An authoritative critique of the established and recently proposed methods thickener design

s. ubstantial advances have been made in recent years

. . .
in defining principles of sedimentation. The prac-
touch one another, and at least some supporting force
is transmitted through interparticle contact. There-
ticing engineer will ask, How do they affect design of fore, layers of solids exert a squeeze or compressive force
thickeners? This paper will seek an answer by on those below, so the regime is named compression.
appraising current theories. Then it will suggest As long as particles are subsiding, some of their weight
design procedures which seem appropriate at the present is borne hydrodynamically, but when subsidence is
incomplete state of the art. complete, this component disappears.
Coe and Clevenger (7) postulated two regimes of
thickening: In one, particles were assumed to settle ZONE SElTLlNG
without any mechanical support from those below. The classic Coe and Clevenger procedure (7) is to
Particle weight is borne solely by reaction with fluid. determine zone settling rate u over a range of concentra-
They named such behavior free settling and we retain tions C by observing initial interface subsidence rate in a
the term. However, they also assumed that under series of batch tests. They deduce that the solids
such conditions settling rate u will vary only with solids handling capacity or flux Goof any concentration layer
concentration C. Their procedure relies upon the existing at steady state must be related to settling ra+-
latter assumption, as does that of Kynch (7), and we will and pulp consistency by the equation
take it as defining a zone settling regime.
In their other regime, particles are close enough to GO --
II
(
K ( D - 0%)
AUTHOR Bryant Pitch is Director of Research for Don- and the maximum capacity possible will be the smallest
Oliver, Inc., Stamford, Conn. He has been with Don-Olivn value of Goobtained by applying the formula to a series
m its predecessor companies since 1944 and is a recognized wmld of tests made upon pulps ranging in consistency from
leader in design of equipment for physical separation processes. that of the feed to that of the thickest free-settling pulp,

18 INDUSTRIAL AND ENGINEERING C H E M I S T R Y

I i

and taking for D. the ratio of fluid to solids desired in state zone settling and the sudden appearance of a zone
the underflow (our notation substituted for theirs). of intermediate concentration C, when a continuous
If the thickener is fed at a greater solids rate, a zone thickener is overfed. However, Kynch (7) in 1952
of critical concentration corresponding to the smallest presented an elegant analysis which explains nearly all
value of Gs will form and build because solids cannot zone settling behavior and launched a modern era of
pass through it as fast as they are fed up to it (3). At a thickening research. What is being done today is
lower feed rate the zone will dwindle and disappear. largely elaboration of principles taught in his illuminating
Because K D = 1/C - l / p . and u = G/C, by simple paper.
substitution
Kynch Theory
G = Go- e ) C Kynchs theory was exciting because it established
quantitative relationships between flux plots and batch
Equation 2 defines an operating line in a plot of G us. C, settling curves. It showed how to construct a flux plot,
as shown in Figure 1. Such an operating line, drawn or at least the important part of one, from a single batch
through any p i n t (I on the flux curve and C. on the C- test. We hoped, therefore (73, that thickener unit
axis, will intersect the G-axis at a value GB. A line area could be determined directly from one such test,
drawn tangent to the flw plot as at 6 will show the rather than from a series of them as in the Coe and
smallest intercept Gemin, which is the one defining Clevenger procedure.
maximum possible thickener throughput. This Yoshioka Kynch deduced three relationships or theorems of
(27) construction, as well as a related one using reduced primary interest:
variables introduced by Wallis (79), is a useful and con- THEOREM I. If a concentration discontinuity is
venient graphical way of applying Coe and Clevenger propagating in the direction of settling with a velocity
zone settling theory. The Yoshioka and Wallis methods 6, then
do not, in themselves, constitute a new concept in
thickener design. 6 = -
AG
(3)
Coe and Clevenger adequately explain both steady- AC

V O L 5 8 NO. 10 OCTOBER 1 9 6 6 19
where AG is the instantaneous settling flux difference through it. By material balance
across the discontinuity (equal to AS in zone settling
regime). AC is instantaneous concentration difference.
Derivation. The flux into one side of a discon- But
tinuity must equal that out the othet. If the dis-
continuity is propagating with a velocity 6, then at any -l @ t = Ha
instant:
Then, if settling rate u. is constant, it is equal to R.
- 8) - 8)
because
Cl(U1

Clul = G1 and CZUZ

=
= Gz
CZ(UZ
&olids = +
ACa (Rata Ha) ACoHet
and if initial concentration Cois uniform
8=
Gi - Gn =-
AG 2.olids = AHoCo ACJi",
CI - Cn AC so
HEO OR EM 11. If a concentration gradient exists in
(5)
the neighborhood of concentration Ci, then the locus of
thii concentration will propagate in direction of settling which is Theorem 111.
with a velocity @ such that While Kynch did not apply his theory to steady-
state operation, many others have, and the following i
(4) corollaries of Theorem 111 have been recognized: The
hbscissa of the Kynch tangent line at an ordinate value
Derivation. Theorem I1 can he deduced from of 1/C. measures unit area 1/G, required in a thickener
continuity equations in.ziny of several equivalent ways to handle a zone of concentration C.. This is the
(7, 77). Perhaps it is simplest to note that as Cn + Ci, Talmage and Fitch construction (75). Intercept of the
a dkontinuity dwindles to the locus of Ci. tangent with the abscissa axis is equal to the value of
Its instantaneous propagation velocity is 1/S corresponding to C. (73). Thus, a batch settling

@
rim AG
*
A-
-=
AC
rg) f
curve appears as a transform or mapping of the flux
plot onto a 1/C, l/Splane under Theorem 111 rules.
It is important to note that Kynch Theorems I and
THEOREM 111. Theorem 111 is best expressed graphi- 11, being derived from material balance or continuity,
cally. Figure 2 is a batch settling cqrve, reduced by are generally valid, whether the pulp is in zone settling.
dividing both scales by COHO(5, 27). The theorem Theorem 111, on the other hand, is highly restricted, and
askrts that if a tangent is drawn to a settling curve at applies specifically to batch settling. It holds only if
same point u, its intercept with the ordinate axis measures settling rate in the zone at all times remains constant,
l/C*, where C. is concentration existing just below and the zone propagates from the bottom of the column.
slurry-supernatant interface at point u. The first condition implies zone settling, the second will
Derivation. If a zone of constant concentration be investigated later in considering the effects of com-
starts propagating from the bottom of a column, hy the pression. Derivation OF the corollaries will be obvious
time it reaches the surface all solids must have passed from Figure 2.

20 INDUSTRIAL AND ENGINEERING CHEMISTRY

Figura 3. Soli& pr@s in conrinrrour lhinkning [after Cmingf ( Z ) ] Figura 4. F l u plots careipmdiwg to proJil$s of Figm 3

Equations 1 and 2 can be deduced from Kynch coincide exactly with a postulated linear segment of a
theory, but this is an unnecessarily roundabout way of flux plot is infinitesimal. But when the u n d d o w
applying material balance. Coe and Clevenger and concentration is changed (profile 2), a new operating
Talmage and Fitch deduce Equation 1 dircctly from line is established and a pmfile of varying concentratbn
fluid flux balance. Yoshioka (27)deduced Equation 2 is stillobtained. This r e q h that the flux plot coincide
from solids flux. All are fimctionally interrelated and with the new operating line $80, over the concentration
equivalent; there is nothing fortuitous (6) in their range observed in profile 2. This is absolutely impossible
c o m p n d e n c e . Kynch theory, in itself, does not if S is a function of C only.
constitute a new concept in thickener dcaign, although Over at least the wide range of concentrations common
it did promise a simpli6ed way of collecting the requisite to both profiles, some other variable must enter to change
flux data. the flux plot. And by logical induction this must be
true for any concentration which can be observed in
Umitalionr of ZOM Theory any such profile.
If wne s d i n g obtained over the entire concentration While the effect of the new variable might conceivably
range between feed and underflow, thickener design be consided secondary (13),it is certainly not minor.
would offer no problems. Coe and Clevenger wne tests Even if the settling rates correspnding to profile 1
would give the answer, whether calculated by their could be considered roughly or approximately a function
equation or worked out graphically on a flux plot made of concentration, thm for profile 2 are far differtnt.
from the data (6, 17, 27). Zone settling does occur It is W e v a n t to note that a conical vessel OF perhap
over much of the concentration range in pulps made up rake action might give thick bed profiles of qualitatively
of uniformly sized, n o d o c d a t i n g particles. But in the shape observed here because the cited tests were
most flocculent industrial p u l p it does not. There carried out in a cylindrical thickener and the p d l e
many evidences of this (4, 5). O d y the most extended far above the zone of rake acdon. Settling
direct and cogent will be cited here. rate of solids is not even appmxitnately a singlavalued
Continuous thickeners operate with a thick bed or function of concentration. The profiles cannot in any
profile of settled solids in which concentration varies sense be explained on the assumption $hat it is.
with height. Two typical profiles from the work of Pmfile concentrations cannot be in zone settling, so
comings (2) are shown in Fire 3. In these two cases maybe they are in compression. That was the elassied
thmughput flux Ga was maintained constant. Under- view before attempts were made (6,9,10,73,78,19,27)
flow rates, and hence C,,, were different. The material to explain all or most s e t t f i behavior un the basis of
was calcium carbonate. wllc settling and Kynch.
Now each and every concentration which exists in the
profile at steaay state must transmit exactly the same COMPRESSION
thickener flux Gl. Its settling flux must then be aa Coa and CIewn#r
sppcified by Equation 2. I n other words, the settling Cue and Clevengers concept of compremion haa been
flux provided by the slurry and that demanded by the widely h n s t n x d . They p-t data h r m n
operating or material balance line must coincide over M e r e n t pulps, six characmized as metallurgicaz and
the entire rangc of concentrations observed in the profile. one chemical. I n their context, metalluqicd p u l p
Therefore, the flux plet for prufile 1 would have to ase those resulting from grinding of ore; the chemical
Id something like Figure 4. This is unlikely enough. one was a precipitate. They observed that after what
The chance that a flux plot would have a linear range they conceived to be compression waa r e d d , then
gcm8 remote. The probability that GIand C. would further elimination of water becomes approximately a
be chasen in such a way as to make the operating line function of time, insofar aa tests for m d u g i e a l prac-

V O L 5 8 NO. 10 OCTOBER 1 9 6 6 21
tice are concerned. Their detention time basis for By force balance around this plug they deduce that
compression thickening is based on this empirical interface subsidence rate R will be :
observation.
With regard to chemical pulps they say: It would
seem that in this type of pulp the law of compression
being a function of time would not hold. By now
that must be one of the most thoroughly reconfirmed And for a batch test
conclusions in sedimentation literature. They give a
different design procedure for such pulps.
Unfortunately, some subsequent literature gives the
impression that their metallurgical procedure is univer-
sally valid. But it wasnt Coe and Clevengers idea, Note that initial changes in floc structure observed in
and in the demiworld of industrial know-how it has some pulps is not taken into account here.
long been recognized as not universally true ( 3 , 5 ) . The Michaels and Bolger approach will now be ex-
The hypothesis Coe and Clevenger advanced to tended to the region below the control plug. Because
explain detention time behavior is as follows: It seems it will be applied to industrial thickeners, d is large and
probable that the relationship of pulp consistency to effects of wall shear may be neglected. Then at all
time of thickening is an indirect one, depending upon +
points the compression stress will be equal to u, and
the effect of compression caused by the depth of pulp will be a function of C.
being counteracted by the resistance in the pulp to the By force balance around a lamina at x at some fixed
escaping water, and to the admixture of the pulp in the time it follows that
upper portions of the thickening layer with the ascending
water from lower regions. This, which has been
accepted as the gospel of Coe and Clevenger theory, is
somewhat ambiguous, and as often interpreted can be
shown to be logically inconsistent.
If the resistance to flow of escaping water exists in the Derivation
overlying layers, then by force reaction or balance the
flow provides hydrodynamic support to the overlying The unbuoyed weight of solids per unit area at time t
layers. Unless there is some other sort of resistance in a lamina of thickness dz equals
arising locally in the floc structure itself, all the resistance
results in hydrodynamic support of overlying layers, and
no compression exists. The pulp remains in zone
settling.
Also in a continuous thickener there is no ascending I n free settling, the entire unbuoyed weight of solids is
water-the water flux is at all points downward below borne by the pressure gradient necessary to push fluid
the feed point. And, in batch settling, the total flow through solids at a superficial velocity or space rate
past any level is zero, so the ascending flux of water relative to solids equal to u. At low Reynolds number
equals in volume the descending flux of solids at that flow, Darcys law should hold so that the fraction of
level. I t is not otherwise related to any trades in posi- solids weight supported by fluid drag at a subsidence
tion between solids and water at any other level. There- rate, u, will be equal to v/u. Because v / u = G/S, the
fore, the idea of admixture requires much further support given by fluid flow through the lamina equals
explanation to be meaningful.
Coe and Clevenger, then observed, but did not
adequately explain, compression behavior.
c g
(y) dz

Michaels and Bolger Hypothesis The support given by d$ is the difference, so

Michaels and Bolger (8) present a hypothesis or
model of compression which lends itself to mathematical =- Z)
(g+J c (7 ) (1
analysis. They postulate that solids structure in a com-
pressing pulp will have some yield value u for com-
pression and another T for shear. They tacitly assume
that deformation of the plastic structure will be so slow
that viscous stresses are negligible compared to yield
values.
At the top of any originally uniform column of pulp But if $ = $(C),
starting in compression there will then always remain
some control plug of initial concentration because
through its height AH, the squeeze, $, generated does not
exceed yield value u.

22 I N D U S T R I A L A N D ENGINEERING C H E M I S T R Y
 If we substitute back into Equation 8,

If we rearrange,
dz = ( L(e)*
)
dC S- G dC
Equation 10 gives a relationship between C and I, in
terms of G, S, and the value o f (dx/dC)* observed in a
batch test after subsidence is complete. The latter
group are functions of C and can be evaluated. Thus,
flux G is determined at any concentration by thickener
material balance. S can be dmrmined fmm deep
' P* c batch tests to be desvibed later. The expression, (dx/
dC) *, is by definition an experimentally determined b
F i p e 5. Porrulorsdpur plot showing Mmprassion ragimr tion.

Combining continuity with Equation 8 gives a CcllculclHon of Compnrribn a p l h

general differential equation for compression thickening.
Integration of Equation 10 leads directly to thickness
It is formidable. Fortunately only steady-state d u - of a thick bed of compressing pulp.
tions are appropriate for design of continuous thickeners,
and they can be obtained simply. At steady state,
W/bz becomes independent of time and, hence, is a
complete differentialdC/dz.
I = JG
c-o ("S)- G ($)*dC (11)

When there is no longer subsidence in a batch test a U m M o n r of Compmrion Theory

steady state has been attained. Equation 8 is valid, W e arc now out of the dilemma pmed in Figure 3,
with G = 0 and aC/& = (dC/dx)*. Thus, and the compression hypothesis seemstenable (Figure 5).
But when TaJmagc (74) tried to check it experimentally,
he ran into an unexpected surprise.
An experimental thickenilig column, as shown in
Figure 6, was operated with thick beds (concentration
varying with height), ranging in depth fmm a b u t 8
to 70 inches, with and without constant conwntration
OWE mor DIAMETERPILOT IHICKIYEI
settling zones backed up above them. The device was
run continuously during most of the years ,1956 and
1957. Concentration profiles observed were in wm-
plete agreement with t h e of Comings (2).
Superelevation tubes measure fluid pressure at their
l o w ends. S U ~ r e I e v a t i ~
levels in die t u b and .in the
ures A& &sed by @.%ion loss agwCaW with flow
through solids a t q e i h e ' m d .
have bcen deacrikd by Hwtt
equipment'& F k n w a l d . or
I n eveq'one of the great number of tests nln, fluid
pnssure at any depth was appr+maMy what would
be predicted if aII solids i n the pmfile Wae' hydro-
dynamically supported. Dati were &a&, but there
yas litde difference defectable
. .
which codd be,&bed
to 9-
Now it might be w ed that squeeze was p-t,
but was small. Hom~er,'& an be s e ~ n.in Figure 4,-
flux,O thmugh any given conccnm+n is sub;stantally
less for a IOW p m a e with a 1
ACTUATED
higb om. Therefore, much
PLUG VALVES carried hydtudyaarnicdly.
be but little mechapid .suppprt high p&, there
still would ha& to be a substantial amount in the u ~ s e
of lowon-. . ..

V O L 5 8 NO. 10 OCTOBER 1 9 6 6 21
 We have shown that such thick beds in which C varies showed the following concentrations: 7-inch column,
with height cannot be in zone settling. But if they are 256 grams per liter solids; 14-inch column, 313 grams
to be explained by compression, substantial squeeze per liter solids; 25-inch column, 324 grams per liter
should have been detectable in lower profiles at least. solids. These are all lower than the 510 grams per
One should not jump to the conclusion that com- liter initial concentration and correspond at least roughly
pression is never a factor. Tory and Shannon (78) to the subsidence rate observed. Lower concentrations
give convincing evidence that it determines final concen- are propagating downward faster than the settling rate
tration in batch settling. But as far as Talmages of the pulp. For this to happen, settling rate must
CaCOn slurries are concerned, it seems necessary to increase with concentration because by Kynch The-
conclude that over the thick bed profile solids arent orem I1
in zone settling and they arent explained by compression
either. We escape from a dilemma but fall into con-
fusion. Something previously uncounted must happen.

ANOMALOUS BEHAVIOR
IN BATCH SElTLlNG
Figure 7 illustrates a type of settling behavior generally du
observed in metallurgical pulps. It shows initial sub-
dCO
sidence rate Ro plotted against initial concentration CQ
in batch tests made at three different initial heights Ho. Thus, the true R us. C plot might run as indicated by the
The solids are gold ore ball-milled to 15% minus 200 dotted line (Figure 7). Such behavior almost has to
mesh. arise from channeling or short-circuiting of fluid through
Initial subsidence rate seems to be more or less in- the higher concentrations. It cannot reasonably be
dependent of initial depth up to the final set at 580 attributed to segregation of pulp because it does not
grams of solids per liter. The final set behaves as happen at lower concentrations, where separation of
might be expected in compression, with deeper columns slower settling solids is most favored (5). It presumably
showing higher initial settling rates. might be attributed to cracking of solids structure
Note, however, that from somewhere around 350 through interparticle forces, except that equivalent
grams of solids per liter through the set of 510 grams per behavior was observed in the flux plot for water-fluidized
liter, subsidence rate RQremains almost constant. There beds of nonflocculent sea sand (4).
is a range of initial concentrations over which initial
settling rate does not vary. Such behavior is de-
scribed by Coe and Clevenger and is shown in three PHASE S m L l N G
of the four pulps for which settling rate data are given by A possible answer is phase settling (4). Kynch
Tahnage and Fitch. arguments show that if a reverse concentration gradient
Samples pipetted from inch below the interface in exists (higher concentration above) it will propagate to a
the 510 grams per liter set after 75 minutes of settling discontinuity if dS/dC is negative and d?S/dcP is positive.
That is, because dS/dC is negative, loci of constant
concentration will propagate upward (Theorem 11).
If @S/dP is positive, lower concentration loci will
propagate upward more rapidly than higher ones.
This gives the usual Kynch behavior with normal
concentration gradients. With a reverse gradient,
lower concentration loci from below overtake higher
ones above. The lower concentration loci do not pass
through higher ones. As shown by Kynch, such a
situation gives rise to a discontinuity.
Now reverse concentration gradients on a large scale
are alien to our thinking, but they must arise locally
due to small random fluctuations in solids concentration.
When they form, they tend to propagate locally to dis-
continuities, with more concentrated pulp above. If
the discontinuity does develop, the faster settling pulp
0.I below may settle away from that above, leaving either a
water-filled parvoid (6) if there is no particle transfer
across the boundary, or a lean phase bubble if there is
limited transfer. Whether or not the bubble will per-
0 100 xx) 300 400 500
sist and grow depends upon the stability of its roof dis-
CONCfNIRATION C, GMS/IITER
continuity, but studies (72) made of voids in fluid beds
Figure 7. Butch seffling fcsfrem/*. on mctnllurgicd+u/p show mechanisms by which bubbles are stabilized.

24 I N D U S T R I A L A N D ENGINEERING CHEMISTRY
 It is concluded, therefore, that in regions of positive As solids transfer to mechanical support at the bottom
flux plot curvature, a lean phase may segregate and of a column, they build up a compression region. The
short-circuit upward through settling pulp. I n ex- upper boundary of this 'region at H, builds up at some
treme cases this is noticeable as intermittent flow of clear rate LI. Any Kynch zones propagating upward at a
liquid through channels. In the case of limited phase lesser rate will be overrun and eliminated ( 1 1, 16, 19).
behavior it will appear as a local swirling or turbulence. If the solids are incompressible, they collapse onto the
The evidences supporting the possibility of phase bottom of the column with some concentration C,, and
settling are many (4). Among the most direct are: have zero flux. The supeqacent Kynch zone feeding
the compression region has some concentration Cd and a
-As discussed above, it can happen. The necessary
flux 8,. By Theorem I, the discontinuity between zone
mechanism exists.
settling and nonsettling pulp will propagate with a
--It is universally observed in gas fluidization. There
velocity a such that
is also evidence of it in water fluidization (4).
-Channeling is observed in batch settling tests.
a=- -S6
? -Something l i e it must happen to explain the CG- cd
anomaliesin settling behavior discussed above.
To be the superjacent zone, C, must propagate with
Phase settling would not occur until past the point of an equal velocity. That is
inflection on a flux plot. In most cases it seems to have
an upper limit, in that there is a maximum amount of
lean phase which can pass through a zone (4). It
probably doesn't extend far into the true compression or
regime because, before a void could form, discontinuity
forces would have to be great enough to overcome
squeeze. Therefore, a flux plot with both phase settling
and compression might look l i e Figure 8, although some 1
Now C, is a constant, dS/dC and Sd are functions of Cd.
qualificationswill appear later. Therefore, from Equation 13, Cd is constant and thus so
is a. The top of the compression zone will build linearly,
as shown in Figure 9. A Kynch zone of concentration
EFFECTS IN BATCH Cdwill propagate upward just above it. Those of lower
SElTLlNC
concentration will propagate from the origin to the
Limitations of Kynch Theorem 111 Constructions pulpsupernatant interface in zone settling all the way.
It is recognized that the Kynch analysis of a single Theorem I11 conditions are fulfilled, and it is valid up to
batch test cannot safkly be used to determine thickener emergence of the compression region at 6.
area (5,13). If the solids are compressible, particles at the top of
T o determine flux as a function of concentration from the compression zone will be subsiding, and S . will not,
a single batch test, Kynch Theorem 111, which is quite re- in general, be zero. It will be so at t = 0, and Equation
stricted, must be used. We now consider the effects of 13 thus holds as t -c 0. However, the top of the com-
these restrictions on batch settling curves. pression region will not build linearly, but will have a

Figure 9. Settling w v c for mmornpessibla pulp showing lincm

Figure 8. Postulatedpux plot showing phase sellling region buildup of compression region

VOL 58 NO. 10 OCTOBER 1 9 6 6 25

slope which decreases with time as more compressing important, are not considered here. We treat two
solids accumulate. questions only: How much area is needed and how
Kynch zones having concentrations below Ca will much depth?
propagate from the origin and for them Theorem 111 There are several possible modes of sedimentation (3,
remains valid. Between Cd and C, zones will form and including: Clarification, in which solids settle either
propagate tangentially from the top of the compression individually or are collected into separated floccules, each
region (Figure 10). They are still in zone settling, but of which settles at its own characteristic rate; zone
do not start from the origin. Theorem 111 construc- settling, in which particles cohere into a structure such
tions are invalid for this part of settling curve. that all in a given neighborhood subside at the same
The above has an interesting consequence. Flux into rate, but the structure does not lend mechanical support
the compression region is determined by the superjacent and u = f(C); compression, in which the structure is
free settling zone and hy the rate at which the top of capable of mechanical support and u = f(C, #); and
the region propagates upward. In the range of initial phase settling, in which gross inhomogeneities form in
concentrations below C,, the zone superjacent to the the solids structure and which may occur in either zone
compression region is independent of initial pulp concen- settling or compression.
tration or depth. Therefore, until the compression Different regimes form one below the other in a con-
region emerges at c (Figure lo), its rate and manner of
growth will be independent of Coand Ho.
Figure 11 shows how to solve Equation 13 graphically
for C, on a free settling flux plot. It also demonstrates
that if solids are taken at an underflow concentration C,,
which is equal to or lower than that, C,, at which they are
capable of supporting compression, Theorem 111 con-
structions are valid because, by Figure 1, the concentra-
tion corresponding to Gamin will then be equal to or
lower than Cd. But in a majority of practical cases
underflow will be desired at a concentration within the
compression regime. In any case, there is no apparent
way to recognize Ca from a single batch test and to know
whether or not one is dealing with the range in which
Theorem 111 is valid. Therefore, it, together with the
Talmage and Fitch corollary, is of little practical value
for thickener design.

H h c l s in Compression Regime

I n his excellent work on batch settling of calcium

carbonate slurries, Tory (77) finds u approximately a
function of C even in the compression regime. How
can this be reconciled with the argument accompanying Figwe 10. Se66/ing c u m for comprcsriblc puQ showing jree-scf6lifig
Figure 4, which shows it cannot be so in continuous m q c in which Thtorm I N is inualid
thickening?
First, from Equation 7, whenever initial column
height Ho is large with respect to yield height H,, and
diameter large with respect to du, initial settling rate Ro
will be approximately u, which is the settling rate with-
out compressive support. Tests made under such
conditions would not give much evidence of compression
even if it exists.
Second, it was argued that the thick bed profiles may
be the result of phase settling rather than compression.
Phase behavior has not been subjected to mathematical
analysis and we dont know in any quantitative way
just what it does.

HOW SHALL WE DESIGN THICKENERS?

To design a thickener, we must know what is needed to
handle a specified flow, keep unwanted solids out of the
overflow, and deliver underflow at the desired concen-
tration. Several design factors enter which, though

26 INDUSTRIAL A N D ENGINEERING CHEMISTRY

h u o u s thickener, although all are not necessarily Zone settling imposes an area requirement. The area
present. For example, zone settling regions will be needed can be determined by the classical Coe and
observed only in overloaded thickeners. Clarification Clevenger batch settling technique, using initial setding
will constitute a design problem only if feed dilution is so rates. Kynch Theorem I11 interpretations of a single
great that a fully cohesive solids structure is not formed. batch test cannot safely be used. Zone settling imposes
A compression region will be absent if underflow solids no depth requirement because such zones have no depth
are withdrawn at a concentration below C,. However, in any but an overloaded thickener.
solids pass downward in sequence through whatever Compression presents an unsolved problem. As
regimes are present. Area provided in the thickener shown by this paper, hypotheses advanced to explain it
must be great enough to satisfy whichever demands the lead to either logical or experimental inconsistencies.
most. Total depth must equal the sum of those re- And much of what has been considered compression
quired by each. Clarification will not be discussed may actually be phase behavior. This leaves us with
here. Although it is limiting in some cases, it is outside no truly rational design procedure based on batch tests.
the context of modern thickening theory. Modern concepts, however, do lead to a tentative
conclusion on area demanded by the thick bed or com-
paction region, whatever it may be (9). All have in
common the idea of a maximum free settling flux for any
given concentration. Actual flux may be less if there
is mechanical support, or if the upper limit of phase
behavior has not been attained, but it will not be greater.
Such free settling flux at any concentration imposes a
minimum demand or lower bound on area. At least
enough must be supplied to accommodate desired solids
rate through whatever concentration shows the lowest
free settling flux, whatever the regime. A corresponding
test procedure is given in the next section of this paper.
The only rationally valid design procedure for com-
paction depth is continuous piloting, or a semicon-
tinuous procedure such as that described by Coe and
Clevenger for chemical pulps. However, such prc-
cedures are time-consuming and are seldom used in
practice.
A frequently used empirical design method is to
determine a detention time for pulp in compaction
according to the procedure recommended by Coe and
Clevenger for metallurgical pulps, and to specify that
the depth of the compaction zone shall not exceed 3
feet. For slowly compacting pulps this results in an
area demand in excess of that required by zone settling.
A procedure based on the Michaels and Bolger
model is given in the last section of this paper. It is
untested, but we feel it should give values for compaction
depth which will be as good a guide as any.
To summarize, modern theory has not basically
changed our design concept for zone settling. It has
rightfully challenged the misconception that underflow
density will always be a function of solids detention only,
but has as yet given no rational alternative basis for
predicting it from batch tests. From an engineering
standpoint modern theory has not answered the pmb-
lems unsolved by Coe and Clevenger, and in this respect
we are not greatly more advanced than they left us half
a century ago.

FREE S E H L I N C TESTS
All hypotheses agree that the initial subsidence rate
in batch tests with large column diameter d and initial
depth HO will be essentially free settling rate u. We
Figwc 73. Plot of p d p daprh os. commotion jromfinol dilution lest don't know aprimi just how large the settling tube should

VOL 5 8 NO. 10 O C T O B E R 1 9 6 6 27
 Note that if point a lies at a concentration whue
(dx/dC)* is not equal to zero, thickeper depth will be
infinite. Even when this is not so, calculated depth
may be greater than acceptable. I n such a case,chwae
a lower value of Ge, draw a new operating line, and
repeat the calculation. By trial and error an acceptable
combination of Ge and depth will be discovered.
Remember, however, that the depth thus calculated
is to be taken only as indicative. There is yet no experi-
mental proof that the compaction depth actually ob-
served in a thickener will equal that calculated by thii or
any other batch procedure.

A
c
--
NOMENCLATURE

-- wnscntration m a s alid./uolumc pulp

d
du - wlumn diameter
yield diameter, diameter at which comant coaccnmtion plug L just u p
D -- ported by &mr

-
dilution, mnrwater/maralid.
G
-
n -
GI
rulingnur sio.oncrttliog
thickener thrmyhput f lu
.
Figure 14. crnphicol infrgrntion for depth of cmpe&m
t h i C k
region in
Hw - height fmm bornm of wivmn
yidd height; height ofmnstmt wncmtrstion plug which cam bc mppmtad
by.

be, but suggest starting with one about 3 feet in working --

&t-
Aii
intmcption adinate axis oflinc d r a m -8-f
Msht of wmtmf wnanmtion plug in wmp-
-
m batch r t t l i i r n
pulp

depth and 4 inches in diameter.

K
ry-- Coc a d o l e v c n p a d i m e ~ l s m n t , L * / M l/pf
p- head nvailpbk w drive E d throvgb " m y -fo

s - -
R ."b.dcnccrateofpvlpnuhecinbatchfat
To make an individual zone test, pour about 3 feet m e *tang flux G i n zone *ttlins

--
--
of slurry at some concentration C, into the tube, mix t -dmc
Y free rrtling rate Of particle. with rupccl w pulp
thoroughly, and then observe subsidence rate of the
sludge-supernatant interface. It will often happen that
n
x -- pchlpl rtuiug *ate O f particla with rcqcet w pulp
diruncc m - 4 in direction of rfflcmcnt in batch lest
in dlstiobn of rtUemenf in mntinvolv th*hna (r -
-
I
rate increases for a time, presumably because floc struc- 8 +nu) w k I L downward mwrmcnt ofpulp Y a whok
ture is changing. It will finally reach a maximum value, a pmpsgadonvslodtyiiodLsrionofrrtlcmcntofmpofmmwrqlion

and this should be used in subsequent calculations. p - in htsb tats; always haJ a rqati"e valve

--
p-ah Vclodly of loem of constant wncmmtion in dlstion o f r t b
mmf
Make zone tests at a series of concentrations ranging 6 pm-donvclodtyofcoaccntradondirondnvityindlshdrttlement

--
from that of thickener feed to that of expected under- p, dcnityofnuid
pa dcodty of -lid parti&
flow. Calculate flux RC for each zone, and plot against w yield valve O f & .rmc- f a wmpraion
Cas shown in Figure 12.
Thickener underflow concentration C . must next be
*
r -yidd"alucof.o~eNcM.fmlhcnr
-squracmcamw.-inlolid.rtrucm

specified. Usually it will be determined with aid of a

final dilution test to be described later. Draw a con- a --
SusacluPTs
rucd valve O f v a r L b L

struction line through C,, on the C axis and tangent to

c
d - valve ofvaiahk at cot~anaw wmplaiDa regime
valucofvnrisbkntpoiofdd~byEquah13
the underside of the flux curve at some concentration C .
It will intercept the RC axis at the value of Gemin which is
0
Y -
-initialvalUc
valveofvariabkinth*haauaddow

an upper bound for the throughput flux possible in a

continuous thickener.
-
SWKFZUUFt
valve ofvariable in batch tat w b m subid- is wmpkte

REFERENCES
(1) Coc,H.S.,Clevmgcr,G.H., T m . M M E 5 5 , 3 5 6 (1916).
COMPACTION DEPTH BASED ON MICHAELS (2) coming,E. W., Im. ENO.Cnau. 32,663 (1940).
AND BOLGER MODEL (3) Fit& B., "Biolo#cal W t m m t of S m w d Industrial Wutq':' Vd. 1.
p. 159, Rdnhold, ov York, 1958.
(4) pitch, B., Im. Jho. Caay. Rnm-*r*u 5, 129 (1966).
Mow pulp to settle in a large deep tube for several (5) Pitch, B., T m . M M E ZS, 192 (1962).
days, or until subsidence is essentiallycomplete. Sample (6) H-tt, N. J., I d , c*nirr 82, 116, 169, 489 (1958); 57. 25 (1961); as%
(1964).
at different depths to determine concentration profile. (7) Kpch, 0.J., T m .ForadqSr. 48,166 (1952).
Plot C os. depth x (Figure 13). From thii test choose (8) A. S.,Bolgcr, J. C., Iw. Em. Caay. p U m * m r r u 8 1.24 (1962).
(9) M o d , A.G., &U. I-,. Mini= Md. 75,729 (1964).
an attainable and acceptable value for C, to use in zone (lo) Ports, J. L.,Suodrrt,H.F.,"E~tr~ti"eMnnUvrsloll\luminum,~"P.95,
tart calculations. Inwrirnee,New Yo*, 1963.
(11) Robmu, E. I., pdvatc wmmuniution (1955).
At a selected value of C determine (dx/dC)* from (12) Rowe, P. N.,H m d , G. A,, T m . Iwl. Clnn. E n p . I r k r J . 4 3 1 (1961).
Figure 13, and the ratio S/(S - G) from Figure 12. (13) Sharam, P. T.,Tory, E. M.,ko. -0. CUaY. 57,18 (1965).
(14) Tal=-, W. P , , p r i v a t ~ c o ~ u n i ~ t i oPm-Oli~cr
n nport) (1959).
The product of these will equal dz/dC for thickener (15) Tal- W. P.,Fiteh, E.B.. Im. Jho. C m r . 47.38 (1955).
operation at the selected value of C. Repeat for the (16) Tory, E. M., Iw. Jho. C a a y . h ~ - 4 , 1 0 6 (1965).
(17) Tory,E.M.,Ph.D.tbd.,RvdveUniwdty,l%l.
full range of concentrations observed in Figure 13. (18) T q , E. M., Shannon, P.T.,Iw. ENS.Cnau. PYIIDUIL*=*LI 4,194 (1965).
Plot dz/dC us. C (Figure 14). Depth of thick bed can be (19) Wall" 0. B. Sympoium on InLpaetion bel- Fluid* .nd P a hSt.
a m . &.,&dm, Junc%ZZ,l962..
determined by graphical integration under the curve (20)Y e , N.,Hofu, Y.,T a d * S., &E& & 'OK IS, 616 (1955).
(21) Y d o k a , N.. Hofu, V., T d a , S., Naito, S.. Twami, S., W.. 21. 66
h m C = OtoC.. (1957).

28 INDUSTRIAL A N D ENGINEERING CHEMISTRY