Professional Documents
Culture Documents
http://www.microporefiltration.co.uk
Fundamentals of
Particle Technology
Richard G. Holdich
Department of Chemical Engineering,
Loughborough University,
Leicestershire,LE11 3TU,
U.K.
Whilst all due care has been exercised in the preparation of this book, the
author and publishers do not warrant that the information contained
therein is completely free from errors.
Registered names, trademarks, etc. used in this book, even when not
specifically marked as such, are not to be considered unprotected by law.
Preface
There are many industrial surveys reporting poor performance of processes involving
particles, whereas fluid only processes often reach 90 to 95% of design capacity. One
study reports only 6% of particulate processes examined reporting no major
performance problems (in R.A. Williams, CAPE-21, EUREKA project 2311). Clearly,
the presence of particles increases process uncertainties and this book is aimed at
reducing some of those uncertainties for a large engineering and scientific audience.
In many engineering disciplines Particle Technology is now a subject worthy of
study on its own, rather than a sub-section of fluid mechanics. Over the last twenty
years, or so, there have been several books authored on aspects of the subject. This is
an inevitable consequence of the subject’s breadth and the interests of those authors
within it. So, there is no lack of information for the serious student to turn to.
However, where does he, or she, start off? Fundamentals of Particle Technology was
written as a starting point for students new to the subject. The philosophy behind the
book is to provide a text that all students, hopefully, regard as accessible, both in cost
and style. Also, the nature of learning has changed significantly in recent years and
this book reflects some of these changes. There are numerous line diagrams and
illustrations included, but very few equipment pictures. Instead, web addresses are
provided for equipment manufacturers and interested students should access these
sites to see the equipment offered. This provides ready access to colour pictures of
modern machinery, rather than pictures that date all too quickly. The web sites are
those of major manufacturers that are likely to be available for many years to come
and, although the exact URL might alter, a small amount of intelligent browsing will
soon find the device, or its more modern equivalent. Likewise, the exercises at the
end of a chapter have further details of their solution via the Internet site
www.midlandit.co.uk/particletechnology. Most practical questions require a series of
calculations following from problem decomposition and the student may perform an
error at any stage; the multiple choice style is designed to take the student through
the calculations so that he, or she, cannot progress if an error is made on the way.
Thus, the student should know instantly if he, or she, has made a mistake. The
disadvantage of this style is that it provides limited training in problem
decomposition; so, the support web site has additional problems available for practice
in this.
Experience shows that journal references at the end of each chapter are rarely
consulted; so, where appropriate, references are given within the text and to named
research workers. A simple literature search will show up their work – it is not usual
for an author to publish only a single paper; hence it should be easier to look for a
name and context, rather than a specific paper. There is also a further reading section,
with recommended books that correspond to each chapter within this book.
Finally, I would like to express my gratitude to colleagues at Loughborough
University, both present and past, for helping me develop this style of teaching and
encouraging my interest of this subject.
Richard Holdich,
Loughborough, November, 2002.
Contents
1 Introduction 1
1.1 Prerequisites and objectives 2
1.2 The micron 2
1.3 Sampling 2
2 Particle characterisation 5
2.1 Particle size functions 7
2.2 Algebraic representation of size functions 9
2.3 Specific surface area per unit volume 12
2.4 Distributions: moments and conversions 13
2.5 Means of a distribution 15
2.6 Image analysis and particle shape 15
2.7 Example interpretation of distribution data 16
2.8 Summary 17
2.9 Problems 17
4 Filtration of liquids 29
4.1 Deep bed and clarifying filtration 29
4.2 Cake filtration 31
4.3 Specific resistance and dry cake mass
per unit volume filtrate 32
4.4 Compressible cake filtration 33
4.5 Filtration modes of operation 35
4.6 Membrane filtration 38
4.7 Filter media 40
4.8 Filter aids 42
4.9 Summary 42
4.10 Problems 43
5 Dilute systems 45
5.1 Weight, drag and Particle Reynolds number 45
5.2 Other forces on particles 48
5.3 Particle acceleration in streamline flow 49
5.4 Settling basin design (Camp-Hazen) 50
5.5 Laboratory tests 51
5.6 Summary 52
5.7 Problems 52
7 Fluidisation 67
7.1 Minimum fluidising velocity 68
7.2 Types of fluidisation 69
7.3 Bed design and bubbling behaviour 70
7.4 Gas flow patterns around bubble and stability 70
7.5 Davidson and Harrison model 71
7.6 Discrete element analysis 73
7.7 Summary 74
7.8 Problems 75
8 Centrifugal separation 77
8.1 Sedimenting centrifuges 78
8.2 Hydrocyclones 80
8.3 Filtering centrifuges 83
8.4 Washing and dewatering 85
8.5 Summary 88
8.6 Problems 88
9 Conveying 91
9.1 Heterogeneous flow in liquids 91
9.2 Dilute phase pneumatic conveying 92
9.3 Dense phase pneumatic conveying 94
9.4 Other conveying equipment 94
9.5 Summary 96
9.6 Problems 96
Nomenclature 164
Further reading 166
Appendix – Heywood Tables 168
Index 170
1 Introduction
Particle technology may be described as being the study of materials
dispersed within a continuous fluid. The particles may be solid, but
they can also be oil droplets in water, water droplets in air, etc. So, by
a particle we mean any dispersed material within a fluid. In many
cases deformable particles have a slightly different behaviour to rigid
ones, but the starting point for the description of deformable particles
is that of the rigid, and simpler, case. Hence, particle technology
includes the understanding of raindrops, oil emulsions, powders,
slurries, etc., and just about every industrial process uses the subject
at some stage. For example, in petrol production the catalytic
cracking of petroleum is achieved in fluidised beds of catalyst
particles (Chapter 7). An understanding of fluidisation relies upon
knowledge of particle characterisation and fluid flow through porous
media (Chapters 2 and 3). The petroleum processing is performed in
the vapour phase, not the liquid, hence the fluidised beds require
appropriate gas cleaning equipment for recycling and retention of the
catalyst particles (Chapter 14). The catalyst is stored and conveyed
into the system (Chapters 10 and 9) and, of course, due care must be
exercised over powder hazards (Chapter 15). So, even in the case of
an obviously liquid product, petroleum spirit, we encounter a
significant proportion of material covered in this book.
An even greater reliance on particle technology is provided by the
increasing trend towards high value batch processing in the chemical The Reynolds number
and pharmaceutical industries. A prime example is the production of Is a measure of the amount
a tablet. In many cases a reactant is provided in a solid form and of turbulence within a
product recovery involves nucleation and then crystallization of the system. It is numerically
the ratio of the inertial to
product. These two processes are not covered here, but the interested
viscous forces. Flow
reader is directed towards other works [J.W. Mullen, 1997,
Reynolds number for a
Crystallization, Butterworth-Heinemann, 3rd edition; R.J. Davey and fluid through a circular
J. Garside, 2001, From molecules to crystallizers, Oxford Chemistry pipe is:
Primers, No. 86]. Most of the remaining aspects of product recovery duρ
are covered in this work. The crystals may be settled, to increase the Re =
µ
slurry concentration going on to a filter, or filtering centrifuge; the
see the Nomenclature for
resulting cake will need washing free of reaction products and
definitions. Values above
unreacted feed material and mechanically dried, to minimise the 2000 are usually taken to
amount of thermal energy required to complete the drying. After indicate very significant
thermal drying (not covered here), there is likely to be a need for turbulence.
product storage, crushing and classification, solid/solid mixing,
conveying and agglomeration for the purpose of forming the tablet.
Any one of these processes may be the cause of a process bottleneck,
or throughput limitation, and the intention of this book is to provide
a sound understanding of the underlying principles behind these
operations to enable reliable operation and appropriate decisions to
be drawn.
2 Introduction
1.3 Sampling
Processing is usually performed on small particles contained in bags,
drums, 1 tonne Intermediate Bulk Containers (IBC) or trucks.
Fig. 1.1 A grab sample Analysis for particle size, and other characteristics, usually requires
only a few grams. The most convenient method of obtaining a sample
of a few grams is to take a grab sample from the bulk, often scooping
the few grams off the top of the pile, or sack, see Figure 1.1. However,
segregation during particle motion is very common; examples include
the tendency for larger particles to rise to the top in breakfast cereals
such as muesli – and finer particles to fall between the gaps to the
bottom of the packet. Hence, a simple grab sample is rarely adequate
for a reliable analysis. The absence of a truly representative sample
invalidates any further work on the particles, be it laboratory tests for
some process (i.e. unit operation) or for particle size analysis and
characterisation.
When sampling from a suspension the viscous nature of the fluid
may assist in representative sampling; e.g. particles settle slower so
that stirring is often effective in suspending the particles for an
adequate time to obtain a good sample. However, at low
Fig. 1.2 A chute riffler concentrations the hindering (Chapter 6) effect of other particles is
not strong and vigorous stirring can lead to centrifugal separation
Fundamentals of Particle Technology 3
(Chapter 8). In air, or any gaseous medium, motion can help the
segregation of particles by differences in size, density, moment of
rotational inertia, coefficient of restitution, etc. Hence, sampling of
free flowing particles should be performed according to accepted
standards (BS 3406) and using well-known techniques such as:
sample cutters, cone and quartering (25 kg solids and above), chute
riffles (50 kg solids down to 25 g), see Figure 1.2, spinning riffler (1
kg down to 25 g), see Figure 1.3. Ultimately, a scoop, or grab, sample
from the suitably reduced mass of powder will be required for use in
an instrument if, as is usual, 25 g is still too much powder to be used.
When sampling a dilute stream of powder entrained in a gas stream,
isokinetic sampling is often recommended; i.e. equal fluid velocities in
the bulk flow and within the sample probe. Isokinetic sampling is
discussed in greater detail below. Fig. 1.3 Spinning riffler with
A sample cutter is used to take a sample from a flowing stream of vibratory feeder from hopper
powder, such as on a conveyer belt. At a transfer station, from one belt
to another, the powder will be dropped from the first belt on to the
second, and the sample cutter will intermittently move into the
falling powder to remove a sample. It is important that the cutter
removes a sample across the entire width of the falling powder, or
width of the belt, because segregation during transit may result in
particles of a certain size preferentially positioned on the belt.
Cone and quartering is a technique used extensively within the
minerals industries for coarse sized material up to several centimetres
in diameter, to obtain a more tractable sample size, and is often
applied before riffling. The entire contents of the container, or bag, to
be sampled are discharged on to a clean flat surface, which could be a
swept floor. The material is then made into a cone by shovelling
material from the sides to the centre to form a cone. The cone is
usually then flattened to minimise the likelihood of pushing material
at the top in any one direction. It is common for the larger particles to
roll to the periphery of the cone and the finer ones to stay near the
centre. However, so long as the cone is made evenly there should be
an equal distribution of larger particles all around the periphery of
the cone, with no bias towards any position. Thus, quartering the
heap by dividing it into four roughly equal portions should provide
samples representative of the original material. It is usual to then take
two quarters at opposite sides of the heap and remove them from the
sample. The remaining two form the sample and the process can be
repeated to reduce further the overall sample size. A similar
technique can be applied to smaller amounts of finer particles, subject
to adequate safety – health and environment considerations, coned
and quartered on a bench, but the process is operator dependent and
it is unlikely that a sample would be as representative as one
provided by the other techniques described above.
When sampling airborne particles from flow within ducts it is
important that the flow entering the sample filter is the same as that Fig. 1.4 Isokinetic sampling
in the duct, or pipe, see Figure 1.4. This is to prevent particles from a duct or pipe
4 Introduction
This uses the property that a sphere has the smallest surface area per
unit volume of any shape. Hence, the value of sphericity will be
fractional, or unity in the case of a sphere. There are a variety of
accepted shape descriptors and some of these are provided in Table
2.1.
n 3 ( x) = Z exp[−b ln 2 ( x / x m )] (2.13)
12 Particle characterisation
x 3n 0 ( x)
n 3 ( x) = (2.19)
∫ x 3 n 0 ( x )d x
n ( x)
S v = 6 ∫ 3 dx (2.20)
x
We can now define another equivalent spherical diameter that may
be used to represent the entire size distribution: the spherical particle
diameter that has the same specific surface as that of the distribution.
Rearranging equation (2.14) provides this value
6
xSv = (2.21)
Sv
This is called the Sauter Mean Diameter, and is sometimes represented
as x(3,2) in some distribution data. It is of great importance in particle
technology because it is the most appropriate equivalent spherical
diameter to represent the size distribution in fluid flow calculations.
An example of this is given in Section 3.4 in the next chapter.
xmax
π
ρ s ∫ x 3 n 3 ( x )d x (2.25)
6 xmin
which represents the mass weighted mean particle mass for a
distribution, that is of some importance in solid/solid mixing. The
weighting is towards the mass distribution because of the term n3(x)
and it is the mean particle size of the distribution multiplied by the
volume shape factor and solid density – thus giving the mean particle
mass.
Conversion between distributions was first introduced in equation
(2.2) and its analogue (2.19). In general, we have
x r n 0 ( x) x r n 0 ( x)
n r ( x) = xmax
= (2.26)
r M r,0
∫ x n 0 ( x ) dx
xmin
Substituting equation (2.26) into (2.23) results in
k r
∫ x x n 0 ( x )d x
M k + r,0
M k, r = = (2.27)
M r ,0 M r ,0
Using equation (2.27) it is easily possible to show that
M 2 ,0
M −1,3 =
M 3, 0
which is consistent with equations (2.18) and (2.20) combined.
Another illustration of conversion between distributions is to
change a mass distribution into a number one, considering both
discrete data and continuous functions as follows. If we substitute
equation (2.25) into (2.2) and use mid-points to represent the
increment we arrive at the following equation for mass fraction
k v xi 3 ρ s f i x i 3 ni
mi = 3
=3
(2.28)
∑ k v xi ρ s f i ∑ x i ni
where ni represents the fraction by number of the distribution within
the increment considered. We must now rearrange for the number
distribution
m
ni = 3i ∑ x i 3 ni (2.29)
xi
At first sight, equation (2.29) is without solution because the number
fraction appears on both sides, but the summation will result in a
constant which is multiplied by all the incremental values of mass
fraction divided by mid-points. Hence, simply dividing the mass
fraction in the increment by the mid-point cubed, summing all these
values up and then dividing through by this summed value will
provide the number fraction for each increment, as the constant value
(i.e. the summation) will cancel. Rearranging equation (2.26) provides
the analogue equation for the continuous functions
n ( x)
n 0 ( x ) = 3 3 M 3, 0
x
Fundamentals of Particle Technology 15
2.8 Summary
When using particle size distribution data the most important
considerations to be aware of are: how the data was obtained, i.e. the
basis of the analysis by number, mass, etc., and how the data will be
used. For routine quality control, or assurance, the first principle is
not so important as simply a change in results will be enough to
signal a change in quality. However, these principles are important
for equipment design, or understanding. For design, the equivalent
spherical diameter, or distribution data, appropriate to the end use
should be used; e.g. settling diameter for sedimentation basin design,
Sauter mean for pressure drop calculations.
From a practical point, the repeated testing of samples of the same
powder in a size analysis laboratory can result in significantly
different particle size distribution data. This is a consequence of
variations in both sampling as well as size analysis. Modern
computer based electronics has resulted in size analysis equipment
providing many channels, 256 and 1024, and the use of spreadsheets
has made manipulation of the data relatively easy to accomplish. Fig 2.13 Example tabular data
Hence, the previous practice of fitting mathematical functions, or output from Malvern – the
undertaking graphical analysis of the data, in order to obtain values original is available for
viewing at:
such as the specific surface, has become less important. The level of
www.midlandit.co.uk/particl
effort required to obtain a graphical solution is rarely justified when
etechnology
the distribution data changes from one analysis to another.
The Results Statistics section
Invariably, the requirement is for a single value to be used for design
is reproduced in Table 2.3
(such as an average particle diameter), or comparison purposes. This
value can be adequately obtained from a spreadsheet analysis.
However, given the possible lack of repeatability of the size analysis,
and sampling, it may be wise to obtain a range of values for the
average diameter; as a small change in its value might significantly
influence a design variable, see Chapter 3 for an illustration of this in
the context of fluid flow.
For a more thorough modelling of particle processes, the particle
size distribution functions covered in Section 2.2 have considerable
utility, as they can be used to investigate, or model, what happens
within a process to the different sizes of particles.
2.9 Problems
1. The following size distribution was obtained for glass spheres by
microscope examination. Complete the table rows showing the
relative number of particles, the cumulative number undersize and
the relative (i.e. fractional) number per micron range.
Size range (µm): <5 5-7 7-10 10-15 15-20 20-30 30-40 40-50 >50
Number in range: 0 50 150 200 55 45 20 5 0
Relative number: 0 0.0952
Cumulative no. 0 0.0952 1.000
undersize:
Relative no. per µm: 0 0.0476
18 Particle characterisation
2.
The mass of a single
Using the above number distribution data, complete the table:
particle is:
π 3
x ρs Size range (µm): <5 5-7 7-10 10-15 15-20 20-30 30-40 40-50 >50
6
(mid point)3.frequency 10800
where x is particle
Relative mass: 0 0.0039
diameter and ρ s is solid
Cumulative mass 0 0.0039 1.000
density. Thus it is possible
undersize:
to convert a number
distribution to a relative
3.
mass distribution by
The mean particle size by number is best calculated by determining
multiplying the number of
the fractional contribution to the mean from each size range, or grade.
particles by the mass of a
single particle and
This is the product of the relative number and the mid-point size in
dividing through by the the grade. Summing all the contributions together gives the mean
total mass of the system. size. Note that this is mathematically equivalent to the more
The range, or grade, mid- conventional:
point is used for the ∑ f i xi ∑ f i
particle diameter. Note
that the volume shape where fi is the frequency (or number) of occasions that xi occurs, e.g.
factor and density cancel you may have used this formula before to calculate an average mark.
out during this operation.
The relative number can select the correct answer from the following
also be used in preference Thus the mean size by number is (µm):
to the actual number for a a: 12.4 b: 13.5 c: 25.0 d: 28.2
similar reason.
4.
The mean size by mass, calculated in a similar way, is (µm):
a: 12.4 b: 13.5 c: 25.0 d: 28.2
mi 5.
S v = 6∑ The complete distribution specific surface area per unit volume (Sv) is
xi
the total surface area divided by the total volume. It may be
where mi is the calculated by several routes, one is shown left. For this distribution Sv
relative mass - see
is (µm−1):
Section 2.3
a: 0.213 b: 0.240 c: 0.264 d: 0.444
6.
You will need the equations relating specific surface (Sv) to number
frequency and also to mass fraction (both per micron). See Section 2.3 Number distribution:
∫ x n ( x )d x
for the derivation of the last relation starting from the former. For 2
π 0
spherical particles the two equations are given in the box on the right. Sv =
π /6
∫ x n ( x )d x
3
These questions are based on an idealised cumulative size 0
distribution that can be represented as a single straight line on the Mass distribution:
upper figure given to the right. π n3 ( x )dx
Sv =
π /6 ∫ x
(i). The cumulative size distribution of a powdered material may be Relation between two
represented as a straight line on a % number undersize versus types of distribution
particle diameter (x) graph passing through the points 0% by number curves (given for a
at 1 µm and 100% by number at 101 µm. The equation for N0(x) is: number distribution):
N0(x)=... dN 0 ( x )
a: x-1 b: x/100-1 c: x/100-0.01 d: x-100 = n0 ( x )
dx
Note, work in fractional terms and NOT percentages.
The relation between the
mass distributions is
(ii). The equation for n0(x) is: n0(x)=...
similar.
a: 1 b: 1/100 c: 100 d: x/100-1
(iii). What is the shape of the n0(x) graph given by your answer to
part (ii)?
(iv). The best equation to use for calculation of specific surface is:
101 2 101 2
∫1 x (1 / 100)dx 6 ∫1 x dx
a: S v = 6 101 3
b: S v = 101
c: S v = 6 101 3
∫1 x (1 / 100)dx ∫1 xdx ∫1 x dx
(vii). If the sphericity of the material is really 0.9 the specific surface
is (µm−1):
a: 0.071 b: 0.088 c: 0.066 d: 0.053
7.
(i). The specific surface of another size distribution having the same
limits as that given in question 6 but on a mass distribution N3(x)
basis is (µm−1):
a: 0.013 b: 0.138 c: 0.277 d: 0.554
(ii). The Sauter mean diameter of the distribution in part (i) is (µm):
a: 50.0 b: 43.5 c: 21.7 d: 10.8
20 Particle characterisation
8.
The attenuation of a parallel beam of light is used to monitor
visibility on an airport approach during fog conditions. It is found,
experimentally, that the droplet size distribution of the fog may be
represented by the equations
for: 50≥ x ≥15 µm N 0 ( x ) = 0.83 + 0.043 ln( x )
15≥ x ≥4 µm N 0 ( x ) = 0.30 + 0.239 ln( x )
4≥ x ≥0.5 µm N 0 ( x ) = 0.21 + 0.304 ln( x )
where N0(x) is the cumulative fraction by number less than the
particle size in µm.
(i). Write down below the equations for n0(x)
for: 50≥ x ≥15 µm n0(x)=
15≥ x ≥4 µm n0(x)=
4≥ x ≥0.5 µm n0(x)=
(ii). The specific surface area per unit volume (Sv) of the fog is the
total surface area divided by the total volume, in number distribution
terms (assuming spherical droplets):
4 15 50
2 2 2
∫ x n 0 ( x )dx + ∫ x n0 ( x )dx + ∫ x n 0 ( x )dx
S v = 6 04.5 4
15
15
50
3 3 3
∫ x n0 ( x )dx + ∫ x n 0 ( x )dx + ∫ x n0 ( x )dx
0.5 4 15
Just considering the size grade 50≥ x ≥15 µm the area integral in the
∫ x n ( x)dx ) is (µm ):
2 2
above equation (i.e 0
3.1 Definitions
By definition, a porous medium consists of pores between some
particulate phase, contained within a vessel, or some control volume,
as illustrated in Figure 3.1. The fluid flow rate through the bed is Q
(m3 s−1) and the bed cross sectional area is A (m2). Thus the superficial
(or empty tube) velocity U0 is the total flow rate divided by the cross
sectional area. The existence of the particles within the bed will
reduce the area available for fluid flow; i.e. to preserve fluid
continuity with the entering superficial flow the fluid will have to
squeeze through a smaller area; hence the velocity within the bed (U –
interstitial velocity) will be greater than the superficial. In Particle
Technology calculations it is the volume fraction that is most
important, and not the mass fraction. The volume fraction of solids
present (i.e. volume solids in bed divided by total bed volume) is
usually referred to simply as the volume concentration, or solids
fraction, and the remaining fraction is that of the voids. The void
fraction is also called the voidage and the bed porosity. It is important
to realise that, in liquid systems, the voids are usually filled with
liquid and not to assume that the bed consists of just solids and air.
The porosity is usually an isotropic property (i.e. the same in all
directions); hence, the interstitial velocity is simply related to the
superficial velocity by the following expression, which comes from a Fig. 3.1 Illustration of
consideration of fluid continuity. fluid flow through a
porous medium and
U
U= o (3.1) consideration of the
ε volume fractions present
Clearly, the resistance to fluid flow through the porous medium is
related to the amount of particles present, or volume concentration,
but it is conventional to work in terms of bed porosity. At one
extreme, when the bed is full of solids (porosity is zero – possible exercise 3.1
with cubic particles placed carefully within the bed) the resistance is Using continuity:
infinite. At the other, when no solids are present and the porosity is i.e. Q = constant,
unity, the interstitial velocity will be the same as the superficial deduce equation (3.1)
velocity. The resistance to fluid flow gives rise to a pressure drop in
22 Fluid flow in porous media
R
(3.9)
ρU 2
Equation (3.9) is the friction factor and R is the shear stress, or drag
force per unit area, on the particle surface. A force balance at the Force balance on a pipe
particle surface can be constructed as follows. wall
surface area of particles= S (1 − ε ) LA (m2) (for pure fluid – no
v particles)
drag force = R. particle surface area (N)
and
pressure drop on fluid= ∆P (N m−2)
force by fluid= ∆PAε (N)
Equating the two forces and rearranging gives
force on wall:
ε ∆P
R= (3.10) Rπd .δx
S v (1 − ε ) L force on fluid:
Note that equation (3.10) is the analogue of that provided for pipe πd 2
flow in the box. Finally, expanding into a friction factor, equation δP
4
(3.9), together with equation (3.1) gives combine and integrate:
R ε3 ∆P 1 ∆Pd
= (3.11) RL = or
ρU 2 S v (1 − ε ) L ρU 2 4
o d ∆P
Alternatively, the pressure drop per unit length is R=
4 L
2
∆P R S v (1 − ε ) ρU o c.f. equation (3.10)
= (3.12)
L ρU 2 ε 3
where the bracketed term is the friction factor. So, given a flow rate, Laminar flow
hence superficial velocity, it is possible to calculate the Modified Carman correlation
without the turbulent
Reynolds number from equation (3.3) and the friction factor from
correction is
Figure 3.4. This can then be used in equation (3.12) to provide the
R 5
pressure drop, or gradient, under conditions of laminar or turbulent =
flow through the porous medium. ρU 2 Re1
Using equation (3.10) and
3.5 Carman and Ergun correlations (3.3) provides
The friction factor plot, with Reynolds number, for fluid flow through ε3 ∆P 1
= ...
porous media is a smoother function than that found in pipe flow. S v (1 − ε ) L ρU 2
This is due to the smooth increase in turbulences within the bed as o
flow rate increases. Thus the friction factor plot can be represented by K (1 − ε ) S v µ
... =
just one, or two, empirical curves. The Carman correlation is ρU o
generally used for solid objects forming a bed So,
R
=
5
+
0 .4
(3.13) ∆P K (1 − ε ) 2 S v 2
2 Re 0 .1 = µ U o
ρU 1 Re1 L ε3
The Ergun correlation is for hollow objects, such as packing rings
R 4.17 i.e. equation (3.7)
= + 0.29 (3.14)
ρU 2 Re1
The form of both equations is similar: with a correction term added to
the laminar flow term to account for resistance due to turbulences.
26 Fluid flow in porous media
3.7 Summary
In this chapter we have seen the importance of specific surface: it
defines the surface area that is present within a porous medium; and
it is the friction of fluid flowing over that area that causes a pressure
drop. Finer particles provide a higher surface area per unit volume
than coarser ones, therefore, a higher flow resistance. Turbulences
within the fluid inside the bed, at higher flow rates, cause an
additional flow resistance, or pressure drop. Thus, a calculated
pressure drop by a laminar flow equation, Darcy’s law or Kozeny-
Carman, will always underestimate the true pressure drop if
significant turbulence is present. Often the temptation to use Kozeny-
Fundamentals of Particle Technology 27
Carman, rather than the procedure described in Sections 3.4 and 3.5,
is too great even when the Modified Reynolds number is high. This
will lead to an under-design, or specification, for equipment such as
pumps and fans. Providing another example of the failure in design
discussed in the Forward.
Even under conditions of laminar flow, the use of the median size
to respresent the size distributed solids rather than the Sauter mean
diameter can cause significant errors, as shown in Section 3.3. Finally,
a brief consideration of particle packing arrangements has been
included, but in most processes the packing arrangement is random
and not structured. Hence, the safest procedure for the analysis of a
flow through porous media problem is to conduct experiments to
deduce characteristics such as the permeability and if it varies with
Equation summary
flow; a possibility if the finer particles become transported within the Under laminar flow:
bed or if suspended solids within the fluid deposit inside the bed.
∆P K (1 − ε ) 2 S v 2
The latter is depth filtration, which is covered in the next chapter. = µ U o
L ε3
3.8 Problems
where K is the Kozeny
1. constant At high values of
(i) A powder is contained in a vessel to form a cylindrical plug 0.8 the 'Modified Reynolds
cm in diameter and 3 cm long. The powder density is 2.5 g cm−3 and Number' (Re1>2):
2.20 grams of powder was used to form the plug. The porosity inside Uoρ
Re1 =
the plug of powder is (-): µ (1 − ε ) S v
a: 0.58 b: 0.42 c: 0.75 d: 0.25 turbulent conditions
pertain. A pressure drop
(ii). Air was drawn through the plug at a rate of 6.6 cm3 per minute. given a flow rate can still
A mercury manometer was used to measure the pressure drop be deduced, but first the
during this process: a pressure drop of 60 mm Hg was recorded. The Modified Reynolds number
specific gravity of mercury is 13.6, thus the pressure drop across the is required, then the
friction factor using say the
plug was (Pa):
Carman correlation:
a: 80 b: 800 c: 8000 d: 80000
R 5 0 .4
(iii). The superficial gas velocity in (ii) was (m s−1): = +
a: 0.0022 b: 3.65x10−5 c: 2.2x10−6 d: 0.00365 ρU 2 Re1 Re 0.1
1
A force balance on the
−5
(iv).The viscosity of the air was 1.8x10 Pa s, using the Kozeny- surface of the solids and on
the fluid gives:
Carman equation, the specific surface area per unit volume of the
RS v LA(1 − ε ) = ∆PAε
powder was (m-1):
where L is the bed height
a: 2310 b: 3.0x1011 c: 5.5x105 d: 1.2x106
or depth. Given a value for
the shear stress on the
(v). The Sauter mean diameter of the powder was (µm): solids (R) calculated from
a: 2600 b: 22 c: 11 d: 5.0 the Carman correlation
then the pressure drop (∆P)
(vi). The air density was 1.2 kg m−3, the Modified Reynolds Number can be calculated from the
of the system was (-): force balance.
a: 0.10 b: 8.4x10−10 c: 4.6x10−4 d: 0.00021
28 Fluid flow in porous media
2.
(i). A cylindrical ion exchange bed composed of spherical particles 2
mm in diameter packed at a bed voidage of 0.45 is to be used to de-
ionise a liquid of density and viscosity 1100 kg m−3 and 0.0075 Pa s
respectively. The design flow rate is 5 m3 hour−1 and the bed height
and diameter are 2 and 0.2 m respectively, using the Kozeny-Carman
equation the pressure drop is (Pa):
a: 99000 b: 1.32x107 c: 4400 d: 44000
(v). Using the Carman correlation the shear stress on the ion
exchange beads is (Pa):
a: 17.2 b: 3.5 c: 13.5 d: 27.0
(vi). Hence the dynamic pressure drop over the bed is (kPa):
a: 84 b: 130 c: 99 d: 150
(viii). If the liquid has a datum height equal to the position at the base
of the ion exchange vessel and, therefore, needs raising to the top of
the column before it enters the ion exchange bed the additional
pressure drop to effect this, i.e. the static pressure drop over the bed,
is (kPa):
a: 2.16 b: 21.6 c: 216 d: 0.22
4 Filtration of liquids
Filtration is the removal of suspended particles from a fluid,
performed by a filter medium, septum, cloth or bed of solids. In this
chapter we will discuss only liquid filtration, the removal of particles
from gases is covered in Chapter 14. Filtration is commonly
encountered in chemistry laboratories on a Buchner funnel and
within the kitchen during the making of filter coffee. It is a very
important industrial process as it is often a key stage in product
recovery: following reaction, precipitation and crystallisation stages,
but preceding thermal drying and packaging (e.g. in pharmaceutical
production). It is more economic to remove moisture from particles
by mechanical means, including filtration, than by thermal means. Fig. 4.1 In a filter increasing
resistance causes liquid
Thus, domestic washing machines provide higher and higher spin
flow to decrease with time
speeds prior to thermal, or evaporative, drying. – at constant pressure drop
There is a vast range of filtration types; depending upon whether
the objective is to produce a clean liquid, as in drinking water
production, or solids retained in a filter cake, as in product recovery.
The former process is called clarification, or clarifying filtration and is
often performed in equipment containing packed beds, which were
introduced in the previous chapter.
Figure 4.1. When the filtrate rate drops to an unacceptably low value,
History
the filter is backflushed at a rate of up to 36 m3 m−2 h−1 and may
In the UK, Dr John Snow
include an air scour and some degree of fluidisation (Chapter 7). So,
first traced a water borne
cholera epidemic to a for continuous production a second filter is required; so that whilst
public water pump in one is filtering the other can be cleaning and put on stand-by. In
Broad Street, London, in general, backflushing uses up to 5% of the previously filtered liquid.
1854. Surrounding areas In some applications the backflush cycle is timed and, for constant
not using the pump were rate filtration – where the pressure drop increases to maintain the
unaffected. The rate, the backflush is initiated by a pressure threshold. Figure 4.2
connection between water shows a schematic illustration of a bed filter with the associated flow
and disease led to mass
lines for operation.
deployment of sand filters
Various media may be used as the bed, including: sand, gravel,
(300 µm grains and 0.75 m
high), which were anthracite, activated carbon and garnet. The carbon based materials
previously just provides removal of some dissolved species as well as filtration of
considered to improve the suspended matter from the liquid. In many instances a mixed bed is
water look and taste. A used: coarse particles of one type on top of fine particles of another.
public house, called the One advantage of this arrangement is that, if the coarse particles treat
John Snow, now stands the down-flowing suspension first, they will remove the coarser
on the site of the old suspended solids – the finer material will pass between the coarse
Broad Street pump.
grains, but will become captured on the finer grains in the lower bed.
This segregation of filter media prevents the finer suspended material
from clogging up the gaps at the top of the filter bed. However,
cleaning by backflushing would normally result in the finer filter
grains transferring to the top of the bed, but this can be overcome by
designing the bed to contain dense fine grains (of say sand) and
coarse light grains (of say anthracite). An illustration of a mixed
media bed is given in Figure 4.3. Filtration performance is also
improved by the use of flocculants and coagulants (Chapter 13), prior
to passage through the bed, to increase the size of the suspended
particles to be captured within the media.
The basic design, or modelling, of deep bed filters can be achieved
using a simple equation
dN
= −λN (4.1)
dz
where λ is a filtration constant (m−1). Unfortunately, the filtration
constant will change with time, as material becomes deposited within
the bed; hence, equation (4.1) can only be treated as an ordinary
differential equation at the start of the filtration and integrated to give
N = N exp(−λ L) (4.2)
o o
where λo is the initial filtration constant, L is the depth within the
filter and N is the concentration of solids in the influent (on any
consistent basis: mass per volume, volumetric, etc.). Thus, the
concentration of solids deposited from the influent suspension
Fig. 4.3 Illustration of exponentially decays within the bed. It is possible to model the
mixed media deep bed performance of a bed filter by allowing the filtration constant to
with cut-away showing become a variable dependent upon the following
particle deposition (4.3)
λ = f (ε ε σλ )
s o o
Fundamentals of Particle Technology 31
AL 1
α= = (4.12)
kLACρ s kCρ s
Filtration modelling and the permeability can be provided by the expression given on
Permeability can be page 24, which demonstrates that it is a function of porosity (or cake
deduced from a particle concentration by volume fraction), as well as specific surface of the
size distribution –
particles. Likewise, the dry cake mass per filtrate volume will also
hence specific
resistance from (4.12), depend on the applied pressure, again this is because the cake
so long as the cake concentration increases with applied pressure. The relation between
concentration can be dry cake mass per filtrate volume is derived from a mass balance on
estimated. Likewise, solids and liquid entering the filter as follows.
equation (4.14) Total mass slurry filtered =M (kg)
provides a value for c. Mass fraction of solids in slurry =s (-)
Both can then be used Mass of solids in slurry = sM (kg)
in (4.11) for filtration Total mass of liquid in slurry = (1-s)M (kg)
modelling. However,
Mass liquid retained in filter cake = (1-C)ALρ (kg)
laboratory filtrations
used to scale-up So, provided that all the solids are retained in the filter cake, the dry
filtration data are much cake mass per filtrate volume (N.B. volume liquid in the filtrate and
more reliable. not total liquid volume) is
sM
c= (4.13)
[(1 − s) M − (1 − C ) ALρ ] / ρ
The solid mass can be represented in terms of both the solids retained
in the cake and the original solids in the slurry
CALρ s = sM (kg)
Filtration constants
The empirical constants which can be rearranged to give
tell us about the nature sM
AL =
of the material to be Cρ s
filtered:
and substituted in to equation (4.13)
vacuum filtration
sMρ 1
region c= = (4.14)
n < 0.3 and α < 1x1011 [(1 − s) M − (1 − C ) sMρ ] / Cρ s (1 − s) /( sρ ) − (1 − C ) /(Cρ s )
m < 0.1 So, with compressible cake filtration, both the specific resistance and
dry cake mass per filtrate volume are functions of the filtering
pressure filtration pressure because it influences the cake concentration. There are
region
several expressions relating the average specific resistance and
0.3 < n < 0.6
average cake concentration (by volume fraction) to pressure
1x1011< α < 2x1012
difference; the easiest to apply are
0.05 < m < 0.3
α = α o (1 − n)∆Pc n (4.15)
See: and
www.filtration-and-
separation.com for a list C = C o (1 − m )∆Pc m (4.16)
of measured values for where n, m, αo and Co are empirical constants. The coefficient ‘n’ is
common materials. sometimes referred to as the compressibility coefficient and the
greater the value the greater the compressibility. A value of zero
indicates incompressible solids, calcite is about 0.2, kaolin clay about
0.6 and fine sludges approach unity. Unfortunately, equation (4.15) is
unusable when the compressibility coefficient approaches unity.
The average dry cake mass per filtrate volume can be deduced
from the average cake concentration by equation (4.14).
Fundamentals of Particle Technology 35
1
c= (4.17)
(1 − s ) /( sρ ) − (1 − C ) /(C ρ s )
In equations (4.15) and (4.16) the pressure forming the filter cake
should be used; this will be less than the total applied pressure drop
because some of the applied pressure will be used to force the fluid
through the filter medium. However, the use of a higher pressure
difference in equation (4.15), than really exists, will result in a higher
specific resistance to filtration which is a safer design option than
under-estimating filtration resistance. The actual difference between
total applied pressure and that over the cake is normally very small.
So, provided that all the solids are retained in the filter cake, the dry
cake mass per filtrate volume (i.e. volume liquid in the filtrate and
not total liquid volume) is
sM sMρ
c= =
[(1 − s ) M − (m R − 1) sM ] / ρ (1 − s) M − (m R − 1) sM
which simplifies to
sρ
c= (4.21)
1 − mR s
Hence, a knowledge of the moisture ratio, slurry solids concentration
by mass fraction and liquid density provides the dry cake mass per
filtrate volume and the gradient from a plot similar to Figure 4.10 will
provide the experimentally determined value for specific resistance
to filtration. If a series of tests are conducted, at different total
filtration applied pressures, then it is possible to assess if the
resistance varies with pressure: a log plot of these is suggested by
equation (4.15), with the gradient providing the exponent in the
equation and the intercept can be used to deduce αo. A similar plot is
Fig. 4.12 Schematic used for concentration variation with pressure, in accordance with
diagram of a plate and equation (4.16).
frame filter When all the constants in equation (4.18) are know, including the
specific resistance and dry cake mass per filtrate volume, possibly
coming from laboratory test data, then filtration simulation can be
performed:
filtrate volume in time t =V (m3)
dry cake mass in time t = cV (kg)
wet cake mass in time t = mRcV (kg)
mass slurry filtered in time t = mRcV+ρV (kg)
The height of the filter cake can be obtained from a volume balance
conducted on the filter cake:
volume solids in cake + volume liquid in cake = cake volume
cV ( m R − 1)cV
+ = LA
ρs ρ
which can be rearranged to give
cV 1 ( m R − 1)
L=
A ρ + ρ (4.22)
s
Industrial filtration equipment operating under constant pressure
includes most vacuum filters, such as rotary vacuum (or drum) filters
and rotary disc filters. One of these is illustrated in Figure 4.11.
Equation (4.11) can be solved under the condition of constant rate
filtration. Under these conditions, the filtration pressure must increase
Fig. 4.13 Schematic in order to overcome the ever-increasing resistance to filtration due to
diagram of a simple cake growth. Equation (4.11) can be rearranged to give
pressure Nutsche filter µRm dV
µαc dV
∆P = 2 V + (4.23)
A d t A∆P dt
Fundamentals of Particle Technology 37
cartridge filter is used whilst the first is taken off-line for filter
element changing. However, at higher concentrations of suspended
solids the frequency of element changing would be too great and
crossflow filtration may be appropriate. In this instance a clean liquid
stream is still supplied, but a second liquid stream containing a
higher concentration of suspended material is recycled. This stream is
the retentate and the process is illustrated in Figure 4.16.
Despite the application of crossflow filtration the permeate flux
rate usually decreases due to the deposition of some solids on the
surface of the filter, as well as the deposition within filters when
using the design illustrated in Figure 4.14. A typical flux rate curve is
illustrated in Figure 4.17. The conventional units of membrane flux Fig. 4.17 Permeate flux rate
are litres of permeate per square metre of filter per hour (l m−2 h−1). It behaviour with backflushing
to restore the rate
may be possible to momentarily backflush some permeate through the
membrane to dislodge the solids retained on the surface of the
membrane, which will then be swept away in the crossflow, thus
restoring some of the original flux. However, it is much more difficult
to remove the solids deposited within the membrane filter matrix and
this may lead to an ever-declining flux rate, even with backflushing
applied.
During crossflow filtration two stages of modelling may be
considered: the period in which the flux declines, and a model to
predict the equilibrium value of the filtration flux. Flux decline
models are often based on pore blocking, or diminishing size,
concepts discussed in Section 4.5 and numerically the data is fitted to
versions of equation (4.26). From a practical perspective, models of
the equilibrium flux rate have greater application because these
enable the calculation of flow and process conditions required to
perform an industrial separation over an extended period in time.
Two general models are often used: cake filtration and film theory.
These are illustrated in Figure 4.18. Film theory is normally applied
to very small particles, less than 5 µm, and is founded in classical
mass transfer analysis. Cake filtration is usually applied to larger
particles and some of the resistance terms are described by the
equations discussed in Section 4.5, with additional forces due to
colloidal interactions as described in Chapter 13.
In MF the concentration of suspended material passing through
the membrane is usually negligible, cp=0, hence the retention (Rt) is
100%
cp
R t = 1 − 100% (4.27)
cb
In the cake filtration model the permeate flux rate (J) is described by Fig. 4.18 Cake and film
a form of the cake filtration equation (4.7) models of membrane
filtration
∆PTMP
J = (4.28)
µ ( R m + R c − Rs )
where the flux is
40 Filtration of liquids
dV 1
J= (4.29)
dt A
and the filtration is run under the conditions of constant pressure.
The filtration resistance terms are: membrane, cake and shear (for
crossflow). The latter term is negative because the crossflow shear
causes a reduction in the resistance. In many cases it is possible to
relate the reduction in resistance due to shear with the flow
condition, such as surface shear stress, flow rate, degree of
turbulence, etc.
In the film model, the convective transport of solids towards the
membrane is balanced, at equilibrium, by the diffusional transport
monofilament away from the membrane. This gives rise to a smooth concentration
gradient approaching the membrane, rather than the step change
seen in cake filtration. This gradient in concentration of material is
called concentration polarisation and, in the film model, it is here that
the resistance to filtration is assumed to be greatest. Thus, a film
equation is appropriate
D cm − cp
J = e ln (4.30)
δ c b − c p
where cm is the concentration at the membrane surface, De is the
effective diffusivity of the filtering species and δ is the distance over
which the diffusion takes place. The last two terms can be replaced by
multifilament
a mass transfer coefficient, as is conventional in film modelling, and
that coefficient can be related to flow conditions using film transfer
correlations. This enables the prediction of permeate flux under
different conditions of shear, crossflow velocity, etc. The
concentration cm is assumed to be a constant, once sufficient material
has arrived at the membrane to form it, so that an increase in filtering
pressure will not give rise to an increase in cm. Hence, once the value
of cm is achieved at the wall, the flux rate determined by equation
(4.30) will not be influenced by further increases in pressure because
the mass transfer resistance is within the diffusional region
represented by the concentration polarisation. Thus, permeate flux
needle felt
will be transmembrane pressure independent, after the threshold of
reaching cm at the membrane. Increasing the pressure difference will
Fig. 4.19 Examples of increase the thickness of the cm deposit, but will not influence the
filter cloths
concentration gradient at the membrane wall, equation (4.30).
bigger than the particles and this helps minimise the long-term
increase in medium resistance. However, the situation described by
Figure 4.5 is too simplistic, a more realistic representation is provided
in Figure 4.20: it is difficult to define where the cake finishes and the
filter medium starts and there is some penetration of the medium by
the particles. One important consequence of this is that it is not
possible for filter cloth manufacturers to define a value for Rm for
their cloth. The value must be determined in-situ, or during Fig. 4.20 More realistic
laboratory tests on the slurry, as it is a function of both the cloth and representation of
the material being filtered. Typically, the in-situ medium resistance is particles on a filter cloth
one or two orders of magnitude greater than the clean value. In many – c.f. Figure 4.5
cases this is also true during membrane as well as cake filtration.
One of the main distinctions between filtration media for cake
filtration is the use of monofilament or multifilament fibres for the
weaved cloth. Monofilaments give better cake discharge properties,
lower resistance due to internal solids deposition and lower flow
resistance in general; however, they are less effective at particle
retention and can be weaker cloths. Plugging of a monofilament is
also possible; whereas flow is possible between the fibres constituting
the multifilament cloth, so complete plugging is unlikely. When
filtering fine particles both types of cloths may not provide a good
clarity filtrate from the start and these first runnings can be recycled
back into the feed stream until the clarity of the filtrate improves.
Factors to be considered when selecting a filter medium, apart
from the ability to retain particles or form a cake, include: good cake
release, chemical compatibility with the liquid and solid to be filtered,
mechanical ability to withstand the filtration cycle (often including
stretching and rolling), cost, economic filtration time, adequate cloth
lifetime and resistance to particle blinding and biological growth. In
operation, it is common to include a cloth washing stage between
cake discharge and cloth reuse, to enhance cloth performance.
Laboratory tests to determine the cake resistance, as mentioned in
Section 4.5, are also used to evaluate different types of filter cloth and
their resistances, for possible use on a production filter. However, as
the interaction between the cloth and solids is significant, see Figure
4.20, it is important that the filtration test is performed following the
same orientation that the production filter will use. A rotary vacuum
filter has a filtering surface facing downwards; thus coarser and
easier to filter particles will settle away from the surface, so the test
should also use a downward facing filter leaf. An example filter leaf
is illustrated in Figure 4.21. Fig. 4.21 Laboratory test
When filtering finely divided material that may blind a cloth, or filter leaf – usually at
filter membrane, it is possible to deposit a sacrificial initial layer of constant pressure and the
solids of an additional material. The filtration takes place on the filter orientation should be
deposited solids and, when the filtration performance becomes similar to how the filter
unacceptable, the layer can be removed and replenished with a fresh will be used
one. This is called pre-coat filtration and secondary membrane
filtration in classic and membrane filtrations, respectively.
42 Filtration of liquids
4.9 Summary
In this chapter, we have extended the principles covered in Chapter 3
Equipment selection to cover the effects of depositing solids, within a packed bed and on
Attempts to specify
top of a filter medium. In most cases we are concerned with the
equipment based simply
pressure drop required for a given flow rate of filtrate, or permeate.
on particle size are not
successful. However, in
This is required for equipment design, or simply to understand the
general finer particles do process. Numerous filter types are available, these are not described
require higher operating in detail, but the hydraulic operating principles are. Filter cake
pressures for filtration. A washing and air blowing, to displace moisture from the cake prior to
free expert system for thermal drying, are not covered but will be considered in Section 8.5.
selection is available at: The reader should appreciate the many similarities between
www.filtration-and- conventional and membrane filtrations, including the tendency for
separation.com the filter medium resistance to be a function of what is being filtered
and not just the medium itself. For most filter mediums it is not
possible to define a true pore size, or diameter, but an equivalent one
provided by a flow test based on hydraulic resistance, or the ability to
displace one fluid. This is met again in Section 13.5 and is called the
bubble point test.
Selection of equipment for solid/liquid separation depends on
many factors, both equipment and process related. Expert systems
exist for this purpose and the reader is referred to one that is freely
available on the left. However, some existing operating data, or
laboratory tests, are essential for reliable equipment specification.
Fundamentals of Particle Technology 43
4.10 Problems
1. Deep bed filtration
A 3 layer filter is used to clarify an effluent containing 60 mg litre−1 The following
(ppm) of solids. Complete the following table: differential equation
describes the
distribution of solids
Depth Bed Initial filtration Inlet Outlet
within the filter bed:
medium constant concentration concentration
to section from section ∂N
= − λN
∂z t
(m) −1 (ppm) (ppm)
(m )
0.45 anthracite 2 60 where N is solid
0.28 sand 8 concentration in any
0.22 alumina 15 consistent set of units
and λ is the filtration
The final treated effluent concentration discharged from the filter will constant. By considering
be (ppm):........................ only the start of the
filtration the above
partial differential
2. Five litres of a 10% w/w slurry of chalk ( ρ s of 2670 kg m−3) in equation may be treated
water (µ of 0.001 Pa s) was filtered under constant pressure on a filter as an ordinary
leaf of 0.0314 m2 area. The specific surface of the chalk is 3x106 m−1 differential equation,
and the cake can be assumed to have had a porosity of 50%. and integrated with
respect to position (z)
i). The volume of solids in the slurry was (cm3):
a: 100 b: 500 c: 200 d: 250
ii). The volume of water retained in the filter cake was (cm3):
a: 400 b: 200 c: 100 d: 0 Cake filtration
3 Resistances in series are
iii). The filtrate volume was (cm ):
a: 4600 b: 4750 c: 4500 d: 4800 additive
iv). The mass of dry solids in the cake was (kg); ∆Ptotal = ∆Pcake + ∆Pmedium
a: 0.533 b: 0.500 c: 1.335 d: 0.668 Hence
v). The dry mass of solids per unit volume of filtrate (i.e. the same as µαc dV µRm dV
∆P = 2 V +
'c') was (kg m−3): A dt A dt
a: 112 b: 100 c: 116 d: 2670 where c is the dry mass of
vi). Using the Kozeny-Carman equation, the permeability of the cake cake deposited per unit
volume of filtrate:
was (m2):
sρ
a: 1.1x10−14 b: 9x1013 c: 3.3x10−8 d: 3x107 c=
−1
(1 − m R s )
vii). The specific resistance of the cake was (m kg ):
where s is slurry
a: 8.3x10−18 b: 22500 c: 2.5x10−11 d: 6.7x1010 concentration as a weight
viii). The time taken to filter the suspension, if the medium resistance fraction and mR is mass wet
can be neglected and the differential pressure forming the cake was cake sample/dried sample
36 cm of mercury ( ρ Hg =13600 kg m−3), was (s): (i.e. the moisture ratio). The
connection between
a: 55 b: 998 c: 1572 d: 1747 specific resistance (α) and
permeability (k) is
3. The same quantity of a similar suspension to problem 2 was 1
filtered on the same filter at constant rate until 2.3 litres of filtrate had α=
k (1 − ε ) ρ s
been recovered. The following empirical equation for pressure was
observed
∆P = 168Pa/second + 6670Pa
44 Filtration of liquids
Galileo (1564-1642) which can be combined with equation (5.3) to provide an expression
Galileo is credited with for terminal settling velocity (Ut), called Stokes’ law
dropping different
x 2 (ρs − ρ)g
sized balls from the top Ut = (5.5)
of the leaning tower of 18µ
Pisa to show that they In equation (5.4) the drag force is related to the particle velocity (U),
fell at the same rate.
for all values of velocity, whereas in equation (5.5) we are referring to
This ignores air drag
the final (terminal) settling velocity of the particle in a static fluid
and Galileo knew better
than this, he had after the period of acceleration (Ut). Clearly, the settling rate of a
worked on wind particle is a function of its size, solid density and physical properties
friction. Note that of the suspending fluid. Equation (5.5) is only valid when the degree
equation (5.5) is not of turbulences within the fluid is negligible, see Figure 5.2. This is
valid for balls in air, but measured by The Particle Reynolds Number
(5.11) is generally valid.
xUρ
Re' = (5.6)
µ
where the threshold for streamline flow past the particle is believed
exercise 5.1 to be about 0.2. The Particle Reynolds Number measures the ratio of
Determine the maximum
inertial to viscous forces within the fluid; hence it is the fluid
particle size at which
properties that should be used in it: fluid density and viscosity. At
Stokes’ law should be
applied for the data: Particle Reynolds Numbers greater than 0.2 the degree of turbulence
solid density: 2500 kg m−3 becomes more significant leading to an additional fluid drag force
liquid density: 1000 kg m−3
due to form drag. Hence, the terminal settling velocity will be lower
viscosity: 0.001 Pa s than that predicted by Stokes’ law, equation (5.5), which considers
i.e. use Re’=0.2 and only viscous drag around the particle.
substitute equation (5.5) in In common with fluid flow in pipes, it is possible to correlate a
to (5.6). friction factor with Reynolds number for the case of fluid flow past a
single spherical particle. This is illustrated in Figure 5.3. The friction
factor is the shear stress in a plane at right angles to the direction of
motion at the particle surface (R) divided by the fluid density and
relative velocity between the particle and fluid squared. The drag
force is the product of the shear stress and the
particle area, which is the projected area to the
fluid flow (Ap). In particle settling it is usual to use a
drag coefficient (Cd), rather than friction factor,
these are related as follows
f R
Cd = = (5.7)
2 ρU 2
where the fluid drag (FD) is
FD = RAp (5.8)
The projected area for a sphere is
π
Fig. 5.3 The drag coefficient or friction Ap = x 2 (5.9)
factor plot for single spherical particles 4
Fundamentals of Particle Technology 47
( )
2
Re' = 14.42 + 1.827 Ar 0.5 − 3.798
0. 5
(5.16)
valid for 2<Re’<20 000, where The Archimedes number is closely
related to equation (5.14) and is
( ρ − ρ ) gρ 2 x 3
Ar = s (5.17)
ρ µ
2
The fluid drag force may be subject to the mean free path correction,
which is required when the particle size is comparable to the mean
free path of the fluid. This is required because the particles can slip
between the fluid molecules - effectively reducing the viscous drag. It
is more prevalent when the fluid is a gas. The correction to the drag
Brownian motion
coefficient is
When small particles
C d = C d(STOKES) [1 + 1.7λ / x ] −1 (5.20) are suspended in
where λ is the mean free path length of the gas. For air λ=0.1 µm liquids, they are
(approximately) so the error in assuming the Stokes drag term is 17% subject to molecular
for a 1 µm particle and 170% for a 0.1 µm particle settling in air. bombardment giving
When particles come to rest on each other, or a surface, there is a rise to Brownian
solids stress gradient, or reaction force. This force can be rationalised motion. Hence, finely
by considering when a particle is at rest at the base of a vessel, as it divided particles may
experiences no drag or inertia, but still possesses a weight (field) not settle. In practice,
force. This force must be balanced as the particle does not accelerate. particles smaller than
The reaction force is due to a pressure, or stress gradient, exerted 2 µm suspended in
from the vessel base. water will settle
slower than predicted
5.3 Particle acceleration in streamline flow by Stokes’ law and
particles less than 1
In the derivation of equations (5.5) and (5.11) the drag force was
µm might not settle at
equated to the gravitational field force, to determine the terminal
all.
settling velocity of the particle. This simple force balance is only valid
if the particle inertial forces can be neglected. Therefore, the time
taken to reach the terminal settling velocity, or 99% of it, is a useful
check on the validity of the simple force balance used to derive these
equations. A force balance of the apparent mass (buoyed mass), drag
and inertia for a spherical particle is
πx 3 dU
( ρ s − ρ ) g - 3πµxU − m p =0 (5.21)
6 dt
where mp is the actual particle mass, not buoyed mass. Equation (5.21)
can be rearranged to give
dU π ( ρ s − ρ ) x g / 6 − 3πµxU
3
= (5.22)
dt mp
but
π
( ρ s − ρ ) g = 3πµxU t (5.23)
6
i.e. the equation for the terminal settling velocity.
Therefore,
dU 3πµxU t
= (1 − U / U t ) (5.24)
dt mp
We need to integrate equation (5.24) to find time taken to reach a
given velocity, or fraction of terminal settling velocity. The equation
can be rearranged to give
−1 / U t 3πµx
∫ dU = − ∫ dt
1−U / U t mp
50 Dilute systems
5.6 Summary
The settling velocity of small particles may be reliably obtained from
Stokes’ law. Larger particles, however, do not obey Stokes’ law.
Alternative correlations between drag coefficient and Particle
Reynolds number do exist – but the settling velocity is a constituent
of the Particle Reynolds number; hence the answer needs to be
known before the appropriate equation to use can be identified! To
overcome this problem Heywood published a set of tables that can be
used over a wide range of settling velocities and particle sizes.
The single particle settling discussed in this chapter is widely used in
engineering calculations. For example, within a spray drier trajectory
analysis is often performed using the drag coefficient and the
difference in velocity between the particle and the gas is of use in
mass transfer calculations. Single particle settling also forms the basis
for understanding the behaviour of more concentrated dispersions,
which is the subject of the next chapter.
Heywood Tables
5.7 Problems
(see Appendix) 1.
4 ( ρ s − ρ ) ρg
1/ 3 i). A solid and liquid has a specific gravities of 2.8 and 1.0,
PH = respectively and the liquid viscosity is 0.001 Pa s, the value of the
3µ 2 function PH is 2.87x104 m−1, the value for QH is (SI units):
and a: 2.87x10−2 b: 2.87x104 c: 2.87 d: 0.490
1/ 3
4( ρ s − ρ ) µg
QH = ii). The SI units of the function QH are:
3ρ 2
a: m−1 b: m s−1 c: m s−2 d: s m−1
Both functions are size
and velocity independent.
When calculating the iii).Use the Heywood Tables to complete the following:
settling velocity given a Particle diameter (µm): 1 10 50 100 1000
particle diameter (x) the log(PHx): ------ ------ 0.
value of log(PHx) is first log(Ut/QH): ------ ------
calculated. Then the first
two significant figures of Settling velocity* (m s−1) ------ ------
−1
log(PHx) are given by the Stokes’ settling velocity (m s ):
first column of the table, * settling velocity using the Heywood Tables
the second and third
come from the scale given iv). Why should the Stokes’ settling velocities of the larger particles
at the top of the table (the always be greater than those found in practice (and given by the
first row). The
Heywood Tables)?
corresponding value of
log(Ut /QH) is then read or
estimated from the table
and converted into a xU t ρ
v). The Particle Reynolds is defined as Re ′ =
value for Ut using the µ
calculated function QH.
which should be below some threshold for Stokes’ law to be
applicable. The maximum particle size at which Stokes’ law is
applicable for the above system is (µm):
a: 59 b: 5900 c: 127 d: 1271
Fundamentals of Particle Technology 53
2.
i). See the continuous settling basin on the right. What will be the
trajectory of particles the same size as the critical particle size, but
which start their descent from a height less than H ?
ii). The solid and liquid densities are 2900 and 1000 kg m−3, the
viscosity is 0.001 Pa s and the critical particle diameter is 50 µm, the
terminal settling velocity (Ut) is (m s−1):
a: 2.6x10−3 b: 2.6x10−5 c: 2.6x10−1 d: 0.52
ii). The effluent solid has the following particle size distribution:
Cumulative mass undersize (%): 100 92 80 62 48 31 18 8 4 0
Particle diameter (µm): 90 80 70 60 50 40 30 20 10 0
This represents the fraction of the total distribution below the critical
size but which still settles because the particles still reach the base of
the vessel in 30 minutes. Add this fraction to the fraction of material
in the size distribution above the critical size (which has all settled
out). See your answer to Part (ii) to help you find this. NB the next
question does not want the fraction settled – it asks for the effluent
concentration going to discharge.
vi). If the effluent discharge consent limit is 20 mg l−1, will you be able
to discharge this suspension?
6 Hindered systems and rheology
On increasing the particle concentration from the low values
considered in the previous chapter, the system properties will change
considerably from those of the continuous phase, usually water, and
the individual particles. In settling, the presence of a large number of
fine particles will hinder the fall of the larger particles and the very
small particles will be dragged down more quickly than under free
settling. There is an important question to be considered in our
treatment of high concentration suspensions: are we interested in the
behaviour of the particles compared to the continuous phase, such as
during hindered settling, or the behaviour of the suspension as a new
homogeneous phase, such as during pumping of a mixture? The
latter is the concern of buoyancy, viscosity and rheology and is
considered in the later sections. Initially, the settling of particles
within a continuous phase will be considered. Another fundamental
concern is the particle concentration at which hindered systems are
appropriate, rather than the models discussed in the previous
chapter. In general, hindered settling is appropriate when the particle
concentration is greater than about 1% by mass. To describe
concentrated suspensions we will need to use some of the definitions
Fig. 6.1 Illustration of
from Chapter 3. An illustration of porosity and solid concentration by porous medium –
volume fraction is reprinted in Figure 6.1. including settling
The industrial equipment in which hindered settling is conducted suspension
is simply tanks, which may be operated batch-wise or continuously.
Settling is a cheap method of concentrating solids, the driving
potential for it is free (gravity), but it provides only a limited final
solid concentration and the process is slow. However, it is very often
used in thickening a suspension before a more capital intensive
operation, such as filtration. The design of industrial thickeners is
covered in Section 6.3.
The technical term for the residual liquid above the interface is the
supernatant and, for the duration of this chapter, it will be assumed
that it is entirely free of solids. It will also be assumed that there are
no wall support effects for the settling suspension, a vessel of
diameter equal to, or greater than, 150 mm is often recommended for
this purpose, but tests in different diameter vessels can show if this
effect is important.
The hindered settling velocity is a strong function of the particle
concentration and, for a given material, nothing else; i.e. it is
independent of vessel diameter, shape, etc. The dependency on
concentration is logical: at the highest possible concentration no
Fig. 6.3 Variation of settling can take place, at low concentration the particle will settle
settling velocity with under free settling conditions. Thus, in hindered settling the settling
starting concentration velocity will be between these two limits and a mathematical
expression for hindered settling velocity could be considered to be a
correction term to the free settling velocity (Ut). The most famous
empirical relation between settling velocity (Uo) and solid
concentration by volume fraction (C) is the Richardson and Zaki
equation
U o = U t (1 − C ) n (6.1)
where n is a variable constant that depends on the Particle Reynolds
number and may be dependent on vessel diameter (d):
In Figure 6.3 all the settling interface plots are straight lines, followed
by curves. This can be explained by considering what takes place
within the settling suspension. At the start, the concentration is Cf and
uniformly distributed within the vessel. At a time δt the concentration
at the base of the vessel is Cf +δC. At the next instance in time, 2δt, the
concentration at the base is Cf+2δC. If we were to track the
concentration Cf+δC we would find that it is now slightly higher up
the vessel than the base. Thus layers of constant concentration appear
to propagate upwards. Of course, all the solids are settling; no solids
are propagating upwards – we are merely looking for a region of a
given concentration of solids. It takes some time before the solid
concentration below the settling interface increases from that of the
Fig. 6.4 Solid concentration original and, according to equation (6.1), settling velocity is a unique
increases starting at the base
function of concentration; hence, the settling velocity must remain
and apparently rising
upwards in the vessel
constant, giving rise to the initial straight lines in Figure 6.3. The
concentration increases from the base are illustrated in Figure 6.4.
Fundamentals of Particle Technology 57
In our settling suspension we can see that there are zones, or regions,
of: supernatant liquid, the original concentration (just below the
settling interface) and it is logical that there will be another zone
consisting of fully settled sediment at the base of the vessel. There is a
fourth zone: that of variable (increasing) concentration lying between
the sediment and initial concentration zones. This is illustrated on
Figure 6.5 and the mathematical description of the concentrations in
this zone was first published by Kynch in the 1951, and is covered in
Section 6.4. At this stage, it is reasonable to suggest that because the
settling velocity is a unique function of concentration the velocity at
which the concentrations propagate upwards will be also, giving rise
to lines emanating from the origin called concentration characteristics.
FC f
A= (6.7)
(U o + T )C u
but the batch flux at the underflow will be negligible compared to the
underflow withdrawal flux; hence we can write
FC f
A= (6.8)
TC u
Equation (6.8) is the design equation and to use it we must determine
the flux at the underflow concentration.
The underflow withdrawal induces a downward velocity within
the thickener which is constant for all the concentrations present from
Cf to Cu. Hence, the underflow withdrawal flux is a straight line on a
graph of flux against concentration, see Figure 6.10. Adding the batch
and underflow withdrawal flux together gives the composite flux
curve; which has a minimum at a critical concentration between Cf and
Cu. It is this concentration that has the minimum solids flux, or
handling, ability. Hence, if too much solids are added to the thickener Fig. 6.10 Underflow
the critical concentration (Cc) will build up within the device and withdrawal flux
eventually overflow. The batch, underflow and composite flux curves
are illustrated in Figure 6.11.
In Figure 6.12 the construction required to determine TCu, for use
in equation (6.8), is illustrated. The minimum composite flux occurs
at Cc and is numerically equal to the underflow flux at Cu; i.e. TCu.
However, the composite flux at the critical concentration does have
two components: that due to batch settling and underflow
withdrawal, which are marked on the figure. Clearly, it is possible for
the thickener to be operated under conditions that require a batch
flux less than the value at the batch flux curve, at Cc, but batch fluxes
greater are not possible. Hence, the limit of operation is where Cc
meets the batch flux curve. So, for a required underflow Fig. 6.11 All the flux
concentration (Cu) a line drawn through the concentration axis at Cu curves
and tangential to the batch flux curve will meet the flux curve at the
value of TCu. This value can then be used in equation (6.8) to
determine the thickener area.
Thickener height is not determined by flux theory and, in general,
thickeners are short vessels with diameters up to 50 metres.
Minimum heights are allocated for the raked zone (0.5 m), solids
storage zone (0.5 m) and clarification (0.5 m). Thus, a thickener is
usually 1.5 to 4 m in height, unless solids compression is important.
An alternative design method to the use of the batch flux curve
construction described above was originally described by Coe and
Clevenger. A flux balance between the feed and underflow provides
FC f Fig. 6.12 Construction to
G = ATC u ρ s = FC f ρ s hence T =
AC u determine the critical flux –
equation (6.8)
which can be substituted in to equation (6.7) and rearranged to give
60 Hindered systems and rheology
FC f 1 1
A= − (6.9)
Uo C C u
ρ m u 2−m d m
Re * = m
(6.19)
K 6m + 2
8 m
For Newtonian fluids, in the turbulent flow regime, the following
equation correlates the friction factor and flow Reynolds number
−1 / 2 f 1 / 2
f
= 2.5 ln Re + 0.3 (6.20)
2 2
The analogous correlation for power law fluids, and suspensions, is
( f )−1 / 2 =
m
4
0.75
(
ln Re* f (1− m / 2 )
) − m0.4
1. 2
(6.21)
6.8 Summary
If a continuous thickener is designed correctly sufficient area will be
present to allow the solids to settle into the underflow, whilst
supernatant liquid reports to the overflow. If the total solids flux is in
excess of the design value, the thickener will have insufficient
capacity for the throughput required. Suspension, at the critical
concentration, will build up inside the thickener and eventually
overflow. Above the hindered settling region there may be some free
settling (clarification) or simply a region of clear liquid. Some special
designs exist that enhance the throughput, such as the lamella
separator illustrated in Figure 6.17, but the principles of hindered
settling are similar to those described in the earlier sections – the
designs provide an enhanced effective plan area for the separation.
Equation (6.11) suggests that it is possible to use permeability,
covered in Chapter 3, to deduce the settling rate (Uo) given that for
incompressible (zone settling) the solids stress gradient will be zero.
This is true, and it is possible to rearrange equation (6.11) for settling
velocity and combine with equation (6.3), for flux, or equation (6.9)
for plan area of a thickener. If the Kozeny-Carman model of
permeability is used, however, it is usual to reduce the Kozeny
constant from the usual value of 5 to 3.3, as this is believed to give a
better fit to the data. However, there is a wealth of evidence to
support this so-called constant being a function of solids
concentration.
6.9 Problems
1.
i). A 3% v/v suspension is to be settled in a batch settling tank prior
to water reuse within a process. A sample of the suspension settled
64 Hindered systems and rheology
vii). At the end of the settlement period, i.e. on day 4 after pumping
the supernatant but before pumping the sediment out, the volume of
sediment is (m3):
a: 90 b: 527 c: 437 d: 1351
viii). Hence, given the vessel area from Part (vi), the height required
for the sediment is (m):
a: 0.049 b: 129 c: 0.129 d: 0.00013
ix). Draw a line from the origin of your settling graph to meet the
settling curve at the height given in Part (viii). This occurs at the time
(minutes):
a: 80 b: 160 c: 220 d: 280
You have now completed one design for this settling vessel: the
height comes from Part (iv), the area from Part (vi), and the time
required to settle from Part (ix).
x). Comment below on your design.
xi). The line you have just drawn from the origin to the settling curve
represents a solid characteristic at a concentration greater than that
given in Part (iii). It can be used to provide another vessel design.
Assuming that the maximum permissible time for settling is 24 hours,
the height of this characteristic after 24 hours is (cm):
a: 12.6 b: 28.0 c: 32.0 d: 56.0
xii). Assuming that this height again represents the depth of sediment
in the vessel, the new vessel area required to accommodate the total
sediment on day 4 is (m2):
a: 1650 b: 3290 c: 6590 d: 13200
xiii). Under these conditions the new vessel height will be (m):
a: 4.02 b: 2.17 c: 0.82 d: 0.28
xiv). Explain below how other characteristics can be used to provide
alternative designs:
2.
i). A continuous thickener is to be designed to deal with the effluent
from the last question. It will treat the 1000 m3 per day of suspension
fed at 3% v/v solids concentration and is to discharge underflow at
13.8% v/v solids. Use the settling curve from question (1) and a mass
balance to complete the following table.
Concn (v/v) 0.03 0.039 0.045 0.049 0.056 0.067 0.074 0.092
Height on axis (cm) 28 21.5
Velocity (m s−1) 3.2x10-5
Batch flux (m s−1) 9.5x10-7
note batch flux is the product of settling velocity and solid concentration
Mass balance, where Cf
ii). Plot the batch flux curve on the grid provided, or on graph paper. is any concentration in a
iii). Now a flux balance on a thickener provides the following result: batch settling vessel and
A(TC u ) = FC f = YC u Hf is the corresponding
where A is the thickener area, (TCu) is the critical thickener flux which height:
is the intercept of a line drawn as a tangent to the batch flux curve C f H f Aρ s = C1 H 1 Aρ s
and going through the desired underflow concentration, F and Y are
66 Hindered systems and rheology
the volume feed and underflow rates respectively, Cf and Cu are the
volume fraction feed and underflow concentrations respectively.
Note that T is, in effect, the velocity of solid movement in the
thickener caused by underflow withdrawal at the solid concentration
Cu. The critical flux in this thickener giving an underflow discharge
concentration of 13.8% v/v solids is (m s−1):
a: 10x10−7 b: 8.5x10−7 c: 7.2x10−7 d: 5.8x10−7
iv). The minimum thickener area for this duty is (m2):
a: 480 b: 29000 c: 16000 d: 960
v). If the thickener is circular in cross-section the minimum thickener
diameter is (m):
a: 25 b: 190 c: 140 d: 35
2 −1
vi). The underflow rate is (m hour ):
a: 1.25 b: 2.4 c: 4.6 d: 9.1
3 −1
vii). The overflow rate is (m hour ):
a: 40.4 b: 39.3 c: 37.1 d: 32.6
4.
i). Rheological tests on milk of magnesia have provided the following
In laminar flow of
data. Calculate the consistency coefficient and the flow index.
Newtonian fluids
pressure drop and flow
rate are proportional – shear rate (s−1): 7.2 16 64 320 720
see how this suspension shear stress (Pa): 7.0 9.1 14.3 24.2 31.6
shear thins?
ii). The suspension must be pumped 2.11 m down a pipe of radius
6.95 mm. Mean suspension density is 1300 kg m−3. Complete the
table and sketch the graph of pressure drop against flow rate. How
does it differ from what is expected of a Newtonian suspension?
solid’s outlet. The cyclone inlet is simply open to the freeboard. Solids
feed into a fluidised bed is often a significant challenge. Screw
feeders used on free flowing materials may form a seal to prevent the
gas escaping but, this cannot be guaranteed and rotary valves may
need to be used. Reaction rates within the fluidised bed are usually
rapid, hence the bed material is often inert, or reaction product, the
actual amount of solid reacting within the bed may be very small; e.g.
in fluidised bed combustion of coal a bed of calcium carbonate may
be used and the coal particles may only be between ½ to 2% of the
bed by mass. The calcium carbonate becomes oxide and then reacts to
form calcium sulphate if sulphur dioxide is present. This provides in-
Fig. 7.3 Up to the fluidisation situ sulphur dioxide emission removal. A high pressure drop over
point pressure with
the distributor plate is usually required to ensure adequate
superficial velocity is linear in
accordance with Darcy’s law:
distribution of fluidising gas over the entire area of the bed. Solids
off-take can be by a simple overflow or via the gas outlet, or cyclone.
µ
∆P = LU o
k
On increasing the flow
7.1 Minimum fluidising velocity
particles rearrange before The fundamental equation for fluid flow through porous media,
fluidisation giving rise to the under laminar flow conditions, is Darcy’s law, equation (3.4). In most
maximum – on decreasing instances the Kozeny-Carman equation (3.7) is preferred because it
flow a lower pressure drop has an explicit expression for permeability in terms of bed porosity
may be found, giving a and specific surface. The pressure at the base of a fluid (due to fluid
hysterisis.
weight) comes from the static component of Bernoulli’s equation
∆P = L ρ g
where L is fluid height, g is the acceleration due to gravity and ρ is
the fluid density. For a suspension, a similar equation is valid but, for
the static pressure due to the solid particles, we must take account of
buoyancy and the proportion of particles present. Clearly, if there are
no solids present there would be zero static pressure due to solids –
but rather than use the solid concentration by volume fraction we use
the porosity, see Figure 3.1, thus
∆P = L( ρs − ρ ) g (1 − ε ) (7.2)
−3
where ρs is the true solid density (kg m ). Combining equations (7.1)
and (3.4) – remembering that we stated that fluidisation occurs when
Fig. 7.4 Pressure gradient
the bed weight (per unit area) equals the fluid drag gives
with superficial velocity up to
and after fluidisation – linear k
U mf = ( ρ s − ρ ) g (1 − ε ) (7.3)
as above whilst L is constant, µ
but drops when L increases:
N.B. the minimum fluidising velocity is a superficial velocity (not
∆P µ
= Uo interstitial). It is common to see equation (7.3) written with the
L k permeability term expanded as provided by the Kozeny-Carman
equation, see equation (3.7), and assuming spherical particles,
equation (3.8), gives an alternative equation for minimum fluidising
velocity
( ρ s − ρ ) gε 3 xSv 2
U mf = (7.4)
180(1 − ε ) µ
Fundamentals of Particle Technology 69
Fig. 7.7 Bubbling behaviour 7.4 Gas flow patterns around bubble and stability
A bubble is stable so long as particles that are drawn up into the
bubble from underneath; i.e. are carried into the bubble by the fluid,
can fall back out again. Unstable bubbles fill in with particles from
the bottom. A simple schematic diagram of a gas bubble, within a
surrounding emulsion phase of the fluidised bed, is illustrated in
Fundamentals of Particle Technology 71
R 3
U θ = b3 (U b / 2 + U ) + (U b − U ) sin θ (7.7)
r
where U is the interstitial gas velocity, Ub is the bubble velocity, Rb is
the radius of the bubble and r is the radial coordinate (i.e. distance
from bubble centre). The gas velocity in the azimuthal direction is
assumed to be zero; i.e. the angular velocity of the gas coming out of
the plane of the paper is zero.
Considering Figure 7.12 representing a fast moving bubble, on the
horizontal plane the gas flow rate upwards though the bubble will be
the same as that going downwards through the cloud because there is
no interchange of gas with the surroundings. Hence, using A for the
Fig. 7.11 Illustration of cloud radius, the gas flow rate through the bubble will be equal to
spherical polar coordinates A
∫ 2πrεU θ dr (7.8)
Rb
At an angle of 90o sinθ is unity, hence the gas flow rate will be
A Rb 3
∫ 2πεr 3 (U b / 2 + U ) + (U b − U ) dr which integrates to give
Rb r
1
πε (U b + 2U )R b 3
1
( )
− + (U b − U ) A 2 − R b 2 (7.9)
Rb A
Fig. 7.12 Fast moving Now considering the periphery of the cloud, where there is no
bubble and cloud interchange of gas with the surroundings, Ur→0 and r=A can be used
in equation (7.6) to provide
Rb 3
=
(U b − U ) (7.10)
A 3 (U b + 2U )
Crossover velocity with rearrangement and substitution into equation (7.9)
At U=Ub gas recirculation 2
A3 R
in the cloud extends over: gas flow rate = − 2πε (U b − U )1 − 3 b (7.11)
A→∞ R b 2
i.e. crossover between the
two types of flow round a Equation (7.10) can be used to relate cloud and bubble radii, thus
gas flow rate = − πεR b 2 (− U b − 2U ) = 3πεRb 2U
bubble shown in Figure
(7.12)
7.10. So, at crossover:
U mf Using the assumption that the gas in the emulsion phase has the
Ub = same velocity as it did at incipient fluidisation, then the interstitial
ε mf
and Umf may come from velocity will be
equation (7.4). This links U = U mf / ε mf (7.13)
the properties of the
particles, bubble size and
and the porosity of the emulsion phase remains substantially
fluidisation bubbling unaltered from that of εmf thus combining equations (7.12) and (7.13)
type.
gas flow rate = 3πRb 2U mf (7.14)
Equation (7.14) suggests that the gas flow rate going through a
bubble is three times the minimum fluidising velocity flowing over
the bubble cross-sectional area. Another significant result from the
concepts used in this modelling is that the gas in excess of that
Fundamentals of Particle Technology 73
needed to fluidise the bed will pass through the bed in bubbles.
Hence, in aggregative fluidisation bed expansion will occur because
of displacement of particulate phase bed by the bubbles, not due to
the homogenous redistribution of solids represented by equation
(6.1); i.e. the Richardson and Zaki equation is not valid for
aggregative fluidisation.
The Davidson and Harrison model has been refined by other
research workers since the 1960’s, but it represents the starting point
for many of these later models. Its advantages include: a relatively
simple model to apply, appears to predict gas flow near a bubble,
shows how gas can by-pass solids in a fluidised bed, explains
stability of a bubble (by predicting gas flowing up through one) and
it predicts pressure variation near a bubble reasonably. However,
there are several deficiencies including: the bubble shape is wrong
(mushroom shape is observed), it doesn’t explain what happens
when particles enter a bubble, the maximum predicted bubble size is
wrong, it assumes incompressible phases and it uses Darcy’s law to
link the phases together. More recently, with the advent of increasing
and readily available computing power an alternative approach to
the analysis of homogeneous phases has been to consider the
behaviour of individual particles themselves. This is known as
Discrete Element Analysis, or Distinct Element Method.
dz dz d2z
= + 2 ∆t (7.16)
dt t + ∆t dt t dt
and
dz dz
z t + ∆t = z t + + 0.5∆t
(7.17)
dt t dt t + ∆t
The drag force used in the analysis usually comes from either the
correlations described in Chapter 3, for the dense phase – such as
Ergun’s correlation, or the drag coefficient described in Chapter 5, for
a region dilute in particles.
The contact force consists of normal and tangential forces, which
can be modelled in terms of a spring, dashpot and a friction slider as
shown in Figure 7.13. The normal and tangential component of forces
are expressed as the sum of the forces due to the springs and dash
pots. The stiffness of the springs and the viscous dissipation
coefficient of the dash pot determine the force with which the
particles interact. A very soft spring would cause the particles to
embed into each other, causing a large overlapping region between
the particles. By allowing some particle overlapping in space it is
possible to overcome the problem of defining how the particles
deform on contact and the mathematical solution is simpler. In the
model, two particles move towards each other until the contact force
changes its sign; i.e. from attraction to repulsion. Contact between a
particle and a wall is modelled in a similar way, except that the wall
is stationary.
Modelling particle interaction with its neighbours and the wall is
very computationally intensive and is restricted to simulations using
several thousand particles, at best. This may be sufficient for a
reasonable indication of the physics of what is taking place, but
Specific surface actual processes normally contain many millions of particles.
For gas phase reaction Nevertheless, DE modelling of processes within particle technology
modelling the specific will become more reliable and extensively used as computing power
surface is often much increases. They also have the advantage that additional forces can be
greater than that added into the solutions, e.g. colloidal interaction, and that a range of
provided by equations particle sizes can be considered.
covered in Chapter 2.
This is due to internal
cracks and pores in
7.7 Summary
catalyst particles. This In a packed bed of static solids the weight of the particles is
area may be determined transmitted to the base by a network of contacts between the
by gas phase adsorption particles. The two forces present are the solids stress gradient
tests. However, for (upthrust reaction force) and the particle weight. During upflow of
calculations of pressure fluid through the bed the equations for flow and drag, Darcy's law or
drop, e.g. equation (7.4),
Carman correlation, are valid and at some point the fluid drag will
the particle envelope
specific surface should
equal the bed weight. This is the fluidising point. After the point of
be used. This is the same fluidisation the bed may expand further, the gaps between particles
value as that covered in getting bigger. During fluidisation, the pressure loss of the fluid is
Chapter 2. equal to the bed weight and this will remain constant irrespective of
the bed expansion. Hence, the pressure drop remains constant.
Fundamentals of Particle Technology 75
When fluidised, the bed adopts all the properties of a usual fluid
including: lighter objects float, surface stays horizontal, solids would
pour out of a hole in the vessel side, the level of two connected beds
would be equal (like a U tube) and the static pressure within the bed
is given by depth x density. These properties assist in some of the
advantages of a fluidised bed, which include: easy to control and
automate, rapid mixing (excellent temperature control), large and
small beds are usable, very high heat and mass transfer coefficients.
Some of the disadvantages include: little scope for temperature
gradients giving thermal shock, a good distributor plate is usually
needed, expensive dust cleaning equipment may be necessary,
particle breakup, erosion and stirred tank behaviour gives non-
uniform residence times.
Fluidised beds are frequently used for reactors, such as the vapour
phase catalytic cracking of hydrocarbons. In this operation multiple Basic equations
beds are used; cracking taking place in one bed under one set of The Kozeny-Carman
equation for fluid flow
reaction conditions and regeneration of catalyst in a separate bed.
through porous media is
Transfer of catalyst is easily performed because of the fluidised
∆P 5(1 − ε ) 2 S v 2
behaviour of the solids being similar to that of a fluid. In liquid = µ U o
fluidisation, segregation between particles based on size and density L ε3
difference is possible and this is used in the regeneration of bed filters where ∆P is the pressure
using mixed beds of sand and carbon. Segregation is not so common drop over the bed, L is the
in gas fluidised beds because solid/solid mixing is encouraged by bed depth, µ is the
entrainment of particles within the wake of the gas bubbles. viscosity, ε is the bed
porosity, Sv is the specific
7.8 Problems surface area per unit
1. Minimum fluidising velocity (Umf) is a superficial velocity. Use volume of the particles and
Uo is the fluid superficial
the Basic equations box to derive an equation for the minimum
velocity.
fluidising velocity where Umf=f[g, ρ s , ρ , µ , ε and diameter (x)]:
During fluidisation the
U mf = pressure drop over a
fluidised bed is constant
2. and equal to the bed
i). A packed bed consisting of 1.96 kg of solids of density 2.8 g cm−3 weight, i.e.
is contained in a cylindrical vessel of 10 cm internal diameter, and the ∆P = (1 − ε ) g ( ρ s − ρ ) L
bed height is 20 cm. The volume of the vessel occupied by the bed is where g is the acceleration
(ml): due to gravity and ρ and
a: 157 b: 1570 c: 3140 d: 6280 ρ s are the fluid and solid
ii). The volume of the solids in the vessel is (ml): densities respectively.
a: 700 b: 1430 c: 70 d: 5490 The modified Reynolds
iii). The porosity of the bed is (-): number (Re1) is:
a: 0.554 b: 0.773 c: 0.446 d: 0.227 ρU o
Re 1 =
iv). The particle size is 500 µm and the liquid density and viscosity S v (1 − ε ) µ
are 1000 kg m−3 and 0.001 Pa s, minimum fluidising velocity is (m s−1): The Froude number (Fr) is:
a: 9.4x10−6 b: 0.021 c: 0.017 d: 0.0094 U mf 2
v). Was the use of the Kozeny-Carman equation justified? Fr =
gx
vi). What type of fluidisation is likely to occur?
76 Fluidisation
3.
Richardson and Zaki i). Another fluidised bed contains 1.8 kg of solids of density 1.3 g
Uo = U tε n cm−3 in a cylindrical 30 cm internal diameter vessel, with a bed height
where Ut is the terminal
of 20 cm. The bed porosity is (-):
a: 0.853 b: 0.976 c: 0.098 d: 0.902
settling velocity (modified for
wall effects if necessary) and ii). The particles used above have a diameter of 200 µm and the fluid
n is an exponent defined as density and viscosity are 1.29 kg m−3 and 2x10−5 Pa s, respectively.
follows The terminal settling velocity of a particle is 0.783 m s−1 (from the
n = 4.6 ; at Re’<0.2 Heywood Tables - check this), the Particle Reynolds number is, see
n = 4.4 Re −0.03 ; at 0.2<Re’<1 Chapter 5:
n = 4.4 Re −0.1 ; at 1<Re’<500 a: 0.101 b: 10.1 c: 10200 d: 1x107
n = 2 .4 ; at Re’>500 iii). For this system, the value of the exponent n in the Richardson
and Zaki equation should be:
a: 3.49 b: 4.60 c: 4.11 d: 2.40
iv). The superficial velocity required to fluidise these particles to
achieve the above conditions is (m s−1):
a: 1.8x10−5 b: 0.71 c: 0.53 d: 0.55
v). The gas flow rate required to fluidise these particles to achieve the
above conditions is (m3 s−1):
a: 0.15 b: 380 c: 0.039 d: 3.9
4.
Hint Q.4 Estimate the bed expansion and bypass fraction for a fluidised bed
See comments at the foot
with incipient fluidisation velocity of 0.1 m s−1 if it is operated at a
of page 72. A volume
balance is needed for the superficial velocity of 0.15 m s−1 and has 2 cm diameter bubbles rising
bed expansion. at 0.5 m s−1. At incipient fluidisation the porosity is 0.4.
5.
A gas bubble in a fluidised bed reactor was found to be typically 100
mm in diameter and to be rising with a velocity of 0.5 m s−1. Using
Hint Q.5 the Davidson and Harrison model, investigate the particle size at
See the box on crossover which the transition in the gas flow pattern occurs: from bubbles
velocity on page 72.
rising slower than the interstitial gas velocity to bubbles with gas
circulation between the bubble and its cloud. Assume typical values
of particle physical properties; take the viscosity of air to be 2x10−5 Pa
s. Finally, draw an estimated graph of the thickness of the cloud
surrounding a bubble of 100 mm diameter as a function of the size of
the fluidised particles.
6.
Using the Davidson and Harrison two phase model, derive an
equation for the rate of flow of gas into and out of a gas bubble in
terms of bubble radius, interstitial velocity and bubble velocity.
Discuss how the model may help in the design of catalytic fluidised
bed reactors.
8 Centrifugal separation
The basic equations for most centrifugal modelling were introduced
in Chapter 5. The liquid drag force was given in equation (5.4), under
streamline flow, and the centrifugal field force was provided in
equation (5.18). It is a simple matter to equate these to arrive at an
analogue equation to the terminal settling velocity, equation (5.5), but
with one significant difference
2 2
dr x ( ρ s − ρ ) rω
= (8.1)
dt 18µ
the distance with time differential is not constant. In a centrifugal
Fig. 8.1 Particle in rotation
field the particle moves radially, see Figure 8.1 and equation (5.18), and definitions
and the radial position is part of the field force – hence the particle
accelerates during its travel in the radial direction. Thus, to determine
the particle position as a function of time integration is required.
It is well known that from a strict physical definition of forces on a
particle, in circular motion, the centripetal force and not the
centrifugal force should be considered. An unrestrained particle
would leave its orbit tangentially if the centripetal force was
suddenly removed. This is what happens with particles in cyclone
separation from gases and this is discussed further in Chapter 14.
However, this chapter is concerned with separation of particles in Buoyancy
rotating flow within a viscous medium, usually water. The particle If a particle floats, rather
will not travel tangentially to one orbit, but to lots of orbits, giving than sinks, then it will
the impression of radial movement outwards (provided the particle is move inwards in a
centrifugal field. Particles
denser than the surrounding continuous phase). Mathematically, we
denser than the fluid will
can use the well-known expressions, such as equations (5.18) and
move outwards. The
(8.1), to describe this travel. centrifugal field acts like
As illustrated in Figure 8.1, the centrifugal acceleration is simply an enhanced
the product of the radial position (r) and the square of the angular gravitational field in
velocity (ω). The SI units of angular velocity are s−1, but calculated by equation (5.3) and it is
converting from revs per minute (rpm) into radians per second – then usual to speak in terms
ignoring the dimensionless radian term. In solid body rotation, such of the equivalent ‘g’
force: i.e. centrifugal
as a centrifuge, this is easily calculated from the rotational speed,
acceleration / 9.81 m s−2.
usually provided in rpm. Thus, 1 rpm is 2π s−1 as an angular velocity.
In free body rotation, such as the hydrocyclone, the angular velocity
is calculated from the tangential velocity (uθ) by
uθ
ω= (8.2)
r
this is also illustrated on Figure 8.1. In the hydrocyclone the principle
known as the conservation of angular momentum is used; in which
knowledge of the tangential velocity at any radial position can be
used to calculate the tangential velocity at another because
u θ1r1 = u θ 2 r2 = constant (8.3)
78 Centrifugal separation
=
2
( 2
Q π ro − rL Lω 2 ) (8.7)
Ut g ln (ro / rL )
It is notable that the left hand side of equation (8.7) is identical to the
left hand side of equation (5.28). Thus, equation (8.7) represents a
centrifuge that has the same settling capacity as the plan area of a
Fundamentals of Particle Technology 79
∑ machine =
( )
π ro 2 − rL 2 Lω 2
(8.9)
g ln (ro / rL )
Both equations (8.8) and (8.9) have the SI units of area and both
represent the theoretical plan area of a gravity settling basin that
would perform the same separation duty on the solids. Introducing
an efficiency factor (EA) to take account of poor flow distribution
within the machine and other factors reducing the separation
capacity gives
Q
= A o
(
E π r 2 − rL 2 Lω 2) (8.10)
Ut g ln (ro / rL )
It is worth noting that, under gravity, particles less than 2 µm in size
might not settle because of molecular bombardment from the liquid
and colloidal forces. However, in a centrifugal field the body force is
much stronger and these particles have a greater chance of settling.
Hence, a separation that might not be possible under gravity might
be possible in a centrifuge. Obviously, a separation that is possible
under gravity will be much quicker in a centrifuge, due to the
enhanced g force. However, if there is little density difference
between the particle and fluid the separation under gravity and in a
sedimenting centrifuge will be slow, or impossible.
In a continuous sedimenting centrifuge there is always the
problem of how to remove the deposited solids continuously and
Fig. 8.3 Some sedimenting
how to enhance the separation. Various designs are used including:
centrifuge designs
scroll discharge decanter, time activated nozzle discharge disc stack
and the continuous tubular centrifuge. These are illustrated in Figure
8.3. In the scroll discharge machine the Archimedean scroll rotates
very slightly slower than the centrifuge, to convey the solids up the
beach and out of the machine. The disc stack centrifuge provides a
lot of parallel settling chambers, similar to the lamella separator in
Figure 6.16. However, solids discharge is usually intermittent from
this machine, limiting its application to low concentration slurries. In
the tubular bowl centrifuge there are no internal structures, so it is
possible to run the machine at very high rotational speeds, up to
Fig. 8.4 The hydrocyclone
30000 rpm. Continuous discharge relies upon flushing material out of and flow patterns inside
the machine using displacement by incoming material and these
devices are usually used on liquid/liquid separations or emulsions.
80 Centrifugal separation
tangential:
8.2 Hydrocyclones
Figure 8.4 illustrates a hydrocyclone, including the flow patterns
found within the device. The tangential inlet causes the fluid to
rotate, rather than mechanically rotating the wall, hence these devices
are described as having no moving parts. Of course, a pump or other
prime mover for the suspension is required. The flow pattern within
the hydrocyclone is complex and there are three velocities that need
to be considered. The tangential velocity gives rise to particles subject
to the centrifugal field force and is, therefore, critical to the operation
of the hydrocyclone. Tangential velocities of the liquid, and solids,
may be up to 20 m s−1. The radial velocity is much lower, usually less
than 0.1 m s−1. However, within the device there is a net flow of
liquid towards the centre and a net flow of solids away from it.
radial: Hence, it is important to distinguish between the radial liquid, or
solid, flow. The third velocity is in the axial direction. This velocity
has to be considered carefully because the hydrocyclone has two
outlets, continuously splitting the feed in to two separate streams.
The overflow contains a suspension that is more dilute than the feed
and has a finer particle size distribution. By contrast, the underflow is
a suspension more concentrated than the feed and has a coarser
particle size distribution. Thus, the hydrocyclone acts as both a
thickener (i.e. concentrates a suspension) and a classifier (i.e. selects
particles of a specific size). The axial velocity must take material to
the two outlets and the suspension near to the wall of the
hydrocyclone flows axially to the underflow. The material near to the
centre of the hydrocyclone flows axially to the overflow. Hence, there
is axial flow downwards, and upwards, within the hydrocyclone as
axial: illustrated in Figure 8.5.
A further understanding of these axial flows can be obtained by
considering the hydrocyclone primary and secondary vortices. The
primary vortex spirals down towards the underflow taking the larger
particles with it: these are centrifugally encouraged towards the wall
of the hydrocyclone. However, the geometry of the hydrocyclone
causes flow reversal towards the central axis of the device, giving rise
to the secondary vortex that spirals upwards to the overflow. It is the
finer particles that are caught within the secondary vortex. A vortex
finder is used on the overflow to minimise short circuiting of solids
from the feed directly to the overflow.
Referring to Figure 8.5, the argument can be made that because
there is axial flow upwards into the overflow and downwards into
Fig. 8.5 Velocities within underflow there must be a shear plane within the hydrocyclone
the hydrocyclone where there is no net velocity in the axial direction. In fact, the shear
plane is a surface, or locus, because the device is three dimensional;
Fundamentals of Particle Technology 81
hence, this has become known as the Locus of Zero Vertical Velocity
(LZVV). However, it is a misnomer because the hydrocyclone could
operate equally efficiently in any orientation and the term should
really be locus of zero axial velocity. The concept of this LZVV is very
important and it can be used to explain some of the observed
behaviour in the hydrocyclone. Within the device it is possible to set
up a force balance between the centrifugal force and the liquid drag
force. The latter pulls the particles inwards, as liquid must flow
inwards in order to enter the overflow. Hence, particles may adopt an
orbit; where the drag force is balanced by the centrifugal force.
Particles that orbit at a radial distance greater than the LZVV will be
in the primary vortex and will tend to report to the underflow.
Particles orbiting at radial distances less than the LZVV will be in the
secondary vortex and will be carried into the overflow. Thus, the Fig. 8.6 Equilibrium orbit at
particle size that orbits at the LZVV will have no preference for the the LZVV with liquid drag
overflow or underflow; i.e. it will have an equal chance of entering and centrifugal forces
either exit. This is defined as being the cut size (x50) of the balanced
hydrocyclone, see Figure 8.7. Again, the term is misleading as it could
be assumed that no particles bigger than the cut size enters the
overflow; which is very significantly different from the true meaning
of equal chance of entering either flows. For a particle to be radially
stationary on the LZVV the force due to liquid drag inwards must be
balanced by the particle centrifugal field force outwards, Figure 8.6
πx 3
3πµxu r = ( ρ s − ρ ) rω 2
6
which can be rearranged to give
18µu r
x 50 = (8.11)
( ρ s − ρ ) Rω 2
where R is the radius of the LZVV. Hence, it is possible to predict the
cut size of the hydrocyclone provided the terms on the RHS of
equation (8.11) are known. This approach is known as the equilibrium
Fig. 8.7 Size distributions
orbit theory. There are three variables that must be deduced before the
of feed, overflow and
cut size can be estimated: the radial liquid velocity, the radial position underflow showing the
of the LZVV and the angular velocity at the LZVV. The remaining cut size. Note that this is
physical constants should be straightforward to obtain. an idealised plot
The radial position of the LZVV is deduced by assuming that the assuming equal volume
LZVV is a shape that is identical to that of the overall hydrocyclone, flow split between the
but at a smaller radius, and that the volumetric flow split of overflow underflow and overflow
rate compared to feed rate is equal to the volume ratio within the – hence the two outlet
hydrocyclone; i.e. curves should equal the
feed curve.
volume inside LZVV volume overflow rate
=
volume inside hydrocyclone volume feed rate
see Problem 8 for a worked example of this.
Once the radial position of the LZVV has been deduced it is
possible to calculate the radial liquid velocity if it is assumed that the
liquid flow entering the overflow is uniformly distributed over the
82 Centrifugal separation
E1 − Rf
Reduced grade efficiency = E 2 = (8.13)
1 − Rf
and a sharpness of separation can be deduced from
x
Sharpness of separation = 25% (8.14)
x 75%
A residence time model for hydrocyclones, similar to that derived for
centrifuges before equation (8.7), provides the following result
x50 2 ( ρ s − ρ ) L∆P
= 3 .5 (8.15)
µρQ Fig. 8.9 Definition of
terms within a filtering
The derivation was first published by Rietema and his work was later centrifuge
extended by Svarovsky. Consideration of equation (8.15) leads to the
conclusion that there should be an optimum design: minimising cut
size and pressure drop and maximising flow rate. Equation (8.15) is
useful as it relates cut size to pressure drop over the hydrocyclone.
The operations in italics in Table 8.1 are considered in the rest of this
chapter, starting with the modification of the filtration theory already
84 Centrifugal separation
Chapter 13. A finite time is required for the cake to drain, hence the
rate of approach to the equilibrium saturation level of the filter cake
is also important. The equilibrium saturation is known as the
irreducible saturation and saturation is usually reported in terms of a
ratio based on the initial value: i.e. after cake formation a fully
saturated cake has a value of 1, but during dewatering the relative
saturation approaches the irreducible saturation if sufficient time is
allowed. This ratio of saturation values is known as the relative
saturation. There are two well-known models for saturation
modelling, one developed by Wakeman and based on particle
Fig. 8.14 Comparison of rotation
properties and another based on the work of Zeitsch, which is based speeds and resulting drainage
on a Boltzman distribution of pore diameters. However, in curves.
application the Zeitsch model applies the data obtained from the Operating data:
filtration stage and will be discussed further here. filter diameter 0.3 m
The relation between filter cake permeability and cake specific solvent density 980 kg m−3
resistance was provided in equation (4.12); permeability is used in solids density 1400 kg m−3
Zeitsch’s dimensionless Drainage (DN) number solvent viscosity 0.001 Pa s
cake concn. 0.3 v/v
kro 2ω 4 ρ 2 ( ro − rc ) 2
DN = (8.29) specific resistance 6x1010 m kg−1
(1 − C )σ 2 cos 2 θ rotational speed 3400 rpm
where σ is the surface tension and θ is the contact angle of the liquid surface tension 0.07 N m−1
contact angle 10 Degrees
on the solid. Zeitsch deduced the irreducible saturation ( S ∞ ) to be
cake thickness 50 mm
−1 πD N 1
S ∞ = 1 − exp + 1 − erf (8.30)
D 2D D
N N
N
The error function (erf) is built in to most modern spreadsheets,
hence evaluation of operating parameters, such as rotational speed,
by equations (8.29) and (8.30) on a spreadsheet is straightforward.
For a kinetic evaluation of drainage, Zeitsch defined a drainage
rate constant (φ), which has the SI units of s−1, as
σ 2 cos 2 θ
φ= (8.31)
2 µ ( ro − rc ) 3 ρroω 2
and, for convenience, it is possible to combine the drainage rate
constant and drainage number into a single dimensionless term (B)
1
B= + φt (8.32)
DN
where t is the time during drainage. According to the Zeitsch model,
the relative saturation with respect to time is
S t = S∞ +
1 1
exp(− B ) −
[ ]
1 − erf ( B ) π
(8.33)
2
DN B B 2 B Fig. 8.15 Effect of cake
thickness on saturation at a
Figure 8.14 provides an illustration of the above theory used to rotational speed of 3400 rpm
compare rotational speeds on relative saturation and drainage times. – all other variables are as
On Figure 8.14, it is noticeable that there is little saturation given below Figure 8.14
reduction by spinning at 1400 rpm; the equilibrium value (irreducible
88 Centrifugal separation
8.5 Summary
A centrifugal field force can be used to speed up sedimentation and
filtration. A sedimenting centrifuge, or hydrocyclone, may be used
instead of gravity settlement and a filtering centrifuge enhances the
rate at which liquid passes through the filter cake and cloth. In
practice, in filtering centrifuges centrifugal sedimentation should be
expected as well as filtration. The models of gravity sedimentation
and pressure filtration were adapted to a centrifugal field force in this
chapter.
8.6 Problems
1.
i). The equation for the rate at which a particle will settle in a
gravitational field, neglecting the acceleration of the particle, is Ut=...
x3 (ρs − ρ )g x 2 (ρs − ρ )g x 2 (ρs − ρ )g
a: U t = b: U t = c: U t =
18µ 9µ 18µ
ii). The equation for the rate at which a particle will move in a
centrifugal field, neglecting the acceleration of the particle, is dr /dt =...
d r x 2 ( ρ s − ρ ) rω 2 d r x 3 ( ρ s − ρ ) rω 2 dr x 2 ( ρ s − ρ ) g
a: = b: = c: =
dt 18µ dt 18µ dt 18µ
Fundamentals of Particle Technology 89
v). If the oil is less dense than the surrounding water an oil droplet Questions 2 and 5
will travel inwards on the application of a centrifugal field force, just Consider the axial and
as oil floats in a gravitational field. In the integration of the radial flow in a tube
centrifugal rate expression with respect to time and radial distance, centrifuge. If the start
the upper or top limit of the integration of radial position is (cm): radius is, in fact, the
a: 30 b: 12.5 c: 15 d: 27.5 inner diameter of the
centrifuge, i.e. between r
L
vi). The oil and water densities are 800 and 1000 kg m−3, respectively and the central axis there
and the viscosity of water is 0.001 Pa s, the time taken (neglecting is only air, and we
acceleration) before a sample of emulsion withdrawn at the inner consider a critical
radius contains no particles bigger than 10 µm in diameter is (s): particle trajectory we can
a: 15 b: 2290 c: -15 d: 150 derive the following
equations
Q
2. In a continuous tube type centrifuge 5.4 m3 min−1 of an aqueous ∑ PROCESS =
suspension is being processed and all the particles of diameter 10 µm Ut
or more are being removed. The solid and liquid specific gravities where Ut is the terminal
are 2.8 and 1.0 respectively, and the liquid viscosity is 10−3 Pa s. settling velocity of the
critical particle under
i). The volume flow rate was (m3 s−1): consideration. This is the
a: 0.0015 b: 0.09 c: 324 d: 5.4 area of a settling basin
ii). Stokes' settling velocity of the 10 µm particle was (m s−1): that would perform the
a: 9.8x10−6 b: 9.8x10−7 c: 1.5x10−4 d: 9.8x10−5 same duty under the
iii). The particle Reynolds number was: given flow rate. The
theoretical capacity of
a: 2.7x10−3 b: 275 c: 9.8x10−5 d: 9.8x10−4
the centrifuge is
2
iv). The sigma process value was (m ); ∑ MACHINE = ...
a: 920 b: 150 c: 9200 d: 590
π ( ro 2 − rL 2 ) Lω 2
…
3. The machine used in the above question had a length and g ln(ro / rL )
diameter of 1.5 and 0.75 m respectively. The pool depth (i.e. ro - rL)
was 0.1 m, and the operating speed was 1800 rpm.
i). The rotational speed was (s−1):
a: 190 b: 380 c: 11000 d: 94
3
ii). The volume of the centrifuge pool ONLY was (m ):
a: 0.31 b: 0.20 c: 0.62 d: 0.05
iii). The 'sigma machine' value was (m2):
a: 19 b: 840 c: 3600 d: 2200
iv). Compare the two sigmas for the efficiency of the machine (%):
a: 26 b: 4.3 c: 110 d: 30
90 Centrifugal separation
5.
i). Derive an expression for the volume fraction of a centrifuge, based
on an imaginary start radius that exists between rs and ro (i.e. p=).
ii). Using rs instead of rL and equating the two sigma terms (see box
above), derive an expression for rs.
iii). Combine the answers to parts (i) and (ii) to give an expression for
the proportion of particles removed as a function of flow and
material properties, i.e. p or the grade efficiency=...
The equilibrium orbit (R) of
8. The feed and overflow rates of the hydrocyclone illustrated are 36
a particle at the locus of
zero vertical velocity is
and 23 litres per minute, the liquid viscosity is 0.0015 Pa s, the solid
given by and liquid densities are 2000 and 1000 kg m−3, respectively. These
x 2 ( ρ s − ρ ) Vi 2 questions determine the separation size for the hydrocyclone using
R= equilibrium orbit theory.
18µ U
i). If Ro is the internal radius of the hydrocyclone, the fractional
where Vi is the tangential
volume inside the locus of zero vertical velocity is:
velocity of the liquid in the
hydrocyclone at the a: (R/Ro) 3 b: (R/Ro) 2 c: πlR 2 d: lR2/lRo2
equilibrium orbit position ii). By assuming that the fractional volume inside the locus of zero
and U is the inward radial vertical velocity is equal to the proportion of the feed flow going into
velocity of the liquid. The the overflow (O/F), the equilibrium radius was (m):
principle of the
a: 0.0052 b: 0.016 c: 0.032 d: 0.0080
conservation of angular
momentum gives
iii). By considering the locus of zero vertical velocity as a surface over
which the flow entering the O/F is uniformly distributed the liquid
Vi ri = constant
and the surface area and
velocity over the LZVV was (m s−1):
volume of a cone are a: 0.025 b: 0.057 c: 0.0186 d: 0.069
πrl πr 2 l / 3 iv). The inlet velocity, hence tangential velocity at Ro, was (m s−1):
where l is height or length. a: 5.97 b: 11.9 c: 7.32 d: 3.0
v). Using the principle of the conservation of angular momentum,
and assuming that all the inlet flow occurs at Ro, the tangential
velocity at the equilibrium orbit radius was (m s−1);
a: 14.9 b: 45.8 c: 9.54 d: 7.47
vi). The separation or cut size of the hydrocyclone was (µm):
a: 2.0 b: 6.0 c: 9.4 d: 12
vii). The particle Reynolds number at the LZVV was:
a: 0.074 b: 0.12 c: 0.15 d: 0.30
0. 8 K2
ρ x85
u LDV = K1 gx85 s − 1 (9.1)
ρ D
where x85 is the particle diameter below which there is 85% by mass
of the distribution, K1 and K2 are constants. The value of K1 is 650,
when viscous forces predominate, and 0.19 in the turbulent region.
The value of the constant K2 is 0, when viscous forces predominate,
and -2 in the turbulent region.
During heterogeneous flow there will be two identifiable regions,
as illustrated on Figure 9.1, a homogeneous region flowing above a
heterogeneous one. It is assumed that the overall pressure drop, or
gradient, for the two phase flow is the sum of the pressure drops in
both of these regions. Using the empirical correlation proposed by
Durand (Wasp, E.J., Kenny, J.P. and Gandhi, R.L., 1979, Solid-liquid
flow slurry pipeline transportation, Gulf Publishing, Houston, Texas.)
gD ( ρ − ρ )
1.5
∆P = ∆Pw 1 + 85C s (9.2)
u2ρ C
d
where ∆Pw is the calculated pressure drop for the homogeneous
phase, u is the mean suspension velocity and Cd is the particle drag
coefficient. The homogeneous phase pressure drop may be calculated
using Newtonian flow equations with equation (6.14) for the
viscosity. The drag coefficient comes from equation (5.13), with the
terminal settling velocity in that equation provided by Stokes’ law or
The Heywood Tables.
where
a = 1440 x + 1.96 (9.4)
b = 1100 x + 2.5 (9.5)
where Ms is the mass flow rate and SI units must be used throughout.
The saltation velocity is a superficial velocity and the required gas
velocity for use in the friction correlations is the interstitial. Also, to
ensure that the flow is in the dilute phase region an excess gas
velocity of 50% may be employed. Thus, with 50% excess the gas
velocity is
u salt
u g = 1 .5 (9.6)
ε
where ε is the porosity of the gas phase and is, in general, close to
Fig. 9.3 Types of flow found
unity. Having established the superficial gas velocity (Uo) the solids in pneumatic conveying
velocity (us) can be calculated using an equation for slip
u s = U o (1 − 0.0638 x 0.3 ρ s 0.5 ) (9.7)
where SI units must again be used for dimensional consistency.
Clearly, the greater the particle size and solid density the greater the
degree of slippage between the gas and particle velocities. The
interstitial gas velocity is used to calculate the friction of the gas in
the pipe, comes from standard methods for single phase pressure
drop calculations. It is assumed that the overall pneumatic conveying
pressure drop is the sum of all the pressure drops due to the following,
which also indicates where the equation originates:
acceleration of the gas (from Bernoulli’s equation)
1
ρεu g (9.8)
2
acceleration of the solids (from Bernoulli)
1
ρ s (1 − ε )u s (9.9)
2
friction of the gas on the pipe wall (from friction factor)
L
2 f g ρu g 2 ε (9.10)
D
friction of the solids on the pipe wall (from friction factor)
L
2 f s ρ sus 2 (1 − ε )
D
which, using continuity, is the same as
L
2 f s Gs u s (9.11)
D
where Gs is the mass flux of solids,
the static head of the gas is (from Bernoulli)
ρεgL sin θ (9.12)
94 Conveying
9.5 Summary
The transportation of particulate solids is an important consideration
of most processes involving solids. Often equipment selection is
based on process considerations such as the need to avoid
contamination from the environment of the valuable solid such as a
pharmaceutical product. If required throughput is high, normally
associated with low value products such as minerals, then a
mechanical device is appropriate such as a belt conveyor. If the
material is sticky, due to moisture or temperature softening, then a
chain conveyor may be required. For most chemical and food
processes pneumatic conveying is preferred because it is enclosed
and controllable. Using a rotary valve to meter the solid input into
the system the solid mass flow rate can be easily relayed to the
control room, and changed by it, if the valve is controlled by an
inverter.
9.6 Problems
Design a positive pressure dilute phase pneumatic conveying system
to transport 800 kg h−1 of sand of density 2500 kg m−3 and mean
particle diameter 80 µm between two points in a plant separated by
30 and 15 m in the vertical and horizontal directions, respectively.
Use ambient air ( µ = 1.84x10 −5 Pa s; ρ = 1.2 kg m -3 ) and assume that
eight 90o bends are required in the circuit. Start your worksheet
solution based on a pipe with an internal diameter (D) of 40 mm, then
try the pressure drop using pipe diameters of 50, 63 and 78 mm.
Which one would you recommend? Provide answers to the following
calculation stages.
Gs
5. The porosity (ε) is ε = 1 − .
ρ s us
(u g − u s ) ρx
8. The particle Reynolds number is Re' = .
µ
12. The pressure drop due to acceleration of the gas is = 0.5( ρεu g 2 ) .
22. The pressure drop due to the gas static head is ρεgLV sin θ .
28. Now calculate the gas flow rate in m3 s−1 for each pipe.
The bulk density is the combined density of the powder and the
void space (i.e. porosity see Figure 3.1). Hence the bulk density is the
same as the mean suspension density, equation (6.12), but as the fluid
is air its contribution to the mean density is ignored. Thus
Values of the required stress to break a stable arch are deduced and
the hopper is designed to provide these conditions with this powder.
This is illustrated in Figure 10.9. Although most of the following text
considers hopper design, the principles are more generally applicable
to powder flow and characterisation than simply hopper design: i.e.
the Powder Flow Function (PFF) or sometimes called the Material Flow
Function, characterises the ease, or otherwise, of powder transport
and storage.
During the discharge of a mass flow silo there is a possibility that
the powder flow may stop, or become intermittent, due to the
formation of a stable arch within the hopper. The reasoning is that the
particles interlock to form a stable arch over the opening much like
bricks in an arched bridge. In practice, sufficient force should be
provided by the system, if correctly designed, to break this bridge if it
forms. Firstly, a force balance can be used to determine the
magnitude of forces present in this powder bridge. Consider a
hopper with an opening of diameter B and a slice of powder of depth Fig. 10.9 Stable arch formed
∆h. If an arch forms there will be air on one side and powder on the above the hopper outlet
other. The powder within the arch will have a yield stress; given
sufficient stress above the arch it will break. This value of stress is
called the unconfined yield stress (fc): unconfined because the powder is
open to air on one side. An illustrative balance (taking moments from
the wall) on the plane forming the arch provides the following result,
see Figure 10.9
B
f c AreaL = W (10.10)
2
where W is the powder weight, B can be taken to be the hopper
opening diameter and L is the linear distance that the arch can be said
to act over. The force from the weight of powder is
W = Area∆hρ b g (Newtons) (10.11)
Combining these two equations gives the following result
∆h ρ b gB
f c = ρ b gB = (10.12)
2 L H (θ )
where L, and hence H(θ), is a function of the geometry of the opening.
Thus the minimum hopper opening diameter needs to be
f c H (θ )
B= (10.13)
ρb g
The next stage is to identify the unconfined yield stress for a powder
inside a hopper, and to know more about the functional relation H(θ).
2
1 1
τ 2 = − σ − (σ x + σ y ) + (σ y 2 − 2σ yσ x + σ x 2 )
2 4
Thus,
2 2
1 1
τ 2 = (σ x − σ y ) − σ − (σ x + σ y ) (10.19)
2 2
i.e. the equation of a circle in σ and τ coordinates, see equation (10.16)
above. So, the principal planes illustrated above can be represented
on a Mohr's circle. Note that the minor principal plane occurs where
the normal stress axis (σ) is cut at the lower end of the circle (τ=0) and
the major principal plane is at the top end of the circle. We call these
σy and σx respectively - see the Mohr's Circle diagram.
The angles between planes and in the Mohr's Circle are related, as
can be seen from the vertical and horizontal force balances on the
principal planes (rearranged for tan θ)
τ σ y −σ σ y −σ
tan θ tan θ = tan 2 θ = =
σ −σ x τ σ −σ x
and using the general result
2 tan θ
tan 2θ =
1 − tan 2 θ
So,
2τ
σ − σx
tan 2θ =
σ y −σ
1−
σ −σ x
Hence,
2τ τ
tan 2θ = = (10.20)
(
2σ − σ x − σ y σ − 1 σ x + σ y
2
) Fig. 10.12 Mohr’s circle with
angle by inspection
Comparing equations (10.20) and Figure 10.12, shows that
2θ = φ (10.21)
i.e. the principal plane angle is half that given by the Mohr's circle.
We have now seen how it is possible to consider an arbitrary
plane inside a powder compact, with both normal and shear stresses
acting on it, and resolve it into two principal planes. A Mohr's circle
represents shear stress (on the y axis) and normal stress (on the x
axis). Hence, we can draw the shear condition of this arbitrary plane
on these axes and obtain the equivalent minimum and maximum
principal plane values. This is of interest if we know the stress
condition that will cause the powder compact to break, or fracture. It
is argued that the stress of interest is the maximum principal stress
(σx): this represents the maximum stress consolidating the powder,
i.e. giving it strength. Also of interest, for the purpose of breaking the
powder compact, is the unconfined yield stress – again a maximum
principal plane stress. The latter is obtained from the Mohr's circle
106 Powder flow and storage
that must also go through the origin. Before we see this in practice we
should consider the Mohr's circle a bit further.
If we increase the normal, or shear stress, on a powder the Mohr’s
circle representing it will become bigger, see Figure 10.13. If we know
where the yield locus lies, then the circle can only become bigger
until it touches the yield locus. Note that if we treat powder compacts
as Coulomb solids the yield locus cuts the shear stress axis (i.e. does not
go through the origin) and usually has a small positive gradient. For
a given powder compact the maximum principal stress can be
obtained by drawing a Mohr circle that is tangential to the yield locus
at the upper end. The corresponding unconfined yield stress comes
from drawing a Mohr circle again tangential to the yield locus, but
Fig. 10.13 Mohr’s circle
with the minor principal plane stress going through the origin (i.e.
with yield locus – when a
circle touches the yield
unconfined with zero stress acting).
locus failure occurs However, the Coulomb solid formed by the powder compact will
have properties that depend upon how it has been treated. The
greater the original consolidating load, the larger will be the yield
locus. Hence, a series of yield loci will exist for a given powder,
dependent upon the consolidation conditions used to form the
compact. This series of yield loci can be used to provide a relation
between the strength of the powder to resist breakage and the
consolidating conditions used to form the compact; this is the
material, or powder, flow function.
Fig. 10.14 Cohesion and In a Coulomb solid there is a limit to the range of stresses that will
friction in a Coulomb solid cause no permanent deformation. A stress equal to the limit causes
plastic flow, see Figure 10.14, where Cy is similar to a yield stress and
is called the cohesion
τ = σ tan θ + C y (10.22)
Fig. 10.19 Jenike design chart for 10.5 The hopper flow factor and hopper design
a powder with an effective To determine the Hopper Flow Factor (HFF) the following procedure is
angle of internal friction of 50o used: the appropriate Jenike design chart is identified using the
effective angle of internal friction. The angle of wall friction is
obtained from tests described later and the hopper slope from the
vertical is either measured, for an existing hopper, or selected for a
new design. The value selected is based on the knowledge that the
design chart, an example is illustrated in Figure 10.19, is split into two
regions. The bottom left region is for a mass flow hopper and the top
right will result in a core flow hopper. Thus, a value of the hopper
slope from the vertical that rests on the dividing line between these
regions is bordering on a mass flow design. For the sake of safety,
Fig. 10.20 Test for flow by
assuming that a mass flow design is required, it is usual to come 3
comparing the PFF and the HFF
degrees back towards the vertical and into the mass flow regime from
this line. On the design chart, the hopper ratio between compacting
stress and applied shear stress (ff) is estimated, and the HFF is plotted
on the same graph as the PFF, as a line with gradient of 1/ff, Figure
10.20.
To test for a stable arch in the hopper the following logic is
applied. Where the PFF lies above the HFF the powder has greater
strength to resist shear and collapse than the hopper/powder system
can provide. A stable arch is, therefore, possible. When the HFF is
above the PFF the system has sufficient stress to break an arch and
reliable flow should exist. The point at which the PFF and MFF meet
gives the critical unconfined yield stress (fc -critical) and this can be used to
determine the minimum hopper opening, from equation (10.13)
f c −critical H (θ)
B= (10.23)
ρb g
Fig. 10.21 H(θ) for types of where H(θ) is a constant dependent upon the geometry of the hopper
hopper opening where L is opening. The factor H(θ) arose in the derivation of the stable arch
one rectangular dimension where the arch was assumed to act over a linear distance L, which
and B the other: if L<3B use acts up to the centre of the hopper. The arch forms above the opening
the square curve but the linear dimension measured, or deduced, is the hopper
opening itself. This is not the same as L, or even 2 L, but bears some
relation to it. Jenike provided another chart, Figure 10.21, for this
relation, dependent upon the opening geometry and the hopper slope
from the vertical.
Fundamentals of Particle Technology 109
σ 11 τ 12 τ 13
τ
21 σ 22 τ 23 (10.24)
τ 31 τ 32 σ 33
10.7 Summary
The hopper design procedure is summarised as follows: when
considering a powder for the first time a series of yield loci can be
Fig. 10.27 Three dimensional obtained by laboratory measurements and the Powder Flow Function
representation on Mohr’s circles (PFF) deduced. The effective internal angle of friction is also obtained
and the appropriate Jenike design chart identified. Tests also give the
wall friction. For mass flow conditions (no arch and uniform flow),
the operating region should be in the bottom left section of the Jenike
design chart. For the sake of safety, it is usual to come three degrees
back into the mass flow regime, i.e. reduce the hopper angle from the
vertical axis. This is, therefore, the hopper half angle. The hopper
flow factor is also deduced from this chart. A plot of the hopper flow
factor and PFF can be used to deduce the critical unconfined yield
stress and, therefore, the minimum hopper opening from equation
(10.23). Hence, the hopper is now specified in terms of angle and
opening diameter; storage capacity per silo can be deduced from the
geometry and the maximum discharge rate can be estimated from the
empirical equation (10.2), but this neglects coherency and should only
be used as a rough estimate.
Fig. 10.28 Hopper fastener The powder flow function is useful for hopper design, but is also a
and discharge arrangements
property of the powder and can be used as a characterisation
to avoid arch formation
technique in itself. For example, it may be used to correlate some
behaviour within a process, or from a product, with the PFF.
Fundamentals of Particle Technology 111
However, changes with the PFF dependent upon humidity and time
should be considered. See:
Apart from powder storage in silos, other common powder www.jenike.com
containers include an Intermediate Bulk Container (IBC), which is a for more information on
large sack that can hold over 1 tonne of powder. For transportation of hopper design
amounts in excess of IBC values trucks and wagons are needed.
10.8 Problems
Explain and answer the following terms:
Dilatancy Segregation Shear stress
Critical Porosity Normal stress Principal stress
angle of internal friction angle of wall friction
effective yield locus powder flow function
hopper flow factor unconfined yield stress
Derive the powder flow function for the powder. If the material is
hygroscopic, outline a series of tests to check on the importance of
changes in humidity on the hopper design.
112 Powder flow and storage
The following results were obtained using a Jenike shear cell with
time consolidation of 0, 7 and 14 days. Obtain the theoretical
minimum opening for a cylindrical silo of hopper half angle 30o, H(θ)
of 2.2 and angle of wall friction of 15o. Use a powder bulk density of
1800 kg m−3.
Efforts to relate the increase in the surface area to the energy required Noise can be used to
for the crushing have met with very limited success. The theoretical control the mill - an
energy required in comminution to create a new surface is often less empty mill 'rings'.
than 1% of the required amount – a lot of energy goes into effects
other than the new surface itself. During crushing, energy is used by
Elastic deformation
the processes of: elastic deformation (not including breakage) of the returns to the
particles, inelastic deformation, elastic deformation of equipment, previous shape but an
friction between particles, friction between particles and machinery, inelastic particles
noise, heat and vibration, and friction losses in equipment drives. bends out of shape or
Together with the energy required to form a new surface. flattens like putty.
where b11 was defined in Figure 11.6. Likewise, the mass flows
entering all the other grades will be
p 2 = b 21 f1 ; p 3 = b 31 f1 ; p 4 = b 41 f1 ; p 5 = b 51 f1
The full mass flow rates for the mill product, and a schematic
representation of it, are illustrated in Figure 11.7. It is possible to
represent this material balance in matrix form, as shown below
p1 b11 0 0 0 0 f1
p b b22 0 0 0 f 2
2 21
p 3 = b31 b32 b33 0 0 f3 (11.8)
p 4 b41 b42 b43 b44 0 f 4
p 5 b51 b52 b53 b54 b55 f 5
Fig. 11.7 Mill product and The matrices may be represented as follows
feed material balance
{p} = M{f} (11.9)
where M is the milling matrix. Figure 11.7 illustrates a single pass of
material through the mill and it is highly unusual that such a single
pass would produce the desired mill product quality. It is common to
recycle mill product back to the mill for a second, or more, chance of
breakage. Thus, Figure 11.7 represents what is called open circuit
grinding. In order to return oversize material back to the mill, but
remove undersize material that is our product, a classifier is required.
A classifier is schematically represented in Figure 11.8 and the
Fig. 11.8 Schematic classifier matrix is shown below
representation of a classifier
c11 0 0 0 0
0 c 22 0 0 0
C=0 0 c33 0 0 (11.10)
0 0 0 c 44 0
0 0 0 0 c 55
It is usual to define grade efficiency for a classifier as being the fraction,
by mass, of material entering the fine cut of the classifier. During
classification no alteration of particle size takes place, it simply sorts
the particles within a grade: some entering the coarse cut and some
the fine. Hence, the classifier matrix is a leading diagonal matrix of
Fig. 11.9 Closed circuit fractional values, where the top left element should be a low value
grinding
and the bottom right element should be approaching unity.
The mill and classifier, with recycle, constitute closed circuit
grinding and are illustrated in Figure 11.9. At steady state, the mass
fed to the circuit {f} must be in balance with the circuit product {q}.
The mass flow rate of material being returned to the mill by the
classifier is large and the circulating load is defined as the ratio of the
amount returned {p-q} to the amount fed {p} or {q}. Typical values of
circulating loads are in the 100’s of percent. Clearly, the circuit
product is the classifier matrix operating on the mill product
{q}=C{p} (11.11)
Using the labelling shown in Figure 11.9, the mill product is
Fundamentals of Particle Technology 119
{p}=M{k} (11.12)
and the recycle is
{p-q}=(I-C){p} (11.13)
where I is the identity matrix. Considering the mass flow entering the
mill, which is the summation of the feed to the circuit and the recycle
{k}={f} + (I-C){p}={f} + (I-C)M{k}
and rearranging for {k} using matrix algebra
{k}=[I-(I-C)M]-1{f} (11.14)
It is possible to substitute equation (11.14) into equation (11.12), to
provide the mill product and then into equation (11.11) to provide the
circuit product.
The practical possibilities represented by equation (11.14) are
significant. Provided the operating matrices remain constant, i.e. C
and M do not change under different loadings, then using equations
(11.11) to (11.14) provides a means to determine what the effect of a
change in the feed condition will be on all the mass flow rates
throughout the rest of the circuit, including the circuit product. In
order to use equation (11.14) the feed column vector {f} must be
known, but this is likely to be the case. It represents the feed mass
flow rates in the size grades selected: 1000 to 800 µm, etc. in the
example quoted. Hence, simulation of the milling circuit under
different loading conditions can be provided: covering both a simple
increase in the overall feed flow rate into the circuit and the influence
of a change in size distribution of the feed. Also, it is possible to check
the change in energy requirement by the mill using the matrix
modelling: e.g. if Rittinger’s constant in equation (11.3) is known,
then the specific surface of the feed and mill product may be
calculated from the mass flow rates in these streams by dividing the
mass flow in each grade by the total mass flow, to give the mass
fraction, and then using equation (2.17) for specific surface. More
complicated mill circuits can be modelled in a similar way, the
objective is always to relate the column vectors within the circuit to
the feed column vector, using only the operating matrices that are
assumed to remain unchanged under the different operating
conditions.
∫
B(V ) = b(V1 , V 2 ) D(V1 )dV (11.16)
where the breakage function has been modified for particle volume
rather than diameter. Equation (11.15) can be introduced into (11.16)
to provide an expression for the births in terms of both the particle
breakage and selection functions. Numerical values may be assigned
to the selection function based on the assumption that it is
proportional to the particle volume.
The birth and death rates may be combined with the appropriate
flow particle equations for the system under consideration; e.g. once
through or recycle, in order to conduct a complete material balance
based on the usual input-output=accumulation.
11.6 Summary
The production of particles by crushing and grinding is a very
Grinding media important operation throughout the world. In most cases mechanical
Common grinding media classification is also required, to return the oversize material to the
such as balls in a ball mill, or
mill. Common classification equipment includes sieves, or when used
attrition mill, include:
on an industrial scale these are called screens. In some cases decks of
alumina (ceramic),
screens may be used to fractionate the product: the largest screen size
carborundum,
on the top deck. Slotted screens are often used to minimise screen
silica,
blinding, or blockage. Alternative classifier designs include cyclones
zirconia,
and hydrocyclones for dry and wet milling operations respectively,
nylon, and
steel.
and wet classifiers using screws and rakes to lift the coarser particles
to one end, whereas the finer particles leave with the liquid flow at
the other. In some cases the screen can be bolted onto the mill itself;
Fundamentals of Particle Technology 121
for example, on a ball mill the feed enters at one end and the product
may leave over an attached screen, called a trommel, at the other.
Grates, or screens, may be used within the mill to retain the grinding
media and some of the oversize, see Figure 11.11.
No mechanical classification will be perfect, normally the
separation curve, see Chapter 14, is very shallow. Hence, a significant
amount of fine material is usually recycled to the mill and oversize
material may be found in the circuit product. For critical applications
it may be necessary to classify the product for a second, or more,
times to remove oversize. One of the most fundamental decisions to
be made is whether to wet or dry mill a material. In most cases a dry
product is required, but dry milling very fine material (<20 µm) is not
efficient. The dry powder tends to coat the grinding media and
vessel. Wet milling provides a viscous medium that absorbs some of
the grinding energy, but it does help to disperse the fine particles so
that they can receive impact, shear and abrasion. However, the need
for wet milling, or otherwise, is a material dependent property and
should be considered for each case individually.
11.7 Problems
1. Rock is crushed in a cone crusher. The feed is a nearly uniform 50
mm sphere. The screen analysis of the product is given below. The
power required to crush this material is 0.429 MW/tonne, of this
11.18 kW is required to drive the empty mill. By reducing the
clearance between the crushing head and the cone, the screen
analysis of the product becomes finer, i.e. second grind product.
a) Using Rittingers Law, what do you estimate the power Rittingers Law
hint: use energy per unit
requirement for the second grind to be?
mass is proportional to
change in specific surface
b) Using Bond’s method, estimate the power necessary for each
grind, refer to Table 11.1 for a suitable Work Index.
122 Crushing and classification
PQ σ Q2
2
σR2 = P wQ 1 + + Qw 1 + σ P (12.3)
P
M wQ 2 wP 2
where WP and WQ are the mean mass by weight of particles in the two
components. i.e.
π x - max
WP = ρ s ∫ x 3 n 3 ( x ) dx (12.6)
6 x - min
σ 2 −σ R2
M = (12.7)
σ o2 −σ R2
12.3 Equipment
It is possible to obtain a good mixer quality, but a poor product
See: quality as there are other operations between mixing and use, e.g.
www.midlandit.co.uk/part emptying, transportation and use, all of which may induce
icletechnology/chapter12 segregation. In a process analysis the permissible variation in the
for further information on product quality must be known and its relation to quality from the
mixer types and links to
mixer deduced. Batch mixing large quantities (up to 2 tonnes)
suppliers web sites
reduces labour costs but as size goes up so does time to reach desired
quality, filling and emptying times per batch. Continuous mixing
depends on metering rates, capacity of mixer, axial and radial
dispersion performance. Loading two, or more, components together
should reduce batch mixing time, but requires metering or practice. A
brief equipment type description follows.
Rotating shapes: tumbling action that induces particles to roll and fall,
material is elevated beyond its angle of repose and it falls to the free
surface. Good mixing of free flowing materials (tea, seeds, etc.). Easy
cleaning and emptying and power consumption and wear are low to
moderate.
Ribbon blade: an agitator mixes material in a trough. Reasonably
gentle mixing but with shear and impaction. Not suitable for very
cohesive materials, unless a dough is required. Used for addition of
small amounts to larger components, but can be difficult to clean.
Orbiting screw: charge is lifted up through mixer and spread out. Can
be modified for paste use. Moderate power consumption and
reasonable to clean.
Pan mixer: blades or ploughs move through mixture with various
angles of attack. Mainly found in food and pharmaceutical use, easy
to clean but has a high power consumption.
Z-blade: two contra-rotating blades, seldom used for dry materials
Fig. 12.6 Blending by ideal for a dough. Very high power requirement and difficult to
metering correct proportions clean.
of components from storage Muller or edge runner mill: crushes aggregates and mixes well, but not
hoppers used for free flowing or too cohesive mixtures – just the right amount
of friction is needed to resist the rollers.
High speed impeller: material hits impeller and is thrown out towards
the wall. Very high power consumption, but rapid mixing.
Fundamentals of Particle Technology 129
Others include fluidised beds and air lifts, blending from hopper
(bunker) discharges and stockpile blending. Continuous mixing of
free flowing powders includes metering the solids out at a
continuous and well defined rate, such as on a conveyor belt, see
Figure 12.6. This avoids the problem of mixing, but requires precise
powder control. This is only sometimes achievable in practice. Also,
this arrangements assumes that it is acceptable to form the product
from two separate streams; e.g. mixing will take place in the bag or
by the consumer. Metering, together with continuous mixing, has
become accepted for cohesive powders, where segregation is less Table 12.1 Comparison of
likely and is sometimes used with free flowing ones. Table 12.1 continuous and batch mixing
compares batch and continuous mixing strategies. of free flowing powders:
italicised description
12.4 Cohesive powder mixing indicates preferred option
12.5 Summary
Even when fully mixed we know that there will be a difference
between samples, or products, made from a powder. This variation is
given by Lacy's or Poole, Taylor and Walls' equations. If this variation
is too great for our specification we must either supply components
separately or decrease the size of the particles to give a more intimate
mix. This may lead to powder handling problems and a strategy of
size reduction, followed by mixing, then granulation before further
processing may be required. In general, mixing is easier when
particle size and other physical properties are the same. When a small
amount of one component is mixed with a large amount of another,
we may need to grind finely and coat larger particles.
130 Solid/solid mixing
12.6 Problems
1. According to a British Pharmacopia standard, the allowable
Consult equations variation in composition of an active ingredient is ±10%. A tablet
(12.1) and (12.5), contains 10% active ingredient and 90% inert diluent. Calculate the
but you will need to minimum number of particles to each tablet so that it may be within
convert between the B.P. specification (assume the standard is based on number):
number and mass a) assume that the particles are all the same size, and
proportions. See
b) assume that the particles are size distributed so that the mean
discussion around
weight of a particle of the active ingredient is 300 µg, and the other
Figure 12.3 for a
discussion on the ingredient is 100 µg.
number out of spec c) Out of a batch of 1 million tablets how many will be outside the
tablets. B.P. specification. How could you improve the situation?
not possible to shear the ions off the particle that are closer to it than
the shear plane. Thus, if the particle surface potential is measured
using a device involving motion, the potential will be measured at the
shear plane, and not the true potential on the particle. This is called the
Zeta potential. So, a particle in an electrolyte solution contains a
strongly bound layer and a diffuse layer of ions close to its surface.
This is known as the Double layer. It is worth noting that ions are
hydrated, often by several solvent (usually water) molecules, so the
distance of these layers is not that of the ions alone, but is a distance
dependent on the hydrated ion. The degree of hydration depends
upon the ionic strength of the solution (I), which is defined as follows
1 2
I= ∑ ci z i (13.1)
2
where ci is the molar concentration of an ion and zi is the ionic charge,
Dimensions or valency; i.e. the ionic strength is based on the sum of contributions
There is some confusion in for all ions in solution. The electrical potential at the particle surface
colloid chemistry with units (ψo) can be deduced from the Nernst equation, based on the
and dimensions. The concentration (or more correctly solution activity) difference between
recommended system for the particle surface and bulk solution. The potential at the Stern layer
equation (13.3) is: (ψd) will be less, and can even have a different charge value (i.e.
−
ci in number of ions m 3, negative or positive) to the potential at the surface, if the counter ions
−19 bound to the surface are polyvalent. The diffuse layer potential (ψ)
e is 1.6x10 Coulombs
− diminishes exponentially, as first derived by Gouy-Chapman, giving
εo is 8.85x10 12 Farad m-1
εd is typically 81 for water ψ = ψ d exp(− Κ∆z ) (13.2)
− −
kB is 1.38x10 23 J K 1
where ∆z is distance from the surface and Κ is the Debye-Hückel
so
function. For a simple electrolyte consisting of a cation of
ci is 1000 x Molarity x
concentration c1 and anion of concentration c2, this function is
Avagadro’s number
1/ 2
c z 2e 2 c2 z 2 2e 2
Κ = 1 1 + (13.3)
ε p k BT ε p k BT
where e is the charge on an electron, no is the concentration of ions
Potential and energy present, εp is the electrical permittivity, kB is the Boltzmann’s constant
The potential is defined as and T is the temperature in degrees absolute. However, in SI units it is
work done in moving a unit usual to calculate the permittivity from the dimensionless permittivity
charge of the same sign as
(εd) and the permittivity of a vacuum (εo)
the surface from infinity to
that point; i.e. if a charge of ε p = 4πε d ε o (13.4)
ze is moved from infinity to a
The inverse Debye-Hückel function has the units of length and it is
point within the electric field
convenient to assume that this is the diffuse layer thickness for
surrounding the particle the
practical use. However, the real diffuse layer extends to infinity.
work done is
zeψ Equations (13.2) to (13.4) can be used to predict the electrical potential
with separation distance from a particle surface. Clearly, on bringing
which is the potential energy
two particles together the electric fields will overlap and this could
possessed by the charge at
that position.
cause the particles to repel each other. This is known as electrical
double layer repulsion.
At the same time, and if the particles are in close enough proximity,
the London-van der Waal’s attractive force will tend to pull the
Fundamentals of Particle Technology 133
curve 3, the primary well is close to the particle surface and from
equation (13.14) the force is attractive (positive gradient) until the
approaching particle sits in the primary well, which is at a low enough
separation distance for settling as an aggregate to take place. Thus, in
case 3 the colloid in unstable as the two particles will settle as an
aggregate. For curve 2, there is a mildly attractive force until the
particle sits in the secondary minimum. In order to approach further
the particles must have sufficient energy to overcome the maximum –
and to overcome the repulsive force at separation distances less than
this maximum. Once over the secondary maximum there is a strong
force pulling the particle into the primary well. Hence, this situation
could be a stable colloid (because of the repulsive force preventing the Fig. 13.4 Clays in freshwater
approaching particle from reaching the primary minimum), or it may and estuary water salinities
be unstable because the maximum to be overcome is fairly low in and stability – conditions:
height. In curve 1 the maximum is very significant and the force, Zeta potential -30 mV
equation (13.14), is strongly repulsive: this is a very stable colloid. particle size 4 µm
−20
One practical application of this theory is to explain the occurrence Hamaker water 5x10 J
−
of silting within estuaries. Colloidal clay is carried down a river in Hamaker particle 6x10 20 J
stable suspension because rivers are freshwater and have a low ionic ion valancies 1
concentration, the electrical double layer extends to a large distance temperature 298 K
around the particles providing a strong repulsion force. This is dimensionless dielectric 81
illustrated in Figure 13.4, the 0.01 M NaCl curve. However, on meeting also:
the saline seawater the ionic concentration increases, and may reach permittivity of a vacuum
− −
0.3 M, which reduces the distance that the electrical repulsion acts 8.85x10 12 F m 1
over, equation (13.3), and the particles can approach each other. This is Boltzmann constant
− −
illustrated as the 0.3 M NaCl curve on Figure 13.4. The colloidal 1.38x10 23 J K 1
particles, therefore, aggregate together and settle out into the estuary electron charge
−19
causing silting. 1.6x10 Coulombs
The DLVO theory is a useful approach to explaining observed Avagadro’s number 6x1023
phenomena, but many of the equations are difficult to apply and have
different forms depending upon the assumptions made in their
derivation. Also, colloidal stability, or instability, may be due to
adsorption of solutes and ions onto the particle surface. So, the above
description is not a complete model of colloidal behaviour.
13.3 Coagulation
In the silting of estuaries example given above particles were brought
together to form an aggregate that is easier to remove from suspension
than the primary particles. This is an example of the process of
coagulation, and the resulting aggregate is called a coagula which is
formed by adding a coagulant (usually a metal salt) in solution. The
process of coagulation forms coagula that may be broken by the
application of shear, such as during pumping, but they will usually
reform when the shear field is removed.
Consideration of equation (13.3) results in the conclusion that the
distance over which the electrical double layer repulsion acts is
Fig. 13.5 Zeta potential of
strongly dependent on the valency of the electrolytes in solution. Thus, iron oxide colloid particles
a tertiary valent salt is more effective at causing coagulation than a
136 Colloids and agglomeration
13.4 Flocculation
Another means by which primary particles are brought together to
form aggregates, that is easier to remove from suspension than
primary particles, is by flocculation, see Figure 13.6. In this process the
aggregation is caused by bridges from one particle to another, usually
formed by high molecular weight polymers in solution. The polymers
may be man-made, or they can be naturally occurring; examples of the
latter are found with several biological compounds. The resulting
Fig. 13.6 A floc is much
aggregate is called a floc which is formed by adding a flocculant
bigger than a particle and is
(usually a polymer) in solution. The process of flocculation forms flocs
easier to remove from
suspension, but a floc
that may be broken by the application of shear, such as during
network settles slowly pumping, and the physical bridges linking the particles do not usually
reform when the shear field is removed. Thus, unlike coagulation, the
process of flocculation is shear sensitive. So, when flocculants are used
in high shear applications, such as to assist dewatering in a scroll
discharge decanter centrifuge (Figure 8.3) a special shear resistant
flocculant is required. These are normally very high molecular weight
polymers. However, the higher the molecular weight the more difficult
it is to dissolve the polymer and it may be more difficult to disperse
the flocculant within the feed suspension to be treated.
Synthetic polymer flocculants (polyelectrolyte) are supplied in three
forms: non-ionic, anionic and cationic. The first has no net charge, the
second is negatively charged in solution and the latter is positively
charged. Hence, when treating mineral suspensions cationic
flocculants are preferred because they can more easily bind on to the
negatively charged mineral surface. For similar reasons anionic
flocculants are applied to biological suspensions. There are many
instances when flocculation follows a primary coagulation stage. The
coagulation is designed to destabilise the suspension and the
flocculation stage to then form aggregates that are even easier to settle,
Fig. 13.7 Floc bed clarifier filter, etc. Hence, anionic polymers may be applied to mineral systems
after coagulation and cationic to biological systems. So, the optimum
flocculant and coagulant dose, and strategy, is one that can only be
Fundamentals of Particle Technology 137
The surface tension force, or moisture force, arises from the well-
known capillary pressure effect (see Figure 13.9). There is a contact
angle (θ), which is the equilibrium angle of a liquid on a surface in the
presence of another phase such as air (i.e. three phases are required:
two fluids and a solid); e.g. as with the meniscus illustrated here. The
Young-Laplace equation is
2γ cos θ
∆P = (13.16)
R
where γ is the surface tension of the liquid and R is the radius of the
capillary. The height of rise of the fluid (h) to balance this pressure is
2γ cos θ
h= (13.17)
R ρg
From a consideration of the Young-Laplace equation Newitt and
Conway-Jones (1958, Trans. IChemE) deduced the following equation
for the force due to the pendular liquid between particles:
2πγr
Fpend = (13.18)
1 + tan(θ / 2)
where r is particle radius. However, the angle formed by the meniscus
will change as liquid is lost from the agglomerate (it dries out). When
θ→0 the particle diameter with the ratio of force divided by particle
weight of one, is 4200 µm. Hence, this force is more significant than
van-der Waal’s, or electrostatic, forces for bringing particles together in
the agglomeration purposes; i.e. for particle diameters less than 4 mm
the moisture force is greater than the force due to gravity. However,
equation (13.18) is for pendular liquid bridges between the particles,
other bridges are not so strong, so some consideration of the states of
dehydration within a particle assemblage is required, see Figure 13.10.
Initially, as a droplet, the strength of the agglomerate is due only to
the liquid surface tension; this forms a weak agglomerate that may be
Fig. 13.10 Stages during the
deformed or broken easily. As some moisture is lost the capillary
formation of a granule as
agglomerate state is achieved, in which all the interstices are filled, but
liquid is removed the agglomerate strength comes from curvature of the liquid alone. As
more moisture is lost the funicular state is achieved, which still forms
only a weak structure; followed by the pendular state of agglomerate.
In the pendular state the liquid bridges are all independent of each
other and the liquid cannot move between bridges, but draws particles
together by the capillary pressure effect. It is a very strong force.
Finally, it is possible to form a dry agglomerate, possibly bound
together by van-der Waal’s or electrostatic forces but, in the absence of
solid bridges between constituent particles, the agglomerate will be
weak. However, if solid bridges have been formed, caused by
precipitation from solution, or drying of a binder, the agglomerate can
be very strong. Such dry granules may be formed when a soluble
binding agent, such as starch dries out to chemically bind the particles
together, or if the particles have been sintered together: a high
temperature causing some of the particle surface to melt and form the
bridge between particles after cooling. Strong bonds are also formed
Fundamentals of Particle Technology 139
an outlet weir) may be used for wet and dry granulation; the latter just
Capillary pressure and using the moisture associated with the powder mix.
pore size
Powder compacts are also formed by pressing and tabletting and
Equation (3.16) can be
by extrusion of pastes. The forces considered earlier in this chapter can
rearranged to give a
method to determine the again be used to assist in the understanding of these processes, but in
pore size (d) of a these instances very high pressures and, sometimes, shear fields are
membrane, cloth or porous also employed. The equipment required for tabletting, for example, is
medium: a highly specialised mould and punch system; generating considerable
4γ cos θ compaction pressure to compress and form the tablet. Usually, many
d=
∆P punches are used simultaneously, to provide high throughput.
Hence, immersing a Another important consideration in the pharmaceutical and food
membrane, in a filter related industries is an ability to comprehensively clean the
holder, within a liquid and machinery, possibly using a Cleaning In Place (CIP) strategy. In many
slowly increasing air instances it may be possible to clean the granulator and additional
pressure in the holder until processing equipment at the same time. However, in highly regulated
air flow is observed can be
production environments it is usual for agreed procedures and
used to determine the
biggest pore size from the
protocols to be in place, under the remit of Good Manufacturing Practice
above equation. This is (GMP), which might restrict the alteration of existing operating
known as the bubble point strategies, or adoption of new equipment, or techniques. Thus, when
test. considering granulation equipment for these industries both GMP and
CIP are factors that need to be considered carefully.
13.7 Summary
Forces on particles, in both gases and liquids, exert important
influences on their behaviour in processes and processing equipment.
Particles less than 10 µm in diameter are considerably influenced by
the colloidal forces described in this chapter and, towards the lower
end of this size range, the weight force may become insignificant in
determining the particle’s behaviour. In wastewater and potable water
treatment, coagulation and flocculation are extensively used processes
for the removal of unpalatable and dangerous finely dispersed
materials. The processing of fine particles is becoming increasingly
important, but fine particles tend to stick, arch in hoppers, cause safety
hazards and are generally difficult to process. Thus, it may be
important to create fine particles for a processing stage, such as
solid/solid mixing, but this may need to be followed by granulation in
order to provide acceptable later processing performance.
Advances in laboratory analytical equipment, such as the Atomic
Force Microscope, is leading to better direct experimental
measurement of colloidal forces, which will help to refine the
theoretical models and lead to a better understanding of particle-
particle interaction.
13.8 Problems
1. Using the data supplied with Figure 13.4, write a computer
spreadsheet to investigate the influence of salt concentration and ion
valency on the separation of two particles with the Zeta potentials, and
other physical properties, assumed to remain constant.
14 Gas cleaning
In the earlier chapters unit operations that cause particles to become
entrained in gas stream were covered; examples include fluidisation,
pneumatic conveying and grinding. In the next chapter the effect of
dust in the working environment is described. In certain workplaces
removal of particles down to very low sizes and concentrations is
essential, such as within operating theatres, during fabrication of
electronic equipment and production of pharmaceutical grade
materials. Hence, a critically important subject is the removal of
particles from gas streams. The subject is significantly different from
solid-liquid separation because the fluid medium is much less
viscous than a liquid and this influences the forces that are most
relevant to the trajectory analysis.
In many industrial processes sedimentation is the primary
mechanism for particle removal from a gas stream. Thus, gas
fluidised beds have the characteristic shape illustrated in Figure 7.2:
the increase in the bed diameter leads to a decrease in the gas velocity
above the bed and particles will fall back into the fluidised bed.
However, finer particles may still be carried over with the gas, i.e.
entrained in the gas flow, and additional particle/gas separation
equipment is required. The material presented in Chapter 5 is
appropriate to calculate the sedimentation rates of particles in gases,
this chapter covers other relevant mechanisms.
possibly contacting it. If it does so, and sticks, it will be removed from
the gas stream. All particles of a critical size starting from r=0 to r=rc
are captured, particles starting their trajectory from r=rc to r=rt will
not be collected. Clearly, rc=f(x) or more accurately a function of the
inertia provided by the particle and fluid system.
The grade efficiency is the collection efficiency for a given particle
size range and would be calculated by dividing the mass of particles
in a grade retained in the device compared to the mass entering the
dust collection equipment for the size range. In most cases, it is easier
to remove larger particles; hence, the grade efficiency curve for dust
collection is at, or approaching, 100% for large particles and lower for
finer particles. An example grade efficiency curve is provided in
Figure 14.2.
The grade efficiency considers collection of particles in as many
grades as are defined, but the overall collection efficiency is a single
value that represents the total mass retained compared to the total
mass challenging the dust removal equipment. Hence, it may be
deduced using the grade efficiency curve if the total dust mass flow
rate entering the device is known. The total mass flow rate of dust
leaving can be calculated and the overall efficiency is the total mass
Fig. 14.2 Grade efficiency retained divided by the total mass entering.
– note a different
definition from a classifier: 14.2 Collection mechanisms
here the intention is to
remove particles hence a Considering Figure 14.1, the dust collection mechanism was inertial
good grade efficiency separation. The inertia is a property of the system, and not just the
occurs with the larger particle; the significance of the direction change of the gas, viscosity
particles. Efficiency is still of the gas and velocity of the gas entraining the particle all influence
based on a size range – or the likelihood of the particle inertially separating from the gas
grade. streamline and hitting the target. Hence, a measure of the inertia of a
system needs to contain all these parameters and it is called the
Stokes Number (Stk)
xu g ρ s
Stk = (14.2)
9 µD
where it is assumed that the particle starts its trajectory at the same
speed as the gas (ug) and D is a characteristic dimension of the system.
In the case of Figure 14.1 the obvious dimension would be the fibre
diameter. The Stokes number is dimensionless and is not to be
confused with the Stokes’ settling velocity.
Another very common dust collection mechanism is diffusional
collection. This is illustrated on Figure 14.3 and is more relevant to
the capture of very small particles. The dust is subject to
bombardment by molecules of gas and, as the particle is small, the
Fig. 14.3 Particle momentum of the gas can induce particle motion (Brownian motion).
captured by diffusional
Hence, it is possible for a dust particle to be small enough to follow
mechanism on a fibre
the gas streamline around the target, i.e. not collected inertially, but it
Fundamentals of Particle Technology 143
N UpL
η =1− = 1 − exp − (14.11)
No ug H
Equation (14.11), when applied to dust collection by means of an
electrostatic precipitator, is known as the Deutch Equation.
dC 4η sα f dL
− =
C πd f (1 − α f )
Integrate from C=Co at L=0 to C=C at L=L to give
C 4η sα f L
η =1− = 1 − exp − (14.12)
Co πd f (1 − α f )
note single target efficiency varies according to Figure 14.4 and
usually has a minimum at approx 0.4 µm.
Similarly, for a spray tower (Figure 14.8), a mass balance on dust
entering and leaving a layer of depth dz in a spray tower gives
∂N
− ug A dz = accumulation
∂z
where u g is the superficial gas velocity, A is the cross-sectional area
and N is the dust concentration (kg m−3). Using
mass flow rate x target area x efficiency of capture
per unit area (or target efficiency)
The mass flow rate per unit area is: Nu g /(1 − α s ) , the projected area of
the spray drops is: Aα s dz 3 / 2 x and the target efficiency is: η s , where
α s is the volume fraction of the liquid in the tower and x is the liquid
Fig. 14.10 A gas cyclone
droplet size.
Combining the equations for accumulation and integrating under
appropriate boundary conditions provides the following expression
for the overall efficiency (η), based on the tower height (H)
3α sη s H
η = 1 − exp − (14.13)
2(1 − α s ) x
Equation (14.13) shows that the collection efficiency for the tower
depends on the liquid droplet diameter and the target efficiency. The
latter is dependent on the dust particle diameter, but it is generally
found that the optimum liquid drop diameter is approximately the
same for all dust diameters and is about 600 µm. This is illustrated in
Figure 14.9. Rain has a diameter close to this value; thus rainwater is
efficient in removing particles suspended in air.
14.5 Summary
There are many different types of gas cleaning devices available for
both high throughput industrial use as well as small scale high purity
applications. The latter usually rely on high efficiency particulate air
(HEPA) filters that may be several centimetres thick, but posses a
very high porosity between the fibres. Deposition takes place within
the fibre matrix. For high throughput industrial use, such as mineral
processing and fossil fuel power generation, gas cyclones may be
used as a pre-treatment technique because they are maintenance free
and have a low pressure drop followed by electrostatic precipitators
or special design filters. Ceramic candle filters are becoming popular
in high temperature applications, such as foundries, as they can
withstand corrosive gases, abrasive particles and high temperatures.
It is also worthy of note that the human body provides efficient
particle removal surfaces in both the nose and upper respiratory
tract: nasal hair helps capture particles by inertial impact and
Fundamentals of Particle Technology 149
diffusion and further inertial collection takes place in the throat and
whenever the air flow changes direction before entering the lungs.
Ans: η = 1 − ...
4. Please refer to Question 3, you will need to apply the same logic:
the flow up a chimney is turbulent but we can assume that particles
are deposited at some velocity within the laminar boundary layer.
i) Using the concept of the fraction of particles removed (-dN/N) is
equal to the fraction of the volume occupied by the differential ring dr
within the slice dz, over which the capture takes place, as you did in
the previous tutorial, derive an equation in which N/No=f(ug,r,L,Q),
where Q is the axial air flow rate and L is the chimney height.
N=No...
ii) The thermophoretic velocity of a particle in a temperature gradient
Considering a chimney
may be assumed to be
of cylindrical symmetry
−0.036 dT
Up = the radial (Up) and axial
T dr (ug) velocities are
where Up is velocity in cm s−1, z is distance in cm and T is temperature dr dz
in degrees absolute. The temperature gradient at the wall of a Up = and u g =
dt dt
chimney is 315oC per cm, and the temperature of the bulk gas is
1000oC (assume T=1273 K). The drift velocity towards the wall is
(feet s−1):
a: 0.00037 b: 0.00055 c: 0.00029 d: 0.00069
iii) The chimney is 3 feet in diameter and 50 feet tall, and gas is
flowing at 500 cubic feet per minute (cfm). The inlet dust
concentration is 1 milligram per cubic foot, the outlet concentration is
(milligram per cf) - NB convert cfm to cubic feet per second first:
a: 0.984 b: 0.967 c: 0.970 d: 0.979
iv) The dust mass deposited on the chimney wall over 28 days is (g):
a: 0.66 b: 13.8 c: 330 d: 655
to (ii) and (iii) to give an equation for the rate of increase in deposit
thickness with time dr/dt=f[ug, N (at z=0.5L), ρ s , (1- ε )]:
s.g. is specific gravity dr
– or relative density:
= ...
dt
relative to water at ix) The dust s.g. and deposit porosity are 2.6 and 0.5, the deposit
−3
1000 kg m thickness after 28 days is (µm):
a: 2.9 b: 5.8 c: 11 d: 5760
5. Spray tower
i) A spray tower reduces the concentration of a dust emission from
Combining the equations 0.009 to 0.00135 g m-3, the overall collection efficiency is (%):
for accumulation and a: 85.0 b: 82.4 c: 15.0 d: 17.6
integrating under 3
ii) The spray tower uses 1.5 m of water per hour and the residence
appropriate boundary time of a drop in the tower is 30 seconds, the volume of water in the
conditions provides the
tower at any instance is (m3):
following expression for
the overall efficiency ( η )
a: 0.75 b: 1.5 c: 0.0125 d: 0.0083
iii) The tower is 15 m high and 5 m in diameter, the volumetric
3α sη s H concentration of the liquid drops is (-):
η = 1 − exp −
2(1 − α s ) x a: 0.0051 b: 0.0025 c: 4.2x10−5 d: 2.8x10−5
iv) The liquid drops are 150 µm in diameter; target efficiency is (%):
a: 1.0 b: 1.1x10−4 c: 30 d: 85
iv) How could the overall efficiency of the tower be improved?
ANS (include a sketch of efficiency against drop diameter if
necessary):
6.
Geometrically similar Air Classified (AC) fine test dust was used in a model 8 inch diameter
cyclones can be gas cyclone and the following collection efficiencies were obtained
compared or scaled when operating with an inlet velocity of 18 m s−1
using the stokes number:
x 2ug ρs Particle diameter (µm): 1 3 5 7 9 11 22
Stk =
Collection efficiency (%): 20 45 60 72 79 84 100
9 µD
i.e. collection efficiency is The specific gravity of the dust was 2.6 and the viscosity of the gas
a unique function of
was 1.8x10−5 Pa s.
Stokes number. If the
Stokes number is
An industrial cyclone of similar geometry to the model but 5 ft in
maintained constant then
diameter is to be used to remove dust of specific gravity 1.8,
the collection efficiency
will also be a constant. contained in a gas stream at 120oC - viscosity 2.2x10−5 Pa s, with an
inlet velocity of 66 feet per second. The size distribution of the
industrial dust is provided in the figure on the next page.
i) Calling the model cyclone test StkA and the industrial application
StkB fill in the missing terms in the following (using subscript A and
B where appropriate):
Fundamentals of Particle Technology 153
Stk A
= unity = /
Stk B
ii) Now using the concept of the fraction of particles removed (-dN/N)
is equal to the fraction of the volume occupied within dL by the
Modelling is again based
differential layer dr, i.e. the part over which capture takes place (as
on the assumption that
you have done in previous questions) derive an equation in which
the flow is turbulent with
N/No=f(ug,R,L,w), where L is the tube length:
particles deposited at
some velocity within the
N=No...
laminar boundary layer.
Considering a tube the
iii) An electrostatic precipitator consisting of 100 tubes each of length
axial (ug) and radial (w) 5 m and internal diameter 10 cm is operating on a power station flue
velocities are gas. The total gas flow rate is 300 m3 min−1. The axial velocity in the
dL dr precipitator is (m s−1):
ug = w=
dt dt a: 1.6 b: 0.11 c: 3.2 d: 6.4
iv) Tests show that the collection efficiency for 5 µm particles is
99.10%, the radial, or drift (w), velocity of particles of this diameter is
(m s-1):
a: 0.0003 b: 0.15 c: 0.3 d: 0.6
v) The force due to an electrostatic field is:
pEC E P x 2 / 4
where p is a constant dependent on the dielectric properties of the
system, x is the particle diameter, EC and EP are the charging and
precipitating electric field strengths. Ignoring all forces other than
hint Q7.(v)
Use the fluid drag
electrostatic and fluid drag, the drift velocity is:
expression given by pE C E P x 2 pE C E P x 12µπ 12 µπ
a: b: c: d:
Stokes 12 µπ 12 µπ pE C E P x pE C E P x 2
dP 1/ 3
V v = K ST (15.1)
d t max
Nomographs relating the vessel volume, value of KST and required
vent area, based on the above equation, are available – see The Chemical
Engineer, January, 1989, pp18 – 21 and ‘Guide to dust explosion prevention
and protection’ – Part 1, IChemE, Rugby, 1984.
regions. The body has several mechanisms for dealing with particles
including:
The cilia, which are very thin hairs up to 4 µm long that line bronchi
and trachea and catch foreign bodies in the respiratory system.
Trapped particles are covered in mucus and passed up into the throat
where they are swallowed, sneezed or spat out; N.B. cilia are
destroyed at an early stage of smoking induced lung cancer, causing
mucous to accumulate in the air passageways and lungs resulting in
smokers cough.
Phagocytes are cells that surround the particle and reduce its irritation.
They can take the particle to the bronchioles, and out by the cilia, as
above, or into the blood stream and eventually excreted, or into the
lymphatic system, and possibly back into blood.
In the UK, the threshold limit value (TLV) applies to the dust in the
respirable range and were published by the Health and Safety
Executive (HSE), but the Control of Substances Hazardous to Health,
COSHH, act 1988, introduced the Maximum Exposure Limit (MEL) and
158 Powder hazards
15.3 Summary
The head and throat regions have very effective means for trapping
dust: nasal hair and splits in the respiratory tract provide targets for
diffusional and inertial collection. Mucus is used to ensure that the
particles do not become re-entrained in the flow. Ultimately, the
particles are discharged from the body or become swallowed. This
protects the lungs from dust ingress. However, the prevention of dusty
environments by good working practices, such as fume hoods and
flow booths is an important part of health and safety provision. The
likelihood of powder deflagrations is also reduced if powders are not
allowed to remain deposited on surfaces within operating
environments.
16 Case study
In this concluding chapter the application of certain aspects of the
preceding knowledge will be illustrated. The case study is entirely
fictional, but it includes many aspects appropriate to a large number
of industries. Recently, process engineering industries have moved
towards high value, relatively low throughput and batch identifiable
compounds and this is the strategy that will be used for this example.
An organic compound is produced in a batch reaction between a
mixture of liquids. The solvent in the reaction is an alcohol. The
organic compound forms a precipitate that needs to be recovered and
cleaned of unreacted species, and the presence of residual solvent
must be minimised, before being stored as a dry powder, or granule.
The solubility of the precipitate in the reactor is temperature
sensitive; dissolving at elevated temperature. Each batch produces
approximately ½ tonne of dry powder product. The product is to be
put into a 2 tonne storage vessel and then mixed with other powder
products to provide a particulate product with total mass of 20
tonnes, per batch. The intention is to sell this product in 50 kg drums,
or kegs, as a chemical intermediate for further processing. The
specifications on the product state that it must flow easily and
reliably, and that it must not provide a serious health risk from dust
during any further processing. Given the nature of the organic
compound produced in the batch reaction, it is desirable to minimise
operator contact with the reactor product and subsequent operations.
Let’s assume that the particle size distribution of the crystallised
product has been analysed and is as shown in Figure 16.1, marked
before recrystallisation. Being a batch system, there is probably the Fig. 16.1 Comparison of two
intention of moving the product out of the reactor as quickly as size distributions provided
possible to allow for another batch, or cleaning, or a different product by initial crystallisation and
to be formed in the reactor. This would make good economic sense as after recrystallisation by
the most expensive capital item could be highly utilised. However, heating and cooling – RRB
downstream processing; i.e. unit operations following on from the parameters are: x63.2% of 8
reactor may provide a bottleneck, if fed with material that is difficult and n of 1.5, and x63.2% of 12
to process. In general, finer particles are more difficult to process than and n of 2.5, respectively, see
bigger ones and one possibility is that the original precipitate could equation (2.5)
be reheated, to dissolve the particles, then gently cooled again to
grow larger crystals. On reheating the finer particles will
preferentially dissolve, as they have a high surface area to volume
ratio. On cooling, the particles remaining in suspension will provide
sites on which the solid material, coming out of solution, can deposit
onto thus forming bigger particles. This is represented by the second
curve shown in Figure 16.1. Appropriate data for the two curves are
included in the caption. The effectiveness of this strategy can be
evaluated by considering the downstream process of filtration.
160 Case study
and 35% by mass, 0.11 to 0.27 by volume – see equation (3.16), the
settling velocities are found to vary between 1.34x10−5 and
1.1x10−5 m s−1. Thus, the time taken to settle one metre is between 21
and 25 hours! So, gravity sedimentation is too slow to provide
thickening in this case. Clearly, the low density difference between
the solid and liquid, and small particle size are factors in this
disappointing result.
As there is a wish to minimise operator contact with the product
and the need for rapid processing of the batch, one suitable method
would be to use an automated filtering centrifuge. One popular type
of machine is the inverting bag centrifuge, which can discharge its
solids contents directly into a dryer. Washing of the filter cake and
spinning dry of residual liquid within the cake is possible prior to
cake discharge. These operations will now be considered with the
feed concentration of 15% by mass.
In the operation of a filtering centrifuge with particles less than 10
µm in diameter it would be unrealistic to expect to have no filter
medium resistance. The fine particles are likely to penetrate the cloth
to some extent and cause a significant resistance. Hence, a value of
Fig. 16.4 Time required for
1x1011 m−1 will be assumed. The cake resistance and physical
a wash ratio of 4 on the
properties are as before and other data is as follows: filter basket
filtering centrifuge
height 0.22 m, filter basket diameter 0.45 m, and rotational speed of
1200 rpm. Equations (8.24) and (8.26) must be solved and a critically
important factor is the cake radius: this is limited by the geometry of
the machine and filtration must stop before the cake becomes too
great. Under the stated operating conditions, an analysis shows that
the radius starts at 0.225 m, the full radius of the device, and after 5
minutes the cake is at a radius of 0.13 m; providing a cake thickness
of 0.1 m. In order to wash unreacted chemicals from the product a
wash ratio of 4 has been selected, see Figure 8.13. Clearly, the thicker
the cake the greater the wash time will be. The wash time is provided
by equation (8.27). A reasonable wash time is provided by the 0.1 m
thick cake, which was formed by filtering for 300 seconds. The filter
has 14.2 kg of solids in it and has filtered 65.6 kg of filtrate. An
additional 15.8 kg of liquid is retained in the cake prior to the
Fig 16.5 Zeitsch’s
dewatering stage. The time taken to spin the cake as dry as possible
dewatering model – with
and the residual moisture value are obtained from equations (8.29) to
a surface tension of
(8.33). One advantage of a low cake resistance is that the dewatering
0.08 N m−1 and 10 degrees
stage is quick and very efficient at removing residual solvent within
solvent to solid surface
the cake. Figure 16.5 illustrates the speed with which the cake contact angle and a very
dewaters at a rotational speed of 3400 rpm. All other properties are as low cake resistance
provided before, or stated below the figure.
The cycle time, on a single small inverting bag centrifuge is,
therefore, 300 seconds for filtration and 500 seconds for washing.
Time is also required for speed changes and discharge; thus a total
cycle time of 1200 seconds would be anticipated, to provide 14.2 kg of
product. This equates to 43 kg per hour and 12 hours to process the
entire reactor product using a single machine. This represents best
162 Case study
in Chapter 9 suggests that a pressure drop of 0.28 bar results from the
operating conditions provided in the box, in a 63 mm diameter pipe.
The gas flow rate required for this operation is 0.05 m3 s−1, operating
at 50% excess gas over the saltation velocity. Dilute phase conveying
should help minimise granule breakup.
Assuming that the powder to be mixed with the granules has a
mean size, by mass, of 100 µm the quality of the batch mixture can be
estimated from equation (12.5). Note, this will represent the best
quality we should expect from the powders; it is possible that the
mixture quality will be worse, due to the performance of the Mean particle mass of
equipment. The mass proportion of our granules will be 0.1, i.e. 2 the granules is:
tonnes going to form an overall mixture mass of 20 tonnes, and the π
Wp = ρs x3
mean particle mass of the granules will be 5.9x10−9 kg, see box. 6
The other powder is likely to be a filler and will have a density where the density is
closer to that of an inorganic material, such as 2000 kg m−3. Thus, the 1400 kg m−3 and
mean particle mass for the other powder will be 1.0x10−9 kg. The diameter is 200 µm, but
appropriate sample mass to use in equation (12.5) is the mass of the this assumes no internal
product leaving the process, which is in 50 kg drums. Hence, the voidage of the granules.
The real density would
between drum variance of our active ingredient will be 9.7x10−12,
be slightly less, but the
giving a standard deviation of 3.1x10−6. Now, to put these values into overall mixing result will
context: using the analysis described in Chapter 12, one drum in 1000 not significantly change.
will have a proportion of the granules greater than 0.1 plus 3 times
the standard deviation, and one in 1000 will have a proportion less
than 0.1 minus 3 times standard deviation. This suggests that 998
drums out of 1000 will have masses of the granules varying from
4.9995 to 5.0005 kg. This analysis suggests that a very good mixture
quality will be achieved by batch mixing these dry powders.
However, segregation as the powder falls from the mixer into the
hopper will have to be assessed. Thus, the batch mixing theory
provides an indication of the best possible quality that can be
obtained, operating procedures and mixer type choice may reduce
the mixture quality obtained in practice.
The 20 tonne storage hopper should be designed to provide mass
flow, in accordance with Chapter 10. This will minimise segregation
during the filling of the 50 kg drums. It is not possible to provide a
design here without laboratory data obtained from a shear cell,
similar to that provided on page 112. In practice, it is likely that the
hopper will discharge directly into the drums which will be
positioned on a load cell to weigh the drum contents. An alternative
would be to discharge into a screw conveyer, which then deposits the
mixture into the drum positioned on a load cell. The latter system is
more controllable and less likely to provide a hazard to the process
operators.
The above simplified example provides some indication of how
unit operations in particle technology are related and how some of
the principles covered in the preceding pages may provide an insight,
or even an outline design, of the process.
Nomenclature
Symbols defined and used locally are not included here.
A area, m
2
−2
a acceleration, ms
Ac cloud radius around bubble, m
C solid concentration by volume fraction, -
−3
c dry cake mass per unit volume filtrate, kg m
Cd drag coefficient, -
Cf feed solid concentration by volume fraction, -
Cu underflow concentration by volume fraction, -
D vessel, or pipe, diameter, m
d pipe diameter, m
df fibre diameter, m
E efficiency, -
−2
F force, kg m s
f friction factor, -
fg gas friction factor, -
fi number of particles within an increment, -
fs solids friction factor, -
−2 −1
Gs mass flux solids, kg m s
h pressure head; i.e. equivalent to height of liquid, m
H height of channel, or vessel, m
Hi height of suspension at some time, m
Ho original height of suspension, m
2
k permeability to fluid flow, m
L length of pipe or bed, m
mi mass fraction within a size range, or increment, -
mp mass of particle, kg
−3
N mass concentration of particles, kg m
−3
No original mass concentration of particles, kg m
−1
PH Heywood settling factor, see page 50, m
3 −1
Q volume flow rate, m s
−1
QH Heywood settling factor, see page 50, ms
−1 −2
R shear stress, kg m s
−1
Rm filter medium resistance, m
r radial coordinate, m
rc radial position of cake in filtering centrifuge, m
rcr start radius for critical particle size, m
Rf recovery of flow to underflow of hydrocyclone, -
rL inner radial position of liquid in a centrifuge, m
Ro equilibrium orbit radius, m
ro radius of a centrifuge, m
rt target radius, m
Rb bubble radius, m
Fundamentals of Particle Technology 165
2 −3
Sv specific surface area per unit volume, m m
s slurry concentration by mass fraction, -
t time, s
−1
U interstitial velocity, ms
−1
u fluid velocity, ms
−1
Ub bubble velocity, ms
−1
ug gas velocity (can include entrained solids with no slip), ms
−1
Umf minimum fluidising velocity, ms
−1
Uo superficial velocity, ms
−1
Up particle velocity, ms
−1
Ur radial gas velocity, ms
−1
ur radial velocity, ms
−1
us solids velocity with slip (us < ug), ms
−1
Ut terminal settling velocity, ms
−1
Uθ angular gas velocity, ms
−1
uθ tangential velocity, ms
3
V volume of fluid, m
W channel width, m
x particle diameter, m
x mean particle diameter, m
xi mid-particle diameter within a size range, or increment, m
xSv Sauter mean diameter, m
z height, or axial distance, coordinate, m
zs stop distance, m
Greek
−1
α specific resistance of filter cake, m kg
αf packing density of fibres, -
αs volume fraction of liquid in spray tower, -
−1 −2
∆P pressure difference, or drop, kg m s
ε void fraction, voidage or porosity, -
η particle removal efficiency, -
ηs single fibre, or target, particle removal efficiency, -
−1
λ bed filtration constant, m
−1 −1
µ fluid viscosity, kg m s
−3
ρ fluid density, kg m
−3
ρb bulk density, kg m
−3
ρm mean suspension density, kg m
−3
ρs solid density, kg m
−1 −2
σ normal stress, kg m s
−1 −2
τ shear stress, kg m s
−1
ω angular velocity, s
Further Reading
As this is an introductory work, the number of references in each
chapter has been kept to an absolute minimum. Where provided, they
are given in the text. An intelligent search of the Internet will result in
excellent equipment illustrations and pictures and these have also been
kept to a minimum. However, for the purpose of further in-depth
study of the topics introduced here the following books are
recommended. All these books are currently available from well-
known Internet retailers.
i.e. within the table are values of log10 Ut/QH which correspond to the values of log10 PHx given
by the left hand column, with the second decimal place coming from the scale on the top row,
interpolation may be used for the third decimal place.
log PHx 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
–0.2 –1.780
–0.1 –1.580 –1.600 –1.620 –1.640 –1.660 –1.680 –1.700 –1.720 –1.740 –1.760
–0.0 –1.382 –1.402 –1.422 –1.442 –1.461 –1.481 –1.501 –1.521 –1.541 –1.560
0.0 –1.382 –1.362 –1.343 –1.323 –1.303 –1.283 –1.264 –1.244 –1.225 –1.205
0.1 –1.185 –1.166 –1.146 –1.126 –1.106 –1.087 –1.068 –1.048 –1.029 –1.010
0.2 –0.990 −0.971 −0.952 −0.932 −0.912 −0.893 −0.874 −0.855 −0.836 −0.817
0.3 −0.799 −0.780 −0.762 −0.743 −0.725 −0.707 −0.688 −0.670 −0.652 −0.634
0.4 −0.616 −0.598 −0.580 −0.562 −0.544 −0.527 −0.510 −0.492 −0.475 −0.457
0.5 −0.440 −0.423 −0.406 −0.389 −0.373 −0.357 −0.341 −0.325 −0.308 −0.292
0.6 −0.276 −0.260 −0.245 −0.229 –0.213 −0.198 −0.183 −0.168 −0.153 −0.138
0.7 −0.123 −0.109 −0.095 −0.080 −0.066 −0.052 −0.038 −0.024 −0.011 0.003
0.8 0.017 0.030 0.043 0.057 0.070 0.083 0.096 0.109 0.122 0.135
0.9 0.148 0.161 0.173 0.186 0.199 0.211 0.224 0.236 0.248 0.261
1.0 0.273 0.285 0.297 0.309 0.321 0.333 0.345 0.356 0.368 0.380
1.1 0.391 0.402 0.414 0.425 0.436 0.447 0.458 0.469 0.480 0.491
1.2 0.502 0.513 0.523 0.534 0.545 0.555 0.565 0.576 0.586 0.596
1.3 0.607 0.617 0.627 0.637 0.647 0.657 0.667 0.677 0.686 0.696
1.4 0.706 0.715 0.725 0.734 0.744 0.753 0.762 0.772 0.781 0.790
1.5 0.800 0.809 0.818 0.827 0.836 0.844 0.853 0.862 0.870 0.879
1.6 0.887 0.895 0.904 0.912 0.920 0.928 0.936 0.944 0.951 0.959
1.7 0.967 0.974 0.981 0.989 0.996 1.004 1.011 1.018 1.026 1.031
1.8 1.040 1.048 1.055 1.062 1.069 1.076 1.083 1.090 1.097 1.104
1.9 1.111 1.118 1.125 1.132 1.139 1.146 1.153 1.160 1.167 1.174
2.0 1.180 1.187 1.194 1.200 1.207 1.214 1.220 1.227 1.233 1.240
2.1 1.246 1.253 1.259 1.265 1.272 1.278 1.284 1.290 1.296 1.302
2.2 1.307 1.313 1.319 1.324 1.329 1.335 1.340 1.345 1.350 1.355
2.3 1.360 1.364 1.369 1.374 1.378 1.383 1.388 1.392 1.397 1.401
2.4 1.406 1.411 1.415 1.420 1.424 1.428 1.433 1.437 1.441 1.445
2.5 1.450 1.454 1.458 1.462 1.466 1.470 1.474 1.478 1.482 1.486
2.6 1.490 1.494 1.498 1.502 1.506 1.510 1.514 1.518 1.521 1.525
2.7 1.529 1.533 1.537 1.541 1.545 1.549 1.553 1.557 1.561 1.565
2.8 1.569 1.573 1.578 1.582 1.586 1.590 1.594 1.598 1.603 1.607
2.9 1.611 1.616 1.620 1.624 1.629 1.633 1.637 1.642 1.646 1.651
3.0 1.655 1.660 1.665 1.669 1.674 1.679 1.684 1.689 1.694 1.698
3.1 1.703 1.708 1.713 1.718 1.724 1.729 1.734 1.740 1.746 1.751
3.2 1.757 1.763 1.770 1.776 1.782 1.788 1.795 1.801 1.808 1.814
3.3 1.821 1.828 1.834 1.841 1.848 1.854 1.861 1.868 1.875 1.881
Fundamentals of Particle Technology 169
i.e. within the table are values of log10 PHx which correspond to the values of log10 Ut/QH given
by the left hand column, with the second decimal place coming from the scale on the top row,
interpolation may be used for the third decimal place.
log Ut/QH 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
−1.7 −0.160
−1.6 −0.110 −0.115 −0.120 −0.125 −0.130 −0.135 −0.140 −0.145 −0.150 −0.155
−1.5 −0.060 −0.065 −0.070 −0.075 −0.080 −0.085 −0.900 −0.095 −0.100 −0.105
−1.4 −0.009 −0.014 −0.019 −0.024 −0.029 −0.034 −0.040 −0.045 −0.050 −0.055
−1.3 0.041 0.036 0.031 0.026 0.021 0.016 0.011 0.006 0.001 −0.004
−1.2 0.093 0.087 0.082 0.077 0.072 0.067 0.062 0.057 0.052 0.046
−1.1 0.143 0.138 0.133 0.128 0.123 0.118 0.113 0.108 0.103 0.098
−1.0 0.195 0.190 0.185 0.179 0.174 0.169 0.164 0.159 0.154 0.148
−0.9 0.246 0.241 0.236 0.231 0.226 0.221 0.216 0.211 0.206 0.200
−0.8 0.299 0.293 0.288 0.283 0.278 0.272 0.267 0.262 0.257 0.252
−0.7 0.354 0.348 0.343 0.337 0.332 0.326 0.321 0.316 0.310 0.305
−0.6 0.409 0.404 0.398 0.392 0.387 0.382 0.376 0.370 0.364 0.359
−0.5 0.465 0.460 0.454 0.448 0.442 0.437 0.432 0.426 0.420 0.414
−0.4 0.524 0.518 0.512 0.506 0.500 0.494 0.488 0.483 0.477 0.471
−0.3 0.585 0.579 0.573 0.567 0.561 0.555 0.548 0.542 0.536 0.530
−0.2 0.649 0.642 0.636 0.629 0.623 0.616 0.610 0.604 0.597 0.591
−0.1 0.716 0.709 0.702 0.695 0.688 0.682 0.675 0.668 0.662 0.656
−0.0 0.781 0.773 0.766 0.759 0.752 0.745 0.738 0.730 0.723
0.0 0.788 0.795 0.802 0.810 0.818 0.825 0.832 0.840 0.848 0.856
0.1 0.863 0.871 0.879 0.886 0.894 0.902 0.910 0.917 0.925 0.933
0.2 0.941 0.949 0.957 0.965 0.973 0.981 0.989 0.997 1.006 1.014
0.3 1.022 1.031 1.039 1.048 1.056 1.064 1.073 1.082 1.090 1.099
0.4 1.108 1.117 1.126 1.135 1.144 1.153 1.162 1.171 1.180 1.189
0.5 1.198 1.208 1.217 1.227 1.236 1.245 1.255 1.265 1.274 1.284
0.6 1.294 1.303 1.313 1.323 1.333 1.343 1.353 1.363 1.373 1.384
0.7 1.394 1.404 1.415 1.425 1.436 1.446 1.457 1.468 1.479 1.490
0.8 1.500 1.511 1.522 1.533 1.545 1.557 1.568 1.580 1.592 1.604
0.9 1.616 1.628 1.640 1.652 1.665 1.678 1.691 1.704 1.718 1.731
1.0 1.745 1.759 1.773 1.786 1.800 1.813 1.827 1.841 1.855 1.870
1.1 1.884 1.899 1.913 1.927 1.941 1.956 1.970 1.985 2.000 2.015
1.2 2.030 2.045 2.060 2.075 2.090 2.106 2.122 2.138 2.154 2.170
1.3 2.187 2.204 2.222 2.241 2.260 2.280 2.300 2.321 2.343 2.365
1.4 2.387 2.409 2.431 2.454 2.477 2.500 2.524 2.549 2.574 2.600
1.5 2.626 2.651 2.677 2.703 2.728 2.753 2.778 2.803 2.827 2.851
1.6 2.874 2.897 2.920 2.943 2.966 2.988 3.010 3.032 3.053 3.073
1.7 3.093 3.113 3.133 3.152 3.170 3.188 3.204 3.220 3.236 3.252
1.8 3.268 3.283 3.298 3.313 3.328 3.343 3.358 3.373 3.388 3.402
Index
aerodynamic diameter 156 colloid stability 134
agglomeration 131-140 comminution 113
Andreasen pipette 6 – mechanisms 116
angle of internal friction 106-108 composite flux 59
angle of repose 99 compressible filter cake 33
angular momentum 77 compressible sediments 60
Archimedes number 48 concentration
atomic force microscopy 133 – by mass and volume 26
attritor mill 113 – characteristics 57
– polarisation 40
backflush 29, 39 cone and quartering 3
ball mill 120 confidence 127
batch flux 57 continuous powder mixing 129
Beer-Lambert law 20 conveying 91-98
between sample variance 124 core flow 101
bin 101 Coulter Counter™ 6
body feed 42 crack propagation 114
breakage function 116-118 critical concentration 59
critical trajectory 50, 78, 144
Broadbent & Callcott 11, 117 crossflow filtration 38
Brownian motion 49 crushing 113
bubble wake 71 crystallization 159
bubbling bed 69
bucket elevator 95 Darcy’s law 23
bulk density 22, 100 Davidson and Harrison 71
buoyancy 45, 61, 77 deflagration 155
densification 139
cake washing 85 dewatering 85
Camp-Hazen 50, 53 diffusional collection 142
capillary rise & pressure 137-140 dilatancy 100
Carman 25 discrete element analysis 73
centrifugal collector, gas 146 DLVO theory 133
centrifugal force 48, 77 Dodge and Metzner 62
centrifugal separation 77-90 double layer repulsion 131-135
centrifuges, sedimenting 78 downstream processing 159
chain conveyor 95 drag coefficient 46
chimney, particle deposit 151 drainage number 87
circulating load 118 drainage rate constant 87
clarification 45 dry cake mass per filtrate 32
clarifier tank 50 volume
clarifying filtration 29 dust sampling 157
classification 113, 118 dust, collection mechanism 142-146
classifier, grade efficiency 118
coagulation 135 efficiency
cohesion 106 – overall 141
cohesive powder mixing 129 – target 141, 145
Fundamentals of Particle Technology 171
sampling 2 ultrafiltration 38
saturation 87 unconfined yield stress 103-107
screw conveyor 94 underflow 80
secondary vortex 80 – withdrawal flux 59
Sedigraph™ 6 uniformity index 10
sedimentation 45-66
segregation 2, 123 van der Waal 132, 137
selection function 116 variance 125
separation span 149 – with mixing time 126
separation, liquid-liquid 79 velocity
settling basin, liquid 50 – angular 77
settling chamber, gas 146 – axial 80
shape coefficient 7 – bubble 71
sharpness of separation 142 – interstitial 21
shear cell 109 – locus of zero vertical 81
sieve 6 – limit deposit 91
sigma theory 79 – minimum fluidising 67, 68
silo 101 – saltation 92
– stress 102 – superficial 21
solids flux 57-60 – tangential 80
solids fraction 21 venturi scrubber 146
specific resistance, cake 32 vibratory conveyor 95
specific surface area 12, 74 viscosity, Newtonian 61
specification 127 voidage 21
spherical diameter 5 volume concentration 21
spherical polar coordinates 72 vortex finder 80
spray tower 145
stable arch 103 Wadell’s sphericity 5
standard deviation 125 Wakeman 87
Stern layer 131 wash ratio & volume 86
Stokes number 142
Stokes’ law 46 yield locus 106
stop distance 144 Young-Laplace equation 138
stress gradient 49
supernatant 56 Zeitsch 87, 161
surface tension force 138 Zeta potential 131-135
Svarovsky 83 zone theory 55-57