Biosystems Engineering (2006) 95 (3), 405412

doi:10.1016/j.biosystemseng.2006.07.005
PHPostharvest Technology
Separation Theory for Palm Kernel and Shell Mixture on a Spinning Disc
O.A. Koya
1
; M.O. Faborode
2
1
Department of Mechanical Engineering, Obafemi Awolowo University, Ile-Ife, Nigeria;
e-mail of corresponding author: afemikoya@yahoo.com
2
Department of Agricultural Engineering, Obafemi Awolowo University, Ile-Ife, Nigeria; e-mail: mfaborod@oauife.edu.ng
(Received 2 November 2005; accepted in revised form 19 July 2006; published online 1 September 2006)
The separation of palm kernel from the shell is an important process in the recovery of the kernel for use in
vegetable oil production. The inherent shortcoming of the fertiliser spinner spreader, resulting in non-uniform
distribution, has led to investigations into the possibility of its use in the separation of palm kernel from shell.
The differences in the physical properties of the kernel and shell provided the basis for separation. Models
describing the motion of the kernel and the shell on a spinning disc with vanes, feeding a segment at a time,
were developed, and comparisons between the predictions and experimental results were made.
Discharge angles obtained from the predictive models and from the experiment were in close agreement.
The experiment, however, indicated a mid-span, between the discharge angles for kernels and shells,
containing the mixture. The models reasonably predict approximate ranges of discharge angles for palm
kernels and shells on a spinning disc with specied diameter, friction coefcient and rotational speed.
r 2006 IAgrE. All rights reserved
Published by Elsevier Ltd
1. Introduction
In spite of the importance of perfect shell separation
from kernels for the process of industrial palm kernel oil
extraction, existing devices and separation methods have
limitations in solving the problem completely. The
greatest difculty is with kernels and shell particles of
comparable sizes, and this is where most existing
separators are defective.
The two most common industrial varieties of the oil
palm fruits in the extraction of palm kernels are the
Dura and Tenera, which are classied as thick-shelled
and thin-shelled fruits, respectively. However, the
densities of the kernel and shell for both fruit types
are close (Hartley, 1977; Opeke, 1997; Koya et al.,
2004), though, Badmus (1991) reported that future
breeding activities might lead to Tenera fruits with
thinner nutshells that will facilitate kernel and shell
separation. Large areas of existing oil palm plantations
(Faborode, 2004; Owolarafe et al., 2002) is a major
motivation for the renewed interest in the separation of
kernel from shell (Akubuo & Eje, 2002; Koya et al.,
2004).
Palm kernel and shell separators are of two types
based on the wet and the dry methods. The wet method
uses clay slurry, brine solution or a hydrocyclone, and
generally exploits the differences in the densities of the
kernel and the shell as the physical basis for separation
(Hartley, 1977; Henderson & Perry, 1980; Opeke, 1997).
However, the densities of the kernel and shell are very
close so that separation efciency depends upon the
quality and proper control of the specic gravity of the
solution used. In addition, the wet kernels must be
sterilised against moulds and dried for 1416 h in silos to
remove moisture picked up by the kernels. Therefore,
separation systems without moisture absorption will
result in savings in the energy used in re-drying the
kernels in the wet methods.
Dry separation methods include systems exploiting
the winnowing principle followed by handpicking,
reciprocating inclined plane, vibrating and reciprocating
screens, rotating screens, indented cylinders, and pneu-
matic conveying in a vertical tunnel. The traditional
hand picking method gives a high separation efciency
but it is time consuming, tedious, and of very low
capacity of about 60 kg h
1
(Akubuo & Eje, 2002).
ARTICLE IN PRESS
1537-5110/$32.00 405 r 2006 IAgrE. All rights reserved
Published by Elsevier Ltd
Efciencies reported in respect of other mechanical
systems need considerable improvements (Hartley, 1977;
Akubuo & Eje, 2002; Badmus, 1991; Olie & Tjeng,
1974). Aerodynamic forces required in pneumatic
systems are quite unstable and uneven because the
densities of kernel and shell are close, while the shell
particles also have irregular shapes and are of varied
sizes. Generally, shell particles and kernels having
comparable sizes make screening and other proprietary
devices inefcient.
Having examined the shortcomings of existing sys-
tems, the possibility of adopting the centrifugal method,
hitherto used in fertiliser spreaders, was considered. This
in essence would imply exploiting what is considered a
disadvantage in fertiliser spreading, that is, the tendency
to unevenly spread and segregate materials (Patterson &
Reece, 1962; Inns & Reece, 1962; Brinseld & Hummel,
1975; Kasza et al., 1995; Aphale et al., 2003; Reumers
et al., 2003). The spinner spreader consists of a rotating
disc with vanes attached to the disc surface. Fertiliser
fed onto the spinner is thrown onto the ground after
colliding with the rotating vanes and moving radially
outwards to the periphery of the disc.
Theoretical and experimental studies of particle
trajectories on and off the spinner spreader have been
widely investigated (Aphale et al., 2003; Reumers et al.,
2003; Patterson & Reece, 1962). Patterson and Reece
(1962) and Aphale et al. (2003) analysed the motion of
fertiliser particles on rotary spreaders and showed that,
particle trajectories depended on particle shape, and
coefcient of friction on the disc. Brinseld and
Hummel (1975) observed, among other things, an
increase in discharge velocity and dispatch angle as the
coefcient of friction increased. However, in the
separation of palm kernel from the shell, the product
will be collected on leaving the disc, hence, particle
motion from the feed position to the edge of the
spinning disc becomes relevant. Shell particles, though
irregular in shape, are mostly at or dish-shaped,
while whole kernels are ellipsoidal or nearly spherical
(Akubuo & Eje, 2002; Koya & Faborode, 2005).
The work reported here formulates appropriate
relations governing the differential dispatch of kernel
and shell from a spinning disc, when both are fed from
the same point. Experiments to validate the theoretical
proposition resulted in the development of a schema for
the design of centrifugal separator, which holds great
promise.
2. Theoretical analysis
The motion of particles on a rotating disc was
analysed, treating the mixture (kernel and shell) as
aggregate of disconnected particles with no interaction
between them. Each particle was assumed to be in
contact with the disc at all times. It was assumed that
the motion of each constituent (kernel or shell) is
represented by the motion of a proprietary single
particle. It was also assumed that the particles were
fed onto to the disc without impact, such that bouncing
does not occur.
The problem, therefore, is to consider a particle
dropping onto a rotating disc at near the centre of the
disc and to describe its motion on the disc. The disc is
turning with constant angular velocity. The discharge
angle, dened as the displacement through which the
disc turns before the particle drops off, depends on
the time upon which the particle remains on the disc.
The principal factors, which control the motion of the
particle, are its coefcient of friction relative to the disc
and the vane, and its shape. Where the disc and the vane
are made of the same materials, the coefcients of
friction on both surfaces will have the same value. The
shape of the particle will decide if it will slide or roll. The
analysis is presented for both pure sliding (to approx-
imate the motion of at shell particles) and pure rolling
(approximating the motion of the spherical kernels).
ARTICLE IN PRESS
Notation
A, B, C,
D
constants of integration
a,b boundary conditions radial distances at
time t is zero, L
F force applied to the rolling particle by the
vane, MLT
2
g gravitational acceleration, LT
2
I mass moment of inertial, ML
2
M,N characteristic roots of differential equation
m mass of particle, M
n particle radius, L
R disc radius, L
r radial distance of particle from the centre of
the disc, L
t time, T
a angular displacement of particle, dimen-
sionless
m friction coefcient between the particle and
the disc/vane, dimensionless
f discharge angle, dimensionless
o angular velocity of the disc, T
1
O.A. KOYA; M.O. FABORODE 406
2.1. Pure sliding
Applying Newtons second law, the motion of a
particle on a rotating disc with a vane, considering
frictional forces generated by its Coriolis and particle
weight, is as follows:
d
2
r
dt
2
2mo
dr
dt
o
2
r mg (1)
where: r is the radial distance of the particle from the
centre of the disc; dr/dt is the radial velocity of
particle at distance r; d
2
r/dt
2
is the radial acceleration
of particle at distance r; m is the coefcient of
friction between the particle and the disc/vane; and o
is the angular velocity of the disc; and g is the
gravitational acceleration. Solving this differential
equation
r
mg
o
2
Ae
oMt
Be
oNt
(2)
where: A and B are constants of integration; and M and
N are characteristic roots given by
M

m
2
1

m (3)
N

m
2
1

m (4)
Hence, for all real values of m, M and N are positive.
From Eqn (2),
dr
dt
AoMe
oMt
BoNe
oNt
(5)
d
2
r
dt
2
Ao
2
M
2
e
oMt
Bo
2
N
2
e
oNt
(6)
Comparing Eqns (2) and (6) d
2
r=dt
2
is zero or
negative if romg/o
2
.
Consequently, for the particle to slide radially on the
disc, i.e., d
2
r=dt
2
40 then, romg=o
2
is required.
Therefore, taking as initial conditions: r a where,
a4mg=o
2
; an dr/dt 0, when t 0:
A
a
mg
o
2

N
M
1
N
M
(7)
and
B
a
mg
o
2

1
N
M
(8)
So that
A B
N
M

(9)
Identical solutions have been obtained by other
researchers such as Patterson and Reece (1962), and
Aphale et al. (2003).
Following from Eqns (7) and (8), Eqn (2) becomes
r
mg
o
2
B
N
M
e
oMt
e
oNt

(10)
If the particle slides progressively, and without
impact, from the feed position to the edge of a disc,
with radius R, rotating with a constant angular velocity
o, the discharge angle f can be estimated from the
following:
N
M
e
f

M
e
f

R
mg
o
2

B
(11)
2.2. Pure rolling
Rolling without slipping will occur when the angular
velocity of the particle matches its radial velocity, so
that:
dr
dt
n
da
dt
(12)
where n is the particle radius and a is the angular
displacement.
If F is the force causing the rotation of the particle
about its centroid, then
Fn I
d
2
a
dt
2
(13)
where I is the mass moment of inertia.
Assuming a solid, spherical particle (I 2/5 mn
2
),
resulting in
F
2
5
m
d
2
r
dt
2
(14)
where m is the main of the particle.
Applying Newtons law of motion and assuming that
F is less than the limiting frictional force mmg then
m
d
2
r
dt
2
o
2
r

F mmg (15)
or
d
2
r
dt
2

5
7
o
2
r mg
Solving Eqn (16)
r

7
5

mg
o
2
Ce

5=7

ot De

5=7

ot (16)
where C and D are constants of integration.
ARTICLE IN PRESS
SEPARATION THEORY FOR PALM KERNEL AND SHELL MIXTURE 407
Applying boundary conditions
R b where b4

7
5

mg=o
2
and dr/dt 0, when t 0
we obtain
r

5
7

mg
o
2
b

5
7
mg
o
2

cos h

5
7

ot (17)
which is identical to the equation derived by Patterson
and Reece (1962) and Aphale et al. (2003).
Consequently, for a disc with radius R turning at a
constant angular velocity o, the angular displacement f
can then be written as
f

7
5

cos h
1
R

7
5

mg
o
2

b

7
5
mg
o
2

(18)
2.3. Application of the theory to separation of kernel and
shell
From the foregoing theory, important parameters
required in predicting the motion of a particle on a
spinning disc are its shape, radial distance from centre to
the location where the particle is fed, and the radius of
the spinning disc. The separation will be feasible if there
are signicant differences in the discharge angles for the
kernels and the shells.
Separation was considered based on the premise that
the kernel slides on the disc and rolls on the vane while
the shell slides all the way. Therefore, a comparison of
Eqns (2) and (17) will describe their relative motion and,
the possibility of separating kernel from shell using this
concept (Koya, 2004). However, agricultural products
are usually not perfectly spherical or at, so that the
actual motion of the kernel and shell may differ from
what is obtained in the theory, thus, requiring experi-
mental verication.
3. Materials and methods
3.1. Physical properties of palm kernel and shell
The Dura variety of the oil palm, characterised as
having thick-shelled fruits, was chosen for study. Values
of the required properties of the kernel and shell, shown
in Tables 1 and 2, were taken from Koya (2004) and
Koya et al. (2004). Some of the parameters were also
reported by Akubuo and Eje (2002) for a mixture of
palm nut varieties.
Half of the geometric mean diameter provided an
estimate for the particle radius n in Eqn (12). Static
coefcients of friction were reported (Table 1) but the
models developed require dynamic coefcient of fric-
tion; consequently, predictions based on the models
will result in larger discharge angles than expected
in practice. As an approximation, therefore, the
lower limits of the ranges of observed static coefcients
were substituted in the equations. The highest coef-
cients of friction as well as, where the largest
difference between the values for kernel and shell were
observed was on the jute bre. The friction coefcients
were, therefore, taken as 068 for kernel, and 095
for shell.
ARTICLE IN PRESS
Table 1
Size-related properties of kernels (Dura variety) relevant in product separation
Property Measurements for kernels on size, mm
10 14 20 25
Size distribution, % retained 25 738 220 14
Axial dimensions, mm 530 (188)

1946 (184) 1700 (174) 1960 (123)

Major diameter 994 (103) 1282(188) 1300 (209) 1322 (111)
Intermediate diameter 832 (113) 934 (105) 918 (224) 874 (100)
Minor diameter 1082 1326 1266 1313
Geometric means, mm 171 068 074 067
Sphericity 074 137 119 133
Average mass, g

Numbers in parentheses are the standard deviations. Sphericity is the ratio of geometric mean diameters to the major diameter. Source: Koya
(2004).
Table 2
Friction coefcient of kernel and shell (Dura variety)
Structural surface Friction coefcient
Kernel Shell
Galvanised sheet metal 039 (001)

038 (001)
Plywood 051 (001) 052 (001)
Jute bre (new) 068 (001) 095 (001)

Numbers in parentheses are the standard deviations Source: Koya

(2004).
O.A. KOYA; M.O. FABORODE 408
3.2. Apparatus for verication
It was necessary to fabricate an apparatus, shown in
Fig. 1, for the verication of the theory. A 600 mm
diameter wooden disc with straight radial vanes was
used. The disc and the vanes were overlaid with jute
bre to provide same friction coefcients on the disc and
the vanes, in consonance with the assumption in theory.
Samples were fed onto the disc through a stationary
100 mm diameter tube, which was nearly touching the
disc. The tube discharges through a sector of its curved
surface: 301 subtended at the centre, and 50 mm high.
The material was, therefore, drawn from the tube to feed
a segment of the disc at a time, as the disc rotates about
the tube.
The disc was driven by a variable speed motor. A
rectangular wooden collection tray, 12 220 mm
2
, was
provided, and placed 50 mm below the spinning disc.
The tray has 12 radial compartments to guide in
estimating the discharge angle, when the material drops
off the disc.
3.2.1. Experimental design and methods
Nut samples were cracked in a machine, and the
mixture was screened using a sieve (Endecotts Ltd,
London) with 10 mm aperture to remove dust, small
shell particles and small-sized or broken kernels. Other
(larger) broken kernels were picked out by hand. This
procedure was adopted to provide for investigation into
separation between whole kernels and shell particles of
comparable sizes mainly. Samples of the mixture were
fed in batches, and the disc driven at 950 min
1
. The
speed was chosen after some preliminary trials to
determine the minimum speed when appreciable percen-
tage of the kernels and shell moved gently, to reach the
edge of the disc in a single rotation. Discharge angles,
relative to the feed-on position, for kernels, shells and
kernel-shell mixture were estimated using the divisions
on the collection tray. Each experiment was replicated
ve times.
In order to determine the discharge angles from the
theory, appropriate values taken from Tables 1 and 2 for
the input parameters were substituted in Eqns (11) and
(18). The equations were then solved numerically, for
disc radii of between 50 and 300 mm; it was assumed
that the disc turned through 0301 before the individual
particle leaves the feeding tube. The mean discharge
angles obtained from the theory were compared with the
values obtained from the experiment, using paired t-test
(Barnes, 1994).
4. Results and discussion
4.1. Predictions from theory
From the theory, the feed-on positions required for a
particle to move radially outward on a spinning disc is
given by a4mg/o
2
in pure sliding, and b47 mg/(2o
2
) in
rolling, yielding a409 mm and b424 mm, assuming
that shells slid and kernels rolled all the way. These
values are practically smaller than the minor diameter of
the kernel, and much smaller than the chosen 50 mm
diameter of the feeding tube. This implies that the
kernel, owing to its physical dimensions, will always
leave the centre of the disc.
Solutions to Eqns (11) and (18) for incremental radial
distances from the centre of the disc are plotted in the
polar chart of Fig. 2. The ranges of the discharge angles
are 1702001 for kernels and 2642941 for the shells.
4.2. Experiment
In the experiment, kernels were discharged between
901 and 1501 of the disc rotation. Shell particles were
discharged from 2101 through a complete rotation of the
disc. Some shell particles, sticking to the disc at the end
of a complete rotation, were swept off using a brush.
The range 1802101 consists of kernel and shell mixture,
which in practice should be fed back into the machine.
Samples of the product (kernel, shell and the mixture)
obtained from the separation on the spinning disc are
shown in Fig. 3.
ARTICLE IN PRESS
Fig. 1. The apparatus for verication of the theory of centrifugal
motion: A, speed control panel; B, spinning disc; C, collection
tray; D, electric motor
SEPARATION THEORY FOR PALM KERNEL AND SHELL MIXTURE 409
As expected, the discharge angles observed in the
experiment are lower than the values predicted from
theory. However, the possibility of separating kernel
from shell using the tested concept is conrmed.
Statistical analysis, comparing mean discharge angles
from the theory with the values from the experiment for
the kernels and the shells is presented in Table 3. The
statistics does not reveal signicant difference between
the results from theory and experiment. The theory is,
therefore, suitable for estimating discharge angles, using
at discs of different diameters, and friction coefcients.
4.3. Proposed schema for palm nut cracking and product
separation
Based on the results of this work and an exploitation
of the benets of existing relevant systems, a possible
schema for integrated palm nut cracking and product
separation is proposed (Fig. 4). The system consists of
three sub-units namely, a multi-stage nutcracker, a pre-
cleaner and a centrifugal separator. The multi-stage
nutcracker is operated at a speed lower than obtained in
the conventional nutcracker, and unbroken nuts are
recycled. This approach results in negligible kernel
breakage while obtaining nearly 100% cracking ef-
ciency (Koya, 2004). The operating speeds of the
nutcracker (1100 min
1
) can then be synchronised with
the speed of the centrifugal separator (driven at
950 min
1
).
The pre-cleaner is basically a set of sieves, where dirt
and small shell particles are discarded while the pre-
cleaned kernel and shell mixture are introduced into the
centrifugal separator. The centrifugal separator, based
on the result of the present work, will be provided
with three discharge chutes; for kernels, shells and
unseparated kernel/shell mixture. The mixture will be
re-introduced into the machine. However, the perfor-
mance test of the prototype and the effect of feed rates
will determine optimum operating conditions. The
ARTICLE IN PRESS
50100
F
e
e
d
30
60
90
120
150
180
210
240
270
300
330
D
i
r
e
c
t
i
o
n

o
f

d
i
s
c

r
o
t
a
t
i
o
n
S
h
e
l
l
s
K
e
r
n
e
l
s
150 200 250 300 Radius, mm
Fig. 2. Theoretical trajectories of kernels and shells on a spinning at disc, rotating at 950 min
1
Fig. 3. Image of separation products from the spinning disc: (a)
palm kernels at 901501 of disc rotation; (b) mixture of palm
kernels and shells at 1802101 of disc rotation; (c) palm shells at
more than 2101 of disc rotation
Table 3
Comparison of mean discharge angles from the theory and the
experiment
Product Mean discharge angles, deg
Theory Experiment Difference

Kernel 185 120 65

Shell 279 285 6

Computed t-value is 0588, which is not signicant at probability

Po01
O.A. KOYA; M.O. FABORODE 410
development of the integrated palm nutcracker and
kernel/shell separator is the focus of future work.
5. Conclusions
The phenomenon of uneven distribution and segrega-
tion of material particles, which is disadvantageous in
the use of fertiliser spinner spreaders, was theoretically
examined for possible application in palm kernel and
shell mixture separation. The differences in the physical
properties of the constituent materials such as shape,
and friction coefcients of the kernel and shell, provided
the motivating physical basis for the separation.
Experimental verication of the concept approximately
corroborated the predictions from theory, thus conrm-
ing the feasibility and viability of using the spinning disc
in the separation of kernels from the shells.
References
Akubuo C O; Eje B E (2002). Palm kernel and shell separator.
Biosystems Engineering, 81(2), 193199, doi:10.1016/biosys-
temeng.2001.0029
Aphale A; Bolander N; Park J; Shaw L; Svec J; Wassgren C
(2003). Granular fertiliser particle dynamics on and off a
spinner spreader. Biosystems Engineering, 86(3), 319329,
doi:10.1016/S1537-5110(03)0062
Badmus G A (1991). NIFOR automated small-scale oil palm
fruit processing equipmentits need, development and
cost effectiveness. A Journal Publication of the Nigerian
Institute for Oil Palm Research, (NIFOR) Benin City,
Nigeria, pp 2031
Barnes J W (1994). Statistical Analysis for Engineers and
Scientists: A Computer-Based Approach. McGraw-Hill,
Inc., New York
Brinseld R B; Hummel J W (1975). Simulation of a new
centrifugal distributor design. Transactions of the ASAE,
18(2), 213220
Faborode M O (2004). Oil palm production improvement
programme. Technical Report, New Nigeria Foundation,
Lagos
Hartley C W S (1977). The Oil Palm. Longman Publishers,
London
Henderson S M; Perry R L (1980). Agricultural Process
Engineering. The AVI Publishing Company, Incorporated,
Westport, CT
Inns F M; Reece A R (1962). The theory of the centrifugal
distributorII: motion on the disc, off-centre feed. Journal
of Agricultural Engineering Research, 7, 7884
Kasza D; Csizmazia Z; Polyak I (1995). A general approach
to computer simulation of fertiliser particle motion on
spreader disc and vane of arbitrary shape. Proceedings of
the Agricultural and Biological Engineering Conference,
Poland, 2023rd September 1995
Koya O A (2004). Studies of the mechanics of cracking and
centrifugal separation of palm kernel and shell mixture.
Unpublished PhD Thesis, Obafemi Awolowo University,
Ile-Ife, Nigeria
Koya O A; Faborode M O (2005). Mathematical modelling of
palm nut cracking based on Hertz theory. Biosystems
Engineering, 91(4), 471478, doi:10.1016/j.biosystemeng.
2005.05.011
Koya O A; Idowu A; Faborode M O (2004). Physical properties
of palm kernel and shell relevant in nut cracking and
product separation. Journal of Agricultural Engineering and
Technology, 12, 2739
Olie J J; Tjeng T D (1974). Kernel Recovery In The Extraction
of Palm Oil, The Netherlands Stork Amsterdam, Amstelv-
een. pp 6372
ARTICLE IN PRESS
Nut
grading
Multistage
nutcracker
Broken
nuts
Pre-cleaners
Dirt and
smaller shell
particles
Unbroken
nuts
Kernels and
shell
Centrifugal
separator
Kernels
Shell
Mixture
Fig. 4. Proposed schema for integrated palm nut cracking and product separation
SEPARATION THEORY FOR PALM KERNEL AND SHELL MIXTURE 411
Opeke L K (1997). Tropical Tree Crops. Spectrum Books
Limited, Lagos, Nigeria
Owolarafe O K; Sanni L A; Edinowe O S; Ogunyomi O T;
Faborode M O; Ajibola O O (2002). Development of an oil
palm fruit stripper. Journal of Agricultural Engineering and
Technology, 10, 5964
Patterson D E; Reece A R (1962). The theory of the centrifugal
distributorI: motion on the disc, near-centre feed.
Journal of Agricultural Engineering Research, 7,
345353
Reumers J; Tijskens E; Ramon H (2003). Experi-
mental characterisation of the cylindrical distribution
pattern of centrifugal fertiliser spreaders: towards an
alternative for spreading hall measurements. Biosystems
Engineering, 86 (4), 431439, doi:10.1016/j.biosystemseng.
2003.09.002
ARTICLE IN PRESS
O.A. KOYA; M.O. FABORODE 412