See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/256352300

Fluid dynamic analysis of the solid-liquid separation process by

centrifugation

Article  in  Journal of Agricultural Engineering Research · January 2000

CITATIONS READS

16 1,083

2 authors:

Riccardo Amirante Pasquale Catalano

Politecnico di Bari Università degli Studi del Molise
129 PUBLICATIONS   1,584 CITATIONS    64 PUBLICATIONS   456 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Ultrasound into the olive oil extraction process View project

Experimental and Numerical Investigations on Particle Number and Mass Emissions from Spark-Ignition Engines View project

All content following this page was uploaded by Riccardo Amirante on 25 February 2014.

The user has requested enhancement of the downloaded file.

J. agric. Engng Res., 2000, 77 (2), 193}201
doi:10.1006/jaer.2000.0586, available online at http://www.idealibrary.com on
PH*Postharvest Technology

Fluid Dynamic Analysis of the Solid}liquid Separation Process by Centrifugation

R. Amirante; P. Catalano
Department of Animal, Plants and Environmental Science, University of Molise, Via De Sanctis, I-86100 Campobasso, Italy;
e-mail of corresponding author: catalano@foodeng.unimol.it

(Received 16 March 1999; accepted in revised form 23 June 2000; published online 15 August 2000)

In this paper, a theoretical analysis and preliminary experimental trials of the solid}liquid separation process by
centrifugation has been carried out to evaluate the modi"cations which would improve the performance of the
centrifuge.
A new dimensionless number, the drag number e , has been de"ned as the ratio between the mean settling
B
time of the smallest particles and the mean drag time necessary to remove all the settled solids.
A high drag number for the centrifugal extraction of olive oil allows the transition zone between the two main
directions of the #uid motion to reach the oil layer. Consequently, it is important to avoid this interaction,
otherwise a mixing between water and oil will occur. In this case, it is necessary to correctly evaluate the optimal
value of the drag number in order to leave a thick water layer over the sediment.
Finally, it is pointed out that about 75% of the solids present in the olive paste, used in preliminary
experimental trials, settle almost immediately, close to the feed opening. This tendency was avoided by allowing
the di!erential velocity to vary as a function of the characteristics of the olive paste and/or by using barriers
correctly located near the feed opening.
 2000 Silsoe Research Institute

1. Introduction conveyed towards the conical part of the decanter,

The decanter centrifuge is commonly used to rapidly
separate the solids suspended in a #uid for a wide spec-
trum of processes, such as de-watering raw or digested
sludge, thickening activated sludge, and extracting the oil
from the paste obtained by olive crushing.
Figure 1 shows the working principle of a decanter
centrifuge: the separation is obtained by rotating the
bowl at a quite high speed generating a centrifugal force
varying from 1000 to 6000 g, or more for some particular
applications ('10000 g), where g is the acceleration due
to gravity (Madsen, 1989a). The #uid enters through
a feed pipe and is accelerated by this centrifugal force.
The suspension #ows towards the end of the cylin-
drical part of the decanter with a gradually less amount
of solids content. In fact, the solid particles, being heavier
than the liquid and subjected to the centrifugal force,
move towards the bowl wall and settle in a relatively
short path with di!erent velocities, according to their
mean diameter and density (Stokes law). The sediment is
0021-8634/00/100193#09 $35.00/0
and discharged through the solids outlet, by a screw
conveyor rotating at a speed slightly slower than that of
the bowl.
In order to correctly evaluate the separation e$ciency
of such machines, many studies were carried out in the
last 20 years (Amirante & Catalano, 1995; Amirante
et al., 1993; Gupta Santosh, 1978; Madsen, 1989b).
However, some important simpli"cations were made, in
particular considering a constant #uid velocity pro"le,
neglecting the di!erent paths followed by di!erent-sized
particles and the drag e!ect due to the screw conveyor
#ights.
All these theories led to the classical &-formula (Am-
bler, 1952) or similar, where & is de"ned as the equivalent
area of a settling tank having the same separation
e$ciency.
In this paper a deeper knowledge of the phenomena
occurring in the decanter is achieved by better develop-
ing the #uid dynamic analysis of the solid}liquid separ-
ation process.
193  2000 Silsoe Research Institute
194 R . AM I RA N TE ; P. C ATA LA N O

Notation

a external centrifugal acceleration equal w longitudinal component of the #uid ve-

A X
to R u, m/s locity pro"le, m/s

C content of solids in the feed section as w angular component of the #uid velocity
 F
volume of total solids suspended in the pro"le, m/s
unit mass of #uid = longitudinal component of the #uid ve-
J
C (r , z ) content of solids at the point having locity evaluated at the free surface, m/s
Q ? ?
co-ordinates (r , z ): as C = screw conveyor dragging velocity equal
? ?  B
D mean diameter of suspended particles, m to ¸ Du, m/s
N A
D ,D minimum and maximum value of the z longitudinal co-ordinate, m
KGL +
particle diameters, m z axial dimensionless co-ordinate equal
?
D ,D average and 25th percentile of the par- to z/¸
 
ticle diameters, m b axial shape factor of the centrifuge
f (C ) Stokes law correction factor depending equal to ¸/R
Q 
on the solid content e drag number
B
g acceleration due to gravity, m/s c"R /R radial shape factor of the centrifuge
 
k coe$cient of the curve used to "t the u(D , r , z ) solid particle diameter distribution
E N ? ?
cumulative diameter distribution U (D ), u (D ) solid particle diameter distribution near
 N  N
m\ the feed opening
¸ decanter length, m U (D ) cumulative distribution of the variable
 +
¸ screw conveyor pitch, m D
N +
M, N coe$cients k #uid viscosity, N s/m
D
p pressure, N/m k liquid-phase viscosity, N s/m
J
Q decanter #ow rate in the feed section, m/s h angular co-ordinate, rad
Q #ow rate of solids, m/s o #uid density, kg/m
Q D
r radial co-ordinate, m o liquid-phase density, kg/m
J
r normalized radial co-ordinate equal to o solid-phase density, kg/m
? Q
r/R p standard deviation of the particle dia-
 "
R ,R inner and external bowl radii of the meter distribution
 
liquid ring, m R equivalent area of settling tank for com-
DR liquid ring thickness, m parable separation e$ciency, m

Dr sediment layer thickness, m u bowl angular velocity, rad/s
Dr normalized sediment layer thickness Du di!erential angular velocity, rad/s
?
S normalized dynamic productivity coef-
"cient Superscripts
S solid #ow coe$cient (50) values of any variable referred to as
Q D "D
t mean drag time, s N 
B (min) values of any variable referred to as
t mean settling time, s
Q D "D
¹, ;, P coe$cients N KGL
v radial component of the solid particle (opt) variables and parameters relative to the
P present theory in the optimal working
velocity, m/s
v longitudinal component of the solid conditions
X (old) variables and parameters relative to old
particle velocity, m/s
v settling velocity due to gravity (modi- theories
QE (¸) values of any variable referred to as
"ed Stokes law), m/s
w radial component of the #uid velocity z"¸
P
pro"le, m/s

2. Theoretical considerations
To correctly develop the #uid-dynamic analysis of the
solid}liquid separation by centrifugation, it is necessary
to consider the decanter as a cylindrical free surface canal
where the solid particles settle under the centrifugal force
instead of by gravity.
The momentum equations, in steady and laminar con-
ditions, may be written, using the cylindrical co-ordinate
system de"ned in Fig. 1, as follows (Navier}Stokes
 S O LI D- L I QU I D S EPA RA TIO N P RO C ES S B Y C E NT RIFU G A TI O N 195

therefore, can be considered as constant: this hypoth-

Fig. 1. Schematic representation of a decanter; R1, R2 and DR2,
inner and outer radii and thickness of liquid ring; r, z and h,
radial, longitudinal and angular co-ordinate; u, bowl angular
velocity; w , longitudinal component of the yuid velocity proxle;
X
= , longitudinal component of the yuid velocity at the free surface;
J
= , drag velocity of screw conveyor; v (D ), radial component of
B P N
solid particle velocity as a function of particle diameter; v ,
X
longitudinal component of particle velocity
esis was justi"ed later, by proving that most of the
solids settle within 10% of the decanter length.
Moreover, the following conditions can be hypo-
thesized:
(a) w "w "0, since the secondary motions of high-
P F
viscosity #uids are negligible (Gao and Hartnett,
1993) and the helicoidal motion in the screw con-
veyor does not a!ect the centrifugal force applied to
the solid particles;
(b) *w /*h"*w /*z"0, because of the almost cylin-
X X
drical symmetry of the problem, and
(c) p"constant, since the motion is equivalent to the free
surface motion in a horizontal canal with an overfall.
Under these hypotheses, the system of equations,
Eqn (1), can be reduced to


1 * *w
r X "0 (2)
r *r *r
equations):
As the co-ordinate system is "xed on the rotating


*w w *w w *w external surface, the settled solids move, in the chosen
o w P# F P! F #w P "
D P *r r *h r X *z co-ordinate system, with a drag velocity = :
B
!= "!¸ Du/2n (3)

 
*p * 1 * 1 *w 2 *w *w B N
! #k rw # P! F# P ,
*r D *r r *r P r *h r *h *z where ¸ is the pitch of the screw conveyor and Du is the
N
di!erential angular velocity between the conveyor and


*w w *w ww *w the external bowl. Neglecting the thickness of the solid
o w F# F F# P F#w F "
D P *r r *r r X *z layer D at the internal bowl radius R , the #uid near this
P 
layer (r"R ) also #ows in the direction opposite to the


 
1 *p * 1 * 1 *w 2 *w *w
! #k rw # F# P# F , main #ow with a velocity != . Therefore, the boundary
B
r *h D *r r *r F r *h r *h *z conditions are


*w w *w *w w (R )"= , w (R )"!=
X  J X  B
(4)
o w X# F X#w X "
D P *r r *r X *z where R is the inner radius of the liquid ring; and the

velocity of the #uid at the free surface = must be evalu-


 
*p 1 * *w 1 *w *w J
! #k r X # X# X (1) ated on the basis that the delivery in a generic cross-
*z D r *r *r r *h *z section is equal to the given entry #ow Q.
Solving Eqn (2) and imposing the boundary conditions
where o is the #uid density in kg/m; k is the #uid
D D in Eqn (4) gives
viscosity in N s/m; w , w and w are the radial, longitu-
P X F


dinal and angular components of the #uid velocity pro"le ln r ln r
w "= ?#= ?!1 (5)
in m/s; r, z and h are the radial and longitudinal and X J ln c B ln c
angular coordinates in m and rad, respectively; and p is
the pressure in N/m. where r is the dimensionless radial co-ordinate given by
?
The use of these equations is justi"ed by the following r/R , and c is the radial shape factor of the centrifuge,

hypotheses. given by R /R .
 
Once the #uid velocity pro"le has been computed, the
(a) Even if most of the organic #uids are non-Newtonian, settling solids path can be evaluated: in fact, the solid
this has a little in#uence on the velocity pro"le particles are carried by the #uid along the Z-axis (see
because of the high value of the apparent viscosity Fig. 1) with a longitudinal velocity v equal to that for the
X
k (Kostic, 1993). #uid velocity w and settle toward the external surface of
D X
(b) The viscosity k , depending on the suspended solids the decanter with a radial velocity v , because of the
D P
concentration, changes little with the position and, centrifugal "eld existing in the #uid.
196 R . AM I RA N TE ; P. C ATA LA N O

The settling velocity can be approximately evaluated
using the modi"ed Stokes law, where a Stokes law cor-
rection factor f (C ), depending on the solids content C ,
Q Q
has to be included. But, if the Reynolds number for the
particles is less than unity, as usually veri"ed in most of
the separation processes by centrifugation because of the
high viscosity of the #uid, the particles settling velocity
Owing to the high viscosity of the #uid, the transition
between the turbulent #ow (near the feed opening) and
the laminar one can be neglected. In this hypothesis, the
distribution of solids u(D , r , z ) becomes independent
N ? ?
of r only near the feed opening (z "0) because of the
? ?
solids mixing in the turbulent #ow:


*u(D , r , z )
v is given by (Stokes law)
P N ? ? "0 (10)
*r
? X?
D (o !o )R u D a
v" N Q J  r "vKGL N A r (6) The last condition allows the estimation of the solids
P 18k ? QE D g ?
J KGL concentration C (r , z ) on the basis of the initial concen-
Q ? ?
where D is the mean diameter of suspended particles; tration C and of the particle distribution at the feed
N 
o is the solid phase density; o is the liquid phase density; opening u (D ) " u(D , r , 0). In fact, because all the
Q J  N N ?
u is the bowl angular velocity; k is the liquid viscosity; settling particles (D *D ) at the point (r , z ) arrive
J . KGL ? ?
vKGL is the settling velocity due to gravity referred to as from the feed cross section, the solids concentration
QE C (r , z ) can be evaluated considering only those par-
D "D , the mean diameter of the smallest settling Q ? ?
N KGL ticles that, in the feed section, have mean diameter
particles; a is the external centrifugal acceleration equal
A D )D (r , z ); therefore, the following expression for
to R u; and g is the acceleration, due to gravity. . + ? ?
 C (r , z ) can be obtained:
Therefore, in steady conditions and laminar motion of Q ? ?
the solids, the average path followed by all the particles "+P? X?
having mean diameter D could be represented by the
N
following equation (see Fig. 1):
C (r , z )"C
Q ? ?  
"KL
u (D )dD "C U [D (r , z )] (11)
 N N   + ? ?

dz R v 1v where U (D ) is the cumulative distribution of the solid

?"  X" X (7)  +
bv particles of diameter D )D in the feed section.
dr
?
¸ v
P P N +
If the concentration C is known, it is possible to
Q
where z is the axial dimensionless co-ordinate relative to evaluate the delivery of the solids Q , across a generic
? Q
the decanter length ¸; and b the axial shape factor of the cross-section:
centrifuge, that is the ratio of the decanter length to outer
 L
radius. The integration of Eqn (7) gives Q (z )"R
Q ?  
A 
C (r , z )v r dr dh
Q ? ? X? ?



1 P? v

gD ln r
z" X dr*" KGL = ?!ln c

A CQ (r? , z?)vX r?dr?
? b v ? 2ba vKGLD J ln c
A P A QE N "2nR

(12)


ln r
#= ?!2ln r #ln c (8) Generally, U (D ) is unknown, making it impossible
B ln c ?  +
to compute C by means of Eqn (11). Owing to the great
Q
where r* is the variable of integration. number of casual events in#uencing the distribution of
?
Equation (8) shows that all the particles, having dia- solid particles in the most common #uids, however, the
meter D and entering at the position r "c, z "0, move assumption of a Gaussian distribution introduces a negli-
N ? ?
longitudinally while settling (c(r (1) towards the gible error.
?
position z (r ). Under this hypothesis, if the diameter D becomes
? ? +
Thus, the diameter D of the solid particles present at small enough to have
N
the point with co-ordinates (r , z ) coming from any
? ? U [D (r , z )];1 ∀c)r )1 (13)
radial position of the feed section is always less than the  + ? ? ?
diameter D of those particles that, settling from position (this condition being veri"ed for D )D !0)5
+ + 
(c, 0), have not yet moved beyond the radius r , that is,
? p
D , where D is the average particle diameters;
using Eqn (8): "  
D is the 25th percentile of the particle diameters; and

p is standard deviation of the particle diameter distribu-



gD ln r "
D) KGL = ?!lnc tion, meaning that about 75% of the suspended solids
N 2ba vKGL z J ln c have been settled), the function U (D ) could be evalu-
A QE ?  +
ated using the following approximation:


ln r
#= ?!2ln r #ln c "D (r , z ). (9)
B ln c ? + ? ? U(D )
k (D!D ) ∀D *D (14)
N E N KGL N KGL
 S O LI D- L I QU I D S EPA RA TIO N P RO C ES S B Y C E NT RIFU G A TI O N 197

where D can be set as D "D !3p . The value of
KGL KGL  "
the correlation coe$cient between the Gaussian distribu-
tion and the "tting curve given by Eqn (14) is 0)98, when
D )D !0)5p and the coe$cient k "0)041/p .
N  " E "
This hypothesis, however, can be accepted only
for those cross-sections where at least 75% of the solids
have already been removed, that is for z *zKGL (see
? ?
Appendix A):
D g =
zKGL" KGL [!ln(c)] J (15)
? D 2ba vKGL = #=
 A QE J B
Moreover, integrating Eqn (12), the following relation-
ship results for the #ow coe$cient of solids S (z ), a
Q ?
dimensionless number related to the delivery, across a
generic section, of the solid particles whose diameter D is
N average, still. This zone plays an important role in many
greater than D (see Appendix B):
KGL
1 and olive oil extraction, both by centrifugation. In the
S (z )" [¹S#(;e !z )S#Pe] (16)
Q ? z B ? B
? the solid content in the puri"ed liquid it is convenient to
where S is the dynamic productivity coe$cient, a dimen-
sionless number related to the feed delivery; e is a dimen-
B B
sionless number, previously (Amirante & Catalano, 1995)
named the drag number, related to the screw conveyor
angular velocity and pitch (see Appendix B); and ¹,
; and P are coe$cients.
In particular, the drag number e is equal to the ratio
B
between the mean settling time tKGL of a particle having
Q
diameter D and the mean drag time t , that is the time
KGL B
necessary to remove the sediment using the screw con-
veyor. It characterizes a set of similarly operating ma-
chines in the following way: the smaller the value of e ,
B
the more negligible is the e!ect of transport by the con-
veyor on the solid particles that are still settling. These
considerations remain acceptable until all the particles
fall with a velocity near the value given by the modi"ed
Stokes law.
If the thickness of the liquid ring is reduced too much
(cP1), or the value of the viscosity k increases too
D
much (this happens, for example, when the initial concen-
tration of solids C is too high), the particles should settle

with a velocity smaller than the value furnished by the
modi"ed Stokes law, eventually independent of their
diameter (mass sedimentation) in the worst case.
In order to maintain the sediment carrying capacity of
the conveyor unchanged (= "constant) under these
B Q
circumstances, the value of the drag number e should be
B MNR
increased to o!set the reduction in the value of vKGL when
QE
the concentration C increases. As signi"cant variations
 (17)
of e could lead to appreciable worsening of the decanter
B
performance, the same e!ect can be obtained by increas-
ing the value of the external radius R or the angular
 by means of the following equation:
velocity u, or, preferably, by extending the cylindrical
part of the decanter through an increase in the value of b.
On the contrary, if the concentration of solids is
continuously and eventually rapidly varying during the
separation process, which is possible in some speci"c
processes, for example in olive oil extraction by centrifu-
gation, it should be necessary to provide the plant with
a control of the di!erential angular velocity Du, based on
the viscous characteristics of the #uid. In other words, as
a continuous variation of the decanter length is imposs-
ible (except by a shift in the feed section, a solution which
has yet to be experimentally veri"ed), the only way to
maintain e constant is to vary Du (so varying = ) in the
B B
function for vKGL, for example, by a measure of the torque
QE
applied to the screw conveyor.
Another important concept is represented by the inver-
sion surface, that is the surface where the #uid is, on
processes and, in particular, in waste water concentration
"rst case, as the main aim is to reduce as much as possible
shift the inversion surface toward the inner radius by an
increase of e , thus improving the e$ciency of the separ-
ation process. However, the inversion surface is not well
de"ned, but it corresponds to a more or less wide
transition zone between the two directions of motion: the
main direction and the dragged one. If the drag number
is increased too much, then it is possible that the
transition zone becomes very near to the free surface with
the consequential formation of undesirable turbulence.
Therefore, an optimal value of e exists and it should be
B
used in the design of the decanter. In the case of olive oil
extraction by centrifugation, this value is smaller than the
corresponding one obtained for waste water concentra-
tion. In fact, in olive oil extraction a second transition
zone between the separated oil layer and the vegetable
water exists. It is important, therefore, to avoid the inter-
action between the inversion surface and the oil layer by
reducing e , otherwise a mixing of oil and water will
B
occur. Another reason to correctly evaluate the optimal
value for the drag number e is to leave a water layer,
B
moving towards the outlet, between the separated oil and
the sediment.
In any case, the optimal value for the dynamic produc-
tivity coe$cient S can be calculated by assuming that
MNR
the content of solids in the #uid leaving the decanter is
negligible (S "0, given z "1).
?
Recalling Eqn (16), S must satisfy the following
relation:
¹S #(;e !1)S #Pe"0.
MNR B MNR B
The absolute value of the feed delivery can be obtained
QMNR"2vKGL RMNR (18)
QE
198 R . AM I RA N TE ; P. C ATA LA N O

where RMNR"nRb(a /g)S is the equivalent area of When the optimal value for the feed delivery is used
 A MNR
the settling tank for comparable separation e$ciency, (S"S ) then Eqn (22) can be re-written as
MNR
known as R-value (Gupta Santosh, 1978).
In the hypothesis for optimal regulation of the feed S 1!z
Dr (z )"CKGL (!ln c) MNR ?#Dr*. (23)
delivery of the decanter, Eqn (16) becomes ? ? Q e z ?
B ?
1!z Therefore, the thickness of the sediment [Eqns (22) or
S (z )" ?S (19) (23)] varies following a hyperbolic curve for z *zKGL.
Q ? z MNR ? ?
?
showing a hyperbolic decrease in the delivery of solids
when moving away from the feed section. 3. Results and discussion
Finally, applying the solids balance principle to a con-
trol volume located near the bowl surface, as shown in The theoretical study of the #uid motion in a decanter
Fig. 2, the height of the solid phase Dr can be evaluated in and the dynamics of the solid particles in the centrifugal
any cross-section located at a distance z *zKGL from the sedimentation lead to results [see Eqns (16) and (18)]
? ?
feed section: similar in form to those of many simpli"ed studies
(Amirante et al., 1993; Ambler, 1952; Gupta Santosh,
2nR = Dr(z #dz )!2nR = Dr(z )#Q (z ) 1978):
 B ? ?  B ? Q ?
dDr 1 dQ (z ) QMJB"2vRMJB (24)
!Q (z #dz )"0, or ?" Q ? (20) QE
Q ? ? dz 2nR= dz
?  B ? where
where Dr is the dimensionless ratio of the thickness of n(R!R)¸u nRba (1!c)
?  
the solids layer to the outer bowl radius (Dr "Dr/R ). RMJB" "  A
?  gln(2R/R#R) g ln(2/1#c)
Solving Eqn (20) gives   
nRba
1 "  A SMJB(c) (25)
Dr (z )" Q (z )#Dr* (21) g
? ? 2nR= Q ? ?
 B
and v is the settling velocity due to gravity referred to
where Dr* represents the minimum value of the solid- QE
? as D "D ; and c is the normalized bowl inner radius
phase thickness evaluated at z "1. N 
? (c"R /R ).
Substituting Eqn (16) into Eqn (21), the sediment thick-  
However, the delivery obtained in optimal operating
ness becomes conditions using Eqn (18) is less than that computed by
Eqn (25), with vKGL(v and RMNR(RMJB (so
¹S#(;e !z )S#Pe QE QE
Dr (z )"CKGL(!lnc) B ? B #Dr*. QMNR(QMJB).
? ? Q ez ? In fact, as shown in Fig. 3, the thin line representing the
B ?
(22) trend for SMNR, evaluated by means of Eqn (17) for di!er-
ent values of the drag e!ect e (e "0, e "2 and e "12)
B B B B
where Dr* is the sediment layer thickness evaluated at is always below the bold line representing the trend of
?
z "1. SMJB. In a previous paper (Amirante et al.,1993), the trend
?
of SMJB was analysed and it was noticed that, considering
the actual values used for c (ranging from 0)3 to 0)6) in
many industrial decanters, while an increase of c (cP1)
leads to an increase in the decanter productivity (RMJB),
reducing c (cP0), so that the variation of RMJB becomes
negligible (bold line in Fig. 3).
On the contrary, the present deeper analysis points out
that an excessive decrease of c (corresponding to an
increase in the thickness of the liquid ring) can lead to an
unacceptable worsening of the decanter performance
(thin line in Fig. 3). Analysing the behaviour of SMNR, it
Fig. 2. Solids balance in a control volume located near the bowl
surface in a generic cross-section; R2, outer radius; z , axial can be seen that it is convenient to reduce the thickness of
? the liquid ring as much as possible, but taking into
dimensionless co-ordinate; Dr, sediment layer thickness; = , drag
B
velocity of screw conveyor; Q , solid yow rate
Q
account that the mean #uid velocity, and, of course, that
 S O LI D- L I QU I D S EPA RA TIO N P RO C ES S B Y C E NT RIFU G A TI O N 199

not well removed by a #uid with an initial high content of

solids.
This condition, in the case of the olive oil extraction
process, can be easily achieved when working without
added water (water saving decanters) and avoided
adopting alternative technical solutions as barriers suit-
ably located and/or automatic control of the variation of
the di!erential velocity Du between bowl and screw
conveyor.
All these considerations, coming out from the theoret-
ical analysis, have led many factories to adopt some or all
of these suggestions with very interesting improvements
in machine performance.

4. Conclusions

Fig. 3. Trend of the normalized dynamic productivity coezcient

In this paper, a comprehensive theoretical analysis of
S depending on the ratio of decanter length ¸ to outer radius the solid}liquid separation process by centrifugation has
R2 (c) and the drag number e ( , old theory; , optimal been carried out to evaluate the modi"cations which
B
value of this theory) would improve the performance of the centrifuge, leading
to new interesting results.
Three dimensionless numbers have been de"ned: the
of the solid particles, increases. Thus, it is not convenient solids #ow coe$cient, related to the delivery, across
to increase c over some limit in order to avoid turbulence a generic section, of the solid particles whose diameter
in the motion of the solid particles which could lead to an D is greater than the minimum particle diameter D ;
N KGL
undesirable mixing of the solids in the #uid. the dynamic productivity coe$cient, related to the feed
Finally, the theoretical analysis has highlighted the delivery; and the drag number, related to the screw con-
importance of the minimum distance zKGL, that is the veyor angular velocity and pitch.
?
distance of the cross-section where 75% of the total In particular, the dynamic productivity coe$cient
amount of solids to be settled has been separated. replaces the R-value, de"ned as the equivalent area of
Equation (23) points out that zKGL mainly depends on a settling tank for comparable operation e$ciency, be-
?
the square ratio between the minimum diameter D of cause it allows consideration of similarity conditions.
KGL
the settled solids and that corresponding to the 25th The need for de"ning the drag number e is derived
B
percentile of the particle diameter distribution. directly from the comprehensive theoretical analysis,
Some experimental trials on the olive oil extraction where the drag e!ect of the screw conveyor on the #uid
process have been carried out in order to evaluate this has been considered. The drag number characterizes a set
ratio and compute the distance zKGL in the hypothesis
?
that the decanter feed delivery has been optimally regu-
lated [see Eqn (18)]. Olives of the Coratina Cultivar and
an industrial hammer crusher were used.
In Fig. 4, the cumulative distribution of the particles
diameter and the trend of a Gaussian integral curve, best
"tting the real distribution, are shown. Using the "tting
curve, whose mean value for D is 2)73 mm and whose

standard deviation p is 0)88, the minimum particle dia-
"
meter D (0)1 mm) and the one corresponding to the
KGL
25th percentile D (2)3 mm) can be calculated. The re-

sulting diameter ratio D /D is 0)043.
KGL 
Supposing c in the range 0)3}0)7 and e (20, zKGL can
B ?
be evaluated (zKGL)0)1). This result is very important
?
because of its quite low value. In fact, it means that the Fig. 4. Cumulative distribution of the particle diameters in the
most of the solids accumulate near the feed section, paste obtained by hammer crushing the olives used in the experi-
leading to possible problems due to an excess of sediment mental trials; D , particle diameter
N
200 R . AM I RA N TE ; P. C ATA LA N O



of similarly operating machines in the following way: the gD lnr
D (r , z )" KGL = ?!ln c
smaller the value of e , the more negligible is the e!ect of + ? ? 2ba vKGLz J ln c
B A QE ?
transport by the conveyor on the solid particles still


ln r
settling. On the contrary, a high drag number, for #= ?!2ln r #ln c )D (A1)
B ln c ? 
example for the centrifugal extraction of olive oil, allows
the transition zone between the two main directions of for any value of r . Solving this with respect to z :
? ?
the #uid motion to reach the oil layer. Consequently, it is


D g ln r
important to avoid this interaction, otherwise a mixing z * KGL max = ?!ln c
? D 2ba vKGL P? J ln c
between water and oil will occur. In this case, it is neces-  A QE


sary to correctly evaluate the optimal value of the drag ln r
#= ?!2ln r #ln c (A2)
number in order to leave a thick water layer over the B ln c ?
sediment.
D g =
Finally, it is pointed out that about 75% of the solids " KGL [!ln (c)] J "zKGL
D 2ba vKGL = #= ?
present in the olive paste, used in preliminary experi-  A QE J B
mental trials, settle almost immediately, close to the feed
opening. This tendency was limited by allowing the dif-
ferential velocity to vary as a function of the character- Appendix B
istics of the olive paste and/or by using barriers correctly
The concentration C [see Eqn (11)] can be computed using
located near the feed opening. Q
D and U evaluated by Eqns (9) and (14), respectively:
In conclusion, the present theoretical analysis repre- + 

 

sents a more accurate model than previous ones of the gD lnr
C (r , z )"k C KGL = ?!ln c
centrifugal solid}liquid separation. The obtained results Q ? ? E  2ba vKGLz J ln c
A QE ?
have been a valid aid in the design of the so-called

 
lnr
third-generation decanters. #= ?!2ln r #ln c !D . (B1)
B ln c ? KGL

Substituting this result into Eqn (11), the solids delivery in any
cross-section can be evaluated:
References
 

Q (z ) CKGLg  lnr
Q ?" Q rv = ?!ln c
2nR 2ba vKGL D z ? X J ln c
Ambler C M (1952). The evaluation of centrifuge performance.  A QE KGL ? A
Chemical Engineering Progress, 48(3), 150}158


Amirante R; Catalano P (1995). Extraction of olive oil by lnr Q
#= ?!2ln r #ln c dr !CKGL
centrifugation: #uid-dynamic aspects and assessment of new B ln c ? ? Q 2nR
extraction plant solutions. OLIVAE, 57, 44}49 
(B2)
Amirante, P; Di Renzo G C; Di giovacchino L; Bianchi B;
Catalano P (1993). Technological developments in olive oil where CKGL"k C D .
extraction plants. OLIVAE, 48, 43}53 Q E  KGL
In order to compute the longitudinal component of the #uid
Gao S X; Hartnett J P (1993). Steady #ow of non-Newtonian velocity at the free surface, = , which is the only unknown in
#uids through rectangular ducts, International Communica- J
Eqn (B2), the feed delivery has to be calculated, using the
tions on Heat and Mass Transfer, 20, 197}210 expression for the #uid velocity given by Eqn (5):
Gupta Santosh K (1978). A new equation for the sigma value of

A r?wX dr?
conical bowl discharge centrifuges. Chemical Age of India, 29
Q"2nR (B3)
Kostic M (1993). In#uence of viscosity function simpli"cation 
on non-Newtonian velocity and shear rate pro"les in rectan-
gular ducts. International Communications on Heat and and that integration gives
Mass Transfer, 20, 515}525



Madsen N F (1989a). Slender decanter centrifuge. Inter- Q 1 1!c 1 1!c
national Chemical Engineering Symposium Series No. 113, "! = 2c# ! = 2# .
2nR 4 J ln c 4 B ln c
pp 281}299 
Madsen B (1989b). Flow and sedimentation in decanter centri- (B4)
fuge. International Chemical Engineering Symposium Series Solving Eqn (B4) with respect to = gives
No. 113, pp 301}316 J
Q
= "M(c) #N(c)= (B5)
J 2nR B

Appendix A where

Recalling Eqns (9) and (14), 75% of the solids have been 4
M(c)"!
settled when D )D : 2c#(1!c)/ln c
+ 
 S O LI D- L I QU I D S EPA RA TIO N P RO C ES S B Y C E NT RIFU G A TI O N 201

 


and 1  ln r ln r
;(c)"! r M(c) ? N(c) ?!ln c
2ln c ? ln c ln c
2#(1!c)/ln c A
N(c)"! (B6)


2c#(1!c)ln c ln r
# ?!2ln r #ln c dr
ln c ? ?
Substituting Eqn (B5) into Eqn (B2) and recalling that

 

1  ln r ln r
v "w , gives
X X P(c)" r N(c) ?# ?!1
2ln c ? ln c ln c
A

 
Q (z ) CKGLg  M(c)Q ln r
Q ?" Q ?


 
r
? 2nR ln c lnr ln r
2nR 2ba vKGL D z ; N(c) ?!ln c # ?!2ln r #ln c dr
 A QE KGL ? A  ln c ln c ? ?


ln r ln r S (z ) is the solid #ow coe$cient, a dimensionless number
#= N(c) ?# ?!1 Q ?
B related to the delivery, across a generic section, of the solid
ln c ln c
particles whose diameter D is greater than D , given by
N KGL





M(c)Q ln r ln r Q (z )
Q ?
; ?!ln c #= N(c) ?!ln c S (z )" (B9)
2nR ln c B ln c Q ? a
 A
2nRbvKGL CKGL
 QE g Q


ln r CKGLQ
# ?!2ln r #ln c dr ! Q (B7) S is the dynamic productivity coe$cient, a dimensionless num-
ln c ? ? 2nR ber related to the feed delivery, given by

Q
Dividing Eqn (B7) by bvKGL (a /g)CKGL gives S" (B10)
QE A Q a
2nRbvKGL A
1  QE g
S (z )" [¹S#(;e !z )S#Pe] (B8)
Q ? z B ? B and e a dimensionless number, named the drag number, related
? B
to the screw conveyor angular velocity and pitch, given by
where
=
e "(!ln c) B (B11)
 

1  ln r lnr B a
¹(c)" r M(c) ? ?!ln c dr bvKGL A
2 ? ln c ln c ? QE g
A

View publication stats