Separation Processes

ChE 4M3

© Kevin Dunn, 2013

kevin.dunn@mcmaster.ca
http://learnche.mcmaster.ca/4M3

Overall revision number: 284 (November 2013)

1
Copyright, sharing, and attribution notice
This work is licensed under the Creative Commons Attribution-ShareAlike 3.0
Unported License. To view a copy of this license, please visit
http://creativecommons.org/licenses/by-sa/3.0/

This license allows you:

I to share - to copy, distribute and transmit the work
I to adapt - but you must distribute the new result under the
same or similar license to this one
I commercialize - you are allowed to use this work for
commercial purposes
I attribution - but you must attribute the work as follows:
I “Portions of this work are the copyright of Kevin Dunn”, or
I “This work is the copyright of Kevin Dunn”
(when used without modification)

2
We appreciate:
I if you let us know about any errors in the slides
I any suggestions to improve the notes

All of the above can be done by writing to

kevin.dunn@mcmaster.ca
or anonymous messages can be sent to Kevin Dunn at

http://learnche.mcmaster.ca/feedback-questions

If reporting errors/updates, please quote the current revision number: 284

3
Sedimentation

4
Last class: Separation factor
As mentioned, we will introduce a number of important principles
we will re-use later.
Separation factor

xi,1 /xj,1
Sij =
xi,2 /xj,2

I select i and j so that Sij ≥ 1

I units of the x terms in the above equation can be mass or
mole fractions (or flows)
I any units can be used, as long as you are consistent

Based on this definition: we can see why solid-fluid separations

often have high separation factors
5
Mechanical separations

We will start with this topic

I It’s easy to understand!
I Requires only a knowledge of basic physics (e.g. 1st year
physics)
I It introduces a number of important principles we will re-use
later
I Mechanical separations remain some of the most widely used
steps in many flowsheets. Why?
I reliable units
I relatively inexpensive to maintain and operate
I we can often achieve a very high separation factor (that’s
desirable!)

6
Units we will consider in depth

Under the title of “Mechanical Separations” we will consider:

I free settling (sedimentation)
I screening of particles
I centrifuges
I cyclones
I filtration

There are also others that go in this category. Deserving a quick

mention are:
I magnetic separation
I electrostatic precipitation

7
Quick mention: Magnetic separation
I used mainly in the mineral processing industries
I high throughputs: up to 3000 kg/hour per meter of rotating
drum
I e.g. remove iron from feed
I Also used in food and drug industries at multiple stages to
ensure product integrity
[Sinnott, 4ed, v6, Ch10]

8
Quick mention: Electrostatic separators
I depends on differences in conductivity of the material
I materials passes through a high-voltage field while on a
rotating drum
I the drum is earthed
I some of the particles acquire a charge and adhere stronger to
the drum surface
I they are carried further than the other particles, creating a
split
[Sinnott, 4ed, v6, Ch10]

9
What is sedimentation?

Sugar video, http://www.youtube.com/watch?v=ZBOou6cahtw at

04:35 to 05:02

DIY:
I concrete powder in water
I drywall compound (calcium carbonate and other particles) in
water
I add vinegar to milk to make it curdle, stir, then settle

10
Definitions

Sedimentation
Removal of suspended solid particles from a fluid ( liquid
or gas) stream by gravitational settling.

Most common to use a liquid rather than gas phase.

Some semantics:
I Thickening: generally aims to increase the solids to higher
concentration; higher throughput processes

I Clarification: remove solids from a relatively dilute stream,

usually aims for complete suspended-solids removal: units are
deeper, and have provision for coagulation of feed.
Perry, 8ed, Ch 18.5

11
Where is it applied?

Most commonly:
I water treatment
I and mineral processing applications

But also chemical, pharmaceutical, nuclear, petrochemical

processes use gravity settling to resolve emulsions or other
liquid-liquid dispersions. [Svarovsky]

12
Topics we will cover

I factors that influence sedimentation

I designing a settler unit
I costs of building and operating a settler unit
I flocculation (coagulation)

13
List any factors that influence sedimentation process

I diameter of the particles

I i.e mass of particle (as long as density is constant)
I strength of gravitational field
I relative density of particle vs fluid
I density of fluid
I viscosity of fluid
I particle concentration (hindered)
I no effect: diameter of the vessel (or area)... to a point

14
List any factors that influence sedimentation process

I diameter of the particles

I i.e mass of particle (as long as density is constant)
I strength of gravitational field
I relative density of particle vs fluid
I density of fluid
I viscosity of fluid
I particle concentration (hindered)
I no effect: diameter of the vessel (or area)... to a point

14
Ideal case: momentum balance on an unhindered particle
Forces acting on a spherical particle in a fluid:

Assuming the fluid is stagnant. 15

Ideal case: momentum balance on an unhindered particle

Forces acting on a spherical particle in a fluid:

1. Gravity: a constant downward force = mg = Vp ρp g
2. Buoyancy: proportional to volume fluid displaced = Vp ρf g
3. Drag: opposes the particle’s motion (next slide)
4. Particle-particle interactions and Brownian motion:
assumed zero for now
πDp3
Vp = particle’s volume = [m3 ]
6
ρp = particle density [kg.m−3 ]
ρf = density of fluid [kg.m−3 ]
µf = fluid’s viscosity [Pa.s]
g = gravitational constant = 9.81 [m.s−2 ]
Dp = particle’s diameter [m]

16
Drag force

ρf v 2
Fdrag = CD Ap
2
where
v = relative velocity between the particle and the fluid [m.s−1 ]
Ap = projected area of particle in direction of travel [m2 ]
CD = drag coefficient (it’s assumed constant!) [−]
ρf = density of fluid (not the particle) [kg.m−3 ]

17
Estimating the drag coefficient, CD
Dp v ρ f
It’s a function of Reynolds number = Re =
µf
[Richardson and Barker, p 150-153]

1. If Re < 1
24
CD =
Re
2. If 1 < Re < 1000
24
1 + 0.15Re0.687


CD =
Re
3. If 1000 < Re < 2 × 105

CD = 0.44

4. If Re > 2 × 105
CD = 0.10

18
Drag coefficient as a function of Re

Geankoplis, 3rd p818, 4th p921

19
Momentum balance (Newton’s second law)

dv
m = Fgravity − Fbuoyancy − Fdrag = 0 at steady state
dt

ρf v 2
0 = Vp ρp g − Vp ρf g − CD Ap
2
πDp3 πDp2
Substitute Vp = and Ap = for spherical particles (Ap is
6 4
the 2-D cross-sectional area) and solve for v :

Terminal velocity of an unhindered particle

s
4 (ρp − ρf ) gDp
v= (TSV)
3CD ρf

20
Stokes’ law

Simplification of the above equation when Re < 1:

(ρp − ρf ) gDp2
v=
18µf

Confirm it for yourself: hint: use the solution for a quadratic

equation ax 2 + bx + c = 0

21
Solving the general equation for v

v = fn(CD ), but CD = fn(Re) = fn(v )

1. Assume Re < 1 (Stokes’ region)
2. Solve for v using equation on slide 81
Dp v ρf
3. Calculate Reynolds number, Re =
µf
4. Was Reynolds number region assumption true? If so: stop.
5. If not, use new Re and recalculate CD (see slide 18)
6. Repeat from step 2 to 5 until convergence

kg
Example: A particle 1mm in diameter, with density of 5000 3 is
m
settling in an unhindered environment of water. Calculate an
estimate of its terminal velocity. [ans: 27 cm/second]

22
Why is the terminal velocity so important?

Design criterion
Terminal velocity of the slowest particle is our limiting design
criterion

We will describe particle sizes soon. But for now, it is apparent

that the feed material will have small and large particles.

We are designing the unit for the smaller particles.

23
Introducing “hindered settling”

http://www.youtube.com/watch?v=E9rHSLUr3PU

Most important points:

I large particles settle faster in low concentration (free settling)
I settling interface forms (independent of particle size)
I Stokes law (free settling) doesn’t apply
I interface’s height vs time plots are formed

24
Hindered settling
Particles will not settle as perfect spheres at their terminal velocity
under a variety of conditions:
I if they are hindered by other particles
I they are non-spherical
I concentrated feeds: particles form clusters that tend to settle
faster
I concentrated feeds: modify the apparent density and viscosity
of the fluid
I upward velocity of displaced fluids
I small particles are dragged in the wake of larger particles
I ionized conditions can cause particle coagulation → larger
diameters → faster settling

Video: http://www.youtube.com/watch?v=h8n3Nt4tPXU shows

some of these issues
25
Hindered settling

For a high concentration of particles we have hindered settling.

Stokes law doesn’t apply in these cases.

How to deal with this:

1. Modify the density, viscosity and other terms in the
momentum balance (slide 20): use correction factors
2. Resort to lab tests on samples that closely match the actual
feed material
I use lab results to design the settler

Let’s see what these large-scale units look like.

26
Large scale settlers
This unit operates on a continuous basis (at/close to steady-state)

27
Settling zones during sedimentation
We can run a batch experiment and observe settling rates

[Svarovsky, 3rd ed, p 135]

28
Settling zones during sedimentation

I initial constant rate of settling is observed

I a critical point is reached: point of inflection
I slow compression of the solids after this point

At least 2 procedures in the literature to design settlers from

settling curves:
I Talmage and Fitch: tends to overdesign the area
I Coe and Clevenger: underdesign of the thickener area
[Svarovsky, 4ed, p 180]

In practice: we will rely on outside consultants and civil engineers,

most likely, to size and design the unit. Else see the references at
end for more details.

29
The effect of particle concentration

More concentrated solutions take longer to settle; sometimes see

clearer supernatants with concentrated solutions: small particles
are pulled down in wake of larger particles.

30
How can we accelerate settling?

I modify the particle shape: spherical vs needle shape (usually

not possible)
I modify the fluid viscosity and density
I not practical in most cases
I e.g. used to separate diamonds in a process called “dense
medium separation”
I raking or stirring: creates free channels for particles to settle in
I flocculation: to increase the particle’s size by coagulating
particles

31
Flocculation
MIT video on water cleaning:
http://www.youtube.com/watch?v=5uuQ77vAV U

Please take notes from the video

32
Flocculation
Small particles (around < 40µm) and some biologically active
particles will take unreasonably long times to settle, if at all.
Flocculated particles cluster together and settle at higher rates

I impossible to predict shape and hence settling rate

I used in clarifiers, where clear supernatant is desired

Flocculation can be “included” with the sedimentation step:

[Perry, 8ed, Ch22]

I important to not disrupt the flocs after contacting with

flocculant: 30 seconds to 2 minutes contact time 33
Feed area: feedwell
I aim to minimize turbulence from entry velocity
I avoid disruption to existing settling
I avoid breaking up existing flocs
I must not get clogged

34
Sludge interface experiments

Since flocculant and concentration effects cannot be derived from

theory, resort to lab settling tests.

35
Further settler terminology
A standard gravitational thickener:

[Svarovsky, 3ed, p141]

36
Unhindered settling: design principle
I Takes place when settling occurs at a constant rate,
independent of other particles.
I Use the equations derived in last class to estimate settling
velocity = v .
I Draw an imaginary horizontal layer through the settler and
observe the mass of solids passing across it per unit time, per
unit area = mass flux.
kg solids meters
I The flux of solids is ψ = C0 v , with units of 3 ·
m feed second
kg solids 1
I ψ = C0 v ·
second meters2
I ψ = mass feed rate per unit area = loading rate = flux
1
I = unit area required per given amount of mass feed rate
ψ
Note: assuming no solids leave the overflow
37
Preliminary settler area estimate
The area required under these ideal conditions:
QC0 QC0 Q
A= = =
ψ C0 v v
where
m3 feed
 
Q = volumetric feed rate
s
 
kg solids
C0 = concentration of solids in feed
m3 feed
 −1 
v = settling velocity m.s
 
kg solids 1
ψ = mass feed rate per unit area · 2
s m area

38
Example
A sample of material was settled in a graduated lab cylinder
300mm tall. The interface dropped from 500mL to 215mL on the
graduations during a 4 minute period.
1. Give a preliminary estimate of the clarifier diameter required
to treat a waste stream of 2100 L per minute. Over-design by
a factor of 2, based on the settling rate, and account for
about 7 m2 of entry area used to eliminate turbulence in the
entering stream.
2. If the feed concentration is 1.2 kg per m3 feed, what is the
loading rate? Is it within the typical thickener range of 50 to
120 kg per day per square meter? [Perry, 8ed, p22-79]
Answers:
1. Settling rate = 171 mm per 4 minutes = 42.8 mm/min.
2.1 m3 .min−1
Area = 1

= 98 + 7m2
42.8 × 10−3 m.min−1
2
kg m 60 × 24min kg
2. ψ = C0 v = 1.2 3 · 0.0428 · = 74
m min day day.m2
39
Settler design: shape, length, width

1. What width and depth should the settler be?

2. How long should the particles be in the settler? Does
residence time matter?
Perry’s, section 22.5.6:
I sedimentation tank can be rectangular or circular
I rectangular: effluent weirs at the end
I circular: around the periphery
I main concern: uniform flow in the tank (no short-circuits)
I removal efficiency = f(hydraulic flow pattern in tank)
I incoming flow must be dissipated before solids can settle
I evenly distributed; minimal disruption to existing fluid
I overflow and underflow draw collected without creating
hydraulic currents
I solids are removed by scraping, and hydraulic flow

40
Concept: the ideal rectangular settling basin

I Inlet zone: feed is assumed to be uniformly distributed across the tank’s

cross-section (if viewed from the top)
I Settling zone: where particles move downwards towards the sludge area;
particles also move horizontally due to fluid flow
I Outlet zone: the supernatant/clarified liquid is collected along the basin’s cross
section and removed in the overflow
I Sludge zone: where the solids collect and are removed in the tank’s underflow
Ensure horizontal fluid velocity (i.e. residence time) is slow enough
that particles at their terminal velocity, v , will reach the sludge
zone and settle out.
41
The ideal rectangular settling basin

Changing depth has no effect in a rectangular basin [Svarovsky, 4ed, p170]

42
Concept: the ideal circular settling basin

I Same zones as before

I Fluid’s horizontal velocity is a function of radial distance
I As before, ensure residence time is long enough for particles
to reach the sludge zone
43
Settler design rules of thumb: size

For wastewater treatment the main design criterion: solids

percentage in underflow
I A volume and mass balance on solids and liquids is then used
to find the liquid overflow rate
I surface overflow rate (SOR) ∼ 40m3 per day per m2 for
primary units
I secondary units as low as 12 up to 30m3 per day per m2
I minimum depth of sedimentation tanks is around 3.0 m
I circular sedimentation: minimum diameter of 6.0 m
I length to width ratio of 5:1

44
Settler design rules of thumb: residence time

I gravity sedimentation tanks normally provide for 2 hour

retention of solids, based on average flow
I longer times for light solids, or in winter times
I organic solids generally will not compact to more than 5 to
10%
I inorganic solids will compact up to 20 or 30%
I why important: we have to design sludge pumps to remove
the solids: high concentration solids require diaphragm pumps

45
Capital costs considerations
Svarovsky 3rd, p179: cost = ax b
I x = tank diameter between 10 and 225 ft
I a = 147 and b = 1.38 for thickeners
Perry, 8ed, section 18.6
I Installation costs will be at least 3 to 4 times the actual
equipment costs.
I Equipment items must include:
I rakes, drivehead and motors
I walkways and bridge (center pier) and railings
I pumps, piping, instrumentation and lift mechanisms
I overflow launder and feed
Installation is affected by:
I site surveying
I site preparation and excavation
I reinforcing bar placement
I backfill
46
Operating costs

These are mostly insignificant

I e.g. 60 m (200 ft) diameter thickener, torque rating = 1.0
MN.m: requires ∼ 12 kW
I due to slow rotating speed: peripheral speed is about 9 m/min
I implies low maintenance costs
I little attention from operators after start-up
I chemicals for flocculation (if required), frequently dwarfs all
other operating costs [Perry, 8ed, Ch18.6]

47
Further self-study

I Designs with peripheral inlets (submerged-orifice flow control)

and either center-weir outlets or peripheral-weir outlets
adjacent to the peripheral-inlet channel.
I Deep cone thickener
I Lamella (inclined plate or tubes): often for gas-solid
applications

48
Practice question

1. Calculate the minimum area of a circular thickener to treat

720 m3 per hour of slurry containing 65µm particles of silica,
whose density is about 2600 kg.m−3 . The particles are
suspended in water at a concentration of 0.650 kg.m−3 . Use
an over-design factor of 1.5 on the settling velocity.
[Ans: v = 3.7/1.5 mm.s−1 and A = 81.4 m2 ]
2. If it is desired to have an underflow density of 1560 kg solids
per m3 underflow; what is the underflow volumetric flow rate
if total separation of solids occurs?
[Ans: Qunder = 0.3 m3 solids.hr−1 ]
3. Calculate the separation factor.

49
References for this section
I Geankoplis, “Transport Processes and Separation Process
Principles”, 3rd or 4th edition, chapter 14
I Perry’s Chemical Engineers’ Handbook, 8th edition, chapter
18
I Richardson and Harker, “Chemical Engineering, Volume 2”,
5th edition, chapter 3 and 5
I Sinnott, “Chemical Engineering Design”, Volume 6, 4th
edition.
I Talmage and Fitch, 1955, “Determining Thickener Unit
Areas”, Ind. Eng. Chem.,47, 38-41, DOI:10.1021/ie50541a022
I Fitch, 1965, “Current theory and thickener design”, Ind. Eng.
Chem., 57, p 18-28, DOI:10.1021/ie50682a006
I Svarovsky, “Solid Liquid Separation”, 3rd or 4th edition.
Particularly thorough regarding the settler’s mechanical
accessories: pumps, scrapers, etc.
50
Particle size distribution and
characterization

51
Some context

Three important characteristics of an individual solid particle are

its composition, its size and its shape.

Composition: affects density, conductivity, and other physical

properties important to separating it.

We will consider shape and size characterization now.

52
Particle shape characterization

A particle may be regular shaped, e.g.

I spherical or cubic objects are capable of precise definition
using equations

Irregular shaped: e.g. broken glass, sand, rock, most solids

I properties of irregular shapes are expressed ito a regular
shaped particle’s characteristics

So we will spend time characterizing spherical particles, then

expressing other particles in terms of anequivalent spherical particle

53
Particle shape: Sphericity

Why use a sphere?

I it has the same shape from all angles
I behaves the same way from all angles
Other particles behave less ideally; we define sphericity as one
metric of a particle’s shape:

surface area of sphere with same volume as particle

ψ=
surface area of particle

For all particles: 0 < ψ ≤ 1

To try: calculate the sphericity of a cube with side length = c

Answer: ψ = 0.806

54
Other shape metrics

Find the diameter of a sphere that has the same as the

irregularly-shaped particle
I volume
I surface area
I surface area per unit volume
I area in the projected direction of travel [drag diameter]
I projected area, but in a position of maximum stability
I settling velocity [Stokes’ diameter]
I will fit through the same size square aperture [sieve diameter]

55
Shape metrics example

For example, how would you quantify yourself if measured by:

1. Circumference around your waist?
2. Diameter of a sphere of the same surface area as your body?
3. Length of your longest chord (height)?

The measured values will have different meanings. e.g.

1. used to size a life jacket
2. (perhaps?) used to estimate heat losses through your skin
3. if you are buying a sleeping bag I suggest the last one.
[Adapted from: George G. Chase, The University of Akron]

56
Particle size characterization

So far we have assumed particles to be separated are of a single

size. This is never true: there is always a size distribution.

Particle size: affects surface per unit volume (mass transfer), rate
of settling in a fluid (separation), etc

Aims

How do we measure this distribution?

How do we describe (characterize) a size distribution?

What is the “average” particle size?

57
Other reasons to consider particle size distributions?

I Understand your colleagues: “After crushing, the feed may be

ground in several stages from a size of 5 to 6 cm to a powder
of 75 to 90 percent passing a 200-mesh sieve.”
I Solid material handling industry: we must deal with
distributions
I What kind of industries are we referring to here?
I e.g. activity of a powdered drug = f (particle size)
I e.g. “hiding power” of a paint/pigment = f (particle size)

We will require this understanding for future sections: filtration,

flow of fluids through packed beds, cyclones, centrifuges,
membranes, and so on.

58
Some typical particle sizes
We typically work in microns. 1 micron = 1 µm

[Richardson and Harker, 5ed, p4] 59

Standard screens

Mesh 10 screen = 2.00 mm opening = 10 openings per linear inch.

60
Standard screen sizes
The US standard (Tyler series). Selected screens from Tyler sequence:
Mesh number Square aperture opening (µm)
3.5 5600 Tyler standard:
.. ..
. . I e.g. 75µm opening:
10 2000 called 200 mesh screen
.. .. I i.e. apertures per inch
. .
20 850 = 200 mesh screen
25 710
30 600
.. ..
I Successive apertures
. . decrease
√ by factor of
140 106 ∼ 42
170 90
200 75
230 63
I Other standards:
.. .. British I.M.M. and
. . U.S. A.S.T.M.
450 32 61
Screens
Stack screens: apertures from largest on top to smallest

[http://en.wikipedia.org/wiki/File:Laborsiebmaschine BMK.jpg] 62
Screens

I Select top screen to (usually) have 100% material passing

through
I Shaken for a predetermined time; or rate of screening levels
out
I Shake intensity must be balanced: not too aggressive to break
particles apart
I Smaller particles tend to stick to each other, so small size
fractions inaccurate
I One can have wet or dry screen systems
I Wet screening: washes smaller particles off larger ones

63
Data analysis from a screen sample
Mesh Aperture [µm] Mass retained [g] Avg size* Cuml. % passing
14 1400 0 - 100
16 1180 9.1 1290 98.1
18 1000 32.1 1090 91.6
20 850 39.8 925 83.5
30 600 235.4 725 35.5
40 425 89.1 513 17.4
50 300 54.4 363 6.3
70 212 22.0 256 1.8
100 150 7.2 181 0.4
140 106 1.2 128 0.1
Pan 0 0.5 53 0.0
Sum 491

∗ average screen size used for differential plots

64
Differential and Cumulative analysis

Theory:

dF (x)
f (x) =
dx F (x) = percent passing curve
x is the avg particle size (bin)
1−F (x) = percent retained curve
65
Theoretical view

[Svarovsky, 4ed, p42]

66
Mean diameter calculations
A number of mean diameters can be calculated. These can be
derived from the cumulative analysis plot:
I Arithmetic mean = 318 µm
I Volume mean diameter =
430 µm
I Surface mean diameter
(Sauter mean diameter) =
565 µm
I Weight or mass-mean
diameter = 666 µm

Our aim is not to calculate all these

(the formulae are messy, and error
prone). Your lab will have these
already set up in their analysis
software.
I Seader, 3ed, p 678 - 679 [for worked example]
I Svarovsky, 4ed, p37 - 43 [for descriptions of many means] 67
Which mean should I use?

I Rather use the distribution curve, if available

I If one has to resort to a single number, use what is
appropriate
I volume mean diameter: used for packing estimation
I surface mean diameter: used for skin friction, and mass
transfer calculations
I The idea is that if two materials had the same “mean
diameter”, that they would behave the same way in the
application being considered.

68
Two distributions, same arithmetic mean

[Svarovsky, 4ed, p59] 69

Sampling a stream

Particle size measurements are strongly dependent on the sample

taken. The “golden” rules of sampling:
1. take sample from a moving stream: dry powders and slurry
2. sample whole stream for many short periods (not part of
stream for whole time)

There are books written on the topic of sampling. Consult an

experienced person if important decisions rest on the sample taken.

70
References

1. Richardson and Harker, “Chemical Engineering, Volume 2”,

5th edition, chapter 1
2. Perry’s Chemical Engineers’ Handbook, 8th edition, chapter
21.1
3. Seader, Henley and Roper, “Separation Process Principles”,
page 675 to 679 in 3rd edition
4. Svarovsky, “Solid-Liquid Separation”, 4th edition.

71
Centrifugal separations

72
References

I Geankoplis, “Transport Processes and Separation Process

Principles”, 3rd or 4th edition, chapter 14.
I Richardson and Harker, “Chemical Engineering, Volume 2”,
5th edition, chapter 9.
I Perry’s Chemical Engineers’ Handbook, 8th edition, chapter
18.8.
I Svarovsky, “Solid Liquid Separation”, 3rd or 4th edition.
I Seader et al. “Separation Process Principles”, page 800 to
802 in 3rd edition.
I Schweitzer, “Handbook of Separation Techniques for
Chemical Engineers”, chapter 4.5.

73
Why consider centrifuges?

I When gravity (freely available) is not fast enough

I Decrease the separation time and increase separation factor
I Much smaller piece of equipment
I Achieve separations not possible by gravity:
I overcome Brownian limits
I overcome convection currents
I overcome stabilizing forces that hold an emulsion together

Why not just apply flocculation?

74
Terminology

I Suspension: the mixed

material added into the
centrifuge tube
I Pellet or precipitate:
hard-packed concentration of
particles after centrifugation
I Supernatant: clarified liquid
[http://en.wikipedia.org/wiki/File:Tabletop centrifuge.jpg]
above the precipitate
75
Uses
Used since 1700’s:
1. separate particles from fluid based on density
2. separates immiscible fluids (liquid and even gases) of different
densities
3. to enhance drainage of fluid from particles for drying
4. enhance mass transfer (look at centrifugal packed bed contactors in your
own time)

Examples:
I Cream from milk (milk is an emulsion)
I Clarification: juice, beer (yeast removal), essential oils
I Widely used in bioseparations: blood, viruses, proteins
I Remove sand and water from heavy oils

76
Interesting use: gas-gas separation
I Uranium enrichment in a Zippe-type centrifuge: U-235 is only
1.26% less dense than U-238: requires counter-current cascade

[http://en.wikipedia.org/wiki/File:Gas centrifuge cascade.jpg]

77
Principle of operation
I items being separated must have a density difference
I it is not a difference in the mass, only density
I Video of emulsion separation at high G’s
I centrifugal force acts outward direction = ma = m(r ω 2 )
I m = particle’s mass [kg]
I r = radial distance from center point [m]
dθ
I ω = angular velocity = [rad.s−1 ]
dt
I recall 2π rad.s−1 = 1Hz
I and 1 rad.s−1 ≈ 9.55 revolutions per minute [rpm]
mr ω 2 r ω2
I G= =
mg g

Example Revolutions per minute G’s

Car going round and round 10 to 15 1 to 2
Washing machine at home 1500 625 (r=0.25m)
Industrial centrifuge < 15000 25000 (r=0.1m)
Laboratory centrifuge 30,000 to 100,000 100,000 to 800,000
Zippe-type centrifuge* 90,000 ∼ 1 × 106
* tangential velocity > Mach 2 ∼ 700m.s−1 78
Laboratory centrifuges

Main selection factors:

1. duration = t [use minutes in the equation below]
2. maximum rotational speed = RPMmax
k
t= I S = Svedberg coefficient of the
S
material (from tables, experiments)
I rmax and rmin as shown above [cm]
 
e.g. S20 = 6.43 for collagen 11 ln (rmax − rmin )
I k = 2.53 × 10
RPM2max
79
Tubular bowl centrifuge
I Most commonly used for small particle separation
I Fluid and suspended solids are fed at the center
I A vertical wall of fluid is formed. Useful video to see this.
I Feed is continually added, forcing fluid out the top, over the
retaining wall. Solids accumulate inside the bowl.

[Geankoplis, Fig 14.4-2; draw top view]

80
Recall particles in a fluid: Stokes’ law
Let’s understand how the solid particles move:
Recall if Re < 1
grav Dp2 (ρp − ρf ) g
vTSV =
18µf

In a centrifuge, we have simply replaced g with a centrifugal force,

r ω 2 (gravity is negligible)

cent dr Dp2 (ρp − ρf ) r ω 2

vhoriz = =
dt 18µf

I The particle is also forced in the vertical direction of fluid flow

at a constant upward velocity, so its net trajectory is curved.
I In centrifuges: particles are likely to have Re < 1 (why?)

81
Theoretical trajectories: tubular bowl centrifuge

Integrate from t = 0 where r = r1 to the outlet, where we require

the particle to be exactly at r = r2 within a time of t = t∗ seconds:
18µf r2
t∗ = 2 2
ln
Dp (ρp − ρf ) ω r1
82
Theoretical trajectories: tubular bowl centrifuge
Consider a particle moving with too slow a horizontal velocity (e.g.
centrifuge is too slow).
I Within the time from t = 0 to t = t∗ , this particle is moving
too slowly, and will not reach the wall at r2
I This particle is then assumed to have left in the supernatant
(liquid discharge)

t = t∗ gives a bound on the time it should take a particle to reach

the wall at r2 , starting at r1 . 83
Calculating the centrifuge’s throughput, Q
Once we know how long a particle should be in the centrifuge, we
can calculate a feed flowrate,
Q. The volume of fluid in the
centrifuge is V = π r22 − r12 h. Calculate the volumetric flow rate

V Dp2 (ρp − ρf ) ω 2
π r22 − r12 h [m3 .s−1 ]


Q∗ = =
t∗ 18µf ln(r2 /r1 )

I What happens if we operate a flow rate slower/faster than

this Q∗ ?
I Alternative interpretation: for a given flow Q∗ , find the largest
particle diameter that will arrive exactly at r2 at height h.
Particles with smaller Dp are expected to leave in supernatant.
I Obviously this is excessive: we have the horizontal discharge
weir to retain particles that might not have reached r2 at
height h
I r2 remains fixed for a purchased and installed centrifuge
(design parameter)
84
Cut-size diameter
So to prevent excessive over design, we rather find the halfway
mark between r1 and r2 , and solve the same equations to find the
time, called tcut , for a particle to reach this cut point:
2
Dp,cut (ρp − ρf ) ω 2
V
π r22 − r12 h


Qcut = =
tcut 18µf ln [2r2 /(r1 + r2 )]

I we design for the cut-point volumetric flow rate Qcut

I and can then solve for the cut point diameter, Dp,cut
I all other terms in the equation are known/set
I We can also design for a given diameter, and solve for the
Qcut .

Note: We could use any reasonable point between r1 and r2 . The 50%
point is convention. It accounts for uncertainties in our flows, physical
properties and idealities assumed with Stokes’ law.
85
Example
A lab scale tubular bowl centrifuge has the following
characteristics:
I r1 = 16.5 mm and r2 = 22.2 mm
I bowl height of 115 mm
I 800 revolutions per minute
It is being used to separate bacteria from a fermentation broth
experiment.
If the broth has the following properties:
I ρf = 1010 kg.m−3 ← note how close these are
I ρp = 1040 kg.m−3
I µf = 0.001 kg.m−1 .s−1
I Dp,min = 0.7 µm ← note how small
1. How many G’s is the particle experiencing at r2 ?
2. Calculate both Q∗ and the more realistic Qcut .
3. Verify whether Stokes’ law applies.
4. What would be the area of the sedimentation vessel that
Q
would operate at this Qcut ? Hint: recall that A = .
vTSV 86
Example
1. Illustrate the trajectory taken by a particle reaching the
cut-point within time tcut
2. In the same duration of time, what trajectory will a smaller
particle have taken?

87
Sigma theory for centrifuges

Take the previous equation for Qcut , multiply numerator and

denominator by g , then substitute Stokes’ law for particles settling
under gravity:
grav (ρp − ρf ) gDp2
vTSV =
18µf
we obtain:
!
2
(ρp − ρf ) gDp,cut grav
Qcut = · (Σ) = vTSV ·Σ
18µf

ω 2 πh r22 − r12


Σ=
g ln [2r2 /(r1 + r2 )]

Σ = f (r1 , r2 , h, ω)

88
Why use the Sigma term?
I Σ = f (r1 , r2 , h, ω)
I it is only a function of the centrifuge’s characteristics; not the
particle or fluid
I Σ has units of m2 : Σ is the equivalent surface area required
for sedimentation by gravity
grav
I Centrifuge A: Qcut,A = vTSV · ΣA
grav
I Centrifuge B: Qcut,B = vTSV · ΣB
I
Qcut,A ΣA
=
Qcut,B ΣB
grav
I Used for scale-up of the same feed, i.e. the same vTSV
I Used for scale-up within the same types of equipment
I Σ equation is different for other centrifuge types
I Question: if I know ΣA for a given centrifuge and for a given
feed; can I calculate the performance, Qcut,B , for a different
feed stream?
89
More on the tubular bowl centrifuge
I Batch operation: stop to clean out solids; restart again; use
paper on wall to assist solids removal [∼ 15 min turnaround]
I Contamination possible, not always suitable for bioseparations
I A high L/D aspect ratio is used (around 8), as it is more
stable to operate
I Minimize D; very high wall stresses are developed at higher
diameters
I Can be used for fluid-fluid separation

ρH r 2 − r12
= 22
ρL r2 − r42

90
Disc bowl (disc stack) centrifuges

[Geankoplis, Fig 14.4-4]

Video to illustrate operation:

http://www.youtube.com/watch?v=YMbaBLpInrc
Another video: http://www.youtube.com/watch?v=bzXUiLajVlg 91
Disc bowl centrifuges

I Recall: Q = V /t∗ (the t∗ will be different for disc bowl

compared to tubular bowl)
I If we increase rate of fluid feed, we get higher throughput, Q
I Adding angled discs gives a greater surface area, hence
greater volume treated, without increasing bowl diameter
I Widely used in bioseparations: no contamination (aseptic)
I Also for: fish oil, fruit juice, beverage clarification
I 3-phases separation: e.g. sand, oil, water mixtures

92
Disc bowl centrifuges
I Discs angled at 35 to 50◦ ; ∼ 50 to 150 discs per unit
I Typically between 0.7 to 0.5m in radius; with rotational
speeds of 0 to 12,000 rpm
I Typically between 0.15 to 1.0m in diameter; with rotational
speeds of 0 to 12,000 rpm
I Typically used to treat up to 15% solids in feed stream
I Can be operated continuously (infrequent cleaning of discs)

2πω 2 N(r13 − r23 )

Σ=
3g tan θ

I N = number of disc plates

I θ = angle of discs
I r1 = outer cone radius
I r2 = inner cone radius
93
Scroll centrifuges

The scroll allows for continuous removal of solids:

[Perry, fig 18-159]

I Sedicanter: biotechnology, vitamin, soy, and yeast separations.

94
Scroll centrifuges

[Perry, fig 18-160]

I Sorticanter: used for plastics recycling

I General scroll centrifuges: used in oil-sands separations

95
Sequencing of centrifuges

[http://www.westfalia-separator.com/products/innovations/oil-sand-bitumen-process.html]

96
Safety

I careful selection of materials of construction: corrosion and

withstand high forces
I heat removal might be required (some units come with
integrated refrigeration)
I rotational equipment requires careful balance
I digital control is critical
I PLC: programmable logical controllers
I SCADA: supervisory control and data acquisition
I safety interlocks
I cameras are increasingly used to monitor sediment buildup:
auto-stop and clean
I flammable fluids (e.g. solvents): nitrogen blanket

97
Choosing a centrifuge unit

[Schweitzer, p 4-58]

98
Selecting a centrifuge
Based on required performance

99
Design a centrifuge for beer clarification

Design a separation plant to remove suspended yeast cells from

beer.

Beer is produced in batches of 100 m3 , with 4 batches per day.

Some data:
I Density of beer: 1020 kg.m−3
I Density of yeast cells: 1075 kg.m−3
I Yeast cell diameters: 4 to 6 µm
I 11.5 metric tonnes of yeast are suspended in each 100 m3
fermenter
I Aseptic operation is vital

100
Further practice questions

1. In a test particles of density 2800 kg.m−3 and of size 5 µm,

equivalent spherical diameter, were separated from suspension
in water fed at a volumetric throughput rate of 0.25 m3 .s−1 .
Calculate the value of the capacity factor, Σ. [Ans: Σ = 1.02 × 104 m2 ]

2. What will be the corresponding size cut for a suspension of

coal particles in oil fed at the rate of 0.04 kg.s−3 ? The density
of coal is 1300 kg.m−3 and the density of the oil is
850 kg.m−3 and its viscosity is 0.01 N.s.m−2 . [Ans: Dp,cut = 4µm]

3. Is Stokes’ law applicable? cent

[Ans: Calculate the vTSV and confirm if Re< 1]

[Richardson and Harker, v2, 5th ed, p482-483]

101
Cyclones

102
References

I Svarovsky, “Solid Liquid Separation”, 3rd or 4th edition,

chapter 6.
I Richardson and Harker, “Chemical Engineering, Volume 2”,
5th edition, chapter 1.
I Sinnot, “Chemical Engineering, Volume 6”, 4th edition,
chapter 10.
I Perry’s Chemical Engineers’ Handbook, 8th edition, chapter
17.2, “Gas-Solid Separations”
I Schweitzer, “Handbook of Separation Techniques for
Chemical Engineers”, chapter 4-135.

103
The cyclone Cyclone

Hydrocyclone
104
Uses
Wide variety of uses:
I dust removal (principal application) in many industries
I cement industry
I sawmills
I catalyst particle recovery in reactors
I mist (droplets) removed from air streams
I recovery of spray-dried particles
I separating immiscible liquids (different densities)
I dewater suspensions: concentrate the product
I remove dissolved gases from liquid stream
I solids-solids separation: very common in mining

Where possible, consider a cyclone before a centrifuge for

solid-fluid separations.

105
Alternatives
A number of alternatives exist; based on the principle of removing
the particle’s momentum relative to the fluid’s momentum. Other
options?

[Richardson and Harker, 5ed, p81]

106
Cyclone operation

[Brown and Associates, Unit Operations, p 119] 107

General path of travel in a cyclone low viscosity, low solids concentration

Generally, flow pattern is

more complex than this.

See, for example, this

video of a PET scan of a
radioactive isotope
labelled particle 18 F

I Vortex and tangential

forces formed by the
fluid

108
Principle of operation
I Same principle as a centrifuge: density difference required
I No moving parts! and no consumable components!
I Low operating costs: essentially only pay for ∆P
I Operated at many temperatures and pressures
I As small as 1 to 2cm to 10m in diameter
I Very low capital costs: can be made from many materials
I Particle sizes 5µm and higher are effectively removed
I Even different particle shapes (due to different settling
velocities) can be separated
I Forces acting on particles: between 5 (large cyclones) and
2500 G (small cyclones)

Videos:
I http://www.youtube.com/watch?v=2bUlytvimy4
I http://www.youtube.com/watch?v=GxA49uVP2Ns
I http://www.youtube.com/watch?v=BicR3JGlE5M
I http://www.youtube.com/watch?v=QfTZUMq-LGI
I and many other videos of people making their own cyclones. 109
Velocity profile: very complex
3 directions of travel:
1. LZVV = locus of zero
vertical velocity (axial l)
2. radial velocity is small
(←→)
3. tangential velocity
I vt r n = constant
I true at all heights inside
cyclone
I centrifugal force (acts −→)
I drag force (acts ←−)
I if Fcentrifugal > Fdrag
particle moves towards wall
I then pulled down in axial
stream and exits in
underflow
110
Velocity profiles
The above description is extremely simplistic; velocity profiles
cannot be theoretically derived for most practical cases.

I it is not gravity that removes the heavier particles in underflow

I it is the slower, boundary layer flow at the walls and air flow
out of the spigot
I particles rotate at a radius where centrifugal force is balanced
by drag force (recall elutriation concept)
I larger, denser particles move selectively towards the wall
I residence time must be long enough to achieve equilibrium
orbits; spiral patterns help
I all of this comes down to a careful balance of radial and
tangential velocities
I velocities: these are our degrees of freedom to adjust the
cyclone’s performance
111
Evaluating a cyclone’s performance

Mass balance: M = Mf + Mc
I both overall balance
I and within each size fraction
[Svarovsky, 3ed, p210]
112
Concept: Grade efficiency
Total efficiency
Mc Mf
ET = =1−
M M

I not too much to interpret here: it is just a definition

I 0% efficiency: all mass is being sent to overflow (fines) stream
I 100% efficiency: all mass to underflow (coarse) stream

More useful though is

Grade efficiency

(Mc )(fraction of size x in stream C, coarse stream)

G (x) =
(M)(fraction of size x in feed)

I calculated for a given particle size fraction x; repeat at all x’s

113
What is “fraction of size x” again?

Percentage area under the (differential) curve, at size fraction x.

114
Back to grade efficiency
Grade efficiency equation
(Mc )(fraction of size x in stream C, coarse stream)
G (x) =
(M)(fraction of size x in feed)

I If G (x) = 0.5 (50%): implies half the material (by mass) in

size fraction x is leaving in the underflow (coarse)
I and the other half in the overflow; 50-50 (mass) split in the
two outlets for particles of size x. Called the “cut size”, x50
I If G (x) = 1.0: implies the particle size that gets captured
100% in the coarse (underflow) stream
I Where G (x) reaches 1.0 means the x = largest particle size
we expect to ever see in overflow (see next slide)
I (advanced)What would G (x → 0) = 10% mean?
[i.e. the G (x) curves don’t always reach 0%]

115
Grade efficiency curve
Calculate efficiency at each size fraction, x, and plot it:

Which is a more desirable cyclone from a separation efficiency

perspective? In general, what shape would be the most desirable?
116
Day-to-day operation

I most important factor: pressure drop = ∆P = difference

between inlet and overflow (fines) pressures and typically we
have ∆P ∼ 500 to 1500 Pa
I increase ∆P, increases efficiency, ET , and recovery into the
coarse stream
1
I ∆P ∝ ρf ∆P ∝ vin2 and voverflow
2 ∆P ∝
dunder
I vin = entry velocity and dunder = diameter of underflow
I efficiency drops off at high solids concentration: try to operate
as dilutely as possible if requiring high solids recovery
I leave the underflow opening diameter, dunder , as an physically
adjustable variable: it is hard to predict its size from theory
I air leaks at this point are disastrous for efficiency [Perry, Ch 17.2, 8ed]

117
Operational advantages and disadvantages
Advantages
I cost of operation: related to ∆P (i.e. electrical cost only)
I cheap capital cost to build cyclones
I small size
I mounted in any orientation (except for very large units)
I versatile: multiple uses
Balanced by some disadvantages:
I subject to abrasion
I cannot use a flocculated feed: high shear forces break flocs up
I limits on their efficiency curves
I requires consistent feed rate and concentration to maintain
efficiency i.e. not suitable for variable (volumetric) feeds
I counteract: use many small cyclones in parallel; bring them
online as needed

118
Selection of cyclones, sedimentation or centrifuges

119
How to select/model cyclones
Given the complex fluid patterns, cyclone selection is best done
with the vendor.

There are some guiding equations though.

pressure forces ∆P
Eu = =
inertial forces ρf v 2 /2

4Q
For a cyclone, the characteristic velocity, v = 2
πDcyc
∆P = pressure drop from inlet to overflow [Pa]
v = characteristic velocity (not inlet velocity) [m.s−1 ]
ρf = density of fluid [kg.m−3 ]
Q = volumetric feed flow rate [m3 .s−1 ]
Dcyc = cylindrical section diameter of cyclone [m]

0.02 < Dcyc < 5.0 m are typical values

120
Euler number for cyclones
I It is relatively constant, under different flow conditions, for a
given cyclone
I e.g. “this cyclone has an Euler number of 540”
I provided solids concentration remains around or below
1 g.m−3
I Eu can be easily calculated found from clean air at ambient
conditions
[Svarovsky]

If you can’t get/calculate it, then use this:

   2
2 Dcyc Dcyc Dcyc
Eu = π
L K M

L = width of rectangular inlet [m]

K = height of rectangular inlet [m]
M = diameter of overflow (gas) outlet [m] 121
Predicting cut size
The cyclone’s cut size, x50 , can be predicted from the Stokes
number. This is a great way to scale-up through geometrically
similar cyclones:
2 ρ v
x50 S
Stk50 =
18 µf Dcyc

x50 = cut size [m]

ρS = solids density [kg.m−3 ]
v = characteristic velocity [m.s−1 ]
µf = fluid viscosity [Pa.s]
Stk50 = Stokes number [−]
Note:
I this only predicts the cut-size, not the shape of the grade
efficiency curve
I as with Eu, the Stk50 must be calculated on an actual feed
I it is relatively constant for changing conditions
122
Example
Outline the process/plan to solve this problem (do calculations at
home!)

What diameter of cyclone do we need to treat 0.177 m3 .s−1 of

feed, given:
I µf = 1.8 × 10−5 Pa.s
I ρf = 1.2 kg.m−3
I ρS = 2500 kg.m−3
I ∆P = 1650 Pa
I x50 desired is 0.8 µm
I Eu = 700
I Stk50 = 6.5 × 10−5

Hint: if we use 1 cyclone, the pressure drop will be too high; so we

must split the feed into multiple, parallel cyclones. So then, how
many cyclones, and of what diameter should we use?
[Ans: 5 cyclones, Dcyc = 0.15m]
123
Circuits of separators
The remaining slides can be applied to any separation system, though most
commonly used for cyclones and other solid-fluid separations.

When one unit is not enough...

I we need a lower cut size

I need a sharper cut (slope of grade
efficiency curve at xcut )
I we need high concentrations
I use lower velocities to reduce
abrasion on equipment, but this
will change efficiency, so then ...

The rest of this section is from Svarovsky, 4ed, chapter 16

124
Units in series: overflow

Grade efficiency curve for the entire sequence

I cut size becomes
smaller with more
units in series
I cut size sharpness
(steepness of curve)
increases
I but there are
diminishing returns
after 3 to 4 units

G (x → 0) = 10%: implies that 10% of the smallest size fractions

are always found in the coarse underflow: we cannot remove these
fines

125
Recycle around a unit

I dilutes feed, which

improves efficiency
I decreases cut size for
increasing recycle
ratio: Q/q
I again diminishing
returns after a ratio
of 3 is exceeded

126
Units in series: underflow

I we get worse
efficiency
I is this useful for
anything?

127
Recycle in the underflow

I Best of both worlds?

128
Filtration

129
Filtration

130
Filtration section
Filtration: a pressure difference that causes separation of solids
from slurry by means of a porous medium (e.g. filter paper or
cloth), which retains the solids and allows the filtrate to pass

[Geankoplis, p 905]
131
References on filtration

I Geankoplis, “Transport Processes and Separation Process

Principles”, 4th edition, chapter 14.
I Perry’s Chemical Engineers’ Handbook, 8th edition, chapter
18.
I Seader, Henley and Roper, “Separation Process Principles”,
3rd edition, chapter 19.
I Uhlmann’s Encyclopedia, “Filtration 1. Fundamentals”,
DOI:10.1002/14356007.b02 10.pub3

132
Why filtration?
Example: alkaline protease, used as an additive in laundry
detergent

[MIT OCW, Course 10-445, Separation Processes for Biochemical Products, 2005]

133
Commercial units: rotary drum filter

[Seader, modified from fig 19-13] [YouTube video] 134

Commercial units: plate and frame

135
Commercial units: plate and frame

[Also see YouTube video]

136
Commercial units: plate and frame (beer clarification)

137
Questions to discuss

1. What characteristics of a filtration

system will you use to judge the
unit’s performance?

2. What factors can be used to

adjust the units’s performance?
Example: a rotary drum
filter

138
Poiseuille’s law

Recall from your fluid flow course that laminar flow in a pipe
(considering no resistance):

−∆P 32 µ v
=
Lc D2

−∆P = pressure drop from start (high P) to end of tube [Pa]

Lc = length being considered [m]
µ = fluid viscosity [Pa.s]
v = fluid’s velocity in the pipe [m.s−1 ]
D = pipe diameter [m]

139
Carmen-Kozeny equation through a bed of solids (cake)
−∆P 32 µ v
= from which we derive the Carmen-Kozeny equation:
Lc D2
 2
−∆Pc v (1 − )S0
= k1 · µ · ·
Lc  

−∆Pc = pressure drop through the cake [Pa]

k1 = 4.17, a constant [−]
 = void fraction, or porosity typical values? [−]
S0 = specific area per unit volume [m2 .m−3 = m−1 ]

I S0 = specific surface area per unit volume is a property of the

solids
6
I Prove in the next assignment, for spheres, S0 = = f (d)
d
140
Solids balance
Mass of solids in the filter cake =
A Lc (1 − ) ρp
Mass of fluid trapped in the filter cake =
A Lc  ρf ≈ small
Mass of fluid in the filtrate = V ρf
What key assumptions are being made here?

ρp = solid particle density [kg.m−3 ]

ρf = fluid density [kg.m−3 ]
V = volume of filtrate collected [m3 ]
A = cross sectional area for filtration [m2 ]
Then define slurry concentration:

mass of dry solids mass of dry solids ALc (1 − ) ρp

CS = ≈ =
volume of liquid in slurry volume of filtrate V
141
Exercise

1. Calculate the mass of solids in the cake for the case when
I ρp = 3000kg.m−3
I A = 8m2
I Lc = 10cm
2. Also calculate the mass of water in the cake.

142
Deriving the flow through the filter
Our standard equation for fluid flow:

1 dV Q
· =v =
A dt A
for a given velocity v , and volumetric feed flow rate, Q.

But, from the Carmen-Kozeny equation:

−∆Pc k1 µv (1 − )2 S02
=
Lc 3
(−∆Pc ) ()3
v =
(µ)(k1 )(Lc )(1 − )2 (S02 )
from our definition for CS we can solve for Lc
CS V
Lc =
A (1 − ) ρp
1 dV (−∆Pc ) (A) (1−)()3 (ρp )
 −∆Pc
· =v = 2
=
A dt (µ)(CS )(V )(1 − )(k1 ) (1 − ) (S0 )
  µCS V α /A
143
The general filtration equation
1 dV −∆Pc
· =
A dt µCS V α /A

CS = slurry concentration [(kg dry solids)/(m3 filtrate)]

α = specific cake resistance [m.kg−1 ]
All aspects of engineering obey this general law:

transfer rate driving force

J = flux = =
transfer area resistance
including the filtration equation:
1 dV −∆Pc
· =
A dt µCS V α /A
1 dV −∆Pc
· =
A dt µRc
CS V α
Rc = resistance due to the cake = [m−1 ]
A 144
Resistance due to the filter medium
In a similar way, we can define the filter
medium’s resistance:

1 dV −∆Pm
· =
A dt µRm

−∆Pm = pressure drop across the medium [Pa]

Rm = resistance due to the filter medium [m−1 ]
Notes:
I From a practical standpoint, Rm is empirical for the given filter
I We wrap up all other minor resistances into Rm also (e.g.
pipe flow into/out of filter)
I The flux through the filter cake is exactly the same as through
the medium
I After filtration gets started, we very often have Rm  Rc
145
Bringing it all together

As with resistances in series (you learned in Physics I), we have:

 
1 dV −∆Pm −∆Pc −∆Ptot
· = add with =
A dt µRm µRc µ (Rm + Rc )

this is called the general filtration equation.

Rc = resistance due to the cake [m−1 ]

Rm = resistance due to the medium [m−1 ]
−∆Ptot = total pressure drop = −(∆Pc + Pm ) [Pa]

146
Questions

How would you use this equation?

1 dV −∆Ptot
· =
A dt µ (Rm + Rc )

1. to determine the medium resistance?

2. to determine the cake resistance?
3. to find the utility cost of operating the filter?
4. to predict the flow for a given filter when your boss wants a
higher throughput?

147
Cake filtration

I most widely applied

I functions only when
particles have been
trapped on the medium,
forming a bed/cake
I the cake is the filtering
element, not the medium
I medium’s pores are larger
than particles often
I e.g. filter press, rotary
vacuum drum filter

148
Cross-Flow Filtration (TFF)

I TFF = tangential flow filtration (parallel to medium)

I used with gel-like or compressible substances
I pores are smaller than particles
I used when absolute exclusion is essential
I inlet flow of suspension provides shear to limit cake build-up
I reduces cake resistance, Rc
149
Basic filtration equation recap
1 dV −∆Ptot −∆Ptot
· = =
A dt µ (Rm + CS V α /A) µ (Rm + Rc )

V = volume of filtrate collected [m3 ]

A = cross sectional area for filtration [m2 ]
−∆Ptot = total pressure drop = −(∆Pc + Pm ) [Pa]
µ = fluid viscosity [Pa.s]
CS = slurry concentration [(kg dry solids)/(m3 filtrate)]
Rm = resistance due to the medium [m−1 ]
Rc = resistance due to the cake [m−1 ]
α = specific cake resistance [m.kg−1 ]
(k1 ) (1 − ) (S02 )
α =
()3 (ρp )

I which entries in the equation are a function of t?

I which entries in the equation are a function of V ?
150
How do we determine cake resistance α?

Recall:
(k1 ) (1 − ) (S02 )
α=
()3 (ρp )

I Measuring S0 is difficult for irregular solids

I  changes depending on many factors (surface chemistry,
upstream processing)
I Is  constant over time? What does it change with?
I α will also change as a function of ∆P
I So we let α = α0 (−∆P)f

151
Constant pressure (batch) filtration

1 dV −∆Ptot −∆Ptot
· = =
A dt µ (Rm + V CS α /A) µ (Rm + Rc )
Invert both sides, separate, divide by A
dt µ (Rm + V CS α /A)
A· =
dV −∆Ptot
dt µ µCS α
= Rm + 2 V
dV A (−∆Ptot ) A (−∆Ptot )
dt
= B + Kp V
dV
Z t Z V
dt = (B + Kp V ) dV
0 0
Kp V 2
t = BV +
2
µ µCS α
B= Rm [s.m−3 ] and Kp = [s.m−6 ]
A (−∆Ptot ) A2 (−∆Ptot ) 152
Constant pressure (batch) filtration

Kp V 2
t = BV +
2

Practical matters:
I how do we get constant −∆P?
I which type of equipment has a constant pressure drop?
I what is the expected relationship between t and V ?
I Plot it for a given batch of slurry.
I If −∆P is constant, then  is likely constant, and so is α

153
Example
A water-based slurry of mineral is being filtered under vacuum,
with a controlled pressure drop of 38 kPa, through a filter paper of
0.07 m2 . The slurry is at 24 kg solids per m3 fluid. Use
µ = 8.9 × 10−4 Pa.s and the following data:
V [L] t [s] t/V [s/L]
0.5 19 37
1 38 38
2 95 48
3 178 60
4 280 70

1. plot a rough sketch of x = V against y = t/V

2. read off the intercept and slope
3. calculate medium resistance, Rm [Ans: 8.11 × 1010 m−1 ]

4. calculate specific cake resistance, α [Ans: 1.87 × 1011 m.kg−1 ]

5. cake resistance Rc at t = 280 s [Ans: 2.56 × 1011 m−1 ]

154
Pressure dependence of α

Recall: α = α0 (−∆P)f
I We can find tables of α0 and f for various solids
I But they almost will never match our situation

Finding the α value as a function of pressure drop:

1. Repeat the example above at several different −∆P levels
2. Calculate α at each −∆P (see example above)
3. Plot x = (−∆P) against y = α (expected shape?)
4. Take logs on both sides of equation:

ln (α) = ln (α0 ) + f ln (−∆P)

Note: if f = 0 the cake is called an “incompressible cake”

155
Constant rate filtration
dV
At a constant rate, i.e. dt = constant

1 dV 1V Q −∆P
= = =
A dt A t A µ (Rm + CS V α /A)
We can solve the equations for −∆P

µRm V µCs αV 2
−∆P = +
At t · A2
µRm Q µCs αQ 2
−∆P = + t
A A2

I For many processes, the constant rate portion is fairly short

I Occurs at the start of filtration, where a plot of x = (t)
against y = (−∆P) is linear
I Then we settle into constant pressure mode
I That’s the part we have focused on earlier
156
Example continued

Kp V 2 µ µCS α
t = BV + B= Rm and Kp = 2
2 A (−∆Ptot ) A (−∆P tot )

157
Example continued
Suppose our prior lab experiments determined that
α = 4.37 × 109 (−∆P)0.3 , with −∆P in Pa and α in SI units.
Now we operate a plate and frame filter press at −∆P of 67 kPa,

kg dry solids
with solids slurry content of Cs = 300 . The cycle
m3 filtrate
time that the filter actually operates is 45 minutes (followed by 15
minutes for cleaning). Based on a simple mass-balance, the
company can calculate that 8.5 m3 of filtrate will be produced.

1. Calculate the area required.

2. If the slurry concentration had to double (still the same
volume of filtrate), what would the required pressure drop
have to be to maintain the same cycle time?
3. Create a plot of the volume of filtrate leaving the press as a
function of time, t.
158
Pressure requirements change with concentration of slurry

I A = 81 m2 stays fixed
I t = 2700 s stays fixed 159
Volume of filtrate produced against time

I A = 81 m2 stays fixed 160

Extend your knowledge

Research the following topics:

I Variable rate filtration: pressure and flow rate both vary along
the characteristic centrifugal pump curve
I Filtering centrifuge (combine the topics from the last 2 weeks)

161