Journal of Biochemical and Microbiological Technology and Engineering

VOL. I, NO 2. PAGES 185-205 (1959)

The Theory of Scaling up Laboratory Data for

the Sedimentation Type Centrifuge

CHARLES M. AMBLER
The Sharples Corporation, Philadelphia, Pa.

Summary. The mathematical theory of the sedimentation type of

centrifuge is developed in terms of unhindered settling of a single particle
that has reached equilibrium velocity. From this is derived the sigma
value for several of the different forms of sedimentation type centrifuge
which is the calculated equivalent area of a settling tank theoretically
capable of doing the same amount of work in a unit gravitational field.
The limitations controlling the accuracy of scale up between centrifuges
of similar geometry and between centrifuges of substantially different
geometry are discussed.

Introduction
Apparatus using centrifugal force for resolving multi-component
systems can be divided into two major categories, and a third
which may be considered a combination of the other two:
(1) solid wall centrifuges in which the separation is by
subsidence,
( 2 ) perforate wall centrifuges in which the solid phase is
supported on a permeable surface through which the liquid
phase is free t o pass, and
(3) combinations of the two in which the primary concentration
is effected by sedimentation followed by subsequent drainage
of the liquid phase away from the solid phase.
Each of these has its gravitational analogue: (1) the settling
tank or thickener; ( 2 ) the vibrating or rotary screen; and (3) the
rake classifier.
I n the case of the solid wall centrifuge, as is true of the settling
tank, there are two types of forces opposing the movement of the
particles of the discontinuous phase through the continuous fluid
185
186 CHARLES M. AMBLER

phase. For small particles, moving not too rapidly, the viscous
resisting force is directly proportional to the velocity of sedi-
mentation. For larger particles moving rapidly, when their
Reynolds number, dvp/q, is greater than the order of unity, the
viscous resistance becomes negligible compared to the turbulent
resistance which is proportional to the square of the velocity of
sedimentation.
General Analysis
I n analysing centrifuge performance, when a high degree of
clarification or separation is important, the behaviour of the
smallest particles in the system is usually the controlling factor.
We may, therefore, consider viscous resistance to be of prime
importance although a knowledge of the factors involved in
turbulent resistance is also of interest and will be given secondary
consideration.
The effective force acting on a given particle in a centrifugal
field is
F1 = ( m - m l ) w 2 r (1)
where m = the mass of the particle, ml = the mass of fluid it
displaces, w = the angular velocity about the axis of rotation,
and r = the distance of the particle from the axis of rotation.
If the particle is a sphere, the force is

where d = the diameter of the particle and A p = p-p1 = the

difference between the density of the particle and that of the fluid
in which it is suspended.
The force opposing the sedimentation of the particle, assuming
that it is small and does not move too rapidly, is given by Stokes’
law
F = 3.rrqdvS (3)
where q = the viscosity of the fluid phase, and vs = the velocity
of the particle moving through it.
When the force causing sedimentation reaches equilibrium with
the resisting force, i.e. when v becomes constant,
 SCALING UP CENTRIFUGE DATA 187

When the particle is in the gravitational field

&qg
APd2Y
= -
1811
The simplest form of a continuous centrifuge, cf. Fig. 2 , is a
rotating cylinder with end caps, provided with means for being
fed fluid a t one end and discharging it from the other. I n such
a device vs is the velocity with which the particle approaches the
bowl wall (or if A p is negative, the liquid surface). If the thick-
ness, s, of the liquid layer is small compared to the radius of the
cylinder, then v6 will be approximately constant and the distance
a given particle will settle during the time the liquid in which it
is suspended is in the bowl

where t = time,
Q = rate of flow of liquid through the bowl,
V = volume of liquid in the bowl = nE (+‘-r12). (64
If x is greater than the initial distance of the given particle
from the wall of such a rotor, it will be deposited against the wall
and removed from the liquid phase. I n an ideal system when
x = s/2, half the particles of diameter d will be removed from
suspension and half will not. This condition will be considered
the cut-off point and a t cut-off, by substituting x = s/2 and
rearranging equation ( 6 ) .

from which the critical diameter

for particles, half of which will be removed, can be calculated.

Referring back to equation ( 7 ) ) it will be noted that the factors
in the first group of the right-hand side of the equation are con-
cerned only with the parameters of the system being processed;
density, diameter and viscosity. The factors in the second group
are concerned only with the parameters of the centrifuge, volume,
rotational speed, radius and settling distance.
188 CHARLES M. AMBLER

Equation ( 7 ) may be rewritten

Q = 2 ~ ~ 2 (9)
in which

vg =
Apd2g
-
1%
and

in which re and se are the effective radius and settling distance of

any centrifuge of this type, and L’is an index of centrifuge size.
It is, in fact, the calculated equivalent area of a settling tank
theoretically capable of doing the same amount of work in a unit
gravitational field.

Theory for Different Types of Centrifuges

Let us consider different types of centrifuges operating on a
system in which viscous resistance to particle motion is the
controlling factor.
(1) The laboratory ‘bottle’ or batch centrifuge in which
cylindrical containers are rotated with the axis of the cylinder a t
right angles to the axis of rotation.
(2) The continuous flow tubular type in which the settling
distance is not small with respect to the radius.
(3) The disc type in which the settling distance is small with
respect to the radius and in which the discs are sloped to permit
radial movement of the sedimented solids across them.
T h e bottle centrifuge, Fig. 1. For this, equations (4)and ( 5 )
may be re-arranged,

Then, if r = distance of the particle from the axis of rotation,

rl = radius of liquid surface in bottle, r2 = radius at surface of
sedimented cake, and t = total centrifuging time, we can re-
arrange equation (11) and integrate :
 SCALING UP CENTRIFUGE DATA 189

(12)
from which

Fig. 1. Particle in bottle centrifuge

At the 50 per cent cut-off point where half of the particles would
be sedimented assuming them all to be the same size
A ( r - r l ) = A (r2-r) (14)
A = the cross-sectional area of the bottle and
r2 + rl
.: r = - 2 (15)
Also,
V
(16)
t=&
Q = Effective ‘flow rate’ through the bottle (ml per min).
Substituting equations (15) and (16) into (13) and re-arranging,

from which
/y=-=
Q W2V

or more conveniently
190 CHARLES M. AMBLER

The time required to sediment all particles of diameter d can

be determined by integrating drlr in equation (12) between the
limits r l and r2. This may be regarded as the special case when
r = r l , and
V,W2t
lnrz/rl = - (20)
9
For spherical particles, from equation ( 5 ) and re-arranging

The continuous $ow tubular type. A typical tubular bowl is

shown in Fig. 2 . A particle a t radius r and distance z from the
bottom (inlet end) has two motions :

Fig. 2 . Particle in tubular bowl centrifuge

(1) A radial velocity due to centrifugal force (equation 1 1 )

dr
_ -- -v p 2 r
dt g
(2) A linear velocity due to the flow of liquid through the bowl,
which is assumed to be uniform across its cross-section,
 SCALING U P CENTRIFUGE DATA 191

from which

Integration of this equation, within the limits r1 = r, at z = 0,

and r = r2 at z = 1, gives

The 50 per cent cut-off corresponds to

47-22 - r 2 ) = n(r2 - r12) (26)
and

By re-arrangement, and using equation ( 2 7 ) , equation ( 2 5 )

becomes

for which the approximation, within a maximum error of 4 per

cent. is

and by comparison with equation (10)

-re _- (%z2+$r12)
se (rz2-r12)
in which the effective values are
3r22+r12
re =
4r2
and

From equations (29), (9) and (6a),

192 CHARLES M. AMBLER

I n a similar manner, the Z value of the tubular bowl in a system

where turbulent resistance to motion is the controlling factor can
be derived from the basic equation of resistance to turbulent
motion

F = K -rr4 p l d Z U g 2 (34)

K a constant having a value of approximately 0.5 and

=
dimensions of l/g
from which

(38)
and

(39)

It will be observed from a comparison of equation (39) with

equation (33) that while the same elements are involved in the
calculation of Z, the exponents are quite different. On the other
hand, in comparing equations (5) and (37) for the elements of vg
it will be observed that not only are the exponents different but
g has been replaced by the reciprocal of the constant K , and 7,
the viscosity, by p l , the density of the fluid phase.

The disc type. Sedimentation in a disc-type rotor can be treated

theoretically by analysing the space between two adjacent discs
on the assumption that the flow in such a rotor is uniformly
distributed in all such spaces. Such a system is shown in Fig. 3.
The discs extend from r l to rz and a particle is shown a t position P.
Its distance from the lower disc is y, and its distance from the outer
edge of the disc, measured along the surface of the disc, is x. The
 SCALISG LTP CENTRIFUGE DATA 193

discs of a commercial centrifuge usually have horizontal flanges

a t their inner and outer diameters. These are present entirely
for mechanical reasons, t,o improx7-e the strength of the discs.

Fig. 3. Particle in disc bowl centrifuge.

They contribute nothing to clarification eEciency and are not

considered in these measurements. The particle P has two
motions,
(1) centrifugal force moves it away from the lower disc
dy
_ -- -vgo2r cos 8 = -
vyw2
(r2 - x sin 8) cos 0
dt 9 9
8 = the disc half angle, and
( 2 ) it is swept along by the flow of liquid from r2 towards r l .
If there are n spaces between the discs in a given rotor and the
total throughput is Q, then the flow between each pair of discs is
&In and the average velocity across the distance a, between and
normal to adjacent discs, a t radius r is QI2nnar. The velocity
will be zero a t the disc surfaces and reach a maximum somewhere
between them. Let f ( y) be the ratio of liquid velocity a t position
y to the mean velocity. Then

Q
- = -f(y)
dx = QfW
dt Lnnar Pnna (r2 - x sin 8)
and, from equations (40) and (41)

- 2nnavp2 (r2- x sin e ) 2 cos 8

dy -
_
dx Qs f(Y)
7+
194 CHARLES M. AMBLER

Iiitegratioii gives

(43)

in which y2 is the value of y a t x = 0 where the particle enters

the space between two discs and yl is the value of y a t x =
( r z - r l ) cosec 0 where the particle would leave the disc stack if it
were not sedimented out and it will be just sedimented if yl = a.
Now
IY:f
(Y)dY (44)

is the fraction of the flow between y2 = y and yl = a and all

particles of diameter d in this portion of the flow are sedimented
against the upper disc surface. Half the particles will be
sedimented and half will not when the fraction is one half, and at
cut-off,
4nnvgw2
Q = - (r23 - r13) cot 8 (45)
3gc
in which C = the average value for the integral off (y),and

z = -Q= zTnw2(r23- r13)

2% 3gC tan 8
The value and significance of C in equation (46) is not completely
understood. If we assume the flow between adjacent discs to be
symmetrical and laminar, then C equals unity only when either all
particles of diameter d are sedimented or a t cut-off when half the
particles of diameter d are sedimented, i.e. when the value of
y a t x = 0 is either zero or aj2. However, the value of C is signi-
ficantly different from unity when the particle starts from any
other position and may be greater than unity even starting from
the a12 position if the flow pattern is disturbed by surface
imperfections or by the accumulation of solids on the disc surfaces.
This may also be true if the symmetry of the flow pattern is dis-
turbed by the vector effect of centrifugal force or a change from
a radial direction of flow due to conservation of angular momentum.
The significance of this, and possibly other effects, has been con-
firmed by experimental data using satin clay in aqueous dispersion
as the test medium.
 SCALIKG UI’ CENTRIFUGE DATA 195

Comparison of Centrifuges
Referring to equation (9) it will be observed that for a given
test medium whose z!, is a constant, Q1.Z = 2v, = constant, and
therefore it should be possible to compare various centrifuges
with gravity settling systems on the basis that

On tubular bowls over a fairly wide range of proportions of

1’2,r l and 1 the determined 2 values were between 97 and 98 per
cent of the calculated values. On disc bowls, when the flow was
introduced at the outside diameter of the disc stack, the observed
value was 55 per cent of theoretical. Both bowls were studied
over the range of 1 to 20 per cent of unsedimented particles. The
above is not to be construed as a criticism of the disc type of
centrifuge. On many commercial applications its ‘efficiency ’
does more nearly approach 100 per cent and conversely that of
the tubular bowl may fall below the above mentioned 97 to 98 per
cent value. The example selected does show that the relative
actual performance characteristics of these two types of centri-
fuges may vary over comparatively wide limits, in a ratio
approaching two to one, from their calculated characteristics.
The significant point is that data, in terms of Q12,should not
generally be extrapolated from the performance of one type of
centrifuge to that of another.
I n this connection, the effect of two other phenomena should be
considered carefully.
The mathematical analysis has concerned itself with the sedi-
mentation of a single particle in a field in which it is not affected
by the pressure of other particles. For small particles, exhibiting
Brownian movement, the diffusion constant

where N = Avagadro’s number, R = t,he gas constant and

rl = the 1:wticlc diameter, may be appreciable and the applied
centrifugal force must be great enough to overcome the ‘threshold
effect ’ before the laws of sedimentation will apply.
IF6 CHARLES M. AMBLER

The second significant point to be gained from this is that, when

dealing with very small particles, the extrapolation should not
be made over too wide a range of centrifugal force even when
dealing with centrifuges of similar proportions.
Referring back t o equation (9) it will be observed that all
succeeding calculations are based on the vg of a given system being
constant. While this is frequently true, the particles in many
systems exhibit a degree of agglomeration with the passage of
time and these agglomerates exhibit varying degrees of stability.
A typical example is the precipitate of ferric hydrate, the stability
of which is influenced to a great degree by the electrolyte concen-
tration of the fluid phase. I n a quiescent field such a precipitate
shows a high value of vg. When introduced into a rotating
centrifuge bowl the abrupt change in angular velocity substantially
changes the diameter of the individual particles. If the retention
time in the centrifuge is short, even though the agglomerates do
reform, in time, there is an apparent shift of Q/Z to a smaller value
for the same degree of clarification as the rotational speed of the
centrifuge, and its shearing effect on the feed, increases.
A third significant point, which may be considered a combina-
tion, and in support of the first two, is that the extrapolation in
many cases should only be made between centrifuges of the same
type operating a t approximately the same rotational speed.
Finally, it can be demonstrated that the plot of Q/Z versus the
proportion of unsedimented solids is theoretically a straight line
when plotted in that order on logarithmic-probability paper when
the particle size of the solids follows a normal Gaussian curve.
No extrapolation should be undertaken in the range where such
a plot deviates from linearity since this is a positive indication
that the Z value of the test centrifuge is deviating from theoretical.

Liquid-Liquid Systems
Up to this point, centrifugal performance has been discussed
only in terms of the sedimentation of particles in a centrifugal
field. Commercial centrifuges are also commonly used for the
separation of immiscible liquids, such as oil and water. I n such
systems one phase is usually dispersed in the other and effective
separation is considered to have been obtained when the required
 SCALIXG UP CEKTRIFUGE DATA 107

proportion of the smaller dispersed particles has been separated

out. Such a separation follows the same theory outlined above.
In order to operate on such a system continuously it is necessary
to keep the separated phases in a condition of hydrostatic balance.
An arrangement for accomplishing this is shown in Fig. 4. It will
be noted that this system is in apparent balance when the dirnen-
sion e - I times the density of the light phase equals the dimension
c - h times the density of the heavy phase. However, this only
applies because the variation in the earth's gravitational field
between points I and h is negligible. The general formula

is applicable. I n the earth's gravitational field the values of f ( r )

on each side of the equation are almost identical and cancel out
so that

I n tlhe centrifuge as illustrated in Fig. 4,this is not the case and

(52)

Centrifuge

Fig. 4 ( a ) . Separation in tuhiilar Fig. 4 (b). Separation in

bowl centrifuge gravitational field
198 CHARLES 31. AMBLER

This general formula is equally applicable to the gravitational

field if the dimensions e , h and 1 are measured from the centre of
the earth, that is the point where r = 0,which is also the reference
point from which these dimensions are measured for the centrifuge.
I n continuous centrifuge rotors of both the tubular and disc
types, provision is usually made t o control the position of the
interface ( e line) by varying either the 1 or the h dimension of the
bowl. The more common method is to vary the h dimension by
the use of different sizes of ring dams or different lengths of heavy
phase discharge nozzles. The rl of the light phase and the r2
of the heavy phase become fixed dimensions of the rotor, while the
r2 of the light phase and the r l of the heavy phase, which generally
correspond to the e line, can be varied t o meet process require-
ments.
A given bowl running a t a given speed therefore has a fixed total
2 value which is the sum of the variable 2 values of the portions
of the bowl occupied by the light and heavy phases, 21 and Z h
respectively.
For optimum use of a given separator bowl the e line must be
controlled in accordance with the following relationships:
( 1 ) QZ/Zl is selected to give the required degree of purification
of the light phase.
( 2 ) & h / Z h is selected to give the required degree of purification
of the heavy phase.
(3) Q (the total flow) = Qz+Qh
(4)Z (the entire centrifuge) = , Z l + Zh
When considering the operation of the tubular bowl as a
separator on a given system (cf. Fig. 4) the optimum position of
the e line can most easily be established by the trial and error
method since variation of the e line within the limits 1 to e only
affects 2
1 and 2 h within the previously defined limits.

Liquid-Liquid Systems in Disc Bowls

When considering the disc bowl as a separator, another com-
plication arises. The feed to such a bowl enters and is distributed
through holes in the disc stack, Fig. 5 . For optimum performance
the e line must coincide with the location of these holes. If it
 SCALING UP C E B T R I F U G E DATA 199

does not, then a t least some of the 2 value of the fraction of the
disc stack lying between the e line and the feed holes is not
available for purification of the light (if the e line is outside of the
holes), or of the heavy phase (if the e line is inside of the holes).

k- e e 4

Fig. 5 . Diagram of disc bowl separator

On a disc type separator bowl, therefore, not only must the

proper ring dam adjustment be established, but the location of
the feed distribution holes in the disc stack must coincide with the
resulting e dimension.
Furthermore, where separation on a disc type rotor is being
studied, either the Q1.E values of both phases must be separately
determined, or the extrapolation should only be made between
centrifuges having similar . E l / 2 h ratios.

Extrapolating Test Data

An examination of equation (46) indicates that the performance
of a disc type centrifuge may be improved :
(1) by reducing the disc angle ;
(2) by increasing the number of spaces between discs, n, which
can be most easily done in a given bowl by reducing the space
between adjacent discs, dimension a of Fig. 3.
I n either case, the dimensions and nature of the solid particles
present may become of critical importance. The disc angle must
200 CHARLES 31. AMBLER

be large enough to permit the sedimented solids to slide off them

freely, and the spacing between discs must be large enough so that
these accumulated solids as they slide along the disc surface do not
interfere with the flow of liquid between discs.
To summarize, for simple tubular and disc clarifier and separator
centrifuges, the following points should be considered before extra-
polating test data from one size or type of centrifuge to another.
(1) Extrapolation should only be made in the area where the
plot of Q/Z versus proportion of unremoved impurity on a log-
arithmic-probability chart lies in a straight line.
(2) Extrapolation should not be made between different types
of centrifuges (such as disc and tubular) unless the correction
factors for the calculated 2 values of the particular centrifuges
and the effect of the particular system on these correction factors
are known.
(3) Extrapolation should not be made between centrifuges
operating a t different speeds unless it is known that the effects
of diffusion and de-agglomeration will be negligible on the
particular system.
(4)Extrapolation should only be made between disc centri-
fuges having a close similarity with respect to disc angle, disc
spacing and, in separation problems, feed hole position.

Disc Centrifuges with Discharge of Sediment

Disc type centrifuges, because of their relatively small height
to diameter ratio, are particularly suited for modification to
permit the continuous or intermittent discharge of the sedimented
solids through ports located around the periphery of the bowl.
The inside of the bowl shell can be tapered or shaped to direct the
solids to hhese peripheral ports without increasing the bowl
diamet,er by an excessive amount, Fig. 6.
Two types of ports are generally used:
( 1 ) Open nozzles through which there is a continuous flow of
the heavy liquid carrying the sedimented solids with it.
(2) Valved ports which may be opened intermittently to permit
the discharge of the accumulated solids with minimum discharge
of the heavy phasc liquid.
 SCALING U P CENTRIFUGE DATA 20 1

Fig. 6. Peripheral discharge disc bowl

I n Type 1, the flow through the nozzles is a function of their

size and number and of the liquid head acting on them.
Q = Klna(2gh)i = K1n~[w'(r22-~12)]1 (53)
where K1 = nozzle coefficient
I n Type 2, the same factors apply multiplied by a coefficient
which is the proportion of the total operating time that the valves
are open.
tl
Q = K1- ~ Z C L [ W ~ ( Y ~~' -1 2 11
) (54)
t
where tl = valve open time and t = total time.
Since these ports are located outside of the disc stack, they have
a negligible effect on the ,Z value of the bowl. The nature of the
solids when centrifugally compacted becomes of primary impor-
tance in the application of centrifuges of this type. The sedi-
mented solids must be of such a nature that they will flow or be
extruded through the discharge ports under the pressures avail-
able. Their angle of repose must be such that any accumulation
of solids between adjacent ports does not build into the disc
stack, an effect which will reduce the ,Z value of the assembly.
The application of these centrifuges is therefore to systems from
which the centrifugally sedimented solids are relatively plastic
in nature.
Test data from small-scale equipment are of value in deter-
mining the Q1.Z ratio at which satisfactory removal of the solid
7*
202 CHARLES M. AMBLER

phase is obtained within the limits described above. To determine

the quantity and physical nature of the sedimented solids the test
centrifuge must be operated a t a speed t o give the same centri-
fugal force as the final proposed machine.
While the quantity of solids to be removed can be determined
quite accurately by this method, no rigorous tests except an
actual full-scale run can be applied to determine whether the
nature of the solids is such that they will satisfactorily discharge
from a peripheral discharge centrifuge of either the nozzle or
valve type. There is a considerable background of operating
experience with such centrifuges on widely different materials
such as yeast, protein, starch, gluten, vegetable oil, mud from wool
scouring, clay and tar available to be drawn on, but each new
application requires careful individual study with particular
reference to the presence and characteristics of minor constituents.
I n general, the following considerations apply:
( 1 ) All particles having a diameter greater than half the dia-
meter of the discharge ports must be removed by pre-clarification
or straining.
( 2 ) It is not practical to use less than 2 discharge ports in a
commercial centrifuge of this type and for many applications the
use of 12 or more is desirable to limit the hold-up of solids between
ports.
(3) On most applications the smallest practical nozzle size is
0.035 in. diameter which delivers between 60 and 80 gal per h
per nozzle from a 6,000 rev/min 19 in. diameter centrifuge depend-
ing on the viscosity of the sludge and the density of the liquid.
On certain specific applications such as soybean protein con-
centration, it may be possible t o use smaller nozzles, t o 0-031 in.
diameter, and on others, such as yeast concentration, special bowls
have been developed t o return the concentrated solids back
towards the centre of rotation, decreasing the net effective head
pressure on the nozzles.
For this type of centrifuge, it may therefore be said that small-
scale tests will:
( 1 ) Establish the throughput rate, &, for effective clarification
on a given centrifuge rotor,
 SCALING UP CENTRIFUGE DATA 203

( 2 ) Indicate the terminal concentration possible in this

centrifuge.
(3) Establish the approximate number and size of nozzles
required to concentrate the solids into the required portion of the
feed stream.
(4) On the basis of the answers to ( 2 ) and (3) indicate the
relative process advantages of the nozzle and valve type rotors.

Centrifuges with Helical Conveyors

Two modifications of a type of centrifuge that utilizes centrifugal
force for both the sedimentation and drainage of the solid particles
are shown diagrammatically in Figs. 7 and 8. I n these, the
sedimented solids are carried to the small diameter end of the

Fig. 7. Conical continuous discharge rotor

rotor by a helical conveyor, which is driven a t a speed slightly

different from that of the rotor itself. The liquid overflows a t
the opposite end through ports that may be adjusted with respect
to their distance from the axis of rotation for control of the liquid
level in the bowl.

f
L c

Fig. 8. Cylindrical continuous discharge rotor

204 CHARLES JI. AMBLER

I n the conical modification, Fig. 7 , the sedimented solids are

carried up a finite length of ‘beach’ where they are given an
opportunity to drain before being discharged. I n the cylindrical
modification, Fig. 8, which because of its greater volume has a
much larger 2 value for a given lengthldiameter ratio, the
length of time the solids are out of the liquid layer before being
discharged is relatively short. I n this, the ‘drying’ is principally
effected by compaction of the solids under centrifugal force as
they are being scrolled.
For the conical conveyor type of centrifuge the 2 value may be
calculated to be
2 = -2 ( ~ ~ 2r22+
Y
1 3r2r1+ 4r12
8 1
where I = length of liquid measured a t its surface, r l = inside
(55)

radius of liquid, and r2 = outside radius of liquid.

From this should be deducted a correction factor of approxi-
mately 6 per cent, representing the volume of the conveyor
flights in the rotor.
For the cylindrical conveyor type of centrifuge the Z value
can be considered as the sum of the 2 values of a true cylindrical
section of length 11 by equation (33) and that of a conical section
of length 12, it being considered that a t the solids discharge end of
the rotor a beach is formed to the contour of the conveyor.
The composite formula therefore becomes

.=-( 9
1 + r12
2 ~ ~ 2 13rz2
4
2 ~ ~ 2 1 r22+
)+T(
2
8 1
3r2rl+ 4 - 1 2
(56)
and again a correction factor of approximately 6 per cent should
be deducted for the volume of the conveyor flights.
An examination of Figs. 7 and 8 also indicates that two other
factors may decrease the calculated 2 values. I n addition t o the
liquid volume displaced by the conveyor, the sedimented solids
that are being conveyed through the rotor displace their own
volume of the feed liquid and therefore must be compensated for.
I n addition, it will be noted that it, is impossible to introduce the
feed stream a t the point where I = 0 without washing the solids
that are about to be scrolled back into the pond.
Experimental data on a variety of materials have indicated
that on the conical conveyor type the actual .Z value approaches
 SCALING U P CENTRIFUGE DATA 205

0.62 times the calculated 2 value and on the cylindrical scroll

conveyor type the actual 2 value approaches 0-67 times the
calculated.
As in other types of solids discharge centrifuges the nat,ure of
the sedimented solids is of prime importance. The solids must
pack to a consistency that will permit them to be scrolled out of
the bowl. I n the simplest possible terms, the solids must have a
high angle of repose and be of the general nature of material that
can be transported by a screw conveyor. Here, as in the peri-
pheral discharge disc machine, tests on a centrifuge that has at
least the equivalent of a conveyor is a final requirement since the
apparent Ql2value for a given degree of clarification will be
affected by the presence in the system of any solids that cannot
be scrolled.
Laboratory units of this type are available that give good corre-
lation of Qi2 values with commercial equipment. Since these
small units scroll the solids in the same manner as their commercial
counterparts. the question of scrollability of the solids is auto-
matically compensated for.
The amount of mother liquid adhering to the discharged solids
in a centrifuge of this type appears to follow the general equation
logy = a+bx (57)
where y = proportion of adhering mother liquor, x = reciprocal
of the time interval during which the solids are out of the con-
tinuous liquid phase, and a and b are constants of the system and
the applied centrifugal force, wZr/q.
If two points for a given system are known it theoretically is
possible to calculate the complete curve, but this has not been
thoroughly explored and resort must usually be made to full-scale
operation on commercial size equipment to establish complete
data.

References
Hebb, hl. and Smith, F. Centrifugal Separation, in Kirk and Othmer,
ed. Encyclopedia of Chemical Technology, Vol. 3, 1st ed. (1949) New York;
Interscience
Flowers. Centrifuges, in Perry, ed. Chemical Engineers Handbook,
pp. 992-1013, 3rd ed. (1950) S e w Y d i ; JlcCraw-Hill.