Chapter 9
Solids separation processes
M. J. LEWIS, Department of Food Science and Technology, The University of Reading,
RG6 6AP
9.1 INTRODUCTION
This chapter will cover the separations involving solid foods, together with the properties
of those solids which will influence that separation. Some mention will also be made of
handling and transporting solids and preparatory processes, such as size reduction. The
separation of solids from liquids and solids from gases is not covered in detail in this
chapter, although a summary of the methods based on sedimentation and filtration is
given in Table 9.1. In these applications, the term solids refers to discrete particles
suspended within the fluid and not those dissolved or in the colloidal form, for which a
range of other operations for their removal or separation is available. The objective may
be to recover the solid for further processing or to remove the solid which may be
contaminating the liquid or gas. The method selected also depends upon whether the solid
is to be retained or discarded.
To illustrate some of the difficulties in selecting solids separation methods, the re-
moval of solids from gases will be further illustrated. This can be achieved by classifiers,
cyclones, bag filters or electrostatic precipitators. In cyclones on milk powder plant,
particles less than 5-10 pm may be lost. Cyclone losses of 0.35-1.0% of total production
have been cited for dairy products. Such losses are now unacceptable for environmental
reasons. High-efficiency cyclones have been used, whereby secondary air is introduced
into the cyclone to increase the efficiency. However, these cyclones are not very success-
ful with powders containing fat, as considerable free fat is generated and the powder
sticks to the interior surface of the drier. Therefore it is not possible to install a milk drier
where the powder recovery system consists of cyclones alone. Wet systems such as
scrubbers have been installed, using the pasteurised milk, prior to evaporation, as the
scrubbing liquid, thereby recovering the fines and heat. From a recovery standpoint, this
would seem an excellent solution. However, from a hygiene and quality standpoint, these
proved almost impossible to operate without bacteriological contamination. Most of these
have now been removed (Knipschildt, 1986). The solution to the problem has been
provided by bag filters, which are capable of reducing the particle concentration from
244 M. J. Lewis
Table 9.1. Summary of mechanical solid separation techniques
Solids from liquids
Sedimentation:
Principles: gravity, centrifugal, electrostatic, magnetic centrifugation
Examples: gravity settlers, centrifugal clarifiers, hydrocyclones; use of chemical floc-
Filtration: (see also Chapter 8; fat fractionation)
Principles: gravity, vacuum, pressure and centrifugal
Examples: sand and cake filters, rotary vacuum filters, cartridge and plate and frame
filters, microfilters (Chapter 5); use of filter aids
culants or air flotation
Solids from gases
Principles: sedimentation and filtration
Examples: cyclones, bag filters, electrostatic precipitators
200 mg m-3 to below 10 mg m-3 air. The powder can be recovered from the bags and the
‘clean air’ can be used for heat exchange. Further details are provided by Knipschildt
(1986).
However, rather than removing all the particles, there may be a requirement to
fractionate the powder, based on particle size (see Sections 9.3 and 9.4). This example
illustrates the theme for this chapter, where the main emphasis is placed on the separation
of components from within a solid matrix. Solids come in many forms, shapes and sizes,
so the first part of the chapter will be devoted to discussion of the main properties of solid
foods which will influence the different types of separation processes.
9.2 PHYSICAL PROPERTIES OF SOLIDS
Solids come in a wide variety of shapes and sizes. All solid foods are particulate in nature
and there are a wide range of sizes and shapes to contend with. Some examples are
illustrated from the different food sectors in Table 9.2. It should be noted that although all
these foods are regarded as solids, their moisture content may range from less than 10%
to greater than 90%. Their moisture content and chemical composition can be found from
foods composition tables, for example Paul and Southgate (1978) (see also Chapter 2).
Indeed, one of the main objectives is often to remove selected components from the food.
Some operations where separations from solids is involved and constitutes an impor-
tant part of the process are:
cleaning of agricultural produce (see Section 9.6.3);
sorting and size grading, particularly for quality grading of fruit and vegetables;
peeling of vegetables, dehulling of cereals and legumes and deboning or shelling of
meat and fish;
fractionation or recovery of the main components within the foods, e.g. proteins, fat,
carbohydrates and minerals.
Solids separation processes 245
Table 9.2. Some examples of solid foods
Fruit: apples, oranges, grapes, blackcurrants, pears, bananas
Vegetables: potatoes, carrots, sprouts, peas
Cereals and legumes: rice, wheat, soyabeans, cowpeas, sorghum
Animal produce: large carcasses, small joints, minced meats, fish fillets, prawns, shrimps
Beverages: coffee beans, tea leaves, instant powders and granules
Other powders: milled products, powders produced by drying and grinding methods
and other shellfish
A special range of operations and an area of increasing interest is concerned with the
separation or fractionation of solids, in their particulate or powder form, and their recov-
ery from other materials. In this chapter, emphasis will be placed on the separation of
powders, based on factors such as size and shape, density differences, flow properties,
colour and electrostatic charge, An important pretreatment for many such operations is
size reduction. Methods of size reduction are discussed in Section 9.3.1. Size reduction
increases the surface area and the surface area to volume ratio, thereby enhancing rates of
heat and mass transfer.
However, in some cases very fine powders provide processing problems, and size
enlargement or agglomeration may be used to improve flow characteristics and
wettability.
Many foods which are solid in appearance, will also flow if the shear force provided is
great enough, e.g. butter, spreads and starch doughs. This behaviour is known as
plasticity. The flow behaviour of powders is also important and is discussed in more
detail in Section 9.2.7. Some of the important physical properties of solid foods are listed
in Table 9.3. These are discussed in more detail by Lewis (1990), Jowitt et al. (1983,
1987), Mohsenin (1984, 1986) and Peleg and Bagley (1983). Many of these properties
are influenced by the chemical composition of the food, and in particular its moisture
content.
Of special interest in this context is the behaviour of particulate systems and the
separation of mixtures. Many such separations are based on density differences. In some
cases the powders may be subjected to various forces, gravitational, which are slow
Table 9.3. Physical properties of solids
Appearance, size, shape, size distribution, colour
Specific gravity, particle density, bulk density, porosity, overrun (for aerated products)
Thermal properties; specific heat, latent heat, thermal conductivity, thermal diffusivity,
Rheologial properties; plasticity, elasticity, viscoelasticity, hardness
Electrical conductance or resistance, electrical charge, dielectric constant, dielectric loss
Diffusion and mass transfer characteristics
specific enthalpy
factor
246 M. J. Lewis
compared to centrifugal forces, drag forces or electrical, electrostatic or magnetic forces.
Also, the flow characteristics and behaviour of food powders are markedly different to
those of fluids.
Some of the physical properties of food powders will now be considered in more
detail, especially those which will influence the effectiveness, quality and nature of the
separation process.
9.2.1 Classification of powders
Powders can be characterised in a large number of ways; Peleg (1983) gives some
examples:
by usage: e.g. flours, beverages, spices, sweeteners;
by major component: e.g. starchy, proteinaceous, fatty;
by process: e.g. ground powders, freeze-dried, agglomerated;
by size: e.g. fine, coarse;
by moisture sorption characteristics: e.g. hygroscopic;
by flowability : free flowing, sticky, very cohesive.
Further classification could be by hardness, by explosion potential or by microbial
hazards. Hayes (1987) summarises a detailed system used for characterising a wide range
of food powders based on density, size, flowability, abrasiveness, a range of miscellane-
ous properties and hazards such as flammability, explosiveness and corrosive nature.
Some important physical, chemical and functional properties of powders are given in
Table 9.4. For products such as beverages, the palatability and sensory characteristics of
the reconstituted products are important and may be variables considered when grading
these products. Care should also be taken to ensure that the microbial count is within
acceptable limits for the products.
Determination of some of these properties for milk powders is described in publica-
tions by the Society of Dairy Technology (SDT, 1980), and Schubert (1987a).
Table 9.4. Factors contributing to
the quality of powders
Appearance
Size and shape
Wettability
Sinkability
Solubility
Dispersibility
Bulk density and particle density
Palatability
Nutrient content
Microbiological quality
Solids separation processes 247
9.2.2 Particle size and particle size distribution
As mentioned earlier, food powders come in a wide range of sizes and shapes. Uniform
shapes, such as spheres, can be characterised by one dimension, i.e. the diameter, whereas
two or more measurements may be required for more complex shapes. Whatever the
shape, there are several methods available to characterise the size and particle size distri-
bution. Virtually all operations that result in the production of a powder, e.g. milling or
spray drying, will give rise to a product with a distribution of particle sizes and this
distribution is of extreme importance and will affect the bulk properties. Particle size may
range over several orders of magnitude, ranging from less than 1 pm to as large as
hundreds or even thousands of microns for some large granules. Particle size can be
measured in principle by measuring any physical property which correlates with the
geometric dimensions of the sample. According to Schubert (1987a) the attributes used to
characterise particles may be classified as follows:
geometric characteristics, such as linear dimensions, areas or volumes;
mass;
settling rates;
interference techniques such as electrical field interference and light or laser scattering
or diffraction.
Based on these attributes, the following methods have been used for food materials:
microscopy or other image scanning techniques;
wet and dry sieving methods;
electrical impedance methods such as the Coulter counter;
laser diffraction patterns, such as the Malvern, Northrup and Cilas instruments.
Since particles can vary in both shape and size, different methods of particle size analysis
do not always give consistent results, both because of the different physical principles
being exploited, but also because size and shape are interrelated. Sampling is also
important to ensure that a representative sample is taken, usually by the method of
quartering.
Whatever method of measurement is used, a large number of particles must be meas-
ured in order to ascertain the particle size distribution. It has been suggested for light
microscopy that 200 measurements are made on each of three separate slides (Cloutt,
1983); this makes the method very tedious. The simplest way to present such results is in
the form of a distribution curve, the two most common being in the form of either a
frequency distribution (histogram) or a cumulative distribution (see Fig. 9.1). The cumu-
lative distribution can be based on percentage oversize or percentage undersize. Percent-
age undersize is used more often. The method used for data collection may give a
distribution in terms of number of particles (for example by counting) or the mass
(weight) of particles (for example by sieving).
If the number of particles is known, the distribution can be represented by a frequency
distribution. Table 9.5 gives some typical figures for the number of particles collected by
microscopical examination, arranged into numbers falling within different size ranges (0-
10 pm) etc., together with the frequency distribution and cumulative number distribution
undersize.
248 M. J. Lewis
100
80
60
10 40
20
0
LL *OB 0 0 20 40 Size (prn) 60 80 100 0
Fig. 9.1. (a) Frequency distribution (F), (b) cumulative distribution (0: see also data in Table 9.5.
Table 9.5. Frequency distribution
Size range Mean diameter Number Frequency Cumulative Cumulative
OLm) of range (pm) in range distribution distribution volume
distribution
0 to 10 5 5 1.8 1.8 0
10 to 20 15 15 5.6 7.4 0
20to30 25 35 12.9 20.3 2.0
30to40 35 50 18.4 38.7 7.0
40to50 45 55 20.3 59.0 20.0
50to60 55 50 18.5 77.5 41 .O
60to70 65 32 11.8 89.3 64.0
70to 80 75 20 7.4 96.7 85.0
80to90 85 8 2.9 99.6 98.0
90to 100 95 1 0.4 100.0 100.0
>loo 0
27 1
-
mean diameter = 45.96 pm; d2/1 = 53.29 pm; d3/2 = 58.86 pm
Cumulative number frequency indicates the percentage of the total number less than the mean diameter of the
range.
Cumulative volume represents the percentage of the total volume less than the mean diameter of the range.
Other values which may be calculated from the distribution include the mean diameter
and the median diameter and the standard deviation, which gives an indication of the
spread.
The simplest is the mean diameter, defined as
C nidi /C ni 7
Solids separation processes 249
where ni is the number of particles in class i and di is the mean diameter of class i. The
median diameter is the diameter which cuts the cumulative distribution in half. The d2/1
ratios and d3/2 ratios are also calculated.
However, one widely used characteristic is the Sauter mean particle diameter (d&
This is calculated from
d3p = xnid:/xnid,? (9.1)
This gives the diameter of the particle having the same surface area to volume ratio as the
entire dispersion.
The surface area/volume ratio = 6/d3/2.
(9.2)
Rates of heat transfer and mass transfer are proportional to the surface area to volume
ratio. Therefore the surface area exposed has a big influence on physical properties, e.g.
wettability, dispersion, dissolution and chemical reactions, such as oxidation, as well as
the forces acting at the surface of powders. Equation (9.2) demonstrates that decreasing
d312 will increase the surface area to volume ratio.
Such data can be converted to frequency or cumulative distribution based on surface
area or volume, by calculating the surface area and volume of each range. These cumula-
tive distributions based on numbers and volume are compared in Fig. 9.2. This distinction
is made because the shape of a numbers distribution and a mass or volume distribution is
quite different because the area and volume distributions are most influenced by the
larger diameter particles, since the volume = 4 m3. For example, it can be seen that only
10.7% of the particles are greater than 65 pm, whereas on a volume basis, 36% by
volume are greater than 65 pm (Fig. 9.2). The weight fraction distribution would be
similar to the volume fraction distribution, provided that the solid density is independent
of particle size. The volume distribution is a common form of presentation in emulsion
science, since it is often the larger particles which are likely to cause separation problems.
Therefore it can be very informative to know what fractions by volume are bigger than a
particular size. For example, in cream separation in milk, problems may arise from a
relatively small number of large fat globules.
100
40
0 liL 20 0 0 20 40 60 80 100
Size (pm)
Fig. 9.2. Comparison of volume distribution (V) and cumulative number distribution (N). See
also data in Table 9.5.
250 M. J. Lewis
Most of the discussion has focused upon spherical particles or those closely approxi-
mating to these. However, the particle shape is also very likely to be important and a
wide variety of shapes are also found. Irregular-shaped objects are more complicated to
define and a number of characteristic dimensions have been used to represent them. Some
are given in Table 9.6.
Table 9.6. Characteristic diameters for irregular shaped particles
4 Surface diameter
dv Volume diameter
dd Drag diameter
The diameter of a sphere having the same surface area
as the particle
The diameter of a sphere having the same volume as
the particle
The diameter of the particle having the same resistance
to motion as the particle in a fluid of the same density
and viscosity
The width of the minimum square aperture through
which the particle will pass.
4 Sieve diameter
Other dimensions include the free-falling diameter and Stokes diameter, the projected
area diameter and the specific surface diameter. In many cases the shape is more complex
and a large number of dimensions would be required to describe the size and shape.
Image analysis methods, whereby an image of the object is transferred to a computer
screen and software is available to do any number of manipulations and calculations on
the shape, are useful for this.
The particle size and distribution has a pronounced effect on interparticle adhesion,
which will affect some of the bulk properties, such as bulk density, porosity, flowability
and wettability (see Section 9.2.5).
9.2.3 Particle density
The density of an individual particle is important as it will determine whether the compo-
nent will float or sink in water or any other solvent; the particle may or may not contain
air. It can be measured using a specific gravity bottle, using a fluid in which it will not
dissolve. Alternatively, it may be measured by flotation principles. However, surface
forces may start to predominate for fine powders.
In the absence of air, the particle density can be estimated from the following
equation, based on the mass fractions and densities of the food components.
P = 1/ [ (M1 lP1+ M2 / P2 + * * * + Mtl lPn ,I (9.3)
where Ml is the mass fraction of component 1, p1 is the density of component 1 and n is
the number of components. Data on mass fractions can be found from the Composition of
Solids separation processes 25 1
Foods Tables (Paul and Southgate, 1978). A simple two-component model can be used
(n = 2; water and solids) or a multicomponent system.
The density of the major components are given as (kg m-3) (Peleg, 1983):
water 1000 salt 2160
fat 900-950 citric acid 1540
protein 1400 cellulose 1270-1 6 10
sucrose 1590 starch 1500
glucose 1560
It is noteworthy that all solid components except fat are substantially more dense than
water. However the differences between protein and the various types of carbohydrates
are less marked, although minerals are much higher. In comparison air has a density of
1.27 kg m-3.
This equation is not applicable where there is a substantial volume fraction of air in
the particle. Any deviation between the experimentally determined value and the value
calculated from the above equation may mean that there is substantial air within the solid.
An estimate of the volume fraction of air (V,) can be made from
(9.4)
P= &pa + V,P~ =VaPa +(l-va)Ps
where pa = density of air; ps = density of solid (estimated using eq. (9.2)) and p = true
solid density, measured experimentally. This volume fraction (V,) of air is sometimes
known as the internal porosity.
Many other foods contain substantial amounts of air, for example mechanically worked
doughs. One solution to determine the unaerated density is to measure the dough density
at different pressures and extrapolate back to zero pressure (absolute) to obtain the
unaerated density. This methodology could then be used to determine the extent of
aeration during the mixing process.
Note that from the compositional data, the calculated particle density of an apple is
about 1064 kg m-3. Most apples float in water, indicating a density less than
1000 kg m-3, Mohsenin (1986) quotes a value of 846 kg m-3, suggesting an air content of
about 20%. One important objective of blanching is to remove as much air as possible
from fruit and vegetables prior to heat-treatment in sealed containers, to prevent exces-
sive pressure development during their thermal processing. Data on the amount of air in
fruits and vegetables are scarce in the food literature. There is evidence that this air is
quickly displaced by water during soaking.
Data on particle densities are provided by Lewis (1990), Mohsenin (1986), and Hayes
(1987). Note that if the food is frozen, the density of ice should be substituted
(916 kg m-3 at 0°C). However, not all the water is likely to be frozen, even at -30°C.
The particle density of dehydrated powders is considerably affected by the conditions
of spray drying. Increasing the solids content of the feed to the drier will result in higher
particle densities and bulk densities. High particle densities will enhance sinkability and
reconstitution properties. Differences in particle densities are exploited for several clean-
ing and separation techniques, e.g. flotation, sedimentation and air classification.
252 M. J. Lewis
9.2.4 Forces of adhesion
There will be interactions between particles, known as forces of adhesion and also be-
tween particles and the walls of containing vessels. These forces of attraction will influ-
ence how the material packs and how it will flow. Some of the mechanisms for adhesive
forces have been described as
liquid bridging by surface moisture or melted fat;
electrostatic charges;
molecular forces, such as Van der Waals and electrostatic forces;
crystalline surface energy.
Schubert (1987a) describes some of the models that have been used to quantify these
forces, and the limitations of such models.
There is some indication that interparticle adhesion increases with time, as the material
consolidates. Flowability may be time-dependent and decrease with time.
9.2.5 Bulk properties
Although the discussion so far has focused on individual particles, the behaviour of the
collective mass of particles or bulk is very important in most operations. The bulk
properties of fine powders are dependent upon geometry, size, surface characteristics,
chemical composition, moisture content and processing history. Therefore it is difficult to
put precise values on them and any cited values should be regarded as applying only to
that specific circumstance, Peleg (1983).
The term cohesive is used to describe the behaviour of powders, as they are influenced
by forces of attraction (or repulsion) between particles. For powders that are cohesive, the
ratio of the interparticle forces to the particles’ own weight is large. This ratio is also
inversely proportional to the square of the particle size, which explains why small
particles adhere to each other more strongly than large particles. Schubert (1987a) states
that the majority of food particles are non-cohesive (and thus free flowing) only when the
particle size exceeds 100 pm. Increase in moisture content makes powders more cohesive
and increases the size at which the transition from cohesive to non-cohesive takes place.
Some of the bulk properties will be considered in more detail.
9.2.6 Bulk density and porosity
The bulk density is an important property, especially for storage and transportation, rather
than separation processes. It is defined as the mass divided by the total volume occupied
by the material. This total volume includes air trapped between the particles. The volume
fraction trapped between the particles is known as the porosity (E), where
E = PS - pb/ps
(9.5)
where ps and pb are measured solid and bulk densities. Methods for determining bulk
density are described by the Society of Dairy Technology (1980) and Niro (1978). Terms
used depend upon the method of determination and include loose bulk density and com-
pacted and compressed bulk densities.
Some bulk densities of powders are given in Table 9.7. Further values are given by
Peleg (19831, Hayes (1987) and Schubert (1987a). Peleg (1983) argues that the relatively
Solids separation processes 253
Table 9.7. Bulk density of some powders
Bulk density Bulk density
Powder (kg m-3) Powder (kg m-3>
Oats 513 Milk 610
Wheat 785 Salt (granulated) 960
Flour 449 Sugar (granulated) 800
Cocoa 480 Sugar (powdered) 480
Coffee (instant) 330 Wheat flour 480
Coffee (ground and roasted) 330 Yeast (baker’s) 520
Corn starch 560 Egg (whole) 340
From the data of Milson and Kirk (1980).
From the data of Peleg (1983).
Taken from Lewis (1990) (with courtesy of Prentice Hall).
low bulk density of many food powders cannot be explained solely by geometrical
considerations. As mentioned, most food powders are known to be cohesive. Therefore
open bed structures supported by interparticle forces are very likely to occur. Such
materials are likely to have a low bulk density and high porosity. Factors that increase
cohesiveness and interparticle forces are likely to decrease the bulk density. Moisture
sorption tends to increased cohesiveness, mainly due to interparticle liquid bridges.
Anticaking agents are believed to work by reducing cohesive forces and thereby
increasing bulk density. Peleg (1983) provides data for the cohesiveness of some powders
together with the effects of some anticaking agents on the bulk density of some food
powders.
Powders can be compressed either by tapping or mechanical compression, as in
tableting. The forces involved in compression are much higher than those in tapping or
mechanical vibration.
The ratio of tapped bulk density to the loose bulk density is referred to as the Hausner
ratio. Hayes (1987) quotes the following ranges, together with some values for some
food powders:
1 .o- 1.1 free flowing
1.1-1.25 medium flowing
1.25-1.4 difficult
>1.4 very difficult.
Hayes also refers to another index, termed the ‘Novadel Tap Test’, which is related to the
percentage volume decrease on tapping. The larger the volume decrease, the poorer is the
flowability.
Peleg (1983) states that the Hausner ratio may be used for flowability index in
powders, where friction is the major obstacle to flow, but that there is no evidence that it
is useful for cohesive powders.
When powders are compressed the powder bed deforms and a number of mechanisms
are involved, including spatial rearrangement of the particles without deformation
254 M. J. Lewis
together with those brought about by fragmentation and plastic deformation of the
particles. For cohesive powders, the open structure supported by interparticle forces is
relatively easily overcome by the compressive force and there is a relatively large change
of bulk density with pressure. Non-cohesive powders show relatively little change of buik
density with pressure.
An empirical relationship of the following form is found to fit experimental data well:
p~ =a + b log s
(9.6)
where p~ is the bulk density, a and b are constants and s is the shear stress. The constant
b is defined as the compressibility. High values of b indicate a cohesive powder, whereas
low values indicate a non-cohesive powder. Some values for different powders are given
by Peleg (1983). The use of anticaking agents was found to reduce the compressibility.
9.2.7 Flowability
The flowability of powders is very important in their handling. Some indices of flow-
ability have already been discussed.
Generally flowability increases with increasing particle size and decreasing moisture
content, As well as compressibility and cohesiveness, other factors used to assess flow-
ability are as follows:
Slide angle. This is measured by placing the powder sample on a flat smooth horizon-
tal surface, which is then slowly inclined until the powder begins to move. The angle
at which movement occurs is known as the slide angle.
Angle of repose. This is useful in the design of powder handling systems. Its value
depends upon the method of determination, which is usually by forming a heap. Other
methods involve a bed rupture or a rotating drum method. Its magnitude is affected by
frictional forces and interparticle attractive forces, which become dominant in cohe-
sive powders.
According to Carr (1976), angles of up to 35' indicate free flowability; 35-45"
indicates some cohesiveness; 45-55' indicates cohesiveness or loss of free flowability ;
>55' indicates very high cohesiveness, very limited or zero flow
These parameters are empirical in nature and often the results are not applicable, when
conditions are changed (Peleg, 1977).
Peleg (1977) and Schubert (1987a) have described a more fundamental method for
looking at the flow behaviour of powders, based on the work of Jenike, described by
Leniger and Beverloo (1975). A flow cell is used, where the powder is first consolidated
to a particular bulk density and porosity (see Fig. 9.3(a)). It is then subjected to a
compressive force (N) and the shear force (S) required to cause the powder to yield and
shear is determined. These readings are converted to a normal stress (0) (N/A) and a
shear stress (7) (SIA). This procedure of determining the shear stress is repeated for a
number of different normal stress values. The information is presented on a plot of shear
stress against normal stress and gives the yield locus, for that particular porosity.
Figure 9.3(b) shows the data obtained for a non-cohesive powder, which can be
characterised by the angle of friction (a). Also in all cases a large angle of friction,
Solids separation processes 255
tN --t Tp Tp
0 0
(4 (b) (c)
(4 Tb Fc (J1 (J
Fig. 9.3. Solid characterisation: (a) Jenike flow cell; (b) normal stress against shear stress, for a
non-cohesive powder, a = angle of friction; (c) yield locus for a cohesive powder for powders
compacted to different initial porosities; porosity 1 > 3; (d) Mohrs circles, showing the
unconfined yield stress (f,) and major consolidation stress (01).
indicating high interparticle friction, does not always mean poor flowability, for example
dry sand has a high value but flows quite well.
Figure 9.3(c) shows the yield locus for a cohesive powder, at a particular porosity.
However, if the porosity of the sample is increased, the yield locus will change. There-
fore there are a family of curves at different porosities. Also the curves do not pass
through the origin. This yield locus data therefore describes the flow behaviour of
pow ders.
This data is used to determine the unconfined yield stress (f,) and the major consolida-
tion stress ((T~), by application of Mohrs circles (see Peleg, 1977; Schubert, 1987a;
Leniger and Beverloo, 1975).
The ratio of ol/fc is termed the Jenike flow function, which has also been used as an
indicator of the flowability of powders. Its values correspond to the following character-
istics:
<2 very cohesive, non-flowing
24 cohesive
4-10 easy flowing
>10 free flowing.
This more fundamental information is extremely useful for designing hoppers, bins,
pneumatic conveying systems and dispensers. Similar measurements can be made using a
more sophisticated annular flow cell, which is capable of reliable shear force determina-
tions at low normal stresses.
The hydrodynamics of powder flow are different to that for liquids. The pressure does
not increase linearly with height, rather it is almost independent. Also they can resist
appreciable shear stress and can, when compacted, form mechanically stable structures
256 M. J. Lewis
that may halt flow. Also any pressure or compaction can increase the mechanical strength
and hence the flowability.
9.3 SEPARATION OF PARTICULATES AND POWDERS
This section will be most concerned with the separation or recovery of solids from within
a solid matrix or from a particulate system. The main emphasis will be those in fine
particulate form, so the production of material in a form suitable for separations is often
crucial for the process. In this respect size reduction and milling equipment is important.
9.3.1 Size reduction
Size reduction is a very important preliminary operation for separation processes for
many cereals, legumes and other commodity crops, as well as for extraction operations,
e.g. tea and coffee, or expression processes, e.g. fruit juice expulsion or oil extraction.
Sugar is one example of a commodity that comes in a range of particle sizes, e.g.
granular, caster, and icing sugars. Some data on sieve size measurements of different
sugars is cited by Hayes (1987).
The term ‘crushing’ is applied to the reduction of coarse material down to a size of
about 3 mm, whereas ‘grinding’ is commonly used for the production of finer powdered
material.
The degree of size reduction can be characterised by the size reduction ratio (SRR),
where
(9.7)
average size of feed
average size of product
SRR =
Several stages may be required if the overall size reduction is large.
The main forces involved are compressive forces, impact forces and shear or attrition
forces. Usually there is a predominant force involved for each type of equipment, al-
though the other forces may be involved to a lesser extent. The fracture resistance in-
creases with decreasing particle size.
Aspects which need to be considered in the selection of the most appropriate equip-
ment for size reduction are the particle size range required and the hardness of the
material. Hardness can be measured in Mohs, whose scale ranges between 0 and 8.5. On
this hardness scale, most foods are either very soft (~1.5 Moh); soft (1.5 to 2.5 Moh) or
medium hard (2.5 to 4.5 Moh). More details are provided by Hayes (1987) and Christison
(1991).
Very soft materials such as dried fruit, dried plant material, meat and fish may be
processed with a Colworth stomacher down to 100 pm, or high-speed cutters, such as a
bowl choppers.
Other mills for processing grain cereals, legumes, salt, and sugar include the follow-
ing:
(1) Hummer mills. These are very much general-purpose mills. Size reduction is
mainly due to impact forces. They are widely used for peppers and other spices, sugar
and dried milk powder.
(2) Roller mills. These can be one or several sets of rollers; size reduction is by
Solids separation processes 257
compressive forces; size reduction ratio is usually below 5. These are widely used for the
milling of wheat and refining of chocolate. (Size range 10-1000 pm.)
(3) Disc attrition mills. These come in a number of designs. Simple disc mills have
two discs, one of which is stationary and the other moving. The speed is relatively slow,
with a peripheral velocity of 4-8 m s-'. The feed material enters at the centre of the discs
and the discs are profiled in such a way as to cause grinding to occur as the material falls
radially across the grinding discs. (Size range down to 100 pm.)
On the other hand, impact pulveriser mills, such as pin disc or stud mills, operate at
high rotational speeds, creating peripheral velocities up to 200 m s-*. In this case the
discs contain pins or studs, which intermesh. In the simple design there is one stationary
and one moving disc, whereas in other designs both discs move.
These types of mill can produce very fine powders, suitable for air classification (see
Section 9.4).
Another high-speed mill is the high-speed rotor mill, which is a variant of the hammer
mill. A rotor with a series of hardened blades rotates at speeds in excess of 15 000 r.p.m.
and the fines pass through a sieve ring, fitted round the circumference.
(4) Bull mills. This is a tumbling mill and is used for very fine grinding processes. It
comprises a horizontal slow-speed rotating cylinder which contains steel balls of flint
stones; the balls are normally 25-150 mm in diameter. The mechanism is by impact and
shear. The optimum speed of rotation is about 75% of the critical speed, which is defined
as the speed which causes the steel balls to centrifuge.
Two or more mill types may be required to achieve the desired level of size reduction.
The size rediction achieved often depends on whether the discharge product is
released immediately or whether it is restricted by use of a screen. In the latter case the
residence time within the action zone is increased until the particle is smaller than that of
the screen. A third alternative is to allow all the particles to leave unrestricted and to
separate them externally, recycling oversize particles for further milling.
The particle size required also affects the cost of milling and the energy requirement:
the latter is based on the following equation:
dE - K,
dD D"
--
-
where dE is the energy required to produce a small change in diameter dD and K, is a
characteristic of the material. The three main equations result from different values of n.
(Note: n is a power-law exponent.)
n = 1: E = K,, ln[Dl/D2]; Kick's law
Bond's law
n = 7: 3 E = 2K,[L--&]
D20.5
n=2: E = Kn,[&-+] Rittinger's law
258 M. J. Lewis
where D2 is the final diameter and D1 is the initial diameter. Energy requirements are
well in excess of those required to produce a new surface, as much energy is lost in
friction and other inefficiencies. Care must be taken to ensure that the sample does not
get too hot during size reduction.
Wet milling can be achieved by wetting the material and the feedstock is ground in a
suspension in the liquid, which is often water. Energy requirements are usually slightly
higher than for dry milling but a finer powder is obtained and dust problems are elimi-
nated. The wear and tear on the mill is also higher. Often wet milling is useful as part of
an extraction process, whereby soluble components are transferred from the solid to the
liquid phase. Wet milling is popular for corn milling.
More information on size reduction is provided by Brennan et al. (1990), Loncin and
Merson (1979), and Christison (1991). One very pertinent comment about milling is that
the weakness of a material may be at the juncture of different components, thereby
initiating a crude form of fractionation process, which can be further exploited.
9.3.2 Sieving
Sieving is probably the easiest and most popular method for size analysis and separation
of the components within powders. A sieve is an open container which has uniform
square openings in the base.
The screen aperture is defined as the space between the individual wires of a wire
mesh screen, and the mesh number is the number of wires per linear inch. However, this
nomenclature has now been deleted from the latest British Standards although it may still
be regularly encountered.
A whole range of standard sieves are available up to 25 mm in size. Sieves for
powders and agglomerates may be from a few millimetres, down to about 20 pm. Sieves
in common use are produced to a number of standards, e.g. BS 410,1969 (see Table 9.8),
IS0 R 565 and ASTM Ell 81. In most cases the screen interval, which is the ratio of
successive sizes in a test series, is either 2, 2°.5 or 2°.25.
The high mesh sieves may be too fragile for some applications. A more robust set
consists of the Institute of Mining (IMM) screens, where the thickness of the wire is
approximately the same as the aperture size. The Tyler series is another used in America.
Coulson and Richardson (1978) summarise these. Hayes (1987) also gives a summary of
the most widely used mesh sizes, the smallest being about 44 pm. These sizes are slightly
higher than both the British and American standards. Some information is also provided
by Christison (1991). A number of special sieves are available for particle size measure-
ment below 50 pn. Complications arise below this size for reasons discussed later.
A single sieve separates a particulate material into two fractions. When sieving
materials of a non-spherical nature, the situation is complicated by the fact that particles
with a size close to that of the nominal aperture of the test sieve may pass through only
when presented in a favourable position. Such particles, sometimes termed ‘near-aperture
particles’, may partially block or blind the sieve aperture and reduce its effective area.
Therefore some particles less than the nominal sieve diameter will be retained by the
sieve. Particles much smaller than the sieve nominal diameter pass through fairly rapidly,
whereas those close to the sieve diameter take a much longer time and a small fraction
may never pass through. The effectiveness of a sieving process depends upon the amount
Solids separation processes 259
Table 9.8. Common mesh sizes
Mesh size Aperture (mm) Mesh size Aperture (pm)
4 4.00
44 355
6 2.80
52 315
7 2.24
60 250
8 2.00
72 224
10 1.80
85 180
12 1.40
100 160
14 1.25 120 125
16 1 .oo 150 112
18 0.90
170 90
22 0.7 1 200 71
25 0.63
240 63
30 0.50
300 56
36 0.45 350 45
400 36
of material placed on the sieve, the type of movement imparted to the sieve and the time
allowed for the process. A small charge will result in more effective sieving, but care
should be taken to ensure that a uniform sample is taken, particularly for analysis. The
criteria may well be different for analysis and separation. The sieving time can be af-
fected by the following factors (British Standards Institute 1796, 1989):
the material characteristics, e.g. fineness, particle shape, size distribution, density;
intensity of sieving;
nominal aperture size of the test sieve;
characteristics of sieving medium;
humidity of the air.
Materials can also be sieved in a liquid, the procedure being referred to as wet sieving.
Wet sieving is used for extremely fine particles, e.g. below 50 pm, or particles that
become electrically charged. One advantage is that it reduces inter-particle adhesion.
Also powders that cannot be dispersed or materials in liquid suspension should be sieved
wet in order to facilitate dispersion of the primary particles, which may cause the coarser
particles to agglomerate or may be difficult to disperse. A wide variety of dispersant
liquids is available, for example ethanol, isobutanol or benzene, for wheat flours, and
octanol for milk powder. More details are provided by Schubert (1987a).
Usually a number of sieves are clamped together, with the largest on the top. The
material is placed in the top sieve and the sieves are subjected to a vibratory mechanism.
In this way a distribution of particle sizes can be determined. This method is
recommended for particle size estimation between 100 and 1000 pm. More sophisticated
equipment available for research and development relies on vibrations caused by electro-
magnetic or sonic mechanisms (Christison, 1991).
260 M. J. Lewis
A special type of sieve is the air jet sieve, in which a reduced pressure is applied to the
underside of the sieve. A jet of air is discharged upwards from a radial slotted arm
rotating continuously under the screen and this helps prevent blinding by fluidising the
particles that are likely to cause blinding. A diagram of the air jet sieve is shown in Fig.
9.4. One separation is performed at a time. This is very useful for analyses involving
smaller particles.
Fig. 9.4. Air-jet sieve (courtesy of Hosokawa Micron Ltd).
Test sieving is camed out on a wide variety of food materials and for many different
purposes, for particle size analysis, for producing materials with a known particle size
distribution or for supplying materials of a specified size range. Problems encountered
with sieving result from sample stickiness, sieve blockage and agglomeration. Strumpf
(1986) contests that these problems increase exponentially as the sieve size decreases.
One of the main applications of sieving is in the flour industry, to separate the differ-
ent fractions of flour. Names of particular sieving processes include scalping, to separate
the break stock from the remainder of the break grind; dusting, bolting and dressing,
which involves sieving flour from the coarser particles; and grading, which is classifying
mixtures of semolina, middlings and dunst into fractions of restricted particle size range.
Semolina, middlings and dunst are terms used to describe fractions of rolled endosperm
of decreasing particle size (as determined by sieving). Definitions are provided by Kent
(1983).
Other terms sometimes used are scalping, for removing large particles, and dedusting,
for removing small particles, Brennan ef al. (1990).
9.4 AIR CLASSIFICATION
9.4.1 Introduction
Air classification is a means of using a gaseous entraining medium, which is usually air,
to separate a particulate feed material into a coarse and fine stream, on a dry basis.
According to Klumpar et al. (1986), classifier designers take advantage of the following
phenomena to achieve this objective:
Solids separation processes 261
small particles fall more slowly in air than large particles;
larger particles have a greater centrifugal force in cyclonic flow than smaller particles;
smaller particles have less inertia and can change their direction of flow easier than
large particles;
larger particles require a higher conveying velocity;
larger particles have a larger probability of colliding with a rotating blade.
Separation is based mainly upon particle size, although other particle properties, such
as shape, density, electric, magnetic and surface properties may play a part. The
procedure of winnowing or aspiration is a traditional process to separate chaff from grain
after threshing and is one of the simplest forms of air classification. The chaff is
dispersed in the wind or by using an air stream. This principle is used on vining and
combining machines for harvesting peas and grain. Another simple form of classification
involves subjecting a powder, containing a range of particle sizes, to an upward airstream
of constant and uniform velocity (Fig. 93a)). Some of the particles will become
fluidised, some will be conveyed and carried away by the air stream and others will
remain stationary. Consequently, some degree of separation is achieved. In principle, the
remaining material could then be subjected to a higher velocity, removing another
fraction, and this process could be continued. The forces involved in this simple process
are the drag forces acting on the particles due to the air stream, which counteract that due
to gravity.
An alternative system is to use a long shallow, slightly sloping tray (see Fig. 9.5(b)).
This type of separation also forms the basis of the simple zig-zag separator, illustrated in
Fig. 9.5(c) and Fig. 9.6, which can be single or multiple tube. This is used for separating
particles in the range 0.1 to 10 mm. In this case the separation is further enhanced by
tortuous passages and collision surfaces, which are particularly effective at removing the
larger and more dense particles. It can be used for dedusting operations and is also
capable of removing shells or hulls from disintegrated peanuts, cottonseed, rapeseed or
cocoa beans.
However, the major interest in air classification is that it provides a means of separat-
ing small particles which cannot be readily achieved by sieving, i.e. below 50 pm. Thus a
4
/-heed <7 Fines
q-TJ 1 -1 F 4,,$ Feed
11
Fines
F
7Farse
Coarse
(4 (b) (C)
Fig. 9.5. Simple clnssifiers: (a) for aspiration F = fan; (b) for fractionation L = large; S = small
particles; (c) zig-zag classifier.
262 M. J. Lewis
Fig. 9.6. Commercial zig-zag classifier (courtesy of Hosokawa Micron Ltd).
powder subject to air classification can be separated into two streams, one primarily
below the required particle size and the other predominantly above it. The required
particle size is referred to as the cut size and is discussed in more detail in Section 9.4.3.
Cut sizes of interest in food processing operations may range between 2 and 50 pm.
9.4.2 Commercial air classifiers
In many commercial air classifiers, the gravitational force used in the examples cited
above is supplemented by a centrifugal force, induced by subjecting the particles to
circular motion. This is essential for separating small particles and speeds up the separa-
tion process. This centrifugal force is produced by a spiral or vortex air flow pattern,
promoted by a tangential air inlet and the use of directional vanes or baffles. In addition
there may also be a rotating disc or turbine, which further accelerates both the air and
particles and increases the centrifugal force produced. This force acts toward the outside
of the chamber, and counteracts the drag force which is produced by the air moving in a
spiral direction toward the centre of the classifying chamber.
The mode of operation of a typical classifier is as follows. The inlet air is mixed with
the material to be separated. The feed particles are subjected to a centrifugal force
originating from a revolving rotor and a drag force produced by the air current, which
moves in a spiral direction toward the central shaft. The separation is based on differen-
tial mass, density and shape. The larger and more dense particles are influenced by the
mass-dependent centrifugal forces and move toward the outside of the chamber, where
they are removed by a discharge worm-screw conveyor or some other means. The
smaller, lighter particles are more subject to the frictional forces of the air current and
move with the air stream, leaving from the centre of the classifying chamber, into a
cyclone, where they are separated from the air. The relative magnitude of the two forces
can be changed by altering the rotational speed of the disc and the air velocity; changing
either will change the cut-size. Classifiers with the facility to change these factors inde-
pendently, will be capable of improved operational flexibility. Although the bulk of the
separation takes place within the classifying chamber, some preliminary removal of the
coarser particles may be achieved outside the main classifying chamber. The disc or
Solids separation processes 263
turbine can be mounted on a horizontal or vertical axis. The latter produces a centrifugal
force in the horizontal plane and favours greater throughput, but less precision in cut-size
(Fedoc, 1993). Some large capacity classifiers have several turbines mounted in one unit.
Air classifiers are categorised by reference to a number of factors, such as:
the forces acting upon the particles; e.g. the presence or absence of a rotor, the drag
force of the air and the presence of collision forces, which hinder larger particles;
the relative velocity and direction of the air and particles, controlled by their respec-
tive feed systems;
directional devices such as vanes, cones or zig-zag plates, which allow a change in
direction and provide collision surfaces;
location of the fan and fines collection device, whether they are situated internally or
externally.
Other important features are the capacity of the classifier and the energy utilisation. A
comprehensive treatment of classifier types and their operating principles has been
reviewed by Klumpar et al. (1986). For some classifiers processing coal dust and cement,
flow rates of over 100 tonnes h-' can be handled. Laboratory classifiers are available
which will handle batches of as little as 50 g of samples and which will separate
2-6 kg h-' on a continuous basis. Larger classifiers handling foods can process more than
5 tonnes h-I .
One machine which has been widely used for food processing is the Alpine Mikroplex
classifier (Fig. 9.7), which is rated up to 1.6 t h-' and a maximum energy consumption of
19 kW. This design uses a rotor on a horizontal axis. Material is fed by gravity between
the fan and the vertical distributor plate. The air flow rate and direction of flow can be
changed by vanes within the machine, and this is sufficient to change the cut size, which
is also affected by feed rate, The cut size range is 3-30pm. The coarse material is
removed by a discharge worm conveyor. It is categorised as a free vortex machine. Such
machines are now largely being replaced by forced vortex machines (Fedoc, 1993),
Coarse fraction
knife edge
Inlet for material
to be classified
Coarse fraction
discharge worm
Air guide vane
Classifying
chamber
Spiral airflow
Fig. 9.7. Alpine Microplex classifier (frontal cross-section view) (courtesy of Hosokawa
Micron Ltd).
264 M. J. Lewis
whereby the force is provided by means of a rotor or turbine, which disperses the
particles into an air stream applied by a suction fan. These types of equipment operate
under a slight vacuum. Some examples are shown in Fig. 9.8.
9.4.3 Process characterisation
In most cases, air classification work is empirical in nature because of the difficulties in
quantifying the forces acting upon a particle, with any degree of accuracy.
One method of characterising the separation is by means of the cut size. Ideally, all
particles below the cut size end up in the fines and all particles above the cut size end up
in the coarse stream (see Fig 9.9(a)). However, there will always be a small fraction of
particles smaller than the cut size in the coarse stream and an equally small proportion of
particles larger than the cut size in the fines stream. The extent of this overlap and the cut
size can be determined by measuring the particle size distribution of the coarse and fine
streams (see Section 9.2.2), and presenting the data for both streams as a weight fre-
quency distribution. The cut size is defned as that size where the weight of particles
below the cut size in the coarse fraction is the same as the weight of coarse particles
above that size in the fines stream (see Fig. 9.9(b)). The yields of fines (Yf) and coarse
(Y,) streams need to be known. If they are equal, the point of overlap (Fig. 9.9(b)) gives
the cut size. If they are not equal, which is most likely to be the case, the frequency
distribution for the fine stream must be multiplied by the yield for the fine stream, and
that for the coarse stream by the yield for the coarse stream (Fig. 9.9(c)). The cut size is
given by the point of intersection of these curves.
Factors which influence the cut size are the dimensions of the classifying chamber,
peripheral forces and the spiral gradient. The cut point can be adjusted by varying the
rotor speed, air velocity, vane setting and feeding rate.
By equating these forces when they are in equilibrium, an equation for the cut size (d)
can be derived. This is based on Stokes’ equation:
d2 = [18pu,r/pu~] (9.8)
where
p = viscosity of air
u, = radial speed of air
r = clearance of classifier wheel
p = particle density
up = peripheral speed of particle (equivalent to rotational speed).
Although it is not possible to predict the movement of a particle by Stokes’ law in an
air classifier, because the forces acting upon the particle depend upon its position in the
classifier, the equation is useful in that it predicts how the two main parameters, air flow
rate and rotational speed, may influence the cut size. It predicts that increasing the air
flow rate (u,) increases the cut size, whereas increasing the rotational speed (up) de-
creases the cut size. The cut size for most operations is in the sub-sieve size range.
Solids separation processes 265
Hosokawa
Turboplex Ultra-Fine Classifier
Fig. 9.8. Selection of air classifiers (courtesy of Hosokawa Micron Ltd).
266 M. J. Lewis
IuuAdhk (4 (b) Size (c)
Fig. 9.9. Cut size determination: (a) ideal separation; (b) real separation, weight frequency
distribution; (c) weight frequency distribution corrected for yield.
Therefore air classification provides an excellent means of separating powders, based on
cut sizes below 40 pm.
However, the cut size alone does not provide information on how sharp the separation
is. An alternative method of evaluation is to determine the grade efficiency, which also
has the advantage of indicating the sharpness of the separation.
The particle frequency distribution is determined by weight for the coarse stream
(qc(x)) and feed material (4L.x)) (see Fig. 9.9(b)).
The yield is determined for the coarse stream Yc.
The grade efficiency T(x) indicates for any particle size x, the mass fraction of feed
material appearing in the coarse fraction. Thus
ycqc (x)
T(x) = ~
4f (XI
Thus grade efficiency can be plotted against particle size (see Fig. 9.10).
The cut size is where the T(x) = 0.5, indicating the size of the particles, half of which
appear in the coarse stream and therefore by difference, half of which appear in the fine
stream.
The sharpness of the separation is measured by the ratio k = [~&x75~], i.e. the ratio of
the sizes giving grade efficiencies of 0.25 and 0.75 respectively. Ideally k = 1 .O.
The best industrial air classifiers achieve k = 0.7, but typically commercial air classifi-
ers show k values from 0.3 to 0.6 (Schubert, 1987b).
qJ7 c
Size
0
Fig. 9.10. Grade efficiency vs particle size: (a) ideal separation; (b) and (c) decreasing sharpness.
Solids separation processes 267
Another factor used to define the separation achieved is the protein shift, which is
widely used in those operations where fractionation of protein occurs. The protein shift
(S,) for a simple separation is defined as
(9.9)
(Cp - Cpo) y
s, =
CPO
where C, = proteins in fines, C,, = protein in flour and Y is the dry weight yield of fines.
Table 9.9. Schubert (1987b) has shown that protein shift is largely independent of yield.
Table 9.9. Protein shifts produced by air classification of
flours from different cereals and legumes
This can be determined by analysis of the resulting streams. Some values are shown in
Flour Processing Protein shift
(%)
Barley Roller-milled 19
Barley Pin-milled 28
Malted barley Commercial grind 8
Malted barley Pin-milled 18
Oats Pin-milled 27-32
Triticale Pin-milled 28-36
Rice Turbo-milled 8-10
Potato Pin-milled 22-25
White bean Turbo-milled 22
Field bean Pin-milled 42
Fababean Pin-milled 45
Taken from Sosulski (1983a) (with courtesy of Chapman Hall).
In an ideal separation all the protein will finish up in one stream, i.e. the fines. Under
these conditions, the ideal protein content in the fines (CPldeA, ) will be C,,/Y, provided
the yield is greater than the initial protein content.
It is possible to compare the measured separation with the ideal case by defining an
efficiency of protein enrichment (epe), where
(CP - CPO)
(CP,deai - CPO)
epe =
Values range from 0 to 1, the higher the value the more efficient is the separation process.
Eliminating (CPideal ) gives
(CP -Cp0)J
cpou - Y)
epe = (9.10)
268 M. J. Lewis
However, this is little used, compared to protein shift.
ing relationship:
Schubert (1987b) gives an alternative method of analysing the data, using the follow-
c, = [CPO/Y"] (9.11)
where m is a measure of the separation. If the protein content in the fines is plotted
against the yield (Y), a characteristic curve is obtained, which can be used to determine m.
m = 1
m = 0
applies to an ideal curve, where C, = [Cpo/Y]
no separation, C, = C,,
This approach can also be used to measure the effectiveness of the separation. Higher
values of m imply better separation.
Examples are given for protein extraction from spent grain by air classification
(m = 0.12) and wet processing (m = 0.48). The reason that wet processing is more
effective is that the protein strongly adheres to the larger husks and is not effectively
moved by air classification. The presence of a liquid breaks down these adhesive forces
and improves the quality of the separation.
9.4.4 Applications
Air classifiers are designed to grade endosperm and cotyledon particles in the subsieve
size range of 2-60 pm, into subgroups based on differential mass, density and shape. This
is below the particle size conveniently handled by sieves. With most cereals the
separation of starch and protein is based primarily on size and shape rather than density,
even though the density difference is significant. However, these subgroups may still
represent subcellular structures which differ substantially in their physical and chemical
properties.
Most of the applications have been concerned with fractionation of the components of
cereals and legumes, in particular the starch and protein fractions. The simplest process
involves a single pass through the air classifier after size reduction.
An alternative procedure is to have a double pass, where the coarse stream from the
first stage of separation is further milled and reclassified to produce a second fines stream
(Fig. 9.1 1). This results in two fines streams, which may be handled separately or recom-
bined, and one coarse stream.
Comparison of results from different workers is not straightforward, because not all
the information is always presented. The quality of the separation will be influenced by
the particle size range, cut point, yield, moisture content, feed rate, classifier type and
operational conditions (see Section 9.4.3).
9.4.5 Cereal separations
Mature endosperms of most cereals are composed of thin-walled cells which are approxi-
mately 100 to 150 pm in diameter, with the cell walls being only 3-7 ,um in diameter
(Kent, 1983). Most of the cell volume is occupied by starch in the form of granules,
which is embedded in a protein matrix. The main starch granules are spherical or lenticu-
lar in nature with a diameter 15-40 pm, although wheat, barley and oat endosperms also
Solids separation processes 269
Feed
M
-
- Fines
M
-
Coarse
Fig. 9.1 1, Double classification process: M - mill; A - air classifier; C - cyclone.
contain a low proportion by weight of small spherical granules with a diameter of
1-10 pm. The protein matrix is often referred to as interstitial or wedge protein. In cereal
flour milling it was discovered at an early stage that the finer flour contained more
protein, which led to the development in some countries of commercial procedures for
protein displacement milling of soft wheat flours, whereby finely ground flours were air
classified into a light fine fraction which has double the protein content of the original
wheat flour and a coarse starch fraction with some specific advantages in certain applica-
tions. Obviously the quality of the separation depends upon the cereal type and variety
and the method of milling.
During conventional roller milling of wheat, the endosperm particles are separated
from the bran while progressively being reduced in size to pass through sieves having
apertures of 100-150pm. The resulting flour consists of particles with a range in
diameter of 2-200 pm. Kent (1983) made some observations on the differences between
hard wheat and soft wheat. Hard wheats contain higher protein contents 12-14% and are
physically more difficult to break. Soft wheats have a lower protein content and are more
easily reduced in size. In hard wheats, the majority of the particles are shattered cell
fragments over 50 pm in diameter, whereas in soft wheats the protein matrix is partially
disintegrated to yield more of the detached starch granules and wedges of protein matrix
containing variable amounts of small starch granules and cell wall fragments. There was
also a difference in the amount of protein bound to starch between hard and soft wheat.
It is now common practice to use impact mills such as the pin disc mill on powders as
a pretreatment to air classification, as these are capable of producing significant
proportions below 20 pm. Impact mills consist of two sets of pins. In the simple design
only one set of pins rotates. Finer flours result when both sets rotate in opposite directions
at different speeds (6000-1 8 000 r.p.m.). An optimum impact velocity for disintegrating
most endosperm is about 200 m s-’, which is not high enough to damage the starch
granules extensively. Damaged starch granules would adversely affect the baking
performance and functionality of the starch fraction. Impact mills are best for soft and
brittle products because of the high wear and tear. Particle size was also found to be
dependent on moisture content, feed rate, nature of the grinding surface and size of the
screen selected.
270 M. J. Lewis
Jones et al. (1959) showed that there were three principal groups of particles in pin-
milled wheat flour: large particles of either cell wall material including starch and pro-
tein, or larger detached starch granules or seed coat (>40pm); medium-sized starch
granules, some with adherent protein (15-40 pm), and small chips of free wedge protein
and detached small starch granules (<15 pm). The main mechanism of separation was
size rather than density differences.
Kent (1965) fractionated hard (13.6% protein) and soft (7.6% protein) wheats, using
17 and 35 pm as the cut-off to distinguish between the fractions. The yield of the fine
fraction (0-17 pm) was 7% for the soft wheat, but only 1% for the hard wheat and the
protein content in the fine fraction had increased much more for the soft wheat, to 14.5%,
compared to 17.1% in the hard wheat. The yield of intermediate fraction for soft wheat
(17-35 pm) was 45% and the protein fraction was reduced from 7.6 to 5.3%. The yield of
the coarse fraction (over 35 pm) was 48%, but the protein content was higher than the
original, at 8.9%. Further size reduction, using an impact mill, increased the fine fraction
yield for the soft wheats from 7 to 20% and the protein content from 14.5 to 15.7%. Thus
the amount of protein associated with the fine fraction had increased by 300%. For wheat
it can be seen that a limited fractionation of its major components can be achieved and
that the protein content can be doubled in the fine fraction for soft wheats.
As cut size increased above 19 pm, the yield of fines increased but the protein content
of the fines started to decrease (Jones, 1960; Sullivan et al., 1960). Repeated classifica-
tion, performed by remilling the coarse fraction and reclassifying it at the same cut point
(four times) produced protein shifts ranging between 20 and 34%, and (eight times)
ranging between 30 and 60%. These improvements were more marked for soft wheat
than for hard wheat (Stringfellow et al., 1963-64).
One problem could be starch damage. Some of the protein was found to be bound to
the starch. Sosulski (1983) summarises the use of air classification for cereals. Since then,
more attention has been devoted to hard wheats. For these, attrition milling has been
found to be more effective than pin disc milling, and protein enriched fractions have been
used as gluten replacers (Sosulski et al., 1988a). Further work on hard wheat classifica-
tion and baking properties of the resulting fractions has been reported by Nowakowski et
al. (1987).
Vose (1978) provides data on some protein shifts for barley, malted barley and oats
(see Table 9.9). Protein shifts were higher for flours produced by pin-milling.
9.4.6 Legumes
Sosulski (1983a) has reviewed the use of air classification for fractionating proteins in
some legumes. He produced data which gave a measure of the efficiency of the process,
in comparison with those for some cereals. Comparison was made using the degree of
protein shift, which is based on the yields of coarse/fine fractions and their protein
contents relative to the parent flour (see eq. (9.9)). Some values are given in Table 9.9.
High protein shifts are indicative of good separations. The results confirm that pin-
milling as a pretreatment improves the quality of the separation and roller-milling and
turbo-milling are less effective because they fail to release much protein. Results with air
classification of potato protein were not very promising, but the protein shifts of greater
than 40% for legumes were very encouraging. Unlike most cereals, the starch granules of
Solids separation processes 27 1
legumes do not exhibit a bimodal distribution or variable distribution and the average
dimensions are 16-21 pm in width and 23-28 pm in length (Biliaderis et al., 1981). Pin-
milling has no effect on the starch granules but reduces other cell structures to below
5 pm.
Trials with field peas, which contain 1627% protein, have proved successful and
Sosulski (1983a) reported results for separation of field peas containing 21% protein, into
25% of fines, containing 60% protein. The coarse fraction was pin-milled a second time
to release more protein and reclassified, giving an additional 10% of fines, containing
46% protein. The resulting final coarse fraction contained only 3% protein, giving an
overall recovery of over 90%. Further pin-milling and reclassification, in order to further
reduce the protein content of the starch fraction and improve its functionality, resulted in
marginal reduction in the protein, but considerable increase in the amount of starch
damage. The residual protein could also be removed by water washing, which gave a
product containing less than 0.06% nitrogen.
Experience has shown that it is not necessary to remove the hulls prior to pin-milling.
However, if not removed, they will finish up in the coarse fraction and modify its
composition, as they comprise a significant part of the seed weight, about 8% for field
peas and 13% for fababeans. They may cause problems if they contain anti-nutritional
factors and it is advisable to remove them for fababeans to avoid problems from con-
densed tannins.
Fababeans contain 28-32% protein, which is substantially more than field peas and
makes them an attractive proposition for air classification. They have also been success-
fully fractionated.
Work with field beans showed that protein fractions at yields of 20-30% and
containing 50-60% protein could be obtained in the fines. However, the coarse fractions
still contained substantial quantities of protein, giving a considerable reduction in the
overall recovery. Results for a number of other legumes are recorded in Table 9.10.
Tyler et al. (1984) examined the effect of cut size on the separation of several
legumes. In general an increase in the cut size resulted in an increased yield of the fines
fraction and protein recovery in the fines, but a decrease in protein concentration. How-
ever, the amount of starch in the fine stream increased, together with a decrease in starch
separation efficiency.
In fact, for legumes, air classification can be regarded as a means of separating starch,
as most of the non-starchy materials finish in the fines (Han and Khan, 1990a). For pea
flour it was found that most of the lipid fraction was found in the fine stream, but fibre
was distributed between the fractions (Wright et al., 1984). Cloutt et al. (1986)
investigated starch size distribution for cowpeas, fababeans and pigeon pea and found
considerable differences between them.
Cloutt et al. (1987) looked at the effects of cut size on the separation characteristics of
the same three legumes. For each one there was a good relationship between the fines
yield and protein content of the fines. This relationship could be established to permit
protein content to be estimated from the dry weight yield.
Han and Khan (1990b) have evaluated the effects of dry roasting on the fractionation
process and functional properties of the separated fractions and found for the protein
fraction that roasting reduced nitrogen solubility index and foaming properties.
h B
r:
y=?z
?g?a$?
Ye
85
x
!i
I
Irnrnrn
rn~rnrnrnrn
Lt:
cd
gg
8
c.l
Z??
"V?Z$Z
2
.3
0 0
cd
.s
m
a
.s
.e
Y
a,
Y !a
a
E
I
I*ww
~wwwww
VI
-0
c cd
TOW
"\ql?W"l?
h
3
0-0
0rn-lO-N
G
I
lior-00
cor-wwww
r( E CI
sz
9.2
$2
i
"1PO
"?".099"
CA%
;;$
e
c
Il0Nrn
rnornm*=
8s
Q
I
IZVZ
"OF""?
;5
rno-ooc1
.s
5
ag
?9TOo9
Zpa,,,
T-Cr?"F
m-g
6
s3zrn-e
sg
.9z
a%
22
oc
C(
TP09
09?09TTl?
&$
q
I
lggz
rnrn-*rnN
5.2
bbrnrlio0
e
2%
%
.s
"r?O\W"
W\q"\qr?? "rnQ\SGG
A2
Q
ZTSZ8
WbW
2LF: 0%
eT
G
gzggg
ZZggZG
a
.%
ea
%5
.s,"
zs
m
m
.Y
g
c
Ilr-li3f3
222222
k-g
>la
c
c1Cu~OQ'
0bCuN-r-
-00
r-w
NN
NNNNNc1
g& c
s
c
$
t'Vi"O\"
909"\q.?V!
5
.z
m
0**
0*W000
$
0
*
Z?bV)F
W09OTb09
5
2
+OQN
rnCurnQO0
0
c
3
scd
e3
s
33Ecaa,a, as
a
8s
z
2
Aa.->p3iz-- 0
3s
a
8%'
g
a
maw22
QJEALE
'3
2
0-0 2- Y oa
2G
e
a,
';j
5-
M
a,3 *-c 0
.e
Y
.e
.o
Y
brnrnbr-
NWbrn~CI
a
00
g 0
"g
.-
c
0a
.e
a,
a
rl*Nc'IN
NNc1NrnN
*
2
Q\
0)
.=
g.5$xs8&PCzPM
s - 6!
B E,
2 ;
2 g s
f .- .E 3 Q' (pI m v .- 2 2
E g
e*
-0 g
8
2
Solids separation processes 273
The functional properties of air-classified legumes have been reported to be very good.
These have been evaluated for field peas and fababeans by Sosulski and McCurdy (1987)
for the fine fraction. They reported good solubility properties and that water holding
capacity and oil absorption increased with protein content. They also showed good emul-
sification, whippability and foam stability. Han and Khan (1990b) evaluated the func-
tional properties for the fine and coarse fractions. In general, they were affected by the
heat treatment, the ratio of protein to starch in the fractions and the presence of other
components such as lipids. Aguilera et al. (1984) considered that there was
potential for using both the protein-rich and starch-rich fractions in extruded products.
Sosulski et al. (1988b) contrasted the relative efficiencies of wet and dry processing
methods; dry processing resulted in 75-85% recovery of protein for field beans and 93-
98% recovery of starch. The protein concentrates from dry processing had higher
whippability and foam stability, but lower water hydration and oil absorption capacity.
For cowpeas, protein fractions produced by dry processing generally produced
superior products to those obtained from wet processing (Ningsanond and Ooraikul,
1989).
The binding of phytic acid to protein was found not to be affected by dehulling or air
classification (Carnovale et al., 1988). Tyler and Punchack (1984) found that milling and
air classification was not influenced by the state of maturity, when peas were harvested at
different dates. Sosulski et al. (1987) reported that dehulling and air classification
improved the storage properties of cowpea products.
9.4.7 Other applications
Air classification and sieving was found useful in the preparation of oat-bran, which is
rich in beta-glucan (Wood et al., 1989). It has also been used to remove gossypol from
cottonseed protein (Quang et al., 1988).
Potato granules produced by spray drying from a wet milling process have been
successfully classified. Granules containing 10% protein were converted to a fine stream
containing 38% protein (Fedoc, 1993).
Air classification of spent brewers grain is described by Schubert (1987b). The
recovery of protein was poor because of the strong forces of attraction between the
protein and the husks.
For rapeseed extracts, a shift of 11.5-17.2% protein in the fines was found. However,
phytic acid and glucosinolates were also concentrated. (King and Deitz (1987)). It was
also found that steaming, crushing and air classification was an effective method for
removing fibre.
Ground amaranth has also been air classified, for addition to breads (Sanchez-
Marroquin et al., 1985).
9.5 WET SEPARATION PROCESSES
As well as dry separation, wet separation techniques are available, most of which are
dependent upon differential solubilities and precipitation methods. Of special interest is
the recovery of protein from a solid matrix.
274 M. J. Lewis
9.5.1 Protein recovery
Food composition tables simply give the protein content of foods, so we may be forgiven
for assuming that protein is a single entity. In fact most foods contain many protein
fractions; therefore there are many objectives in recovering them, the main ones being as
follows:
(1)
(2)
(3)
(4)
to recover all the protein from foods or their by-products to improve functional
properties and reduce waste;
to separate proteins from toxic components within the food;
to recover specific biologically active proteins, such as enzymes, insulin and hor-
mones;
to fractionate proteins; for example albumins are soluble in water and globulins in
salt solutions.
Deutscher (1990) provides a thorough review of protein purification methods.
In terms of selling price, there may be up to seven orders of magnitude difference
between some of the proteins currently available.
The solubility of a protein in solution depends primarily upon the properties of its
exposed surface groups, the type of solvent, its temperature, pH and polarity level, Le.
dielectric constant, and the type and concentration of dissolved ions. Water is the sim-
plest extractant, and its pH is adjusted to be well away from the isoelectric point of the
protein, which is in the region of pH 4 to 5 for many proteins. Dilute solutions of neutral
salts are also used for salting-in, as these are thought to interact with surface charged
groups, thereby improving solubility. Once in solution, it may then be required to aggre-
gate and precipitate the protein. Methods available are:
lowering the temperature to reduce protein solubility;
adjustment of pH to the isoelectric point;
addition of non-polar solvents to reduce the attraction of surface polar groups with
water, to encourage hydrogen bonding between surface polar groups;
unfolding (denaturation) and hydrophobic interactions;
addition of large quantities of very polar solvents, which also causes unfolding, hydro-
gen bonding between surface polar groups and hydrophobic interactions;
increasing the levels of salts (salting out), whose ions bind more readily to water and
allow hydrogen bonding between exposed polar groups;
raising the temperature to cause thermal denaturation to take place.
The principles are reviewed by Brocklebank (1987). Some specific examples will be
taken from legume and cereal processing.
9.5.2 Soya processing
The terminology used for the different grades of proteins includes flours, grits,
concentrates and isolates. Concentrates and isolates are protein enriched, where
concentrate applies to products with greater than 70% protein on a dry weight basis
(dwb), whereas isolates applies to greater than 90%.
Solids separation processes 275
Soyabeans contain about 40% protein and 20% fat (dry weight basis). The beans are
cracked, dehulled and flaked prior to oil removal by solvent extraction; the defatted
flakes are toasted to remove solvent and inactivate the antinutritional compounds.
Undesirable changes may also take place, such as colour development (darkening) and
protein denaturation. These are reduced by minimising retention times and operating at
low temperatures. Desolventisation is by direct steam heating or by passing superheated
solvent vapour through the flakes, which evaporates the bulk of the remaining solvent
from the flakes. Protein flakes contain about 50% protein. Flakes are further milled to
produce grits or flour, the only difference being the final particle size. Concentrates (70%
protein) are produced by one of three methods, which involve the washing out of non-
protein material:
extraction with alcohol (60-80%);
extraction with water acidified to pH 4.5 to minimise protein loss;
water extraction with highly toasted flakes to minimise protein extraction.
In all cases the aim is to maximise the extraction of sugars and other soluble
components and minimise the loss of protein.
Isolates are produced by aqueous extraction of proteins at elevated pH, followed by
isoelectric precipitation using acid, producing a protein curd, which after washing con-
tains over 90% protein. Greater than 90% protein is extracted at pH 8. The pH of
minimum solubility is 4.2-4.6, where about 10% of the protein is soluble and not
recovered. Functional properties such as solubility, whipping ability, emulsification
capacity and gelation can be modified by various chemical, enzymatic and thermal treat-
ments, before drying. The proteins in soyabeans, and many other legumes and
cereals are characterised by a variety of means, for example their solubilities in different
solvents (see Section 9.4.5), or according to their molecular weight, determined by
ultracentrifugation or SDS electrophoresis (Pearson, 1983; Deutscher, 1990). Note that
isolates can be produced by other methods, such as ultrafiltration and diafiltration
(Chapter 4), or ion exchange (Chapter 6).
The principle of minimum solubility has also been applied in the procedure for the
production of spun soya fibres, which involves solubilisation at low pH, followed by
extrusion into an acid bath. The fibres produced are further stretched in a heated bath.
Texturisation of soyabeans can also be achieved by thermoplastic extrusion, by use of
high temperatures and pressures. At the end of the extruder barrel the pressure is released
and expansion occurs.
Similar principles of washing out or solubilisation and isoelectric precipitation can be
applied to the production of concentrates and isolates from rapeseed, cottonseed, lupins
and other protein sources (Hudson, 1983). Sosulski (1983b) noted that rapeseed meal
gave lower extraction rates, a poorer recovery at low pH and a darker product, compared
to soyameal, with only about a 50% overall yield of protein. Glucosinolates and some of
the other toxic components can be removed by 80% ethanol. Some lupin seed varieties
have the highest protein contents amongst the legumes (up to 45%), but are extremely
variable in their composition. They can also be high in alkaloids, which are bitter and
need removing. In some cases the fat content may be greater than 15% (Cerletti, 1983).
276 M. J. Lewis
There has also been considerable work done on extracting protein from leafy materials
(Humphries, 1982).
9.5.3 Wheat protein
An interesting example for cereals is the separation of the protein fractions in wheat.
Traditionally, wheat proteins have been classified according to their extractability in
various solvents. Schofield and Booth (1983) describe five fractions which can be
produced by sequential fractionation:
Albumins (1) and globulins (2) are extracted in dilute saline; on dialysis with water the
albumins remain soluble and the globulins precipitate.
Gliadins (3) are soluble in concentrated aqueous alcohol.
Glutenins (4) are extracted in dilute aqueous acid or alkali.
A residual fraction (5) is not extracted under any of these conditions.
All fractions are heterogeneous, with overlap between the samples. The main fractions
are the glutenins and gliadins, which each constitute about 35% of the total protein. Both
these fractions are insoluble in water and a crude preparation can be produced by washing
in water or saline.
The protein ‘fraction’ of greatest technological significance is gluten, produced by
washing a flour dough with excess water. From the above discussion it would be
expected to be heterogeneous in nature, comprising mainly glutenins, gliadins and
residual protein, with small amounts of albumins and globulins. Although characterised
as a protein fraction, it may also contain up to 20% of other components, such as starch,
lipid and hemicelluloses, depending upon the production conditions. The gliadins confer
extensibility while the glutenins and residue protein confer elasticity. More detailed
analysis of these fractions is provided by Schofield and Booth (1983).
There is considerable demand for gluten products and production has expanded
regularly as new uses for gluten and starch develop. Techniques involved in producing
gluten are mostly based on extraction processes, from either a dough or a batter made
from flour, rather than whole grain. In these processes, the gluten network is allowed to
develop during the extraction process.
Requirements for raw materials are that the protein content should be high and the
flour should be of consistent quality to ensure a uniform end-product. The other
important raw material is water; with soft water the gluten is soft and slimy and the starch
is removed less easily. Therefore hard water is preferred, perhaps suggesting a role for
calcium in protein stability. Extraction from dough or batter uses considerable water and
several alternative procedures have been investigated to reduce water utilisation.
Some processes rely on separating some of the starch before the gluten network fully
develops. Centrifugation of flour slurry, not fully hydrated, produces (based on density
differences) a starch-rich fraction and a second fraction containing most of the protein,
with a protein content of 2040%. This can be spray dried to produce a protein enriched
fraction, but with a much lower protein content than normal gluten. Alternatively, the
protein rich fraction can be further sheared and milled, which allows the gluten to
develop, and this is removed by screens and further washed. Wet milling of whole wheat
Solids separation processes 277
has also been practised. Wheat grains are macerated with water, the bran screened-off
and starch and gluten separated.
Other proposed methods involve chemical dispersion, for example in weak acids such
as acetic acid, or in dilute ammonium hydroxide solutions. Non-aqueous separations have
been proposed, using fluids with different densities, such as fluorinated hydrocarbons.
Drying of gluten is important as gluten deteriorates rapidly if kept wet. Heat damage
should be avoided to ensure good quality; quality is based on its protein content, physical
characteristics and end-use performance.
9.5.4 Other applications
These examples illustrate some of the techniques that have been investigated to concen-
trate or recover protein from cereals and legumes. Similar approaches can be used for
animal protein. Mackie (1983) gives a comprehensive review on the recovery of fish
protein from a wide range of raw materials. Early fish protein concentrates were
produced by solvent extraction to remove fat and water, followed by air drying to remove
residual solvent. Bones may or may not be removed, leading to higher ash contents if
they were not. One of their major problems was almost a complete lack of any functional
properties. A similar procedure to that for soya, for the production of fish protein fibre
from alkali extracts of fish processing wastes is also described. The handling of by-
products from meat, fish and poultry are considered in more detail by Ockerman and
Hansen (1 988). Topics covered include mechanical deboning, rendering for fat extraction
and waste meal production, and the extraction of gelatine and other food and non-food
materials.
Another basis for separation is the ability of the water or solvent to reduce interparticle
adhesion. Schubert (1987b) reports that interparticle adhesion is about one order of
magnitude less in liquids than in gases. Therefore for strongly cohesive materials, there is
potential for improving separations by dispersing the material in a liquid, followed by
sieving or separating by sedimentation under those conditions. However, water costs and
dewatering costs may be high. One example investigated was the extraction of protein
from spent brewer’s grain. This material contains about 20% dry matter, of which up to
28% is protein which has been heat denatured. This tends to adhere strongly to the much
larger husks which contain little protein.
Wet processing using water to reduce interparticle adhesion was evaluated. Additional
water was added and the mixture subjected to moderate shear and then separated. The
protein enrichment was about 65%. Water utilisation was high. In an alternative process a
specially designed screw press was used to separate the husks from the protein fraction.
The relationship between initial protein content, final protein content and yield could
be described by the characteristic equation (see eq. (9.1 1))
c, = [cpO/ytn]
where m = 0.48. In both cases the degree of protein enrichment was much higher than for
dry processing, whereby the whole mass was dried, milled and subject to air classifica-
tion; the protein enrichment was between 30 and 40% and m = 0.1.
278 M. J. Lewis
9.6 SOME MISCELLANEOUS SOLIDS SEPARATIONS
9.6.1 Dehulling
In many countries of the world, legumes are initially processed by removing the seed coat
or hull and splitting the seed into its dicotyledenous components. Removal of the hull
brings about some of the following advantages: a reduction in fibre and tannin content,
and improvements in appearance, cooking quality, texture, palatability and digestibility.
Removing the hulls from many legumes is a tedious task. Legumes are often soaked
and dehulled manually and redried. This method is probably the only one which removes
all the hulls and consequently is used for estimating the hull content of seeds and also the
theoretical yield of dehulled product, which usually range between 85 and 95%. Some
values for hull content, expressed on a dry weight basis, are given in Table 9.1 1 (Reichert
et al., 1984).
Table 9.11. Dehulling performance of some different legumes
Hull Dehulling Hull Intact
content % Yield efficiency (DE) adhesion seeds %
(a) (b) (c) (d) (e)
soya bean 8.27 88.7 0.72 1 91.3
fababean 11.92 83.3 0.71 1 59.9
field pea 7.74 87.3 0.61 1 47.6
lentil 8.47 85.4 0.54 1 98.2
mung bean 8.95 74.2 0.33 2 18.1
cowpea (black-eyed) 5.24 79.6 0.25 2 18.7
cowpea (brown) 3.24 78.3 0.11 2 19.3
(a) Dry weight basis.
(b)
(c)
(d)
(e)
Adapted from tables in Reichert et nl. (1984).
kidney bean 8.47 84.2 0.51 2 0
The yield of dehulled grain when 90% of the hull has been removed from the seed.
DE = hull removed (g/ 100 g seed)/( 100 - yield) (g/100 g seed).
1 designates loose adhesion between hull and cotyledon, whereas 2 denotes a tight binding.
The weight percentage of seeds which have cotyledons bound together after dehulling.
For most commercial dehulling applications, abrasion or attrition mills are used, with
attrition mills, e.g. plate mills, being favoured where the hull is less firmly attached to the
seed coat. One problem with abrasion mills is that the yields are much lower than the
theoretical yields and losses are higher because cotyledon material is lost with the hulls,
sometimes as high as 30%. Soaking methods and residual mechanical hull removal are
methods still widely used to evaluate the efficiency of hull removal. One problem with
dehulling comes from the different size and shapes of the legumes, with dehullers de-
signed specifically for each crop. There is some interest in a universal dehuller.
Reichert et al. (1984) also examined the dehulling performance of a multipurpose disc
attrition mill with a variety of legumes. Their characteristics are compared in terms of
yields and dehulling performance (see Table 9.1 1 footnotes for definitions).
Solids separation processes 279
Although there was little difference in the yields, which was the weight of dehulled
product recovered when 90% of the hulls were removed, the dehulling performance,
which gives the proportion of hulls in the abraded fines, gives a better indication of
performance (see definition). Values ranged from 0.72 for soya bean down to 0.11 for
brown cowpeas. Thus for soyabeans the abraded fines contains 72% hull materials,
whereas for brown cowpeas it was only 11 %, indicating big losses of cotyledon materials
in the hulls.
Statistical analysis showed that the factors most responsible for differences in
dehulling performance were seed hardness and resistance to splitting.
9.6.2 Peeling
Peeling is an important process for many processed and convenience fruit and vegetables.
Mechanisms involved in peeling are abrasion, chemical cleaning, including caustic (lye)
or brine and thermal peeling. Often more than one mechanism is involved and often spray
washing is required to remove any loosely attached peel. All peeling operations generate
solid waste and may cause damage to the material.
Abrasive peeling. The food is fed into a rotating bowl, which is lined with an abrasive
material, such as carborundum. Rollers can also be used. The abrasion rubs off the skin,
which is removed by water. Claimed advantages are (Fellows, 1990) low energy costs,
minimal thermal damage, low capital costs. Drawbacks include higher product losses (up
to 25%), production of large volumes of dilute wastes and relatively low throughputs.
Some irregular shaped materials, for example potatoes with eyes, may need some manual
inspection and finishing. Onion skins are easily removed by abrasion peeling. Knives
may be used for citrus fruits.
Chemical peeling. A dilute solution of sodium hydroxide (1 to 2%) is heated to 100-
120°C and contacted with the food for a short time period. Water sprays are then used to
dislodge the skin. This was once popular for root vegetables, but it can cause some
discoloration. It has now been largely replaced by steam peeling. The use of a more
concentrated lye solution (10%) is known as dry caustic cleaning (Fellows, 1990) and
reduces water consumption and produces a more concentrated waste for disposal. Brine
solutions are also sometimes used.
Thermal peeling. The food is fed in batches into a pressure vessel, which rotates
slowly. High-pressure steam is fed into the vessel and rapid heating occurs at the surface,
within 15-30 s, but not in the bulk, due to the low thermal conductivity of the food,
thereby minimising chemical reactions, including cooking, in the bulk of the food. The
pressure is suddenly released, causing boiling of the liquid under the skin and flashing-off
of the skin, which is removed with the condensed steam. Additional water sprays may be
required. This method is increasing in popularity; it produces good quality products, with
little damage, at high throughputs. There is minimum water utilisation and minimum
losses. Flame peeling, using temperatures of 1000°C, has been used for onions.
9.6.3 Cleaning of raw materials
Contaminants on food raw materials can be of various origins:
mineral - soil, stones, sand, metal, oil;
plant - twigs, leaves, husks, skins;
280 M. J. Lewis
animal - faeces, hair, insects, eggs;
chemical - pesticides, fertilisers, other contaminants;
microbial - yeasts, moulds, bacteria and metabolic by-products, e.g. mycotoxins, e.g.
patulin in apples.
One of the first preliminary operations must be to remove these. Important considera-
tions are high efficiency of removal, combined with minimising loss and damage and
further recontamination of components. A combination of methods is used, including dry
cleaning and wet cleaning. Aspiration to remove dust and light contaminants has already
been described.
Screening is widely used for removing contaminants considerably different in size to
the food being treated. Sieves are available up to an aperture size of 25 mm and screens
for larger sizes (see Section 9.3.2). Also disc separators, where the shape of the disc
matches the shape of the food, can be used for separating seeds from grain.
It is important to be able to detect the presence of metal fragments and remove them
from the raw material, to prevent damage to the food processing equipment and contami-
nation of the final product. Metal detectors may also be incorporated at the end of the
packaging line. Magnetic materials can be removed by powerful magnets, which can be
permanent magnets or electromagnets. Non-ferrous metals, such as aluminium, are
detected by passing the material through a strong electromagnetic field. This field is
distorted and initiates a warning signal. X-rays have also been used for products in sealed
containers. It has been reported that modem cocoa processing leads to iron contamination
levels of 200 mg/kg in cocoa mass and greater than 300 mg/kg in cocoa powder; this
arises from hammer mills, impact mills and the agitator blades of rotating ball mills.
Between 5 and 15% of the metal was greater than 75 pm. Improvement in design of this
equipment was considered to be the best way of reducing this (List and Thiede, 1987).
Krishnan and Berlage (1984) looked at the principle of separating walnuts by a magnetic
field. Iron dust with gelatin or a magnetic solution was added to the whole nut. This was
cracked and the shell separated from the meat using a permanent magnet.
Electrostatic methods for cleaning materials are available, which take advantage of the
differences in electrostatic charge of materials under controlled humidity conditions. The
solid is fed from a hopper onto a drum, rotating at 70-350 r.p.m., which is either charged
to a potential of 5-20 kV or earthed and the oppositely charged particles are separated as
they are more strongly attracted to the drum. They are removed from the drum by a
scraper. This method can be used to remove dust and stalk from tea fannings and also
some unwanted seeds from cereals and oilseeds (Brennan et al., 1990). The Dodder mill
uses a roller coated with a velvet-type material, which will attract particles, such as seeds,
and remove them from cereals, due to differences in their surface properties.
Wet methods are also widely used for cleaning purposes. Heavily soiled vegetables
can be simply presoaked in water; the process helped by agitation. A more efficient
process is spray washing, which uses high-pressure sprays and requires smaller volumes
of water. The principles of flotation are used for cleaning vegetables; heavy particles such
as soil, metal or glass sink, the vegetables are neutrally buoyant and straw and grass float.
Screens may also be incorporated to remove oversize material. In wet-processing
Solids separation processes 28 1
applications, the microbial quality of water supply and the additional costs of the water
and effluent treatment need to be considered.
Peeling or dehulling, as well as removing the outer layers, will also remove any
disorders associated with them.
9.6.4 Sorting and grading
Sorting and grading are important preliminary operations. Sorting is normally reserved
for processes which separate foods into categories based on a single physical property,
such as size, shape, weight or colour. Grading, on the other hand, is a quality separation
and a number of factors may have to be assessed. Some examples are colour, absence of
blemishes, flavour and texture. Food grading is usually done manually, by trained
experts, because it is not usually possible to link quality with one physical property.
Some examples are meat grading and inspection, fish grading, horticultural products,
tea and cheese. However, the food analyst is always seeking for instrumental techniques
for assessment of these sensory attributes which contribute towards the character of the
food. For example the dielectric properties of fish have been found to change as the fish
becomes less fresh. Consequently a wide range of instrumental methods has been
evaluated for measurement both on-line and in the laboratory, of properties of foods that
correlate with the sensory characteristics of appearance, colour, flavour and texture.
Appropriate instrumentation and sensors have been reviewed by Kress-Rogers (1993).
Equipment for size sorting based on rollers and screens, which provide either a fixed
or a variable size aperture, are discussed in more detail by Fellows (1990) and Brennan et
al. (1990). Sorting by weight is important for high value products such as eggs, and some
tropical fruits. Image analysis is being increasingly investigated in this respect.
Colour sorting and grading
Foods can be sorted on the basis of their colour, for example removing discoloured baked
beans, prior to them being blanched. One of the most common applications is to pick out
miscoloured pieces and the simplest method is by manual inspection, as the food passes
by the inspectors, on conveyor belts.
Colour sorters have been available for over 40 years and one widely used application
range is for particles in the range 2-10 mm. Some examples are: rice, baked beans and
other legumes, peanuts and roasted coffee beans. Throughputs range from about
10 kg h-' up to 10 000 kg h-', with many applications between 100 and 1000 kg h-'.
These are based on a sensor located above the conveyor. The feed is divided into lanes
or channels. Mohsenin (1984) summarises the sequence of operations as singularisation
into discrete units, acceleration to present a substantial number of units to the system per
unit time, presentation of each unit before the sensors, evaluation and comparison to
some predetermined standard, and segregation to separate each unit according to its
colour or other specified standard. Figure 9.12 illustrates the layout of such a sorter.
Colour sorting relies on the optical properties and reflectance of the samples. The
principle is that the light source is directed on the material and the reflected light is
measured by a photodetector and compared with preset standards. Materials outside the
range are rejected. The incident light and reflected light may or may not be filtered, to
282 M. J. Lewis
Chute
dust
n
t
product
Reject
product
Fig. 9.12. Diagram of colour sorter (with courtesy of Sortex Ltd).
allow only selected wavelengths to reach the detector. Monochromatic sorting uses only
one selected waveband range and is used where there is sufficient difference between the
reflectivity of acceptable and unacceptable products within the selected waveband.
Unfortunately it is not always possible to find a single section of the spectrum where this
is the case, so it is necessary to resort to a more complicated procedure which involves
measuring at two selected wavebands. This is known as bichromatic sorting. Usually a
ratio of the signals from the two wavebands will facilitate sorting of the materials. In
some cases dual monochromatic sorting is used, where it might be important to reject
more than one type of defect. A discrepancy in either signal will cause the item to be
rejected. More detail is provided by Low and Maughan (1993). Other factors to be
considered are the natural variations in colour that are found for each product. The feed
rate is also important and there may be problems operating in a dusty or humid environ-
ment.
As in most applications, such machines are never 100% efficient in terms of either
removing all defective items or rejecting acceptable items. Improving the sensitivity of
the detection unit increases the efficiency for removing defective items, but also means
that a greater proportion of acceptable items will be rejected. There is also a
recommended flow rate range for each machine. Increasing the flow rate within this
range usually leads to a greater loss of acceptable material. In most cases, the overall
performance improves if the material has been cleaned and size graded prior to colour
sorting. Some typical examples of removal efficiencies for defective items are
Solids separation processes 283
green coffee: removing defective beans and foreign material, 90% at 900 kg h-l;
white beans: removing discoloured beans and foreign material, 97% at 1500 kg h-l;
frozen peas: removing foreign material, 99% at 10 000 kg h-'; (Low and Maughan,
1993).
Other applications are for sorting of fruit, picking out bruised, damaged or mouldy
fruit. Colour measurement is used for control purposes; for example controlling the
energy input into baking ovens to ensure a product of uniform colour from the oven.
Transmittance methods form the basis of egg inspection and have been used to distin-
guish between cherries with and without pips.
Future developments will combine colour sorting with vision analysis, whereby sort-
ing will be based on colour differences, size and shape.
REFERENCES
Aguilera, J. M., Crisafulli, E. B., Lusas, E. W., Vebersax, M. A. and Zabik, M. E. (1984)
Biliaderis, C. G., Grant, D. R. and Vose, J. R. (1981) Structural characterization of
Brennan, J. G., Butters, J. R., Cowell, N. D. and Lilly, A. E. V. (1990) Food Engineering
British Standards Institute 1796: Part 1: 1989, IS0 8156: 1988, Test Sieving.
Brocklebank, M. P. (1987) Large-scale separation and isolation of proteins. In Food
Biotechnology - 1, King, R. D. and Cheetham, P. S. J. (eds.), Applied Science
Publishers, Barking.
Carnovale, E., Lugaro, E. and Lombardi-Boccia, G. (1988) Phytic acid in faba bean and
pea; effect on protein availability, Cereal Chem., 65, 114-117.
Carr, R. L. (1976) Powder and granule properties and mechanics. In Gas Solids Handling
in the Processing Industries, Morchello, J. M. and Gomezplata, A. (eds.), Marcel
Dekker, Inc., New York.
Cerletti, P. (1983) Lupin seed proteins. In Developments in Food Proteins - 2, Hudson,
B. J. F. (ed.), Applied Science Publishers, London.
Christison ( 1991) Equipment Catalogue, Albany Road, East Gateshead Industrial Estate,
Gateshead, England.
Cloutt, P. (1983) A study of component fractionation of grain legume flours by air
classification. PhD thesis, University of Reading.
Cloutt, P., Walker, A. F. and Pike, D. J. (1986) Air classification of flours of three
legume species: effect of starch granule size distribution, J. Sci. Food Agric., 37, 173-
184.
Cloutt, P., Walker, A. F. and Pike, D. J. (1987) Air classification of flours of three
legume species: fractionation of protein, J. Sci. Food Agric., 38, 177-186.
Coulson, J. M. and Richardson, J. F. (1978) Chemical Engineering, Vol. 2, 3rd edn,
Pergamon Press, Oxford.
Deutscher, M. P. (ed.) (1990) Guide to Protein Purification, Methods in Enzymology,
Academic Press, San Diego.
Air classification and extrusion of navy bean fractions, J. Food Sci., 49, 543-546.
legume starches I1 studies on acid treated starches, Cereal Chem., 58, 502-507.
Operations, 3rd edn, Elsevier Applied Science, London.
284 M. J. Lewis
Fedoc, P. (1993) Air Classifcation, Encyclopaedia of Food Science and Technology,
Academic Press, London.
Fellows, P. J. (1996) Food Processing Technology, Principles and Practice, Woodhead
Publishing Ltd, Cambridge.
Han, J. Y. and Khan, K. (1990a) Physiochemical studies of pin-milled and air classified
edible bean fractions, Cereal Chern., 67, 384-390.
Han, J. Y. and Khan, K. (1990b) Functional properties of pin-milled and air-classified dry
edible bean fractions, Cereal Chenz., 67, 390-394.
Hayes, G. D. (1987) Food Engineering Data Book, Longman Scientific and Technical,
London.
Humphries, C. (1982) Towards leaf protein as a human food. In Developments in Food
Processing - 1, Hudson, B. J. F. (ed.), Applied Science Publishers, London, pp. 263-
288.
Hudson, B. J. F. (ed.) (1983) Developments in Food Proteins - 2, Applied Science
Publishers, London.
Jones, C. R. (1960) Developments in milling process: protein displacement, Milling, 135,
494-498.
Jones, C. R., Halton, P. and Stevens, D. J. (1959) The separation of flour into fractions of
different protein contents by means of air classification, J. Biochem. Microbiol.
Technol. Eng., 1, 77-98.
Jowitt, R., Escher, F., Hallstrom, B., Meffert, H. F. Th., Speiss, W. and Vos, G. (eds.)
(1983) Physical Properties of Foods, Applied Science, London.
Jowitt, R., Escher, F., Kent, M., McKenna, B. and Roques, M. (1987) Physical Properties
of Foods - 2, Elsevier Applied Science, London.
Kent, N. L. (1965) Effect of moisture content of wheat and flour on endosperm break-
down and protein displacement, Cereal Chem., 42, 125-139.
Kent, N. L. (1983) Technology of Cereals, 3rd edn, Pergamon Press, Oxford.
King, R. D. and Deitz, H. M. (1987) Air classification of rapeseed meal, Cereal Chem.,
Klumpar, I. V., Currier, F. N. and Ring, T. A. (1986) Air classifiers, Chenz. Eng., 93, 77-
Knipschildt, (1986) Drying of milk and milk products. In Modern Dairy Technology,
Kress-Rogers, E. (ed.) (1993) Instrumentation and sensors for the food industry,
Krishnan, P. and Berlage, A. G. (1984) Separation of shells from walnut meats, using
Leniger, H. A. and Beverloo, W. A. (1975) Food Process Engineering, Reidel, Dordrecht.
Lewis, M. J. (1996) Physical Properties of Foods and Food Processing Systems,
List, D. and Thiede, A. (1987) Contamination during cocoa processing, causes and reduc-
Loncin, M. and Merson, R. L. (1979) Food Engineering, Academic Press, New York.
64,411-413.
92.
Vol. 1, Robinson, R. K. (ed.), Elsevier Applied Science, London.
Butterworth-Heinemann, Oxford.
magnetic methods, Trans. ASAE, 27, 1990-1992.
Woodhead Publishing Ltd, Cambridge.
tion, Gordiun, 87, 199-204.
Solids separation processes 285
Low, J. M. and Maughan, W. S. (1993) Sorting by colour in the food industry. In
Instrumentation and Sensors for the Food Industry, Kress-Rogers, E. (ed.),
Butterworth, Heinemann, Oxford.
Mackie, I. M. (1983) New approaches in the use of fish proteins. In Developments in
Food Proteins - 2, Hudson, B. J. F. (ed.), Applied Science Publishers, London.
Milson, A. and Kirk, D. (1980) Principles of Design and Operation of Catering Equip-
ment, Ellis Horwood, Chichester.
Mohsenin, N. N. (1984) Electromagnetic Properties of Foods and Agricultural Products,
Gordon and Breach, New York.
Mohsenin, N. N. ( 1986) Structure, Physical Characteristics and Mechanical Properties
of Plant and Animal Materials, Gordon and Breach, London.
Ningsanond, S. and Ooraikul, B. (1989) Chemical and nutritional properties of dry and
wet milling products of red cowpeas. Canadian Inst. Food Sci. and Technol, J., 22,
147- 155.
Niro Atomiser (1978) Analytical Methods for Dry Milk Products, 4th edn, Copenhagen.
Nowakowski, D. M., Sosulski, F. W. and Reichert, R. D. (1987) Air classification of pin-
milled break and middling flours from hard spring wheat, Cereal Chern., 64, 363-370.
Ockerman, H. W. and Hansen, C. L. (1988) Animal By-product Processing, Ellis Hor-
wood, Chichester.
Paul, A. A. and Southgate, D. A. T. (1978) McCance and Widdowsons: The Composition
of Foods, HMSO, London.
Pearson, A. M. (1983) Soy proteins. In Developments in Food Proteins - 2, Hudson, B. J.
F. (ed.), Applied Science Publishers, London.
Peleg, M. (1977) Flowability of food powders and methods for its evaluation - a review,
J. Food Process Eng., 1, 303-328.
Peleg, M. (1983) Physical characteristics of food powders. In Physical Properties of
Foods, Peleg, M. and Bagley, E. B. (eds.), AVI, Westport, Connecticut.
Peleg, M. and Bagley, E. B. (eds.) (1983) Physical Properties of Foods, AVI, Westport,
Connecticut.
Purchas, D. B. and Wakeman, R. J. (1986) SolidLiquid Separation Equipment Scale-Up,
Uplands Press and Filtration Specialists Ltd, London.
Qiang, Z., Posner, E. S. and Deyoe, C. W. (1988) Production study of a low gossypol
protein product from cottonseed meal, J. Agric. Food Chem., 36, 153-155.
Reichert, R. D., Oomah, B. D. and Youngs, C. G. (1984) Factors affecting the efficiency
of abrasive-type dehulling of grain legumes investigated with a new intermediate-
sized batch dehuller, J. Food Sci., 49, 267-272.
Sanchez-Mamoquin, A., Domingo, M. V., Maya, S. and Saldana, C. (1985) Amaranth
flour blends and fractions for baking applications, J. Food Sci., 50, 789-794.
Schofield, J. D. and Booth, M. R. (1983) Wheat proteins and their technological signifi-
cance. In Developments in Food Proteins - 2, Hudson, B. J. F. (ed.), Applied Science
Publishers, London.
Schubert, H. (1987a) Food particle technology, Part I: Properties of particles and particu-
late food systems, J. Food Eng., 6, 1-32.
286 M. J. Lewis
Schubert, H. (1987b) Food particle technology, Part 11: Some specific cases, J. Food
Eng., 6, 83-102.
Shah, F. H., Shah, W. H. and Bano, S. (1984) Vegetable protein for food and feed I.
Production of high protein, low fibre detoxified mustard seed meal, Qualitas
Plantarum, Plant Foods for Human Nutrition, 34, 75-79.
Society of Dairy Technology (1980) Milk and Whey Powders, SDT, Wembley, Middle-
sex.
Sosulski, F. W. (1983a) Legume protein concentration by air classification. In Develop-
ments in Food Proteins, Hudson, B. J. F. (ed.), Applied Science, London.
Sosulski, F. W. (1983b) Rapeseed protein for food use. In Developments in Food Protein
- 2, Hudson, B. J. F. (ed.), Applied Science, London .
Sosulski, F. W. and McCurdy, A. R. (1987) Functionality of flours, protein fractions and
isolates from field peas and faba bean, J. Food Sci., 52, 1010-1014.
Sosulski, F. W., Kisser-Alemu, E. N. and Sumner, A. K. (1987) Microscopical, nutri-
tional and functionai properties of cowpea flours and protein concentrates during
storage, J. Food Sci., 52, 700-706.
Sosulski, F. W., Sosulski, K. and McCurdy, A. R. (1988a) Wet milling versus air classifi-
cation of field peas to isolate protein, starch and fibre, Acta Alimentaria Polonica, 14,
41-49.
Sosulski, F. W., Nowakowski, D. M. and Reichert, R. D. (1988b) Effects of attrition
milling on air classification properties of hard-wheat flours, StarcWStaerke, 40, 100-
104.
Stringfellow, A. C., Pfeifer, V. L. and Griffin, E. L. Jr. (1963-64) Air classification
response of flours from hard winter wheats after various premilling treatments, North
West Miller, 269, 12; 270, 12.
Strumpf, D. M. (1986) Selected particle size determination techniques, Manufacturing
Confectioner, 66, 11 1-1 14.
Sullivan, B., Engebretson, W. E. and Anderson, M. L. (1960) The relation of particle size
to certain flour characteristics, Cereal Chem., 37, 436-445.
Tschubik, I. A. and Maslow, A. M. (1973) Warmephysikalische Konstanten von
Lebensmitteln und Haljabrikaten, Fachbuckverlag.
Tyler, R. T. and Panchuk, B. D. (1984) Effect of seed maturity on the air classification of
field peas, Cereal Chem., 61, 192-193.
Tyler, R. T., Youngs, C. G. and Sosulski, F. W. (1984) Air classification of legumes; cut-
size effects, Canadian Inst. Food Sci. and Technol. J., 17, 71-78.
Vose, J. R. (1978) Separating grain components by air classification, Separation and
Purification Methods, 7, 1-29.
Wood, P. J., Weisz, J., Fedec, P. and Burrows, V. D. (1989) Large scale preparation and
properties of oat fractions enriched in ('sH3) (l'sH4) beta D glucan., Cereal Chem.,
Wright, D. J., Bunsted, M. R., Coxon, D. T., Ellis, H., DuPont, M. S. and Chan, H. W. S.
(1984) Air classification of pea flour-analytical studies, J. Sci. Food Agric., 35, 531-
542.
66,97-103.