari- flocculant at each stage, half or one third of
fier feed so that the treated particles, now the original dose being fairly typical. The ac-
ready to adhere to each other, grow into tual quantity required can be adjusted by
floes and then into larger aggregates to observing the performance of the clarifiers
cause rapid sedimentation and to incorpo- involved.
rate the finer particles which otherwise
cause residual turbidity after settling. 13.4.4 Final Settled Volume

It may appear at first to be paradoxical, but

It must, however, be borne in mind that in the greater the attraction between particles,
designing flocculation equipment the detailed brought about by pretreatment, the further
mechanism of polymer flocculation are gener- they stay apart in the final sediment. This is
ally not as well understood as the theory, and related to the stability of an open structure of
the process cannot be applied without some adhering particles compared to the same
empiricism and resort to pilot scale experi- structure when the particles are dispersed. The
ments, and even then the results cannot be latter, as they descent individually from sus-
scaled up with certainty.88 pension, cannot remain in contact with the
first two or three particles they encounter in
13.4.3 Floe Strength the build-up zone, because of the repulsive
charge, and continue to descend by sliding or
Excess shear of the formed floes should be
rolling until they reach a point of maximum
avoided, especially those produced by irre-
stability where they fill the lowest remaining
versible polymer bridging. In one difficult case,
gap in the structure. Dispersed sediments are
dropping from the end of the launder into the
therefore compact while flocculated sediments
feed well was sufficient to rupture the floes.
are voluminous49 and more readily break down
Removal of the launder from the feed well
on shear.
and lowering its end a few centimeters below
the water surface of the clarifier provided a
13.4.5 Theory and Design
smoother entry for these fragile floes and solids
throughput was increased two to three times Clarifiers are sized on the basis of the permit-
as a result. Smith and Kitchener89 present ted (or desired) upper limit of solids in the
three techniques whereby the strength of par- overflow and the settling rate of these solids.
ticle adhesion in flocculating media may be If the size distribution of the solids in the feed
measured. They found, as a quick check, that is known, it can be converted to a settling
an increased floe size was a good indication of velocity distribution according to Eq. (13.8). A
an increased strength of adhesion. typical plot is shown in Figure 13.24.
The processes of coagulation, flocculation, As the feed enters the clarifier the particles
and then settling of the floes can be combined with settling velocities greater than the upflow
in a single vessel90 on a large scale. It is rate settle to the bottom while the rest are
670 HANDBOOK OF POWDER SCIENCE

a size distribution of this material, 96.4%Of

the solids have settling velocities exceeding
0.052 mm/s. The maximum upflow rate, ws? j s
therefore 0.052*3600/1000 = 0.187 m/h.
As throughput qF = us* area of clarifier (A)

600 m 3 / h = 0.187 m / h * A m2
QUARTZ PARTICLES SUSPENDED
IN TURBID LIQUOR
A = 3200 m2
and diameter of circular tank = 64 m.
(2) Clarification of a liquid after flocculation
99,99 99,9 99 90 50 10 I 0,1 0,01
CUMULATIVE MASS PERCENTAGE OF PARTICLES SETTLING GREATER THAN INDICATED VELOCITY Assuming that 64 m diameter tank is too
Figure 13.24. Settling velocity distribution of sus- large for the site, what steps can be taken to
pended quartz particles. reduce it? Flocculant tests in the laboratory
indicated good settling behavior of the solids
in the feed after addition of 5 mg/liter of
carried over. The proportion of feed particles
ferric chloride coagulant followed by 0.2
appearing in the overflow can be determined
mg/liter of a polyacrylamide flocculant.
from Figure 13.24 for any given overflow rate
For sizing the clarifier a 10-liter sample of
and their actual concentration then depends
the pretreated turbid gold solution was then
on the total quantity of solids in the feed. The
gently added to a transparent settling tube
settler area is governed by the maximum al-
~ 65 mm diameter and 2 to 3 m deep. Provi-
lowable overflow rate.
sion was made for periodically sampling the
When the "particles" to be settled are floes
liquid at known depths either by lowering a
or agglomerates of the originally dispersed
siphon tube or through suitably spaced side
fine solids, their size characterization is not
ports.2'92
simple. It is best to directly measure in the
At time zero, the tube was immediately
laboratory the essential parameter, settling
filled, the contents were sampled at the top,
velocity distribution. This is illustrated in the
middle, and near the bottom to determine the
examples below.
original (feed) suspended solids concentration
and check on even solids distribution. When
13.4.5.1 Two Examples of Estimating
the liquid started to clear, the time was noted
Clarifier Areas
and samples taken at all levels from top to
(1) A liquid containing dispersed solids bottom in that order, and analyzed for sus-
A pregnant gold solution, obtained by ro- pended solids. The sampling was repeated af-
tary vacuum filtration of cyanide-leached ore, ter four or five similar periods until all
contains 550 mg/liter of 1 to 30 /xm quartz samples indicated a suspended solids con-
fines and requires clarification to a limiting centration below the desired limit. Results of a
concentration of 20 mg/liter. What size sedi- typical test are shown in Table 13.7.
mentation tank is required for a flow of 600 Sampling at H m below the surface after
m 3 /h? t h static settling is exactly equivalent to
The solids to be removed are (550 - 20) sampling the overflow liquid from a contin-
mg/liter or 96.4% of the mass of incoming uous clarifier operating at an upflow velocity
solids. As only solids that settle faster than the qF/A = H/t. It can be seen from the results,
upward velocity of the overflow can be col- as is to be expected, that the suspended solids
lected, the maximum overflow rate must be (SS) at any level decreases with time and at
low enough to collect this percentage of the any time increases with depth. These data
incoming solids. From Figure 13.24, based on provide the maximum permissible upflow ve-
Table 13.7. Suspended Solids Values at Various Depths at Different Times.

TIME OF SAMPLING (min)

nCDTU

FROM 0 5 10 15 20 25
SURFACE ss SS "8 SS "8 SS M8 SS u8 SS
(m) (mg/liter) (m/h) (mg/liter) (m/h) (mg/liter) (m/h) (mg/liter) (m/h) (mg/liter) (m/h) (mg/liter)

0.25 3 10 1.5 3 1 2 0.75 1 0.6 2

(4.8) (20)
0.5 550 6 30 3 7 2 3 1.5 2 1.2 Nil
1.0 12 140 6 23 4 7 3 3 2.4 3
1.5 560 18 380 9 65 6 15 4.5 5 3.6 4
(6.5) (20)
2.0 24 520 12 140 8 35 6 9 4.8 5
2.5 540 30 550 15 250 10 65 7.5 15 6 7
672 HANDBOOK OF POWDER SCIENCE

locity for any desired suspended solid in the which represents a narrower tank but much
overflow. By interpolating the SS data after 5 deeper. This apparent wide choice of possibili-
min settling, it may be estimated that for a ties is, however, limited by considering the
clarifier product containing no more than 20 standard sizes available as settling tanks of
mg/liter of suspended solids, the upflow must this diameter normally come with a standard
not exceed 4.8 m/h. For a volumetric through- depth of ~ 2.5 m, and therefore us * tB = 2.5
put qF = 600 m 3 /h this means a tank of diam- m. The products us * tu for various times taken
eter =• 12.6 m. from Table 13.7 are shown in Table 13.8.
If the exercise is repeated with the 15 min From the final column the required value of
results, however, the diameter is found unex- 2.5 m occurs at about 19 min; therefore:
pectedly to decrease to 10.8 m. This could not
fD = 19/60 = 0.315 h
occur unless there were a change in the nature
of the settling solids with time. For the various us = 2.5/0.315 = 7.9 m / h
times shown in Table 13.7 it can be seen, and
however, that the SS value for a fixed upflow D = 9.8 m
rate, say us = 6 m/h, decreases steadily with
time from 30 mg/liter at 5 min to 7 mg/liter In actual practice a converted 10 m diame-
at 25 min. Therefore, both upflow rate and ter pachuca tank was used133 for this duty, in
detention time are important in clarifier de- this case probably by reversing the procedure
sign. (Note: This aspect of detention time is and adding the necessary quantity of floccu-
quite different from the idea that it may be lant to suit the size tank available. As for
required to achieve maximum sludge thicken- thickeners, the calculated value of D is
ing.)90 Particles or floes continue to grow dur- rounded up to the next largest standard diam-
ing settling, either as a result of faster settling eter tank manufactured and this has the dou-
units overtaking and coalescing with slower ble advantage of further lowering the upflow
ones or due to velocity gradients in the fluid.143 velocity and increasing the detention time,
In a cylindrical tank of diameter D the which is always useful for contingencies.
detention time t0 = irD2H/4qF while the up-
flow rate is the detention time t0 4qF/irD2.
Therefore the dimensions of the tank are 13.5 NONCONVENTIONAL
SEDIMENTATION PROCESSES
AND EQUIPMENT
D=
7TUC

and Normal thickeners and clarifiers are large rel-

atively empty tanks or basins that provide vir-
tually quiescent conditions and sufficient sur-
In the example qF = 600 m 3 /h and if we face area and volume for optimum settlement.
select a detention time This simple process has changed little since
tD = 5 min, us must be 4.8 m / h for SS the turn of the century, and, based on the
< 20 mg/liter
D = 12.6 m Table 13.8. Values of tD * us Calculated
and from Table 13.7
H= 0.40 m
TIME ws FOR 20 mg/liter tD*us
If a longer detention time is selected, say (min) SUSPENDED SOLIDS (m)
tB = 25 min, then us becomes 10.8 m / h (by
5 4.8 0.40
extrapolation): 10 5.6 0.93
D = 8.4 m 15 6.6 1.65
and 20 8.3 2.77
H = 4.5m 25 10.8 4.50
 SEDIMENTATION 673

percentage tonnage currently handled, ap- FLOOR AREA =

pears to be as well established as ever.
However, situations do arise where the con- SETTLING AREA = W . N . —
tana

ventional design is not ideal for special condi- FLOOR : SETTLING AREA RATIO = — •

tions. These include sedimentation of solids

whose density is only marginally greater than
that of the fluid; the clarification of high- CLARIFIED WATER

temperature or very volatile liquids that must

in consequence be kept under pressure during
treatment, or when new or additional sedi-
mentation equipment must be installed in ex-
isting premises, and floor space has become
severely limited. In some cases the cost of
creating the additional space may be many
times greater than that of the equipment it-
self. In underground mining for instance, 10 to
20 m3 of virgin rock has to be blasted out for
every m2 of settling area95 required. Such situ-
ations have led to the introduction of a num-
ber of innovations, some of which have reached
the commercial exploitation stage.
' SOLIDS TO
Their common purpose is to increase solids COMPACTION

throughput per unit of superficial area. This

Figure 13.25. Principles of the lammella settler.
can be done either through an improvement in
actual particle sedimentation rates whereby
more solids are settled per square meter of
actual settling area or due to a multiple design closely spaced, and at as shallow an angle a as
where, for the same projected floor area, a possible. The minimum value for a is pre-
larger net settling area is provided. Mechani- scribed by the flow properties of the sludge.
cal innovations such as improved designs of For plates of H = 1.5 m, at the usually
feedwell and rake are also presented. accepted angle of a = 60° and spaced d = 30
mm apart, a 10-plate tank has an actual set-
13.5.1 Lamella Settlers tling area about seven times greater than a
conventional settler in the same space, whereas
According to Hukki et al,96 the use of inclined with 20 plates, this increases to 11. In the
baffles for settling fine solids dates back to limit, as N -> °°, the advantage becomes equal
1882 when a French patent was granted to to H(d/cos a), which in this case is 25. As the
Gaillet and Huet. The principle has since then denominator d/cos a represents the vertical
"been rediscovered many times in many coun- distance between adjacent plates, the maxi-
tries."96 Figure 13.25 shows the angle and mum advantage is, as would be expected, the
general arrangement of the baffles in the early number of plates that overlap in the vertical
Hukki design, the mode of settlement of the direction.
solids, the method of calculating the floor area This theoretical figure is rarely achieved in
required. The ratio of floor area to settlement practice, however, for, besides the obvious end
area is effect, factors contributing to the lower effi-
1 d \N-1 ciencies in the earlier designs are:
N H cos a [ N
and so to minimize R', the number of plates 1. Manifolding problems of feeding a large
should be large, and they should be deep, number of plates
674 HANDBOOK OF POWDER SCIENCE

2. Entrainment of solids in the rapidly rising Willis103 reviews the practical design factors
liquid stream due to internal mixing97 giv- for tubular settlers including important points
ing rise to an increased overflow turbidity such as sludge collection, how to specify over-
3. Accumulation of solids on the plates. flow rate, and presents various shapes of tube
that may be used. Much of it applies also to
These shortcomings, which are not always seri- conventional plate settlers but tubes are con-
ous, have been recognized by various manu- sidered to overcome the hydraulic instability
facturers and the following modified designs of "wide horizontal plates."
are available commercially: An alternative view,97'104 that lamella set-
tlers offer the advantage of rapid sedimenta-
1. Corrugated plates to avoid a continuous tion or additional clarifier capacity because of
curtain of descending solids, and guard gut- the decreased vertical fall height, can lead to
ters to separate the flow of these solids confusion between a reduced throughput time
from the incoming feed and a real increase in total solids throughput.
2. Feeding the plates from the side to avoid Only the former applies here as solids
feeding and discharging solids at the same throughput is governed by available area and
point not settling height.
3. Feeding from the top with overflow return
pipes extending to clear water zones near
13.5.2 Upflow Solids Contact
the base
4. Stacking set of inclined plates, one above In contrast to the lamella settler, which in-
the other, each inclined in opposite direc- creases sedimentation "efficiency" by provid-
tions to present a vertically zigzag profile ing a multiplicity of settling planes, the upflow
for which settling solids should present the principle operates by improving the actual sed-
minimum disturbance to the clarified liquid imentation characteristics of the feed. It is
5. Continuous raking or low-frequency vibra- confined to flocculated suspensions or metal
tors to assist both removal of solids from hydroxide precipitates in which the freshly
the plates and promote their compaction formed loosely knit, voluminous floes settle
6. Use of flexible textile materials or rubber slowly owing to their smaller initial mean size,
for the "plates" to permit periodic dis- and their greater fragility, or their decreased
lodgement of solids or cleaning. density differential compared to compact
agglomerates.
The lamella settler has been successfully Such floes may, however, mature with time
used in many fields; in coal preparation resulting in larger, stable, denser, fast settling
plants98'99 for removal of fine mill scale from units, and when these aged floes contact freshly
hot rolling mill wastes for reuse of the water100 formed material the settling characteristics of
and also for separating metal hydroxides, fly the latter are promoted.80 The upflow princi-
ash, nickel, catalyst fines, cement dust, clarifi- ple exploits this phenomenon by adding the
cation of phosphoric acid, lime kiln scrubber freshly flocculated feed, not into a normal
water and paint booth water curtain.100 feedwell above a body of clear liquid, but
The tubular settler101'103 and the rotating below the pulp interface of a bed of aged floes.
spiral thickener104 operate on the same general These floes are thereby kept in a state of
principle of providing increased settling area fluidization, and feed rates of up to 35 times
per unit of floor area. In the latter, a "Swiss higher than for conventional units are claimed.
roll" provides a number of flow channels be- This is possible if the matured floes have a
tween curved walls through which the slurry settling rate 35 times faster than freshly formed
flows and concentrates while the unit is slowly floes. This represents a special case of the
rotated. benefits of detention time except that in this
 SEDIMENTATION 675

case the mass of the solid rather than the

volume of the liquid is used to calculate
detention.
Relatively long residence times are possible
because of the high concentration of the -•-SOLIDS BED

solids in the fluidized zone. Escape by short-

circuiting of the slower settling fresh floes with
the water from this zone is reduced by provid-
ing a zone sufficiently deep and uniform. There
must therefore be some "filtering" action
whereby the new floes or their fragments col-
lide with and are held by older floes until they THICKENED PULP

too mature and trap fresh material. The need

to maintain a stock of fluidized solids means, Figure 13.27. Upflow solids contact in thickening.
however, a relatively long start-up period
(while running at low efficiency) and the
inability to recover quickly following flow gangue, in this context it refers specifically to
reductions or shut-downs. the removal of all solids suspended in a liquid
The original application appears to have to achieve complete clarification.
been for water clarifications105'106 and a sim- Where the density difference between solid
plified design is shown in Figure 13.26. More and liquid is slight, as for example, the sus-
recently it has been applied to conventional pended organic solids in sugar juice and in
thickener feeds such as 20% uranium sewage, sedimentation rates are low and sludge
slurries107 or zinc mine tailings108 in units volumes usually high. Attachment of gas bub-
shown schematically in Figure 13.27. bles to the particles or floes reduces their
A useful comparison of solids contact clari- density still further to such an extent that the
fiers, lamella separators, and inclined tube set- density different between "particles" and liq-
tlers is given by Mace and Laks109 including uid is reversed and increased. Therefore, the
cost data. driving force for separation is also increased.
Owing to the reversal in actual densities,
13.5.3 Flotation the floes now rise but at a rate more rapid
Although froth flotation is used on a wide than they descended previously and collect as
scale for separating between solids, particu- a froth at the surface. They form a solids zone
larly the flotation of mineral particles from partially above the liquid surface so that the
solids can drain to produce a "sludge" of con-
FEED LAUNDER
siderably reduced liquid to solids ratio.110
The bubbles are normally generated by re-
leasing the pressure of an air-saturated liquid
stream111 or by electrolysis.112
Further aspects of the process and detailed
design of air flotation systems will be found
elsewhere.113

13.5.4 Feedwell Design

Unconventional feedwell designs that have
Figure 13.26. Upflow solids contact in water clarifica- been proposed aim to improve the hydraulics
tion. of feed entry or the settling characteristics of
676 HANDBOOK OF POWDER SCIENCE

the solids.4'73'114 Tests of feedwell designs per- ACKNOWLEDGEMENT

formed on model thickeners should be inter-
preted with caution and if possible repeated The author extends his gratitude to Professor
on a full-scale plant before making claims.4 Frank M. Tiller of the University of Houston
The dynamic effects are accentuated on mod- for his review and recommendations during
els, and whether good114 or bad122 generally the preparation of this manuscript.
decrease on scale-up.
In clarifiers, internal circulation (caused by
density currents) is to be avoided because it LIST OF SYMBOLS
leads to short-circuiting and higher local up-
flow velocities near the overflow. Various ways A Area of horizontal cross-section of
of overcoming this defect are discussed and a thickener or clarifier
new type of feedwell presented by Firch and Ah Area of batch test cylinder
Lutz.115 Within the normal cylindrical well are a Divisor for calculation of Re defined
two adjacent, inwardly open, submerged chan- in Table 13.2
nels between which the feed is split and intro- BF Buoyancy force or Archimedes' up
duced to each tangentially but in opposing thrust = V - p L *g
directions. When these streams meet in "open b Exponent for calculation of Re de-
water" near the center of the well their veloci- fined in Table 13.2
ties are instantaneously dissipated as turbu- C Mass concentration of particles, that
lence and residual velocity streams are thereby is, mass of oven dry solids per unit
eliminated. volume of suspension
Ccrit Concentration of the minimum flux
zone
13.5.5 Rake Design
CD Dimensionless drag coefficient de-
In common with other materials throughout fined in Eq. (13.5)
the world, the gold ores mined on the East CF Concentration of thickener feed
Rand in South Africa tend to be especially ^max Maximum concentration attained by
"sticky."116 The clayey solids accumulate117 on random close packing of spheres, that
the arms of conventional rakes and often cause is, in the absence of compression
disastrous shock overloads when they eventu- Co Initial uniform concentration in a
ally fall off in massive chunks. It is then also batch test
extremely difficult at the next half turn to rake Cu Concentration of thickener underflow
the resultant "islands" to the discharge point. D Diameter of settling vessel
The Klopper Dragrake118 was developed to d Distance between plates in lamella
overcome this type of problem. It consists of a settler (normal to plates)
tubular steel arm to which are welded the dp Particle diameter
raking blades. This arm is pivoted at the base dA Diameter of a sphere having same
in the center and suspended by means of surface area as a nonspherical parti-
cables or chains from a drag arm rotating cle
above the liquid. Solids build-up on the arm is dM Diameter of circle with same pro-
minimal and it has the ability to swing and lift jected area as image of particle in
automatically over any unusual loads. The microscope
great usefulness of this device is confirmed by dns Diameter of sphere with same termi-
its being used elsewhere119 and by being taken nal settling velocity as nonspherical
up by international thickening manufacturers particle
for the general treatment of thixotropic dscr Diameter of sphere just passing same
slurries and tacky muds. square aperture as particle
 SEDIMENTATION 677

Fv Drag force on particle moving relative lamella settler to vertically projected

to liquid settling area of all its plates
G Solids flux Re Reynolds number of a particle
Solids flux in a clarifier (= J p wp L //x)
GT Gravitational force of attraction = V • t Time
Ps' g 'D Detention time in a clarifier
Minimum value of total flux G T tu Time at which Hu intersects exten-
G7 Total flux sion of tangent to batch settling curve
g Acceleration due to gravity drawn before compression point
Height (1) to suspension-supernatant t1 Period during which initial settling
H rate u{ increases to a maximum steady
interface in a batch test; (2) Vertical
height of lamella settler value us; the induction period
Ho Height of suspension in batch test at u Absolute velocity of liquid in a thick-
zero time ener
Underflow line = HOCO/CU uF « 0 for a suspension of floes
/ Property of a particle-liquid system u{ Initial settling rate of an intermediate
independent of w s (= C D R e 2 ) suspension
K Permeability of a particular bed wns Terminal settling velocity of a non-
Permeability under null stress spherical particle
Shape factor = dp/dA u0 Settling velocity of a single particle,
Effective volume of a sedimentation representative of a suspension, that is,
entity (single rough particle, cluster or suspension settling velocity at C = 0,
floe) per unit volume of contained equivalent to ww
solids, that is, ratio volume of parti- us Settling velocity of a particle, a con-
cles plus fixed water to volume of centration zone, or a suspension-
particles supernatant interface relative to the
k' Effective volume of a floe per unit container
mass of contained solids ( = kv/ps) usB Absolute solid velocity at the bottom
N Number of plates in a lamella settler of thickener
n Exponent in the Richardson and Zaki
*s,crit
Settling velocity of minimum flux zone,
equation = 4.65 + 19.5 d/D in ^crit
streamlined flow, ~ 4.7 for most slur- uST Relative velocity between particle and
ries in gravity sedimentation practice liquid
Compressibility parameter in Eq. ww Settling velocity in presence of con-
(13.49) tainer walls as opposed to settling in
PL Hydraulic pressure an infinite liquid
Ps Solid compressive pressure in the sed- u^ Terminal settling velocity of a sphere
iment in an unbounded liquid in streamlined
Q Volumetric underflow discharge rate flow
from a continuous thickener w^t As for u^, but in transitional or tur-
Volumetric flow rate of liquid bulent flow
Volumetric feed rate V (1) Volume of particle; (2) Volume of
Volumetric flow rate of solids in a thickener
thickener W (1) Retardation factor due to pres-
Minimum value of total solid flux ence of container walls, Eq. (13.31);
Volumetric flow rate of solid at the (2) Width of lamella settler
underflow x Distance up from bottom of the set-
R' Ratio of floor area required to locate tling column
678 HANDBOOK OF POWDER SCIENCE

a (1) Slope of "underflow flux line," 6. B. Fitch, "Choosing a Separation Technique,"

tan a = Q/A. Equals rate of rise of Chem. Eng. Prog. 70(12):33-37 (1974).
7. D. B. Purchas, "Solid/Liquid Separation Equip,
C crit in batch settling test; (2) Angle ment: A Preliminary Experimental Selection Pro-
of plates in a lamella thickener to the gramme," Chem. Eng. No. 328:47-49 (January,
horizontal 1978).
p Compressibility parameter in Eq. 8. B. S. Okunev, I. I. Morosova, Yu. A. Tisunov, and
(13.49) N. F. Kotov, "Automatic Control of Radial Thick-
ener Operation," Coke Chem. USSR (12) 49-52
8 Compressibility parameter in Eq.
(1977).
(13.49) 9. C. G. Stokes, "On the Effect of the Internal
e Porosity Friction of Fluids on the Motion of Pendulums,"
es Solidosity, solid volume fraction in the Trans. Cambridge Philos. Soc. 9(II):8-106 (1851).
thickener 10. I. Proudman and J. R. A. Pearson, "Expansions at
Small Reynolds Numbers for the Flow Past a
e s0 Solidosity under null stress
Sphere and a Circular Cylinder," /. Fluid Mech.
esu Solid volume fraction at the under- 2:237-262 (1957).
flow 11. G. K. Batchelor, An Introduction to Fluid Dynam-
fi Liquid viscosity ics, Cambridge University Press, London, p. 233
v Characteristic velocity; slope of set- (1967).
tling flux curve 12. R. H. Perry and D. Green, Chemical Engineers'
Handbook, 6th edit., McGraw-Hill, New York. 12a,
pL Liquid density pp. 19-64, 12b, pp. 5-63, 64, 12c, pp. 1101 [Third
ps Particle or solids density Ed.-(1950)] (1984).
<£s Volume fraction of particles in a sus- 13. M. B. Sonnen, "Subroutine for Settling Velocities
pension = C/ps of Spheres," Am. Soc. Civil Eng. J. Hydraulics Div.
</>scrit Solid volume fraction at a minimum 103 (HY9) 1097-1101 (1977).
14. V. F. Swanson, "A Free-Settling Equation Valid
flux
for All Particle Sizes," in Conference: Proc. Tech.
</>sF Solid volume fraction of the feed Program: Int. Powder Bulk Solids Handl. Pr., pp.
</>s0 Initial solid volume fraction in a set- 82-89 (Published 1978). (Available Ind. Sci. Con-
tling column ference Man., Inc., Chicago, IL)
(f)su Solid volume fraction at underflow 15. P. G. W. Hawksley, "Survey of the Relative Mo-
tion of Particles and Fluids," Br. J. App. Phys.
5:S1-S5 (1954).
16. E. S. Pettyjohn and E. B. Christiansen, "Effect of
Particle Shape on Free-Settling Rates of Isomet-
REFERENCES ric Particles," Chem. Eng. Prog. 44(2):151-112
(1948).
1. M. J. Pearse, "Gravity Thickening Theories: A 17. A. I. Medalia, "Dynamic Shape Factors of Parti-
Review," Warren Spring Laboratory, Stevenage, cles," Powder Technol. 4:117-138 (1970/71).
England, No. LR 261 (MP) (1977). 18. J. Tsubaki and G. Jimbo, "A Proposed New Char-
2. R. W. Moore, Jr., "Liquid-Solids Separation," acterization of Particle Shape and Its Application,"
Presented at a symposium organized by Hunslet Powder Technol. 22:161-178 (1979).
Taylor-Eimco Sales, Johannesburg (6 August 19. S. T. Fong, J. K. Beddow, and A. F. Vetter, "A
1971). Refined Method of Particle Shape Representa-
3. H. W. Campbell, R. J. Rush, and R. Tew, "Sludge tion," Powder Technol. 22:17-21 (1979).
Dewatering Design Manual," Res. Rep. No. 72. 20. J. M. Ziegler and B. Gills, Tables and Graphs for
Res. Program Abatement Munic. Pollut. Provi. the Settling Velocity of Quartz in Water, Above
Canada-Ontario Agreement Great Lakes Water the Range of Stokes' Law. Woods Hole Oceano-
Quality (1978). graphic Institution, Reference No. 59-36 (July,
4. K. J. Scott, Unpublished work of co-worker (when 1959).
mentioned). CERG-CSIR Internal Reports (1965- 21. R. H. Richards and C. E. Locke, Textbook of Ore
1970). Dressing, 3rd edit., McGraw-Hill, New York, Table
5. F. M. Tiller (ed), "How to Select Solid-Liquid 30, p. 129 (1940).
Separation Equipment," Chem. Eng. 81:(9) 22. A. W. Francis, "Wall Effect in Falling Ball Method
116-136 (April 29), (10) 98-104 (May 13, 1974). for Viscosity," Physics 4:403-406 (1933).
 SEDIMENTATION 679

23. J. Garside and M. R. Al-Dibouni, "Velocity Theories of Hindered Settling," Powder Technol.
Voidage Relationships for Fluidization and Sedi- 73:123-132 (1976); 27:147-152 (1977).
mentation in Solid-Liquid Systems," Ind. Eng. 39. G. B. Wallis, "A Simplified One-Dimensional
Chem. Proc. Des. Dev. 26:206-214 (1977). Representation of Two-Component Vertical Flow
24. B. H. Kaye and R. P. Boardman, "Cluster Forma- and Its Applications to Batch Sedimentation," in
tion in Dilute Suspensions," in Proc. Symp. Inter- Proc. Symp. Interaction Between Fluids and Parti-
action Between Fluids and Particles, Inst. Chem. cles, London, Institute of Chemical Engineers,
Engr. A17-21 (1962). pp. 9-16 (1963).
25. R. Johne, "Einfluss der Konzentration einer 40. L. A. Adorj'an, "Determination of Thickener Di-
monodispersen Suspension auf die Sinkge- mensions from Sediment Compression and Per-
schwindigkeit ihrer Teilchen," Chemie Ing Tech. meability Test Results, Trans. Inst. Min. Metal. C,
35:428-430 (1966). S5:C157-163 (1976).
26. B. Koglin, "Dynamic Equilibrium of Settling Ve- 41. C. E. Capes, "Particle Agglomeration and the
locity Distribution in Dilute Suspensions of Spher- Value of the Exponent n in the Richardson-Zaki
ical Irregularly Shaped Particles," in Proceedings Equation," Powder Technol. 20:303-306 (1974).
of the Conference on Particle Technology, IIT Re- 42. A. E. Fouda and C. E. Capes, "Hydrodynamic
search Institute of Chicago, pp. 266-271, August Particle Volume and Fluidized Bed Expansion,"
21-24 (1973). Can. J. Chem. Eng. 55:386-391 (1977).
27. E. M. Tory and D. K. Pickard, "A Three- 43. L.-G. Eklund and A. Jernqvist, "Experimental
Parameter Markov Model for Sedimentation," Study of the Dynamics of a Vertical Continuous
Can. J. Chem. Eng. 55:655-665 (1977). Thickener-I," Chem. Eng. Sci. 30:597-605 (1975).
28. J. F. Richardson and W. N. Zaki, "Sedimentation 44. A. S. Michaels and J. C. Bolger, "Settling Rates
and Fluidization: Part 1," Trans. Inst. Chem. Eng. and Sediment Volumes of Flocculated Kaolin Sus-
32:35-53 (1954).
pensions," Ind. Eng. Chem. Fundam. 2:2433 (1962).
29. H. Watanabe, "Voidage Function in Particulate
45. E. K. Obiakor and R. L. Whitmore, "Settling
Fluid Systems," Powder Technol. 18:211-225
Phenomena in Flocculated Suspensions," Rheol.
(1978).
Acta 6:353-359 (1967).
30. A. D. Maude and R. L. Whitmore, "A General-
46. P. G. Cooper, J. G. Rayner, and S. K. Nicol,
ized Theory of Sedimentation," Br. J. Appl. Phys.
"Flow Equation for Coagulated Suspensions,"
9:477-482 (1958).
/. Chem. Soc. Faraday Trans. I, 74:785-794 (1978).
31. W. C. Thacker and J. W. Lavelle, "Stability of
47. K. J. Scott, "Theory of Thickening: Factors Af-
Settling of Suspended Sediments," Phys. Fluids
fecting Settling Rate of Solids in Flocculated
27:291-292 (1978).
Pulps," Trans. Instn. Min. Metal 77:C85-97 (1968);
32. P. T. Shannon, E. Stroupe, and E. M. Tory, "Batch
75:C116-119, 244-245 (1969).
and Continuous Thickening. Basic Theory. Solids
48. K. J. Scott, "Thickening of Calcium Carbonate
Flux for Rigid Spheres," Ind. Eng. Chem. Fundam.
2:203-211 (1963). Slurries. Comparison of Data with Results for
Rigid Spheres," Ind. Eng. Chem. Fundam.
33. K. J. Scott and W. G. B. Mandersloot, "The Mean
7:484-490 (1968).
Particle Size in Hindered Settling of Multisized
Particles," Powder Technol 24:99-101 (1979). 49. K. J. Scott, "The Water Content of Floes," /. S.
34. B. Fitch, "Sedimentation Process Fundamentals," Afr. Inst. Min. Metal. 65:357-367 (1965).
Trans. AIME 223:129-137 (1962). 50. C. C. Harris, P. Somasundaran, and R. R. Jensen,
35. J. P. Mogan, R. W. Taylor, and F. L. Booth, "The "Sedimentation of Compressible Materials: Anal-
Value of the Exponent n in the Richardson and ysis of Batch Sedimentation Curve," Powder Tech-
Zaki Equation, for Fine Solids Fluidized with nol. 22:75-84 (1975).
Gases Under Pressure," Powder Technol. 4:286- 51. P. A. Vesilind, "Evaluation of Activated Sludge
289 (1970/71). Thickening Theories," /. San. Eng. Div. Proc.
36. D. K. Vohra, "Sedimentation in a Viscous Sus- ASCE 94:SA1, 185-191 (1968).
pending Medium," Inst. Eng. (India) I. Chem. Eng. 52. H. S. Coe and G. H. Clevenger, "Methods for
57:97-100 (1977). Determining the Capacities of Slime-Settling
37. A. K. Korol'kov, M. N. Kell, and A. A. Vasil'eva, Tanks," Trans. AIME 55:356-384 (1916).
"Dependence of the Rate of the Hindered 53. B. Fitch, "Batch Tests Predict Thickener Perfor-
Settling of Grains on Their Specific Surface," mance," Chem. Eng. 7S(19):83-88 (August 23,
Obogashch. Rud 25:24-26 (1970). (Russ.) [Chem. 1971).
Abstr. 74:89097 k (1971).] 54. B. Fitch, "Current Theory and Thickener Design,"
38. L. Davies, D. Dollimore, and J. H. Sharp, "Sedi- Ind. Eng. Chem. 5<5(10):18-28 (1966).
mentation of Suspensions: Implications of 55. R. H. Bretton, "The Design of Continuous Thick-
680 HANDBOOK OF POWDER SCIENCE

eners," Ph.D. Thesis, Yale University (August, Slurries," Ph.D. Thesis, Purdue University (June
1949). 1961).
56. J. A. Cole, "A Model for Activated Sludge Thick- 72. N. Yoshioka, Y. Hotta, S. Tanaka, S. Naito, and S.
ening," Ph.D. Thesis, University of Wisconsin Tsugami, "Continuous Thickening of Homoge-
(January, 1970). neous Flocculated Slurries," Chem. Eng. (Jpn)
57. K. J. Scott, "Continuous Thickening of Floccu- 27:66-674 (1957).
lated Suspensions," Ind. Eng. Chem. Fundam. 73. E. H. D. Carman and D. P. Steyn, "Some Obser-
9:422-427 (1970). vations on Thickening," in 8th Commonwealth
58. G. Sarmiento and P. H. T. Uhlherr, "The Effect Min. Metall. Congress Australia 6:443-454 (1965).
of Temperature on the Sedimentation Parameters 74. Anon., "Electrostatic Separation of Solids from
of Flocculated Suspensions," Powder Technol. Liquids," Filtration Separation 74:140-144
22:139-142 (1979). (March/April, 19-77).
59. R. H. Weiland and R. R. McPherson, "Accel- 75. Anon., "Rectangular versus Circular Settling
erated Settling by Addition of Buoyant Particles," Tanks," Am. City SS(10):98-99 (1973).
Ind. Eng. Chem. Fundam. 18:45-49 (1979). 76. J. D. Walker, "Sedimentation Maximizes Clarity,
60. R. L. Whitmore, "The Sedimentation of Suspen- Minimizes Turbidity in Potable Water Treatment,"
sions of Spheres," Br. J. Appl. Phys. 6:239-245 Water Sewage Works. Reference number
(1955). 1978:R136-150 (11 pp.) (1978).
77. J. H. Tay, "Study of Settling Characteristics of
61. P. Somasundaran, E. L. Smith, Jr., and C. C.
Physical-Chemical Floes in Sedimentation Tanks,"
Harris, "Effect of Coarser Particles on the Set-
Diss. Abstr. 38:1921 B (1978).
tling Characteristics of Phosphatic Slimes," in
Proceedings of the Conference on Particle Technol- 78. C. A. Lee, "Increasing Settling Tank Efficiency,"
ogy, IIT Research Institute of Chicago, pp. Plant. Eng. 126-127, April 19 (1973).
145-150, August 21-24 (1973). 79. C. E. Hubbell, "Hydraulic Characteristics of Vari-
ous Circular Settling Tanks," Am. Water Works
62. J. B. McVaugh, Jr., Mathematical and Experimen-
Assoc. J. 30:335-353 (1938).
tal Investigation of Nonsteady State Thickening of
80. Anon. "Sludge Settling Tank More Than Halves
an Ideal Slurry," M.Sc. Thesis, Delaware Univer-
Retention Time," Chem. Proc. 73(5):4-7 (May,
sity (May, 1975).
1967).
63. W. P. Talmage and E. B. Fitch, "Determining
81. C. G. Bruckmann, "Economic Aspects of Thick-
Thickener Unit Areas," Ind. Eng. Chem. 47:3841
ener Operations," Presented at Symposium S 19
(1955).
on Thickener Design and Operation organized by
64. K. J. Scott, "Mathematical Models of Mechanisms
the NCRL-CSIR, Pretoria (28 June, 1966).
of Thickening," Ind. Eng. Chem. Fundam. 5:109-
82. G. H. Matheson and J. N. W. MacKenzie, "Floe-
113 (1966).
culation and Thickening Coal-Washery Refuse
65. N. J. Hassett, "Mechanism of Thickening and Pulps," Coal Age 67:94-100 (December, 1962).
Thickener Design," Trans. Instn. Min. Metal. 83. D. G. Hall, "The Role of Bridging in Colloid
74:621-656 (1964/6S). Flocculation," Colloid Polymer Sci. 252:241-243
66. D. J. Slagle, Y. T. Shah, G. E. Klinzing, and J. G. (1974).
Walters, "Settling of Coal in Coal-Oil Slurries," 84. W. E. Walles, "Role of Flocculant Molecular
Ind. Eng. Chem. Proc. Des. Dev. 77:500-504 (1978). Weight in the Coagulation of Suspensions,"
67. T. M. Keinath, M. D. Ryckman, C. H. Dana, and /. Colloid Interface Sci. 27:797-803 (1968).
D. A. Hofer, "A Unified Approach to the Design 85. Committee Report, "State of the Art of Coagula-
and Operation of the Activated Sludge System," tion Mechanism and Stoichiometry," Am. Water
Presented at the 21st Annual Industrial Wastes Works Assoc. J. 63:99-108 (1971).
Conference, Purdue University (4 May, 1976). 86. S. L. Daniels, "A Survey of Flocculating Agents:
68. W. Rudolfs and I. O. Lacy, "Settling and Com- Process Descriptions and Design Considerations,"
pacting of Activated Sludge," Sewage Works J. AIChE Symp. Ser. No. 136, 70:266-281 (1973).
6:647-675 (1934). 87. F. M. Tiller, J. Wilensky, and P. J. Farrell, "Pre-
69. H. H. Steinour, "Rate of Sedimentation: Concen- treatment of Slurries," Chem. Eng. S7(9):123-126
tration Flocculated Suspensions of Powders," Ind. (April 29, 1974).
Eng. Chem. 36:901-907 (1944). 88. R. J. Akers, Flocculation. Publ. by I. Chem. E.
70. A. M. Gaudin and M. C. Fuerstenau, "On the Services for Inst. Chem. Eng., 15 Belgrave Square,
Mechanism of Thickening," in International Min- London, England SWIX 8PT (1975).
eral Processing Congress, Institute of Mining and 89. D. K. W. Smith and J. A. Kitchener, "The Strength
Metallurgy, London, pp. 115-127 (1960). of Aggregates Formed in Flocculation," Chem.
71. E. N. Tory, "Batch and Continuous Thickening of Eng. Sci. 33:1631-1636 (1978).
 SEDIMENTATION 681

90. R. N. Kovalcik, "Single Waste-Treatment Vessel sign of Underground Settlers," /. S. Afr. Inst. Min.
Both Flocculates and Clarifies," Chem. Eng. Metal. 67:501-521 (1961).
S5(14):117-120 (June 19, 1978). 106. A. W. Bond, "Upflow Solids Contact Basin,"
91. D. A. Dahlstrom and C. F. Cornell, "Thickening /. San. Eng. Diu. Proc. ASCE (SA6) 73-99 (1961).
and Clarification," Chem. Eng. Deskbook Issue 107. Private communication from T. M. Stielau, Delkor
7S(4):63-69 (February 15, 1971). "Sedimentation Technik, Randburg, S.A.
Systems from Laboratory Data," Chem. Eng. 108. Anon., "Revolutionary Thickener Design Tackles
6S(19):167-170 (September 18, 1961). nearly Flow of Zinc Mine Tailings," Eng. Min. J.
92. R. A. Conway and V. H. Edwards, "How to De- 179:18-19 (April, 1978).
sign Sedimentation Systems from Laboratory 109. G. R. Mace and R. Laks, "Developments in Grav-
Data," Chem. Eng. 6£(19):161-170 (September 18, ity Sedimentation," Chem. Eng. Prog. 74(l):ll-83
1961). (1978).
93. H. C. Bramer and R. D. Hoak, "Design Criteria 110. J. R. Bratby, "Aspects of Sludge Thickening by
for Sedimentation Basins," / EC Proc. Des. Deu. Dissolved-Air Flotation," Water Pollut. Contr.
2:185-189 (1962). (Lond.) 77:421-432 (1978).
94. E. B. Fitch, "The Significance of Detention in
111. M. T. Turner, "Use of Dissolved Air Flotation for
Sedimentation," Sewage Ind. Wastes 29:1123-1133 the Thickening of Waste Activated Sludge," Ef-
(1957). fluent Water Treat. J. 75:243-251 (7 p.) (May, 1975).
95. W. H. Mitchell, "The Preparation and Treatment
112. E. R. Ramirez, "Dewatering Skimmings and
of Mine Water Prior to Pumping and Notes on
Sludges with a Lectro-Thic Unit," Water and
Maintenance of Pumping Plant," Inst. Certif. Eng.
Wastewater Equip. Manuf. Assoc. Pollut. Control
(S. Afr.) J. 26:86-99 (1953).
Conf., 3rd Annual Ind. Solutions '75: Air-Water-
96. R. T. Hukki, G. Diehl, and P. Vanninen, "Princi- Noise-Solid Waste, Proc, pp. 467-78, ANN, April
ples of Construction and Operation of the Chan- 1-4, 1975.
nel and Syphon Thickener," in 7th Int. Mineral
113. V. Gulas and R. Lindsey, "Factors Affecting the
Processing Congress, Vol 1, Gordon and Breach,
Design of Dissolved Air Flotation Systems."
New York, pp. 115-123 (1965).
/. Water Pollut. Control Fed.
97. R. F. Probstein and R. E. Hicks, "Lamella Set-
114. H. E. Cross, "A New Approach to the Design and
tlers: A New Operating Mode for High Perfor-
Operation of Thickeners," /. S. Afri. Inst. Min.
mance," Ind. Water Eng. 15:6-8 (1978).
Metall. 63:271-289 (1963).
98. R. L. Cook, "Compact Lamella Thickeners in Coal
Preparation Plants," in Second Symposium on Coal 115. E. B. Fitch and W. A. Lutz, "Feedwells for Den-
Prep. (Washington, DC), at NCA/BCR (Natl. Coal sity Stabilization," /. Water Pollut. Control Fed.
Assoc./Bitum. Coal Res.) Coal Conf., and Expo. 32:141-156 (1960).
3, Louisville, KY (October 19-21, 1976). (Availa- 116. Anon., "Improved Thickener Rake," S. Afr. Min.
ble from NCA.) Eng. J., p. 415 (August 23, 1968).
99. R. L. Cook and J. J. Childless, "Performance of 117. F. R. Weber, "How to Select the Right Thickener,"
Lamella Thickeners in Coal Preparation Plants," Coal Min. Proc. 14(5):98-100, 104, 116 (1977).
Min. Eng. 30:566-511 (1978). 118. V. S. Dillon, "Special Features of the Kinross
100. L. C. Meitzler and G. H. Weyermuller, "Compact- Mines, Ltd., Reduction Plant," in Proc. 9th Com-
ness of Thickener Permits Installation in Limited monwealth Mining Metall. Congress, London, Miner.
Space," Chem. Proc. 37(5):18 (1974). Proc. Extr. Metal. 3:485-508 (1969).
101. F. C. McMicheal, "Sedimentation in Inclined 119. R. P. Plasket and D. A. Ireland, "Ancillary Smelter
Tubes and Its Application for the Design of High Operations and Sulphuric Acid Manufacture at
Rate Sedimentation Devices," /. Hydraul. Res. Impala Platinum Limited," /. S. Afr. Inst. Min.
10:59-15 (1972). Metal, pp. 1-10 (August, 1976).
102. D. Davis, S. Rogel, and K. Robe, "Increases Clari- 120. F. Concha and E. R. Almendra, "Settling Veloci-
fier Capacity 85% at l/20th Cost of New Unit," ties of Particulate Systems. 1. Settling Velocities
Chem. Process (U.S.) 36:10 (January, 1973). of Individual Spherical Particles," Int. J. Miner.
103. R. M. Willis, "Tubular Settlers-A Technical Re- Proc. 5:349-367 (1979).
view," /. Am. Water Works Assoc. 70:331-335 121. K. J. Scott, "Experimental Study of Continuous
(1978). Thickening of a Flocculated Silica Slurry," Ind.
104. Axel Johnson Institute, "A New Separation Tech- Eng. Chem. Fundam. 7:582-595 (1968).
nique," Trade Brochure Box 13, Nynashamn, 122. L.-G. Ecklund, "Influence of Feed Conditions
Sweden. on Continuous Thickening," Chem. Eng. Sci.
105. J. W. de Villiers, "An Investigation into the De- 34:1063-1066 (1979).
682 HANDBOOK OF POWDER SCIENCE

123. J. H. Masliyah, "Hindered Settling in a Multi- 129. D. C. Dixon, "Effect of Sludge Funneling in Gray.
species Particle System," Chem. Eng. Sci. ity Thickeners," AIChE J. 26:471-471 (1980).
3*1166-1168 (1979). 130. W. Chen, "A Study of the Mechanism of Sedi-
124. F. M. Tiller, "Revision of Kynch Sedimentation mentation," M.S. Thesis, University of Houston
Theory," AIChE J. 27:823-839 (1981). (1984).
125. B. Fitch, "Kynch Theory and Compression Zones," 131. R. D. Paradis, "Application of Alcan's Deep
AIChE. J. 29:940-947 (1983). Thickener Technology for Thickening and Clari-
126. F. M. Tiller and W. Chen, "Limiting Operating fication," in American Filtration & Separation
Conditions for Continuous Thickeners," Chem. Society Annual Meeting, Chicago (1993).
Eng. Sci. 43(7):1695-1704 (1988). 132. G. J. Kynch, "A Theory of Sedimentation," Trans.
127. J. L. Chandler, "Design of Deep Thickeners," Faraday Soc. 48:166-176 (1952).
Preprint, Institute of Chemical Engineers Sympo- 133. L. E. Kun, R. O. Oelofsen, and E. J. J. Van
sium, Rugby, U.K. (1976). Veuren, "Hopper Clarification of Gold Pregnant
128. P. Kos, "Fundamentals of Gravity Thickening," Solution at Vaal Reefs South," /. S. Afr. Inst. Min.
Chem. Eng. Prog. 73:99-105 (1977). Metall 75:201-206 (1979).
 14
Filtration of Solids from
Liquid Streams
Larry Avery

CONTENTS

14.1 INTRODUCTION 683

14.2 PHYSICAL MECHANISMS OF FILTRATION 685
14.3 FILTRATION THEORY 686
14.4 FILTER MEDIA 688
14.5 MEMBRANES 690
14.6 FILTER AIDS 695
14.7 STAGES OF THE FILTER CYCLE 696
14.8 LITERATURE AND INFORMATION REVIEW 698
14.9 TYPES AND DESCRIPTION OF LIQUID FILTER EQUIPMENT 701
4.10 CENTRIFUGES 719
4.11 FILTER EQUIPMENT SELECTION 723
REFERENCES 723

14.1 INTRODUCTION quire clean water, sewage treatment, waste

reduction, and sludge dewatering before dis-
In the modern filtration world, there are few posal, all of which are controlled by various
industries that do not depend at least in part liquid-solid filtration separation process
on the liquid filtration process. Current con- operations.
sumers demand products that are clear, free The diversity and rather little known ubiq-
from contamination, attractive, and healthful. uity of the many common products made partly
Filtration helps to accomplish these objectives. by filtration operations was the subject of a
Industry likewise demands commercial most interesting article entitled "Do We Need
chemical and pharmaceutical products that are Filtration?" by Carl Jahreis, a research filtra-
pure, meet stringent standards, and are of tion engineer of the Shriver Co. He described
high quality. Pollution control regulations re- more than 100 common products that one
683
684 HANDBOOK OF POWDER SCIENCE

encounters daily that would not exist if it were 14.1.3 Clarification

not for liquid filtration operations.1
Clarification involves hundreds of liquid food
In light of competitive market conditions,
products, juices, household products, water,
increased energy cost, and government regula-
coolant liquids, chemicals, hydraulic oils, gaso-
tions, the continuing profitable manufacture of
line, and even molten metals. Generally, the
products requires a knowledge of how to filter
liquid has a very low percentage of solids,
particles suspended in the liquid phase. Also
usually 0.1% or less. Typical media are in the
needed is a knowledge of how best to select a shape of cartridges, bags, sheets, pads, and
suitable filtering device, how to program and nonwovens. Operation is by pressure, and it is
operate it efficiently, and, finally, how to per- usually a small-scale batch process.
form needed postfiltration steps to provide the
final quality product.
It is the aim of this chapter to explore this 14.1.4 Cake Filtration
complex subject of particle removal from liq- Cake filtration provides pigments, dyestuffs,
uid streams, and to make understandable the minerals, chemicals, food such as yeast, phar-
mechanics involved in the practical modern maceuticals, and catalyst. Filtration is a most
filter devices used to accomplish this goal. important process in pollution control for de-
watering waste water streams. It provides an
14.1.1 Definition economical way to solidify waste for regulated
landfill and hazardous waste disposal facilities.
Liquid filtration is a two-phase physical sepa- Cake filtration may use pressure, vacuum, or
ration of particulate solid matter from the centrifugal force. Equipment used includes
liquid in which it is suspended. The means is a filter presses, rotary vacuum filters, and cen-
filter medium properly selected to retain the trifuges. Large-scale chemical, mineral, and
solid particles. The driving force may be grav- waste sludges are usually cake filtrations.
ity, vacuum, pressure, or centrifugal force.
14.1.5 Concept
14.1.2 Purposes The concept of filtration seems simple as
The general purposes of filtration can be shown in Figure 14.1. Ideally, all of the parti-
broadly stated according to a frequently used cles would be removed and the filtered ef-
fluent, or filtrate, would be perfectly clear. In
classification as follows:
practice, this never happens because the filter
medium always permits some particles to pass
1. To clarify, purify, or sterilize a valuable through. Also, some liquid is always retained
liquid end product where the contaminants in the collected cake, which is the greater
are discarded usually with the filter medium problem of the two.
2. To collect a solid valuable product with the
liquid discarded, reused, or recycled
3. To save both phases as useful products SLURRY

4. To dispose of as waste both the solids fil-

tered out as dewatered sludge, and also the
liquid portion as a waste material.

From the above it is apparent that there are

two major divisions of filtration: clarification FILTRATE FLOW

and solids or cake filtration. Figure 14.1. Basic filtration device.

 FILTRATION OF SOLIDS FROM LIQUID STREAMS 685

14.1.6 The Four Basic Components the mechanism of the particle collection as
regards the specific media. In the first case, as
Important considerations of the four basic
components are given in the following sub- shown in Figure 14.2, the particles are re-
sections. tained from the suspension exactly on the face
of the medium as the particles approach the
Liquid. The liquid contains the suspended medium at right angles. The principle is that
particles, and is called the feed, suspension, the pores or openings in that medium are
or slurry. The types of particles—the size char- smaller than the particles contacting it, thus
acteristics, the density, the settling rate, preventing them from passing through. The
the shape, softness, quantity, and chemical medium must be physically and mechanically
nature—and the viscosity of the liquid deter- strong enough to resist any pressure deforma-
mine the filterability of the feed or slurry. tion preventing the pores from enlarging. The
particles must be rigid or firm enough so as
Medium. The medium is the porous material not to compress or squeeze through the open-
for collection of the particles. It determines ings. If all these conditions are met, we would
the efficiency of the filtration, the mechanism have complete or absolute retention of the
involved, and the suitable operation of the
particles. With metal screens, perforated met-
filter itself.
als, porous ceramics, and some membrane fil-
Solids. Mostly, it is desirable for the solids ters, this condition can approach reality. In
to have as low a residual moisture content as many cases it does not have to be perfect
possible, and to be free from the mother liquor; because the filtrate can be recirculated and
hence the need to wash the filter cake. trapped by the solids already built up on the
medium. This cannot be done, however, in
Filtrate, or Effluent. A high degree of clarity critical microbe filtrations of pharmaceutical
or purity is required for liquid products. products where a single pass must be complete
Wastewater streams should be low in TSS to what is called a log-reduction value of 7,
(total suspended solids). which indicates that there were 10 microbes
Filters can operate in either a batch or a found in the filtrate for a filtration efficiency
continuous mode. Most batch filters operate of 0.9999999 (seven nines).
on a small scale. However, some batch filter
This same surface filter medium is also the
presses, for example, can handle solids loads
desired type for cake filtrations, where the
up to 300 ft3 per batch or about 9 tons de-
solids built up to 1-in. thickness or more based
pending on the density of the wet cake. Con-
tinuous belt or rotary vacuum filters can pro- on the type of filter. Media with a smooth slick
cess up to 120 tons per day, again depending surface and a pore size in the 1 to 10 micron
on the percent solids in the feed stream, and
the density of the solids. Of course, batch
filters can be operated in multiple parallel SUSPENDED SOLIDS
stages to produce essentially a continuous
production output. I i 1 1 1 I i

14.2 PHYSICAL MECHANISMS OF

FILTRATION FILTER_
MEDIUM
14.2.1 Surface Filtration
The two almost classic concepts of liquid fil-
tration are called surface filtration and depth CLEAR FILTRATE

filtration. They refer to the filter medium and Figure 14.2. Surface filtration sketch.
686 HANDBOOK OF POWDER SCIENCE

range will accomplish this. Over time, there the longer the cartridge life and lower related
will be some penetration of the medium known filtration costs.
as progressive plugging, but yet these filter
cloths can last for hundreds of filtration cycles 14.2.3 Pore Blocking
before they have to be replaced.
Another mechanism closely related to the
above is pore blocking and particle bridging of
14.2.2 Depth Filtration the pores. The first is undesirable because it
stops the flow. It occurs most often with rela-
The other basic separation mechanism is tively small particles, high viscosity, and low
known as depth filtration, and as the name solids concentration.
implies, the solids are captured under the sur- Particle bridging results from particles col-
face, and within the depth of the filter medium. lecting around the pore openings, and gradu-
This concept is shown in Figure 14.3. This can ally closing over the opening. An increase in
apply to membrane media that may be only 50 the suspension's particle concentration favors
microns thick or fabrics and filter sheets that this mechanism. Once this occurs, cake filtra-
can be as much as 0.125-in. thick. This is not tion can take place.
to say that very large particles may all collect
on the surface so that it performs like a sur-
face medium. This can happen with string 14.3 FILTRATION THEORY
wound filter cartridges, for example, which are
layered to provide a porosity gradient from the Filtration has long been considered more of a
outside feed side to the internal core for practical art still being developed than as an
filtrate discharge. engineering science. Likewise, the theory of
The advantage of depth filters is that they filtration operations has itself been the contin-
can trap particles smaller than the average uing subject of much study in the academic
pore size in the medium. This is done by field. Many of the basic approaches for the
electrostatic forces, molecular forces, direct last 75 years have been most important in
impingement on fibers, and attachment to the developing fundamental theoretical relation-
sidewalls of the interstices within the medium. ships. The real beginning was the work of
This entrapment of particles within the depth Darcy on capillary and pressure relationships
leads to an important property of filter car- in 1856.
tridges called dirt-load capacity. Even though His work was recently translated not with-
cartridges are used for feed streams of under out some difficulty from the French to English
0.1% solids, the higher the dirt-load capacity, by J. B. Crump and critiqued by Tiller as
related to our current theory.43
The equations expressing relationships be-
tween filtration variables have been applied to
SKETCH FOR DEPTH FILTRATION certain designs of equipment, but mostly they
SLURRY are helpful in interpreting pilot and laboratory
tests and determining the specific cake resis-
tance which is unique to each slurry. This
specific cake resistance is affected by the basic
factors plus the porosity and the specific sur-
face of the particles in the suspension to be
"I "I T'J 1" 1" 1 filtered.
CLEAR FILTRATE The fundamental theory begins with the
Figure 14.3. Depth filtration sketch. basic Darcy equation relating the flow rate Q
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 687

of viscosity /x through a bed of thickness L may increase where cake builds on tubular or
and area A and driving force p: rotary drum surfaces. The viscosity stays con-
AAp stant if the temperature is likewise constant
Q =K (14.1) and the liquid is Newtonian.
The specific cake resistance a should be
where K is a constant referred to as the constant for incompressible cakes, but could
permeability of the filter bed. This equation is vary slightly because of possible cake consoli-
often written in the form dation or feed approach velocity. However,
most cakes are compressible, so the specific
(14.2) cake resistance changes with Apc. Then the
Q=
average specific cake resistance aav should
replace a in Eq. (14.5). It can be determined
where R is called the medium resistance and
by
is equal to L/K.
If the suspension were a clean liquid, the APc d(Apc)
parameters in Eq. (14.1) would be constant, (14.6)
Apc
and the relationship between the flow and the
pressure drop would produce a cumulative fil- if the function a = f(Apc) is known from test
trate volume that would increase linearly with data. If not, an experimental empirical rela-
time. When the suspension contains particles, tionship can be used over a limited pressure
the resultant cake formation takes up more range:
pressure so the flow decreases with time.
With cake forming, there are two resis- a = ao(Apc)n (14.7)
tances to flow, the cake and the filter medium
as per the following equation: where a0 is the resistance at unit applied Ap
and n is a compressibility index (equal to zero
AAp for incompressible substance).
Q= (14.3)
Rc) Using Eq. (14.7), the average cake resist-
ance aav can be shown to be:
This assumes the filter medium resistance to
be constant, which in practice is not always = (l -n)ao(Apc)n (14.8)
precisely true because of particle impingement
on the medium surface, and also progressive The mass of cake deposited per unit area w is
plugging of the media. Assuming the cake (if a function of time in batch filtrations, and it
incompressible) is proportional to the amount can be related to the cumulative volume V in
of cake deposited, it follows that time t by
Rc = aw (14.4) wA = cV (14.9)
where w is the mass of cake deposited per unit where c is the concentration of solids in the
area and a is the specific cake resistance. suspension.
Substituting Eq. (14.4) in (14.3) gives From the above initial analysis, basic equa-
ApA tions for filtration operations for incompress-
Q = — ^ (14.5) ible cakes using constant pressure and con-
stant rate filtrations have been developed.
This relates the flow rate Q to the pressure From pilot tests, the specific cake resistance
drop Ap, the mass of cake deposited w, and can be determined. Likewise, equations for
other parameters, some of which can be as- compressible cake filtrations and relationships
sumed to be constant. However, Ap may be between the specific cake resistance and
constant or variable with time. The face area porosity and specific surface of the particles
688 HANDBOOK OF POWDER SCIENCE

have been made known. These are expressed 8. Should not have loose fibers that shed into
as the classic Kozeny-Carman equation. the cake or liquid being filtered.
The above is only a very basic outline of 9. Should be capable of being fabricated,
simple theory as based on an excellent presen- sewed, fused, or adaptable to other types
tation by L. Svarvosky on filtration fundamen- of converting operations.
tals in his recent book. All of the basic equa- 10. Should have an economic service life.
tions mentioned above are included in detail
in Ref. 44. Not all the above will be found in a single
As research workers explore troublesome medium so that certain compromises will have
assumptions in the classic theory, new con- to be made regarding cost, medium life, and
siderations are presented. Work by Tiller, performance.
Wakeman, Rushton, Willis, and others is
adding to the field. Studies on formulas for 14.4.1 Types of Filter Media
constant pressure filtration and compaction of
filter cakes were presented at the recent The most common types of media are woven
American Filtration meeting in Hershey, PA.4 fabrics, papers, and felts. Yarn types for woven
media are shown in Fig. 14.4. Physical and
chemical characteristics of the most frequently
14.4 FILTER MEDIA used fibers are shown in Table 14.1. In recent
years, there has been an increasing interest in
nonwoven textiles and also membranes, lami-
Filter media are available in many different
nates, finemesh woven metal wire, and
forms, and being the essential element of a
photoetched metals.
filter, they should have as many of the follow-
ing characteristics as possible. These pertain Also considered as media are screens, wedge
mostly to woven fabrics, but can apply to some wire, see Figure 14.5 grids, sand, perforated
nonwovens such as felt as well. They: steel plates, porous ceramic, see Fig. 14.6 plas-
tic, and carbon sheets and tubes. Thus, it can
be seen that some media are flexible, some
1. Should have particle retention suitable for
rigid, and some even granular. Pore size and
the application, generally no more that
porosity can vary considerably. Selecting the
actually required because of increased
right filter fabric was covered by Clark.45
cost.
2. Should have low flow resistance.
3. Should be resistant to chemical degra-
dation and any subsequent cleaning
chemicals.
4. Should have enough physical strength to
adapt to the type of filter equipment used
and avoid problems from creasing.
5. Should not change form, stretch, or shrink
during filtration or be susceptible to bacte-
rial growth.
6. Should offer resistance to the maximum
temperature of liquid to be filtered or
subsequent washing or steam cleaning of
media.
7. For cake filtration, should have a smooth
and slick surface to facilitate unaided re- Figure 14.4. Yarn types. (Zurich Bolting Cloth Mfg.
lease of the filter cake. Co. Ltd.)
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 689

Table 14.1. Typical Characteristics for Common Fibers Used for Filter Cloths for Liquid Filtration.
Temperatures are Approximate. Resistances Depend on Strength and Temperature of Acid or Alkali.

MAXIMUM
OPERATING
TEMPERATURE ACID ALKALI WET HEAT FLEX AND
FIBER °F °C RESISTANCE RESISTANCE RESISTANCE ABRASION
Acrylic 275 135 Excellent Fair Good Good
Aramid 400 205 Fair Good Excellent Very Good
Cotton 210 99 Poor Good Fair Good
Nylon 250 121 Fair Very Good Good Excellent
Polyester 284 140 Good Fair Good Very Good
Polypropylene 200 94 Excellent Excellent Fair Very Good

Ceramic media for severe corrosive environ-

ments are discussed by Sheppard.46

14.4.2 Selection of Filter Media

Generally the type of filter equipment selected
will determine the appropriate filter medium
based on years of experience by the filter
manufacturers and users. For example, filter
presses use filter clothes that are mostly syn-
thetics such as polypropylene and polyester;
pressure leaf filters use fine mesh stainless
Figure 14.5. Wedgewire media. Wound to form tubular steel woven wire cloth; rotary vacuum filters
filter element. (Johnson Screens) use lighter weight, 5 to 6 oz/yd, 2 more open

60|im
thickness

Pore sizes are

distances between
non-porous
alumina
crystals

SEM image shows cross-section of support and membrane layers that make up a 0.2 micron PI9-40 element.

Figure 14.6. Porous ceramic. U.S. Filter Membralox® (Trade Mark)

690 HANDBOOK OF POWDER SCIENCE

cloths than filter presses; and belt filters use other major use is on continuous horizontal
heavy-duty, 22 to 25 oz/yd 2 rugged synthetic plate filters for coolants in the metal roll-
woven cloths. Many clarifying filter presses use ing industry and D & I can manufacturing
filter paper and sheets. Cartridges and filter operations.
bags are also used widely for clarifying filtra- In a study comparing pleated nonwoven me-
tions. Once a medium has been established in dia in a filter cartridge against a conventional
practice, and there is a change in the process, wound cartridge design, it was found that the
or if a problem develops, such as insufficient nonwoven media had a greater efficiency in
particle retention, improvements can be made particle removal than the wound yarn design.65
by gradually selecting a similar filter medium The media used was a polyester material,
for test on a pilot scale. Then the new medium Other materials are the cellulose, rayon, and
can be tried on plant equipment, usually with nylon used in early nonwovens. More recently,
good results. aramids such as Nomex and Kevlar and
Where an untried application develops, the fluorocarbons such as Teflon are being used.
selection has to be made with a more critical
look at all the desired characteristics. Here,
lab tests will be required to determine a final 14.5 MEMBRANES
choice.
Filtration and separation media characteris- A most important field of liquid filtration is
tics along with advantages and typical uses are the one that benefits from membrane filters.
shown in Table 14.2.47 These are very thin microporous polymeric
film sheet media from 10 to 100 microns thick.
The range of separations is shown in Figure
14.4.3 Nonwoven Media
14.7. The four basic types of membrane
A newer type of media showing increasing use processes are:
is the nonwoven or bonded material. It has a
web structure of entangled fibers made by a 1. Reverse osmosis (RO), with an osmotic
mechanical, thermal, or chemical bonding pro- pressure driving force separating a solvent,
cess. The filtering properties of these media usually water, from a dissolved monovalent
are controlled, such as strength and uniformity salt
of fiber orientation. A recent article explains 2. Nanofiltration (NF), which rejects divalent
the advantages and applications of the four salt, sugars, and disassociated acids
basic technologies for bonding nonwoven webs, 3. Ultrafiltration (UF), which separates or
which are chemical, ultrasonic, needle punch- fractionates dissolved molecules by molecu-
ing, and adhesive melt.48 The various types are lar weight and size
designated by the method of formation such as 4. Microfiltration (MF), which is actually
card webs, air laid, wet formed, spunbonded, particle removal of very fine or colloidal
and melt blown. This nonwoven technology is particles.
explained by Shoemaker.49 In addition, bond-
ing mechanisms are given by Pangrazi.48 There is some overlapping of the separation
Major uses for nonwovens are for mem- range, and since we are concerned only with
brane supports and cartridge filters, especially particle filtration, we will discuss MF and the
for swimming pool water and other liquid fil- related range of UF. From their limited initial
trations. Advantages are pleatability, resis- use 50 years ago in removing microorganisms
tance to damage, good retention values, and in drinking water, they have had rapid growth
flow rates. In roll form from 18 in. to 45 in. to sales of over $900 million annually. They
widths, they are widely used in filtering ma- meet critical applications in gas, liquid, and
chine tool coolants in deep bed filters. An- solvent separations. Major uses are in desali-
Table 14.2. Filtration and Separation Media Characteristics.

RELATIVE GENERAL RELATIVE CURRENT

MEDIA EFFICIENCY MEDIA MARKET
COST RATINGS PERMEABILITY PENETRATION
TYPE RANGE RANGE RANGE TREND AND USE ADVANTAGE DISADVANTAGE COMMON USE
Ceramic 8-10 5-10 1-4 Chemical compatibility Expensive and brittle. Biotechnology-
and high temperature pharmaceutical,
capabilities. reusable.

Filter aid 1-2 2-7* 8-10 Same Inexpensive, excellent filter Disposal, mostly limited Precoat for large volume
cake base. to pressure filtration. pressure filtration e.g
leaf pressure filters.

Glass 1-2 2-8 4-8 Same to — High temperature, chemical Limited media processing Baghouse filtration,
compatibility, low stretch, capabilities, yarn laboratory filters,
low cost. weakness. HEPA filters.

Membranes 6-9 9-10 1-2 Narrow pore size High cost, low flow rates. Pharmaceuticals,
distribution below Somewhat hard to semiconductors, medical
one micron, many process. devices, ultrapure
polymer choices. water.

Metal 3-8 4-9 3-10 Same to - Reusable, high temperature, Expensive, high cleaning Vibratory sifting, aerospace,
diverse properties, costs. polymer filters, reusable
narrow pore applications.
size distribution.
Table 14.2. Filtration and Separation Media Characteristics. Continued
RELATIVE GENERAL RELATIVE CURRENT
MEDIA EFFICIENCY MEDIA MARKET
COST RATINGS PERMEABILITY PENETRATION
TYPE RANGE RANGE RANGE TREND AND USE ADVANTAGE DISADVANTAGE COMMON USE

Nonwoven 1-3 1-8* 4-8 ++ Low cost, dirt holding Random pores, particle Chemical, medical, water,
fabrics capability, diverse unloading, fiber baghouse filters,
constructions. migration strainer bags.
disposal of media.

Paper 1-2 1-6 4-8 Dirt holding capabilities, Fiber release possible, Automotive, laboratory,
diverse polymers, poor wet strength, air, and general
moldable. particle unloading, process industries.
disposal of media.

Porous 3-7 4-8 2-5 Same Dirt holding capabilities, Restricted to rigid Medical, battery vents,
plastics diverse polymers, forms, limited uses. water.
moldable.

Precision 4-7 7-9 4-10 High flow with minimal Especially expensive at Dewatering, medical,
woven resistance, precision lower size pore ratings aerospace, automotive,
synthetic pores, wide choice. (5-30 micron). process filtration
screen including belts.
fabrics

Woven 3-5 2-6 2-7 Same High wet/dry strength. Lower flow rates, Filter presses, RO
fabrics Lower cost. Dirt random size pores, channel separators,
holding capabilities. particle unloading. vacuum belts
Wide choice.

* High range includes special conditions/processing and circumstances.

Range Ratings: 1—Represents lower cost or performance. 10—Represents higher cost or performance.
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 693

RANGE OF SEPARATION BY MEMBRANE PROCESSES

REJ ECTS SUSPENDEI) PARTICLES

REJECTS SMALL ;R PARTICLES

REJECTS MACR< MOLECULES

REJECTS DIVALE vlT SALTS, SUGAF S, DISSOCIATED j tCIDS

REJEC TS MONOVALENT SALTS AND UNO SSOCJATED ACID 5

lA lOA 100A lOOOA 1 um 10 jam 100

PORE SIZE IN ANGSTROM UNITS OR MICROMETERS
Figure 14.7. Range of separation by membrane processes.

nation, fluid sterilization, and waste water ceramic application is for large discs in a
treatment, such as separating emulsified oily rotary vacuum disc filter used in the min-
wastes. Microfiltration uses include removal of ing industry. These filters can have several
suspended particles from effluent waters, clari- hundred square feet of filter area.
fication of fruit juices and vinegars, and har- Membrane filters can be designed with flat
vesting bacterial cells. sheets in a plate and frame support. Car-
Cellulose esters were first used for making tridges are made in a spiral wound or pleated
microfiltration membranes, but now various membrane configuration. Hollow fiber mem-
polymers including nylon, polyvinyl chloride branes are in a tubular design.
(PVC), polypropylene, polysulfones, and poly- Flat sheet media or pleated cartridges have
tetrafluorethylene (PTFE) are used. The latter a special application in the pharmaceutical
can be made by stretching a thin sheet of the industry for the purpose of sterilization of
polymer carefully and bonding it to a porous certain liquid batch products. The membrane
substrate.50 See Fig. 14.8. The membrane film media and its holder or housing must be steril-
can be from 100 to 250 microns thick. Other ized, and tested for integrity. Since it is critical
membrane preparation methods are track- that no microbes or contaminants pass through
etching and phase inversion casting. Details of the filter assembly, the pores in the membrane
these processes are given by Porter.51 must not be larger than the microbes or parti-
Most of these membranes have physical and cles to be retained. To verify this, a bubble
temperature limitations and may be subject to point test must be done. The apparatus for
chemical and solvent attack. Recently ceramic performing this test is described by ASTM
membranes have become commercially avail- F136. The factors involved are absolute filtra-
able. Originally developed by the French nu- tion, average size pore, and filtration effi-
clear industry, and now declassified, they are ciency. The variables and interpretation of this
being used as tubular membrane filters with subject are discussed in detail by Johnston.52
the membrane surface on the inside. Reten- One aspect of membrane filtration, and dif-
tion values down to less than 0.2 micron are ferent from sterilizing nitrations, is that the
available as shown in Figure 14.6. Another flow patterns are not at right angles to the
694 HANDBOOK OF POWDER SCIENCE

1000X
Figure 14.8. Illustration of stretched polymer membrane media at 1000 X magnification. Courtesy
W. G. Gore and Associates.

membrane, see Fig. 14.9, but tangent to the membrane filtration, it is essential that the
media, which is called crossflow filtration. (See crossflow be of sufficient velocity to offset a
Fig. 14.10.) This concept has been used in phenomenon known as "concentration polar-
conventional cake filtration as a means to limit ization" in which the solute builds up on the
cake growth and increase output. However, in surface of the membrane in concentrations

CONVENTIONAL FILTER

Feed Filtrate Flow

Collected Solids or Cake Filter Media

DIRECT-FLOW (PERPENDICULAR TO MEDIA)

Figure 14.9. Direct flow to media.
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 695

MEMBRANE FILTER
Concentrate
Feed or
Retentate

mbrane
y y y

Permeate Flow

CROSS-FLOW (TANGENTIAL TO MEMBRANE)

Figure 14.10. Cross flow filtration.

much higher than in the bulk flow of the feed cle sizes determines the grade and the practi-
stream. There are ways of overcoming this cal application. Flow rates of different grades
problem, and, among others, Van den Berg are shown in Figure 14.11. Diatomite filtration
and Smolden developed mathematical models systems can remove particles under 1 micron
to study it. They concluded that besides cross- and at flow rates of from 0.2 to 2.0 gal/min/ft 2
flow filtration, reducing scaling of membranes, on rotary vacuum filters.56 This type of use is
chemical treatment of the membrane surface, called precoat filtration; a 5- or 6-in. layer of
using corrugated membranes, and using ap- filter aid is formed on the filter drum, and is
propriate pretreatment methods to increase gradually scraped off with a sharp knife edge
the mass transfer coefficient are helpful.53 along with a thin layer of the filtered solids.
In the field of biological membrane separa- Precoat can also be used on sheet media and
tions, Gyure discusses in qualitative terms the filter cloths as a porous layer, and also serves
many practical considerations in using cross- to facilitate cake removal from the medium.
flow filtration. Continuous versus batch sys- Another common method of using filter aids,
tems are compared and methods for effective called admix or body-feed, is to add them to
cleaning of membranes are given.54 the suspension being filtered. The amount and
grade used can be determined empirically, but
generally it must be equal to or more than the
14.6 FILTER AIDS weight of the suspended solids, and it can
exceed this by up to 10%. If this optimum
Filter aids are loose powders, such as diatoma- amount is not maintained, it is apparent that
ceous earth (DE) and expanded perlite, that the formed filter cake will plug to end the
are used to facilitate and improve the filtration cycle.
of difficult to filter products, such as gels, Although filter aids are inert, and up to
hydroxides, and very fine particles. Their rigid- 95% silica, they do have impurities such as
ity and high porosity make them suitable for iron, copper, etc. and the possible effect on
this purpose. They are added to the slurry, the filtrate should be considered in their selec-
thus forming a more permeable filter cake. tion. Also, the amount of filter cake produced
Occurring as natural minerals, they are pro- is greater, and this could add to disposal costs.
cessed into about 10 different grades with The permeability of diatomite filter aids is
ranges of particle sizes from 40 microns down specified in Darcies, which is defined at unity
to under 2 microns. The distribution of parti- if a liquid material passes 1 cc/cm 2 per sec-
696 HANDBOOK OF POWDER SCIENCE

/
Levels Off At 280 Gallons

>t Celite 560* - —

cel ite 5 45

/
r
y
y r
v

/I • H •OB

ss
««• Jelit }53 5 -

r Celite 503
SSI

/ **>
***

. Celite 501

t
mm • i

mm » I-Hyflo buper-Cel
***
• — mm • Celite 512
i I Stand ard 3r-C
Iter Cel /Celite 500
- i i i i
No Filter Aid
0 1 2 3 4 5 6 7 10
Hours Elapsed Time

Figure 14.11. Flow rate properties of filter aids.

ond through a layer of 1 cm thick, the viscosity 14.7 STAGES OF THE FILTER CYCLE
being 1 cp and the pressure drop 1 atm.
14.7.1 Pretreatment
Another material used as a filter aid is
cellulose fiber. Besides serving as a filter aid, Before the filtration actually starts, the condi-
particularly over screens on pressure leaf fil- tion of the slurry can be modified to a certain
ters, it is combustible in case the filter cake extent for the purpose of making the separa-
has to be incinerated for disposal or product tion easier by increasing the size of the parti-
recovery. cles to be filtered out. Larger particles settle
A recent addition to filter aid materials is faster, and also make more porous and less
rice hull ash (RHA), which is from 92% to resistant filter cakes. Both chemical and physi-
97% amorphous silica dioxide. These calcined cal methods can be used.
curved rigid particles have a porosity that Pretreatment, also called conditioning in
makes them suitable for body feed filtrations. water treatment, is done in several ways. For
Examples and filtration characteristics are clarifying operations, with cartridge filters, for
described by Reiber.57 example, the use of a coarser filter before the
Major uses of diatomite filter aids include final filter is a common approach. Choices as
food and chemical processing, brewing, phar- to the relative retention values need to be
maceutical, metalworking, and electric power determined by tests for the most economical
industries. Recently, they have been used more results.
in the municipal water treatment field and In the case of cake filtration the most fre-
also for clarifying drinking water supplies. quent method used is to thicken the solids
More than 170 plants using DE have been content in the slurry. This has a great effect
installed since 1949. They are also effective in on the performance of cake filters. It affects
controlling the waterborne disease giardiasis.58 the capacity and cake resistance. For example,
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 697

for the same cycle time, if the concentration is electrokinetic force that controls this process
increased by a factor of four, the production is called the zeta potential.
capacity is doubled. Or alternately, filtration
area can be cut in half for the same capacity.44 14.7.2 Filtration
Physical pretreatment can include heating After the pretreatment step, the slurry is fed
the slurry to lower the viscosity and improve to the filter by gravity, pressure pumps, or
the flow rate. It may be cooled to chill and vacuum sources. For pressure filters, slurries
solidify waxes, for example, so they will filter are usually fed by diaphragm pumps. They are
out. Other physical means are ultrasonic and easy to control by compressed air, and when
mechanical vibrations, magnetic treatment, the filter is loaded with filter cake, they reach
and ionizing radiation. maximum pressure and stop. Gear pumps are
By far, the most frequently used methods often used for small clarifying filters. Higher
are addition of chemicals to coagulate or pressure progressive-cavity pumps are used
flocculate the particles by changing the parti- for sludge filtering up to 225 psi. Pumps
cle charges. This is particularly helpful in fil- frequently are automatically controlled to
tering colloidal suspensions, usually consid- increase pressure gradually as the cake
ered as containing particles from 0.001 to 1.0 resistance increases.
microns in size. Natural electrolytes such as
alum, lime, ferrous sulfate, and ferric chloride 14.7.3 Cake Washing
decrease the surface charge and are called For cake products, such as pigments, the
coagulants. mother liquor must be removed. Formerly
Flocculants can be either natural or syn- simple displacement washes were used that
thetic chemicals, which cause dispersed parti- were inefficient owing to large volumes of
cles to form relatively stable aggregates of wash fluid required. Also cracks were formed
particles. These settle and filter more easily. in the filter cake, causing bypassing of wash
Higher molecular weight long-chain organic liquids. Recently, the membrane or diaphragm
chemicals called polyelectrolytes are widely filter press has prevented this problem by
used in this process. They are available com- squeezing the filter cake before single- or
mercially in liquid, powder, or emulsion forms, multiple-wash cycles.
and also anionic nonionic and cationic types. Washing on continuous belt or vacuum fil-
The science of selecting them has been highly ters is done by spray washes over the collected
developed.59 Some modern filters such as belt solids either in a single pass or in a counter-
pressure filters for sewage sludge filtrations current mode. Multiple washes are possible
would not be cost effective without the use of where needed and are effective.
suitable polyelectrolytes.
Although there has been some confusion 14.7.4 Solids Discharge
about the terms coagulation and flocculation, In small polish and clarifying filters, the col-
they are better thought of in terms of func- lected contamination is disposed of with the
tion. A good explanation is given by the publi- spent cartridge or filled filter bag. If haz-
cation by Zeta-Meter, Inc.60 Coagulation takes ardous, the volume of either can require com-
place when the energy barrier between parti- pacting to save space and reduce disposal costs.
cles is lowered so that the net interaction is In filter presses, cakes are removed manu-
always attractive. This is also referred to as ally in small units. Larger filters have plate
destabilization. Flocculation refers to the suc- shifting devices that separate the plates, per-
cessful collisions that occur when the destabi- mitting the solids to fall into a receptacle
lized particles come together and form ag- below the filter. Conveyors can also be used
glomerates and then visible floe masses. The under the filter to transfer the cakes to dis-
698 HANDBOOK OF POWDER SCIENCE

posal containers, or to a downstream process industry. Many conferences have been held
such as a dryer. Rotary and belt continuous both in North America and in Europe and
filters discharge over a roll or from a scraper. more than 1000 articles have been presented.
It is encouraging to note that many younger
14.7.5 Drying of Filter Cake engineers are becoming more active in the
Drying of filter cake in filter presses can be field, especially in research and development
done by compressing the cake to remove mois- programs, many funded by the U.S. Depart-
ture, in many cases up to 75%. If additional ment of Energy, the National Science Founda-
drying is needed, air can be blown through the tion, and the U.S. Environmental Protection
filter cake through the wash plates in the Agency.
press. On rotary vacuum filters, air or dry The research covers many areas such as
steam can be used for drying. Some filters also cake compressibility, expression of solids, per-
have compression mechanisms on the top side meability studies, and recently an entire con-
of the filter drum. ference was devoted to the pore and porosity
upon which all relationships in filtration ulti-
14.7.6 Downstream Drying mately depend. This was the Hershey Con-
ference held in May 1991. The proceedings
A number of different cake dryers are avail-
were not published but some papers were
able for waste sludge drying. Typical are coun-
summarized.3
tercurrent hot gas dryers, and paddle or heated
blade type dryers. Product dryers utilize con- Having been able to survey most all of the
ventional spray drying equipment suitable for work done, I will be selective and subjective
the crystals or solids collected. Resultant while trying not to omit any important papers.
fine-dried powders are then packaged as Of course, the development of new theories to
completed product. add to the already extensive literature contin-
ues. From the Pore Conference, Tiller gave a
tutorial on the parameters of pipes and pores.
14.8 LITERATURE AND INFORMATION A mathematical analogy was used in which
REVIEW pipe flow equations for friction factors and
Reynolds numbers are modified for flow in
At the time the first edition of this handbook porous media. Hypothetical pores are ana-
was published, there was a paucity of informa- lyzed showing how the void ratio times the
tion in the general field of liquid filtration. specific surface relates to a channel in a porous
There is a journal called Filtration and Separa- bed. Permeability and equivalent pore diame-
tion started in 1964, and the Filtration Society ter are shown as a function of the fractional
was organized in England the following year. distance in both moderately compactable cakes
4
A series of Filtech conferences began in 1967 and also highly compressible ones.
and have continued. Even so, at the time, Another study tried to resolve theoretical
concerned filtration engineers and academe and experimental problems relating filtrate
were calling for more basic teaching and volume to time in constant-pressure filtrations.
courses in filtration and separation. Problems arise in interpreting theoretical
However, in the last 10 years, a great deal derivation and experimental techniques such
more information has been published and as nonuniform cake deposition, variable slurry
many conferences were held. The new Ameri- concentration, degradation of floes, and clog-
can Filtration and Separations Society pub- ging of cake and supporting media due to
lishes the Fluid/Particle Separation Journal de- migrating fine particles. Reviewed are basic
voted to all phases of the subject. Pioneered planar filtration theory, simplified equations
by Dr. Frank Tiller of the University of for constant pressure filtration, parabolic data
5
Houston, it has gained wide acceptance in our analysis, and determining instantaneous rates.
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 699

One of the important areas of cake filtra- Tiller also presented two papers, one relat-
tion is the compression or expression of filter ing to relative liquid removal in filtration and
cakes by mechanical means after the filtration expression detailing experimental techniques,11
part of the cycle is completed. Prof. M. Shirato and the other concerning improved formulas
from Nagoya University in Japan has done for constant pressure filtration and com-
much work in this field. One of his recent paction of filter cakes.12 Dick et al. wrote
research reports focussed on compression fil- about how capillary forces are related in com-
ters using hydraulic expression with a perfo- pressible filter cake filtrations.13
rated membrane.6 He has recently retired, but Because rating of filter cartridges is a timely
his successor, Prof. Murase, is continuing the and sometimes controversial subject, many ar-
work. He recently explored the problem of the ticles have been written on it. Johnston says
filter cake expression being stopped before the micron rating of a membrane or filter
reaching equilibrium state, causing the cake cartridge can frequently be misleading to the
stress to decrease as the material relaxes. The user. Because filtration is not a pure sieving
study analyzes this condition. It was found that process, its efficiency can be affected by
the cake stress does not depend on either the medium thickness, the nature of the fluid,
constant pressure or constant rate filtration.7 and the fluid flow rate. He emphasizes that
Willis looked at the mechanics of non- no single factor can characterize a filter med-
ium—at least five are necessary: porosity, per-
Newtonian fluids on nonstationary particles to
meability, thickness, material of construction,
determine the applicability of Darcy's law.
and whether or not pores on one face are
They identified the physical significance and
larger than on the opposite face.14
the limitations of this law under these circum-
Many studies on cartridge filters address
stances.8
filter test methods rather than theory because
Willis and Chase considered multiphase
critical applications depend on filtrate or prod-
processes in filtration. They proposed a gen-
uct analysis with emphasis on final particle
eral strategy for developing a fundamental
count. Williams wrote on testing performance
framework and a systematic approach for eval-
of spool-wound cartridges.15 Verdegan et al.
uating any multiphase porous media process.
covered recent advances in oil filter test meth-
Concepts of scale, analogy, and averaging,
ods for cartridge filters.16 The effects of tem-
along with the characteristics of basic princi- perature and volume on filter integrity tests
ples and scientific analysis are used.9 were studied by Scheer et al.17 Another study,
One of the most interesting pursuits of Prof. by Bentley and Lloyd,18 concerned interpreta-
Frank Tiller, who at this writing is 76 years tion of ratings of cartridge filters.
old, is a historical review of papers on filtra- Chiang wrote on the interfacial phenomena
tion theory that were presented at technical in fluid-particle separation. This article gives
meetings some 50 to 75 years ago. These early a complete and detailed study of the most
filtration researchers frequently raised perti- important area of surface-interface relations.
nent questions that could not be answered at The degree and rate of separation are influ-
that time because of lack of instruments to enced by this behavior. The four basic selected
make as precise determinations as we have points covered are: surface of solid particles,
now. However, Tiller reviews and comments fluid-solid particle interface, application of in-
on their questions and provides current theory terfacial surface tensions, and experimental
in explanation of these early investigations. techniques and instruments.19
This is a very valuable contribution for new New filter media were the subjects of many
students of filtration and even experienced articles. Gregor updated media selection re-
engineers involved in filtration process sulting from more demanding environmental
development.10 regulations. Finer filter media and specialized
700 HANDBOOK OF POWDER SCIENCE

media are also covered in this article along Reducing levels of impure chemicals that cause
with options, cost, advantages, and efficiencies yield losses can be controlled by submicron
of existing and new media.20 Mayer explains filtration along with molecular sieve drying
the use of spun-bonded polyolefin nonwovens and fractional distillation as a high-purity sol-
for micro-filtration.21 vent reprocessor. The hazardous waste result-
Bergmann details a new growing market of ing is minimized to about 3% of the original
filter media for blood and medical applica- volume.29 A new membrane process called
tions.22 New uses of new nonwoven filters nanofiltration is expected to widen the use of
made by the melt blown process are presented membranes in liquid-phase separations in the
by Manns. This new method produces mi- chemical process industries. An article com-
crofine fibrous webs with fibers as fine as 1 to pares properties and performance characteris-
10 microns in diameter. The material is made tics of commercially available NF mem-
directly from thermoplastic resins and has a branes.92 A current review of membrane
number of uses such as media for pleated separation technologies for wastewater treat-
cartridges.23 ment is presented by Cartwright and includes
Membrane filters, one of the fastest growing options and comparisons for selecting the best
segments of the entire separation spectrum, method.95
recently estimated in Ref. 97 at 10% per year Pretreatment of slurries by chemical poly-
and reaching $2 billion per year in 1996, was electroytes is essential in many filtrations, and
the subject of many articles. In fact, the an- selecting the proper chemical is a task that
nual membrane conference has had its tenth frequently has to be done by testing. A good
meeting. Membrane fouling in RO systems overview of the use of polymers and inorganic
was discussed by Kronmiller; use of high- coagulants is presented by Mangravite.31
purity water for the semiconductor industry Scheiner discusses the removal of toxic metals
was described by Parekh; crossflow filtration in from waste water by testing 21 different floc-
food applications such as fruit juices was culents. The testing procedures determined the
evaluated by Short; and pervaporation future optimum flocculating agents used to achieve
markets were outlined by Bartels.24 allowable levels of cadmium and lead where
The market for microfiltration membranes hydroxide treatment did not work.30
for environmental purposes was covered by Probably more articles and papers were de-
Cartwright for system design in pollution con- voted to equipment design, performance, and
trol with emphasis on the crossflow applications than anything else. We will men-
technique.25 An article by Duran explained a tion only a few that are new or cover impor-
new water treatment technology involving tant improvements in existing equipment. Fil-
nanofiltration membranes in a spiral-wound tration has been combined recently with dry-
configuration that function at 75 to 130 psi. ing and other processing operations. An inter-
This method replaces conventional lime treat- esting review of this area, in which filtration is
ment.26 In the food industry, a method of used with as many as 16 different processes
using BASIC computer programs to solve relating to heat and mass transfer operations,
problems of the effects of transmembrane was made by Yelshin. Robotic principles, au-
pressure on orange juice concentration was tomation, and a unique concept of using rota-
described by Toledo.27 The development of a tional machines and conveyor systems in the
special asymmetric membrane for hazardous filtration process is presented.32
waste removal in waste water in the electron- A new type of water screen filter is de-
ics industry was discussed by Sternberg.28 Also scribed that is self-cleaning by using a pres-
in the microelectronics field, where liquid pu- sure senser to activate a back-flushing action.
rity is critical, a method of point-of-use treat- No shut-down is required and particles can be
ment of chemical baths was given by Carr. removed down to the 10 to 15 micron range. It
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 701

can be used for any water that needs to be celeration. Selection data and applications are
cleaned or recycled. Individual units can also given.40
handle up to 5000 gal/min. 33 Ekberg describes a vacuum disc filter, long
Continuous belt and belt filter presses con- used in the minerals field, that uses a new
tinue to increase in usage as manufacturers sintered alumina disc medium with average
make improvements. Besides mineral and pore sizes of 1.5 to 2.0 microns. See Figure
chemical processes, many applications are in 14.35. When the pores in the medium are
waste water treatment. Schonstein discusses a filled with liquid, they prevent air passage in
vacuum belt press for paper dewatering use.34 the vacuum cake drying part of the filter cycle
Mau shows how a vertical automatic pressure because of the pressures created in the pores
filter equipped with horizontal filter plates with due to capillary action. Thus the filter discs
squeeze diaphragms can improve sludge filtra- are easier to back-flush, and do not require
tion.35 Deutsch explains the operation, fea- filter cloths.41
tures available, and options for selecting belt An entirely different type of filter is the
filter presses.36 tube press, which was invented 20 years ago
In waste water treatment, centrifuges have a for clay filtrations. It has recently been im-
unique advantage over other conventional fil- proved, with larger filter modules. It is now
ters in that they are enclosed, odor-free, safe, being used in the mining, chemical, and other
and require only minimum labor. One draw- fields as surveyed by Johns.42
back, that of lower solids output, has been
addressed by manufacturers and considerable 14.9 TYPES AND DESCRIPTION OF
improvements made. Leung describes a high LIQUID FILTER EQUIPMENT
solids decanter centrifuge that gives cake solids
above 30% in dewatering mixed primary/sec- Starting with batch equipment, then continu-
ondary sludge.37 Albertson also writes about ous, the various types of current filters in use
improved designs for high cake solids and also will be described. Emphasis will be on most
use of centrifuges for mechanical dewatering recent developments in design and application
processes in general.38'39 Morgenthaler as- while still considering the older types, many of
sesses decanter centrifuges for environmental which are still widely used in industry. Param-
applications using feed rate, polymer addition, eters such as cake washing capabilities, driving
and concentration and the suspended solids in forces, settling rates, types of discharge, and
the feed, cake, and effluent. Equations are cake compression will be added where they
given for calculating polymer consumption, re- relate to the particular filter.
covery of suspended solid particles, and deter-
mining the specific gravity based on density 14.9.1 Batch Filters
and weight percents.93 De Loggio reviews re- 14.9.1.1 Filter Presses
cent design innovations in centrifuges for the
The filter press is the most common type of
chemical process industries. For example, new
pressure batch filter and the oldest, originating
vertical decanter models can handle process in the early 19th century. Its development into
streams up to 700° F and 150 psig. A good a modern efficient, versatile, and flexible filter
selection table of different types of centrifuges has kept pace with technological improve-
is presented.94 ments. As shown in Figure 14.12, it is a series
West discusses the disc-bowl centrifuge in- of plates and frames, or recessed plates
cluding centrifugal settlers, solid-bowl nozzle mounted on side bars and supported by a
types, and conventional nozzle types. He also suitable structure. The plates are held to-
explains sigma theory in regard to the rela- gether during filtration by hydraulic or me-
tionship between geometry and centrifugal ac- chanical pressure. The slurry is fed into a
702 HANDBOOK OF POWDER SCIENCE

Sizes can range from laboratory filters up to

large production units of 6000 ft2 of filter area
and 350 ft3 of cake capacity. Plates can be
made from metals or polypropylene. Figure
14.13 shows typical polypropylene plates. The
feed can be from the center or a corner of the
plate. Plates and frames are mostly used for
clarifying filtrations, where filter paper can be
used over the filter cloth for easy removal of
the solids, and clean papers can be inexpen-
sively used for each batch.
More filter presses are used for cake opera-
tions, and the filter cloth frequently lasts many
Figure 14.12. Filter press showing plates and frames.
(Avery Filter Co. Laboratory filter press) cycles, as many as 500 to 600 times repeat
usage. Cake filters use recessed plates in which
the solids are collected between plates in the
specific port in parallel, and flows through the recess on each adjacent plate, producing final
filter medium covering the drainage area of cake thicknesses of from 25 mm to 50 mm,
the plate. The filtered liquid is discharged although 32 mm is most common. The major
through another suitable port, and the solids advantage of recessed plates is that when the
are collected within the filter chamber. They filter press is opened for discharge, there is no
are released at the end of the filter cycle by frame to retain part of the cake, and it falls
separating the plates, either manually for small free by gravity from the open chamber. Auto-
filter presses, or automatically for larger units. matic plate shifters can thus be used to facili-

Figure 14.13. Typical polypropylene plates. (Klinkau GmbH and Co.)

 FILTRATION OF SOLIDS FROM LIQUID STREAMS 703

tate and automate the cake discharge. If nec- maximum operating pressures may be 50 psi.
essary, all of the functions of the press cycle A typical unit is shown in Figure 14.14.
can be controlled by programmable computer
systems including feed flow rate, mass solids in 14.9.1.3 Membrane Filter Presses
the feed, closing and operating pressures,
change of feed pump pressures during the Also called diaphragm presses, membrane fil-
filter cycle, opening to discharge the filter cake, ter presses (Fig. 14.15) utilize a special plate
and closing the filter to start another cycle. with an impermeable flexible drainage area on
Typical flow rates are from 0.1 to 1.0 gal/min the filter surface of the filter plate (Fig. 14.16).
per ft2 of filter area. Cake solids content This is separated from the body of the plate,
usually range from 25% to 40% depending and can be inflated by air or water after the
on pressures and nature of the solids being end of the cake-forming part of the filter press
filtered out. cycle. This compresses or squeezes the cake to
remove more liquid. This is the most impor-
tant development of the filter press in the last
14.9.1.2 Sheet Filter Presses
decade. The improvement in the performance
Sheet filter presses are so called because they is shown in Figure 14.17. A diagram showing
use cellulose filter sheets of about 0.125 in. operation of the membrane filter press is
thick. Frequently they contain a charged pow- shown in Figure 14.18. Generally, solids con-
der to effect an attraction for submicron parti- tent of up to 75% may be achieved with sav-
cles. These are depth media and are used in ings of downstream drying and sludge cake
beverage, cosmetic, pharmaceutical, plating disposal costs. Examples are given by Mayer61
solution, electric discharge machining (EDM), for the waste disposal area as compared to
and transformer oil filtrations. The filters are rotary vacuum filters and centrifuges, and by
of stainless steel structural construction, and Avery for the food and chemical industry.62

Figure 14.14. Sheet filter presses installed in a beverage plant. (Seitz Werke)
704 HANDBOOK OF POWDER SCIENCE

Figure 14.15. Membrane filter press showing overhead manifold and compressed air connections to membrane
filter plates. (Avery Filter Co.)

Automation may also be applied to the cy- sugar, and chemical industries. Good washing
cle times; the squeeze function including time of filter cakes is made easier because the
and pressure; blowdown, wash, and discharge cakes are in a horizontal position, thus permit-
actions; and cooling or heating the filter plates. ting the wash liquid to uniformly flow through
Membrane plates are polypropylene, but some the filter cake. The wash water is readily re-
have steel bodies and replaceable neoprene moved by the squeeze action of the mem-
diaphragms. brane, and a second wash can be done if
needed.
14.9.1.4 Vertical Automatic Pressure Filter
14.9.1.5 Batch Filters Using Closed
This filter has a plate stack similar to a filter
Pressure Vessels
press except it is in a vertical position as
shown in Figure 14.19. The plates are horizon- These filters have in common a closed tank or
tal, with a membrane on the upper side of the housing containing the filter leaves, plates,
plate only. This limits the capacity of the filter, bags, tubes, or cartridges. Pressures usually do
but it can have areas up to 11 m2. Cake not exceed 100 psi and the size can range from
volumes per cycle can be up to 30 ft3. Because single-element cartridge filters to large hori-
of short 10- to 25-min cycles, the overall out- zontal tank leaf filters of up to 3000 ft2 in
put can be large. The filter cloth is a continu- filter area. The vessels can be made from
ous belt passing in between the plates, and stainless steel or other suitable materials in-
capable of being washed after every filter cycle cluding plastic for smaller sizes.
(Fig. 14.20). The cake is removed from the
3 in. diameter discharge rollers by knife 14.9.1.5.1 Pressure Leaf Filters. Pressure
scrapers. The filters are automatically oper- leaf filters can be vertical or horizontal tank
ated and are used extensively in the mining, designs, the latter capable of larger areas. A
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 705

• Initial water content

10 20 30 40 50 60 70 80 90100
- Filtration »»| |<4- Squeezing

Conventional filtration

Figure 14.17. Conventional filtration versus membrane

filter press operation showing reduction in cycle time
with cake squeezing. (Lenser America, Inc.)

enclosed in a vertical tank. The plates are the

same size and are stacked vertically, the num-
ber relating to the size required. Generally the
maximum area is about 200 ft.2 The slurry
Figure 14.16. Center feed membrane filter press plate flow is either up the center tube, or in from
cross-section showing separation of membrane from
body of filter plate. (Lenser America, Inc.)
the side, then through the filter medium, usu-
ally filter paper. Pressures are moderate, rarely
exceeding 60 psi. A typical unit is shown in
typical vertical unit is shown in Figure 14.21 Figure 14.24.
and a horizontal one in Figure 14.22. A cut- This is primarily a clarifying filter, fre-
away view of a filter leaf is shown in Figure quently using activated carbon decolorize par-
14.23. The center coarse drainage member is ticles, and a filter aid to assist the filtration.
covered by a fine mesh, usually 60 mesh, or a One advantage is that the filter cake formed
24 X 110 dutch twill weave wire cloth, and the on the horizontal surface is stabilized so that it
frame riveted or bolted together. The leaves is uniform and not affected by intermittent
discharge the filtrate through a connection to operation. Because of stacking of plates, the
a manifold pipe at the bottom of the filter. filter is compact, taking up little floor space.
Filter cloth can be used over the leaves, but Inexpensive filter paper can be used and re-
most often the filters operate with a diatoma- placed for each batch, providing uniform flow
ceous earth filter aid, as most of the applica- conditions for each successive cycle because
tions are for clarifying liquids. The solids col- the filter medium resistance is the same each
lected on the leaves can be sluiced off with time.
water, or can be removed as a dry filter cake
manually or by vibrators. Major uses are for 14.9.1.5.3 Horizontal Pressure Plate Filters.
fruit juices, beer, sugar solutions, wines, and Horizontal pressure plate filters with centrifu-
chemicals. gal cake discharge permit multiprocessing
stages such as filtering, washing, drying, and
14.9.1.5.2 Horizontal Plate Batch Filters. discharging of the filter cake automatically. A
The horizontal plate batch filter consists of typical filter is shown in Figure 14.25. The
multiple round filter plates of metal or plastic filter leaves have a drainage member, with a
706 HANDBOOK OF POWDER SCIENCE

1. Relaxed Diaphragm 2. Chambers Filling

Avery's Empty Chamber Membrane

Filter Press Operation

1. Avery's Empty Chamber membrane design, but

membrane filter plate, allows squeezing of the
before filling. Plate is cake into partially filled or
empty, polypropylene empty chambers with no
membranes, as in standard minimum cake thickness.
plates, are relaxed. While still offering improved
cake washing, the Empty
2. Feed pressure compresses Chamber design provides
impermeable membrane important safeguards against
(a) against the plate core operator error and assures
as solids collect on the long plate life.
filter cloth (b) and form
the cake (c). Empty Chamber filter
3. In the Squeeze phase of plates produce consistent
the Empty Chamber solids content cake of any
membrane plate, the design final squeezed thickness
provides the same high cake allowing flexibility for
solids as the standard varying batch sizes.

Diaphragm Squeezed
Into Empty Chamber

Figure 14.18. Membrane filter press operation with new empty chamber membrane filter plate. (Avery Filter Co.,
Inc.)

fine stainless wire cloth over it. They may also catalyst and bleaching earths from edible oils
be fitted with filter cloth. Once the filter cycle and fatty acids. Units are available of up to a
is complete, and the solids are on the plates, 1000 ft2 area. Feed with medium to slow
the feed is stopped, and the stack of plates is settling rates are typical.
rotated at speeds up to 300 rpm to dislodge
the filter cake. It can be done dry, or sluiced 14.9.1.5.4 Single Plate Pressure Nutsche
with a liquid. A different design is used for Filters. Single plate pressure nutsche filters
each type of discharge. can have diameters up to 15 ft, areas of 135
The closed filter vessel is safe in a haz- ft2, and with a 12 in. thick filter cake, a volume
ardous environment and protects workers of 135 ft3. In some cases, filter cakes can be
when they are filtering toxic chemicals. In thicker, giving more cake capacity. These fil-
addition, product purity is maintained and au- ters serve the need for filtrations that can
tomation improves production and reduces isolate the final product to maintain purity and
labor costs. Typical uses include recovery of avoid toxic exposure to the environment. Such
previous metal catalyst, gold precipitate recov- needs are prevalent in the chemical and phar-
ery, separation of antibiotics, and removal of maceutical industries. Thus there are several
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 707

reslurry the cake, refilter it, smooth and com-

LAROX PF19-Q filter with press the cake, dry it, and discharge it from
203.4 ft.2 fflraiion arm
the filter without opening it. All of these steps
can be done automatically. They are made
mostly from stainless steel, but other alloys
can be used; there is even a glass-lined unit
with agitation (Fig. 14.27).

14.9.1.5.5 Pressure Tubular Filters. Used

for clarifying or solids thickening operations,
pressure tubular filters are closed pressure
vessels operating up to 100 psi, and feature a
tubular filter element. There are a number of
types used—wire wound perforated tubes,
wedge wire tubes, porous ceramic, plastic,
metal, and plastic tubes; and membrane-
Figure 14.19. Automatic vertical membrane filter with
laminated felt covers over perforated tubes. A
203.4 ft2 of filter area. (Larox) typical pressure tubular filter is shown in
Figure 14.28. They are sometimes called can-
dle filters because of the vertical tubes. They
manufacturers supplying this highly special- are placed in vertical vessels, in multiples usu-
ized filter. A typical unit with agitation is shown ally on a tube sheet. The feed can be to the
in Figure 14.26. inside or outside of the filter tube. In either
Being totally enclosed, they can operate un- case, once the filter cycle is complete, a re-
der inert gas pressure, and even vacuum if verse hydraulic pulse of air, gas, or liquid
needed. Their versatility comes from their dislodges the solids in about 5 min. In catalyst
ability not only to filter and wash the filter recovery and recycling, 100% can be captured
cake by displacement washes, but also to and recycled. In some cases, tubes are pre-

1. Filter cloth washing

2. Counter pressure roller
3. Filter cloth centering roller
4. Filter cloth driving roller
5. Filter cloth length compensator
6. Guide roller
7. Cake scraper
8. Filter cake
Figure 14.20. Schematic of filter cloth path through filter in Figure 14.19. (Larox)
708 HANDBOOK OF POWDER SCIENCE

Figure 14.22. Pressure leaf horizontal tank filter, 950

ft2. (Duriron, Filtration Systems Div.)

volumes of liquids. Many cartridge filters con-

sist of a suitable metal or plastic housing that
holds 1 to 16 cartridges for flow rates up to 10
gpm per 10 in. cartridge length. Larger pres-
sure vessels may hold over 100 cartridges, and
operate up to 150 psi. High-pressure metal
cartridges can operate up to 500 psi. A group
Figure 14.21. Pressure leaf vertical tank filter. (Duriron, of typical cartridges is shown in Figure 14.30.
Filtration Systems Div.)
The original string-wound cartridge depth
type was developed about 75 years ago, and
coated with a filter aid for polishing liquids since then many other designs have appeared
such as water, solvents, beer, wine, reagents, including pleated, grooved, sintered, resin
boiler condensates, or acids. bonded, thermally bonded polypropylene mi-
Whether feeding to the inside or outside of crofiber, controlled gradient density, and solid
the filter tube, the time of the backwash deter- porous cylinders. There is also a new com-
mines the overall filter capacity. A new design pound radial pleat that increases loading ca-
using a membrane-laminated medium over the pacity. The dirt holding capacity is one of the
support tube can back-pulse the tube as often basic selection factors. Performance and parti-
as every 2 to 3 min and remove particles down cle size retention are the other most important
to 0.5 microns. The discharged solids settle parameters to consider. Multiple layers of me-
into the cone-shaped bottom of the filter dia may hold activated carbon or diatomites
vessel to be concentrated, and periodically
removed (Fig. 14.29).

14.9.1.5.6 Cartridge Filters. Cartridge filters

are a widely used and important class of liquid
batch pressure filters. They maintain or im-
prove cleanliness of many liquids including
water, hydraulic fluids, food and beverage
products, oils, paints, and many other prod-
ucts. Generally they remove small amounts of
particulate contaminants, usually in amounts Figure 14.23. Cutaway view of typical filter leaf.
of less than 100 ppm, from relatively large Liquid-Solids Separation Corp.
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 709

Figure 14.24. Typical horizontal pressure tank filter. (Ketema Inc.)

between them for organic material removal, tests showed that pleated cartridges, after cake
special decolorizing, or finer filtering. formation begins, achieve 98 + % efficiency
Although some cartridges are interchange- equal to about 1 micron performance regard-
able with other manufacturers' housings, there less of the medium.
are no current standards that relate ratings Other investigators have attempted to clar-
and performance among different suppliers in ify time-dependent variations in cartridge per-
the chemical and related industries. On the- formance. Juhasz questioned Beta, Beta Prime,
other hand, critical hydraulic filter systems and Epsilon rating procedures and suggests
have standards set by ASME and the API. that three components of downstream level
Much work has been done on comparing contamination should be considered, namely,
different cartridges. For example, Sandstedt instantaneous efficiency, unloading, and
and Weisenberger65 report on the confusion leakage.66
about micron ratings, saying there is frequent
failure to specify the level of efficiency as part 14.9.1.5.7 Bag Filters. The concept here is
of the performance rating, that there is no a sewn filter bag of fabric, felt, or mesh, of a
acceptance of a single, standard test procedure specific retention value that is made to fit into
to predict performance in many applications, a filter housing with proper seals to prevent
and that the difference between clarification bypassing of the liquid to be filtered. Typically,
and classification is often overlooked. Their a No. 1 standard filter bag size has 2.5 ft2 of
710 HANDBOOK OF POWDER SCIENCE

Typvft
Completely dosed filter with
automatic extraction, for recovery
of the residue dry
Figure 14.26. Pressure nutsche filter 0.6 m 2 for phar-
maceutical pilot plant. Filter is open to show single
plate which supports the filter medium (Rosenmund,
Inc.).
Figure 14.25. Horizontal plate filter with centrifugal
cake discharge feature. (Steri-Tech, Inc. Funda)
capacity of a standard filter cartridge. This
2 means fewer changeovers and lower disposal
filter area, and a No. 2 bag has 5 ft . Filter
costs. This is an advantage over cartridges
bags and housings are shown in Figure 14.31.
because, for equal filter areas, the disposal
Generally, the percent solids in the feed stream
value is less for bags than for cartridges.
are low, but may be as much as 1%, which is
Generally, bag filters offer higher flow rates
much greater than for cartridge filters. Multi-
with lower pressure drops. Cartridges with
ple bag units are available and flow rates can
depth or extended surface area offer greater
reach 1000 gal/min. The original idea was
reliability and efficiency of particle removal.
developed by Wrotnowski, who found that the
felt media did an excellent job of classifying
14.9.2 Vacuum Filters
paints and inks.67 Subsequent developments
have made the bag filter a popular clarifying Using a vacuum as the driving force for filtra-
filter. New and more retentive materials have tion is common in rotary vacuum filters which
increased the acceptance of this filter. An ap- have the advantage of continuous cake dis-
proach to increase the filter area and keep a charge. Forces up to 0.8 bar can be enough to
small size was reported by Johnson,68 using produce improved filtration rates. Vacuum
multiple layers with bypass openings. This de- sources are simple, and the filters from small
sign of filter bag has five times the filter area nutsches to large-scale equipment have been
of a standard bag, and 15 times the dirt load well developed since the first rotary vacuum
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 711

TOP VENT

DOME
DRAIN

"WEDGE WIRE'
ELEMENTS

The ClaRite Filter showing configuration

for "air bump" design.

Figure 14.28. Pressure tubular filter. (Croll-Reynolds)

the filtration, washing-drying, and discharge

Figure 14.27. Special pressure nutsche with glass-lined steps take place. Cake discharge is done by
vessel and agitator. (Zschokke Wartmann, Ltd.) different methods such as scraper, air blow,
rollers with strings, and belt. Filter cloths,
mostly synthetic but sometimes metal, are used
drum filter was introduced in England over with air flows in the range of 25 to 100 cfm on
100 years ago. Some disadvantages are that the Frazier scale. Rotation speed is about 1
the production rate per unit area is low, so rpm. Maximum size can be 1700 ft2, although
that larger equipment is needed for large vol- from 200 to 600 ft2 is much more common.
ume. Some rotary vacuum filters can be 16 ft Outputs on the very largest units can be up to
in diameter and 33 ft long. Solids content can 120 tons per day.
be relatively low as compared to those ob-
tained using pressure filters, and limits apply
to volatile liquids. 14.9.2.2 Precoat Drum Filter. The precoat
rotary vacuum filter is completely dependent
on the use of a filter aid, which in almost all
14.9.2.1 Drum Filters
cases is diatomate. A typical filter system is
Vacuum filters are used for both solids recov- shown in Figure 14.34. The diatomate slurry,
ery and fine clarifications that utilize diatomite from 4% to 8% by weight, from the precoat
and are called precoat filters. The vacuum tank is applied to the filter drum to form a
drum filter is widely used in industry as shown thickness of up to 6 in. After successful pre-
in Figures 14.32 and 14.33. The drum has coating, the filtration begins, using admix, also
sections controlled by a rotary valve on which called body feed, which is fed along with the
712 HANDBOOK OF POWDER SCIENCE

1. Filtration 2, Back-pulse & Cake discharp

While filtration takes Back-Pulse Filtration Cake removed by momentary
place, cake previously pulsed flow reversa! settles rapidly
from the filter tubes for discharge as a highly
settles into the filter cotrn concentrated slurry.

flB Plant Air

( ! | Filter Cake/Underflow

W. L. Gore & Associates, Inc.

Figure 14.29. Tubular back-pulse filter. (W. L. Gore and Associates)

influent liquid to the filter bowl. As the solids data indicate initial optimum conditions. Plant
collect on the precoat, an advancing knife runs can then be observed to see if test results
shaves off as little as 0.001 in. of solids for are effective. The angle and the desired rate
each revolution of the filter drum. Filtering of knife advance can often be determined only
factors are given by Smith69 as: by trial and observation.

Grade of precoat Precoat cutting

efficiency
Vacuum Cake filterability
Drum speed Cake drying time
Knife advance time Liquid viscosity
Drum submergence Precoat thickness
Solids in precoat Continuous addition
of filter aid

The most important objective is to optimize

production output and to control cost by mini-
mizing filter aid consumption. Determining the
right combination of the above factors will
require leaf test and careful observation of
plant runs. Test filter leaves of 0.1 ft2 of filter
area are used with different grades of filter Figure 14.30. Cartridge filters — various types.
aids, cake thickness, and vacuum. Evaluating (Parker-Hannifin Corp.)
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 713

Figure 14.31. Bag and cartridge filters. (Commercial Filter Div. Parker-Hannifin Corp.)

14.9.2.3 Vacuum Disc Filters 14.9.2.4 Horizontal Continuous

Vacuum Filters
Vacuum disc filters are continuous rotary fil-
ters with circular vertical filtering discs The filter surface is formed as a table, a belt,
mounted on a horizontal hollow central shaft. or multiple moving pans in a line or a circular
(See Fig. 14.35.) The slurry fills a trough into arrangement. As with other filters with hori-
which the filter discs are submerged. The discs zontal surfaces, they maintain cake stability
are partitioned into sectors with suitable and thickness, and permit easy cake washing
drainage members to permit the filtrate to be including countercurrent. On the other hand,
fed to the rotating control valve. The filter they take up more space, and cost more than
leaves are usually covered with a filter cloth. drum filters. Their principal use is filtration of
As the leaves rotate and become submerged, gypsum and phosphate rock residues, metal-
the automatic rotary valve applies vacuum and lurgical sludges, pulp washing, and solvent
the solids form on the filter disc. Cake is extraction of oil seeds.
removed by scraper or roller and it drops The horizontal belt vacuum filter is made in
between the tank divisions. A recent develop- two basic designs, one with a heavy rubber
ment in porous ceramic technology has made underbelt carrying the filter cloth. This endless
possible filter discs that do not require filter belt rides on a vacuum or suction trough and
cloth as covers. The disc surface has a very has lateral drainage grooves for the filtrate.
fine pore size, permitting fine filtration and The slurry is fed at one end, and filtering and
easy cake release. Several of these new designs washing take place at the end of the filter,
are reported to be operating successfully in where the solids are dropped from the moving
Australia for zinc, lead, and copper concen- belt.
trate filtrations.70 The second variation of the belt filter uses
fcNDBOOK OF POWDER SCIENCE

Figure 14.32. Vacuum continuous drum filter. (Kromline-Sanderson)

Figure 14.33. Vacuum precoat drum filter. (Witco, Kenite Div.)

 FILTRATION OF SOLIDS FROM LIQUID STREAMS 715

PRECOAT TANK (1)1

ROTARY VACUUM
PRECOAT FILTER
DRUM
FILTER SEPTUM
DIATOMITE PRECOAT
FILTER CAKE (4)
TRAPPED SOLIDS
AND BODY FEED (6)

Figure 14.34. Complete precoat filtration system. (Arthur Basso, Ref. 56)

vacuum pans underneath the filter cloth with 14.9.3 Continuous Compression
no rubber supporting belt. The vacuum boxes Belt Filter
move forward intermittently as the vacuum is
applied as the cloth and pan move together. Originally developed in Germany in the 1960s
This type is generally of lighter weight con- for dewatering pulps in the paper and food
struction than the first design, and is used industries, the continuous compression belt
more on chemical, pharmaceuticals, and food filter quickly became adaptable for waste
products. Widths can be up to 2 meters and sludges. It was called a sewage sludge concen-
lengths to 40 meters. trator, although the common name now is a

Figure 14.35. Vacuum disc filter with ceramic filter discs. (Outokumpu Mintec USA Inc.)
716 HANDBOOK OF POWDER SCIENCE

Figure 14.36. Continuous belt filter press. (Komline-Sanderson)

belt filter press. A typical unit is shown in chanical plows or rollers may be used here
Figure 14.36. It is a heavy-duty mechanical to assist in the drainage.
machine that dewaters sludges that have been 2. The wedge or low-pressure zone. The sludge
properly conditioned with polymers. There are flows onto a carrying filter belt, becoming
various methods to do this, one of which is sandwiched between this and another over
shown in Figure 14.37. This shows a modern filter belt. By converging they apply gradual
controlled system that tends to optimize the increases in pressure to the sludge.
polymer-to-sludge ratio to reduce polymer 3. The dewatering continues as the two belts
costs. A study of compressible sludge proper- enter into a high-pressure or shear zone
ties in belt presses was done by Wells.71 around pressure rolls. These high shear
The process in this filter takes place in forces maximize the cake dryness. From
three steps:
here, the dewatered cake is continuously
removed by a doctor blade on a discharge
1. Gravity settling, in which the free water roller. The general configuration of the
drains from the treated sludge. Some me- press is shown in Figure 14.38.

Figure 14.37. Typical polymer control systems for belt filter presses. (Andritz Ruthner, Inc.)
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 717

-FLOCCULATION/DISTRIBUTION BOX
BELT WASH HOUSING GRAVITY SECTION

BELT TRACKING WEDGE SECTION-

SYSTEMS

BELT WASH
HOUSING

STEEL TUBULAR FRAME

Upper belt
Lower belt
Envirex SERIES 2000-x BELT FILTER PRESS

Figure 14.38. Schematic of a belt filter press. (Envirex Corp.)

Improving performance is detailed in an units installed in a plant in Camden, NJ

article by Lecey in which the mechanical vari- (Fig. 14.39).
able, the sludge characteristics, and optimum
operating conditions are discussed.72 Recent 14.9.4 Screw Presses
developments have provided a number of new Screw presses provide another way of continu-
features and they are described by Deutsch.73 ously compressing or squeezing a sludge mate-
Some installations in municipal treatment rial, particularly of coarse type materials such
plants involve multiple units such as the seven as organic waste, pulps, and fibrous materials.

Figure 14.39. Belt presses installed in a muncipal sewage treatment plant. (Enviroquip. Inc.)
718 HANDBOOK OF POWDER SCIENCE

It too benefits from the use of polymeric treat- cessed, and solvent cake washing performed in
ment chemicals to agglomerate the particles to pharmaceutical operations, dewaxing paraffin
be separated from the liquid. A schematic of a from oil-water mixtures, or removing extrac-
typical unit is shown in Figure 14.40. The tion agents from food processes. A sketch is
screw is an extruder type with a tapered center shown in Figure 14.41. Sizes are available from
shaft that compresses the product gradually as 0.12 to 7.68 m2.
it moves toward the discharge end. Feed can
be as low as 0.5% dry solids, and depending on
14.9.5.2 KDF Filter
feed composition, the solids discharged can
range from 15% to 70% dry solids content. Another continuous pressure filter is the KDF
Throughput can reach 2 tons of dry solids per from Amafilter (Fig. 14.42) in Holland. Its
hour in the largest unit available. design is a horizontal tank in which filter axles
are mounted on a rotating main shaft, each
with a particular number of elements at-
14.9.5 Continuous Pressure Filters
tached. Both the elements and the main shaft
Along with continuous vacuum and belt filters, rotate, using constant air pressure at 6 bar to
there are several continuous pressure filters effect the filtration. The air pressure gradient
that are quite unique in design, providing spe- provides the driving force and is also used for
cial applications that make their relatively high displacement dewatering of the cake. A chain
cost acceptable. type conveyor is used for cake discharge. With
50 m2 of filter area, it can produce filter cakes
of very low moisture content at capacities up
14.9.5.1 BHS Fest Filter
to 1750 kg/m 3 per hour. Developed in the
The first of these is the BHS Fest filter, devel- early 1980s, the principles are detailed by
oped in Germany in the late 1930s. It is an Kleizen and Dosoudil.74 Applications have
entirely enclosed low-pressure (up to 50 psi) been for coal fines, cement slurries, and coal
rotary drum filter in which the slurry is fed flotation concentrates.
into a filter chamber which is a segmented
part of the drum. Subsequent chambers com-
14.9.5.3 Ingersoll-Rand Filter
plete the filtration, wash the cake, then dry it
and prepare for discharge. In this self- The Ingersoll-Rand continuous pressure filter
contained environment protected unit, toxic, is a third commercial device of this category as
hazardous, or solvent materials can be pro- shown in Figure 14.43. It is used not only for

FEED
A D U S T ABLE INLET HOPPER ~x I
PRESSURE CONE

CAKE DISCHARGE

FILTRATE OUTLET
Figure 14.40. Schematic of a typical screw press. (Bepex Corp.)
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 719

DRYING One drawback is that the close tolerances

GAS
cause wear with highly abrasive substances.

14.10 CENTRIFUGES
14.10.1 Use of Centrifuges
The use of centrifuges for liquid-particle sep-
Figure 14.41. BHS Fest continuous pressure filter.
(Komline-Sanderson) aration is widespread in the chemical, food
processing, mining, and pharmaceutical indus-
tries. More recently, they are being used more
in pollution control, especially in municipal
filtering, but also as a slurry thickener. It uti- waste water treatment plants. They utilize the
lizes the concept of limiting the growth of the strong G-forces caused by high-speed rotation
filter cake by rotating filter cloth covered discs up to 10,000 rpm. In general, the power needed
adjacent to stationary filter plates on a hori- is proportional to the square of the operating
zontal shaft inside a horizontal vessel. The speed, and the maintenance may even relate
cake thickness may be reduced to 1.0 mm with to the cube of the speed. Larger machines
3 mm clearance. This feature added remark- with higher capacities running at slower speeds
able flexibility to this continuous filter. High can thus show power and maintenance sav-
pressure in the range of 300 psi and thin cakes ings; whereas smaller machines with lower ca-
combine to produce high filter rates.75 pacity can effect higher G-forces for separat-
This filter is also called an Artisan dynamic ing more difficult-to-separate materials. Parti-
filter, a rotary filter press,76 a crossflow filter cle separations can range from 50 to 1000
with rotating elements,77 and an axial filter, micron sizes for perforated basket types, and
developed at Oak Ridge National from 0.5 to 10 microns for disc types. Some
Laboratory.78 An ultrafiltration module has decanter solid bowl models are capable of
also been described based on this principle.79 separating particles from 1 micron up to 1/4
The filter offers automated, continuous op- in. in size. Flow capacities can be up
eration, compact design, a totally enclosed sys- to 1000 gal/min with solids loading up to
tem, clear filtrates, and low operating costs. 100 tons/h.

Wash
Discs Conveyor

Feed

Figure 14.42. Schematic showing continuous KDF filter. (Amafilter)

720 HANDBOOK OF POWDER SCIENCE

14.10.2 Basic Types of Centrifuges

There are two basic types of centrifuges-—
filtering and sedimentation. The first type, as
shown in Figure 14.44, uses a filter cloth or a
screen element for fine particle separation in a
perforated basket, with either vertical or hori-
zontal configuration. The filtrate has low sus-
pended solids and the trapped solids can be
removed manually, or by mechanical devices.
A sedimentation centrifuge, so called because
it greatly accelerates the normal settling rate
of particles by subjecting them to high cen-
trifugal forces, is shown in Figure 14.45. There
is no filter cloth, and the solids are forced up
Figure 14.43. Automatic continuous rotary disc filter. against an imperforate bowl, allowing the liq-
(Ingersoll-Rand, Inc.)
uid to decant off the top. These types are also

HYDRAULIC
UNLOADER
MECHANISM

PATENTED
CENTER SLUNG®
SUSPENSION

FEED
CONE

CARTRIDGE TYPE
BEARING ASSEMBLY
FULLY ISOLATED
ENCLOSED BELT
TUNNEL

Figure 14.44. Filtering basket centrifuge. (Ketama, Inc.)

 FILTRATION OF SOLIDS FROM LIQUID STREAMS 721

Figure 14.45. Typical solid bowl centrifuge. (Bird Machine Corp.)

called solid bowl, screen bowl, decanters, and the cloth, there is no manual operation, and
disc machines. the centrifuge is not opened during discharge.
Solids can be loaded into a suitable container
14.10.2.1 Filtering Centrifuges without being exposed to the environment.
The machine comes in four sizes from drum
Filtering centrifuges, also called basket cen- diameters of 300 to 1000 mm. Throughputs
trifuges, are commonly used in batch feed vary from 100 to 300 kg/h. Up to 90 psig gas
mode in the fine chemical and pharmaceutical pressure can also be added to the bowl, main-
industries for filtering and washing organic taining liquid head and increasing filtration
crystals, inorganic salts, and fine particles. They rates.
are available in sizes up to 40 ft3 cake capac-
ity, and with top or bottom drive. They can be
14.10.2.2 Solid Bowl Centrifuges
automated and solids can be removed mechan-
ically by plow or peeler devices. Stainless steel The increasing use of solid bowl decanter cen-
sanitary and vapor-tight designs are available. trifuges in waste water treatment plants is due
A new design offers ASME code for 35 psi to their good solids dewatering capability of up
steam for sterilization. A recent innovation to 35% cake solids for a mixed feed of primary
has the basket mounted in a horizontal posi- and secondary sludge. They operate continu-
tion with the filter cloth fastened at both ends ously, and because of their enclosed operation,
of a movable drum. At the end of the filtering reduce or eliminate odor problems. They tend
cycle, the drum insert moves axially and hy- to be favored for very large scale plants. For
draulically into a discharge chamber, carefully example, at the second largest municipal waste
turning the filter cloth inside out so that the water dewatering facility in the United States,
solids are then on the outside of the cloth, and seven of these high-solids centrifuges have
can be discharged by the continuing rotation been installed. They are dewatering 350 dry
of the drum. No residual product remains on tons per day. The centrifuges were chosen
722 HANDBOOK OF POWDER SCIENCE

over filter presses and belt filters in the selec- where d is the particle diameter in meters, ^
tion process described by Lipke.80 is in (kg/m • s), Ap is the difference in density
At another large municipal plant in Los between the particle and fluid, g is the gravi-
Angeles, the sludge dewatering process has tation constant (9.81) m/s 2 , and the ratio G/g
been optimized each year to reduce operating is defined by the equation
cost and yield dryer cakes. The three process
variables changed were solids retention time, G/g = Cl2bRb/g (14.12)
hydraulic loading rate, and polymer injection where fl| is the rotation speed of the bowl in
rate. How this was done is explained by radians per second and Rb *is the bowl radius
Zschach.81 in meters. This ratio measures the centrifugal
However, a smaller unit recently introduced acceleration developed in units of gravity. The
is a modular centrifuge especially designed for required rate from Eq. (14.10) can be equated
treatment plants processing up to 5 MGD. with the available rate from Eq. (14.11) and
The unit is compact and can also serve to rearranged to give
thicken and dewater waste streams separately
by being converted from one mode to the Q = 2Vs(lg)(nlRaw/g)(LA/h) (14.13)
other in minutes.82
where Ks(lg) is the settling velocity (Stokes
Norton discusses applications of decanter
velocity) and Rav is the average radius of the
centrifuges in the oil drilling industry to re-
bowl and the pool. This equation shows that
cover barites and control viscosity in drilling
the throughput Q increases with the Stokes
fluids. Recent design changes permit increases
settling velocity, the intensity of centrifugation
in clarification, solids retention time, and
G/g, and the surface area for settling. This
general performance.83
approach to sedimentation centrifuging is from
Bershad et al.84
14.10.2.3 Sizing Sedimentation Centrifuges
Further analysis of batch filtering cen-
The key factors for sizing sedimentation cen- trifuges is given in this chapter considering the
trifuges is the minimum required settling rate following mechanisms of compaction of the
for the solids material if it is not to leave with solids cake:
the overflow. This can be expressed by the
equation: • Centrifugal force acting on the solids (minus
the effect of buoyancy)
^s(req.) = (h/2)/t = \{h/L)(Q/A) (14.10) • Viscous drag on the solids due to liquid flow
• Resistance mechanisms due to the solids
where h/2 is the distance that an average
stress developed as the cake deforms
particle must travel radially while settling, t is
• An arching effect due to the radial
the residence time, L is the distance between
geometry.
the feed inlet and the overflow, Q is the
volumetric throughput, and A is the cross-
For a compressible cake, both permeability
sectional area of the annulus, the liquid pool
and porosity of the cake are functions of the
adjacent to the bowl wall. This suggests that
solids stresses, These can be measured in the
the required settling rate for the average par-
laboratory by using a hydraulic press85 or a
ticle is the throughput divided by the settling
compression-permeability cell.86
surface area, a very familiar result in sedimen-
tation. To determine the rate available, by
14.10.2.4 High-Capacity Oscillating and
using Stokes' law, the settling velocity Vs can
Tumbler Centrifuges
be obtained from the equation
High-capacity oscillating and tumbler cen-
Vs = (14.11) trifuges (up to 250 t / h solids) have relative
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 723

low clarity of overflow. They are used on polishing and solids contaminant removing fil-
rapidly filtered products such as fire-coal, ore, ters, most often small batch quantities which
sand, and coarse salts. To improve their opera- are best done with cartridge and bag filters.
tion, an increase in the residence time would Where laboratory facilities are available,
be beneficial. A recent article explored the much information can be gained by simple
concept of using a step drum in a tumbler Buchner funnel and vacuum leaf tests. The
centrifuge. It was demonstrated that this leads basic lab test for coagulation clarification is
to much greater improvement over conical the jar test, which permits testing a water or
basket machines. The drum should be de- waste water with various coagulant chemicals.
signed with at least three steps for best The CST (capillary suction time) test is used
results.87 to evaluate filter ability of waste sewage sludge.
All these are described in detail with proce-
dures in several texts, the most exacting of
14.11 FILTER EQUIPMENT SELECTION which is Purchas.91 Methods using test results
for scaling filters up to production size are also
With such a wide choice of many various kinds given.
of filtration equipment, it would at first appear Pilot plant and in-plant test are more com-
that choosing the optimum for the specific plete and often more decisive than lab tests.
application would be confusing and frustrat- In my estimation, the optimum program in-
ing. This is not usually the case, however. volves a test filter placed in the production
Many guidelines exist for the initial category plant and set up to filter a side stream from
of choice, and then more specific and well- the existing process. This eliminates any possi-
defined parameters exist for narrowing the ble variations in the liquid slurry that can be
choice to a very few appropriate filters. In fact caused by shipping to a test facility, time fac-
a recent article uses a best and worst choice of tors, or chemical changes in the material. If
factors related on a scale from - 2 to +2, with the product has not yet been made in produc-
the final best choice indicated. There are, tion, the pilot plant approach is desired. One
however, some warnings on borderline cases.88 advantage of the pilot plant is that the test
More recently, a complete software pro- runs can be made on a 24-h basis, giving more
gram has been developed based upon the positive test data than laboratory testing.
above system, but with the added input of After all the above have been done, a care-
practical or heuristic values so that the out- ful evaluation of last minute considerations
come becomes a real workable basis for mak- must be made. Not yet mentioned, but obvi-
ing a very good first selection.89 ously of major concern, is the relative cost of
The study of particle settling rates has been capital equipment and installation, of multiple
the classic approach to initial filter selection. choices if such exists. In some demanding and
A number of tables and guides have been critical choices I have seen a very high price a
published using this technique. A review of secondary factor. In the final analysis, the ulti-
the most important guides was made by mate desired quality of the final product is
Mayer.90 decisive.
The magnitude of the planned operation
easily eliminates many small and batch filters
and indicates continuous belt and rotary vac-
uum filters. Automatic vertical plate filters with REFERENCES
short cycles and decanter type centrifuges pro-
1. C. Jahreis, "Do We Need Filtration?" Filtration
vide volume production. Vacuum pan and table Separation. Cited in Fluid Particle Sep. J 5(l):53-54
filters are economical only with large-scale (March 1992).
operations. At the other end of the scale are 2. N. P. Cheremisinoff and D. S. Azbel, Liquid Filtra-
724 HANDBOOK OF POWDER SCIENCE

tion, pp. 93-97, Ann Arbor Science, Woburn, MA 19. S.-H. Chiang, "Interfacial Phenomena in Fluid Par-
(1983). ticle Separation," Fluid Particle Sep. J., pp. 168-181,
3. Fluid Particle Sep. J., The Hershey Conference on 5-4 (Dec. 1992).
the Pore, pp. 12S-13S, 4-3 (Sept. 1991). 20. E. C. Gregor, C. Wait, and J. R. Mollet, "Consider-
4. F. M. Tiller, "From Pipe to Porous Media Parame- ations in the Selection of Media of Process Appli-
ters," in Conference on the Pore, Hershey, PA, cations," Fluid Particle Sep. J., pp. 163-167, 5-4
May, 1991. Fluid Particle Sep. J., pp. 139-146, 4-2 (Dec. 1992).
(June 1991). 21. E. Mayer and H. S. Lim, "Tyvek for Microfiltration
5. F. M. Tiller, "Tutorial: Interpretation of Filtration Media," Fluid Particle Sep. J., pp. 17-21 2-1 (March
Data I & II," Fluid Particle Sep. J., pp. 85-94, 3-2 1989).
(June 1990) and pp. 157-164, 3-3 (Sept. 1990). 22. L. Bergmann, "Filter Media for Blood Products,"
6. M. Shirato, T. Murase, E. Iritani, and M. Iwata, Tech. Text. Int., pp. 14-16 (Sept. 1992).
"Advanced Studies in Solid/Liquid Separation, Fil- 23. J. Manns, "The Melt Blown Process: Its Role in
tration, Sedimentation and Expression," Fluid Par- Filtration," Nonwouens Industry, pp. 36-40, 22-10
ticle Sep. J., pp. 131-135, 2-3 (Sept. 1989). (Oct. 1991).
7. T. Murase, M. Iwata, T. Adachi, and M. Wakita, 24. D. Kronmiller, "Fouling Control in RO Systems;"
"Stress Relaxation of Expressed Cake," / . Chem. Parekh, B. et al., "New Developments in the Filtra-
Eng. Jpn. 22:655-659 (1989). Cited in Fluid Particle tion of High Purity DI Water for Semiconductor
Sep. J., p. 17D, 4-4 (Dec. 1991). Manufacturing;" Short, J., "Cross Flow Filtration:
8. M. S. Willis, F. N. Desai, G. G. Chase, and I. The Solution is Clear;" and Bartels, C. R., "The
Tosun, "A Continuum Mechanics Analysis of Future for Pervaporation: Technology and Applica-
Darcy's Law," Fluid Particle Sep. J., pp. 137-142, tions," in 10th Annual Membrane Conference, Busi-
3-3 (Sept. 1990). ness Communications Co., Norwalk, CT (1992).
9. M. S. Willis and G. G. Chase, "A Strategy for 25. P. Cartwright, "Membrane Technology for Pollu-
Multiphase Processes," Fluid Particle Sep. J., pp. tion Control, A System Design Primer," FILTECH
55-59, 5-2 (June 1992). 91, Proceedings, pp. 105-125, Filtration Society,
10. F. M. Tiller, "Out of the Past." Fluid Particle Sep. W. Sussex, England (1991).
J. Various issues since 2-1, March 1989. 26. F. E. Duran, "Florida Treatment Plant Uses Low-
11. F. M. Tiller and C. S. Yeh, "Relative Liquid Re- Pressure Membranes," Water Eng. Mgmt., pp. 26-28
moval in Filtration and Expression," in Proceed- (Jan. 1933).
ings American Filtration Society Meeting, AFS, 27. R. F. Toledo, Fundamentals of Food Process Engi-
Kingwood, TX (March 1988). neering, 2nd edit., pp. 533-538. Van Nostrand
12. F. M. Tiller and S. Hsyung, "Improved Formulas Reinhold, New York (1991).
for Constant Pressure Filtration" and "Compaction 28. S. Sternberg, "Membrane Utilization in Hazardous
of Filter Cakes," in American Filtration Society Waste Removal from Wastewater in the Electron-
Meeting, Atlanta, GA, AFS Kingwood, TX ics Industry," Environ. Prog., pp. 139-144, 6-3 (Aug.
(October 1991). 1987).
13. R. I. Dick, S. A. Wells, and B. R. Bierck, "Role of 29. G. Carr, "Continuous Purification of Process
Capillary Forces in Compressible Cake Filtration," Chemicals," Natl. Environ. J., pp. 20-24
Fluid Particle Sep. J., pp. 32-34, 1-1 (Sept. 1988). (Sept./Oct. 1992).
14. P. R. Johnston, "The Micron Rating of a Filter 30. O. C. Carter and B. J. Scheiner, "Removal of Toxic
Medium: A Discussion of the Performance of Fil- Metals from an Industrial Wastewater Using Floc-
ter Media," Fluid Particle Sep. J., pp. 157-161, 2-3 culants," Fluid Particle Sep. J., pp. 193-196, 4-4
(Sept. 1989). (1991).
15. C. J. Williams, "Testing Performance of Spool- 31. E. Mangravite, "Overview of the Use of Poly-
Wound Cartridges," Filtration Separation, pp. 167 ff meric Flocculants and Inorganic Coagulants in Fil-
(Mar./Apr. 1992). tration," Fluid Particle Sep. J., pp. 95-99, 2-2 (1989).
16. B. M. Verdegan, K. McBroom, and L. Liebmann, 32. A. I. Yelshin, "The Development of Filtering
"Recent Developments in Oil Filter Test Methods," Equipment For Hazardous Materials," Fluid Parti-
Filtration Separation, pp. 327-331, 29-4 (Jul./Aug. cle Sep. J. pp. 126-129, 5-3 (Sept. 1992).
1992). 33. Anon., "Self-Cleaning Filters Help Keep Cooling
17. L. A. Scheer, W. C. Steere, and C. M. Geisz, "Tem- Tower Free of Corrosion, Scale Buildup," Plant
perature and Volume Effect on Filter Integrity Eng. 45-16 (Sept. 1991).
Tests," Pharmaceut. Technol., pp. 22-32 (Feb. 1993). 34. P. J. Schonstein, "Horizontal Vacuum Belt Filter
18. J. M. Bentley and P. J. Lloyd, "Interpreting the for Chem. & Mining Industries," in Conference,
Rating of Cartridge Filters," Filtration Separation, Brisbane, Australia 1989. Filtration Separation, 28-2
pp. 333-325, 29-4 (July/Aug. 1992). (Mar./Apr. 1991).
 FILTRATION OF SOLIDS FROM LIQUID STREAMS 725

35. R. W. Mau and G. S. Miller, "Improved Waste 55. B. Culkin, "Vibratory Shear Enhanced Processing:
Water and Sludge Filtration in Fully Automated An Answer to Membrane Fouling," Chem. Proc,
Pressure Filters," Fluid Particle Sep. J., pp. 158-162, pp. 42-44 (Jan. 1991).
5-4 (Dec. 1992). 56. A. Basso, "Vacuum Filtration Using Filter Aids,"
36. N. D. Deutsch, "Options in Belt Filter Presses," Chem. Eng. (April 19, 1982).
Water Eng. Mgmt., pp. 34-37 (Sept. 1987). 57. R. Reiber, "RHA: A Surprisingly Effective Filter
37. W-F. Leung and R. Havrin, "High Solids Decanter Aid," in Proceedings of the American Filtration Soci-
Centrifuge," Fluid Particle Sep. J., pp. 44-48, 5-1 ety, Vol. 6, Chicago, pp. 227-232, AFS, Kingwood,
(March 1992). TX (1992).
38. O. E. Albertson, "Improved Centrifuge Design for 58. R. Rees, "Diatomites Cut Filtration Costs," Pollut.
High Cake Solids," Fluid Particle Sep. J., pp. 28-35,
Eng., pp. 67-70 (April 1990).
3-1 (1990).
59. L. Svarovsky, Solid-Liquid Separation, 3rd edit.,
39. O. E. Albertson, "Mechanical Dewatering Pro-
pp. 334-335. Butterworth-Heinemann, Oxford,
cesses," Fluid Particle Sep. J., pp. 56-70, 3-2 (June
England (1991).
1990).
60. Anon., Everything You Want to Know about Coagu-
40. J. West, "Disc-Bowl Centrifuges," Chem. Eng.,
lation and Flocculation, 3rd edit. Zeta-Meter, Long
pp. 69-73 (Jan. 7, 1985).
Island City, New York (1991).
41. B. Ekberg and J. Haarti, "Capillary Filtration,"
Fluid Particle Sep. J., pp. 116-120, 5-3 (Sept. 1992). 61. E. Mayer, "Membrane Press Sludge Dewatering,"
42. F. Johns, "Tube Press," in FILTECH 1991 Confer- in American Filtration Society Proceedings,
ence, Karlsruhe, Proceedings of Filtration Society, Pittsburgh Conference. AFS, Kingwood, TX (1989).
pp. 183-195, W. Sussex, England (1991). 62. Q. D. Avery, "Membrane Filter Presses," in
43. H. Darcy, "Determination of the Laws of Flow of American Filtration Society Meeting, Chicago, Vol.
Water through Sand," Translated by J. R. Crump, 6, AFS, Kingwood, TX (1992).
Fluid Particle Sep. J., pp. 33-35, 2-1 (March 1989). 63. Q. D. Avery, "Automated System Design for Mem-
44. L. Svarovsky, Solid-Liquid Separation, 3rd edit. brane Filter Presses," in American Filtration Society
Chap. 9, Butterworth-Heinemann, Oxford, England Meeting, Chicago, Vol. 7. AFS, Kingwood, TX (May
(1991). 1993).
45. J. G. Clark, "Select the Right Fabric for Liquid- 64. G. R. S. Smith and C. Rinschler, "Back-Pulse
Solid Separation," Chem. Eng. Prog., pp. 45-50 Liquid Filtration Enhances Tubular Filter Role,"
(Nov. 1990). Chem. Proc, pp. 48-53 (Jan. 1991).
46. L. M. Sheppard, "Corrosion Resistant Ceramics for 65. H. Sandstedt and J. Weisenberger, "Cartridge Fil-
Severe Environments," Am. Ceram. Soc. Bull. pp. ter Performance and Micron Rating," Filtration
1146-1158,70-7(1991). Separation (Mar./Apr. 1985).
47. E. C. Gregor, C. Wait, and J. R. Mollet, "Filtration 66. C. Juhasz, "Total Filter Performance," in Proceed-
and Separation Considerations in the Selection of ings American Filtration Society, Ocean City, MD,
Media for Process Applications," Fluid Particle Sep. AFS, Kingwood, TX (March 1988).
I., pp. 163-167, 5-4 (Dec. 1992). 67. A. C. Wrotnowski, "Final Filtration with Felt Bag
48. A. Pangrazi, "Nonwoven Bonding Technologies," Strainers," Chem. Eng. Prog., p. 89 ff (Oct. 1978).
Nonwovens Industry, p. 32 ff, 23-10 (Oct. 1992).
68. T. W. Johnson, "Large Surface Area in Small Pack-
49. W. Shoemaker, "Nonwoven Technology in Filter
age," in American Filtration Society Proceedings,
Media," Nonwovens Industry, p. 34 ff (Oct. 1984).
Ocean City, MD, pp. 671-678. AFS, Kingwood, TX
50. G. R. S. Smith, "Membrane Filter Cloth," in
(1988).
American Filtration Society Conference Proceedings,
69. G. R. S. Smith, "How to Use Rotary, Vacuum,
pp. 665-669, AFS, Kingwood, TX (1988).
Precoat Filters," Chem. Eng. 55:84-90 (Feb. 16,
51. M. Porter, "Membrane Filtration," Sect. 2.1,
1976).
pp. 2-103, in Handbook of Separation Techniques
for Chemical Engineers, edited by P. A. Schweitzer, 70. Anon., Case Study, Chem. Equip., p. 29 (June 1992).
McGraw-Hill, New York (1979). 71. S. A. Wells, "Two-Dimensional, Steady-State Mod-
52. P. R. Johnston, Fluid Sterilization by Filtration, eling of Compressible Cake Filtration in a Later-
Interpharm Press, Buffalo Grove, IL (1992). ally Unconfined Domain," Fluid Particle Sep. J.,
53. G. B. Van den Berg and C. A. Smolden, "Flux pp. 107-116, 4-2 (June 1991).
Decline in Membrane Filters," Filtration Separa- 72. R. W. Lecey and K. A. Pietila, "Improving Belt
tion, pp. 115-121, 25-2 (Mar./Apr. 1988). Filter-Press Performance," Chem. Eng. (Nov. 28,
54. D. C. Gyure, "Set Realistic Goals for Crossflow 1983).
Filtration," Chem. Eng. Prog., pp. 60-66, 88-11 73. N. D. Deutsch, "Options in Belt Filter Presses,"
(Nov. 1992). Water Eng. Mgmt., pp. 34-37 (Sept. 1987).
726 HANDBOOK OF POWDER SCIENCE

74. H. H. Kleizen and M. Dosoudil, "Continuous Pres- 87. F. Deshun and R. J. Wakeman, "Effects of Step
sure Filtration: From Theory to KDF," in Proceed- Drums on Solids Residence Times in Conical
ings of the American Filtration Society Conference, Basket and Tumbler Centrifuges," Filtration Sepa-
Ocean City, MD. AFS, Kingwood, TX (1988). ration, 29-2 (Mar./Apr. 1992).
75. A. Bagdasarian and F. M. Tiller, "Operational Fea- 88. M. Ernst, R. M. Talcott, H. C. Romans, and
tures of Staged, High-Pressure, Thin-Cake Filters," G. R. S. Smith, "Tackle Solid-Liquid Separation
Filtration Separation p. 594-598 (Nov./Dec. 1978). Problems," Chem. Eng. Prog., pp. 22-28, 87-7 (July
76. T. Toda, "Recent Advances in the Application of 1991).
the Rotary Filter Press," Filtration Separation, 89. R. J. Wakeman and E. S. Tarlton, "Solid/Liquid
pp. 118-122 (Mar./Apr. 1981). Separation Equipment Simulation and Design—An
77. L. Svarovsky, Solid-Liquid Separation, 3rd edit., Expert Systems Approach," Filtration Separation,
p. 582, Butterworth-Heinemann, Oxford, England 28-4 (May/June 1991).
(1991). 90. E. Mayer, "Solid/Liquid Separation—Selection
78. I. I. Irizarry and D. B. Anthony, Ornl-Mit-129, Oak Techniques," Fluid Particle Sep. J., pp. 129-139, 1-2
Ridge National Laboratory, 28 (April, 1971). (Dec. 1988).
79. B. Hallstrom and M. Lopez-Leiva, "Description of 91. D. Purchas (ed.), Solid/Liquid Separation Equip-
a Rotating Ultra-Filtration Module," Desalination ment Scale-Up, Uplands Press, Croydon, U.K.
24:213-219 (1978). (1986).
80. S. Lipke, "High-Solids Centrifuges Turn Out to be
92. L. P. Raman, M. Cheryan, and N. Rajagopalan,
Surprise Dewatering Choice," pp. 22-24 Water Eng.
"Consider Nanofiltration for Membrane Separa-
Mgmt. (June 1990).
tions," Chem. Eng. Prog., pp. 68-74, 90/3 (March
81. A. Zschach et al., "Hyperion's Recipe for Dry
1994).
Cake," Operations Forum, pp. 16-19, 9-8 (Aug.
93. M. Morgenthaler, "Understanding Decanting Cen-
1992).
trifuges and their Environmental Applications," in
82. G. S. Sadowski, "Modular Centrifuges," Paper pre-
American Filtration Society Workshop, University of
sented to the Texas Water Pollution Control Asso-
Houston, Houston, TX (Jan. 4, 1994).
ciation (June 5, 1992).
83. V. K. Norton, "Centrifuges for Solids Control," 94. T. De Loggio and A. Letki, "New Directions in
Fluid Particle Sep. /., pp. 180-181, 3-4 (Dec. 1990). Centrifuging," Chem. Eng., pp. 70-76 (Jan. 1994).
84. B. C. Bershad, R. M. Chaffiotte, and W-F. Leung, 95. P. Cartwright, "Membranes Meet New Environ-
"Making Centrifugation Work for You," Chem. mental Challenges," Chem. Eng., pp. 84-87 (Sept.
Eng., pp. 84-89 (August, 1990). 1994).
85. F. M. Tiller and C. S. Yeh, "Relative Liquid Re- 96. F. W. Schenck and R. E. Hebeda, "Starch Hydroly-
moval in Filtration and Expression," Filtration Sep- sis Products" p. 495, VCH Publishers, New York,
aration, 17-2 (1990). 1992.
86. H. P. Grace, "Resistance and Compressibility of 97. Survey Report #282. Membrane Separation Equip-
Filter Cakes," Chem. Eng. Prog. Parts 1 and 2 ment, 1995. Future Technology Surveys Inc. Nor-
(1953). cross, GA (see pg. 700 of this document).
 15
Cyclones
David Lelth and Donna Lee Jones

CONTENTS

15.1 INTRODUCTION 727

15.2 PERFORMANCE CHARACTERISTICS 728
15.3 PERFORMANCE MODELING 731
15.4 CYCLONE DESIGN 743
REFERENCES 751

15.1 INTRODUCTION high dust loadings, and can handle large gas
flows. Although standard cyclone designs are
A cyclone is a device without moving parts inefficient for collecting particles smaller than
that spins a gas stream to remove entrained about 5 microns, high-efficiency cyclones used
particles by centrifugal force. Cyclones are alone or in series can collect particles between
simple and inexpensive to make, relatively 2 and 5 microns. The standard design cyclones
economical to operate, and are adaptable to a are probably the most frequently used dust
wide range of operating conditions. Cyclones collectors in industry.
have been used throughout industry since the The first published efforts to predict cyclone
1880s for the removal of dust from gases.1 By performance did not appear until about 1930.
the turn of the 20th century, they were used to Extensive studies of the gas flow pattern in
collect sawdust and shavings in woodworking cyclones made during the 1940s led to the
shops. Ten years later, cyclones began to con- development of many models for predicting
trol dust from cement kilns. Shortly thereafter, cyclone pressure drop and dust collection ef-
they were first used to remove fly ash from ficiency; efforts at modeling cyclone perfor-
flue gas. mance have continued to the present. Al-
Cyclones can be made to withstand extreme though our knowledge of what goes on inside
temperatures and pressures, can accommodate a cyclone has increased over the years, the
727
728 HANDBOOK OF POWDER SCIENCE

basic cyclone design shown in a 1885 German

patent (No. 39,219) looks a good deal like a
cyclone that might be used today.

15.2 PERFORMANCE
CHARACTERISTICS
15.2.1 Types
Over the years, many different types of cy-
clones have been built. However, the
"reverse-flow," or cone-under-cylinder design
shown in Figure 15.1, is the type used most
often for industrial gas cleaning. In this de- Figure 15.2. Cyclone entries. (^4) Tangential; (B) swirl
sign, aerosol enters the cyclone at the cylinder vane; (C) half scroll; (D) full scroll.
top, where the shape of the entry causes the
gas to spin. Tangential, scroll, and swirl vane
axis, and from here up, out the gas exit duct.
entries have been used as shown in Figure
Collected dust descends the cyclone walls to
15.2; tangential entries are most common. Af-
the dust outlet at the bottom of the cone,
ter entering the cyclone, the gas forms a vor-
primarily due to the downward component of
tex with a high tangential velocity which gives
the gas velocity at the cyclone wall rather than
particles entrained in the gas a high centrifu-
due to gravity.
gal force, throwing them to the cyclone wall
Figure 15.3 shows a "straight-through" cy-
for collection. Below the bottom of the gas
clone. Here, dusty gas enters at one end while
exit duct, the spinning gas gradually migrates
cleaned gas and separated dust exit separately
inward, to a "central core" along the cyclone
at the opposite end. Again, the entry shape
causes the inlet gas to spin. Swirl vane entries
are used most often on straight-through cy-
clones. Within the cyclone, centrifugal force
pushes particles to the wall. Cleaned gas leaves
from a central exit duct while separated parti-
cles flow out with a small amount of purge gas
through the annular dust discharge. The re-
mainder of this chapter is devoted to the
INLET reverse-flow cyclone, as it is the type used by
far the most often for industrial dust control.
Particle collection theory for straight-
through cyclones3 is not as well developed as
that for conventional reverse-flow cyclones.

15.2.2 Standard Designs

Figure 15.4 shows the eight dimensions of a
reverse-flow cyclone. It is convenient to ex-
press cyclone dimensions as multiples of diam-
eter, D. The "diameter ratios," a/D, b/D,
De/D, S/D, h/D, H/D, and B/D, allow
Figure 15.1. Reverse-flow cyclone. comparing the shape of two or more cyclones
 CYCLONES 729

GAS
INLET OUTLET

DUST
SWIRL OUTLET
VANES
Figure 15.3. Straight-through cyclone.

that might differ in size. Several sets of dimen- The high-efficiency and general-purpose stan-
sion ratios, or "standard designs," are given in dard designs have tangential gas entries
Table 15.1. whereas the high-throughput designs have
A comparison of the designs in Table 15.1 scroll entries. High efficiency can be traded
reveals that cyclone shape varies with recom- against high throughput for cyclones operating
mended duty. A high-efficiency cyclone has a at the same pressure drop.
smaller inlet area (a/D and b/D) and exit Because cyclone design changes with rec-
area (De/D) than does a high-throughput cy- ommended duty, no single optimum cyclone
clone. Gas outlet length (S/D) is less in the design exists that will work best for all possible
high-efficiency designs, probably because inlet applications. Design of a cyclone appropriate
height (a/D) is less. Outlet length should be for a particular task involves compromises
greater than inlet height to be sure that a among a number of cyclones, throughput per
stable vortex is formed within the cyclone body. cyclone, pressure drop and efficiency.

15.2.3 Application Areas

Standard design cyclones collect particles
larger than 5 to 10 microns in diameter with
reasonable efficiency. For smaller particles,
efficiency declines rapidly. With customized
high-efficiency cyclones, particles less than 5
microns can be collected, although collection
is limited to 1 micron particles or greater. For
dust streams with particles larger than several
hundred microns, a settling chamber can often
by used with lower installation and operating
costs than a cyclone system.
Cyclone pressure drop is similar to that
SECTION found in other particle collection devices, ex-
A-A cept for high-energy scrubbers which can re-
quire much higher pressure drops. For cy-
clones, there generally is a trade-off between
efficiency and pressure drop, with higher pres-
sure drops associated with higher efficiency
and vice versa. Values for cyclone pressure
drop range from about 0.1 to 2 kPa (0.5 to 8
Figure 15.4. Dimensions of a reverse-flow cyclone. in. of water).
730 HANDBOOK OF POWDER SCIENCE

Table 15.1. Standard Designs for Reverse-Flow Cyclones.

RECOMMENDED
SOURCE DUTY D a/D b/D De/D S/D h/D H/D B/D AH Q/D2 (m/h)
Stairmand 4 High-efficiency 1 0.5 0.2 0.5 0.5 1.5 4.0 0.375 6.4 5500
Swift5 High-efficiency 1 0.44 0.21 0.4 0.5 1.4 3.9 0.4 9.2 4940
Lapple 6 ' 7 General-purpose 1 0.5 0.25 0.5 0.625 2.0 4.0 0.25 8.0 6860
Swift5 General-purpose 1 0.5 0.25 0.5 0.6 1.75 3.75 0.4 7.6 6680
Stairmand 4 High-throughput 0 1 0.75 0.375 0.75 0.875 1.5 4.0 0.375 7.2 16,500
Swift5 High-throughput 0 1 0.8 0.35 0.75 0.85 1.7 3.7 0.4 7.0 12,500
a
Scroll type gas entry used.

A properly designed cyclone can process housed together and operating in parallel,
effectively dusts in very high concentrations, sometimes called multiclones, are frequently
and in practice loadings of over 2000 g/m 3 connected to the same dust bin without valves
(1000 gr/ft 3 ) have been accommodated.4 Cy- on the dust discharge of each unit. Unequal
clones have the fortunate ability simultane- inlet pressure distribution across the inlet and
ously to increase efficiency8"11 and decrease exhaust manifolds may cause gas to flow out
pressure drop12"14 with an increase in dust the exit duct of some cyclones, through the
loading. This may come about due to the dust bin, then up and out through the dust exit
increased number of particles that move radi- and gas exit of other units. This flow pattern
ally outward through the cyclone vortex when will adversely affect the performance of the
dust loading increases. This movement might cyclone system. The performance of multi-
hinder the formation of the vortex and thereby clones is almost never as good as that of each
decrease pressure drop. Despite vortex sup- small cyclone operating individually. However,
pression, efficiency might increase owing to multiclone performance should be better than
the increased opportunity for larger particles that of a single large-diameter cyclone operat-
to strike and collect smaller particles while ing at the same pressure drop and handling
they move toward the cyclone wall. the same gas flowrate as the manifolded de-
Cyclones are available in many sizes, and sign. The small-diameter cyclone, manifolded
can be made from materials able to withstand design does offer the advantage of compact
extreme operating conditions. They are com- installation.
mercially available in sizes to process 50 to Industrial processes use cyclones for un-
50,000 m 3 /h. Although smaller diameter units loading material from process gas streams, and
generally are more efficient,4'15'16 a manifold for controlling particulate emissions to the
must be used to connect many small cyclones atmosphere. The dry product collected in a
together to process a large gas flow. Refrac- cyclone can often be recycled to the plant for
tory lined cyclones have been operated at tem- further processing. Among the processes using
peratures of 1000°C, while other units have cyclones are coal driers, grain elevators, grain
run at pressures of several hundred atmo- driers and milling operations, sawmills and
spheres.17 However, special materials of con- wood-working shops, asphalt plant rotary rock
struction chosen to allow operation under ex- driers, and detergent manufacturing processes.
treme conditions may not always have good Teams of cyclones operating in series are used
resistance to erosion of the cyclone walls by under high temperature and pressure condi-
collected dust. Sticky, hygroscopic dusts may tions to collect catalyst dust from catalytic
not discharge readily through the dust outlet18 cracking units at oil refineries and to collect
and these dusts may be better suited to collec- fly ash generated from coal combustion in a
tion in a scrubber. Small-diameter cyclones pressurized fluidized bed.
 CYCLONES 731

15.3 PERFORMANCE MODELING He gives another expression to allow adjust-

ment of the vortex exponent for gas tempera-
15.3.1 Flow Pattern
ture variations.
To understand pressure loss and particle col-
lection in a cyclone, it is important to under- 0.3
stand the cyclone gas flow pattern. Cyclone (15.3)
flow pattern has been studied in some detail;
the overall trend of gas motion has been gen-
erally confirmed by all workers. However, no Figure 15.5 shows the relationship between
generalized model of the flow pattern is avail- vortex exponent, n, cyclone diameter, D, and
able that will allow the prediction of all veloc- gas temperature as given by Eqs. (15.2) and
ity components at any point in the cyclone. (15.3).
Each experimenter has concentrated on at If Equation (15.1) is applied throughout the
most several cyclones of similar design, none cyclone, tangential gas velocity would increase
of which is particularly similar to the units with decreasing radius to infinity at the cy-
studied by others. Although each set of results
clone axis. Actually, within the portion of the
is internally consistent, ambiguities arise when
cyclone above the bottom of the gas outlet
comparing the trends reported in several stud-
duct, tangential velocity is limited by the gas
ies. The most complete review of the gas flow
outlet wall. At the cyclone body wall, and at
patterns in cyclones is probably that by
the gas exit wall, the tangential velocity falls
Jackson.1
off rapidly. If the gas exit is large relative to
body diameter, wall effects may hinder the
15.3.1.1. Tangential Gas Velocity
formation of the vortex to the point that tan-
After entering the cyclone, the gas stream gential velocity becomes almost constant
forms a confined vortex such that the tangen- rather than increasing with decreasing radius
tial gas velocity, vt, is related to the distance, as is normally the case.13
r, from the cyclone axis by Eq. (15.1): Below the gas exit, tangential velocity in-
creases with decreasing radius, up to a maxi-
utrn = constant (15.1)
mum at a point defined as the radius of the
The vortex exponent, n, is +1 for an ideal cyclone central core. Within this core, tangen-
fluid, 0 for velocity which is constant regard- tial gas velocity decreases with decreasing ra-
less of radial position, and - 1 for rotation as a dius, and falls ultimately to a value near zero
solid body. While theoretical descriptions of at the cyclone axis. Within the core, Eq. (15.1)
cyclone flow patterns 16 ' 19 ' 20 employ values of still describes the tangential gas velocity but
n from +1 to - 1 , data13'21"23 indicate that the with a revised value for the vortex exponent
usual range is from 0.5 to 0.9, with most values ranging from about -0.5 to less than -2. 2 3
around 0.5. Tangential gas velocity, therefore, The diameter of the central core ranges from
increases from a minimum at the cyclone wall about 0.3 to 1.0 times the gas outlet
to higher values approaching the cyclone axis. diameter.4'25
Tangential velocities may be lower than the
Jackson1 believed that the core diameter
gas inlet velocity at the cyclone wall, but can
decreases from a value close to that of the gas
exceed the inlet velocity by several times at
exit diameter directly below the gas exit to a
some distance from the wall. Alexander24 has
value of about a third of the gas exit diameter
given an empirical expression for dependence
near the bottom of the cyclone. Iozia (Jones)
of vortex exponent on cyclone diameter in
and Leith26 found that the core diameter is
meters, D m , at temperature of 283 K.
relatively constant throughout the cyclone, giv-
n = 0.67Z)°14 (15.2) ing a cylindrical shape to the core region.
732 HANDBOOK OF POWDER SCIENCE

0.1
0.05 0.1 0.2 0.4 0.6 1.0 2 3
CYCLONE DIAMETER, METERS
Figure 15.5. Vortex exponent n as a function of cyclone diameter and gas temperature, according to Eqs. (15.2)
and (15.3).

Iozia (Jones) and Leith26 found that the but is limited by the size of the core diameter
core diameter can be estimated from the cy- relative to the dust outlet diameter.26 Figure
clone inlet and outlet dimensions. The core 15.6 presents their anemometer measurements
length is calculated from geometry using the of tangential velocity at different positions in a
core diameter and other cyclone dimensions, reverse-flow cyclone with tangential entry.26

Figure 15.6. Tangential gas velocity in reverse-flow cyclone.

 CYCLONES 733

15.3.1.2 Vertical and Radial Gas Velocity self-consistent, as radial velocity must be
greater near the central core from conserva-
In general, the gas within the cyclone flows
tion of mass principles. Radial gas velocity is
downward near the cyclone wall and upward
the most difficult velocity component to mea-
near the cyclone axis; these vertical velocities,
sure experimentally. Still, knowledge of this
both downward and upward, are much less
component is essential for determining parti-
than tangential gas velocities. The radial posi-
cle collection efficiency through the "static
tion at which vertical velocity changes from
particle" approach described below. Lack of
down to up is relatively closer to the cyclone
data on this point has led to unproven specu-
wall at the top of the unit than at the base of
lation on the variability of the radial velocity
the cone. At all vertical locations, the velocity
which is used to explain the inadequacies of
changeover point appears to be outside the
this efficiency theory. Figure 15.8 shows ter
central core. However, once within the core
Linden's measurements of radial gas velocity
the upward gas velocity increases substantially.
at different vertical positions in a cyclone.22
Figure 15.7 shows measurements of vertical
gas velocity made by ter Linden at different
15.3.1.3 Pressure Distribution
positions in a reverse-flow cyclone with scroll
entry.22 The total pressure at any point in a cyclone is
The radial component of gas velocity has the sum of the static pressure and velocity
not been measured as extensively as have the pressure at that point. Total pressure slowly
tangential and vertical components. Data show decreases from a maximum value at the cy-
that inward radial velocity is low, constant with clone wall to a minimum value near the cy-
radial position, and approximately equal at all clone axis. With the high tangential gas veloci-
vertical positions within the cyclone below the ties present in a cyclone, velocity pressures
gas outlet duct. However, these data are not can be so high that static pressure becomes
negative relative to the atmosphere. The static

GAS
OUTLET
GAS
OUTLET

INLET INLET

L
\ 7

VELOCITY VELOCITY
M/SEC M/SEC

DUST DUST
OUTLET OUTLET
Figure 15.7. Vertical gas velocities in a reverse-flow Figure 15.8. Radial gas velocities in a reverse-flow
cyclone. cyclone.
734 HANDBOOK OF POWDER SCIENCE

pressure within the central core can be nega- First23 showed that after the gas has en-
tive, even when the cyclone is installed on the tered the cyclone and makes one full revolu-
discharge side of a fan. The zone of negative tion, it is not entirely displaced downward by
static pressure can extend from the core gas entering subsequently. Some older gas is
through the dust outlet and if a suitable valve forced toward the cyclone axis in an inward
is not used at the dust outlet, into the dust spiral, a phenomenon First calls "lapping." As
collection bin. If no valve or a leaky valve is the newer gas squeezes the older gas toward
fitted and the dust collection bin is not air- the cyclone axis, the tangential velocity of the
tight, dusty air from the bin will be drawn into older gas increases through conservation of
the central core, up and out of the cyclone. momentum.
For this reason the cyclone dust hopper should
always be airtight, ter Linden's measurements 15.3.2 Pressure Drop
of static and total pressure22 in a cyclone are Factors13 that contribute to cyclone pressure
shown in Figure 15.9. drop, static pressure differential across the
cyclone, include:
15.3.1.4 Overall Gas Flow Pattern
1. Gas expansion as it enters the cyclone
As gas enters the cyclone, it forms a vortex in 2. Formation of the vortex
the annulus above the gas outlet duct. Below 3. Wall friction
the gas outlet, spinning gas gradually migrates 4. Regain of the rotational kinetic energy as
into the central core. Near the cyclone walls, pressure energy.
gas flows downward, whereas gas closer to and
within the central core flows upward toward The first three factors are probably the most
the gas outlet duct. At the narrow end of the important. Controversy exists over the impor-
cyclone, all the gas flows into the central core. tance of wall friction on pressure drop, as
Iinoya14 has shown that sand glued onto the
cyclone walls, increasing wall roughness, actu-
ally decreases the pressure drop. If this is
GAS correct, then energy consumption due to vor-
OUTLET
tex formation plays a greater role in pressure
INLET drop than does wall friction. First 23 also found
that wall friction makes an insignificant contri-
bution to overall pressure drop.
Devices such as an inlet vane, an extension
of the inner wall of the tangential gas entry
within the cyclone body up to a position close
to the gas exit duct, and a cross baffle in the
gas outlet duct will lower pressure drop. How-
ever, these devices probably suppress vortex
,0 formation,1'18'27 and so decrease efficiency as
well as pressure drop. Because a cyclone is a
ho device for vortex generation, it is not logical to
PRESSURE — STATIC put attachments within it that inhibit vortex
CM WATER — TOTAL
formation. Cyclones can be designed for low
pressure drop without resorting to internal
OUTLET attachments that may impair efficiency.
Figure 15.9. Static and total pressures in a reverse-flow Many investigators have developed expres-
cyclone. sions to predict pressure drop; some are em-
 CYCLONES 735

pirical, some theoretical, and most a mixture collected by the cyclone. Experience in dealing
of both. Despite the complexity of some pres- with cyclones has shown that collection effi-
sure drop relationships, no single expression ciency increases with:
has been developed that will give a reliable
estimate of pressure drop for all cyclones 1. Increasing particle diameter and density
operating under all conditions. 2. Increasing gas inlet velocity
Cyclone pressure loss is expressed most 3. Decreasing cyclone diameter
conveniently as a number of inlet velocity 4. Increasing cyclone length
heads, AH. Velocity heads can be converted 5. Drawing some of the gas from the cyclone
to loss in pressure units, AP, by Eq. (15.4): through the dust exit duct
AP = AH(\pGvf) (15.4) 6. Wetting the cyclone walls
The number of inlet velocity heads, AH, will
be constant for any cyclone design although A plot of collection efficiency against parti-
the pressure loss, AP, varies with different cle diameter is called a fractional efficiency
operating conditions. Pressure drop for a cy- curve or grade efficiency curve. A typical frac-
clone can best be established by determining tional efficiency curve for a cyclone is shown
AH experimentally for a particular cyclone in Figure 15.10. Fractional efficiency rises
design. The static pressure loss, AP, for geo- rapidly at first, then flattens out and ap-
metrically similar cyclones can then be found proaches unity for very large particles.
from Eq. (15.4) for different operating condi- Particles are separated from the gas stream
tions. Values of AH are listed in Table 15.1 in a cyclone by spinning to the cyclone wall
for the standard design cyclones listed there. through centrifugal force. Figure 15.11 shows
Many analytical expressions for determining the forces acting on a particle rotating with
AH from cyclone geometry have been pre- tangential velocity ut at radial position r. The
sented in the literature. Several are listed in particle moves radially outward with velocity
Table 15.2. One review30 found that the uT. The tangential velocity of the gas and that
Barth,25 Stairmand,29 and Shepherd and of the particle will be assumed equal, ut = uv
Lapple28 equations work better than those This is probably a reasonable assumption for
by Alexander24 and First.23 The Barth and small particles, for which efficiency is most
Stairmand approaches are complex and re- difficult to determine.
quire knowledge of all cyclone dimensions. The centrifugal force, Fc, acting on the
The Shepherd and Lapple approach, Eq. (15.5), particle is given in Eq. (15.18):
is simpler to use, and while it does not include
all cyclone dimensions it nevertheless gives
results about as good as those produced by the F = (15.18)
6r
more complex calculation methods.
ab The drag force, Fd, acting on the particle as it
AH =16—T (15.5)
moves rapidly outward can be given by Stokes'
Values of cyclone pressure drop calculated law; for larger particles with higher radial ve-
from theory may give results in error by 50% locities Stokes' law becomes a progressively
or more. There is currently no alternative to poorer approximation.
experimental testing when cyclone pressure
drop must be known accurately. Fd = ur - vT) (15.19)
15.3.3 Efficiency
Equation (15.1), which describes gas tangential
Collection efficiency, rj, is defined as that frac- velocity as a function of radial position, gives
tion of particles of a certain size that are tangential velocity at position r as a function
736 HANDBOOK OF POWDER SCIENCE

Table 15.2. Equations for Predicting Pressure Loss at Number of Inlet Velocity Heads, AH.

SOURCE PRESSURE LOSS EQUATION

ab
Shepherd and Lapple 2 AH = 16- 1 (15.5)
De

ab 24
First 23 AH = 1/3
(15.6)
(h(H-h)\
2
[ D )
Alexander 24 (15.7)
14
n = 0.67 D^ at 283 K (15.2)
i _ (M* (15.3)
2 \T2)

(15.8)

J2(D-b) \ / 4ab
Stairmand 29 AH = 1 + 2cf>2\ - 1 +2 2r (15.9)
\ De / \ irDe
De De 4G*s
2(D-b) ab
(15.10)
2G*A
ab
2 2
A = —(D - De ) + irDh + irDeS
4 1/2
TT
B)\(H-hr + (15.11)

G* = 0.005 (15.12)

/ AabB \
Barth 25 AH= r e (15.13)
\ irDe2 )
De
1 4, 2/3
- l ) + 4Ad~ (15.14)
- 2d(H - S)(X/De)r
irDe(D - b)
6= (15.15)
4aba* + 2(H - S)(D - b)irk
1.2b
„.-!-_ (15.16)
A = 0.02 (15.17)

of cyclone wall radius, rw, and the tangential describes the tangential velocity near the wall
velocity at the wall, v^. with little error. The sum of the centrifugal
and drag forces acting on the particle will
Vtrn = constant = v ^ (15.1) equal its mass times its acceleration.

Strictly speaking, the gas tangential velocity at d2r

the cyclone wall must be zero. However, the 37Tfld(uT — Vr)
boundary layer at the cyclone wall is thin, as 6r
can be seen in Figure 15.6; and Eq. (15.1) (15.20)
 CYCLONES 737

1 i 1 | 1 1 1
1.0 —

|oB
UJ
o /
it 0.6
UJ
~ /
<0.4
o " /
o /
<0.2 —
a:
u. iI
A i 1 i 1 1 1 | 1 1
10 20 30 40
PARTICLE DIAMETER, MICROMETERS
Figure 15.10. Typical cyclone fractional efficiency curve.

Simplifying and making the substitutions ur The relative importance of each term will
dr/dt and vt = vtwr^f/rn yields Eq. (15.21): change with each cyclone design and particle
diameter. It is unlikely that any approximation
d2r 18/x, dr will yield good results for all applications.
'dt1 The theoretical efficiency of a cyclone can
be characterized in terms of a "critical parti-
cle" diameter, d100. The critical particle is that
Equation (15.21) describes radial particle which, according to theory, is collected with
motion within a vortex and underlies many 100% efficiency. Since collection efficiency in-
approaches used to calculate cyclone collec- creases gradually with increasing particle
tion efficiency. Unfortunately, Eq. (15.21) has diameter and approaches 100% only as a
not been solved analytically. Approximate so- limit, the critical particle is not observed
lutions can be found by postulating various experimentally.
flow conditions within the cyclone, allowing A more easily verified theoretical construct
deletion of some terms in the equation. All is the "cut diameter" or d50, the particle size
these approximations are open to criticism. that is collected with 50% efficiency. Calcula-
tions of critical or cut diameter can be used to
generate the cyclone fractional efficiency curve
shown in Figure 15.10.

15.3.3.1 Critical Diameter: The Static

Particle Approach
Gas at the edge of the central core will have
maximum tangential velocity, vmaK. Gas flow-
ing radially inward to the cyclone central core
will flow past the core edge with an average
inward radial velocity given by Eq. (15.22):
CYCLONE uf
AXIS
Figure 15.11. Forces on a particle in a cyclone.
738 HANDBOOK OF POWDER SCIENCE

For particles of the critical diameter, the in- Core length, used in Eq. (15.26), depends on
ward drag force caused by the inrushing gas the value of the core diameter dc.
will just balance the outward centrifugal force
caused by their rapid rotation about the cy- -0.25
clone axis. These "static particles" will theo- dr = 0.52£»| —j
retically remain suspended at the edge of the
central core. Larger particles should spin out •De^1-53
to the cyclone wall and become collected, and (15.28)
smaller particles should flow past the static
particles into the central core and out the
cyclone. As they are stationary, the critical When dc > B, the core intercepts the cyclone
particles will have no radial acceleration or walls and the core length is calculated from
velocity (d2r/dt2 = dr/dt = 0). From Eq. geometry.
(15.1), v^r2n = ^maxrccTre> which when sub-
stituted into Eq. (15.21) yields the critical zc = (H - S) - ((H - h)/(D/B - 1))
particle diameter.
X«dc/B)-1) (15.29)
1/2

(15.23) When dc < B, the core length extends to the

" 1 0 0 ~~
ir(H - S)PJU
p max bottom of the cyclone.

According to static particle theory, "frac- z=H - S (15.30)

tional degree of dust separation in the case of
a (critically sized) particle suddenly increases
from 0 to 100%."25 The shape of experimental 15.3.3.2 Critical Diameter: The Timed
fractional efficiency curves, which show a Flight Approach
gradual increase in efficiency with increasing
particle diameter, is explained as being due to The timed flight approach is another way to
variations in the gas inward drift velocity, vY, calculate critical diameter, and involves a dif-
along the cyclone axis. ferent set of assumptions for solving Eq.
Barth25 and Stairmand4 used different as- (15.21). Let r{ be the innermost radial position
sumptions in the static particle approach to at which particles enter the cyclone. Particles
develop equations for critical diameter. Iozia entering at this point must cross the distance
(Jones) and Leith26 used experimental data to from r{ to the cyclone wall to be collected, and
develop an equation to predict cut diameter if a particle is not in the cyclone long enough
using the static particle approach. These equa- to travel this distance it will escape collection.
tions are shown in Table 15.3. In the timed flight approach, the particle's
radial acceleration and the gas radial velocity
Iozia (Jones) and Leith26 predicted maxi-
are arbitrarily set equal to zero and neglected
mum tangential velocity from inlet velocity and
(d2r/dt2 = uT = 0). The gas velocity in the
a dimensionless geometry parameter.
entrance duct is assumed equal to the velocity
at the cyclone wall, vT = t;tw. These assump-
tions allow solution of Eq. (15.21) for particle
(15.26) radial position as a function of particle resi-
if) (D
dence time within the cyclone, t. The particle
Fmax = 6.1FA (15.27) with critical diameter will travel from its initial
 CYCLONES 739

Table 15.3. Equations Derived from Eq. (15.20) for Predicting Coiiection Characteristics.

d2r 18/JL dr ( vLr2n lSfxvA

dt 2 + 2
dt
l tw w ( A~ r l -o (15.21)
d pp

ASSUMPTIONS
STATIC PARTICLE d2r dr
APPROACH '"core «W RESULTANT EQUATION
~dS It
1/2

Barth 25 0 De/2 (15.23)

irDe2
1/2
1/2

Stairmand4 0 De/4
90/*
(15.24)
ab De
Iozia (Jones) 26 1/2

and Leith 0 dc/2 6.1 V{ - C,

4 (15.25)

ASSUMPTIONS
TIMED FLIGHT d2r
APPROACH n t RESULTANT EQUATION

1/2
20
D TTDN
Rosin et al. 0 0 b 0 (15.32)
2 v{
1/2
D - De
Lapple and Shepherd- 0 0 - j - 0 (15.33)

1/2

Davies33 0 0
De
1
H
—
V;
-(fj (15.34)

De\
i*f
D - b TTDN ( De
Lapple 6 0 0 — — 0 (15.35)
2 v{ ID J
34 1/(2 + 2)
Leith and Licht 0 0 0 n depends on T) = 1 - exp(-2(C L i/O " ) (15.36)
geometry and
throughput

position at r{ and just reach the cyclone wall t, is sometimes defined in terms of an empiri-
in time t. cal "number of turns," N, that the gas stream
1/2 makes within the cyclone. The value for TV
2n + 2
reportedly varies from 0.3 to 10, with a mean
value of about 5.31 Table 15.3 gives several
(15.31)
sets of assumptions for rv n, and t along with
the resultant equations for either critical or
cut diameter. In the timed flight approach,
Investigators have made assumptions about particles the size of the cut diameter theoreti-
the initial particle radial position, r{, and the cally enter the cyclone at the midpoint of
value for vortex exponent, n. Residence time, gas entry.
740 HANDBOOK OF POWDER SCIENCE

15.3.3.3 The Fractional Efficiency Curve dimensionless geometry parameter, C^, that
depends on inlet dimensions:
Critical particle diameter is useful only as a
rough estimator of cyclone efficiency. For Cp = ab/D2 (15.38)
more precise work, as when estimating overall
cyclone efficiency on a dust with a range of ln^ = 0.62 - 0.87mW50-cm)
particle sizes, the entire fractional efficiency + 5.21 In C«
curve is necessary. Lapple6 and Barth25 have
developed generalized plots of efficiency ver- + 1.05(lnCfl) (15.39)
sus a dimensionless particle parameter.
Efficiency data from the literature36 were
Lapple's parameter is defined as particle di-
used to compare the prediction of efficiency
ameter over the cut diameter calculated from
using Eqs. (15.25) and (15.37) through (15.39)
Eq. (15.35). This plot is given in Figure 15.12
against other theories. 17 ' 25 ' 34 ' 37 Equations
and is valid for cyclones of the Lapple design
(15.25) and (15.39) were found to predict
listed in Table 15.1. No figures for cyclones of
efficiency significantly better than the other
other design are available.
theories.35
Iozia (Jones) and Leith26'35 developed an
Leith and Licht34 combined an approximate
equation to predict fractional efficiency from
solution to Eq. (15.21) with the assumption
the dimensionless particle parameter of cut
that uncollected dust is remixed within the
diameter calculated from Eq. (15.25) over par-
cyclone gas stream due to gas stream turbu-
ticle diameter. The fractional efficiency curve
lence. The assumptions they made for solving
is defined by using the particle parameter in a
Eq. (15.21) are listed in Table 15.3. The resul-
"logistic" equation.
tant equation predicts the fractional efficiency
1 curve:
(15.37)
(d50/df 2rt + 2)
) (15.36)
The logistic slope parameter, /3, is estimated Here, the vortex exponent, n, can be calcu-
from cut diameter (in centimeters) and a lated from Eq. (15.2) and (15.3), or found from

I I I I I I II
0.3 0.5 0.7 1.0 2 3 5 7 10

a
50
Figure 15.12. Fractional efficiency versus d/d50 for Lapple design cyclone.
 CYCLONES 741

Figure 15.5. The influence of particle and gas the theories discussed here may not apply to
properties are combined into I/J, a dimension- smaller cyclones. For these small cyclones, al-
less inertia parameter or Stokes' number: ternative collection efficiency expressions may
2 be more appropriate.39"42
d pvv{(n
(15.40) Although theoretical calculations of critical
particle diameter and fractional efficiency are
useful, they, like theoretical pressure drop cal-
The effect of cyclone geometry is consolidated
culations, may predict performance substan-
in C L , a dimensionless geometry parameter.
tially in error from that experienced in the
The geometry parameter depends only on the
field. All efficiency theories discussed have
cyclone dimension ratios, and is independent
been for tangential entry, reverse-flow cy-
of size
clones as shown in Figure 15.1. Their applica-
bility is unknown to other cyclone designs,
r ^ ( 2 ( 1 (De\2\(S °\ such as those with either scroll or swirl vane
entries, or to straight-through cyclones of the
1 / S + zc - h type in Figure 15.2. The best way to determine
cyclone fractional efficiency characteristics is
D
to test the cyclone in the laboratory or in a
De\2zc S pilot test program.
(15.41) Once an experimental fractional efficiency
~D ~ ~D
curve has been developed for a cyclone oper-
Core length, zc, is found from an equation ating under known conditions, the fractional
developed by Alexander:24 efficiency curve can be determined for a cy-
2\V3
clone of the same design under different oper-
D2 ating conditions by adjusting the efficiency
ze-23De\- (15.42)
curve of the test cyclone. According to one
theory,34 two cyclones will have the same ef-
The diameter of the core, dc, can be deter- ficiency when their Stokes numbers are the
mined from Eq. (15.43). same. If the test cyclone has known efficiency
/ S + zc - h \ on particles of size dv a similar cyclone will
dc = D-(D-B)i R_h (15.43) have the same efficiency on particles d2,
where:
Equation (15.36) implies that a cyclone with 1/2
a high value of geometry parameter, C L , D
2
should have a higher efficiency than a unit = dA^^ — -^\ (15.44)
Ql Pp2 Ml E
with a low value of C L for particles of all sizes
and for all operating conditions. The efficiency This analysis assumes that the diameters of
capabilities of alternative cyclone designs can the two cyclones are close enough that the
be evaluated by comparing their values of C L value of the vortex exponent, n, does not
in the same way that pressure drop require- change appreciably. The fractional efficiency
ments are evaluated by comparing values of curve for the similar cyclone can be con-
AH. Equation (15.36) was tested against ex- structed from the curve for the tested cyclone
perimental data from the literature38 and was by picking a series of coordinates from the
found to predict the data reasonably well. experimentally derived efficiency curve and
The equations discussed in Table 15.3 may calculating the analogous coordinates for the
be useful for determining the efficiency of similar cyclone from Eq. (15.44).
industrial-sized cyclones, a few meters or less The accuracy of this procedure decreases as
in diameter. The gas flow assumptions used in each of the ratios in Eq. (15.44), Qx/Q2, etc.,
742 HANDBOOK OF POWDER SCIENCE

departs more and more from unity. The proce- cyclone inlet velocity above which saltation
dure is especially suspect when predicting the occurs, vis, is:
performance of cyclones with much greater
diameter and throughput than the test model.
Also, when adapting results based on an ex-
perimental dust to a different dust, particle
SI units (m, kg, s) must be used in this equa-
shape may change as well as density. Never-
tion. Cyclone efficiency increases with inlet
theless, a fractional efficiency curve calculated
velocity up to about 1.25 vis; further increases
using this procedure is strongly preferred over
in inlet velocity cause a decrease in efficiency
one determined strictly from theory.
as saltation and reentrainment of collected
dust become more important.
15.3.4 Other Variables Affecting
Stairmand4 showed that the overall effi-
Performance
ciency of a well-designed cyclone increases
Although cyclone performance theories ex- from its normal value of 92% to an increased
press the effect of many variables on cyclone value of 93.6% when about 10% of the gas
performance, several variables known to in- flow is drawn through the dust outlet. A simi-
fluence pressure drop and efficiency are not lar "base purge" increased the efficiency of a
considered. poorer cyclone design from 89.1% to 92.2%.
Increasing inlet dust concentration, ci9 si- Stairmand believes this efficiency increase is
multaneously increases collection efficiency due to a reduced reentrainment of separated
and decreases pressure drop. Briggs quantified dust in the dust outlet region. The disadvan-
the influence of dust loading on pressure drop. tages of this practice are that it requires the
use of otherwise unnecessary auxiliary fans
clean and ducts to draw off the purge, and that if the
1/2
(15.45) purge is recycled to the cyclone inlet, the
0.0086(ci)
cyclone must be sized to handle the purge air
Here, ct has the dimensions of grams per as well as the process air. In practice, base
cubic meter. The effect on efficiency of chang- purge is seldom used.
ing inlet loading from c{1 to c i2 can be found11 Stairmand4 also reported that efficiency in-
from: creases from a normal value of 92% to 93.7%
for the well-designed cyclone and 89.1% to
100 - 0.182 93.2% for the poorer design when these cy-
(15.46) clones operate with wetted walls. The wetted
100 -
walls may reduce reentrainment of collected
dust throughout the cyclone. Disadvantages of
Presumably the values of efficiency and con-
this practice are that water piping is required
centration in Eq. (15.46) are for poly disperse
and that the collected dust is in a slurry.
dusts and the relationship applies to overall
dust concentration and efficiency rather than
15.3.5 Overall Efficiency on
to values for any one particle size.
Polydisperse Dusts
If the tangential velocity of the gas near the
cyclone wall is too high, saltation will occur; Industrial dusts contain particles of many sizes.
particles will bounce along the cyclone wall To calculate the overall cyclone collection ef-
and not be separated effectively from the gas ficiency, T7overaii, on such a dust one must mul-
stream. Kalen and Zenz 43 have examined this tiply efficiency for each particle size by the
phenomenon, and its implications for cyclone fraction of particles in the dust that are of that
design are discussed by Koch and Lict.44 An size. The sum of these products is the overall
empirical equation (15.47), which gives the fractional efficiency for the cyclone. Table 15.4
 CYCLONES 743

Table 15.4. Overall Collection Efficiency Calculation Using Numerical Integration of Eq. (15.48).

(5)
(1) (2) (3) (4) FRACTIONAL
SIZE RANGE MEAN SIZE FRACTION IN EFFICIENCY ON EFFICIENCY
(MICROMETERS) (MICROMETERS) RANGE MEAN SIZE COLUMNS (3) X (4)
0-2 1 0.10 0.03 0.00
2-5 3.5 0.10 0.38 0.04
5-10 7.5 0.10 0.81 0.08
10-20 15 0.15 0.96 0.14
20-30 25 0.10 0.99 0.10
30-40 35 0.10 1.00 0.10
40-60 50 0.15 1.00 0.15
60-76 68 0.10 1.00 0.10
76-104 90 0.07 1.00 0.07
104-150 127 0.03 1.00 0.03
Total 1.00 0.81

illustrates this process for the cyclone whose Here, nd is the efficiency on particles of a
fractional efficiency curve is shown in Figure certain size, d, and dG is the fraction of all
15.10. The dust size distribution is plotted in particles of that size in the dust. The overall
Figure 15.13. Equation (15.48) is the formal efficiency for this cyclone on this dust is found
mathematical statement of this process. to be about 85%.

O verall (15.48)

15.4 CYCLONE DESIGN

200 15.4.1 Necessary Design Information
Before attempting to design a cyclone or cy-
clone system it is important to consider cy-
clone limitations, to be sure that an alternative
control device might not work better. Cyclones
may be unsuited for collection of particles less
than about 5 microns in diameter, as efficiency
falls off rapidly for particles smaller than this.
Other types of collectors such as fabric filters,
electrostatic precipitators, and some kinds of
scrubbers will be able to collect these small
particles more efficiently. If the inlet dust
loading is high and the desired outlet concen-
tration is low, it may be necessary to use a
higher efficiency collector such as a fabric
filter either instead of, or in conjunction with
(usually after) a cyclone system. Sticky or hy-
0.02 0.1 03 0.5 0.7 0.9 0.98 groscopic dusts may stick to the cyclone walls,
FRACTION LESS THAN STATED and not discharge into the collection hopper.
DIAMETER For dusts of this type, a scrubber may be a
Figure 15.13. Particle size distribution from Table 15.4. better collector choice than a cyclone.
744 HANDBOOK OF POWDER SCIENCE

To design a cyclone or any collection device, clone. This is because collection efficiency de-
the inlet dust concentration and size distribu- creases with increasing cyclone diameter and
tion must be known. Although preliminary es- also because of possible problems with space
timates of expected dust properties are avail- or headroom requirements for very large
able from the literature,45"47 this information cyclones.
should always be obtained by stack sampling A fractional efficiency curve for the selected
when possible. Of course, when designing con- design can be determined by one of the meth-
trol equipment for a plant that has yet not ods discussed above. The overall collection
been constructed, stack testing is impossible efficiency for the selected cyclone design, inlet
and in this case the design will have to be dust size distribution and concentration to be
based on data obtained from similar plants in processed, and outlet dust concentration
conjunction with the design plans for the desired can then be determined from the
process to be controlled. methods describe previously.
Design criteria such as gas flow rate, tem- A cyclone can be custom designed to per-
perature, and particle density—material den- form a specific dust collection job.48 This
sity, not apparent or bulk density—special approach will give a cyclone with a greater
conditions of corrosivity, particle abrasiveness, collection efficiency, smaller size, or lower
and fluctuations in gas flow should be noted. pressure drop than a cyclone with a standard
These data requirements are summarized in design. The "optimized" cyclone design proce-
Table 15.5. All the data necessary for design dure requires trial and error calculations that
of a cyclone system can be obtained from a are better suited for a microcomputer or pro-
stack test performed on the gas stream to grammable calculator than by hand.
be cleaned. First, determine a preliminary cyclone di-
ameter from Eq. (15.49):
15.4.2 Cyclone Specification
n
p ^cyclone '
Usually cyclones are not custom designed. Dm = Pp ^ (15.49)
Rather an accepted standard design is se- 275
lected, such as one listed in Table 15.1 or a Particle density and flow must be in units of
manufacturer's proprietary design. Cyclone di- m-kg-h in this equation. If the diameter cal-
ameter can be determined from gas flowrate culated from Eq. (15.49) is greater than 2 m,
Q, using the value for Q/D2 tabulated for then the flow should be divided to accommo-
each standard design given in Table 15.1. Once date at least two cyclones from the start. In
diameter is known the remaining seven dimen- most situations, two or more cyclones should
sions can be determined from the dimension be used to allow flexibility in operation and
ratios of the standard design selected. For maintenance, and to avoid a system shutdown
volumetric gas flows larger than about 20,000 if one cyclone becomes plugged.
m 3 /h it is often better to use several smaller The flow going to each cyclone is calculated
cyclones in parallel rather than one large cy- by Eq. (15.50):

£ system
Table 15.5. Data Necessary for Cyclone Design. 'cyclone (15.50)
Particle size distribution
Inlet dust loading (g/m3) Next, pick a target value for outlet concen-
Particle density (kg/m3)
Gas flowrate (m 3 /h) tration or overall cyclone efficiency, which is
Gas temperature (°C) determined by Eq. (15.51):
Special conditions of corrosivity, abrasiveness,
fluctuations in gas flow, etc. c0 = - T?overall) (15.51)
 CYCLONES 745

Using the design parameters in Table 15.6 Eq. (15.48) These calculations were shown
calculate the overall cyclone collection effi- previously in Table 15.4.
ciency of the dust stream with three different Plot the K values against the predicted
cyclone designs that correspond to design pa- overall efficiency. From the line joining the
rameter K values of 1.5, 3, and 4.4. The cy- three points, determine the closest K value
clone diameter calculated from Eq. (15.49) from Table 15.6 that corresponds to the target
and the cyclone flow calculated from Eq. efficiency. The design in Table 15.6 that corre-
(15.50) are needed in the efficiency calcula- sponds to this K value is the optimized cy-
tions. The highest efficiency cyclone design clone design. At this point the cyclone design
will correspond to K equal to 1.5 and the is fixed.
lowest efficiency cyclone design will corre- The pressure drop for the system will be
spond to K equal to 4.4. The overall efficiency determined from the number of cyclones and
for collection of polydisperse dust is found flow going to each cyclone. The cyclone pres-
from the fractional efficiency curve generated sure drop is calculated from Eq. (15.4) using
using Eq. (15.37) and numerical integration of the AH values from Table 15.6 and the inlet

Table 15.6. Optimized Designs.

DESIGN
PARAMETER
K a/D b/D De/D H/D h/D S/D B/D AH
1.5 0.16 0.30 0.26 6 1.5 0.16 0.26 11.4
1.6 0.18 0.30 0.28 6 1.5 0.18 0.28 11.0
1.7 0.20 0.30 0.30 6 1.5 0.20 0.30 10.7
1.8 0.22 0.30 0.31 6 1.5 0.22 0.31 11.0
.1.9 0.25 0.30 0.32 6 1.5 0.25 0.32 11.7
2.0 0.27 0.30 0.33 6 1.5 0.27 0.33 11.9
2.1 0.29 0.30 0.34 6 1.5 0.29 0.34 12.0
2.2 0.31 0.30 0.35 6 1.5 0.31 0.35 12.1
2.3 0.34 0.30 0.36 6 1.5 0.34 0.36 12.6
2.4 0.38 0.30 0.37 6 1L.5 0.38 0.37 13.3
2.5 0.40 0.30 0.38 6 ]L.5 0.40 0.38 13.3
2.6 0.43 0.30 0.39 6 ]L.5 0.43 0.39 13.6
2.7 0.48 0.30 0.41 6 )L.5 0.48 0.41 13.7
2.8 0.51 0.28 0.42 6 1L.5 0.51 0.42 12.7
2.9 0.54 0.26 0.43 6 ]L.5 0.54 0.43 12.1
3.0 0.57 0.25 0.44 6 L.5 0.57 0.44 11.8
3.1 0.60 0.25 0.44 6 L.5 0.60 0.44 12.4
3.2 0.63 0.25 0.45 6 L.5 0.63 0.45 12.4
3.3 0.66 0.25 0.46 6 L.5 0.66 0.46 12.5
3.4 0.69 0.25 0.47 6 •L.5 0.69 0.47 12.5
3.5 0.73 0.25 0.48 6 L.5 0.73 0.48 12.7
3.6 0.76 0.25 0.48 6 1.5 0.76 0.48 13.2
3.7 0.79 0.25 0.49 6 1.5 0.79 0.49 13.2
3.8 0.82 0.25 0.50 6 1.5 0.82 0.50 13.1
3.9 0.85 0.25 0.50 6 1.5 0.85 0.50 13.6
4.0 0.89 0.25 0.51 6 1.5 0.89 0.51 13.7
4.1 0.93 0.25 0.51 6 1.5 0.93 0.51 14.3
4.2 0.96 0.25 0.52 6 1.5 0.96 0.52 14.2
4.3 0.99 0.25 0.52 6 1,5 0.99 0.52 14.6
4.4 1.00 0.25 0.52 6 1.5 1.00 0.52 14.8
746 HANDBOOK OF POWDER SCIENCE

velocity of the cyclone with flow equal to Since the cyclone diameter calculated with Eq.
Gcycione calculated from Eq. (15.50). If the (15.52) and a flow of 40,000 m 3 /h is greater
calculated pressure drop is too high, then the than 2 m, two cyclones should be used in the
number of cyclones should be increased in Eq. preliminary design calculations. The flow to
(15.50) until an acceptable pressure drop is each cyclone then, calculated from Eq. (15.50),
obtained. As discussed earlier, pressure drop is 40,000 divided by 2 or 20,000 m 3 /h.
calculated from theory can be considerably The other dimensions of the cyclone can be
higher or lower than actual. The limitations of calculated from the dimension ratios given in
the system being designed should be consid- Table 15.1 and are given in Table 15.7. Inlet
ered before a final decision is made. gas velocity will be found from Eq. (15.53).
Once the number of cyclones is fixed, the
design diameter of the cyclone is calculated (20,000 m 3 /h)(l h/3600 s)
V: =
from Eq. (15.49) with Q equal to <2cycione- (0.95m)(0.38m)
In some cases, no cyclone system will pro-
= 15.3 m / s (15.53)
vide adequate collection efficiency or suitable
pressure drop; in this case, teams of cyclones The pressure loss for the system can be
in series or alternate control devices should be calculated from LH for the Stairmand design
considered. from Table 15.1 and Equation (15.4).

15.4.3 Design Example AP = 6.4(1/2X0.944 kg/m 3 )(15.3 m/s) 2

The design procedures described in the previ- AP = 710kPa (15.54)
ous section can be illustrated by the following
design example. Suppose 40,000 m 3 /h of air The expected efficiency and outlet dust
containing rock dust comes from a rotary dryer loading concentration can be calculated using
at 100°C. It is desired to collect as much dust the method outlined previously. The maximum
as possible for recycle to the process. The tangential velocity is calculated from Eqs.
effluent from the cyclone system will go to a (15.26) and (15.27).
scrubber for final control before release to the
2 61
atmosphere. The maximum loading to the Cj = [(0.95 m)(0.38 m)/(1.91 m)
m; ]°'
scrubber should be 10 g/m 3 ; however, 8 g/m 3
or less is preferable. A stack test finds the X[(0.95m)/(1.91m)]"
1-0.74

dust loading from the dryer is 50 g/m 3 . The X [(7.63 m)/(l.91m)]"

1-0.33
pressure drop for the cyclone must be less
than 2 kPa. C, = 0.26 (15.55)
Vmax = 6.1(15.3 m/s)(0.26)
15.4.3.1 Example Using Standard Designs (15.56)
Fmax = 24.2 m / s
Any of the standard cyclone designs shown in
Table 15.1 may be used. Example calculations Core diameter is calculated using Eq. (15.28).
using the Stairmand design will be shown here.
The diameter of the cyclone can be found dc = (0.52X1.91)(0.ir°- 25 (0.5) 1 - 53 5 5?)
from cyclone flow and the value for Q/D2 in dc = 0.61 m
Table 15.1 as in Eq. (15.52).
Since core diameter is less than the outlet
1/2
(40,000(m 3 /h)/2) diameter (0.72 m), core length is calculated
(15.52) with Eq. (15.30) rather than Eq. (15.29)
(5500) 1/2
DM= 1.91m z = 7.63 - 0.95 = 6.68 m (15.58)
 CYCLONES 747

Table 15.7. Cyclone Specifications for Design Example.

STAIRMAND STANDARD DESIGN OPTIMIZED DESIGN

DESIGN PARAMETERS TWO CYCLONES FOUR CYCLONES TWO CYCLONES
Diameter 1.91 1.35 1.34
a 0.95 0.67 0.92
b 0.38 0.27 0.33
De 0.95 0.67 0.63
H 7.63 5.39 8.03
h 2.86 2.02 2.01
S 0.95 0.67 0.67
B 0.72 0.51 0.63
A// 6.4 6.4 12.5
Inlet velocity (V{), m / s 15 15 18
Maximum tangential
velocity (Kmax), m / s 24 24 36
Maximum inlet velocity
for no saltation, m / s 30 28 39
Core diameter, m 0.61 0.43 0.34
Core length, m 6.7 4.7 7.4
d50, microns 5.9 5.0 3.8
Logistic slope
parameter (/3) 1.9 2.2 4.8
Overall efficiency (7?)OVeraii 0.81 0.84 0.84
Pressure drop, kPa 0.71 0.71 1.91
Outlet loading, g / m 3 9.4 8.1 8.1

Next, the value for d50 is calculated from Eq. Cyclone efficiency can now be found for
(15.25). the Stairmand design using Eq. (15.37) for
particles of any size, d.
d50 = {[9(20,000 m 3 /h)(l h/3600 s)
1
X (2.17 X 10" 5 kg/m • S)]/TT(6.68 m) (15.63)
1.93
1 + [(5.9 X 10~6 m)/d]
X (2500 kg/m 3 )(24.2 m/s)2}1/2
(15.59) The relationship between particle diameter
and collection efficiency for the Stairmand de-
*50 = 5.9 X 10~6 m
sign given by Eq. (15.37) is plotted in Figure
The value for the logistic slope parameter 15.14. The overall efficiency for this cyclone
is calculated from Eqs. (15.38) and (15.39). on particles with the distribution given in Fig-
ure 15.13 is determined through calculations
Cp = (0.95 m)(0.38 m)/(1.91 m) shown previously in Table 15.4. Overall effi-
(15.60) ciency was found to be 81% for this design;
Cp = 0.1 outlet dust concentration found with Eq.
(15.51) then will be 9.4 g/m 3 . Although this
In p = 0.62 - 0.871n(5.9 X 10~4 cm)
concentration meets the minimum require-
+ 5.211n(0.1) + 1.05(ln0.1)2 ( 15 - 61 > ments, it is higher than the target efficiency.
Therefore, the number of cyclones is in-
In p = 0.659 creased until the target efficiency is reached.
P = 1.93 (15.62) The calculations performed above for the
748 HANDBOOK OF POWDER SCIENCE

<
o
o

0.2 0-6

Log d/d50

Figure 15.14. Fractional efficiency versus d/d50 for Stairmand design cyclone.

two-cyclone system are repeated for a three- 15.13 and the two-cyclone system. The method
and four-cyclone system. The four-cyclone sys- to calculate overall collection efficiency for
tem reaches the goal of 8 g/m 3 and 84% this system is analogous to the calculations
control. The results of the calculations for a shown in Eqs. (15.53) through (15.63) for the
four-cyclone system are also shown in Table Stairmand design.
15.7. The four-cyclone system also meets the From the plot in Figure 15.15 a K value of
design objective in terms of pressure drop. 3.4 is found that corresponds closest to 84%
collection efficiency, the target efficiency of
15.4.3.2 Example Using Customized Design the system. The optimized design dimensions
As in the standard procedure, the system will corresponding to this K value and cyclone
initially consist of two cyclones operating in diameter calculated from Eq. (15.49) are shown
parallel. The diameter of each cyclone is found in Table 15.7.
with Eq. (15.49) and cyclone flow calculated The pressure drop for the optimized design
with Eq. (15.50). is calculated from a A # value of 12.5 given in
Table 15.6 for the optimized design (K = 3.4)
1/3 using Eq. (15.4). Since the efficiency and pres-
[(2500 kg/m 3 )(20,000 m 3 /h)]
sure drop of the optimized design meet the
275 design objectives, only two cyclones will be
(15.64) necessary.
= 1.34 m

15.4.4 Other Design Considerations

Next a plot of design parameter K versus
collection efficiency is used to find the opti- After the shape and size of a cyclone or cy-
mized cyclone design for the situation. Figure clone system have been decided, it is impor-
15.15 shows the plot of design parameter, K, tant to consider the additional design criteria
of 1.5, 3, and 4.4 versus collection efficiency that will ensure long, trouble-free operation of
for the particle distribution given in Table the system.
 CYCLONES 749

0.95
0.94
0.93
0.92
0.91
0.9
0.89
\
0.88
0.87
0.86
0.85
0:84
0.83
0.82
0.S1
0.8
1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

Design Parameter, K

Figure 15.15. Design parameter, K, versus collection efficiency for example design problem.

When small cyclones, less than 300 mm or It is essential that air not be allowed to leak
so in diameter, are used wall erosion may pose into the cyclone through the dust outlet. Leak-
a serious problem8 if the dust is abrasive. age at this point can keep the cyclone from
Larger dust particles strike the cyclone wall discharging dust to the dust bin, and if suffi-
more forcefully and have more effect than ciently severe, can lower collection efficiency
smaller particles. Abrasion can be especially to zero.1 Leakage through the dust exit can
troublesome around welded seams, and occurs even occur with the cyclone on the pressure
whether the seams are horizontal or vertical. side of a fan owing to the low static pressure
The seam itself may not be as susceptible as in the cyclone central core, although the prob-
the cyclone wall around the seam, which may lem is more pronounced if the cyclone is on
have been softened through annealing during the suction side. When possible, cyclones are
the welding process.8 To minimize the effect mounted on the upstream or suction side of a
of wall erosion, several steps can be taken. fan to minimize wear of the fan impeller from
Often, commercially available small-diameter the dust in an uncleaned gas stream.
cyclones are cast rather than fabricated from If the cyclone operates on the effluent from
sheet metal. Casting eliminates the problem a batch process, the unit is often directly con-
with erosion around weldments, and may pro- nected to a dust bin below the dust exit with-
vide a thicker wall. Replacement wear plates out an intervening valve. When using this ar-
are sometimes installed on the cyclone wall rangement, it is essential that the dust bin be
opposite to the tangential gas inlet. When emptied before it fills and blocks the cyclone
installing a wear plate, it is important that the dust exit. In this case, the gas flowing to the
plate be fitted to maintain a smooth interior cyclone must be diverted before the dust bin
wall. Failure to maintain a smooth wall will can be emptied. The dust bin must be airtight
hasten the erosion of the wear plate or the to prevent leakage from the bin entering the
wall around the edges of the plate, and may cyclone dust exit.
also adversely affect cyclone efficiency. Wear A better solution to dust exit sealing is
plates and entire cyclone interior walls have through use of a valve between the exit and
been rubber coated to reduce erosion. dust bin. The valve must allow for the continu-
750 HANDBOOK OF POWDER SCIENCE

ous discharge of collected dust, but not permit b Gas entry width
backflow of air. A valved dust discharge ar- B Dust outlet diameter
rangement is essential for a continuous pro- Cj Inlet dust concentration
cess as it allows the collection bin to be emp- c0 Outlet dust concentration
tied at any time. Rotary values are often used 8 Cl Cyclone geometry parameter, Iozia
when the negative pressure at the dust exit is (Jones) and Leith
less than about 1 kPa (100 mm of water CL Cyclone geometry parameter, Leith
column). and Licht
Cyclone gas inlet velocities are frequently of Cp Logistic cyclone geometry parame-
the order of 15 m / s while duct velocities are ter
usually lower than this. To minimize pressure d Particle diameter
drop through the cyclone system it is impor- dc Diameter of cyclone core
tant to provide a good transition between the d50 Cut particle diameter, theoretically
inlet ductwork and the cyclone inlet. collected with 50% efficiency
Attempts have been made to regain some of d50-cm Cut diameter, in centimeters
the rotational energy in the outlet gas stream d100 Critical particle diameter, theoreti-
by modifying the shape of the gas outlet. A cally collected with 100% efficiency
thorough review of these devices is provided D Cyclone cylinder diameter
by Stern et al.18 Reverse scrolls mounted above Dm Cyclone cylinder diameter, meters
the gas outlet duct, and curved or straight De Gas outlet diameter
vanes within the gas outlet duct have been
/ Factor in Eq. (15.8)
used. These devices usually provide a reduc-
Fc Centrifugal force acting on particle
tion in pressure drop in the 10% range, but
Fd Drag force acting on particle
despite careful design, collection efficiency
g Acceleration of gravity, 9.81 m / s 2
may be adversely affected. Pressure recovery
G Dust cumulative size distribution
devices are not generally used.
G* Friction factor, 0.005
Build-up of collected dust on the cyclone
h Cyclone cylinder height
walls can be a problem, especially where soft
small-diameter, hygroscopic particles are col- H Cyclone overall height
lected. When build-up occurs, it can some- K Optimum design parameter
times be scoured out by feeding some large- n Vortex exponent
diameter, hard particles as an abrasive. Wall N Number of turns gas makes within
deposition of hygroscopic dusts is aggravated cyclone
by condensation of moisture from the gas Nc Number of cyclones
stream on the cyclone walls, when the cyclone Q Volumetric gas flowrate
F l o w oin
is mounted outdoors in winter. If the problem Gcycione g g t o o n e cyclone
occurs only on start-up, preheating the cyclone Goveraii Total flow of the system
by either warming the inlet gas stream or r Radial distance from cyclone axis
running gas through without dust may help. r core Radial distance from cyclone axis to
Cyclones whose inlet walls are smooth and edge of central core
that operate at inlet velocities in excess of r{ Radial distance from cyclone axis to
15 m / s will be less prone to wall build-up. innermost particle at entry
rw Radial distance from cyclone axis to
cyclone wall, D/2
S Gas outlet height
LIST OF SYMBOLS
t Time
a Gas entry height T Absolute temperature, K
A Inside surface area of cyclone u Particle velocity
 CYCLONES 751

Radial component of particle veloc- 5. P. Swift, Steam and Heating Engineer 38:453 (1969).
ity dr/dt 6. C. Lapple, Chem. Eng. 55:144 (1951).
Tangential component of particle 7. R. H. Perry and C. H. Chilton, Chemical Engineer''s
Handbook, 5th edit., McGraw-Hill, New York
velocity (1973).
Gas velocity 8. H. J. van Ebbenhorst Tengbergen, De Ingenieur.
Gas inlet velocity, Q/ab 77th Year of Publication, Wl (1965).
Gas inlet velocity above which salta- 9. H. J. van Ebbbenhorst Tengbergen, Staub 25:44
tion occurs (1965).
10. W. A. Baxter, in Source Control by Centrifugal Force
Maximum gas tangential velocity
and Gravity. K. J. Caplan, in Air Pollution, Vol. 3,
Radial component of gas velocity 2nd edit., edited by A. C. Stern, Academic, New
Tangential component of gas veloc- York (1968).
ity 11. L. C. Whiton, Chem. Met. Eng. 39:150 (1932).
Gas tangential velocity at cyclone 12. L. W. Briggs, Trans. Am. Inst. Chem. Eng. 42:511
outer wall (1946).
13. C. B. Shepherd and C. E. Lapple, Ind. Eng. Chem.
Length of the core
31:912 (1939).
Factor in Eq. (15.16) 14. K. Iinoya, Mem. Fac. Eng. Nagoya Univ. 5 (Sept.
Logistic slope parameter 1953).
Loss factor 15. E. Anderson, Chem. Met. Eng. 40:525 (1933).
Vd Fractional collection efficiency of 16. M. A. Lissman, Chem. Met. Eng. 37:630 (1930).
particles of one size, d 17. C. E. Lapple, Amer. Ind. Hyg. Assoc. Quart. 11:40
Overall collection efficiency for (1950).
Coverall
18. A. C. Stern, K. J. Caplan, and P. D. Bush, Cyclone
polydisperse dust Dust Collectors, American Petroleum Institute, New
AH Pressure drop expressed as number York (1956).
of inlet velocity heads 19. M. Seillan, Chal Ind. 10:233 (1929).
AP Pressure drop expressed as static 20. P. Rosin, E. Rammler, and E. Intelmann, V.D.I.
pressure head (Ver. Deut. Ing.) Z. 76:433 (1932).
21. F. Procket, Glasers Ann. 107:43 (1930).
Ratio of maximum tangential gas
22. A. J. ter Linden, Proc. Inst. Mech. Engrs. (London)
velocity to gas velocity in gas outlet, 160:233 (1949).
see Eq. (15.15) 23. M. W. First, Sc.D. thesis. Harvard University,
Friction factor, 0.02 Boston (1950).
Gas viscosity 24. R. McK. Alexander, Proc. Australas, Inst. Mining
Gas density Met. N.S. 152-153:203 (1949).
PG
Particle density 25. W. Barth, Brennst.-Waerme-Kraft 8:1 (1956).
26. D. L. Iozia (Jones) and D. Leith, Aerosol Sci. Tech-
Ratio of maximum tangential gas
nol 10:491 (1989).
velocity to velocity within gas entry 27. K. J. Caplan, in Air Pollution, Vol. 4, 3rd edit.,
Inertia parameter edited by A. C. Stern, Academic Press, New York
(1977).
28. C. B. Shepherd and C. E. Lapple, Ing. Eng. Chem.
32:1246 (1940).
REFERENCES 29. C. J. Stairmand, Engineering (London) 168:409
(1949).
1. R. Jackson, Mechanical Equipment for Removing 30. D. Leith and D. Mehta, Atmos. Environ. 7:527
Grit and Dust from Gases, Cheney and Sons, (1973).
Banbury, England (1963). 31. S. K. Friedlander, L. Silverman, P. Drinker, and
2. W. Barth, Staub 21:382 (1961). M. W. First, Handbook on Air Cleaning. U.S.A.
3. J. I. T. Stenhouse and M. Trow, in Proceedings of E.C., AECD-3361, NYO-1572, Washington (1952).
Second World Filtration Congress, 1 Katharine St., 32. C. E. Lapple and C. B. Shepherd, Ind. Eng. Chem.
Croydon CR9 1LB, England (1979). 32:605 (1940).
4. C. J. Stairmand, Trans. Inst. Chem. Eng. 29:356 33. C. N. Davies, Proc. Inst. Mech. Engrs. (London)
(1951). 10:185 (1952).
752 HANDBOOK OF POWDER SCIENCE

34. D. Leith and W. Licht, A.I.Ch.E. Symposium Scr. 43. B. Kalen and F. A. Zenz, A.I.Ch.E. Symposium Ser.
68:196 (1972). 70(137):388 (1974).
35. D. L. Iozia (Jones) and D. Leith, Aerosol Sc. and 44. W. Koch and W. Licht, Chem. Eng. 84:&0 (Nov. 4,
Technol 72:598 (1990). 1977).
36. J. A. Dirgo and D. Leith, Filtration Separation 45. Midwest Research Institute. Handbook of Emis-
22:119 (1985). sions, Effluents and Control Practices for Stationary
37. P. W. Dietz, Assoc. Ind. Chem. Eng. J. 27:288 (1981). Paniculate Pollution Sources. NAPCA contract CPA
38. N. A. Fuchs, The Mechanics of Aerosols, Pergamon, 22-69-104, NTIS Publication No. PB 203-522,
New York (1964). Springfield, VA (1970).
39. T. Chan and M. Lippman, Environ. Sci. Technol.
46. U.S. Environmental Protection Agency. Compila-
11:317 (1977).
tion of Air Pollution Emission Factors, 2nd edit.,
40. W. Licht, T. Chan, and M. Lippman, Environ. Sci.
Publication No. AP 42 (April 1973).
Technol. 11:1021 (1977).
41. W. B. Smith, D. L. Iozia, and D. B. Harris, 47. J. A. Danielson, Air Pollution Engineering Manual,
/. Aerosol Sci. 14:402 (1983). 2nd edit., EPA. Publication No. AP 40 (May 1973).
42. W. B. Smith, R. R. Wilson, D. B. Harris, Environ. 48. D. L. Iozia (Jones) and D. Leith, Filtration Separa-
Sci. Technol. 13 (1979). tion 24:212 (1989).
 16
The Electrostatic Precipitator:
Application and Concepts
Jacob Katz

CONTENTS

16.1 INTRODUCTION 753

16.2 FACTORS AND EFFECTS 757
16.3 RESISTIVITY 759
16.4 OPERATION AND MAINTENANCE 763
16.5 GAS CONDITIONING 768
16.6 DESIGN AND PERFORMANCE CONCEPTS 768
16.7 EFFECT OF PARTICLE SIZE 769
REFERENCES 770

16.1 INTRODUCTION on the particle, rather than on the carrier gas

stream. However, the gas stream or process
16.1.1 General Comments
characteristics will generally determine
The electrostatic precipitator uses electrical whether the particle will be easily collected or
forces to capture either liquid or solid particu- prove difficult to contain by electrical forces.
late matter from a flue gas system. We tend to An interesting facet of precipitation is that
classify the precipitator as a high-efficiency even after many years of application, simple
collector, comparable to the fabric bag-house fundamental knowledge has eluded personnel
or high-pressure venturi scrubber. As such, involved with this equipment. That is why we
collection efficiencies of 99.5% plus are within continually have to face the same field prob-
a design range for most applications. A prime lems and why performance of the precipitators
characteristic separating the various methods often varies greatly from its design criteria.
for high-efficiency collection is that the precip- This lack of understanding cannot be over-
itator concentrates its primary energy forces come with more theoretical coverage, but
753
754 HANDBOOK OF POWDER SCIENCE

rather there is a need for practical concepts handle. Some typical industrial processes that
and field information to be clearly identified have successfully employed the precipitator in-
and distributed. clude:
For that reason, this chapter attempts to
provide a brief description for some of the key PRINCIPAL MATERIAL
areas of precipitation without regard for de- PROCESSES COLLECTED
tailed theory. It actually consists of excerpts Utility Fly ash (SiO 2 , A1 2 O 3 ,
from the book The Art of Electrostatic Precipi- Fe 2 O 3 )
tation written for the practitioner. The bibliog- Industrial boiler houses Fly ash
Oxygen steelmaking Iron oxide (Fe 3 O 4 )
raphy at the end of this chapter also includes furnaces
sources of literature that can be used to Cement kilns Calcium oxide, silicon
upgrade the theoretical knowledge of the oxide
precipitator. Pulp and paper Sodium sulfate

16.1.2 Performance Comments A large number of precipitators have also

Each industrial process presents its own spe- cleaned process gases from blast furnaces, sin-
cific problems or application factors when pre- ter plants, open hearths, coke ovens, gypsum
cipitators are considered. Subtle changes in plants, catalytic cracking, smelters, sulfuric
the process, of raw materials and equipment, acid, and phosphoric acid plants. Other pro-
can often produce a wide band of performance cesses that use precipitators include electric-
characteristics in a specific precipitator. These arc furnaces, scarfing machines, incinerators,
subtle changes probably produce the greatest and the carbon black industry. Other precipi-
number of installations that fail to meet ex- tators have also recovered valuable metals in
pected collection efficiencies because of re- special situations.
duced electrical power input. A secondary
contribution to subperformance levels is a 16.1.4 Precipitator Arrangements
combination of design shortcomings including, A convenient and logical method of comparing
among others, gas distribution, gas sneakage, precipitators is the total collecting surface area
and insufficient sectionalization. A third group and the amount of mechanical and electrical
of factors that keeps the collector from per- sectionalization. The basic terminology used
forming consistently on a satisfactory basis is to describe sectionalization follows and an
the balance of reliability, either in the evacua- illustration is shown in Figure 16.1.
tion of material from hoppers or in the failure
of precipitator components.
Transformer- Rectifier
T-RSet

16.1.3 Applications
The use of precipitators has been applied in
all the basic as well as some exotic industries
over the years. Collection of particulate mat-
ter in a dry-type precipitator with flue gas
temperatures between 250 to 700°F has been
the most popular application. However, spe-
cific process characteristics will usually deter-
mine the design and type of precipitator uti-
lized. There are process situations where the 1 Precipitator showing 12 Bus Sections with 6 Power Supplies
either F-W or Double Half-Wave.

effectiveness of the electrical collector is ques- Figure 16.1. Typical precipitator arrangements showing
tionable because the material is difficult to terminology and method of applying power supplies.
 THE ELECTROSTATIC PRECIPITATOR: APPLICATION AND CONCEPTS 755

Precipitator. A single precipitator is an ar- the function of ionizing the gas and creating
rangement of collecting surfaces and discharge the electric field.
electrodes contained within an independent
housing. Collecting Surface Rapper. A device for im-
parting vibration or shock to the collecting
Bus Section. The smallest portion of surface to dislodge the deposited particulate.
the precipitator that can be independently
deenergized. Aspect Ratio. The length of the precipitator
divided by its height.
High-Voltage Power Supply. The power
supply unit to produce the high voltage re- 16.1.5 Basic Concepts of Precipitation
quired for precipitation, consisting of a trans-
Probably the best way to gain an insight into
former-rectifier combination and assorted
the process of precipitation is to study a rela-
controls. Numerous bus sections can be ener-
tionship generally known as the Deutsch-
gized by one power supply.
Anderson equation. This equation and adapta-
tions of it are well covered in several books.1'2
Field. A field of a precipitator is an arrange-
It describes the factors involved in the collec-
ment of bus sections in the direction of gas
tion efficiency of the precipitator as shown in
flow that is energized by one or more power
its simplest form:
supplies situated laterally across the gas flow.
Collection efficiency
Collecting Surfaces. The individually ground A y
1- e-( /W
components that make up the collecting sys-
where
tem and that collectively provide the total area
of the precipitator for the deposition of partic- A = effective collecting electrode area of the
ulate. precipitator (m2)
V = gas flow rate through the precipitator
Collecting Surface Area. The total flat pro- (acm/s)
jected area of collecting surface exposed to W = migration velocity (m/s).
the electrostatic field (effective length X
effective height X number of sides). This equation has been used extensively in
the above form in past years. Unfortunately,
Effective Height. Total height of collecting while the relationship is scientifically valid,
surface measured from top to bottom. there are a number of operating parameters
that can cause the exponent to be in error by
Effective Width. Total number of gas pas- as much as a factor of two or more. It is well
sages multiplied by the center to center spac- to remember that the basic D-A equation can
ing of the collecting surfaces. (Disregard shape be used as an indicator or tool, but has limita-
of collection surface.) tions more often than not unless equated with
some practical and empirical considerations by
Effective Cross-Sectional Area. Effective the designer. Values used can either be in the
width times effective height. English or metric systems.
The exponent term W, known as the migra-
Gas Passage. Formed by two adjacent rows tion velocity, actually represents the speed of
of collecting surfaces. movement of the particle toward the collector
surface under the influence of an electrical
Discharge Electrode. The component that is field. While we would consider it more an
installed in the high-voltage system to provide indicator than actual velocity, it does have a
756 HANDBOOK OF POWDER SCIENCE

finite value that can be used for comparison that small changes can produce substantial
purposes. This migration velocity is comprised changes in power, and hence in the efficiency
of: of the collector. This is especially true at the
lower levels of power input. It is therefore
ITTO important to understand the factors that
where affect the electrical characteristics of the pre-
cipitator.
a = particle radius, microns
Eo = strength of field in which particles are 16.1.6 Main Factors Affecting Electrical
charged, statvolts/cm (represented by Characteristics
the peak voltage)
Ep = strength of field in which particles are Optimum power input to the precipitator
collected, statvolts/cm (normally the varies among processes and even changes on a
field close to the collecting plates) minute to minute basis for certain applica-
6 = viscosity of frictional resistance coeffi- tions. There are seven basic factors that
cient of the gas. directly affect the electrical characteristics.
These are:
High levels of voltage and useful corona
power in the precipitator, all other conditions 1. Design of power supply
being equal, are synonymous with high collec- 2. Physical design of precipitator
tion efficiencies. Figure 16.2 shows a typical 3. Design of electrode system
performance curve of the effect on efficiency 4. Characteristics of gas stream
by changes in the peak voltage of a precipita- 5. Effect of process changes
tor. This simple curve can represent only one 6. Characteristics of particulate
situation because each precipitator will have 7. Maintenance factors.
its own characteristic curve based on many
factors. The important point to remember is The power supply must be matched cor-
rectly for the precipitator section or service
expected, or several difficulties can arise:
100

1. The impedance of the power supply, includ-

95
ing a ballast resistance or reactor in the
/
primary winding of the transformer, may
90
not be sufficient to dampen the severity of
/ electrical breakdowns in the precipitator.
This condition is especially likely if the
power supply rating is much larger than the
/
Note Effect of actual operating level.
Small Increase of
/
Voltage on Efficiency 2. If the physical size of the precipitator is too
75
large compared to the size of the power
/
supply, then lower than desirable precipita-
70
tor voltages may exist because the current
rating of the supply becomes the limiting
factor.
/ :
. Increases in Two (2) Kilovolt Steps. 3. The gas and particulate matter conditions
Typical Range 36 to 60 kVP.
PRECIPITATOR PEAK VOLTAGE (Kilovolts)
can drastically alter the voltage-current re-
Figure 16.2. A typical electrostatic precipitator peak
lationship and produce lower voltage fields
voltage versus dust collection efficiency curve shows than expected because a small power supply
how efficiency increases with voltage. becomes current limited.
 THE ELECTROSTATIC PRECIPITATOR: APPLICATION AND CONCEPTS 757

The average precipitator can be sensitive to often less critical than other factors. This com-
process changes in the following ways: ment is especially true with the larger designs
that exist today. Even with the 1.8 m / s
1. Changes to gas temperature (effect on (6 ft/s) or more velocity designs, it is often the
density) quality of gas that weighs most importantly.
2. Changes in gas pressure (effect on density) Whether or not the relationship of higher
3. Changes in gas flow rate gas flow rate to reduced efficiency becomes
4. Changes in gaseous composition critical is dependent in large part on the char-
5. Changes in particulate chemical character- acteristics of the particle. Certainly, large
istics porous particles such as combustible grit found
6. Changes in particulate concentration or in fly ash applications will be sensitive to in-
loading creased velocities. On the other hand, fine-
7. Changes in the size distribution of the sized particulate matter that tends to agglom-
particulate erate in the deposited layer of the collecting
8. Changes in the electrical conducting char- surface will resist easy reentrainment into the
acteristics of the particulate. gas stream. With low levels of power input and
low aspect ratios, high gas flow rates can often
It is difficult to separate the effect of one be observed in reduced performances of the
process change on another. If the rate of pro- precipitator.
cess change is rapid, the readings can change
almost instantaneously. On the other hand, 16.2.2 Gas Flow Distribution
rapid changes of temperature may not be seen
Gas distribution problems are of concern from
readily on the meters because of the heat sink
the standpoint of velocity, temperature, and
effect of the precipitator. Some changes in the
concentration of material as well as particle
process will cause large variations of voltage-
size. If one area of precipitation has become
current readings, while others will cause subtle
worse in recent years, it is in gas distribution.
effects.
The trend toward larger collectors has meant
The size distribution of the particulate mat-
greater difficulty in transferring the gas leav-
ter can have a bearing on electrical readings.
ing the inlet nozzle to an acceptable pattern at
For example, iron oxide fume from a basic
the face of the precipitator. Granted optimum
oxygen vessel contains a predominance of sub-
gas distribution is not as critical in the larger
micron particles that will react like a space-
units with all fields serviceable, but the margin
charge in a vacuum tube. This can actually
can be quickly lost with outages of equipment.
impede the flow of precipitator current and
Probably one fallacy in gas distribution is
thereby elevate the voltage potential across
placing too much emphasis on the results of
the space. This condition can become serious
model studies. The model cannot foresee the
enough to completely nullify the precipitator
fallout of material during periods of reduced
process depending on the electrode geometry
operation that will often distort the actual flow
and the concentration level of the submicron
pattern.
particles.
16.2.3 Gas Temperature
16.2 FACTORS AND EFFECTS The level of gas temperature in the precipita-
tor opens up many areas of interest, especially
16.2.1 Gas Flow Rate
the effect on the viscosity of the gas stream.
While the true measurement of the gas flow But the major effects of temperature lie in the
rate commands an important place in specifi- modification of the electrical characteristics
cations and performance tests, in practice, it is and the reactions of the particles as they de-
758 HANDBOOK OF POWDER SCIENCE

posit on surfaces. The effect on metal corro- 16.2.4 Rappers and Reentrainment
sion by changes in flue gas temperature must
also be considered. Nowhere in the original Deutsch-Anderson
equation is an allowance made for the losses
Practically all of the particulate matter han-
that occur in transferring the collected mate-
dled in precipitators will go through a wide
rial from electrode to hoppers. The interplay
spectrum of electrical characteristics for the
between the electrical forces holding the ma-
temperature range of 200 to 750°F. Much of
terial on the collecting surface and the rapping
this has to do with condensation effects and
device attempting to remove it provides a real
surface leakage at the lower range and con- challenge for effective precipitation. But this
ductivity changes in the bulk material at the challenge does not merit the priority some
higher temperatures. The true effects at any people have placed on high rapping forces.
given temperature will depend on the mois- This statement is valid as long as the rapper
ture level and chemical composition of the mechanism is sufficient to impart at least 10 to
particles. Of greater interest would be whether 25 Gs to the support structure holding a group
the precipitator is operating in critical temper- of collector plates. With many process condi-
ature zones for that particulate material. For tions, even a substandard rapper system will
example: not effect performance adversely. But when
the build-up on the collecting surface reaches
over 1.9 cm (f in.) it is prudent to assess
1. High sulfur coal for pulverized coal-fired whether the rapper system is sufficient. Great
precipitators would be critical in the 250 to emphasis should be placed on the reliability of
280°F zone. the rapper and control circuitry components.
2. Lower sulfur coal for this same precipitator This has become more important as collection
might find its most critical zone between efficiency levels have increased.
310 to 360°F. The effect of rapping on precipitator perfor-
3. Cement precipitators might find its most mance is whether puffing losses are observed
critical range in the 350 to 400°F. or measured since this can denote a significant
reentrainment of material from the collector
surfaces caused by the rapper operation. While
A variation in electrical readings may occur the reentrainment puff is usually a mechanical
with as little as 10 to 15°F movements in the occurrence, the operation of the rapping de-
process gases. In some fly ash installations a vice can sometimes effect the electrical char-
15°F change has meant a three to fourfold acteristics at the same time, thus aggravating
increase in emissions. the magnitude of the problem. The vibration
The ability to change flue gas temperatures of the high voltage frames could produce an
from critical zones is as important to success- electrical disturbance dependent on the struc-
ful precipitator performance as any other de- tural integrity of the discharge electrode
sign feature. As with variations in gas flow, system.
short-term variations in flue gas temperature
should be controlled in order to minimize
losses from the collector. In fact, it is usually 16.2.5 Power Supply Characteristics
better to operate at a less than optimum uni- As precipitators have grown in size so have the
form temperature rather than experience vari- power supplies grown in kva ratings. This trend
ations. The heat sink effect of the internal to larger transformer-rectifier capacities has
structure will tend to mask effects of the tem- introduced some difficulties in stability, and
perature cycle if it is less than 10 minutes in yes, even in the performance of the precipita-
duration. tor if a gross mismatch occurs between the
 THE ELECTROSTATIC PRECIPITATOR: APPLICATION AND CONCEPTS 759

size of the power supply and the field to be How the material is handled on the job site
energized. can be important to prevent a tendency to
It is well to understand a basic concept of distort long electrode elements. Weather pro-
precipitation that each field of any installation tection is of primary concern if long storage
will only effectively absorb the amount of time is required.
power that the existing gas, dust, and internal There are advantages for the user to assign
structure integrity allows. Therefore, the ac- an inspector during the actual erection phase.
tual voltage-current requirements of a precipi- Cross-checks of the actual erection procedures
tator field may be drastically different than are important.
shown by the full load rating of the power
supply. 16.2.8 Personnel Assignment

16.2.6 Operation and Maintenance Just who to assign to the precipitator system
Factors should be given much thought. The value of
the initial check-out and contacts with the
Many success stories of the high collection manufacturer can be lost if the user represen-
performance of precipitators are well docu- tative is moved to another assignment.
mented. But to many users, a constant battle A person who can be assigned long term to
is waged to maintain these performance levels. oversee the precipitator system and monitor
A major reason for this situation lies in the the process as to how it affects the collector is
basic design of components for the overall probably the best investment a company can
system that produces sensitivity for break- make. Recent years have shown the advantage
downs. A concerted effort must be made by of close supervision for the large precipitator
the user to understand all the inputs to the installations.
potential problems of maintenance.
Obviously, it is exceedingly difficult to pre-
dict where some of the maintenance troubles
16.3 RESISTIVITY
may occur, but there are eight key areas that
can be emphasized: 16.3.1 Introduction
Much emphasis has already been placed on
1. Raw material and operation forecasts—
the fact that effective precipitation coincides
original design
with the occurrence of optimum amounts of
2. Design concepts
electrical power input in the corona process.
3. Construction phase
While power input is sometimes limited by
4. Initial check and training
structure or individual component defects, the
5. Personnel assignment
performance of limited power installations oc-
6. Control of process
curs under conditions of excessive-electrical
7. Record keeping
resistivity of the collected material, usually
8. The actual maintenance program.
expressed in ohm-centimeters.
All finely divided particles that are gener-
16.2.7 Construction Phase
ated in the basic industrial processes have
The best precipitator design can be adversely critical temperature zones that can affect the
affected in the fabricating and erection phases. electrical operation of the precipitator. The
Just how the quality of welding is controlled, chemical composition of the bulk of dust parti-
or the shaping of the component is finally cles contain common constituents even if the
accomplished in the shop, could have a signif- make-up varies somewhat in weight fractions.
icant effect on the final operating characteris- Given similar gas conditions, it might be hard
tics of the precipitator. to differentiate electrically whether it was fly
760 HANDBOOK OF POWDER SCIENCE

ash, cement dust, or iron oxide being handled Second, the current limit is observed more
in the collector. Of course, that is a simplistic often. Adjusting the conduction angle of the
statement because in practice there are an secondary current to approximately 86% by
infinite number of process conditions where the correct application of linear reactors will
differences in raw material can alter the elec- generally produce a match of the primary and
trical characteristics of the particles. secondary currents. As the conduction angle
Fortunately, high moisture contents in the decreases from 86%, the primary current will
flue gas stream (such as found in wet process trend toward higher readings relative to the
cement applications or other applications secondary current reading. The rated sec-
where spray water is used to cool the gases) ondary current will sometimes be achieved
will usually nullify the subtle chemical particle before the primary current limit if conduction
composition and provide ample power inputs. angles rise past the 86% point when higher
You can call water vapor a primary condition- levels of impedance are applied in the primary
ing agent that will control resistivity problems circuit.
if the quantity of water used is effectively This brings us to the third limit, which is a
matched to the gas temperature levels of the spark-over between the discharge electrode
flue gas entering the collector. and collecting surface. When this occurs, the
When moisture levels in the flue gas are power supply voltage must be reduced to keep
low^-usually below 10% by volume—the the breakdowns within a reasonable level. This
chemical make-up of the particle becomes a level could range from a nominal 150
dominant factor in controlling electrical char- sparks/min for the inlet field to the occasional
acteristics. spark for the outlet field.
The classification of this characteristic of However, the designer predicated his pre-
the particle is simply related to its ability to cipitator performance on a power parameter
conduct or resist the passage of electric cur- that now may not be attained because of the
rent. This ability is not critical for the individ- limitation imposed by spark-over. Basically,
ual particle as it drifts in the gas stream, but precipitation spark-over can occur by one of
becomes important after it deposits on the two mechanisms:
collecting surface.
One of several power input limits can occur: 1. The impressed voltage is greater than the
spacing and the physical contour or conditions
between electrode surfaces will allow, regard-
1. The voltage limit of the T-R set can be
less of the characteristic of the particulate
reached before any other limitation.
matter. This condition is often observed with
2. Either the primary or secondary winding
electrode misalignment where one or more of
current limit could be reached.
the discharge electrodes has moved too close
3. Or spark-over can occur within the field
to the collecting surface.
limiting the available power from the T-R
Another important factor that can cause
set.
premature breakdowns is the presence of se-
vere discontinuities on the collector surface
These three limits should be well understood. opposite the corona-emitting zones of the dis-
First, the voltage limit is rarely observed on charge electrode. This type of breakdown tends
normal precipitator applications, but it can to provide a greater electrical disturbance
occur with an appreciable mismatch of the compared to the spark-over caused by high
T-R set to the load requirement. That is, it resistivity.
can occur when a large capacity of supply is 2. Spark-over caused by high resistivity lev-
connected to a small surface area field and is els is the most common reason for low power
combined with high concentrations of finer inputs to the precipitator. The resistance of
sized particles. the layer of collected material on the collect-
 THE ELECTROSTATIC PRECIPITATOR: APPLICATION AND CONCEPTS 761

ing surface is the prime reason spark-over will lower power level. Once the dust resistivity
occur. This layer will develop a voltage drop reaches a critical point, its deposition on a
based on three factors: the resistivity value X sharp edge, or for that matter any kind of
the layer thickness X the current density. If discontinuity on the collecting surface, will
the voltage drop is greater than the dust layer cause a localized electrical stress build-up
can withstand, then breakdown within the layer point that will draw the spark. This is why
occurs. uniformity of alignment and elimination of all
This phenomenon is not unlike the break- internal irregularities becomes more impor-
down of a capacitor. Reduction of the layer tant as the resistivity moves up from the
resistivity can be achieved by a number of moderate range.
methods including flue gas additives and pro- It is difficult to specify various resistivity
cess modifications. Sufficient reduction of the levels as denoting good or bad operation. The
layer thickness is often difficult to obtain be- poor physical design of the precipitator com-
cause as resistivity increases, so does the ponents from a high-voltage standpoint can
tenacity of the particles to stick together and alter spark-over levels. It is advantageous to
adhere to the collecting surface. group resistivity into three basic zones; low,
The third component of the voltage drop is moderate, and high. The moderate range
the current density. This means that the would generally encompass a resistivity from
amount of corona current attempting to pass 109 to 1011 ohm-cm and is considered the best
through the dust layer must be reduced if zone for effective precipitation. A finer group-
either the resistivity value or layer thickness ing might show the following:
increases. Otherwise, spark-over can occur. For COMMENTS RESISTIVITY RANGE
example, 0.43 ma/m 2 (40 ma/1000 ft 2 ) may 4 7
10 to 10 ohm-cm Usually high conductive
occur with a material resistivity of 1010 ohm- material—hard to
cm. But if the actual resistivity was 10 n ohm- retain— low-voltage
cm the current density might have to decrease fields present.
to 0.27 ma/m 2 (25 ma/1000 ft2) to keep 108 to 109 ohm-cm Sensitive stage where
lack of resistive
spark-over at a reasonable level. In other
characteristics can
words, the higher the resistivity level, the lower sometimes hurt—
the current density must be to keep the pre- especially in
cipitation process functional. It is not uncom- fly ash cases.
mon to see current densities below 0.054 1010 to 10 11 ohm-cm Appears to be the best
range to shoot for—
ma/m 2 (5 ma/1000 ft2) on certain fly ash should show some spark-
applications. over in precipitator.
1012 to 1013 ohm-cm Range usually associated
with low sulfur coals—
16.3.2 Effects of Resistivity on reduced power in all
Power Levels fields can exist.
Over 1013 ohm-cm Not commonly observed in
What this means in a practical sense is that an basic industries with
infinite number of voltage and current read- normal moisture
contents. Can produce
ings can occur in the precipitator that will not severe electrical
in any way match the name-plate data of the disturbances.
T-R sets. The important thing to remember is
that higher resistivity conditions will decrease The description of spark-over can be de-
power inputs because of the spark-over limita- fined as an electrical breakdown through an
tion. Superimposed on a resistivity problem is isolated gas path between the negative and
the possible condition of the internal structure positive electrodes. The case of the threshold
causing the spark-over to occur at a much resistive spark-over where the discharge
762 HANDBOOK OF POWDER SCIENCE

streamers occur from the deposited dust layer primary voltmeter, a current may indicate 750
is considered the start of back corona. This ma with low resistivity, 400 ma in the moder-
situation is not unlike that which occurs in ate range, or 150 ma in a higher resistivity
atmospheric lightning when positive streamers range. In this example, the 400 ma condition
from earth actually draw the localized stroke would probably provide the better collection
from the negatively charged clouds. But in the performance on a higher velocity precipitator.
case of very high resistivity, a severe back- It is always advantageous to work toward
corona condition can occur characterized by higher voltage gradients and take whatever
greatly reduced voltage and high current den- corona current results. The only exceptions
sities without spark-over. would be current suppression caused by dis-
When low resistivity exists with low-voltage charge electrode build-up or excessive space
conditions, it is difficult to achieve high collec- charge caused by high concentration levels of
tion because of dust reentrainment losses. fine-sized particles.
Power consumption is high because of the One important concept is that each process
high current flow through the dust layer caused will produce a particulate matter whose resis-
by a practically nil electrical resistance. Inter- tivity will usually decrease rapidly on the low
nal inspections usually show collecting sur- temperature side of a peak, while decreasing
faces devoid of any buildup. During this low at a lesser rate on the high temperature side.
resistivity, it will be difficult to achieve the Figure 16.3 shows a typical plot of resistivity
guaranteed efficiency at even half the design obtained by laboratory analysis for dust enter-
gas velocity. ing a cement precipitator. A typical fly ash
Moderate resistivity will allow dust particles from an eastern bituminous coal source with
to bond together in the dust layer by forming a
charged dipole relationship not unlike those
found in a magnet. The opposite polarities
provide good adhesiveness at the tangent con-
tact points of adjacent particles and even aid
in holding these particles together as they \
<
1010

become dislodged by the rapping procedure.

Reentrainment losses are at a minimum level.
When the resistivity increases into the high
range, the bonding can become severe and 109

layer dislodgment by rapping is difficult. How-

ever, at some point the reentrainment portion
of the total precipitator losses can be higher
)
CM

than in the moderate range and are caused by

Resistivity - OHM -

the reduced power levels. A sizeable part of

1
the precipitator dust loss in high-resistivity
cases can also occur during high spark-over \ \
conditions because each spark blows out a
small volume of dust from the layer.
The Deutsch equation showed that the mi-
\
gration velocity contained the product of two
precipitator voltage gradients without regard
to the corona current component. The magni- 400 500
Temperature *F
tude of the current flow will partially depend
Figure 16.3 Resistivity curve for dust at inlet to cement
on the resistivity conditions for any given volt- precipitator: Lab measurement at 4000 V and 25%
age. For a fly ash example showing 280 V on a
 THE ELECTROSTATIC PRECIPITATOR: APPLICATION AND CONCEPTS 763

2% to 3% sulfur might show a similar pattern The degree of misalignment allowable in a

at approximately 6% moisture by volume. precipitator is dependent on a number of fac-
tors. In highly resistive conditions, deviations
of 6 mm (\ in.) in some discharge electrodes
from the center of a passage could provide
16.4 OPERATION AND MAINTENANCE
sensitive electrical breakdowns at slightly re-
duced voltages. Other installations can stand
This section intends to provide many practical
electrode misalignment of 25 mm (1 in.) and
items and advice for personnel who continu-
more, and the effect on performance cannot
ally work with the precipitator. All precipita-
be easily observed. Nevertheless, significant
tors have common trouble areas, yet all
misalignment between the high voltage and
precipitator designs will perform well given
collecting electrodes, whether with weighted
certain operating conditions. However, it is
wire or rigid frame, should be cause for alarm.
the recurring trouble areas that nullify good
performance and often cause costly produc- 16.4.3 Particulate Removal from
tion losses because of the outages required to Hoppers
correct these difficulties.
Difficulties with the evacuation of material
from hoppers can provide a continuous
16.4.1 Internal Inspection headache for plant personnel. The large-sized
The internal inspection of the precipitator is precipitator can have a good potential perfor-
an important operating and maintenance tool mance nullified by hopper troubles. This does
that can provide many benefits if done thor- not have to occur. Unfortunately, in the quest
oughly. This is where a person who is trained to produce high-performance units in order to
to do the job on a periodic basis can build up a meet stringent regulations, the emphasis on
knowledge of observations and data that will good hopper design was less than the concerns
help define causes of difficulties and even catch with SCA or power input. Hopper difficulties
areas of impending trouble. The internal con- are often fought for years because key modi-
dition of the small precipitator with only one fications are not applied, especially to correct
field may require this careful inspection more the underlying causes. This lack of commit-
than the multifield unit. Internal defects in ment to eliminate hopper troubles has pro-
any one field of the bigger collectors can often duced a sad reputation for precipitators. The
be detected electrically during operation, while problems are many, the solutions often hard
it might be difficult in a small unit to ascertain to come by, but do not hesitate to expend time
whether a change in the voltage-current and money to obtain a satisfactory perfor-
characteristics is due to the process or to an mance of the hopper system.
internal defect. Process characteristics, the type of material
being handled, and moisture and gas tempera-
ture levels are important inputs in understand-
16.4.2 Alignment of Electrode Systems
ing the reasons for hopper problems. Without
The net effect on the electrical readings by the a reliable evacuation system from the hopper
characteristics of the gas and particulate mat- flange to the storage area, the burden on the
ter will often be contingent on the proper hopper components may be too much. Based
spacing or alignment between the electrodes. on many field experiences, here are several
Meter readings may indicate a resistivity prob- main factors.
lem, while close spacing or even a specific
electrode design may be causing the spark- 16.4.4 Insufficient Heat and Insulation
sensitive precipitator. There is no substitute Lack of proper insulation and heat input, es-
for careful measurements and inspections. pecially at the bottom apex of the hopper, has
764 HANDBOOK OF POWDER SCIENCE

probably accounted for most of the troubles effect of the hoppers. The detector should
experienced. Regardless of the gas and dust recognize this effect and be placed a little
characteristics, the ability to keep the wall lower at the center of the wall or a little
surface temperature of the lower hopper no higher near the corners.
less than 250°F is most important.
16.4.7 Outage Clean-Down of Electrodes
16.4.5 Air Inleakage into Hoppers
If good shut-down procedures are followed on
Entry of outside air into any part of the hop- most installations, the degree of build-up on
per system is considered poor practice. Aside electrode surfaces will usually require no fur-
from the effect on performance, excessive air ther cleaning during the outage. That does not
can cool down the inside wall surfaces, or mean than 6 to 10 mm (\ to f in.) mounds of
condense moisture in some of the high water- deposit will not exist, but the build-up will be
vapor installations. Unfortunately, most of the spotty with most of the surface holding less
screw conveyor installations coupled with than 3-6 mm Q to \ in.) thick mounds. There
process precipitators are conducive to this are exceptions, especially caused with high-re-
condition. sistivity materials or other operating charac-
teristics.
Whether any manual cleaning is imple-
16.4.6 Level Detectors
mented during the outage depends on several
While it is always better to place time and factors. If it is an annual outage with certain
money in the prevention of hopper difficulties, planned work on the electrode system, then a
the detection of build-ups by some method is water wash of the unit might be considered. It
desirable in most applications. These devices is not recommended to use this type of clean-
can utilize gamma radiation, sound, capaci- ing unless it is necessary to perform major
tance, pressure differential, temperature, or work on the system. Depending on the time
even paddle-wheel methodology for the detec- of year and thoroughness of the washing,
tion of excessive build-up. Any method that some rusting and corrosion pockets can be
does not require components within the accelerated.
hopper appears the most desirable.
Several comments: Use detectors as mainte-
16.4.8 Important Troubleshooting
nance tools rather than to identify full hop-
Approaches
pers. For example, if an automatic batch cycle
allows a maximum 90 cm (3 ft) of build-up in Because of the high-voltage danger, familiarity
the inlet hoppers, it is well to locate the detec- with all the safety aspects of the system cannot
tors no more than 150 cm (5 ft) above the apex be overstressed. Even portions of equipment
flange. The object is to alert the operator inside each control cabinet will be at a 480 V
before a major hopper fill-up exists, yet mini- potential, so care must be taken in any mea-
mize frequent detector alarms. A rule of thumb surement procedure. The manufacturer's man-
would allow the lapsed time from normal dust ual should be well studied for guidelines in the
height to alarm level to equal the same length handling of certain control difficulties.
of time it takes the dust to rise from the apex If any troubles occur initially with control
to the normal height. Remember that the circuit components, fuses, or any other low
pyramid design allows for a greater volume of voltage trouble source, correct these problems
material to accumulate in each foot of hopper post-haste, since the high-voltage portion of
height. One problem arises in the uneven the precipitator tends to supply enough poten-
build-up that occurs by the slope and corner tial difficulties of its own.
 THE ELECTROSTATIC PRECIPITATOR: APPLICATION AND CONCEPTS 765

16.4.9 Normal Versus Abnormal readings will generally tell a story of normal
Power Characteristics precipitator performance as well as abnormal
An understanding of the electrical readings of conditions. A sound knowledge of the effective
the precipitator must be a starting point in voltage-current characteristics can allow one
coping with this collector. It was already to judge emissions on reproducible process
stressed that the name-plate electrical values operations almost as well as actual stack tests.
will not be observed on many fields of the The primary winding voltage and current
precipitator, so the patterns of meter readings readings should provide valid reflections on
become an important tool of evaluation. what is occurring in the secondary circuit of
The key word is "uniformity" of patterns of the high-voltage transformer. However, the
each precipitator because gross power values presence of secondary as well as primary me-
used to compare one unit to another some- ters does provide added monitor capability. In
times provide questionable evaluation results. all difficulties within the precipitator, the two
It is recommended that comparison of the voltmeters and the two ammeters will work in
voltage to current flow value of each field be unison for specific characteristics. That is,
ascertained under normal conditions as well as when the primary voltage is low, the secondary
process variations. voltage should also be low, while the amme-
By now, you should well understand that the ters could both be showing relatively high
control panel readings are a reflection of ev- values (see Fig. 16.4).
erything that is occurring in the precipitator. Probably 80% to 90% of the problems that
The magnitude as well as the trends of the occur in precipitators will tend to reduce volt-

Date
Tirr eof
Re; ding Set
No.
Load

Gas
Temp.
V 1

300 0.3 ^- Plot of Voltage

#— —c^

A
^^ Plot of Current
250 0.2

X"" ^x
200 0.1

300 0.6
—*^.

B
-X-*.
250 0.4 -X-
x- -X-.

200 0.2

280 This Rise in Current and

0.6
Drop in Voltage would
X-. - X - - X - -x- - X - - X -
-x Signal Trouble
C
230 0.4

180 0.2

Figure 16.4. Suggestion for a daily plot of voltage and current to help detect start of troubles.
766 HANDBOOK OF POWDER SCIENCE

ages and raise currents at the same time. current patterns between fields and adja-
These problems are usually associated with a cent cells.
multitude of difficulties with electrode failure, 6. Effects of resistivity can completely nulify
dust build-ups in hoppers, and electrical leak- the effects of the space charge on the inlet
age over insulator surfaces. to outlet patterns.
As each field of a multifield precipitator
does its work, the reduction of suspended par-
ticulate matter in the flue gas will alter the 16.4.10 Reentrainment
voltage-current relationship from inlet to out- Some additional comments in the area of
let. This phenomenon is best observed in the reentrainment of material are warranted. The
moderate resistivity range. A change in the effects of resistivity on reentrainment have
resistivity of the material in each field can already been mentioned, but the subject is
alter the patterns, but the slope of the pattern
much more complex. How far the material
is mostly space charge oriented. For example:
moves out into the gas system, where it rede-
PRI. SEC. SPARKS/ posits in relation to its original position, and
VOLTS MA MIN whether it changed its physical character are
but some of the unknown factors in the reen-
Inlet 360 400 50 trainment syndrome.
Center 330 550 20
A common description of the dust layer
Outlet 300 750 occ.
sliding down the collecting surface does not
usually occur. The shock or tremor imparted
In other words, for identically sized fields, we
by the rapper appears to more often dislodge
are generally looking for a stepped decrease in
voltage and stepped increase in current from some percentage of material from its resting
inlet to outlet. Some key concepts in meter place. If the particles have had a chance to
observations include: agglomerate, the adverse effects of reentrain-
ment are minimized. It is when the particles
1. Patterns in voltage and current readings bounce back into the gas stream in the same
from inlet to outlet should form some type condition as they were collected that troubles
of recognizable pattern unless internal begin to mount. This is where proper resistiv-
defects cloud the issue. ity and the timing between raps can play an
2. The high-resistivity range can produce a important part in this interesting phase of
relatively low flat voltage and current pat- precipitation.
tern, but only in the very high resistivity Some key concepts include:
zones (1012 and above). Generally you 1. Always use the internal inspection and
should see an increasing pattern of current other visual means to help ascertain the lowest
flow in the direction of gas flow. rapping intensity possibly commensurate with
3. The moderate range of resistivity would other performance observations.
show the greatest magnitude of change from 2. Always attempt to match the rapping to
inlet to outlet. the dust characteristics or resistivity. For ex-
4. As the resistivity becomes low enough so ample, a low resistivity requires soft rapping,
that all sparkover ceases from this cause, the moderate range requires a harder blow,
the voltage and current readings tend to and the high-resistivity zone means real trou-
flatten out again from inlet to outlet, but at ble. Remember that hard rapping with high-
a much higher power level. resistivity materials usually exhibits limited
5. It is only in the moderate to high resistivity success and changing the resistivity is usually
ranges that internal electrode defects will a much better way to achieve a satisfactory
show major distortions in the voltage- performance.
 THE ELECTROSTATIC PRECIPITATOR: APPLICATION AND CONCEPTS 767

3. Do not feel that the inlet field must be than 1.8 m / s (6 ft/s) through the precipi-
overrapped because it handles the bulk of tator.
material. Be aware that puffs out of the stack 3. Reentrainment dust losses above 50% of
or other signs of reentrainment are not unique the total ESP emission losses.
to the outlet fields. Actually, the material col- 4. Low-resistivity characteristics are apparent
lected in the inlet field of some precipitators is with an absence of spark-over.
often easier to dislodge, and excessive carry- 5. High-resistivity characteristics are apparent
over adds to the reentrainment potential of with less than an average of 0.11 ma/m 2
the following fields. (10 ma/1000 ft2) of collecting surface.
4. Rapping loss is not usually uniform across
the precipitator, even discounting the effects Field observations have pointed out a number
of resistivity gradients. When reentrainment of concepts:
losses are observed, investigations into possi- 1. The gas flow vectors in a dynamic system
ble problems with gas distribution is highly will tend to keep going in the direction pointed,
recommended. until striking another obstruction. This con-
5. When reentrainment losses are severe, cept is fundamental to the understanding of
lengthening the time between raps on the col- why some installations have problems.
lecting surfaces in the direction of inlet to 2. The velocity of the gases entering an
outlet is usually recommended. As a first ad- expansion plenum will determine the final pat-
justment it might be advisable to double the terns at the face of the precipitator. If there is
rap time on succeeding fields. For example, if a poor vector pattern at the entry of the
the inlet field rapped every 5 min, then the nozzle, then higher flow rates will usually ag-
second field would be rapped every 10 min, gravate the distribution by the time the gases
the third field every 20 min, and so on. reach the precipitator.
6. The more the power characteristics are 3. The 40% to 50% open area diffuser
improved, the better the chance for reentrain- plates will provide little correction of a poor
ment losses to diminish. Rappers should al- gas pattern if the pressure drop across the
ways operate across one field before the cycle plate is less than 13 mm (0.5 in.) H 2 O. How-
moves on to the next field. As discussed in the ever, these plates will reduce the rolling action
text, excessive dust disturbances on the col- of the gas, and most of the kinetic energy will
lected layer can lead to adverse electrical field be transferred into smaller jets. Generally, be-
activity in certain cases. Allowing the surface low the 3.0 to 4.6 m / s (10 to 15 ft/s) range,
contour to smooth out slightly between raps only minimal benefits will accrue in the gas
can relieve localized stress points and reduce spreading effects of the low pressure drop
the spark-over potential. Usually 5 to 10 min diffuser.
time duration is needed to observe this 4. With a 40% to 50% open area diffuser
phenomenon where it will occur. plate, which is commonly used, any gas vectors
striking the plate at 45° or more from the
16.4.11 Gas Distribution perpendicular will have a sizeable fraction of
Whether gas distribution is effecting the pre- that gas flow slide across the plate.
cipitator performance adversely can be related 5. Any flue expansion with more than about
to many factors. Efforts are usually worthwhile an 8° slope will generally have some separa-
in exploring improvements in the gas distribu- tion of gas from the surface. The common
tion pattern if one or more of the following practice of 30 to 45° plenum expansions tends
conditions exist: to present distribution problems for that rea-
son.
1. Aspect ratios of 1.0 or less. 6. Any process whose flue gases contain
2. Average calculated gas velocity of more particles over 30 microns in diameter could
768 HANDBOOK OF POWDER SCIENCE

get into distribution troubles by the settling of of gas conditioning. Water addition to many
dust in the expansion plenum. This condition process gas streams is often part of the opera-
becomes worse during long periods of reduced tion. Water forms part of the raw material in
process operation with its attendent low gas some cases, or in others, is primarily used to
velocities. control gas temperature levels at the discharge
of the process. Fuel supplies another source of
moisture. As previously discussed, moisture
16.5 GAS CONDITIONING contents over 20% by volume tend to nullify
resistivity problems depending on the gas tem-
The preferred method to improve the perfor- perature range at the precipitator. The use of
mance of existing precipitators involves the steam is much less utilized because of the cost
use of higher power inputs. Poor resistivity factor, but it is useful on a short-term basis
levels can be overcome by the modification of where water may present condensation
the flue gas characteristics. The term gas con- problems.
ditioning normally refers to the various meth- The use of chemical additives offers a sec-
ods used for injection of chemical constituents ondary approach if the time factor or eco-
into the flue gas stream, primarily to help alter nomics dictates that any modification of the
resistivity levels in the precipitator. This term process is not a satisfactory route. Fly ash
should include any method, whether or not it collection has been the greatest area of imple-
is inherent in the process or supplied from an mentation for this method in recent years.
external source. There are a number of companies and tech-
niques available in the gas conditioning field
16.5.1 Concepts and success has been achieved on difficult
installations.
The process should be first explored to deter-
mine if inherent changes in operation equip-
ment can modify resistivity levels. Some of 16.6 DESIGN AND PERFORMANCE
these techniques include: CONCEPTS

1. Substitute, blend, or prepare some of the The simple Deutsch equation is a valid way to
bad actors in the raw materials or fuel in a understand how the various critical inputs can
manner conducive to precipitation. This affect the performance of the precipitator. As
may require some modification of the mentioned earlier, the exponent can be low, as
handling equipment. much as a factor of two, because of a number
2. More efforts on the maintenance of mois- of problems that the designer did not foresee.
ture levels in the flue gas is important. Just Excessive reentrainment and poor gas distri-
the elimination of inleakage air will have bution were two of the prime reasons for the
this net effect. disparity between the theoretical and actual
3. Awareness of the temperature effect on results.
resistivity must be uppermost for any pro- Recent designs have taken the migration
cess change. Even the elimination of high exponent to another \ power or less to correct
to low gas temperature zones may help for previous problem areas and provide addi-
moderate a poor performance to one that is tional margin for the fine-sized particles exist-
acceptable. ing in the latter fields of the precipitator.
What this means is that a doubling of physical
Probably the use of additional moisture in precipitator is indicated compared to what was
the flue gas by way of water sprays or steam considered a standard design of the early
injection can be considered a primary method 1970s. Whether this is warranted is based on
 THE ELECTROSTATIC PRECIPITATOR: APPLICATION AND CONCEPTS 769

two factors: whether or not good design con- remove from the flue gas stream. Although
cepts are applied and, second, the confidence there is some validity to this concept, too
of the user that he can exercise some control much is made of this point in the practical
over the process gas and dust characteristics. application of the precipitator. It is difficult to
In retrospect, the design of 8 to 12 years ago analyze effectively because the particle size,
could meet its guarantees if conditions were shape, and chemical make-up interact in many
optimized. Designs of 40% to 50% greater diverse ways.
surface area over those of the past now appear Each basic industry tends to produce partic-
quite reasonable if you weigh all the factors of ulate matter from some form of a grinding,
today's environment. This still means carefully combustion, or condensation process. Nor-
addressing the process characteristics and ap- mally, the discrete larger particle of material
plying a commitment to proper operating and found in a flue gas will be more irregular in
maintenance techniques. shape and will be more chemically associated
Each precipitator field should be considered to that of the process raw material. Particles
a separate collector unit, and for that matter, formed by condensation in the process tend to
each gas passage of a field must perform well be submicrometer in size and more spherical
to attain the best bottom line of the overall in shape, while often deviating from the chem-
system. For high collection efficiencies to be ical characteristics of the larger particle found
achieved, the inlet field must perform near in the gas stream of the same process.
design levels usually in the nominal 80% range. The effect of the particle size on the electri-
Theoretically, that means about 80% of the cal precipitation can be identified in a number
particulate matter would deposit in the front of ways:
hopper. This is why the inlet field looms im-
portant in any upgrading program. 1. The larger the particle the more electrical
I would stress a few points: charge can be accumulated on its surface,
1. Each succeeding field works on the resid- and this condition provides an increased
ual of the preceding field, but the potential velocity of the particle toward the collector
collection efficiency tends to decrease in the surfaces of the precipitator.
direction of flow. Part of the reason is that 2. Electrical precipitation probably performs
collection values are harder to achieve as the the least on a particle size about one-half
magnitude of particles decrease. micron in diameter. Collection of particle
2. Another reason is that the particles that sizes less than one-half micron improves
are left in the gas stream in the latter half of with benefits of Brownian motion in the
the precipitator are more difficult to collect vicinity of the collection site, while the
since they usually consist of the finer sized larger sized particles benefit from the
segment. Unless current densities above 0.22 greater levels of charging.
ma/m 2 (20 ma/1000 ft2) are observed in these 3. However, it appears that a large number of
latter fields, their collection efficiencies can the smaller particles tends to adhere to the
deviate substantially from design. larger particles, so that it is difficult to
separate the practical effect of the sizing
segments on the overall efficiency of
16.7 EFFECT OF PARTICLE SIZE collection.
4. A population of particles that is more ho-
The effect of the particle size on precipitation mogeneous in sizing will often make the
is seen in the component relationship that deposited layer of material on the collector
represents the migration velocity of the surfaces more difficult to dislodge by rap-
Deutsch equation. This exponent indicates that ping forces. As a rule, the larger particles of
the smaller sized particle is more difficult to material, because of the effect of greater
770 HANDBOOK OF POWDER SCIENCE

porosity in the layer, will allow for easier 2. M. Robinson, "Electrostatic Precipitation," in Air
removal by the mapping mechanism. Pollution Control I, edited by W. Strauss, Wiley-
Interscience, New York, NY (1971).
5. Both size and chemical segregation of parti- 3. J. D. Cobine, Gaseous Conductors, Dover Publica-
cles will tend to occur throughout the length tions, New York, NY (1958).
of the precipitator. The outlet electrical
fields will often contain a greater percent- Operation and Maintenance
age of the finer sized particles as well as the
J. Katz, The Art of Electrostatic Precipitation, Scholium
more chemically active material, such as International Inc. Port Washington, NY (3rd print-
condensed alkali and acidic ingredients. ing 1989).

Manuals and Publications on ESP's

REFERENCES 5. The National Technical Information Service,
Theoretical Background Springfield, VA.
6. The Mcllvaine Company, Northbrook, IL.
1. H. J. White, Industrial Electrostatic Precipitation, 7. Air Pollution Control Association, Pittsburgh, PA.
Addison-Wesley, Reading, MA (1963). 8. Industrial Gas Cleaning Institute, Alexandria, VA.
 17
Granular Bed Filters
PART 1. THE THEORY

Gabriel I. Tardos

CONTENTS
17.1.1 INTRODUCTION 771
17.1.2 TOTAL BED EFFICIENCY 772
17.1.3 COLLECTION MECHANISMS IN DEEP-BED FILTRATION 773
17.1.4 EXPERIMENTAL VERIFICATION 776
17.1.5 CONCLUDING REMARKS 778
REFERENCES 780

17.1.1 INTRODUCTION drop. This in turn causes a compression of the

deposited layer and hence results in a higher
The separation of airborne dust in granular nitration efficiency as more and more dust
beds takes place in either a "cake" or a "non- accumulates at the surface of the cake. The
cake" (deep-bed) filtration mode depending on efficiency of the filter in cake filtration is over-
the region in the filter in which particle depo- whelmingly a function of the deposited layer's
sition actually occurs. During cake filtration, pore size and increases dramatically with pres-
as the name implies, initially deposited dust sure drop. If the dust particle size is larger
layers serve as collection media for subsequent than the pore size, dust is filtered and the
filtration and the granules in the bed serve efficiency is very high (practically 100%). How-
only as a support for the separated dust. The ever, if the dust is smaller than the open pore
main mechanism of particle separation is siev- size, a cake is not formed and deep-bed filtra-
ing: incoming dust particles are retained on tion takes place.
the already deposited dust. This results in a During noncake or "deep-bed" filtration,
significant increase in thickness of the de- dust particles are captured on each and every
posited layer as filtration proceeds and is ac- one of the granules or collectors of the filter.
companied by a large increase in pressure As filtration progresses, deposits of dust slowly
771
772 HANDBOOK OF POWDER SCIENCE

fill the interstices of the granular bed starting where L/2a is the number of collector layers
with the contact points between granules with- in the filter, e is the relative void volume, and
out drastically altering the geometry of the E is the so-called single collector efficiency.
filter or the pressure drop through the bed. The quantity E is defined as the ratio of the
The filtration in this case is overwhelmingly number of all airborne (dust) particles cap-
influenced by the size of dust particles and by tured by a single collector in the bed to the
the thickness of the granular filter in the di- total number of dust particles flowing toward
rection of the flow. it in a circular tube of cross-sectional area
The theoretical calculations presented in ira2. The implicit assumption in Eq. (17.1.4) is
this section pertain only to the case of deep- that all collectors act as if they were indepen-
bed (noncake) filtration in granular packed, dent within the filter as shown in Figure 17.1.1
moving, or fluidized beds. In these cases dust and hence experience similar filtration phe-
is collected either inside the filter or distinct nomena. Equation (17.1.4) can be used in a
collectors or particles deposit on each other predictive way provided the single collector
without significantly altering the geometry of efficiency E can be calculated from first
the filter as dust collection proceeds. The pres- principles.
sure drop in the filter, A/?/L, under these A somewhat different but in principle
conditions can be calculated from the well- equivalent way of computing the total bed
known Ergun correlation,1'2 which in dimen- efficiency is to use the concept of the unit cell
sionless form is given as: efficiency, e, so that:
/ 0 [ e 3 / ( l - £)] = 180(1 - e ) / R e 0 + 1.8 v=l-[l-e]
n
(17.1.5)
(17.1.1)
The quantity e is defined as the ratio of the
The actual pressure drop per unit thickness of number of airborne dust captured by a collec-
filter is then evaluated from the equation: tor (granule) to the total number of dust parti-
Ap/L=foPU2/2a (17.1.2) cles flowing toward it in a square duct of
cross-sectional area, I2, where the length / is
where the Reynolds number is expressed as
given by:
Re 0 = 2aUop/fjL, L is the thickness of the
filter in the direction of the flow, a is some /= 2[TT/6(1 ~ e)]1/3a (17.1.6)
average granule radius, and Uo is the superfi-
The quantity n is the number of layers of unit
cial gas velocity in the filter.
cells in the filter, n = L/l. Comparing Eqs.
(17.1.4), (17.1.5), and (17.1.6), the ratio of the
17.1.2 TOTAL BED EFFICIENCY

The efficiency with which dust is collected in a

filter, 77, can simply be calculated from the
concentration of airborne dust entering nin, porticle trojectory
and leaving the filter nout as: limit troiectory

V = 1 - "inAout (17.1.3)
Extensive studies of deep-bed filtration in both
granular and fabric filters have revealed that
the total efficiency is an exponential function
of the filter thickness and this can be ex- = {b/a) z

pressed by the equation: Figure 17.1.1. Schematic representation of dust parti-

cle deposition on a sphere. (Copyright AIChE. Vol. 31,
77 = 1 - exp[-1.5(1 - e)(K/2a)E] No. 7, p. 1095, July (1985). (Reproduced with per-
(17.1.4) mission.)
 GRANULAR BED FILTERS 773

single collector and the unit cell efficiencies is impact and stick to the collector. The forces
given by: that are most frequently associated with filtra-
2/3
tion in granular beds are inertia, diffusion,
e/E = 1.2(1 - e) (17.1.7) gravity, and electrical effects. While inertial
Whereas the definition of the single collec- and gravitational forces are characteristic of
tor efficiency is somewhat arbitrary and its large particles of the order of microns and
value can exceed unity in some cases (this may tens of microns, diffusion becomes important
be difficult to justify on purely mechanistic only for very fine particles in the submicron
grounds) the unit cell efficiency has a clear range;4 electrical forces, if present, are effec-
physical meaning. For a detailed discussion of tive in the whole range of particle sizes. For
the different efficiencies and their definitions, relatively small particles and in the absence of
the reader is directed to the exhaustive mono- electrostatics, the so-called interception effect
graph on granular filtration by Tien.3 The becomes important. This is a purely geometric
remainder of this section is dedicated to ways "mechanism" and is due to the finite size of
of calculating the single collector efficiency E the dust particles, that is, even if the particles
which hence allows the prediction of the total follow the fluid stream lines exactly some
efficiency, rj. stream lines will approach the collector to a
distance smaller than the radius, rp, of the
dust particle, as can be seen in Figure 17.1.1,
17.1.3 COLLECTION MECHANISMS IN thereby causing deposition.
DEEP-BED FILTRATION Table 17.1.1 presents a summary of the im-
portant mechanisms that cause deposition in a
Collection of small airborne dust by granules granular bed; each of the mechanisms is gov-
(collectors) in a packed or fluidized bed is due erned by a characteristic dimensionless num-
to external forces that cause the dust to devi- ber that is defined in the second column of the
ate from the fluid stream lines and thereby to table. Because electrical effects are caused by

Table 17.1.1. Collection Mechanisms in Granular Beds.

CHARACTERISTIC
DIMENSIONLESS
MECHANISM NUMBER EQUATION REMARKS
3 2
Interception Rp = rp/a ER= 1.5g (e)R p Re o < 1
Interception ER ^ O/e)Rp Re0 < 30
parameter Geometric effect6
Diffusion Pe = 2aU0/DB ED = 4g(e)Pe 2 / 3 Re0 < lb
Peclet number ED = 4.52/UPe) 1 / 2 Re0 < 30
Gravity Ga = ag/Ui EG = GaSt Independent of flow to
Galileo number a first approximation
Inertia St = 2CPpUor^/9fjLa Ex = 2St'3-9 St' = St[l + 1.75 Re 0 /
Stokes number (4.34~ 6 + St'3-9) 150(1 - e)]c
0.1 < St' < 0.03

Electrical Ke Eel= -4Ke Coulombic force only

effects Electrical number0 Eel = Kex/(1 + * e x ) External electric field
only
a
See definitions in Table 17.1.2
See expressions for g(e) in Table 17.1.3.
c
See other expressions in Table 17..1.4.
774 HANDBOOK OF POWDER SCIENCE

a combination of charges present on the parti- tive by using the expressions in Table 17.1.1, in
cle, the collector, or both, the characteristic reality a combination of effects almost always
electrical number Ke is given separately in exists. A general practice in this case is to add
Table 17.1.2. Table 17.1.3 is a summary of the predicted values for each individual mech-
expressions for the correction factor g(e) that anism by using the equation:
appears in the equations describing intercep-
tional and diffusional efficiencies, while Table E = 1 - (1 - £ R )(1 - £ D )(1 - EG)
17.1.4 contains theoretical and experimental X(l -E.Kl -£el) (17.1.8)
relations to calculate the efficiency due to
inertial effects. As seen in Table 17.1.1, the which, if all efficiencies are small compared to
Reynolds number Re 0 = 2aU0/v enters ex- unity, simply becomes:
plicitly only in the expression of the inertial E1 Ee{ (17.1.9)
E s ER
disposition; one has to note, however, that
expressions for interception and diffusion are The assumption behind Eq. (17.1.8) is that
different for low and high Reynolds number different mechanisms act independently; this
flows as shown in Table 17.1.1. was demonstrated to be true for the case of
Although it is quite simple to predict filtra- diffusion, interception, and inertia;24 intercep-
tion efficiencies if only one mechanism is ac- tion and gravity; and interception, diffusion,

Table 17.1.2. Electrical Forces Between Particles and Characteristic Parameters, Ke.

FORCE, F e DESCRIPTION PARAMETER Kp

Coulombic force Both collector and
particle are
charged.

Charged-particle Particle only is Kic = ycCQ2p/247r2e{rpa2fjiU0

image force charged. Charge
separation induced
in collector.

Charged-collector Collector only is

image force charged. Charge
separation induced
in particle.

External electric Particle only is Kex=CQpE0/67rrpfjLU0

field force charged. Charge
separation in
collector induced
by external
electric field.

icp Electric dipole Neither body is Kicp =

interaction force charged. Charge
separation in both
bodies induced by
external electric
field.

ef, Dielectric constant of fluid;

e p , dielectric constant of particle;
ec, dielectric constant of collector.
y c = ( e c - € f ) / ( e c + 2e f ).
yp = ( e p - e f ) / ( e p + 2e f ).
 GRANULAR BED FILTERS 775

Table 17.1.3. Values of Correction Factor g(c).

AUTHOR RANGE
5 5/3 1 3 5 3 2 1 3
Pfeffer {2[1 - (1 - 6> ]/[2 - 3(1 - 6) / + 3(1 - e) / - 2(1 - e) ]} / Re0 < 0.01
Pe > 1000
6
Tardos et al. {e/[2 - e - (9/5X1 - e) 1 / 3 - (1/5X1 - e) 2 ]} 1/3 Re0 < 0.01
Pe > 1000
Sirkar7'8 {[2 + 1.5(1 - e) + 1.5[8(1 - e) - 3(1 - e)2]^2]/e[2 - 3(1 - e)]}1/3 Re0 < 1
£ > 0.33
Pe > 1000
Tardos et al.9 1.31/c 0.3 < e < 0.7
Re0 < 0.01
Pe > 1000
Tan et al.10 1.1/6 Re0 < 1
0.35 < e < 0.7
Wilson and Geankoplis11 1.09/e Re0 < 10
0.35 < e < 0.7
Thoenes and Kramers12 1.448/6 Re0 < 10
e = 0.746
Karabellas et al.13 1.19/6 Re0 < 10
e = 0.26
Sorensen and Stewart14 1.104/e e = 0.476
1.17/6 6 = 0.26

gravity, and weak electric effects.25 Strong has to be mentioned. It was observed by many
electric effects due to Coulombic attraction researchers that at relatively high gas veloci-
and strong external electric fields (see Table ties or large particle sizes, while inertial ef-
17.1.1) cannot be combined with inertial ef- fects ensure that dust particles collide with
fects and have to be considered separately.26"28 collectors following their tortuous way through
To complete the picture of collection of the filter, the dust is in fact not collected and
small airborne dust by a granule (collector) in instead bounces off on contact and is, in the
a granular bed, the phenomenon of bounce-off end, not retained by the filter. This behavior

Table 17.1.4. Empirical Correlations for Single-Sphere Efficiency Due to Inertial Effects.

AUTHOR RANGE
15 1 13
Paretsky 2 X St - St < 0.01
Meisen and Mathur16 0.00075 + 2.6 X St St < 0.01
Doganoglu17 2.89 X St dc < 100 micron
0.0583 X Re X St dc < 600 micron
Thambimuthu et al.18 105 X St3 0.001 < St < 0.01
Schmidt et al.19 3.75 X St St < 0.05
Goren10 1270 X St9/4 0.001 < St < 0.02
Pendse and Tiena21 (1 + 0.04 ReXSt]
D'Ottavio and Gorenc21 St3j|5/(1.67 + st3^5 0.33 < e < 0.38
Gal, Tardos and Pfeffer (1985)*23 2St'39/(4.3 X 10"6 + St'3-9) 0.01 < St' < 0.02
a
Interception neglected.
fe
St' = St[l + 1.75Reo/15O(l - e)].
c
Steff = /(Re,e)St.
/(Re, e) = (1 - /* 5/3 )/(l - 1.5/*1/3 + 1.5/z5/3 - h2) + 1.14Re£/2/<-2/3 where h = 1 - e.
776 HANDBOOK OF POWDER SCIENCE

results in a reduced efficiency at particle Stokes are performed at different gas flow rates and
numbers larger than about St > 0.01. Tien3 at different electric fields if electric effects are
introduces the coefficient of adhesion proba- present.
bility T given by Figures 17.1.3 and 17.1.4 show measured28'29
and calculated filter efficiencies using the
T = 0.00318 St~ 1248 (17.1.10) equations give in Table 17.1.1. The dust parti-
cles used in these experiments are of the latex
to account for this effect. For practical calcu- aerosol type, which are commonly used in
lations, the efficiency E obtained from Eq. industry to test filters as mentioned previously.
(17.1.8) has to be multiplied by the factor T if Figure 17.1.3 shows filtration efficiencies, E, as
the Stokes number exceeds the value St = 0.01 a function of gas superficial velocity in a sand
even if the deposition is overwhelmingly in- bed of grain average size of 450 fim. The
fluenced by electrostatic effects. calculated values are for large Reynolds num-
bers (upper line in the figure) and very low
Reynolds numbers (viscous flow) using the
equations of Table 17.1.1 for diffusion, inter-
17.1.4 EXPERIMENTAL VERIFICATION ception, gravity, and inertia. The experimental
values of the Reynolds number are depicted
A schematic representation of the experimen- with arrows on the lower side of the figure. As
tal apparatus to test a granular bed filter is seen, the data follow the calculations as ex-
depicted in Figure 17.1.2, a schematic of the pected: for Reynolds numbers below about
test section is also shown. In the case of an Re 0 = 3, the data fit viscous flow calculations
electrically enhanced filter, a wire mesh elec- well, whereas for values of the Reynolds num-
trode is added to the top of the bed where the ber of the order of Re 0 = 30 and higher, the
electric field is applied and a radioactive measured data follow the calculations for po-
source, used to neutralize the generated
tential flow. One can clearly see the effect of
aerosols (dust particles), is followed by a parti-
bounce-off at superficial gas velocities larger
cle charger (not shown in the figure). The
than about 2 m/s. One has to note here that
complexity of the set-up is required by the
the data presented above are an exceedingly
need to very carefully control the dusty gas
exaggerated case in which the limits of the
flow, the particle and granule electric charge
theoretical calculations are being checked.
(or lack of it), and the granule-wall interac-
Granular filters are usually operated at gas
tion in the bed. Additional problems are also
velocities of the order of 2 to 30 cm/s, where
generated by the sensitivity of the particle
counter (Royco counter in the figure). Filtra- it is clearly seen that calculated values fall
tion experiments usually require the genera- quite close to the measured ones.
tion of a dilute stream of test aerosols (usually Figure 17.1.4 shows results for an electri-
latex particles of known size) which are subse- cally enhanced filter operated with an external
quently passed through the filter at known electric field. The shape of the efficiency curves
flow rate and the concentration in and out of (total efficiencies 77 in this case) are typical of
the bed is carefully measured. These experi- granular filters: efficiencies are high for small
ments are repeated with a whole range of dust particles below 0.1 fim and large dust
specially manufactured test dusts or aerosols particles above 1 jam in diameter and are
of different sizes and sometimes composition lower between these two limits. Increasing the
and electrical properties. To control electric applied electric field results in a significant
charges, the test particles are first neutralized improvement in efficiency even at the high
and then electrically charged to the appropri- gas velocity of Uo = 0.5 m / s as shown in
ate level before entering the bed. Experiments the figure.
 COMI EXCESS
DISCHARGE

TEST
SECTION
MICRO-
MANOMETER

- 9 . 5 mm

Horizontal sand bed test section.

VACUUM
ROTAMETER PUMP

VACUUM
GAUGE

Figure 17.1.2. Schematic of experimental apparatus. (Copyright Academic Press, Inc. Journal of Celloid and Interface Science, Vol. 71, No. 3, October (1979).
(Reproduced with permission.)
778 HANDBOOK OF POWDER SCIENCE

1 on the assumption that on the average each

0.5
Experimental granule in the bed plays the role of a collector
• 3 0 - 4 0 mesh sond <2o = 507.5M) and that overall, the effects of all collectors
0.2- x 40-50 mesh sand (2o
can be integrated to yield an exponential de-
• 1 0 - / -0.1 cay in concentration along the filter. The fil-
tration process is then divided into individual
^***i'-w., * Potential #u i mechanisms by which airborne dust deviates
2- from the fluid streamlines and can, at least in
-0.01 principle, collide with a granule and stick to it.
o Some experimental evidence is given to show
5-
IS that this model is at least somewhat realistic
"o
and that carefully measured efficiencies can in
10
10.001 fact be predicted theoretically with some
1 2 5 10 20 50 100 200 5001000
SUPERFICIAL GAS VELOCITY, Uo [CM/SEC] — degree of confidence.
Figure 17.1.3. Single-sphere efficiency versus superfi- Dust filtration in a cake clearly does not fit
cial gas velocity. Filtration of 1.1 /im latex aerosols. the above model and hence the reader is re-
Theoretical values computed for: bed porosity e = 0.4, ferred to the pertinent literature 30 " 32 for fur-
granule diameter 2 a — 0.45 mm, and particle density
p p = 2g/cm 3 . (Copyright Academic Press, Inc. Journal
ther information. Fortunately, deep-bed filtra-
of Celloid and Interface Science, Vol. 71, No. 3, October tion is almost always the important mode of
(1979). (Reproduced with permission.) separation of small particles whereas cake fil-
tration becomes important when the granular
17.1.5 CONCLUDING REMARKS filter is overwhelmingly clogged with dust and
under those conditions the efficiency becomes
A theoretical approach was outlined to predict very high. Because of pressure drop considera-
both the single collector efficiency E and the tions the operation of granular bed filters in
total bed efficiency 77 of a deep-bed granular the clogged regime is not economically and
filter from first principles. The method is based technically attractive.

100

0 90 /
' a = 15 x 10*3m
2
L = 5 0 x 10* m
1
Uo = 0 5 ms*
0 80 -

E
0 70 - • = 0 6 kV/cm
Eo = 1 0 kV/cm
Eo = 2.0 kV/cm

0 60 --

0 50
0 1 10 100
PARTICLE DIAMETER (/urn)
Figure 17.1.4. Comparison of model predictions with theory (Uo = 0.5 ms" 1 , 2a = 3 mm, L = 5 cm).
 GRANULAR BED FILTE

LIST OF SYMBOLS

a Collector (granule) efficiency

C Cunningham correction factor
Dust particle diffusion coefficient
dc = 2a Collector (sphere) diameter)
e Unit cell efficiency
E Single-sphere (collector) efficiency
Single-sphere efficiency due to diffusion
Single-sphere efficiency due to electrical effect
Single-sphere efficiency due to gravity
Single-sphere efficiency due to inertia
Eo Applied electric field
EK Single-sphere efficiency due to interception
/o dimensionless pressure drop defined in Eq. (17.1.2)
8 Acceleration of gravity
Ga = ag/Ui Galileo number
q(e) Porosity-dependent function given in Table 17.1.3
K Boltzman constant
Kc Electrical number defined in Table 17.1.2
I Unit cell size defined in Eq. (17.1.6)
L Filter bed height
"in Inlet aerosol concentration
"out Outlet aerosol concentration
n = L/l Number of unit cell layers
Pe = 2af/ 0 /D B Peclet number
Qo Collector (sphere) electric charge
Particle (dust) electric charge
Dust particle radius
Interception parameter
Re0 = 2aU0/v Reynolds number
St = 2C P p £/ 0 r p 2 /9 M a Stokes number
T Absolute gas temperature
Superficial flow velocity
Ap Pressure drop through the packed bed

Greek letters
Bed porosity
Dust particle density
Gas viscosity
v = Gas kinematic viscosity
V Total filtration efficiency
P Gas density
y Adhesion probability coefficient defined in Eq. (17.1.9)
780 HANDBOOK OF POWDER SCIENCE

REFERENCES 16. A. Meisen and K. B. Mathur, "Multi-Phase Flow

Systems," Inst. Chem. Eng. Symp. Ser. 38, Paper K3
1. S. Ergun, "Fluid Flow Through Packed Columns," (1974).
Chem. Eng. Prog. 48:89-94 (1952). 17. Y. Doganoglu, Ph.D. Dissertation, McGill Univer-
2. I. F. Macdonald, M. S. El Sayed, K. Mow, and sity (1975).
F. A. L. Dullien, "Flow Through Porous Media— 18. K. V. Thambimuthu et al., Symp. Deposition and
The Ergun Equation Revisited," Ind. Eng. Chem. Filtration of Particles from Gases and Liquids, Soc.
Fund. 73:199-208 (1979). Chem. Ind. (London) p. 107 (1978).
3. C. Tien, Granular Filtration of Aerosols and Hydro- 19. E. W. Schmidt et al., "Filtration of Aerosols in a
sols, Butterworths, Boston and London (1989). Granular Bed," /. APCA 2S(2):143 (1978).
4. G. I. Tardos, N. Abuaf, and C. Gutflnger, "Diffu- 20. L. S. Goren, "Aerosol Filtration by Granular Beds,"
sional Filtration of Dust in a Fluidized Bed," EPA Symp. in Transfer and Utilization of Paniculate
Atmos. Environ. 70:389-394, April (1978). Control Technology (Rpt.) EPA-600-7-79-004 (1978).
5. R. Pfeffer, "Heat and Mass Transfer in Multi- 21. H. Pendse and C. Tien, "General Correlation of
Particle Systems," IEC Fund. 3:380 (1964). the Initial Collection Efficiency of Granular Filter
6. G. I. Tardos, C. Gutflnger, and N. Abuaf, "High Beds," AIChE J. 28(4):677 (1982).
Peclat Number Mass Transfer to a Sphere in a 22. T. D'Ottavio and L. S. Goren, "Aerosol Capture in
Fixed or Fluidized Bed," AIChE J. 22:1146-1149 Granular Beds in the Impaction Dominated
(1976). Regime," Aerosol Sci. Tech. 2:91 (1983).
7. K. K. Sirkar, "Creeping Flow Mass Transfer to a 23. E. Gal, G. I. Tardos, and R. Pfeffer, "Inertial
Single Active Sphere in a Random Spherical Inac- Effects in Granular Bed Filtration," AIChE J.
tive Particle Cloud at High Schmidt Numbers," 57:1093 (1985).
Chem. Eng. Sci. 29:863 (1974). 24. C. Gutflnger and G. I. Tardos, "Theoretical and
8. K. K. Sirkar, "Transport in Packed Beds at Inter- Experimental Investigation of Granular Bed Dust
mediate Reynolds Numbers," Ind. Eng. Chem. Filters," Atmos. Environ. 73(6):853 (1979).
Fund. 14:73 (1975). 25. R. Pfeffer, G. I. Tardos, and L. Pismen, "Capture
9. G. I. Tardos, N. Abuaf, and C. Gutflnger, "Dust of Aerosols on a Sphere in the Presence of Weak
Deposition in Granular Bed Filters—Theories and Electrostatic Forces," IEC Fund. 20 (1981).
Experiments," JAPCA 2S(4):354-363 (1978). 26. K. A. Nielsen and J. C. Hill, "Collection of Inertia-
10. A. Y. Tan, B. D. Prasher, and J. A. Guin, "Mass less Particles on Spheres with Electrical Forces,"
Transfer in Nonuniform Packing," AIChE J. IEC Fund. 75:149 (1976).
27(2):396 (March 1975). 27. K. A. Nielsen and J. C. Hill, "Capture of Particles
11. E. J. Wilson and C. J. Geankopis, "Liquid Mass on Spheres by Inertial and Electrical Forces,"
Transfer at Very Low Reynolds Numbers in Packed Chem. Eng. Commun. 72:1/1 (1981).
Beds," Ind. Eng. Chem. Fund. 5:9 (1966). 28. G. I. Tardos and R. W. L. Snaddon, "Separation of
12. D. Theones and H. Kramers, "Mass Transfer from Charged Aerosols in Granular Beds with Imposed
a Sphere in Various Regular Packings to a Flowing Electric Fields," AIChE Symp. Ser. 235(80):60
Fluid," Chem. Eng. Sci. 8:271 (1958). (1984).
13. A. J. Karabellas, T. H. Wegner, and T. J. Hanratty, 29. G. I. Tardos, C. Gutflnger, and R. Pfeffer, "Experi-
"Use of Asymptotic Relations to Correlate Mass ments on Aerosol Filtration in Granular Sand
Transfer Data in Packed Beds," Chem. Eng. Sci. Beds,"/. Col. Int. Sci. 77(3):616 (1979)
26:1581 (1971). 30. R. P. Donovan, Fabric Filtration for Combustion
14. J. P. Sorenson and W. E. Stewart, "Computation of Sources, Marcel Dekker, New York (1985).
Forced Convection in Slow Flow through Ducts 31. R. C. Flagan and H. H. Seinfeld, Fundamentals of
and Packed Beds—I, II, III, IV," Chem. Eng. Sci. Air Pollution Engineering, Prentice-Hall, Englewood
29:819 (1974). Cliffs, NJ (1988).
15. L. C. Paretsky et al., "Panel Bed Filter for Simulta- 32. F. Loeffler, "Collection of Particles by Fiber Fil-
neous Removal of Fly Ash and Sulfur Dioxide," ters," in Air Pollution Control Part I, edited by W.
J.APCA 27:204(1971). Strauss, John Wiley & Sons, New York.
 GRANULAR BED FILTERS 781

PART 2. APPLICATION and DESIGN

Frederick A. Zenz

CONTENTS

17.2.1 INTRODUCTION 781

17.2.2 PURPOSES AND APPLICATIONS 781
17.2.3 POROUS SINTERED GRANULE BEDS 783
17.2.4 CONTINUOUS MOVING-BED FILTERS 784
17.2.5 INTERMITTENT MOVING-BED FILTERS 785
17.2.6 FLUIDIZED BED FILTERS 788
17.2.7 GRANULAR BED FILTERS MECHANICALLY CLEANED 789
17.2.8 GRANULAR BED FILTERS PNEUMATICALLY CLEANED 791
17.2.9 TECHNOLOGICAL STATUS OF SYSTEMS UNDER DEVELOPMENT
AND UNDER COMMERCIALIZATION 792
REFERENCES 801
BIBLIOGRAPHY 801

17.2.1 INTRODUCTION bed, granular, filtration devices in various

forms have been proposed, explored, tested,
The suggestion that extraneous solids could be and even marketed over the period from at
removed from gaseous streams by passage least 1896 to the present day.
through a bed of particles is to this day re-
garded by many as novel, if not improbable,
despite the fact that the average engineer is 17.2.2 PURPOSES AND APPLICATIONS
well aware of the technology of treating liquid
streams by passage through beds of particles The principal interest in development of gran-
for purposes of removing undesirable ele- ular bed filters today rests in removal of the
ments via ion exchange and more simply for particulates from hot pressurized process gases
removing any suspended solids. The sugges- which could then be utilized to drive turbines
tion becomes more acceptable and credible by and other equipment for efficient recovery of
considering the analogy between the well ac- otherwise lost energy. A secondary interest
cepted bag filter, which represents a tortuous lies in ensuring that exhaust gases meet local
path through the interstices in a layer, or bed, and national emission standards for particu-
of interlaced cylinders (fibers), as opposed to late matter, and a third in the simple incentive
the granular bed which represents a tortuous for an efficient filter not subject to high main-
path through the interstices in a layer of inter- tenance costs.
laced spherical or angular shapes. Although no universally accepted standards
The novelty of the concept, in principle, have been specified or agreed upon for the
vanishes rapidly when a search of the patent allowable particulate loadings of gases driving
and technical literature reveals that shallow turbines, it is generally agreed that the life-
782 HANDBOOK OF POWDER SCIENCE

time and performance efficiency of such tur- tages for dry collection to avoid the handling
bines can justifiably be increased by reducing of slurries from wet scrubbers. With cyclones
the particulate content of the feed gas stream not likely to meet the standards, with electro-
via a reasonable cost filter. As an example, static precipitators representing high installa-
some typical particulate loading specifications tion and maintenance costs, and with bag
are summarized in Table 17.2.1. Electrostatic houses at 2 to 10 CFM/ft 2 representing large
precipitators, scrubbers, and fabric filters have installations requiring considerable space, the
been successfully used for removing particu- potential of a 40 to 100 CFM/ft 2 high-ef-
lates from exhaust gases at moderate levels of ficiency, low-maintenance, granular bed filter
temperature and pressure but not under high is an inevitable attraction even in conventional
temperature (e.g., 1000° to 2000°F) and pres- electrical utility applications.
sure conditions. Lowering the temperature and Other than power recovery and meeting
pressure of the turbine inlet gas stream for the emission standards there exist a host of indus-
purpose of facilitating particle removal by us- trial applications in which a "sand" medium
ing proven techniques would result in large would offer considerable process advantage
losses of energy. The increased prospective over a fabric. Condensibles in a gas stream can
utilization of combined cycle gas/steam tur- cause severe bag failures, particularly in sys-
bine/electrical generating systems, coal gasi- tems carrying cement kiln and similar calcined
fication, fluidized bed combustion, and syn- effluents, which can solidify on the bags. Sand
thetic gaseous fuels has intensified the interest media can be effectively dried and even washed
in granular bed filters. Unfortunately the ef- in situ, and light accretions removed by the
fect on turbine blades appears to be a function grinding action of the media. Where collected
of the physical properties of the impinging particulates can represent a fire hazard as in
particles. In catalytic cracking of petroleum the collection of carbon black, cellulosic solids,
fractions, carryover silica-alumina catalyst fines and similar materials, a noncombustible filter
remaining in the exhaust gases after passing medium would have immeasurable advantage.
through three stages of cyclones have shown In the popular literature are found such
power recovery turbine blade erosion to a terms as "gravel bed filters," "panel bed fil-
degree necessitating reblading only at 5-year ters," "expandable bed filters," "moving bed
intervals.1 The severity with coal combustion filters," "sand filters," "loose-surface filters,"
or gasification fines is reportedly far greater. "porous bed filters," "MB filters," and a host
Other than the incentives of power recovery of others all of which pertain to versions of
the increasingly stringent EPA emissions stan- what are generically referred to as granular
dards in the U.S.A. create definitive advan- bed filters.

Table 17.2.1. Typical Turbine Specifications.3

SOURCE PARTICULATE LOADING (MAXIMUM ALLOWABLE)

United Aircraft 0.8 lb/million SCF low-BTU fuel gas (or approximately 12 ppm)
Westinghouse 0.03% (in fuel oil) (or 300 ppm by weight)
0.002 grains/SCF all smaller than 6 /xm (for fluidized bed coal
combustors)
General Electric 30 ppm by weight in fuel gas (10 /xm maximum)
(for aircraft-type turbines)
Brown Boveri 1-2 ppm by weight (in gases entering turbine)
DOE (estimates) 0.75 grains/SCF (or approx. 2.6 ppm by weight) in 0-2 fim range
0.001 grains/SCF (or approx. 35 ppb by weight) in 2-6 /xm range
Exxon (estimates) 45-1 mg/m 3 (0.02 to 0.0004 grains/SCF)
a
NTIS-BP 266 231 Feb. 77.
 GRANULAR BED FILTERS 783

All potential commercializations of the con- In the late 1940s the Pall Corporation (now
cept rest on four basic interrelated factors: Pall Trinity Corp.) supplied such porous metal
in tubular form with the suggestion that their
1. Collection efficiency application would be analogous to a cloth bag
2. Cleaning or regeneration capability filter in which the cloth is replaced by a rigid
3. Capacity thin bed of sintered metal particles.2 As illus-
4. Competitive cost. trated in Figure 17.2.1, in normal operation
valve 1 would be open, allowing dusty inlet gas
Their practicality lies simply in the opera- to flow through the elements, depositing a
tional details of their technology which might filter cake on their outer upstream surfaces
best be illustrated by a relatively chronological and leaving through the exhaust plenum. When
review of the principal attempts at commercial the cake has grown to a thickness exhibiting
development over the past 30 years and the high pressure drop, valve 1 is closed, valve 2 is
status of such work today. opened, and a short, high-pressure pulse of air
is admitted in reverse flow through the porous
elements to dislodge the filter cake which falls
17.2.3 POROUS SINTERED to the bottom of the containing vessel, to be
GRANULE BEDS eventually withdrawn through valve 3. In con-
tinuous use, the valves operate on a timed
Particle filtration via porous membranes rests cycle and the containing vessel is provided
primarily on the formation of a filter cake, with a multiplicity of porous elements and
removable by a reverse flow of fluid when the separate plenums for localized reverse
resistance of the cake (or the pressure drop) cleaning.
exceeds any desired level. The porous mem- There are a number of inherent disadvan-
brane may take the form of a bed of particles tages to this form of filter. In order to obtain
or for that matter a mat of fibers such as is the structural strength, the granules making up
case with cloth bag filters. The thinner the the porous element must be small in size to
membrane the lower the overall resistance
with or without a filter cake and hence by
analogy to the near monolayer of fibers in a Clean gas Compressed air
cloth collector one could construct a thin bed
or sheet of granules by sintering a shallow
layer of metallic granules in a high tempera-
ture furnace. Such sheets of sintered metal
granules in various forms are sold commer- Dusty gas
cially for filtration purposes. Their principal
application is in liquid systems, such as the
maintenance of dirt-free fuel lines in aircraft
engines. They have had limited application in
the recovery of carryover from fluid bed reac-
tors and similar fine-particle processing, but
are not broadly acceptable. Their commercial-
ization stemmed from the work of Dr. David
Pall whose interest in the early 1940s lay in
the development of a gaseous diffusion barrier
for isotope separation in connection with the
Manhattan District project during World
War II. Figure 17.2.1. Porous sintered granule filter.
784 HANDBOOK OF POWDER SCIENCE

present sufficient bonding surfaces. This re- hopefully carried downward with the gravel,
sults in high-pressure drop or low gas capacity was trapped within its interstices, as illustrated
despite the only approximately | in. wall thick- in Figure 17.2.2.
ness of the elements. Simultaneously, the Four units, each of 17,000 CFM air capacity
thin-wall tubes are very subject to cracking as and consisting of two cells each, were installed
a result of repeated thermal shocks between in a plant to collect asbestos rock dust from a
normal high-temperature operation and rela- stream of flue gas coming from a direct-fired
tively colder reverse-flow cleaning blasts. In dryer in which the rock was dried prior to
addition, over extended operation, the submi- milling. The dust was 100% finer than 100-
cron fines penetrate the interstices of the ele- mesh and 60% finer than 10 jam. The concen-
ments. These fines become trapped and are tration entering the collector was approxi-
unable to be blown out with the reverse clean- mately 6 grains/ft 3 and that leaving about 0.2
ing blast. Thus, these elements build up a grains/ft.3
residual pressure drop that further limits their Dorfan's filter circumvented the thermal
capacity. They are not recommended where shock failure of Pall's sintered beds by avoid-
only very fine particles are to be collected; ing reverse-flow cleaning and instead circulat-
they operate most satisfactorily in handling ing the bed granules, plus collected dust,
streams containing sufficient coarse particles, through a vibrating screen. However, this con-
in effect, to build up their own precoat. Such sequently required continual replenishment of
instances are rare, since where they occur the the bed and hence an enormous and costly
particle loading is usually so high that the cost gravel circulation system. In order to provide
of the elements is prohibitive even if they were reasonable gas throughput capacity, the bed
never to fail by thermal shock. granules had to be relatively large so as not to
be blown off the retaining louvers; in practice
"gravel" of plus \ to minus \\ in. was recom-
17.2.4 CONTINUOUS MOVING-BED
FILTERS

As early as 1924, Cramp 3 ' 4 proposed a "dust

curtain" system for filtration of blast furnace
effluent gases. The dust curtain was a vertical
layer of the collected dust itself, held in place
by an arrangement of grids and louvers packed
under its own weight. Dusty gas passed through Inter-granular movement
causes penetrated dust to
this curtain horizontally, adding more dust to be dislodged from gravel
surfaces and reentrained
it. The excess dust gradually overflowed the
louvers and fell to the bottom. The curtain was
Dusty
renewed periodically by dropping dust out of
the bottom of the grids and adding more at
Gravel
the top. It is not clear whether this concept conveying
system
was ever put to industrial use.
In the early 1950s the late Morton Dorfan Continuous
screening
(Mechanical Industries, Inc.) introduced the system
Dorfan Impingo Filter,5'6 which consisted ba-
sically of a downflowing bed of gravel retained
between louvered walls. Dust-laden gas was
filtered by blowing through the bed normal to
the direction of gravel flow, and filtered dust, Figure 17.2.2. Continuous moving-bed filter.