c 2005 Wiley-VCH Verlag GmbH & Co.

KGaA, Weinheim
10.1002/14356007.b02 10
Filtration 1
Filtration
See also Solid Liquid Separation, Introduction
Walter G osele, Heidelberg, Federal Republic of Germany (Chaps. 1 7)
Christian Alt, M unchen, Federal Republic of Germany (Chaps. 8 11)
1. Terminology . . . . . . . . . . . . . . . 3
2. Filtration Models . . . . . . . . . . . . 4
2.1. Cake Filtration: Calculation of the
Pressure Drop . . . . . . . . . . . . . . 4
2.1.1. Denition of Filter Resistance and
Cake Permeability: The Darcy Equa-
tion . . . . . . . . . . . . . . . . . . . . . 4
2.1.2. The Cake Filter Equation . . . . . . 5
2.1.3. Evaluation of Experiments with Linear
Diagrams . . . . . . . . . . . . . . . . . . 6
2.1.4. Compressible Cake Filtration . . . . . 8
2.2. Blocking Filtration and other Modes
of Filtration . . . . . . . . . . . . . . . . 10
2.3. Deep Bed Filtration . . . . . . . . . . . 11
2.4. Cross-Flow Filtration . . . . . . . . . 14
3. Washing of Filter Cakes . . . . . . . . 15
3.1. Basic Effects, Mass Balances . . . . . 15
3.2. Example of Experimental Results . 15
3.3. Test Procedures and Pitfalls . . . . . 17
3.4. Intermediate Deliquoring before
Cake Washing . . . . . . . . . . . . . . 18
4. Deliquoring of Filter Cakes . . . . . 18
4.1. Deliquoring by Gas Pressure . . . . 18
4.1.1. Equilibrium Saturation of Filter Cakes 19
4.1.2. Kinetics of Deliquoring by Gas Pres-
sure . . . . . . . . . . . . . . . . . . . . . 20
4.1.3. Approximate Solution for Coarse, In-
compressible Cakes . . . . . . . . . . . 21
4.1.4. Practical Scale-Up of Deliquoring by
Gas Pressure . . . . . . . . . . . . . . . . 22
4.1.5. Shrinking and Cracks in Filter Cakes 23
4.2. Deliquoring by Expression . . . . . . 23
5. Optimal Cycle Time . . . . . . . . . . 24
6. Understanding the Filter Resistance 26
6.1. The Equation of Carman and
Kozeny . . . . . . . . . . . . . . . . . . . 26
6.2. Interparticle Forces, DLVO Theory 27
6.3. Mathematical Simulation of Cake
Formation . . . . . . . . . . . . . . . . . 28
7. Solving Filtration Problems in
Small-Scale Tests . . . . . . . . . . . . 29
7.1. Laboratory Tests . . . . . . . . . . . . 29
7.2. Handling of Unlterable Suspen-
sion . . . . . . . . . . . . . . . . . . . . . 30
8. Filtration Equipment . . . . . . . . . 32
8.1. Bag Filters . . . . . . . . . . . . . . . . 33
8.2. Belt Filters . . . . . . . . . . . . . . . . 33
8.3. Candle Filters . . . . . . . . . . . . . . 38
8.4. Deep-Bed Filters . . . . . . . . . . . . . 40
8.5. Disk Filters . . . . . . . . . . . . . . . . 42
8.6. Drum Filters . . . . . . . . . . . . . . . 43
8.7. Leaf and Plate Filters . . . . . . . . . 50
8.8. Nutsche (Pan) Filters . . . . . . . . . . 53
8.9. Pressure Plate Filters . . . . . . . . . 54
8.10. Tubular Filters . . . . . . . . . . . . . . 59
8.11. Special Filter Types . . . . . . . . . . . 60
9. Filter Selection . . . . . . . . . . . . . . 62
10. Filter Media . . . . . . . . . . . . . . . 63
11. Filter Aids . . . . . . . . . . . . . . . . . 65
12. References . . . . . . . . . . . . . . . . . 66
Symbols
a intercept in Figure 5, s/mL
A cross-sectional area, lter area, m
2
b slope in Figure 5, s/mL
B creep constant
C creep constant
d
pore
capillary diameter, m
d
S
Sauter mean diameter of particles, m
E electrostatic potential at the solids sur-
face, V
H cake thickness, m
J Bouchers lterability index, m
3
I ionic strength, mol/L
K
H
proportionalityvolume of cake/volume of
ltrate, m
3
/m
3
K
m
proportionality mass of cake/volume of
ltrate, kg/m
3
K
N
proportionality number of parti-
cles/volume of ltrate, m
3
k permeability =1/
h
, m
2
L distance from the inlet face of the lter
bed, m
m mass of dry lter cake, kg
m
i
molar concentration of ion i, mol/L
2 Filtration
N number of pores in the lter medium
n compressibility index
P dimensionless pressure difference
p
c
capillary pressure, Pa
p
ce
capillary entry pressure, Pa
p
ci
capillary suction pressure, Pa
p
L
pressure in the liquid, Pa
p
S
compressive stress on the solids, Pa
q exponent describing blocking behavior
r distance from the solids surface, m
r
D
Debye length, m
S saturation (volume of liquid/volume of
pores)
S
R
reduced saturation
S

irreducible saturation
t
reg
regeneration time for cleaning and
preparing a run, s
U
c
consolidation rate
V volume of ltrate, m
3

V volumetric ow rate, m
3
/h
z
i
valency of ion i

H
cake resistance relative to cake thickness,
m
2

m
cake resistance relative to dry mass, m/kg
resistance of lter medium, m
1
porosity =pore volume/total volume
viscosity, Ns/m
2
dimensionless time
lter coefcient of deep bed lters, m
1
surface potential, V
Indices
av average
e end of ltration
H related to cake thickness (height)
L liquid
loc local
m related to cake mass
N number
S solid
w wash liquor
pore pore (volume)
1. Terminology
Filtration is the separation of solid particles or
liquid ones (droplets) from liquids and gases
with the help of a lter medium also called a
septum, which is essentially permeable to only
the uid phase of the mixture being separated. In
earlier times, this process was carried out with
felts, and the word lter has a common deriva-
tion with felt.
Often however, purication of a liquid or gas
is called ltration even when no semipermeable
medium is involved (as in electrokinetic ltra-
tion). This chapter deals only with the ltration
of solid liquid mixtures (suspension, slurries,
sludges). For the treatment of gases by ltration,
Dust Separation.
The liquid more or less thoroughly separated
fromthe solids is called the ltrate, efuent, per-
meate or, more rarely, clean water. As in other
separation processes, the separation of phases is
never complete: Liquid adheres to the separated
solids (cake with residual moisture) and the l-
trate often contains some solids (solids content
in the ltrate or turbidity).
The purpose of ltration may be clarication
of the liquid or solids recovery. In clarication
the liquid is typically the valuable product and
the solids are of minor quantity and are often
discarded without further treatment. If however
the solids are to be recovered, they very often
have to be washed, deliquored and dried (see
Fig. 1). In this article, washing means the clean-
ing of a product (lter cake) and it is distin-
guished from cleaning parts of the lter itself,
which will be called rinsing (e.g., rinsing a lter
screen or a lter cloth by jets of water). A fur-
ther distinction is to be made between washing
and extraction or leaching. Washing eliminates
liquid contaminants from the pores between the
particles of a lter cake. Extraction recovers sol-
uble matter from the solid particles themselves
(Liquid Solid Extraction). The term drying
means thermal drying, while the elimination of
liquid from the lter cake by mechanical forces
is called deliquoring or dewatering, e.g., deli-
quoring by gas pressure or by expression.
Filtration is effected by application of vac-
uum, pressure, or of centrifugal force (see
Fig. 2). Vacuum ltration requires a vacuum
pump. The pump evacuates the gas froma ltrate
receiver, where the ltrate is separated from the
gas. The ltrate is drained either by a barometric
leg of at least 8 to 10 m or by a pump that is able
to run on snore (i.e. with a deciency of feed
liquid so that it tends to draw in air). Pressure
ltration typically only requires a pump for de-
livering of the suspension and the lter is placed
Filtration 3
Figure 1. Solids processing chain
within a pressure vessel, hence less easily acces-
sible. Centrifugal ltration is done in perforated
centrifuge rotors (Centrifuges, Filtering).
Figure 2. Driving forces in ltration
Vacuum lters have the great advantage that
the cake is freely accessible. This facilitates au-
tomatic cake handling. However, vacuum lters
cannot handle hot liquids, or solvents with high
vapor pressure. The pressure difference across
vacuum lters is very limited, and the resid-
ual moisture of the lter cake is higher than
with pressure lters. Pressure lters in turn are
preferred when the product must be kept in a
closed system for safety reasons, or if the resid-
ual moisture content is important. The handling
of the lter cake is obviously more difcult in
a pressure lter. Filtration by centrifugal force
requires more technical equipment, but as a gen-
eral rule it yields solids withlower residual mois-
ture (Centrifuges, Filtering).
Various models candescribe the physical pro-
cess of ltration. This chapter concentrates on
four idealized ltration models depicted in Fig-
ure 3.
Cake ltration is the most frequently used
model. Here it is assumed that the solids are
deposited on a lter medium as a homoge-
neous porous layer with a constant perme-
ability. Thus, if the ow rate dV/dt is con-
stant, the pressure drop will increase lin-
early, proportional to the quantity of solid
deposited. This model can be applied par-
ticularly well for all hard, particulate solids.
Blocking ltration: The pressure drop is
caused by solid particles blocking pores.
Soft, gelatinous particles retained by a sieve
exhibit such a behavior. If the owrate dV/dt
is constant, the pressure drop increases expo-
nentially with the quantity ltered, the num-
ber of open pores asymptotically approach-
ing zero. The pores may belong to a lter
medium (screen or lter layer) or it may be
pores within a lter cake of coarse particles,
which are blocked by migrating ne parti-
cles.
Deep bed or depth ltration: Solid particles
are retained in a deep lter layer. This takes
place for example in sand lters for clari-
cation of drinking water, which retain even
colloidal particles. The typical effect of deep
bed ltration is adhesion of solids to the
grains of the lter layer, comparable to char-
coal adsorption. Only rather big particles are
retained by the screening effect. When the
lter bed has been saturated with solids, the
solids concentration in the ltrate leaving the
bed progressively approaches that of the in-
coming suspension.
Cross-ow ltration: In cross-ow ltration
the suspension ows with high speed tan-
gentially to the lter surface, preventing the
formation of a cake. Only a small owof liq-
uid passes through the lter medium. A cer-
tain layer of solids accumulates in the bound-
ary layer on the lter surface, and reduces
the ow of ltrate. After an initial period, a
dynamic equilibrium is established between
convective transport of solids to the lter sur-
face and removal of solids by turbulence and
by diffusion.
4 Filtration
Figure 3. Filtration models
Surface ltration is the antonym to depth l-
tration. The solids are retained on the surface of
a lter medium. Generally the models of cake
ltration or of blocking ltration can be applied.
Screening designates a classication process,
which retains the particles below a certain size
and lets pass the smaller ones (Screening).
Often the term screening ltration is also used
to designate a surface ltration with a screen
as a lter medium. Its mode of action resem-
bles screening (or straining) as long as the lter
mediumis clean, but it is clearly a cake ltration
as soon as a layer of solids has formed.
2. Filtration Models
2.1. Cake Filtration: Calculation of the
Pressure Drop
2.1.1. Denition of Filter Resistance and
Cake Permeability: The Darcy Equation
The resistance to ow of a lter cake can be de-
scribed by Darcys law[1] (see Fig. 4). Consider
a liquid owing through a lter cake (or a stream
of water percolating through soil as considered
by Darcy). The pressure drop p of this ow is
proportional to:
1) The ow rate per unit area V/A
2) The cake thickness H
3) the viscosity of the liquid
4) A constant
H
describing the specic lter
resistance of the cake:
p
1
=

V
A

H
H
(1)
The SI-units are

Pa =

m
3
/s
m
2

mPasm
2

The unit of
H
must therefore be m
2
inorder
to satisfy the Darcy equation (1). The reciprocal
of the lter resistance
H
is also called perme-
ability k of the lter cake:
k =
1

m
2

(2)
Sometimes it is more convenient to dene cake
thickness in terms of solid mass per unit area
(unit kg/m
2
). This leads to a slightly different
denition of lter resistance, the factors being:
1) The ow rate per unit area V/A
2) The cake thickness m/A
3) The viscosity of the liquid
4) Aconstant
m
with the unit m/kg describing
the resistance of the cake.
Then the following expression is obtained in-
stead of Equation (1)
p
1
=

V
A

m
A

m
(3)
Filtration 5
For practical reasons the viscosity is very of-
ten not measured separately. Then it is legitimate
to include it in a term
H
(unit mPa s/m
2
) or

m
(unit mPa s m/kg), respectively.
Using this latter term
H
or
m
lter re-
sistances lie between 10
11
mPa s/m
2
(ltering
very rapidly) and 10
16
mPa s/m
2
(nearly unl-
terable), or between 10
8
and 10
13
mPa s m/kg,
respectively.
Figure 4. Denition of lter resistance

H
=Cake thickness H related lter resistance;

m
=Solid mass related lter resistance
2.1.2. The Cake Filter Equation
The pressure drop in a lter is composed of a
pressure drop p
1
across the cake according to
Equation (1) and a pressure drop p
2
across the
lter medium, which can be written as:
p
2
=

V
A

(4)
where b (unit m
1
) is the resistance of the lter
medium.
The total pressure drop is therefore:
p = p
1
+ p
2
=
H
H

V
A
+

V
A
(5)
or
p = p
1
+ p
2
=
m

m

V
A
2
+

V
A
(6)
If the suspension is homogeneously mixed, the
cake height H (or m/A) will be proportional to
the quantity of ltrate. The concentration is de-
scribed by a factor K:
K
H
=
HA
V
(7)
or
K
m
=
m
V
(8)
this gives
p =

H
K
H
A
2
V
dV
dt
+

A
+
dV
dt
(9)
or
p =

m
K
m
A
2
V
dV
dt
+

A
+
dV
dt
(10)
Equations (9) and (10) are identical under the
generalization
H
K
H
=
m
K
m
. Nevertheless
the distinction between both equations is useful
for clarity.
The differential Equations (9) and (10) can
be integrated either for constant ow rate or for
constant pressure. Integration for constant ow
rate dV/dt =const gives the trivial solution:
p =


V
A

H
K
H
A
V +

(11)
or
p =


V
A

m
K
m
A
V +

(12)
For p =const. the integration yields:
dt =
K
A
2
p
V dV +

Ap
dV (13)
t =

H
K
H
2A
2
p
V
2
+

Ap
V (14)
or
t =

m
K
m
2A
2
p
V
2
+

Ap
V (15)
2.1.3. Evaluation of Experiments with
Linear Diagrams
Linear Diagram in the Differential Form.
It is often very helpful to plot the instantaneous
resistance to ow as a function of the quantity
ltered. The resistance is characterized by the
pressure drop p related to the instantaneous
ow rate dV/dt. Using Equation (5) or (6) this
can be written as:
dt
dV
=

H
K
H
pA
2
V +

pA
=

m
K
m
pA
2
V +

pA
= bV + a (16)
6 Filtration
As shown in Figure 5 left diagram [2], inter-
polation gives a straight line with slope b and
intercept a. The intercept
a =

p
represents the resistance at the very rst mo-
ment of ltration, before a cake is formed, hence
the resistance of the lter medium including the
boundary layer to the cake.
The slope b contains the lter resistance
according to
b = slope =

p
dt
dV

V
=
K
A
2
(17)
Inserting K
H
or K
m
respectively, gives

H
=

p
dt
dV

A
2
K
H
=

p
dt
dV

AV
H
=

p
dt
dV

A
H
e
(18)
and

m
=

p
dt
dV

A
2
K
m
=

p
dt
dV

A
2
V
m
=

p
dt
dV

A
2
m
e
(19)
Linear Diagram in Integrated Form. An-
other approach starts from the integrated lter
Equations (14) or (15). The experimental results
are plotted in the form of
t
V
= f (V )
If the pressure is constant duringthe experiment,
a straight line should be obtained
t
V
=

m
K
m
2A
2
p
V +

Ap
=

H
K
H
2A
2
p
V +

Ap
=
b
2
V + a (20)
This kind of plot is still very popular in prac-
tice because the evaluation is often easier than
with Equation (16). Experimental values mea-
sured with a bucket and a stopwatch can be in-
serted directly into Equation (20), while for a
differential diagram they have to be converted
into momentary ow rates dV/dt.
On the other hand Equation (20) is correct
only for p =const. Furthermore, the resulting
diagram shows less clearly the deviations from
linearity than a diagram based on instantaneous
ow rates. If for example a cake stops growing
and keeps a constant resistance this gives a well
discernible plateau in a differential diagram, but
only a gradual hyperbolic bend in the integrated
diagram.
Example. The interpolationof the test results
shown in Figure 5 leads to the intercept a and the
slope b
a = 3.6s/mL = 3.6 10
6
s/m
3
(21)
and
b =
28s/mL
150mL
= 1.8710
11
s/m
6
(22)
The following parameters are known:
Filtration pressure p =1 bar =10
5
N/m
2
Filter area A=20 cm
2
=20 10
4
m
2
Viscosity =0.001 Pa s =10
3
Ns/m
2
Cake height H
e
=37 mm
Cake mass m
e
=24.3 g
Volume of ltrate V
e
=154 cm
3
Solids in suspension
K
H
=
AH
e
V
e
=
2010
4
0.037
15410
6
= 0.481 (23)
or
K
m
=
m
e
V
e
=
24.310
3
15410
6

kg
m
3
= 158
kg
m
3
(24)
This gives for the lter medium resistance
=
pA

= 3.610
6

10
5
2010
4
10
3
m
1
= 7.210
11
m
1
(25)
From Equations (18) or (19) follows the cake
lter resistance

H
= b
A
2
p
K
H

= 0.18710
11

40010
8
10
5
0.48110
3
m
2
= 1.5610
14
m
2
(26)

m
= b
A
2
p
K
m

= 0.18710
11

40010
8
10
5
15810
3
m
2
= 4.7310
11
m
kg
(27)
Filtration 7
Figure 5. Linear plots according to Eqs. (8) and (11) [41]
Deviations from Linearity. Experimental
data very often differ from the linear charac-
teristic shown in diagrams I or II of Fig-
ure 5. The kind of deviation indicates the nature
of secondary effects (see Fig. 6):
A: Theoretical linear curve without sec-
ondary effect.
B: Some solids have settled before the ltra-
tion started, apparently increasing the resis-
tance of the lter medium.
C: At the beginning of ltration the ltrate
has been turbid. The solids loss diminishes
the amount of cake, apparently indicating a
negative medium resistance. (When no tur-
bidity is found in the ltrate, this curve may
indicate that solids accumulate inside the l-
ter medium, blocking it in the long run!).
Similar curves may be observed when the
solids oat and clear liquid is ltered in the
beginning.
D: The solids settle out completely, increas-
ing the speed of cake growth. At the end of
ltration clear liquid ows through the cake
of constant resistance.
E: Only coarse particles settle. After some
time only the remaining nes are ltered and
the resistance increases progressively.
F: Fine particles trickle through the cake
and block the pores of the cake or of the
lter medium. The process can perhaps be
described as a blocking ltration (see Sec-
tion 2.2).
Further causes of deviations might be:
Determination of the beginning or the end
point of ltration may not be correct. The
end point is particularly difcult to deter-
mine if there is no sight glass in the lter.
It is therefore common practice to neglect
some points at the end of the measure-
ment.
Unusual properties of the suspension
like foam bubbles or oil droplets, non-
Newtonian rheology of the liquid or non-
isometric, plate-like particles with special
orientation can produce a great variety of
effects.
The sediment may be so voluminous that
it lls the whole suspension. The process
then resembles rather an expression than
a ltration (Section 4.2).
Freshly precipitated solids sometimes
continue to agglomerate or to ripen dur-
ing ltration, changing their effective par-
ticle size.
2.1.4. Compressible Cake Filtration
Most lter cakes are compressible, which means
that their resistance increases with growing pres-
sure. An x-fold increase in ltration pressure
therefore normally gives rise to a less than x-
fold increase in ow rate. The compression of
the cake is caused by the compressive stress p
s
on the solids which is caused by the drag of the
owing liquid (see Fig. 7). The loss in liquid
8 Filtration
Figure 6. Typical deviations from linearity [41]
A) Differential plot according to Eq. (16); B) Integral plot according to Eq. (20)
pressure p
L
translates into solids pressure p
S
,
the sum of both being constant:
p
S
= p
L
(28)
p
S
+ p
L
= p
0
(29)
Figure 7. Hydraulic pressure p
L
and solids pressure p
S
as
function of cake height. Curve (a) represents a more com-
pressible cake than curve (b)
The local lter resistance
loc
in the cake is
a function of the local compressive stress p
S
and
it is therefore low at the surface of the cake and
high near the interface with the lter medium.
This
loc
=f (p
S
) can be measured in the Com-
pression Permeability cell (CP cell) described in
Chapter 7, where a sample of the cake is subject
to a known compressive stress.
However, only the average resistance
av
of
a lter cake, all layers combined, is important for
practical scale-up purposes. It can be calculated
from
loc
measured in the CP-cell [3]

av
=
p
p

0
dp

loc
(30)
Normally however it is much easier to mea-
sure
av
directly with ltration experiments in
a pressure lter as described in Section 7.1. The
uneven distribution of local porosity and lter
resistance is then ignored.
The dependence of lter resistance
av
or

loc
on pressure can be approximated over a lim-
ited pressure range by
=
0

p
S
unitpressure

n
(31)
where
0
is the resistance at unit pressure drop
and n is a compressibility index (equal to zero
for incompressible cakes). Both parameters can
be determined easily from a logarithmic plot
=f (p
S
), where the slope indicates the com-
pressibility factor n. Introducing
loc
according
to this approximation into Equation (19), it can
be shown that for 0 <n <1
n
av
= n
loc
(32)

0,av
= (1 n)
0,loc
(33)
Thus the approximation Equation (31) has the
advantage that it applies to both the local and the
average resistance. However, such an approxi-
mation is valid only for a limited range of pres-
sures and Equations (32) and (33) are restricted
to 0 <n <1.
Filtration 9
Tiller and coworkers [4] give an integration
of the lter equation also for n >1. According
to their equations the ow rate quickly reaches
a constant value with increasing pressure. For
practical purposes it can be approximated that
such supercompactible lter cakes with n >1
(e.g., a great number of waste water sludges)
yield the same ow rate independently of the
pressure applied: increasing pressure only in-
creases the compressed layer adjacent to the l-
ter cloth. (The hypothetical cake which yields
the highest ow at an optimal pressure and
less owat higher pressures apparently does not
exist. In practical application, however, the out-
put of a ltration may be best with moderate
pressure, since this avoids excessive blocking of
the lter medium).
For practical purposes Equations (31) (33)
are often ignored and compressible cakes are
treated with the same equations as incompress-
ible cakes, provided is dened as average spe-
cic cake resistance under the conditions of op-
eration.
Concerning deliquoring of compressible
cakes, it is important to know that the porosity
is unevenly distributed after ltration. The lter
resistance often is concentrated in a thin, com-
pressed layer facing the lter medium, the rest
of the cake being very porous and wet. Deliquor-
ing by expression (see Section 4.2) will then be
very effective, even at moderate pressures, since
it acts also on the upper layer of the cake.
2.2. Blocking Filtration and other
Modes of Filtration
According to the idealized idea of blocking l-
tration, every particle in the suspension is re-
tained on the lter mediumand blocks one pore.
The remaining number of open pores is then
given by the equation
N (t) = N
0
(1 K
N
V
L
) (34)
where N
0
is the number of pores in the clean
lter medium, N(t) is the remaining number
of open pores and K
N
(m
3
) is the concentra-
tion of particles by number in the suspension.
At constant pressure drop the volume of ltrate
V(t)
p
=const will follow the equation [5]
V (t) =
1
K
N

1 exp

K
N

V
0
t

(35)
with

V
0
(m
3
/s) being the ow rate at the start
of ltration with a clean lter medium.
In real ltration processes not every particle
in the suspension will block a pore of the l-
ter medium. Some particles will be retained by
adhesion to the walls of a pore without block-
ing it totally; others will pass through the lter
medium without being retained at all. Different
formulae (or ltration laws) exist to describe
such moderate blocking. Ageneralized model is
based on the equations
d
2
t
dV
2
= const

dt
dV

q
(36)
for constant pressure ltration and
d (p)
dV
= const (p)
q
(37)
for constant rate ltration.
The exponent q in this equation varies bet-
ween 0 and 2 and describes the blocking
speed:
q =0 cake ltration, slow blocking
q =1 so-called intermediate ltration
q =3/2 so-called standard blocking ltration
q =2 complete blocking.
A summary of the corresponding equations
for ow rates and ltrate quantities is given
in [6], [7]. Figure 8 visualizes some examples:
Surface straining represents the classical exam-
ple of complete blocking (q =2), but also depth
straining follows the same characteristic. Nor-
mal depth ltration in contrast (with particles
much smaller than the lter pores) leads to pres-
sure drop by diminishing pore size and can
be described by the standard blocking equation
(q =3/2).
In most ltrations it is highly probable that
more than one ltration mode occurs. Cake l-
tration for example often is preceded by an ini-
tial period before cake build-up which can be
described as standard blocking. This period is
characterized by depth ltration (particles pen-
etrating into the lter medium) and/or by nes
passing through the lter medium (creating ini-
tial turbidity of the ltrate). For scientic inves-
tigations it may then be helpful to identify the l-
tration mechanismmore precisely by parameter-
tting the experimental data and quantifying the
above-mentioned blocking speed.
10 Filtration
Figure 8. Three possible modes of ltration: Surface strain-
ing and depth straining are described as complete block-
ing (q =2). Depth ltration of particles much smaller
than the pore size is best described as standard blocking
(q =3/2) [6]
Simplied Evaluation of Experimental
Data. A very pragmatic approach is often used
to describe the mode of ltration. The pressure
drop at constant ow rate is depicted on a semi-
logarithmic scale as a function of the volume l-
tered. Normally the values can be approximated
by a straight line according to [8]
log

p
p
0

= JV
L
(38)
For the ow rate at constant pressure drop the
same approximation holds
log

V
0

= JV
L
(39)
Here p
0
is the initial pressure drop of the clean
lter; and the gradient J, which represents some-
thing like the above-mentioned blocking speed,
is Bouchers lterability index. If J =1 Equa-
tion (38) and (39) are identical to an intermedi-
ate ltration with q =J =1. For other values of
J however the physical meaning of J is unclear.
For scientic purposes this approach is there-
fore not so well accepted. For practical purposes,
however, it is very useful.
2.3. Deep Bed Filtration [9], [10]
In deep bed ltration (as opposed to cake ltra-
tion and cross-ow ltration), the solid particles
are separated mainly by deposition within the
pores of the lter medium. The lter medium
may consist of
a 0.5 3 m layer of coarse grains (e.g., sand
0.3 5 mm) or
a layer of a few centimeters of bers (e.g.,
wound or resin bonded cartridge lters) or
sheets of a few mm thickness (e.g., made of
cellulose).
a layer of granular lter aid (e.g., precoat
layer)
All these lter elements have pores that are
larger than the particles to be retained. The par-
ticles stick in the pores by adhesion and their
accumulation weakens the ltering action and
increases the pressure drop, so that the lter must
be cleaned or replaced periodically. This is why
deep bed ltration should be used not to recover
solids from a suspension such as cake ltration,
but instead to produce a very clean efuent from
suspensions with low solids loading (typically
<0.1 g/L) and with very ne particles. Depend-
ing on the particle size, the prevailing effects of
retention are summarized in Figure 9:
1) Particles larger than the pores are trapped
mechanically. This is typically true for par-
ticles 10 m.
2) Particles 1 10 m in diameter hit the solid
surface mainly due to inertia effects and stick
there due to surface forces.
3) Particles <0.1 m reach the solid surface
mainly because of diffusion and again they
stick there due to surface forces.
In the range between 0.1 and 1 mthe effects
of both inertia and diffusion are small and a min-
imumin transport and hence in the effectiveness
of deep bed lters is observed.
Filtration 11
Figure 9. Retention mechanisms in a deep bed lter
a) Particles >10 m are retained by mechanical intercep-
tion; b) Particles of ca. 1 m size are subject to inertial im-
pact +adhesion; c) Particles <1 m follow mainly Brown-
ian diffusion +adhesion
Thus the transport of particles to the solid
surface is rather well explained and can be de-
scribed mathematically. The mechanism of ad-
hesion to the lter grains however and the re-
sulting sticking probability is less well under-
stood. (Some aspects will be discussed in Sec-
tion 6.2). Consequently the clarication effect
must be summarized in an empirical lter co-
efcient describing the local decrease in con-
centration of the suspension owing through the
bed:

C
L
= C (40)
where c is the concentration of the suspension
and L is the distance from the inlet face of the
lter. For uniform conditions, this differential
equation can be integrated:
c = c
0
exp (
0
L) (41)
here
0
is the initial lter coefcient of a clean,
homogeneous lter bed. As soon as the bed is
loaded with solids, its efciency will diminish
and that is why the solution of the differen-
tial equation becomes rather difcult. Different
models exist to describe the process, but they
are rarely used for practical purposes. To nd a
suitable lter medium in a deep bed lter which
shows good retention efciency over a long cy-
cle time, laboratory tests over a realistic cycle
time have to be carried out with lter layers of
realistic depth.
The pressure drop in a deep bed lter can be
interpreted as an effect of two different phenom-
ena of blocking ltration:
1) The lter media exhibit a resistance to ow,
which is increased by the solids deposit in
the pores. The quantity of deposit is gener-
ally small compared to the pore volume, and
the additional pressure loss per unit depth
can be described as proportional to the local
specic deposit. For constant solids concen-
tration at the inlet and constant retention the
increase in this pressure drop with time is
therefore approximately constant.
2) In addition there is often a pressure drop due
to deposits on the surface of the lter bed.
This is a typical blocking ltration and can
be described by Bouchers law (Eqs. 38 and
39).
The total pressure drop of a deep bed lter at
constant ow rate can then be approximated by
an equation of the type
p (t) = const
1
t + const
2
e
Jt
(42)
The pressure drop should be measured in a
test lter with a vertical height close to that
of the full-scale unit (Fig. 10 A) [11]. For the
case that the driving force is gravity, the pres-
sure prole is shown in Figure 10 B. In the static
equilibrium, at ow zero, 1 m of pressure head
is gained for every meter of depth. The down-
streamconsumers may then drawa constant ow
rate. When ow has started, the pressure drop
within the media increases linearly with depth.
As solids are deposited in the pores, the local
pressure loss will increase in the upper layers
and the pressure line becomes distorted. When
the pressure line touches the atmospheric pres-
sure value, the required ow cannot be main-
tained and the lter should be cleaned.
Cleaning of Deep Bed Filters. Sand lters
can be cleaned by backushing. The ow is re-
versed to wash off deposited solids and ush
them, e.g., to a wastewater station. Often air
scour is used to increase the turbulence and re-
duce the amount of water necessary for uidiza-
tion. A typical set of data of a sand lter for
cleaning of water could be [12]:
Medium depth 1 m
Medium size 0.5 1 mm (as uniform as possible)
Flow rate during
ltration
5 10 m
3
m
2
h
1
Flow rate air scour 60 m
3
m
2
h
1
for 10 min
Flow rate water
backow
15 30 m
3
m
2
h
1
for 3 8 min after air
scour
12 Filtration
Figure 10. A) Laboratory test equipment for sand lters
B) Pressure proles in test lter [11]
During backow the sand grains are classi-
ed, the smallest ones are entrained to the top,
the coarse ones sink to the bottom of the lter
bed. This effect is undesired, since it leads to
premature blocking of the top layer. Therefore
it is important to use sand with grain sizes as
uniform as possible.
Sizing of a Deep Bed Filter. The sizing of a
deep bed lter consists of the following steps:
Preselection of the lter material by com-
parative laboratory tests with small layers of
different materials under standard conditions
and with the ltrate quality as a criterion.
Selection of the right grain size in tests with
a lter column of 1 m height and a realistic
ow rate and pressure drop. If necessary the
grain size must be adapted so that the bed
is saturated (the ltrate gets turbid) at the
same time as the maximum pressure drop
is reached. If the maximum pressure drop
is reached rst, the grains should be coarser
and vice versa.
Special Variations of Deep Bed Filters. As
mentioned above the sand grains are classied
during backush and the uppermost layer is
most efcient in collecting and becomes clogged
rapidly by captured solids. This disadvantage is
diminished when a second layer of coarse mate-
rial with lower specic weight is added on top of
the sand layer. During backow these low den-
sity grains collect at the bed surface and form
a coarse layer able to retain a great quantity of
dirt (so-called multimedia lters or dual media
lters, e.g., with a layer of anthracite particles
1 2 mm in size on top of a main layer of sand
grains 0.5 1 mm in size.)
Other designs of sand lters (especially in
the United States) convert the disadvantage of
classication into an advantage by operating in
the upow mode with ltering and backush-
ing from below. This offers the advantage that
large quantities of incoming dirt are retained
in the lower layer of sand, which contains the
coarser grains and larger pores. These upow
lters however need special safeguards against
breakthrough of dirt when the pressure drop be-
comes high.
Filtration 13
In order to increase the retention capacity in
a given volume, it has also been proposed to
use porous grains instead of sand grains. Small
pieces of polymeric foamhave indeed this effect
[13]. Cleaning by simple backush is, however,
not possible withthis material, it must be cleaned
by backush combined with compression, and
this method has therefore not yet found practical
application.
2.4. Cross-Flow Filtration
In cross-owltration the deposition of particles
on the lter membrane is prevented by a strong
ow parallel to the lter surface. This so called
cross-owis achieved in most cases by pumping
the suspension through bundles of membrane
tubes (see Section 8.10, Figs. 77, 78). Alterna-
tively the cross-ow is achieved by rotating in-
serts (Section 8.9, Fig. 76). The elimination of a
cake enables to lter very ne particles, which
otherwise would form a cake with prohibitively
high resistance. Even submicron, nonparticulate
matter can be retained by semipermeable mem-
branes according to this principle. In this chap-
ter a short overview will only be given, covering
both the ltration of particles and the separation
of non-particulate matter. For more details see
Membranes and Membrane Separation. De-
pending on the size of the particles retained, a
distinction is made between microltration, ul-
traltration, nanoltration, and reverse osmosis.
Typical parameters of cross-ow ltration are
summarized in Table 1.
Microltration retains solid particles (see
Section 8.11). According to the usual denition
of solid particles these are visible in the mi-
croscope, hence >0.1 min size. In steady-state
cross-owltration these particles are conveyed
onto the membrane by convection due to the l-
trate owand transported away fromit by hydro-
dynamic lift forces due to the parallel shear ow
(forces F
Y
and F
L
in Fig. 11). For particles be-
low a certain size the lift force becomes smaller
than the convection, F
L
<F
Y
, and they are de-
posited on the membrane [14]. After deposition
they are retained by van der Waals adhesion
forces F
A
andnot easilyswept awayevenif there
is no ltrate ow (irreversible cake formation).
Steady state microltration therefore requires a
parallel owstrong enough to ush off the nest
fraction of particles in the suspension. (In most
practical applications colloidal particles are also
present and they are transported according to the
mechanisms of ultraltration.)
Figure 11. Forces and transport effects onto a particle in
cross-ow ltration
Br =Brownian motion; F
A
=adhesion to the membrane;
F
D
=drag from the cross-ow; F
F
=friction force;
F
L
=hydrodynamic lift force; F
Y
=drag from then ltrate
ow
Ultraltration retains colloidal particles or
macromolecules. The cut-off sizes are ca.
<0.1 m. Such particles are subject to Brow-
nian diffusion (Br in Fig. 11) and only weakly
inuenced by the hydrodynamic lift force [15].
The smaller these particles are, the more they are
transported away from the lter medium by dif-
fusion. This is why in ultraltration the smaller
particles are preferably swept away and the big-
ger ones are collected on the membrane (con-
trary to microltration). The cut-off size of an ul-
traltration is usually dened as the molar mass
(in g/mol) of a test suspension (generally protein
or dextran molecules of dened molar mass in
the range of 10
3
10
5
g/mol). In practical appli-
cation the cut-off size depends not only on the
pore size of the membrane, but to a large degree
on a gel layer on the membrane consisting of
retained colloids.
Nanoltration is a relatively new technique
combining features of ultraltration and reverse
osmosis with a high selectivity. Its name is de-
rived from its approximate cut-off size of some
nanometers or more exactly molar masses of
200 1000 g/mol. This is achieved with special
nanoltration membranes which still have pores
of a dened size, but their retention depends
14 Filtration
on the electrostatic charge of the molecules to
be separated (bivalent anions are typically re-
tained).
Reverse osmosis retains molecules or ions
using selective membranes without pores. Cer-
tain molecules permeate through the membrane
because they are soluble in the membrane mate-
rial. Other molecules are not (or less) soluble and
are retained (or concentrated) on the upstream
side of the membrane. A particular feature of
reverse osmosis is the high pressure required to
overcome the osmotic pressure of the retained
molecules. An important application is desalina-
tion of seawater. In the food industry it is applied
to concentrate juices at low temperatures.
3. Washing of Filter Cakes
3.1. Basic Effects, Mass Balances
The purpose of washing is to remove the mother
liquor left behind in the cake after ltration when
this liquid is regarded as contaminant or must be
recovered as a valuable component. Wash liquid
ows through the lter cake and displaces the
mother liquor, a mixture of both liquids leaving
the cake as a wash ltrate. The concentration in
this ltrate may help to explain some basic facts.
In Figure 12 this concentration is reported as a
function of the ltrate volume V. In the begin-
ning it is c
0
=1 (concentration of contaminant
in the mother liquor). Four characteristic curves
are reported:
Curve 1 represents the idealized piston ow.
The pore volume V
pore
is ideally displaced and
leaves the cake with the concentration c
0
=1. In
this idealizedcase the washingwouldbe nished
with a quantity of wash liquid V
w
equal the pore
volume or a
washratio

=
V
w
V
pore
= 1 (43)
Curve 2 takes into account that even in an ide-
alized cake the ow passes through streamlines
with different lengths and speeds. Because of
this axial dispersion the required wash volume
is several times the pore volume or
washratio

=
V
w
V
pore
> 1 (44)
Curve 3 is still more realistic; taking into ac-
count that a lter cake contains stagnant zones
or dead-end pores, that are not reached by the
ow. They deliver their contaminant content by
diffusion to the efuent. Diffusion is an asymp-
totic process and depends on the time elapsed,
not on the quantity of liquid. This curve is typ-
ical for most washing processes. The rst part
depends on the wash ratio, the end however is
an asymptotic process depending on time.
When the curves 1, 2 and 3 were drawn, it was
supposed that the mass m of contaminant elim-
inated was contained in the pore volume V
pore
.
The integral below the curves is therefore the
same for all three curves:
m =

0
cdV = c
0
V
pore
(45)
Curve 4 nally depicts an example
where the washing resembles an extraction
(Liquid Solid Extraction) because the con-
taminant is not contained in the pores only,
but also in the solid (liquid inclusions or solu-
ble solid matter). The integral under curve 4 is
therefore not related to the pore volume V
pore
but to a higher total quantity of contaminant.
Theoretically such curves could be used to
determine the residual mother liquor content in
the cake from analyses of the ltrate. In prac-
tice this is impossible since the required residual
contents in the cake are much smaller than the
precision of such a mass balance. Nevertheless
analyses of the ltrate are often used to judge
the progress of washing. This is possible if an
empirical correlation between the two concen-
trations in the ltrate and in the cake exists for
the particular process. The validity of such em-
pirical correlation is, however, restricted to the
particular equipment and operating parameters.
3.2. Example of Experimental Results
An illustrative example of a washing process is
given in Figure 13 Aand B[16]. Akaolinite sus-
pension contaminated with NaCl was ltered to
varying cake thickness and washed with varying
quantities of water. After every test the residual
salt content in the cake was measured and the re-
sults are reported as a function of washing time
Filtration 15
Table 1. Typical parameters of cross-ow ltrations
Microltration Ultraltration Nanoltration Reverse osmosis
Cut-off size >100 nm 10 100 nm >1 nm <10
3
g/mol
10
3
10
5
g/mol 200 10
3
g/mol
Transmembrane pressure 0.02 0.5 MPa 0.2 1 MPa 0.5 3 MPa 2 20 MPa
Permeate ow 50 1000 Lm
2
h
1
<100 Lm
2
h
1
<100 Lm
2
h
1
10 35 Lm
2
h
1
Cross ow speed 2 6 m/s 1 6 m/s 1 2 m/s <2 m/s
Important mechanism of
retention
screening by the pores of
the membrane
screening by the membrane
and the gel layer
electrostatic repulsion and
screening
solubility and diffusion in
the membrane
Important mechanism of
transport
hydrodynamic lift force back-diffusion back-diffusion back-diffusion
Figure 12. Concentration of contaminant in the efuent of a lter cake during washing
a) Idealized piston ow; b) Effect of axial dispersion; c) Effect of diffusion from dead-end pores; d) Effect of contaminant
extracted from the solid
Figure 13. Residual concentration of salt in lter cakes of kaolin. The same test results are reported either as a function of the
quantity of wash water (A) or as a function of washing time (B) [16]
16 Filtration
and of wash ratio. (In contrast to Fig. 12 the wash
ratio is related to the solids mass, not the pore
volume. Also the residual content of the cake is
reported, not the concentration in the ltrate.)
Figure 13 A shows how at the very beginning
of the washing process the residual concentra-
tion depends on the wash ratio, disregarding the
cake thickness. In this example this is true up to
the dash-dotted line which corresponds to less
than a one-fold displacement of the pore vol-
ume. As the wash ratio increases to several times
the pore volume, the mechanism of axial disper-
sion becomes more important, and the effect of
a given quantity of wash liquid becomes better
for thick cakes than for thin ones.
Figure 13 Bshows howat the end of an exper-
iment the washing time is the most relevant pa-
rameter for the result. In this example this seems
to be true below a residual concentration of
c
c
0
10
3
3.3. Test Procedures and Pitfalls
As can be seen in Figure 13, washing of lter
cakes gives scattering results even under labo-
ratory conditions. This explains why only few
experimental results are found in the literature:
manyexperimentators are discouragedbyincon-
clusive test results and never publish them. The
reason of scattering results are stochastic inu-
ences:
Fingering. In most washing processes the
wash liquid has a lower viscosity than the mother
liquor. It tends to owthrough the cake in nger-
like streams past isles of viscous uid, as shown
in Figure 14 [16]. This is a stochastic process,
and the local concentration in the cake and the
momentary concentration in the efuent will
vary stochastically. Scale-up from a laboratory
test lter with only small-scale ngering can be
misleading as well as the results of few small
samples taken from a big cake.
Instabilities, Shrinking. If small electrodes
are placed in a lter cake during the washing pro-
cess, they indicate wildly varying local conduc-
tivity. This reveals the existence of small cracks
in changing size and positions even in cakes with
no visible ssures or shrinkage. This is not sur-
prising because the contact forces between the
particles change dramatically when the ion con-
tent of the surrounding liquid is washed off (see
Section 6.2). The phenomenon has not yet been
investigated thoroughly, but its stochastic effects
on the washing results should be similar to the
above-mentioned ngering.
Figure 14. Visualization of ngering owin a transparent
laboratory nutsche lter. A wash liquid of lower viscosity
displaces dark mother liquor [16]
If shrinking during washing presents a seri-
ous problem, it canbe helpful toreslurrythe cake
in wash liquid and lter it again. Many nutsche
lters (see Section 8.8) are used this way. The
second cake (with less ions in the liquid during
cake build-up) is often less porous and less prone
Filtration 17
to shrinking (but also less permeable) than the
original one.
In order to get meaningful results fromwash-
ing experiments in spite of all these pitfalls,
the following procedure is useful (see Fig. 15).
Wash liquid ows at constant pressure through
the cake. The quantity of ltrate and its compo-
sition are recorded automatically. Often the pH
and the conductivity Q of the ltrate are used as
indicators for concentration; sometimes its color
may be more informative. The washing is con-
tinued until the ltrate reaches a pre-set crite-
rion for purity (chosen according to some pre-
liminary experience so that cake purity is near to
the required specication). At this point the ow
of wash water is automatically stopped. This
should be done automatically, since it will prob-
ably happen during the night. The next morn-
ing, the cake is deliquored by air blowing and
its residual content is analyzed. In small-scale
experiments the whole cake should be analyzed,
in large-scale experiments several samples must
be taken from different parts of the cake.
Figure 15. Test arrangement for washing test
The following results must then be evaluated:
The required washing time,
The required quantity of wash liquid, and
The resulting purity of the cake.
Each of these criteria taken alone will scatter
stochastically, but all three together will often
allow a meaningful interpretation.
3.4. Intermediate Deliquoring before
Cake Washing
Normally it is advantageous to deliquor the
cake partly before washing, thus reducing the
amount of mother liquor to be washed out.
Sometimes however this intermediate deli-
quoring has detrimental effects:
If deliquoring is done by gas pressure, cracks
may appear which let pass the washing liquid
(see Section 4.1.5). In this case any deliquoring
before the washing has to be avoided carefully.
On belt lters (see Section 8.2) an overlap of
ltering and washing zones is therefore often
accepted, even if mother liquid and wash liq-
uid get mixed. Some nutsche lters (Section 8.8)
have even been equipped with optical sensors to
avoid premature deliquoring [17]. They measure
the light reectedbythe liquidsurface anddetect
when the cake surface runs dry, so that the gas
pressure can be released before cracks appear.
If deliquoring is done by compression (see
Section 4.2), the cake may get rather imperme-
able. Thus, according to a rule-of-thumb the ap-
plied pressure should be raised monotonously
during the sequence of ltering intermediate
deliquoring washing nal deliquoring, i.e.,
the squeezing pressure for intermediate deli-
quoring should be lower than the pressure for
washing.
4. Deliquoring of Filter Cakes
4.1. Deliquoring by Gas Pressure
The moisture in the pores of a lter cake can be
displaced by gas owing through the cake under
pressure. The residual saturation with moisture
as a function of time then asymptotically ap-
proaches a nal equilibriumvalue as represented
in Figure 16. Thus, an equilibriumis established
with regard to mechanical displacement of the
liquid. Thermal drying by airow will of course
further reduce the moisture down to total dry-
ness if the air ows long enough. But gener-
ally thermal effects are small as compared with
mechanical effects because of the comparably
small mass of air owing through a lter cake
[18].
The nal equilibrium moisture for mechan-
ical dewatering depends on the nature of the cake
and the applied pressure difference. The initial
deliquoring speed however depends in addition
on the cake thickness. For both values the ow
rate of gas is apparently without importance, if
18 Filtration
it is not to produce the pressure difference. In
the next sections rst the equilibriumconditions
will be examined.
Figure 16. Deliquoring by gas pressure residual saturation
as a function of time
4.1.1. Equilibrium Saturation of Filter
Cakes
A plot of saturation of the cake versus pressure
difference in an equilibriumstate is called a cap-
illary pressure curve. It can be measured in a
device according to Figure 17 [19]. The satu-
rated lter cake (b) is placed in the lter (a) on
a semipermeable membrane (c) (the membrane
is permeable for liquid, but it is impermeable
for gas at the applied pressure). Gas pressure is
applied via port (e). The quantity of liquid dis-
placed from the cake is drained by the valve (d)
and collected. (The standpipe is used to obtain
an exact liquid level when the pressure gauge is
adjusted to zero).
A typical capillary pressure curve is shown
in Figure 18 (incompressible cake of glass beads
with a mean diameter of 79 mand water). Start-
ing from the completely saturated cake (S =1)
the gas pressure is slowly increased, at each
set point the equilibrium is established and the
quantity of displaced liquid is registered, yield-
ing a point of the deliquoring curve. Liquid is
displaced from the pores when the pressure ex-
ceeds a certain threshold pressure. This is the
capillary entry pressure p
ce
(in this example
0.062 bar). Each time when a pore is emptied,
the gas pressure overcomes the capillary pres-
sure at the neck of the pores with a diameter
d
pore,neck
:
p
ce
=
4cos
d
pore,neck
(46)
Further increase of the pressure displaces liq-
uid from ner pores and reduces the moisture
down to an irreducible saturation S

after in-
nite time beyond which no further reduction is
reached. This corresponds to the amount of liq-
uid, which is trapped in isolated domains within
the cake.
Figure 18 also shows the imbibition curve.
When the pressure is reduced again, and pro-
vided the bottom of the cake is still in contact
with the expelled liquid, then the cake will be re-
imbibed by capillary suction. At equal moisture
content the suction pressure will be smaller than
the capillary pressure difference for deliquoring.
This is easily explained because the capillary
pressure for imbibition depends on the larger
waist-diameter of the pores (see Fig. 18):
p
ci
=
4cos
d
pore,waist
(47)
Also imbibition does not go to full saturation,
because some air remains trapped in the pores.
The followingconclusions canbe drawnfrom
the capillary pressure curve:
1) Deliquoring is caused by pressure difference
between gas and liquid. Gas owis normally
required to maintain this pressure difference,
but not for the deliquoring itself. No gas
ow is necessary if the cake is placed on a
semipermeable membrane, which is imper-
meable for gas. (This principle can also be
used technically for dewatering without gas
ow [20]).
2) A certain threshold pressure must be ex-
ceeded before liquid is displaced from the
pores; this is the capillary entry pressure p
ce
.
This implies technical consequences:
Vacuum lters are restricted to pressure dif-
ferences <1 bar. This means that cakes can-
not be deliquored by suction if their p
ce
is
near or above this limit. A pressure lter
with p>1 bar has to be considered for such
cakes, even if its installation is much more
expensive than that of a vacuum lter. Theo-
retically suction can dewater lter cakes with
pores >ca. 3 m: withpure water anda small
contact angle ( 0) Equation (46) yields
p =1 bar for d
pore
=2,9 m.
Filter cakes of submicronic particles like
wastewater sludge have such a high capillary
entry pressure that their pores cannot be dewa-
tered mechanically: with pure water and a small
Filtration 19
Figure 17. Test installation for measuring the capillary pressure p
k
[19]
a) Filter vessel; b) Filter cake; c) Semipermeable membrane; d) Draine; e) Entry for gas
contact angle Equation (46) yields p >10 bar for
d
pore
<0.29 m.
Figure 18. Example of a capillary pressure curve for imbi-
bition and for deliquoring [19]
4.1.2. Kinetics of Deliquoring by Gas
Pressure
In most cases there is no semipermeable mem-
brane below the cake but a normal lter
medium that is permeable for gas, like a l-
ter cloth. Gas and liquid ow simultaneously
through the cake. The gas exerts capillary pres-
sure onto the adjacent liquid and both gas and
liquid ow in the same direction, however with
different speeds and different pressures (the lo-
cal pressure difference between gas and liq-
uid is equal to the local capillary pressure).
The difference in saturation between bottom
and top of an incompressible lter cake is nor-
mally rather small, see Figure 19 left. This is
in contrast to the deliquoring by gravity (or by
a centrifugal eld), shown in Figure 19, right.
In this latter case always a saturated bottom
layer exists, corresponding to the capillary suc-
tion height of the cake. Even in equilibrium this
layer remains saturated. Sometimes this height
is negligibly small. In scraper-type centrifuges
(Centrifuges, Filtering) for example, the sat-
urated bottom layer must in any case be smaller
than the residual layer which the scraper does
not remove.
The permeability of a completely saturated
cake is the same if either liquid or gas ows
through the pores. When, however, gas and liq-
uid ow simultaneously through the cake, the
pores lled with liquid and the pores lled with
gas form two separate capillary systems, each
of them with reduced permeability. The local
permeabilities (for liquid or air) in relation to
the permeability of the saturated cake are called
relative permeabilities k
rel
. They depend on
the local saturation of the cake and can be repre-
sented as:
for the wetting uid (the liquid)
20 Filtration
Figure 19. Kinetics of deliquoring by gas blowing (left) and by gravity or a centrifugal eld (right)
k
rel,w
=
k
w
k
=

w
(48)
for the non-wetting uid (the gas)
k
rel,n
=
k
n
k
=

n
(49)
Such relative permeabilities are shown schemat-
ically in Figure 20 as a function of cake satura-
tion [21]. The permeability for the liquid (the
wetting phase) declines to zero at a nite satu-
ration. This value corresponds to the irreducible
saturation S

. At the other extreme the ow of

air ceases when the saturation is above 90 %.
Some of the voids are lled with air without al-
lowing an appreciable airow.
Figure 20. Relative permeability as a function of saturation
[21]
If the curves for capillary pressure and for the
relative permeabilities as a function of saturation
are known, the local values of capillary pressure,
gas permeability, and liquid permeability are de-
ned as functions of the local saturation. The
deliquoring kinetics by gas ow can then be de-
scribed mathematically. The resulting system of
interdependent differential equations is however
extremely complex and not suited for practical
application. For scientic purposes approximate
solutions have been veried by comparing them
to experiments [19].
4.1.3. Approximate Solution for Coarse,
Incompressible Cakes
For the particular case of incompressible cakes
with threshold pressures much smaller than the
applied gas pressure p
ce
p (hence for rather
coarse solid particles), Wakeman has described
the process of deliquoring by gas pressure by
three dimensionless parameters [22]:
reduced saturation
S
R
=
S
1 S

(50)
dimensionless time
=
p
ce
t

H
H
2
(1 S

)
(51)
dimensionless pressure difference
P =
p
entry
p
ce

p
exit
p
ce
(52)
According to this approach, the cake is charac-
terized by its porosity e, the lter resistance
H
(of the saturated cake) and the capillary pressure
curve which itself is approximately described by
Filtration 21
the capillary entry pressure p
ce
and the irre-
ducible saturation S

(see Fig. 18). The rela-

tive permeabilities are described by generalized
interpolation formula. With these approxima-
tions the residual saturation and the theoretical
gas owrate can be read fromthe dimensionless
charts, shown in Figures 21 and 22. An exam-
ple how to apply these charts is given in [23].
However, the validity of these charts is limited
to p
ce
p and to incompressible cakes. Also the
gas owrate indicated on the chart does not take
into account the possibility of cracks in the cake
which can increase the gas ow tremendously
(see Section 4.1.5).
Figure 21. Reduced cake saturation vs. dimensionless time
[23]
Figure 22. Reduced air (gas) ow rate vs. dimensionless
time (Wakemanand Purchas 1986; reproduced from[23])
4.1.4. Practical Scale-Up of Deliquoring by
Gas Pressure
The deliquoring kinetics are normally investi-
gated by measuring the residual moisture as
a function of cake thickness, blowing time,
and pressure difference. The results are interpo-
lated to give the parameters necessary for scale-
up. Extrapolation beyond the investigated range
of cake thickness, blowing time, and pressure
should be avoided. In addition it is important to
measure and to scale-up the gas owfromrealis-
tic experiments, since in most cases the capacity
of the compressor is the limiting factor, not the
pressure. Particular attention has to be paid to the
possible formation of cracks. In case of doubt,
a lter test with sufciently large lter area is
recommended, because the cracks may not be
pronounced and obvious on a small laboratory
lter. Typically the cracks appear at a charac-
teristic liquid saturation which then represents
the limit for deliquoring by gas blowing with
technically reasonable gas ow rates (see Sec-
tion 4.1.5).
Interpolation of the test results with a mini-
mum of tests is made easier if the residual mois-
ture is reported as a function of a dimensionless
time . The results from a given kind of l-
ter cake are then supposed to t approximately
a straight line in a logarithmic chart. As a rst
approach, when nothing is known about the del-
iquoring behavior of the cake, the experimental
values are reported as a function of

0
=
pt
H (H + )
const
pt
H
2
(53)
Here p =p
entry
p
exit
is the pressure dif-
ference applied. This approach gives gener-
ally good results for centrifugal dewatering, as
shown in Figure 23 (the pressure difference be-
ing calculated from the centrifugal force) [24].
For dewatering by gas blowing, the t is how-
ever often unsatisfactory (see Fig. 24) [24]. A
dimensionless time taking into account the cap-
illary entry pressure p
ce
will always permit a
good t. Two equations are proposed in the lit-
erature [24]:

1

pt
H (H + )

p
entry
p
exit

1
p
ce
p

const
pt
H
2

p
entry
p
exit

1
p
ce
p

(54)
22 Filtration
and [25]

1

pt
H
2

1
p
ce
p

const
pt
H
2

1
p
ce
p

(55)
Here p
entry
and p
exit
are the pressures applied
above and below the cake and p
ce
is the capil-
lary entry pressure. This p
ce
is not measured, but
derived as a best t to the experiments (this is
why the t is generally rather good, see Fig. 25
[24]).
4.1.5. Shrinking and Cracks in Filter Cakes
Deliquoring of compressible lter cakes by gas
pressure very often produces cracks in the cake,
a problem of great practical impact. The mech-
anism of crack formation is represented in Fig-
ure 26. During ltration the viscous forces from
the owing liquid compress the cake. As de-
scribed in Section 2.1.4, only the layer near the
lter medium is subjected to the full amount of
pressure, while the upper layer of the cake re-
mains uncompressed and porous. When ltra-
tion is nished, the liquid surface reaches the
cake surface. Before gas enters into the capil-
laries of the cake, the capillary entry pressure
p
ce
must be overcome. The surface tension of
the liquid thus exerts a pressure p p
ce
onto the
cake, comparable to an elastic membrane spread
over the cake. As long as p
ce
is not exceeded, the
cake is compressed monoaxially like by action
of a piston, the cake thickness is reduced, but no
cracks appear. However, when p
ce
is exceeded,
gas penetrates into the largest capillaries, and the
solid structure is compressed by lateral forces.
This may lead to lateral compression of the cake
and hence to cracks. The criterion for cracks
to occur thus probably depends on the lateral
compressive strength of a cake after monoaxial
consolidation. But until now the precise mech-
anism of crack formation is not understood and
the great variety of crack patterns shown in Fig-
ure 27 [26] cannot be explained.
Cracks in a lter cake are very detrimental to
further deliquoring and also to subsequent wash-
ing (see Section 3.4) because gas and wash liq-
uid ow through the cracks without great resis-
tance and without effect. The only sure remedy
against cracks is compression of the cake before
gas blowing. When the remaining shrinking po-
tential (this is the remaining shrink when the
cake is reduced to total dryness) is small enough,
no cracks will appear anymore [27]. Some l-
ters are equipped for this purpose with mem-
branes or with pressure belts (see Sections 8.6
and 8.10). According to a proposal of Shirato
and coworkers this compression can also be
achieved by adding a layer of very ne mate-
rial (hence a cake layer with high p
ce
) onto the
surface of the cake [28].
Some remedy against cracks is also possible
with nutsches equipped with agitators (see Sec-
tion 8.8). With their paddles they smear over the
surface of the cake during deliquoring so that the
cracks are closed as soon as they appear. This
does, however, not eliminate the cracks totally
but only down to a certain depth below the sur-
face of the cake, so that some cracks can exist
invisibly belowthe surface. Without such equip-
ment the formation of cracks can be reduced
(but only to a very limited degree) if the cake
is formed with high nal ltration pressure, and
hence with somewhat less porosity.
4.2. Deliquoring by Expression
When soft lter cakes are compressed, the void
volume of the cake is reduced by expelling liq-
uid, while the voids remain saturated with liquid.
Such compression often is applied in combina-
tion with a subsequent deliquoring by gas pres-
sure; it reduces the subsequent gas owand pre-
vents the formation of cracks (see Section 4.1.5).
But there are many applications where deliquor-
ing by gas pressure is not possible, and expres-
sion is the only way to reduce the liquid con-
tent by mechanical means. Classical examples
for applications of mechanical forces are juice
and oil pressing from fruits and also the dewa-
tering of wastewater sludge in membrane lter
presses (see Section 8.10) or in pressure belt l-
ters (see Section 8.2). Such sludges form lter
cakes with very high capillary entry pressure p
ce
so that deliquoring by gas blowing is technically
not feasible.
Mathematical description of expression gen-
erally starts from modications of the Terzaghi
model for soil mechanics. The ratio of liquid
collected to the amount of liquid, which can be
expressed, is [29]
Filtration 23
Figure 23. Residual saturation of lter cakes dewatered by centrifugal force as a function of dimensionless time
0
[24]
Figure 24. Residual saturation of lter cakes dewatered by gas blowing as a function of dimensionless time
0
[24]
Figure 25. Residual saturation of lter cakes dewatered by centrifugal force as a function of the modied dimensionless time

1
[24]
U
c
=
H
1
H (t)
H
1
H

= 1 Bexp (Ct) (56)

where U
c
is the consolidation rate, H(t) the
cake thickness, H
1
the original cake thickness
and H

the thickness after innite time. B and

C are creep constants and t
c
is the consoli-
dation (=compression) time. A comparison of
different compression models can be found in
[30].
5. Optimal Cycle Time
When sizing a lter installation, the right cycle
time has to be chosen (example: the same task
24 Filtration
Figure 26. Formation of cracks by gas blowing through a compressible cake
Figure 27. Patterns of shrinking cracks [26]
can be performed in a small lter which must be
cleaned frequently or a bigger one with longer
cycles). With a long ltration time the lter cake
becomes thick and the ow rate per lter area
declines. On the other hand for very short l-
tration the downtime for frequent cleaning will
reduce the capacity. This leads to two questions:
(1) What is the optimal cycle time with the high-
est overall throughput? (2) What is the optimal
cycle time with the lowest overall cost?
1) Figure 28 shows the quantity of ltrate pro-
duced as a function of time. Before the start
of ltration a time t
reg
is needed for regen-
erating the lter (extracting the cake from
the previous cycle, cleaning the lter, rear-
ranging the lter elements, etc.). The high-
est overall throughput is obtained when the
straight line from the start of this regenera-
tion to the end of ltration is tangential to
the yield curve. In other words: the ltration
should be stopped when the instantaneous
ow (as indicated by a ow meter) falls be-
low the mean throughput according to the
denition:

V
mean
=
V
t + t
reg
(57)
This criterion can for example be imple-
mented in a computerized process control.
For design purposes it can be useful to cal-
culate the optimal cycle time in advance. For
this purpose a rule-of-thumb will be derived
Filtration 25
assuming cake ltration and negligible resis-
tance of the lter medium ( =0). The quan-
tity of ltrate according to Equations (14) or
(15) is then
V = c
1
t
1/2
(58)
and Equation (57) becomes

V
mean
=
c
1
t
1/2
t + t
reg
(59)
The highest

V
mean
is obtained if the ltra-
tion is stopped at t
opt
, which is dened by
the differential equation
d
dt

V
mean

= 0 =
1
2c
1
t
1/2
opt
(60)

t
reg
2c
1
t
3/2
opt
This leads to the rule-of-thumb for maxi-
mum throughput:
t
opt,1
= t
reg
(61)
2) Quite analogous considerations are valid for
the optimal cycle time in terms of cost per
quantity ltered. During ltration, the run-
ning cost is proportional to time (capital
cost and pumping energy), cost =c
2
t. In Fig-
ure 28 all costs are converted with this pro-
portionality factor c
2
into an equivalent run-
ning time: t =cost/c
2
.
Regeneration represents a considerable cost,
including downtime, labor, new lter ele-
ments, and disposal of waste. If this cost is
called c
reg
, the mean specic cost is
cost
ltrate
=
c
2
t + c
reg
V
(62)
With the approximation of Equation (58)
this is
cost
ltrate
=
c
2
c
1
t
1/2
opt,2
+
c
reg
c
1
t
1/2
opt,2
(63)
The minimum is dened by
d
dt

cost
ltrate

= 0 =
c
2
2c
1
t
1/2
opt,2
(64)

c
reg
2c
1
t
3/2
opt,2
As result the rule-of-thumb for maximal
yield per cost is obtained:
c
2
t
opt,2
= c
reg
(65)
or cost for the lter run =cost for regener-
ating the lter.
Figure 28. Optimal cycle time derived fromthe yield curve.
t
opt,1
for maximal ow per time and t
opt,2
for maximal
yield per cost
Figure 29. Filter resistance vs. particle size
The range of usual values is compared to the calculated lines
according to Carman Kozeny, Equation (66)
These rules have been derived here for con-
stant pressure ltration. They can also be derived
for lter runs with different but constant ow
rates until a pre-set nal pressure. Practitioners
therefore apply the rule-of-thumb rather gener-
ally (e.g., also for ltration with a pump with a
characteristic curve anywhere between constant
pressure and constant ow).
6. Understanding the Filter
Resistance
6.1. The Equation of Carman and
Kozeny
The resistance of a lter cake depends on the
size and number of pores in the cake. In a rst
26 Filtration
approximation it can be related to the size of
particles and the porosity of their arrangement
according to the classical Kozeny equation [31].
(The theoretical basis of the Kozeny equation is
subject to criticism, see for example [32], never-
theless the Kozeny equation is a very useful ap-
proximation. More recent correlations are given
in [33].) For spherical particles this equation is:

H
5
(1 )
2
36

3
d
2
S
(66)
where is the porosity of the cake and d
S
is the
Sauter mean diameter of the particles, i.e., the
diameter giving the same specic surface. The
lter resistance resulting from this equation is
depicted in Figure 29 as a function of particle
size d
S
and porosity . In practical applications
the lter resistances cover a very wide range of
porosity. Fine particles, especially dry dust, of-
ten form cakes with surprisingly high porosities
(see Fig. 30). The same quantity of powder (l-
ter aid, mean particle size 5 m) has settled in
water with different pH and in air. The settling
volume (and the clarity of the supernatant) is
quite different. Filter resistance depends there-
fore to a large degree on cake porosity and hence
on the surface forces producing this porosity!
6.2. Interparticle Forces, DLVO Theory
According to the theory of Derjaguin,
Landau, Verwey, and Overbeek (=DLVO
theory after the researchers initials), the forces
between suspended particles of equal material
are described by repulsion due to electrostatic
charges and attraction due to van der Waals
forces:
Particles in suspension (especially in po-
lar liquids like water) have an electrically
charged surface, due to adsorption and disso-
lution of ions. The order of magnitude of this
charge is 120 mV< <120 mV, and it
depends on the pH of the liquid. The charge
reaches relatively far, depending on the ion
content of the surrounding liquid. The reach
is characterized by the Debye length r
D
.
In pure distilled water this length is excep-
tionally long with 0.9 m. In most technical
aqueous solutions it is much smaller, in sea
water for example only 0.4 nm [34].
Van der Waals forces are part of the cohe-
sive forces between the molecules within the
solid particle. They reach also a certain dis-
tance beyond the surface, but their range is
less than that of electrostatic forces. The de-
cline is proportional to the distance accord-
ing to
E
attraction
=
const
r
(67)
When two particles approach to distance
zero, these attractive forces become very
strong and should in principle become equal
to the cohesive force inside the particle. The
reality is, however, more complicated be-
cause there are always absorbed molecules
on the surface.
An example of the superposition of both
forces according to DLVO theory is represented
in Figure 31. The graph shows the resulting en-
ergy potential. A positive gradient of the curves
describes attraction, negative gradient repulsion.
At very small distances, attraction prevails. At
larger distance electrostatic repulsionprevails (if
it is not zero).
Figure 30. Settling volumina of the same quantity of l-
ter aid (mean particle size ca. 5 m); left: in water at pH7;
center: in water at pH2; right: dry powder compacted by its
own gravity [44]
Filtration 27
Figure 31. Energy potential vs. particle distance (DLVO-
theory)
Attraction and repulsion between two particles in suspen-
sion. The superposition of electrostatic repulsion and attrac-
tion by van der Waals forces is represented by the resulting
energy potential. High ion content in the surrounding liquid
reduces the reach of electrostatic repulsion (Debye length).
Concerning interparticle contact this means:
Electrostatically charged particles repulse
one another. Seen fromtheir equally charged
neighbors they show some kind of smooth
and slippery repulsive skin. (The smooth re-
pulsive skincanbe made visible withmodern
atomic force microscopy, see, e.g., [35].) The
Debye length characterizes the thickness of
this skin (see Fig. 32).
Without electric charge (as for example in
air), there is no such skin. The particles touch
their neighbors with their rough surfaces and
tend to adhere due to van der Waals forces.
This explains why the dry powder in Fig-
ure 30 has the lowest packing density.
Particles in aqueous suspension with weak
electric charge most probably have oppo-
sitely charged patches on their surface what
may enhance mutual adhesion and the for-
mation of loose ocs.
The surface charge of suspended particles can
be measured as a zeta potential (Colloids,
Chap. 4.; Emulsions, Chap. 11.6.). This po-
tential varies with the suspending liquid, espe-
ciallywithits pH: HighpHmeans a highconcen-
tration of OH

ions, which can be absorbed, cre-

ating a negative surface charge. Low pH means
high concentration of protons and a positive sur-
face charge. At a certain pH the surface is neu-
tral (point of zero charge, isoelectric point). At
this particular pH, the particles can more easily
agglomerate and are easier to lter. This opti-
mal pH is often determined empirically without
knowledge of the zeta potential.
Figure 32. A particle in suspension
Seen from a particle with the same surface charge, it is co-
vered with a repulsive skin of a thickness near the Debye
length. This skin has no roughness and no friction; hence it
is rather slippery.
6.3. Mathematical Simulation of Cake
Formation
Different attempts have been made to simulate
numerically the structure and porosity of lter
cakes. Until nowsuch calculations are restricted
tospherical particles withhomogeneous surface.
Starting fromrandomly chosen locations the tra-
jectories of particles in a lter oware calculated
(generally in two dimensions). When they touch
the cake surface, the deposition is simulated us-
ing either a sticking angle [36], [37], or a fric-
tion angle and an adhesive force [38], [39], or by
estimating the interparticle forces from van der
Waals and electrostatic forces according to the
DLVOtheory [40]. After calculating the deposi-
tion of a large number of particles, the packing
28 Filtration
density of the resulting cake is found. These cal-
culations help to understand and explain some
empirical facts:
1) High adhesion forces and large friction an-
gles make the particles adhere at the rst con-
tact and produce a randomly stacked house
of cards with high porosity and permeabil-
ity. Low adhesion forces or even repulsion
between the particles make themslip to more
stable positions and produce dense packing.
Figure 33 shows as an example of the cake
structure calculated for different friction an-
gles and adhesive forces [39].
2) An opposite effect can be observed when the
particles are very small (colloidal). Here the
Debye length is often not negligible com-
pared to the particle size and the repulsion
can be strong enough to keep the particles at
a distance. In this case the porosity and per-
meability will increase by repulsion and de-
crease byadhesion, contrarytowhat was said
above. This is for example observed with
clay soil in marshland near the sea. The per-
meability of such soil is reduced after ood-
ing with salt water and increases again when
the salt is washed out, i.e., the Debeye length
increases with falling ion content. For indus-
trially lterable suspensions however the ef-
fect is not yet reported in the literature. It
seems that for particles >0.1 m either the
Debye length is to small (at high ion content)
or the zeta potential is not strong enough (as
in distilled water) for this effect.
A comprehensive simulation of cake forma-
tion has not yet been proposed. The above men-
tioned approaches are restricted to spherical par-
ticles with uniform surface charge. Often, how-
ever, the electric charge is unevenly distributed,
and there may even exist oppositely charged
spots on the same particle. This would explain
why many suspensions form loose ocs when
the stirrer is stopped. This is quite often ob-
served even if the pH is not exactly at the iso-
electric point (the termisoelectric range would
therefore be more appropriate than isoelectric
point). The ocs disappear when the stirrer
starts again, but nevertheless they probably have
an impact on lterability because the particles
are prearranged in a way that they touch with
their rough, adhesive spots and have less ten-
dency to slip into densely packed positions when
deposited on the lter cake.
Figure 33. Simulation of the cake build-up for two differ-
ent adhesion forces and friction angles (upper diagram for
4 10
8
N adhesion force and 20

friction angle, lower

diagram for 1 10
8
N and 5

) [39]
7. Solving Filtration Problems in
Small-Scale Tests
7.1. Laboratory Tests
The most common technique for laboratory l-
ter tests uses a laboratory pressure lter accord-
ing to Figure 34 [41]. It consists of a pressure
Filtration 29
vessel about 16 cm long with a ltration area of
20 cm
2
(corresponding to a diameter of 50 mm).
Suspension (250 300 cm
3
) is charged into the
lter, and the lid is closed. Filtration starts when
pressurized gas (air or nitrogen) is applied from
above. This should be done quickly (within a
few seconds) to prevent sedimentation. The l-
trate is collected in a vessel on a balance or in a
graduated cylinder and its quantity is registered
as a function of time. When the ltration is n-
ished, the compressed gas is applied for about
another minute to deliquor the cake. The quan-
tity of gas applied during deliquoring is mea-
sured with a ow meter or it is estimated from
the capacity of the feed line and its pressure
drop on gas withdrawal from the feed. After l-
tration and deliquoring the lter is opened, the
cake thickness is measured and the cake is re-
moved to determine its wet and dry mass and
calculate the moisture content. The quantity of
ltrate obtained is interpreted in a linear dia-
gram (see Section 2.1.3) and the corresponding
lter resistance is calculated. Several measure-
ments at different pressures (e.g., 0.5 +1 +6 bar)
provide the compressibility coefcient n from
Equation (31). For more details see [42], [43].
Another very simple test is the bench leaf test.
The lter leaf includes a lter cloth xed on a
plate (100 cm
2
) with grooves and sealing around
the edges. It is connected to an evacuated ltrate
receiver via a rubber hose with a valve. The lter
leaf is plunged into the suspension and ltrate is
sucked through the leaf by applying of vacuum.
It requires, however, considerable skill to get
well reproducible results with this arrangement,
since the solids are sucked to the leaf against the
effect of settling, and their composition depends
on the way the leaf is plunged into the suspen-
sion. This is why the bench leaf test is used less
frequently nowadays, only for sizing rotary vac-
uumlters, which operate in a quite similar way.
For research purposes it may be interest-
ing to measure the lter resistance as a func-
tion of compressive pressure. To this behalf the
Compression-Permeability Cell (CP cell) has
been developed (see Fig. 35). This is a cylin-
drical cell with a porous bottom and a porous
piston from above. Mechanical load is placed
on the piston and the lter cake is compressed.
Then liquid is percolated through the cake with
moderate pressure and the permeability is calcu-
lated frompressure drop and owrate according
to Equation (1) or (3).
Figure 34. Bench pressure lter with balance and recorder
for yield curve
a) Compressed air; b) Heating uid; c) Filter medium; d) Fil-
trate; e) Balance; f) Recorder
7.2. Handling of Unlterable
Suspension
In industrial ltration often suspensions are en-
countered with a high lter resistance. Several
possibilities for handling of these unlterable
suspensions are given below.
Optimization of Upstream Steps (Crys-
tallization, Precipitation). Considerable im-
provements in lterability are achieved by
producing coarser particles. Good knowl-
edge is available about crystallization
(Crystallization). If, however, the solids are
formed by precipitation, i.e., by mixing two
reactants, this quick process often depends
strongly on the mixing conditions and the reac-
tion speeds. In this case the lterability must be
improved empirically. Improvement has been
achieved with the following parameters:
1) Continuous mixing instead of batch mixing,
30 Filtration
Figure 35. Compression permeability cell
a) Filter cell; b) Filter cake; c) Filtrate receiver; d) Pressure recorder; e) Pump; f) Slurry tank; g) Recirculation; h) Movable
plate
2) Recycling of suspended solids to the zone of
solids formation,
3) Variation of temperature, mixing ratio, stir-
ring speed etc.
Application of Flocculants (Polyelec-
trolytes). Synthetic, water-soluble polymers are
highly effective as occulating agents. They ag-
glomerate ne particles in aqueous dispersion
to voluminous ocs, which settle and lter eas-
ily. As a general rule, however, the resulting
lter cake or sediment has a higher porosity and
residual moisture than without occulation.
Synthetic occulants are commercially avail-
able as nonionic, anionic, or cationic grades with
more or less ionic character. A charge opposite
to the zeta potential of the solids to be separated
gives theoretically the best effect. In practice it
is, however, easier to observe the occulation vi-
sually without knowledge of the surface charge.
Samples of the suspension are prepared in grad-
uated cylinders and gently mixed with different
occulants in 0.1 % solution. The high dilution
is necessary to reduce the viscosity and facilitate
mixing. The required quantity of polyelectrolyte
is generally in the range of 0.5 10 kg per ton dry
solids, equivalent to 10 100 mL of solution for
1 L suspension.
Criteria for the occulation effect are settling
speed, clarity of the supernatant, and aspects of
the ocs (small, dense ocs are sometimes pre-
ferred, because they are more stable).
Adaptation of pH. As explained in Sec-
tion 6.2, the interparticle forces depend on the
pH of the surrounding liquid. With a pH in the
isoelectric range the particles can aggregate and
their lter resistance and settling speed is in-
creased.
Checking of Alternatives to Cake Filtra-
tion. Possible alternatives to ltration are pro-
cessing in:
Settling centrifuges (Centrifuges, Sedi-
menting) or static settlers. Even poorly l-
terable suspensions sometimes settle readily,
either the solids have high specic weight
or big, soft ocs make an impermeable cake
(e.g., wastewater sludge).
Cross-ow ltration (see Section 2.4). This
alternative however yields the solids as a
thickened suspension only.
Evaporation of the liquid (Evaporation).
This alternative may be attractive if the re-
quired energy is not to high (highly concen-
trated suspension or a liquid with moderate
Filtration 31
heat of evaporation, like many organic sol-
vents).
Selection of Filter Aids. Filter aids are inert
powders added to the liquid to be ltered and
increasing the porosity and permeability of the
cake (see Chap. 11). They are very helpful, pro-
vided the presence of lter aid in the solid can
be accepted. Filter aids are used in two ways:
As a precoat layer to protect the lter
medium and improve ltrate clarity and
As a body feed to increase ow rates.
Preliminary laboratory tests help to identify
the proper kind of lter aid. The selection pro-
cedure is done in three steps:
1) Selection of the right material according to
the chemical resistance and purity (perlite,
diatomaceous earth, cellulose, carbon).
2) Selection of the grade (particle size) of the
lter aid. The particle size should be as
coarse as possible to give low lter resis-
tance, but ne enough to prevent the dirt par-
ticles from trickling through the pores of the
cake. For selection of the grade a precoat
layer is prepared by ltering a diluted sus-
pension of lter aid. Then the solution to be
cleaned is ltered through this layer. After
ltration the precoat layer is broken in half.
The dirt should form a separate layer on the
top and should not have penetrated into the
precoat.
3) Selection of the quantity of lter aid. As a
rst guess the quantity of lter aid is calcu-
lated which gives a cake volume equal to the
volume of dirt to be separated. It is then ad-
mixed to the suspension and the lterability
of this mixture determined. If the ow rate
is to low, the next trial should be made with
a higher quantity of lter aid.
8. Filtration Equipment
Filters can be classied in accordance with dif-
ferent criteria:
Solids retention: surface, deep-bed
Filtration technique: cake, cross-ow, screening
Driving force: pressure, vacuum, expression,
magnetic or electric eld, capillary
Operation: discontinuous, continuous,
quasicontinuous, automatic
self-controlling, programmed,
precoating
Filter element:
1) bag, belt, candle/cartridge, disk,
drum, leaf/plate, nutsche, press
plate, tube
2) horizontal, vertical,
single-element, multi-element,
stationary, rotating
3) open, closed
Solids discharge: scraper removal, vibration,
centrifugation, back blow, tipping,
manual
Application: rapid settling systems, moderate and
slow settling systems, clarication
of liquids, pastes, pulps, sludges,
and nonuid systems.
Because of numerous overlaps with respect
to process engineering and design, any system-
atic arrangement entails a loss of clarity. In the
descriptions that follow, the design element is
given top priority.
The principles of operation and the design
of lter equipment have changed little for many
years. Only with recent efforts to improve the
economics of processes have some crucial de-
sign changes and additions taken place. The
most important of these involve
1) automation of operating modes and cycles
2) increases in specic throughput capacities
and separation efciencies, especially aimed
at reducing residual moisture
3) optimization in the context of the overall pro-
cess
4) improvement of operational reliability
5) increased concern for environmental protec-
tion and worker safety
6) adaptation to new, more difcult separation
tasks
7) development of new separation methods for
existing ltration problems
For example, classical lter presses have been
replaced by completely automatic, unattended
equipment. Mechanical dewateringdevices have
been installed on belt and drum lters. Classi-
cal open nutsches have been modied into pro-
cess lters, which combine several downstream
operations (e.g., washing, dewatering, and dry-
ing) in a single apparatus. For difcultly l-
terable suspensions, dynamic lters with mem-
branes have been designed; together with elec-
32 Filtration
trokinetic lters, these open up newpossibilities
for mechanical liquidsolid separations.
Important advances in the performance of ex-
isting lters have come about through better un-
derstanding of the theoretical relationships in all
areas, such as ltration, washing, and dewater-
ing and expression. In particular, the automation
and optimization of processes, based on knowl-
edge about the interactions of the many process
parameters, are currently being pursued in ex-
tensive programs.
8.1. Bag Filters
The lter element in this type of lter is a bag
made of the lter medium. A typical bag has an
opening at one end. The bags are mounted in var-
ious kinds of containers; they allow an uncom-
plicated design, are easy to service, and are rel-
atively versatile in view of the large selection of
lter media available. For example, solids in the
range of 80 800 m are successfully collected
with bags made of fabrics woven from monol-
ament yarns; solids between 1 and 200 m are
separated with needled felt bags or by precoat
ltration (in which deep-bed ltration is active
at the same time).
As batch lters, are opened by hand, they are
employed only where relatively small charges
with extremely lowsolids contents are to be clar-
ied. They therefore nd use chiey in clarify-
ing and polishing ltration (sterilization), where
the solids to be removed are impurities or unde-
sirable byproducts. Applications include chemi-
cal and pharmaceutical products, paints and var-
nishes, juices, edible oils, waxes and resins.
Bags open on one side are charged from the
inside, so that the solids are collected as a l-
ter cake in the bag. Service times range from
hours to days. In pressure bag lters, where the
bags are inserted in supporting cages, pressures
between 0.5 and 1 MPa are common; in special
designs, the pressures are up to 2.5 MPa. Ca-
pacities are in the range of 1 m
3
m
2
min
1
at
1.6 MPa. Filter units with several bags in hous-
ings are available at total capacities of up to
2.4 m
3
/min (4.8 m
3
/min maximum) and down
to quite small values.
Pocket-shaped lter bags are charged from
outside; they are slipped over collapsible lter
inserts attached to the ltrate pipe (Fig. 36). The
open corner of the lter bag is sealed against
the ltrate pipe. In order to increase the ltra-
tion area, the bags are made up to three times
wider than the inserts, so that folds are pro-
duced when they are slipped on; in this way,
the space requirement per square meter of l-
tration area is greatly reduced. Filter units up to
250 m
2
are available. After ltration, the bags
can be cleaned by spraying and blowing with
compressed air.
Figure 36. Pocket-type bag lter
8.2. Belt Filters
Abelt lter resembles an ordinary belt conveyor
driven by one of the guide rollers. Slurry is fed
onto the lter belt from above and the ltrate
drains or is pulled by suction fromthe bottomof
the belt. Belt lters are operated as simple grav-
ity, vacuum and pressure lters. Whether oper-
ation is continuous or intermittent (semicontin-
uous), slurry is fed onto one end of the belt. The
ltrate drains into vacuum tanks, which can be
stationary or can move intermittently with the
belt. The cake can be washed and dewatered in
downstream zones. If necessary, the lter belt
is washed during recycling. In another design,
the belt is made up of a number of individual
cells, which bear the lter cake. In this way, a
more careful separation of the ltration stages
(see also Fig. 1) is achieved.
In general, belt lters are applied for read-
ily lterable suspensions that contain coarser,
and therefore easily sedimented, solids (groups
F and M in Table 2), where the ltrate and the
wash liquor are to be collected separately and the
solids must have gentle handling. The maximum
Filtration 33
Table 2. Filterability of cakes
Filtration characteristics Group of suspension, symbol
Fast Medium Slow Dilute Very dilute
(F) (M) (S) (D) (VD)
Initial cake formation rate, min/cm 0.005 0.1 0.1 1 1 10 10 100 no cake
Slurry concentration, % >20 10 to 20 1 to 10 <5 <0.1
Settling rate rapid medium slow slow
Leaf test rate, kg/m
2
h >2500 250 to 2500 25 to 250 <25 no test
Filtrate rate, m
3
/m
2
h >10 0.5 to 10 0.0025 to 0.05 0.025 to 0.05 0.025 to 0.05
Typical slurry crystalline solids salts pigments wastewater water
ltration area is 120 m
2
. Fluctuations in product
call for control of the belt speed or slurry rate.
For especially sensitive ltrations, belt lters
are used with lter media that remove the liq-
uid from the cake by capillary action alone. The
lter cloth runs over felt or similar lter media.
Vacuum Belt Filters. A continuous vac-
uum belt lter consists of the moving lter cloth
supported on a proled elastomer transport belt
(Fig. 37). To provide better sealing for the vac-
uum space, sliding belts can be placed under
the transport belt, or a specially shaped belt can
travel along with the lter cloth. The difcul-
ties met with in ensuring a good vacuum with
moving seal systems are overcome when the l-
ter belt or ltrate box move intermittently. Fig-
ure 38 shows an example. When the belt is sta-
tionary, slurry is fed to the lter and at the same
time the lter cake is washed, dewatered and,
if necessary, pressed. In the intermittent move-
ment of the belt, the discharge roller shifts the
belt, while the retaining roller holds the belt in
place. The movement is compensated by move-
ment of the compensating roller in the opposite
direction. The vacuum is let down when the belt
is moving.
Another continuous vacuum belt lter has
vacuumtanks of pan-shaped sections each 1.4 m
long. Each section is divided into two parts with
exible connections to the vacuum system (vac-
uum pumps). In this way the individual ltra-
tion stages are divided into arbitrary lengths as
needed. The vacuum tanks are designed with
grids that allow the ltrate to drain freely. Dur-
ing ltration, the tanks move along with the belt
at the same speed, as the vacuum builds up. At
the end of the co-moving path, the vacuum is
broken and the tank vented, while the lter cloth
moves on. Slurry and wash liquors used are fed
continuously.
Vacuum belt lters are usually installed in
open frames, or in closed housings if no va-
por is to be released into the environment. Good
ltration capacities and washing effects can be
achieved only with uniform loading of the lter
medium or lter cake. For this reason, the feed
and wash liquor are distributed over the width of
the lter belt by at nozzles or washing grooves.
Downstreamdewatering of the lter cake is done
by double-belt expression (see twin-belt press,
page 36) or a plate press.
Because of the versatility of vacuum belt l-
ters, they have found use in nearly every eld of
liquidsolid separations where slurries are rela-
tively easy to separate.
Even difcultly lterable slurries with com-
paratively nely dispersed solids can be sep-
arated on belt lters if suitable belt materials
are employed and the solids do not immedi-
ately block the medium. In such cases, pres-
sure is most commonly used instead of vacuum
(see page 36).
Belt lters for gravity and suction ltration
offer low equipment costs. With a suitable lter
medium (nonwovens are generally used), they
can successfully clean up slurries that are not too
highly concentrated (solids down to 50 mg/m
3
)
and of the lterability group F in Table 2. Feeds
include machine-tool cooling and cutting lu-
bricants (emulsions), electrolysis cell slimes,
beer (for removal of turbidity and yeast), waste-
waters, and chemical solutions.
A gravity or hydrostatic lter with a contin-
uous belt (Fig. 39) consists of a tank, in which
a moving support belt made of coarse woven
fabric or a exible belt forms a depression. The
lter belt proper rests on this support belt. Op-
eration is intermittent: slurry is fed in until the
34 Filtration
Figure 37. A), B) Vacuum belt lter (reproduced with permission of Dorr Oliver)
a) Filter cloth take-up assembly; b) Feed box; c) Cake wash; d) Vacuum pan; e) Filter cloth; f) Filter cloth wash; g) Drip pan
drain; h) Drainage belt; i) Tensioning device; j) Filter cloth aligning; k) Drainage belt
Figure 38. Jerking-type vacuum belt lter (BAS Sonthofen)
a) Filter cloth take-up assembly; b) Cake wash; c) Vacuum trays; d) Filter cloth; e) Movable discharge roller; f) Cloth wash
assembly; g) Blocking roller; h) Tensioning assembly; i) Floating roller
increasing resistance to ow causes the liquid
above the lter belt to reach a predetermined
depth. The hydrostatic head promotes the l-
tration rate. The belt is then advanced, the l-
ter cake carried out of the tank, and new lter
mediumtransported into it. The ltrate owrate
is 0.2 m
3
m
2
min
1
.
There are two designs of suction belt lters,
also called at-bed lters. One, used chiey for
wastewater treatment, consists of a receiving
tank for slurry, with a drag chain conveyor. The
perforated or slotted oor of the tank holds the
ltrate receiver, which is connected to the pump.
The lter medium is guided between the scrap-
ers and the tank oor. After a lter cake has built
Filtration 35
Figure 39. Hydrostatic belt lter
a) Sludge bin; b) Float control; c) Filter trough; d) Filter medium; e) Carrier belt; f) Filtrate tank
up to a predetermined thickness, the increasing
pressure drop causes the conveyor to begin mov-
ing. It advances some 20 50 cm along with the
lter cloth. These lters are made with ltration
areas up to 21 m
2
.
The other design consists of a vacuum tank
with lid; the paper lter tape is fed between the
tank and the lid. This type of lter also operates
intermittently. At a lter area of 1 m
2
a through-
put of 0.1 0.5 m
3
/h can be achieved, depending
on the amount of solids in the suspension and the
presence of lter aids.
Pressure Belt Filters. A pressure belt lter
of the at bed type consists of a ltrate receiver
and a pressure chamber. The lter medium is
guided between the two chambers, which are
situated in a housing. Operation is necessarily
intermittent, since the pressure chamber, which
is provided with slotted gates, is pressurized dur-
ing ltration. In smaller units, the use of slotted
gates is replaced by pivoting up the pressurized
section. Pressure belt lters are also made with
expression devices (up to some 2 MPa).
A twin-belt pressure lter is a belt lter with
a second, co-moving belt, which exerts an addi-
tional mechanical compressive load on the lter
cake with pressure rollers. Such a lter can be
employed provided the cake is deformable. This
condition holds especially for bulky occulated
slimes such as occur with lter cakes made up of
relatively ne solid particles (mainly of organic
origin) having a broad particle size distribution.
Twin-belt pressure lters have proved them-
selves for the dewatering of sewage sludges that
have been treated with polyelectrolytes (based
on poly(acrylic acid) and polyacrylamide) to
produce relatively large, stable ocs.
Figure 40. Schematic processing in twin-belt pressure l-
ters
This type of lter has made it possible to re-
duce the relatively high 70 vol % residual mois-
ture in municipal sewage sludges by as much
as 10 vol %, corresponding to a decrease of
35.7 % in the moisture content per cubic meter
of sludge. Further applications have also been
tested, and this class of lter has been adopted
for some dewatering jobs, such as slimes from
coal washing, large-scale animal husbandry, cel-
lulose and pulp production, electrolysis, and off-
gas treatment. An average dewatering capacity
for designs now current is 6 m
3
of municipal
sewage sludge per meter of belt width and hour.
With sludges that are readily dewaterable, the
inuent rate can be increased to 15 m
3
/h.
Machines with multistage dewatering are
used in nearly all these sample applications. Fig-
ure 40 shows a simplied diagram. The occu-
lated sludge is put through preliminary dewater-
ing in a simple screen drainage (rotating or at
screens), since part of the liquid bound up in the
36 Filtration
Figure 41. Twin-belt pressure lter (reproduced with permission of Flottweg)
sludge is released in the occulation step. Vac-
uum ltration usually follows before the lter
cake reaches the expression zone.
A special feature of twin-belt pressure lters
is that the belts change direction several times,
on rollers that shear and press the cake, thus
promoting its compaction. Usually the ltrate
produced in the several stages, which is not en-
tirely clear, is recycled to the rst stage, where
the solids still dispersed in the ltrate are re-
occulated. The ltrate resulting here is largely
free of solids and can be passed on to, for exam-
ple, biological treatment.
Quite a variety of arrangements of the indi-
vidual process stages and design features has
come into being. The performance of these ma-
chines thus varies widely, being inuenced as
well by the properties of the inuent sludges.
Twin-belt pressure lters will nd additional
use in food treatment, for instance in juice ex-
traction from fruits and vegetables. Figure 41
shows a typical arrangement of that twin-belt
press lters including horizontal press section
and several turns of the belts. Here the recov-
ery of juice may reach 65 70 % in a 2500 mm
machine and 7 10 t/h throughput.
In the modication of Figure 42 A, the sludge
enters the expression zones frombelow, through
a straining zone (gravity dewatering). Passing
over rolls that vary in diameter, the cake is
sheared and compressed (up to somewhat over
100 kPa). The modication of Figure 42 B ap-
plies a further increased compressive load with
a third moving at belt.
Figure 42 C shows a lter in which an over-
ow (for gentle handling of the occulated
solids) directly connects the occulation vessel
to the gravity ltration and dewatering stages;
the lter cake passes downward through the suc-
cessive expressing zones. The last stage is sup-
plemented by a device that exerts a controlled
linear pressure.
Cake compression in twin-belt pressure l-
ters is rather modest incomparisonwiththe com-
pressions achieved in membrane lter presses
(see page 58). The twin-belt devices have the ad-
vantage of continuous operation and relatively
low cost. A recent development demonstrates
the possibility of reaching pressures in twin-
belt expressing lters that are similar to those
in membrane lters (up to 2 MPa). A ltration
zone with lter belts converging in wedge fash-
ion is followed by compaction as the cake turns
around a drum pressurized by hydraulically ac-
tuated elements.
Experience and theoretical analyses (see Sec-
tion 4.2) lead to the conclusion that optimal per-
formance is attained in twin-belt pressure lters
Filtration 37
Figure 42. Modications of twin-belt pressure lters (A, B, and C)
a) Press zone; b) Gravity strainer zone; c) Pre-compression zone; d) Press belt high compression; e) Low compression zone;
f) Flocculation; g) Cake break-up; h) Line compression
when the cake is relatively thin, the ltrate drains
on both sides, the expressing times are long (es-
pecially for readily compressible cakes), and the
expressing force is as great as possible.
8.3. Candle Filters
Candle lters (cartridge lters) consist of tubular
elements, mostly covered by sleeves of a lter
medium. They are widely used in the ltration of
weakly to extremely weakly concentrated slur-
ries that contain solids more or less as impuri-
ties at concentrations down to the ppm range.
This class includes all lters with cylindrical el-
ements mounted, singly or multiply, more or less
vertically in tanks (Fig. 43 A).
All candle lters have the fundamental ad-
vantages that the lter elements are relatively
simple in construction and cheap to fabricate,
and that the devices are adaptable to require-
ments imposed on separation efciency and
throughput capacity. The elements are mounted
either suspended froman overhead tube sheet or
upright on a plate (except where each element
has its own housing). There are advantages to
the suspended design, but great care is required
with elements that break easily.
Most candle lters are operated as pressure
lters. The elements can be cleaned by back-
washing, or they may be replaceable. Some
multiple-element lters offer ltration areas of
far more than 100 m
2
. As typical clarifying l-
ters, these devices have a broad eld of applica-
tion (groups D and VD in Table 2). The ltra-
tion surface can take a wide variety of shapes,
depending on the job being done.
Sheet Candles. A simple candle of this type
consists of wire mesh or perforated screen made
of AISI 316 stainless steel, phosphor bronze or
man-made bers supported on a tubular metal
core (Fig. 43 B). Slottedscreens andwire-wound
lters can also be employed. Such elements
are used mainly to collect weakly-concentrated
solids from liquids such as oils and soap and
38 Filtration
Figure 43. A) Sectional view of candle lters and B) tubular lter elements (reproduced with permission of Faudi Filter)
a) Deaeration; b) Opening device; c) Filter candle d) Discharge; e) Tensioning plate
dyestuff solutions; mesh openings range from
0.01 mm to 0.5 mm (down to 2 m in special
meshes). Backwashing is desirable for wire-
mesh elements.
Disposable sleeves of paper or nonwoven
synthetics are used to retain ner solids. In part,
these lter media simultaneously act as deep bed
lters, the solids being deposited in the medium
itself. These simple lter elements offer com-
paratively high ow rates. Available lter units
cover the range from a few liters per minute to
several hundred cubic meters per hour. Enclosed
in pressure-tight vessels, they can be operated at
absolute pressures of up to 10 MPa. The pressure
drop across such a lter is around 10 kPa.
Gap-Type Candle Filters. Gap-type candle
lters can be made of disks, proled wire, or
tubes (Fig. 44). Most are held together by straps.
In one design, the gap-forming elements are
slipped or welded onto a tube; lters of this type
are used as well strainers. A special drinking
water lter element is the gravel-lled prepack
cartridge.
A continuously operating gap-type cartridge
lter has several lter elements, charged on the
inside, arranged on a rotating mount that brings
them one after another to a stationary backwash
device.
Cartridge Filters. Cartridges are designed
for deep-bed ltration and are thus used mainly
for ne ltration of dilute slurries, concentra-
tions of less than 0.01 % and particle sizes less
than 20 m. They usually comprise replaceable
cartridges made of a variety of materials such
as sintered metals (AISI 316 stainless steel), ce-
ramics (also usable at high temperatures), plas-
tics, cellulose or glass ber (resistant to most liq-
uids). Wound cartridges of cotton, cellulose or
manmade polymer bers, often combined with
nonwovens, are distinguished by relatively large
spaces (honeycomb) for retaining solids. The
simple construction and low material costs open
Filtration 39
up technically and economically interesting pos-
sibilities.
Large-Area Filters. Large-area lter ele-
ments are made by folding the lter medium
along the direction of the cartridge (Fig. 45) or
by using pouch and disk constructions. The l-
ter medium is again mounted on a center tube.
Elements can also be built up in several layers.
Folded-paper cartridges of this type offer 170 m
2
of ltration area per unit. Pouch- and disk-type
lters have far less area per unit (up to 0.5 m
2
)
but can be cleaned and are therefore reusable.
Up to 3.45 m
2
of ltration area per unit is avail-
able in pouch lters. A disk lter consists of
disks of pulp, metal screen, or membrane lter
mediumslipped on a central support, alternating
with support disks, which have slots to admit the
slurry into the elements.
Figure 44. Gap-type candle lter
A: a) Steel disk package (turned for solids discharge);
b) Fixed gap cleaner
B: a) Steel wire spiral (turned for solids discharge); b) Fixed
scraper
Other Designs. A number of other cartridge
designs have been devised for special tasks.
For clarifying difcultly lterable liquids, such
as some pharmaceutical and biological liquids
(blood, among others), elements made of brous
bases or materials coated with activated carbon
give satisfactory performance in the micrometer
range (microltration).
A recently developed element is assembled
from seven individual tubes having slots for the
passage of ltrate (Fig. 46). A tube bundle is co-
vered with a suitable high-pressure fabric, which
ts tightly around the curved surfaces during l-
tration. During the backwash cycle, the fabric
is lifted off the tubes; this movement causes the
cake residue to be thrown off. The lter medium
has a tubular weave, that is, it forms a seamless
tube that can withstand several bars of backwash
pressure. Therefore, its cleaning does not cause
any concern. This large-area lter element is sta-
ble and can be used especially well as a back-
wash lter with a cake thickness between 3 and
50 mm. With up to 200 m
2
of ltration area, this
candle-bag lter is suitable for thickening and
clarication of suspension as well as for solids
treatment, such as washing, extraction, or steam
drying. Finally, the cake can be reslurried or dis-
charged as dry solids. Typical applications are
catalyst or activated carbon removal, ltration
of pigments or impurities and treatment of poly-
mer intermediates.
Design and Selection. Candle lters are
generally designed to the ow capacity, ltra-
tion pressure and separating power stated by the
manufacturer. The selection of the lter medium
calls for some experience.
It can often be advantageous to use several
units with small ltration areas in place of one
larger-area unit. With several units in parallel,
for example, one element at a time can be cut
out of the line for washing while the remaining
ones stay in the ltration stage.
8.4. Deep-Bed Filters
This section deals only with depth lters having
beds of sand, gravel or similar loose materials.
A depth lter is usually a simple container
with a porous bottom, on which a bed of lter
medium between 0.5 and 1.5 m deep is placed.
If the bed material is gravel or sand, such a l-
ter is also called a sand lter. Figure 47 shows
40 Filtration
Figure 45. Candle-type strainer lters (reproduced with permission of Mann & Hummel)
A) Radial fold lter; B) Disk lter; C) Multitube lter
a simple form of a gravity sand lter. The l-
ter bed (c) is supported by the screen (a), and
the ltrate is discharged through pipe (e). Wa-
ter, steam or air can be admitted through pipes
(g) for cleaning. Wash water overows at (b),
if at all. When agitation is used in cleaning, the
wash liquor is discharged at (f ). The cooling or
heating coils (e) make it possible to temper the
lter. Depth lters are also operated as pressure
lters, either in order to increase the throughput
or when they are installed in pressurized lines.
To increase the ltration area of pressure sand
lters, several beds can be arranged one above
another.
Most lters of this type are used in the treat-
ment of drinking water and process water. Basin-
type sand lters have been introduced where
quantities are very large, since these xed struc-
tures offer relatively large areas and depths.
Flow in a depth lter is usually from top to
bottom. The ow is reversed in order to clean
the bed. The basic operating modes differ little
even though a variety of techniques are possible,
especially for cleaning.
With ner-grained bed materials used to col-
lect nely dispersed solids, the topmost layer of
the bed tends to become plugged, so that clean-
ing may become necessary before the entire bed
is loaded with solids. In such cases, layers differ-
ing in particle size can be used; a new cleaning
method is needed, however, to avoid mixing the
fractions. In such multilayer lters, the liquid
can be introduced at different points. This ap-
proach gives a 70 % increase in ow velocity,
substantially longer service time, and enhanced
sludge capacity.
When a depth lter is employed to clarify
juices, solutions and similar valuable liquids, en-
closed tanks are used instead of open vessels,
and higher pressures can also be applied. Pres-
sure sand lters usually permit two to three times
as great a pressure drop as gravity lters, and so
ner-grained bed materials can be used as well.
With a few exceptions, bed depths in such lters
are much less than the depths usually found in
water-treatment plants. In industry, depth lters
with beds of special materials such as activated
carbon or ion-exchange resins are often installed
so that dissolved substances can be removed si-
multaneously.
In a continuous process (Fig. 48), the inuent
is initially in an unpacked internal space, whence
it enters the bed through conical internals. The
claried liquid (ltrate) drains from the bed into
an external space. The lter mediumis hydrauli-
cally removed at the bottom and cleaned in a
regeneration tank at the top. The clean sand is
returned to the lter vessel.
For smaller units (up to roughly 600 m
3
/h),
a backwash-cleaning lter has been developed
for largely unattended operation. When solids
deposited in the lter bed cause the differential
Filtration 41
pressure to increase, the liquid level in the tank
will also rise until its hydrostatic pressure initi-
ates backwashing. Because the cross-sectional
area of the backwash pipe is larger than that
of the inuent pipe, the ow velocity can be
roughly four times as great (44 m/h).
Figure 46. Candle-bag FUNDA-BAC lter (reproduced
with permission of Dr. M uller, M annedorf)
a) Filtrate/blowback-air; b) Central tube; c) Concentric
tubes; d) Horizontal slots for ltrate entryblowback; e) Fil-
ter medium; f) Slurry; g) Filter cake; h) Filtrate comingdown
concentric tubes and rising in central tube
Figure 47. Open low sand lter
a) Screen; b) Efuent; c) Sand bed; d) Wash water; e) Heat-
ing or cooling system; f) Filtrate; g) Cleaning pipes
Figure 48. Continuous sand lter
a) Sludge; b) Sand and sludge; c) Slurry; d) Sand bed; e) Fil-
trate; f) Sand and sludge
8.5. Disk Filters
The element in a disk lter is a rotating verti-
cal disk (Fig. 49); one or more such disks are
arranged on a shaft, with roughly half of each
disk submerged in a slurry tank (Fig. 50). The
device allows continuous ltration. The disks
are made up of 8 10 individual sectors, each
of which is an independent lter. The interior
space of a sector is connected to the ltrate dis-
charge by a rotary valve, so that a cake can form
on the outside when the sector is submerged in
the tank. After the submersion part of the cycle,
air is drawn or blown through the sector in order
to effect dewatering; nally, the cake is removed
by simultaneous application of a scraper and an
air blast, or by spraying.
The disks are 1.2 3.6 m in diameter; in a
special design, disks up to 5 m in diameter were
used. Multi-disk lters have ltration areas of up
to 280 m
2
. In devices with relatively large diam-
eters, hydrostatic effects may cause nonuniform
cake thickness.
When several disks are mounted in a single
tank, the tank is tted with an agitator (usually a
paddle mixer) or an air turbulizer in order to pre-
vent sedimentation of the solids, which would
lead to uneven cakes and poor dewatering. If not
with an agitator, the multi-disk lter is built with
a divided tank. The slurry level in the tank dic-
tates the disk rotation speed.
The disk lter is commonly operated as a vac-
uum lter. The tank is open and readily accessi-
ble. If vapors generated at elevated temperatures
42 Filtration
Figure 49. Vacuum disk lter
must be kept from escaping into the environ-
ment, the lter can be mounted in a pressure-
tight housing and operated under pressure; for
hyperbaric operation, see also Section 4.1.4.
Figure 50. Vacuum disk lter
a) Scraper; b) Filter disk; c) Trapezoidal sectors; d) Outlet
nipples ; e) Automatic valve; f) Filter tank; g) Overow
In principle, the lter cake in a vacuum disk
lter can be washed, but it is certainly less effec-
tive than in, e.g., belt lters. The basic advantage
of the continuous vacuumlter is that it requires
little space per unit of ltration area. Another
advantage is that the lter elements are readily
accessible. In some cases, this type of lter has
the drawback that cracks develop in the cake and
cannot be smoothed.
Applications center on high-tonnage prod-
ucts where large ltration areas are needed, e.g.,
iron ore slimes, otation concentrates, white wa-
ter from papermaking machines, blast-furnace
dust from scrubbers, and aluminum hydrate
(groups M and S, Table 2).
8.6. Drum Filters
Wrapping the ltration surface around a cylin-
der yields signicant advantages that have made
drum lters one of the most widely used types.
Their advantages are continuous operation, easy
charging, a choice of cake-removal methods,
low sensitivity to suspension variations, good
possibilities for washing in some cases, satisfac-
tory residual moisture, relatively simple designs,
reliability, and economy.
In one revolution, the drum passes through
the processing stages, thus allowing continuous
movement of lter cake and ltrate. Filters can
be operated as gravity, vacuum or pressure l-
ters.
Gravity DrumFilters (Revolving Screens).
Applications of gravity drum lters are limited
to simple screening, where cake formation is not
Filtration 43
crucial to the ltration operation, that is, where
the cake does not greatly increase the ltration
resistance. Thus the nature of the solids gov-
erns the application: gravity drumlters are suit-
able for ltering coarse-grained or brous mix-
tures, for example, treating surface waters, pa-
permaking wastewaters, and the like. They can
be charged on the inside or the outside. When
the lter is internally charged, the untreated liq-
uid ows through a distributing channel to the
inside of the drum. As the drum turns, the solids
collected by gravity move upward, separate and
are removed, for example with a screwconveyor.
The drum is covered with a screen fabric.
With microscreens, it is possible to collect solids
that are somewhat more difcult to lter, such
as occur in sewage treatment.
A lter charged from outside is used in the
treatment of recycle waters and wastewaters in
paper and pulp manufacturing. The drum is al-
most completely submerged in the slurry. While
the solids are held up on the screen surface
(an easily removable cake forms in the case
of brous solids), the virtually ber-free ltrate
runs into the lowermost part of the drum. For a
drum roughly 2 m in diameter turning at a pe-
ripheral speed of 3 15 m/min, throughputs of
0.5 5.0 m
3
/min per meter of drum width are
possible. Activated sludge ocs are collected
with a drum 3 m in diameter and 3 m long, lined
with a microscreen having a mesh opening of
10 m; the throughput is 630 m
3
/h.
Vacuum Drum Filters. In order to increase
throughputs and extend the range of applica-
tions, the inside or outside of the drum is en-
closed in a housing connected to a vacuum
pump. In the simplest case, the drum has heads
on either end, so that the entire inside space
is under a uniform vacuum. Roughly 30 % of
the drum is submerged in a tank, which ts
fairly closely around it. Slurry in the tank is
held at a constant level by means of an ad-
justable overow. The feed is distributed by a
stirrer, whose oscillating movement along the
drumis intended to prevent partial settling of the
suspended solids. In the submerged part of the
drum, the vacuum pulls liquid through the lter
medium while a cake builds up on the outside.
The cake exposed on the top part of the drum
is dewatered by air pulled through it. Shortly
before the drum re-enters the tank, the cake is
removed by an air blast from inside the drum,
falling through a chute into a collector.
Single-cell vacuumdrumlters are built with
ltration areas of up to 40 m
2
. They are suit-
able for all slurries that can be held suspended
in the tank, that is, slurries that do not contain
readily sedimenting solids (group S in Table 2),
and where the cake need not be washed. They
feature a high working vacuum, since there are
no high losses; good dewatering, since relatively
long blowing times are possible; and simple con-
struction.
Drum lters generally have continuously
variable drives so that they can accommodate a
range of product properties. The lter medium(a
nely-woven or ne-pored fabric) is supported
on a widely spaced wire mesh.
Because of its simple design, the single-cell
drumlter is well-suited to precoat ltration (see
also page 65 and Chap. 11). In a special ver-
sion where the common path of ltrate and air is
short, it is preferred in the processing of cloudy
wine and beverages. Throughputs varies from
0.2 1.2 m
3
m
2
h
1
, depending on the kind of
beverage. In order to produce a uniform pre-
coat, which may be 10 12 cm thick (this layer
is continuously scraped off during ltration), the
concentration of precoat material (diatomaceous
earth) is low, the depth of submersion is slight,
and the drum is rotated rapidly. If the drum shell
is divided into sections 50 60 mm wide and a
tank is added to receive the wash liquor, cake
washing is also possible in single-cell drum l-
ters.
A special variety of this lter is the suction
roll used in papermaking. It has a suction box,
which restricts the vacuumto a relatively narrow
zone at the top. Suction rolls make it possible
to increase the speed of the papermaking ma-
chine from 250 m/min with couch rolls to over
1000 m/min.
Multi-compartment Drum Filters. Washing,
as well as the separate recovery of ltrates
from the several process steps, are made easier
in multi-compartment drum lters. The ltra-
tion surface in such a device is subdivided into
at, closed cells, which have their ltrate sides
connected through separate pipes to vacuum
pump(s) (Fig. 51). The compartments usually
have perforated inserts bearing, for example,
drainage channels on the cake side and ltrate
collection channels on the vacuum side. The
44 Filtration
Figure 51. A) Vacuum drum lter (reproduced with permission of Dorr Oliver)
B) Vacuum drum lter
a) Agitator; b) Valve; c) Drum; d) Filter pipes; e) Division strips; f) Tank
compartment insert bears a fabric, which sup-
ports the actual lter medium. While this fabric
is installed separately for each compartment,
the lter cloth itself covers as much of the drum
periphery as possible.
The ltrate pipes rotate with the drum and
have a sliding seal connection (rotary valve) to
the xed lines leading to the supply tanks and
pumps. The valve also serves as control means,
allotting the desired process stages to the sev-
eral sectors of the drum. The control arrange-
ment makes it possible to set up any desired l-
tration cycle; this is an undisputed advantage of
the compartment lter.
The slurry feed is on the ascending side of
the drum. An overow pipe provides level con-
trol and thus maintains steady ltration condi-
tions; the pipe can be adjustable. A stirrer in the
bottomof the tank has the function of preventing
the solids fromsettling in order to keep fromup-
setting the ltration. It is not always desirable to
start ltration immediately after the drumis sub-
merged, since with sedimenting feeds this would
mean picking up the ne fraction rst and thus
impairing the permeability of the lter medium.
The cake emerging from the tank should rst be
dewatered and then washed.
If cracks appear in the cake on dewatering,
they will interfere with uniform washing. The
cake surface can then be smoothed with pres-
sure rolls. Washing should be spread over a large
portion of the drum circumference, and so with
large quantities of wash liquor there is a dan-
ger that the cake will slip off. To prevent this,
washing belts are placed around the drum. Like
the expressing belts described further on (see
Filtration 45
page 36), these surround the drum and have the
dual function of closing up cracks and effecting
good distribution of the wash liquor. The wash
liquor is usually delivered to the cake through
drip tubes, channels or nozzles.
Figure 52. Cake discharge assemblies in vacuum lters
A) Roller discharge; a) Filter drum; b) Filtrate pipe;
c) Scraper; d) Roller; e) Slurry tank;
B) Belt discharge; a) Filter cloth; b) Tensioning device;
c) Cloth wash; d) Wash liquid receiver
A variety of devices and techniques are used
to remove various cakes from the lter medium.
Scraper discharge is the earliest and simplest
method (Fig. 51 B); it is best suited to rigid, non-
deformable cakes at least 4 mm thick. Lifting of
the cake can be reinforced by an air blast sup-
plied belowthe lter mediumthrough the ltrate
pipes.
In string and chain discharge, parallel strings
or chains are passed over the drum in the cir-
cumferential direction. After the last dewatering
step, these pass over a deecting roll with a rel-
atively small diameter, and the cake then drops
off. Chain discharge is suitable for relatively
thick cakes such as are produced in the ltration
of coarsely crystalline materials. Felted cakes,
on the other hand, call for string discharge.
Roller Discharge. Thin lter cakes, down to
1 mm and thinner, can be taken up by rollers
(Fig. 52 A). Contact between roller and drum
should not be too rm. The roller turns at the
same speed or slightly (3 5 %) faster. The cake
is removed from the roller by a scraper, comb or
other means; part is left on the roller to promote
separation from the drum.
Figure 53. Combined vacuum drum pressure belt lter
(Dorr Oliver, Germany)
a) Press belt drive; b) Filter cloth aligning assembly; c) Press
rolls; d) Press belt tensioning device; e) Vacuum drum lter
Belt discharge involves leading the lter
cloth with the cake over small rollers (Fig. 52 B).
This methodcanbe usedfor continuous ltration
of slurries containing ne solids that would gen-
erally block the lter cloth too quickly. Spray-
ing the cloth from both sides makes it possible
46 Filtration
to handle nely-dispersed suspensions without
precoating.
When ltrates are volatile or tend to produce
vapors, or when an improvement in lter capac-
ity is desired (see page 48 and Fig. 58), it may
be useful to enclose the lter and pressurize the
housing slightly.
Expression devices have been added to drum
lters (Fig. 53). Besides lowering the cake mois-
ture by 20 200 %, compression also promotes
cake removal. The rubber expressing belt is held
against the drum with several pneumatically ad-
justable rollers; forces per unit length of up to
some 250 N/cm are used. The belt path is such
that the cake gradually enters the region of max-
imum pressure.
For the ltration of nely dispersed feeds, the
drumlter can also be operated as a precoat lter
(Chap. 11). Before ltration proper, a 5 6 cm
layer of lter aid is applied from a slurry. It may
be advantageous to smooth the precoat layer.
During ltration, the lter aid must remain ad-
hering to the lter cloth over the entire periphery
of the drum, so as to maintain the suction in all
the compartments of the drum. No air blast is
used to help in cake removal. During ltration,
the scraper is advanced 0.05 3 mm per drum
revolution until a permanent precoat thickness
of a few mm is reached.
If the suspension is one that settles rapidly
(group F of Table 2), it can be supplied to the
tank from a box located at the top of the drum
(top-feed lter). This arrangement creates seal-
ing problems with respect to the drum and the
cake. In general, the lower tank is either elimi-
nated or lled with wash liquor. The inner-face
vacuum drum lter (Fig. 54) is also suitable for
this type of feed; sedimentation and ltration
in such a device are in the same direction, so
that no stirring is needed. The lter medium is
on the inside of the drum shell, while the com-
partments are on the outside. The drum is open
at one end to allow slurry feeding and cake re-
moval; the closed end bears the ltrate pipes and
rotaryvalve. Byvirtue of gravity, the coarse frac-
tion of solids is deposited rst on the lter cloth,
forming a more permeable ground layer, which
separates more easily from the cloth.
Vacuum multi-compartment lters are made
in standard series (about 30 sizes) with ltra-
tion areas up to 100 m
2
and drum diameters up
to 4.2 m. The smaller sizes, roughly 0.25 m
2
,
do good service as laboratory lters. The lters
are irreplaceable for many continuous ltration
jobs, being distinguished by reliability, low op-
erating costs, versatility and adaptability.
Rotation speeds depend on lter size and
desired cake moisture content and range from
0.1 3 revolutions per minute.
Figure 54. Inner face vacuum lter
a) Dewatering zone; b) Rotary valve; c) Settling zone;
d) Suction zone
Sizing of Vacuum Drum Filters. The design
of such lters is normally directed to the
throughput of solids as the signicant compo-
nent of the feed not least of the fact that deliquor-
ing and washing of the produced cake herein are
more or less uncomplicated procedures.
Based on the ltration equation (Eq. 15) (Sec-
tion 2.1.2) the material balance using mass ratios
of the slurry, the cake and the ltrate and the sub-
stitution of the effective ltration time (s) by
t = 60
a
n
(68)
where a is the submerged to the total drum area
and n (min
1
) the number of drum revolutions,
leadtothe available throughput of drysolids (kg)
per unit lter area (m
2
) and unit time (h)
(M
S
)
A
= C

aKnp

(69)
here (M
S
)
A
is the specic mass of solids
(kg m
2
h
1
), C = 3600

60

a conversion fac-
Filtration 47
Figure 55. Pressure drumlter (Fest-Filter for antibiotics production, 1 m
2
ltration area (reproduced with permission of BHS
Sonthofen)
tor (s
1/2
min
1/2
h
1
), K =
1
c
M

1
c
M
c
Mc

a concentration constant (kg/m

3
),
1
the density
of the liquid (kg/m
3
), c
M
and c
Mc
the mass ra-
tios of solids in the slurry or in the cake after
the ltration cycle,
m
the average specic re-
sistance of the cake (m/kg), the viscosity of the
liquid (Pa s) and p the pressure drop through
the cake (Pa). The specic ow resistance of the
lter mediumis disregarded because of the com-
paratively low inuence.
The equation gives a parabolic increase in
cake production with increasing rotation speed;
in other words, the maximum solids output cor-
responds to the highest possible rotation speed.
On the other hand, the cake thickness decreases
exponentially thereby. Since this thickness is
limited by the cake discharge device, the maxi-
mum solids production is also limited. The limit
further depends on the material values contained
in Equation (69).
Table 3 presents some examples illustrating
the performance range of the vacuumdrumlter.
Pressure Drum Filters. To realize the ad-
vantages of continuous drum operation in pres-
sure ltration, with its higher throughputs, and
to make the ltration of volatile liquids more
economical, Fest developed a drum lter with
a pressure housing around the drum (Fig. 55).
The drumturns at a continuously variable speed.
The annular space between drum and housing is
sealed with stufng boxes and partitioned into
four or ve chambers. The drumshell consists of
compartments (Figs. 56 and 57), which are con-
nected to a rotary valve by ltrate pipes. Slurry
under pressure is continuously pumped into the
ltration space. The cake builds up in the lter
compartments; as the drum rotates, the cake is
mechanically compacted by passing through a
wedge, then moves on to the washing chamber.
If no provision is made for multistage washing,
the next step is dewatering by blowing in a sepa-
rate chamber, which is followed by the cake re-
moval section. In the unpressurized region, the
cake is removed by back-blowing and pivoting
scrapers. Washing of the lter cloth can also take
place in this zone.
The advantages are: continuous operation,
relatively high throughputs, ltrations up to
near the boiling point of the liquid, no vapor
losses, very good washing, countercurrent wash-
ing when multistage arrangements are used, low
cake moisture, no cracking before washing, and
mechanical compaction. The lter is naturally
more expensive than a conventional rotary drum
lter, so that additional ltration costs are cal-
culated as some 50 %. It nds use mainly in the
chemical industry (ltration of dyes, vitamins,
inorganic and organic chemicals, pesticides, in-
secticides), in the production of pharmaceuticals
48 Filtration
Table 3. Performance of vacuum drum lters (DorrOliver).
Material Feed concentration, Filtrate rate, Cake moisture, Discharge
% solids kg m
2
h
1
%
Calcium carbonate 20 34 50 40 50 scraper
Coal (otation concentrate) 25 360 500 25 scraper
Coal (refuse) 30 120 145 25 30 scraper
Blast furnace ue dust 45 45 25 30 scraper
Iron oxide (pigment red) 13 17 30 roll
Corn starch 12 28 130 165 30 40 belt
Wine lees low 82 820 (L/m
2
h) precoat
(antibiotics), and in vegetable processing (ex-
traction). Table 4 gives some throughput gures,
which showsubstantial gains in some cases. Be-
cause cleaning is more complicate, it is recom-
mended that the product not be changed.
Table 4. Some throughputs of pressure drum lters, kg m
2
h
1
,
based on dry cake (BHS, Sonthofen, Germany)
Throughput Cake moisture, %
Antibiotics 385 14
Copper sulfate 525 24
Sodium hydrosulte 700 20
Zinc stearate 45 160 60 70
PVC (suspension) 420 51
Titanium dioxide 150 45
Figure 56. Sectional view of Fest-Filter (reproduced with
permission of BHS Sonthofen)
1) Filter drum; 2) Pressure tight housing; 3) Scraper; 4)
Partitioning element pressed by compressed air
Increased capacity sometimes results from
the use of pressure drum lters created by hous-
ing conventional vacuum drum lters in pres-
sure tanks (Fig. 58). These devices have long
been known and have recently come to be called
hyperbaric lters. Many such designs provide
better dewatering than the straight vacuum l-
ter.
Figure 57. Schematic operation of Fest-Filter
Figure 58. Pressurized drum lter Hyperbar lter
Filtration 49
8.7. Leaf and Plate Filters
The element in a leaf or plate lter is essen-
tially a leaf-shaped structure covered with lter
medium on one or both sides. Slurry is supplied
to the leaf from the outside, and the ltrate can
drain through the void inside, while the cake col-
lects on the outside. Filter elements are usually
assembled into units of varying size. They can
be operated as open-tank devices, immersed in
the feed, with the ltrate being withdrawn by
suction (open-tank leaf lter, Moore lter). Be-
cause the cake often drops spontaneously into
the feed tank, these devices are also used as l-
ter thickeners. Alternatively, the lter elements
are enclosed in pressure vessels (which may be
heatable) and operated as pressure lters, espe-
cially for inuents of relatively low concentra-
tion (clarifying ltration). They are also well-
suited to precoat ltration.
The shape and arrangement of the lter leaves
or plates are adapted to the properties of the cake
and the lter operating requirements. If the cake
is easily detached from the lter medium, a ver-
tical leaf conguration can be chosen, so that
the cake falls off the medium by gravity (re-
moval can be promoted by an air blast or vibrat-
ing). Horizontal leaves are suitable for poorly-
adhering lter cakes.
Possible element shapes are innumerable,
from simple perforated and mesh constructions
to multi-layer designs with internal stiffening.
The design of Figure 59 is typical for bed ele-
ments which are preferred because of their pres-
sure resistance. Each element consists of ve
fabric plies withgradedmeshopenings. The core
ply, which allows the ltrate to drain, bears a
supporting fabric on either side, with the lter
cloths proper on the outside.
In another type, the effective ltration area in
each element is made up of a base with perfo-
rated plates, which is inserted in a massive frame
to withstand greater mechanical loads. The lter
cloth or a wire mesh is placed over the base.
For clarifying ltration with lter beds, wo-
ven fabrics, lter paper or membranes, the lter
media are placed right on the horizontal plate
structure. The stack of lter elements to retain
solids downtosome micrometers is placedabout
a cylindrical pipe enclosed within a pressure-
tight vessel (see Figs. 60 and 61).
Figure 59. Wire lter leaf
a) Binding; b) Filter medium; c) Metal mesh; d) Wire mesh
support (Chamber screen); e) Filtrate manifold
Leaf lters are made in a smooth design with
up to 60 m
2
of ltration area. In large plant units,
the leaf assemblies can be moved mechanically
or hydraulically (Fig. 62), or the housings are de-
signed for easy cake loosening and removal. Be-
cause the pressure on the cake is relieved when
the tank is opened, a vacuum is often pulled on
the ltrate pipes at this time so that the cake will
not drop off the leaves until the desired moment.
In the Kelly lter, vertical lter leaves were ar-
ranged axially in the cylindrical pressure hous-
ing, so that the leaf size decreases to either side.
The leaves are vertical in smaller units or hori-
zontal in larger ones (up to 140 m
2
). The cylin-
drical housing of the Sweetland lter is divided;
the lower half can be pivoted to the side to allow
cake removal by hosing off.
Where process conditions do not allow the
slurry heel (slurry remaining in the pressure
housing after ltration is complete) to be recy-
cled to the lter feed or otherwise utilized, some
50 Filtration
designs have one or more lter leaves in the bot-
tom of the pressure space to lter this residual
feed. These leaves do not function during the
main part of the cycle. The heel is usually driven
through those lter leaves by pressurized gases
(compressed air).
Figure 60. Horizontal plate lter (deep-bed lter) (repro-
duced with permission of Seitz-Filter-Werke)
Figure 61. Sectional view of horizontal plate lter
Many designs address the problemof cake re-
moval. Modern ltration systems are generally
provided only with completely automatic opera-
tion; with vertical leaves as shown in Figure 63,
automatic removal is often a simple step.
Figure 62. Horizontal-tank pressure leaf lter (Niagara)
Figure 63. Vertical pressure leaf lter
a) Valve security; b) Overow
In horizontal systems, owable and non-
owable cakes must be distinguished. Flowable
cakes are the rule in precoat ltration for clar-
ication; here the cake is commonly discarded
as waste. The round lter leaves or plates, with
slurryfeedfromthe top, are arrangedona hollow
shaft in a pressure vessel and positioned a few
centimeters apart with spacing disks (Fig. 64).
Cake thickness is controlled through the ltra-
tion pressure or with special sensors. Cakes in
these devices are nearly always so loose that they
Filtration 51
can be discharged by briey spinning the stack
of plates; separation can be helped by backwash-
ing if necessary. Cake removal, required when
leaves cease to function, is effected by vertical
vibration in addition to rotation. Filters of this
type are also called centrifugal-discharge lters
up to 170 m
2
. They are often classied as plate
lters.
A further variation in operation of pressure
plate lters is shown in Figure 64. If the cake
does not ow off the rotating plates, it is dis-
charged by xed scrapers during cake formation
on the rotating plates.
Figure 64. Continuous pressure plate-lter (reproduced
with permission of Schenk Filterbau)
1) Feed; 2) Scraper; 3) Plates; 4) Filtrate discharge; 5) Cake
discharge
To empty the lter completely before remov-
ing the cake, most lters of this group have
leaves at the bottom of the stack (scavenger l-
tration elements) to allow separate ltration of
the slurry heel.
For relatively dry lter cakes, a conical dis-
charge is preferred (Fig. 65, left). The stack of
plates in such a device must be driven from the
top of the tank, which, in addition facilitates the
sealing of the driving shaft.
Nonowable cakes are generally obtained
in cake ltration where lter aids must be
avoided. Help in these instances comes only
from mechanical cleaning devices. Washing in
these pressure lters usually involves some dif-
culty, because the slurry heel must rst be re-
moved and the housing lled with wash liquor.
Special devices have been developed to reduce
the consumption of wash liquor.
In the pressure leaf lters described above,
the feed is pumped into the housing, and its dis-
tribution over the leaves is naturally somewhat
random. The slurry can also be supplied to the
leaves through a hollow center shaft, with the
ltrate draining into the housing. In this way,
uniformloading of the lter elements is ensured.
Figure 65. Centrifugal-discharge pressure plate lter
(STAWAG BIOTECH/Switzerland)
a) Filter drive; b) Upper bearing; c) Upper seal; d) Filter
plates; e) Lower bearing and seal assembly; f) Filter vessel
Range of Application. The applications of
pressure leaf lters are extraordinarily broad,
52 Filtration
and extension to special ltration problems is
relatively easy. As cake lters, these devices
are employed for ltering all kinds of chem-
ical products, such as glycerol, caustic soda
solution, phosphoric acid, or pharmaceuticals.
Run times between cleaning periods are sev-
eral hours, and throughputs of 1 4 m
3
m
2
h
1
(up to 6 m
3
m
2
h
1
in extreme cases) have
been achieved. Pressure leaf lters are applied
in group S and D of Table 2.
When used as precoat lters, these units are
especially well-suited for the collection of rel-
atively ne-grained solids from low-concentra-
tion feeds (beverages, cane syrup, or water). Av-
erage throughputs are 0.3 1 m
3
m
2
h
1
. Run
times are strongly dependent on the level of im-
purities to be removed; in mixed and thick juices
in sugar rening, runs of 7 16 h (depending on,
e.g., the quantity of diatomaceous earth used as
lter aid) are achieved.
Centrifugal-discharge lters are available
with ltration areas of 1 60 m
2
(up to
170 m
2
from one manufacturer); the leaves are
420 1500 mm in diameter.
8.8. Nutsche (Pan) Filters
The simplest type of nutsche lter is an open
tank or box with a porous bottom, into which
the slurry is charged or ows continuously. Fil-
ters of this kind are common everywhere, espe-
cially in laboratories. They do not impose strin-
gent requirements on suspension properties, and
they can function even without difculty with
highly-concentrated or rapidly-settling slurries.
As a rule, they are operated manually. For poorly
lterable feeds, suction can be applied to the l-
trate side.
Nutsches are widely applied in the chemical
industrybecause all the steps inltration, includ-
ing washing, dewatering, and drying if practiced
(with hot air), can be performed in a clean, dis-
tinct manner. There has accordinglybeennolack
of efforts to automate the operation of nutsches
or make it continuous.
In order to increase ltrate ow rates, an
enclosed nutsche can also be pressurized with
air or inert gas. Pressure nutsches are espe-
cially good with volatile solvents. They can be
drained mechanically through a retractable bot-
tom. Several process steps can be carried out
with no change of apparatus when nutsches
in process service are tted with a variety
of accessories (e.g., agitator, mixer, heating
and cooling devices); in this way operations
such as mixing, stirring, reacting, crystalliza-
tion ltration, extraction and drying can be
performed in sequence. Low equipment costs,
non-polluting operation, and optimal techniques
make these devices suitable for small product
batches. Nutsches can be placed on pivots and
repositioned for the various process steps (Fig.
66).
Figure 66. Pressure nutsche lter multi purpose lter
in the crystallization step (reproduced with permission of
Schenk Filterbau)
Nutsches are good candidates for ltration
with relatively thick cakes. To keep cracks from
appearing, some lters have smoothing attach-
ments, which can also provide advantageous
mixing of wash liquor and cake and can assist in
cake discharge.
Tanks with internals such as agitators and
solids-discharge devices (plough) are more ver-
satile nutsches. These accessories can homoge-
nize the lter cake and render it especially wash-
able (Fig. 67).
In some advanced developments, intended
for comparatively high throughputs, a number
of individual nutsches circulate on either a belt
or a carousel. (A system of the rst type is usu-
ally classed as a belt lter.) This kind of nutsche
has proved useful especially in phosphoric acid
Filtration 53
Figure 67. Pressure Nutsche lter
a) Agitator; b) Wash device; c) Screw conveyer; d) Heating pipes; e) Filter plate
manufacturing, where it is employed to remove
gypsum in the sulfuric acid digestion of phos-
phate rocks. Careful countercurrent washing is
practiced in order to attain a high degree of sep-
aration. Equipment sizes range up to a ltration
area of some 200 m
2
.
Automated nutsches share the advantages
statedearlier: Theyare insensitive toslurryprop-
erties and thus suitable for groups F, M, and S
in Table 2, and they provide a clear separation
of the process steps and thus insure a uniform,
reliable ltrate composition.
For the ltration of slimy feeds, it is advan-
tageous to keep the lter cake in motion, for ex-
ample by mechanical vibration or an appropriate
slurry ow, so that the cake will remain perme-
able for a longer time.
8.9. Pressure Plate Filters
The lter press is one the simplest of all pressure
lters; it allows relatively easy cake removal af-
ter the individual ltration steps. It consists of a
number of ush plates, up to 100 or more, and
frames whose width (between 25 and 150 mm)
can be matched to the service conditions. The
plates and frames formrather at pressure ltra-
tion spaces with a relatively large ltration area
(plate-and-frame lter press, Fig. 68 A). During
ltration, the plates andframes must be squeezed
by mechanical or hydraulic means in order to
seal the ltration spaces (Fig. 69). The faces of
the plates are studded or grooved to permit l-
trate drainage.
54 Filtration
Figure 68. Filter press sectional view
A) Plate-and-frame lter press
a) Filter cloth; b) Filter plates; c) Hollow frames; d) Mov-
able end half plate; e) Fixed end half plate
B) Recessed-plate lter press
a) Fixed end half plate; b) Filter cloth fastening; c) Recessed
lter plates; d) Filter cloth; e) Movable end half plate
The plates canalsobe made witha raisededge
(recessed-plate lter press, Fig. 68 B). There are
several essential differences between the two
types: In the plate-and-frame press, slurry enters
through channels in the corners of the plates and
frames. The lter cloths can therefore be shaped
as at strips and pulled over the plates; this ar-
rangement is advantageous when the cloths are
changed. In recessed-plate presses, slurry is in-
troduced into a center orice of the plates. The
lter cloths must be made in two parts and in-
serted in the slurry feed channels; this arrange-
ment involves much more effort. Further, it is
harder to vary the cake thickness in recessed-
plate devices than in plate-and-frame presses,
where all that is necessary is to replace the
frames. The recessed-plate press has, however,
the advantage that only about half of the lter
elements need be moved to open the press; the
equipment also costs much less for a given ltra-
tion area. Because of the deection of the lter
cloths at the edge of the recessed plate, thinner
cakes (up to about 35 mm) are generally pro-
duced than in plate-and-frame devices. With ei-
ther type of press, it must be noted that capillary
ow takes place despite the relatively high den-
sity of the lter cloths, and this ow at least
in ltration with solvents causes the undesir-
able escape of ltrate through the seals between
plates.
The ltrate is either discharged separately
from each plate through cocks (open discharge)
or collected in a channel (closed discharge),
which is preferred when toxic or volatile ma-
terials are handled.
Figure 69. Filter press (reproduced with permission of Eberhard Hoesch & S ohne)
Filtration 55
At cake washing time, valves are changed
over and special lines carry wash water only to
every other plate, so that the liquor can ow
transversely through the cake to the opposite
plate. Occasionally the wash water is also de-
livered through the slurry channel; this arrange-
ment is, of course, not as effective. After wash-
ing, the cake can be dewatered by passing steam
or (possibly heated) air through it.
For cake discharge, the press is opened and
the plates are moved, one by one, through a pre-
determined distance. A variety of mechanical
devices are nowavailable for this purpose. If the
cake is to drop out by itself, it should not adhere
too tightly to the lter cloth and should not be
too densely ltered. Cake separation can be as-
sisted by rapping, pulling out the lter cloth, or
other means. An automatic lter press for high
throughputs has more than 100 chambers with a
ltration area of 1000 m
2
and more.
Filter presses are generally insensitive to
slurry ltration properties, with perhaps one ex-
ception: In large presses with long slurry chan-
nels, varying sedimentation may cause a slurry
of groupS(see Table 2) toll the ltrationspaces
unevenly. Under some conditions, an undesired
classication occurs, with a detrimental effect
on washing, and may even lead to pressure dif-
ferences between chambers, causing the plates
to deform or break. In such cases, it is desirable
to provide a hole for pressure equalization or a
supporting button.
In addition to lter presses with vertical
plates, some designs also feature horizontal
plates.
Filtration Cycle (see Chap. 5). Cake ltra-
tion in lter presses involves two indistinctly
separated stages. At the start of ltration there
is no cake, and the hydraulic resistance is there-
fore very low, so that the ltration pressure can-
not rise very high at this point. The process is
similar to constant-volume ltration (see Sec-
tion 2.1.2). The pressure increases as the cake
thickness grows, until the rated pressure (of the
pump) is reached. Constant-pressure ltration
then begins, with a declining quantity of ltrate.
The exact sequence of ltration in lter presses
is still unclear and cannot be calculated, because
not enough research has been done, but a rough
estimate suggests that the highest throughput is
achieved if
t
f
=t
k
Z (70)
where t
f
is the ltration time, t
k
is the time for
lling and emptying per chamber, and Z is the
number of chambers.
On the basis of a cost calculation, the maxi-
mum number of chambers is
Z=

t
f
(C
M
C
p
)
t
k
C
p
1
2
(71)
where C
M
is the cost of mechanical equipment
(xed and movable end plates, adjusting device)
and C
p
is the cost per plate (chamber). For con-
ventional designs, Z is found to be around 60.
To determine t
f
, the total cycle time includ-
ing washing and other steps, experimental data
must be used. Unfortunately, experience has
shown that measurements in laboratory-scale l-
ter presses cannot generally be extended to arbi-
trary sizes, since there may be discrepancies in
the lling of the ltration chambers and thus in
the pressure relationships.
Filter presses are commonly charged at pres-
sures of up to 2 MPa, more in special cases. Cen-
trifugal or diaphragm pumps are preferred for
the cake ltration of highly-concentrated slur-
ries. The possibility of employing relatively high
ltration pressures, along with the mechanical
advantages in charging and discharging, lend l-
ter presses some advantages over other types of
pressure lter, even though these devices are al-
ways operated batchwise:
1) Insensitivity to feed uctuations
2) Relatively high throughputs, even with dif-
cultly lterable slurries, so that far smaller
amounts of lter aids are needed
3) Low cake moisture
4) Denser cakes that are easier to handle
5) Clear ltrate
6) Good separation of different ltrates (wash
liquor, etc.)
7) Arbitrary ltration steps (washing, steam or
air blowing, etc.)
8) Lowmaintenance costs, fewspare parts, low
depreciation
9) Simple operation
10) Comparable equipment costs per square me-
ter of ltration area
The drawbacks are obvious: Batch operation;
personnel requirements incases where cake does
56 Filtration
not drop off by itself; relatively long cleaning
times.
Applications. Even though they are batch
devices, the good washing and low cake mois-
ture available in lter presses have kept them in
wide use. Contributing factors, along with com-
paratively long service times and good reliabil-
ity, include mechanical plate movement, used al-
most exclusively for the opening and discharge
of large units; automatic lter-cloth cleaning
with high-pressure water sprays; and the avail-
ability of special plate designs. Plates up to 2.6 m
on a side give far more than 1000 m
2
of ltration
area in one press. The design of the plates and
the use of polypropylene as the main plate mate-
rial have remedied many of the shortcomings of
lter presses, such as plate fractures, corrosion,
and plate mass.
Filter presses can be employed almost any-
where, even if the mode of operation and the
equipment must be adapted to the feeds in ques-
tion. Filter presses are used successfully in ei-
ther cake or clarifying ltration, depending on
the concentration and properties of the solids; in
the latter case, lter aids are usually added, or
lter beds are employed instead of cloth media.
In the production of chemicals (e.g., dyes),
ceramics and raw materials, and in wastewater
treatment, nuclear technology and other elds,
cake ltration is applied chiey to groups Mand
S of Table 2. Slurry feed rates in these elds de-
pend on the widely varying slurry composition
and range from 0.2 1 m
3
m
2
h
1
on the aver-
age.
Figure 70. Filter press for sheet ltration deep bed
pressure lter (reproduced with permission of Seitz-Filter-
Werke)
Clarication is carried out in sheet lter
presses, which resemble plate-and-frame lter
presses in design (Figs 70, 71, and 72). Because
of the very much lower solids concentration, ser-
vice times are much longer (several hours). For
this reason, automation is commonly dispensed
with. Filtersheets include ordinary precoats and
special papers (see page 64 and Chap. 11).
Figure 71. Sheet lter press opened (reproduced with permission of Seitz-Filter-Werke)
Filtration 57
Figure 72. Clarifying lter press schematic operation
Top: Filtration with lter aids; Bottom: Sheet ltration
The most important applications of clari-
fying ltration include beer, wort, wine, fruit
juices, pharmaceutical liquids, or water steriliza-
tion (group Dof Table 2). Here too, lter loading
varies widely, depending both on the properties
of the solids being removed and their concentra-
tion in the feed and on the nature of the liquid.
They range from 0.05 m
3
m
2
h
1
for viscous
varnishes up to 10 m
3
m
2
h
1
for dilute sus-
pensions, such as beverages.
Membrane Filter Presses. In a membrane
press, the lter plate is coated with an elastic ma-
terial (rubber). After ltration, the membranes
are pressurized so that they mechanically com-
press the lter cake. In this way the cake mois-
ture is reduced by ca. 1 % 20 %, depending on
the cake compressibility. Figure 73 shows an ex-
ample of a recessed-plate membrane press. A
similar expedient is also possible in a plate-and-
frame press. The pressing either shortens ltra-
tion time or the ltration pressure can be made
smaller, which is an eminent practical advan-
tage.
Membrane lter presses can be operated at up
to 2 MPa; they produce a uniformly dewatered
cake.
Plate Press with Belt Discharge (Auto-
matic Press). This type of lter press, orig-
inally developed in Russia, combines positive
automatic cake discharge with cake compres-
sion. The horizontal lter plates are stacked and,
with the lter frames, combined into a three-
chamber system: ltrate chamber, cake cham-
ber, and high-pressure water space. The water
space is separated fromthe cake by a membrane.
For cake discharge, a cloth belt on the frame is
advanced until the cake from each plate is com-
pletely stripped off as the belt is deected into
the next lower plate (Figs. 74 and 75).
Filter Press with Stationary and Rotating
Plates. This lter contains alternating station-
ary and rotating plates on a shaft (Fig. 76). A
velocity difference is produced in the cake bet-
ween the plates, leading to breakdown of the
cake. A thin layer of lter cake, whose thick-
58 Filtration
ness depends on the lter medium employed, is
left on the ltration surface. In the extreme case
where cake formation is totally prevented, this is
pure cross-ow ltration (see Section 2.4). The
intermediate states are referred to as dynamic
ltration.
Figure 73. Membrane lter press plate assembly
a) Filter cloth; b) Slurry corner feed; c) Recessed plate;
d) Membrane recessed plate; e) Press medium; f) Cake;
g) Filtrate; h) Air or water inlet (Pressuring membranes)
If the lter cake is thixotropic, the shearing
makes it owable, so that it can be conveyed by
pressure through the system of plates. Accord-
ing to studies of some dye suspensions, this lter
can collect the same quantity of solids as a con-
ventional lter press with roughly 20 times as
large a ltration area.
8.10. Tubular Filters
Tubes as lter elements are similar in structure to
cartridges. This section deals only with tubular
lters whose dimensions do not usually corre-
spond to the concept of a lter cartridge (diam-
eter either too large or too small). This group
includes, on the one hand, tubular pressure l-
ters and, on the other, tubular membranes used
in cross-ow ltration.
Tube Filters. The tubular pressure
lter the tube press developed in England,
consists of a pressure tube with a lter tube in-
serted in it. After cake ltration in the annular
space, a membrane placed inside the pressure
tube compresses the cake. A pressure of up to
10 MPa can be applied. Next (or immediately
after ltration), the cake can be washed. For cake
discharge, the lter tube is swung downward and
the cake is loosened from the ltration surface
with compressed air. This lter has the advan-
tage of the cylindrical ltration chamber, which
is economical for relatively high pressures.
In another type of tubular pressure lter, l-
tration and compaction are similar but the tube
is positioned horizontally, so that the cake can
be scraped out of the lter tube.
A third type of tubular pressure lter has a
cylindrical rotor inside the lter tube. Slurry is
continuously compacted in the annular space.
The solids are retained in the annular space on
the ltration surfaces of the tube and the rotor,
where they are subjected to shear. In this way,
the cake is caused to move downward to the dis-
charge. As in the lter press with rotating plates
(see preceding section), the lter cake must re-
main owable after substantial dewatering; that
is, it must be thixotropic. In one or more washing
zones, wash liquor can be fed radially through
the lter tube. Several lter sizes with ltration
areas from 0.25 1.6 m
2
are tested.
Cross-Flow Tubular Filters. The lter ele-
ments in the tubular cross-ow lter are small
tubes or hoses, a few millimeters in diameter,
which are bundled into systems of tubes carry-
ing parallel ows (Fig. 77). The ltrate passes
the system through walls of the tubes whereas
the suspension ows tangentially to them drag-
ging the retained particles, see Chapter 2.4. The
Filtration 59
Figure 74. Automatic lter press systemic processing
a) Cloth drive; b) Cloth wash; c) Cake removal; d) Pressure control; e) Filter cloth; f) Filter cloth aligning assembly; g) Ten-
sioning device; h) Yoke plate; i) Yoke anchor; j) Filter plate with inatable membrane; k) Lifting table; l) Hydraulic closure
device; m) Filter preframe
tubes can be made of various porous materi-
als, preferably polymers or ceramics, and have
quite narrow pore-size ranges, from a few mi-
crometers down to hundredths of a micrometer.
They are, above all, microlters for the concen-
tration of dilute suspensions and for clarifying
and sterilizing ltration. Such lter systems with
the nest membrane pore sizes are also suitable
for ultraltration and reverse osmosis; multi-
layer or asymmetric membranes are usedinthem
(Membranes and Membrane Separation Pro-
cesses).
Throughput rates range up to
3000 Lm
2
h
1
, depending on pore size and l-
tration pressure. The possible applications vary
with capacity. The principal users are chemical,
pharmaceutical and foodstuffs producers and,
increasingly, bioengineering concerns.
Figure 78 shows an example of a complete
microltration unit including cross-ow ele-
ments and the programmed control unit for batch
and continous operation.
8.11. Special Filter Types
Coarsely-dispersed systems with relatively high
solids contents are dewatered with conventional
screens like the vibrating screens used for classi-
cation. One such lter is the vertical sieve con-
taining a bent screen plate, which is employed
to separate coarse and brous solids at lower
concentrations from process waters and waste-
waters with screen ltration. The sieve, which
is equipped with an adjustable screen slope, is
charged from the top, so that the slurry ows
60 Filtration
Figure 75. Automatic lter press (reproduced with permission of Eberhard Hoesch & S ohne)
downward by gravity. The screen, with slots
varying from 0.125 2.5 mm wide, retains the
solids, which drop downward. The simple, in-
expensive design has made this device a widely-
used one in food preparation, the textile and
manmade ber industry, paper and pulp man-
ufacturing, and the chemical industry.
Several new ltration mechanisms are
being explored with an eye to ltration
problems that have not yet been solved.
These include, in particular, the dewatering
of suspensions having relatively high con-
centrations of extremely ne solids (below
10 m down to colloidal sizes). The conven-
tional separation techniques occulation and
microltration generally yield lter cakes or
concentrates that require considerable energyfor
complete phase separation (e.g., by drying). Re-
search on ionized phases has shown that the elec-
trokinetic effects of electro-osmosis for liquid
transport and electrophoresis for solids transport
along an electric eld allow separation, even of
difcultly lterable suspensions. Both these ef-
fects have been known for a long time.
A lter based on the above processes con-
sists of separating chambers with anodes, where
the ionized solids collect, and cathodes made as
lter media, where the electrolytic ltrate is dis-
charged [45]. Several separating chambers form
lter systems similar to lter presses.
Difcultly lterable suspensions are also
ltered by high-gradient magnetic separation
(HGMS) (Magnetic Separation). This type
of lter employs a steel wool medium. In a
very strong magnetic eld, high eld gradients
are generated at the individual laments. These
gradients attract ne magnetic particles. This
HGMS lter thus resembles a depth lter. Suc-
cesses have been achieved in the purication of
kaolin and wastewaters.
Figure 76. The dynamic lter press
a) Filtrate discharge pipe; b) Main shaft; c) Filter plate;
d) Agitating disk; e) Cake discharge valve
Filtration 61
Figure 77. Schematic of tubular membrane lter
a) Tube bundle; b) Cross section
Figure 78. Microltration equipment (Seitz-Microow, re-
produced with permission of Seitz-Filter-Werke)
9. Filter Selection
Because certain qualities, largely related to de-
sign, characterize the ranges of application of
lters, some guidelines can be set forth. These
together with a few other points can provide a
basis for lter selection. In general, a good se-
lection takes account of factors such as slurry
properties, throughput, or cake formation rate
(see Table 2). Other points to be considered are
the density and viscosity of the liquid, the size,
size distribution and shape of the solids parti-
cles, the sensitivity of the product to impurities,
corrosion problems, the sensitivity of the equip-
ment to sedimentation and to product uctua-
tions, the cost of operation, and so on. Table 5
can be used in the selection of special vacuum
lter types under a number of criteria. The tabu-
lated data should, however, be regarded only as
rough guidelines. Filter types are shown in Fig-
ure 79 together with the particle sizes for which
they can be used.
Figure 79. Filter selection as a function of solids particle
sizes
The higher cost of pressure ltration in com-
parison with vacuum ltration is usually not
justied except when the cake has a resistance
of 3.5 10
11
m/kg. Flocculated solids, which
should not lose their structure since it is advan-
tageous for ltration, require careful handling,
especially in feeding to the lter. Because most
such solids are somewhat compressible, not too
much pressure should be applied, and so vac-
uum ltration is commonly preferred for these
materials.
A safe selection without trials or appropri-
ate experience is possible only in the rarest in-
stances.
Little general information and few computa-
tional formulas can be given in the eld of l-
tration, which is one of the oldest operations in
process engineering. Regrettably, ltration (like
other essential processes involving the handling
of bulk solids, dumped and packed beds, etc.)
must deal with the effects of material proper-
ties that can be determined only with difculty
or cannot be measured in a predictable way, and
these effects are so large and pervasive that there
is no comprehensive, unied way of allowing for
62 Filtration
Table 5. Applicability of lters
Filter type Suspension of Operation
ne particles Pressure Washing Compression Continuous (c)
automatic (a)
Belt c
Belt with compression c +a
Nutsche a
Tilting pan c
Leaf/plate a
Drum c
Pressure drum c
Membrane lter press a
Twin-belt pressure lters c
Tube press a
them. There is no doubt, however, that selection
criteria exist and that they imply at least certain
lter categories, so that in many cases the num-
ber of open options is not very large.
10. Filter Media
The outcome of ltration, that is, the degree to
which the lter separates the two phases, is gov-
erned above all by the lter medium. Where the
available lter media are not suited to the prop-
erties of the phase (solid) being separated, lter
aids can close the gap.
The lter medium should be selected to pro-
vide the desired separating effect and to t the
special properties of the suspension (such as the
particle size of the solids and the viscosity of
the liquid). The selection of a medium should
also take account of service conditions such
as cleaning, washing, reaction to back-blowing,
and chemical and mechanical stability, which
are no less important than the slurry properties.
In cake ltration, cake removal must also be con-
sidered.
In line with the multiple purposes of ltra-
tion, media must perform the following tasks:
1) Optimal retention of all solid particles out-
side the lter medium
2) Retention of the solid particles, mainly in-
side the lter medium, by virtue of effects
generally lumped together as entrapment,
which essentially have to do with adhesion
forces and electrical forces.
Often these functions cannot be unambigu-
ously separated; this is particularly the case with
slurries having a very broad particle-size spec-
trum. In such cases, some concession nearly al-
ways has to be made in terms of separation ef-
ciency, or further operating practices, such as
careful cleaning or frequent replacement of the
lter medium, have to be introduced.
Filter media can be characterized in terms of
the following overall criteria:
1) Cut size, that is, the particle size that just
passes through the lter medium.
2) Permeability; high permeability means low
pressure drop.
3) Chemical stability with respect to the ltrate.
4) Blocking tendency, particularly for fabrics in
cake ltration.
5) Mechanical strength in relation to loads im-
posed in back-blowing or the movement of
lter cloths.
6) Smooth surface to promote cake removal.
As is appropriate after many years experi-
ence in ltration, a great many types of lter
media have been introduced in practice. New
materials and processing methods continue to
offer a wide range of improvements. The diver-
sityof lter-media properties has ledtoa number
of classications in the literature. The following
seems to be an advantageous one:
Perforated plates, screens, slotted screens
Fabrics
Beds, felts and other nonwoven materials
Packed beds and precoat layers
Porous materials
Membranes
Within these groups, there are differences in
properties, especially with regard to chemical
Filtration 63
stability(includingaging). However, oftenphys-
ical and biological requirements are also im-
posed, such as high-temperature stability, wa-
ter uptake, swelling capacity, weathering behav-
ior (stability to light), and resistance to insects
(moths) and bacteria (rotting).
Aside from material properties, the overall
ltration qualities can be represented only in an
approximate way, because they depend heavily
on the properties of the suspension being l-
tered. Even the often-used permeability and/or
mean pore size gures should be considered at
best rough reference values for the behavior
during ltration. A reliable assessment is usu-
ally possible only if practical tests or, within
limits, laboratory trials (e.g., with the test leaf
lter, Chap. 7) are performed.
Perforated Plates, Screens, Slotted
Screens. Aside from screens used in screen
ltration, lter media in this group serve mainly
to support lter cloths or papers, which as a rule
have ner pores.
With perforated plates, the largest possible
free surface area is usually desired so that the
pressure differences to be overcome in ltra-
tion will be small. Free areas range from some
20 30 % for ordinary plates with circular holes
toover 90 %for high-qualityscreens withsquare
openings in the millimeter range. The free area
drops to less than 10 % for very ne sieves with
openings a few micrometers across.
Slotted screens with very small openings,
used chiey as strainers and candle lters (see
Section 8.3), are made by stacking plates, rings
or other elements around a permeable cylinder.
The ne gaps between the elements allow only
the liquid to pass. Strainers can be fabricated
more simply by winding proled (wedge) wires
or strings on a core. These units have little free
area, roughly 10 %. As a result, they are eco-
nomic only when solids concentrations are very
low.
Fabrics. By far the most widely used lter
media are woven fabrics. Manmade bers dom-
inate, but bers of glass, minerals, and metals
are also used to an increasing extent, especially
in view of their good chemical stabilities.
The advantages of multilament-yarn fabrics
include the depth action of the lter medium,
which permits slight penetration of solids into
the medium, thus effecting some reduction in
pore size and also promoting bridge formation
on the medium. The ease of working with fabrics
is an incomparable advantage, of course.
The fabric weave is important for ltration
proper as well as service properties. Another es-
sential factor is the structure of the ber, which
varies widely in natural and synthetic materials.
Where particularly high strengths are needed,
as in twin-belt pressure lters, monolament-
yarn fabrics are preferred. High precision in fab-
rication allows mesh openings in the micrometer
range, especially with metal bers.
Within some limits it is possible to start with
typical fabric data, such as weight per unit area,
numbers of warp and lling yarns, ber diame-
ter, fabric thickness and ber density, and calcu-
late the pore size and hydraulic resistance. The
resistance calculation is based on the pore radii
of the fabric and the yarn, and Poiseuilles law
is used as a rst approximation.
The permeability of a clean, unused fabric
can be calculated (with about the same reliabil-
ity) from Kozenys equation (66); the starting
data are the densities and porosities of the yarn
and the fabric. But empirical relations should
give better agreement with experimental values.
However, it is vital to know how the fab-
ric inuences cake formation and the narrowing
and possible blocking of the pores. The result-
ing change in hydraulic properties of the lter
medium is described in Section 2.2.
Nonwovens, Papers, and Felts. These l-
ter media consist of relatively short bers (sta-
ple bers a few centimeters long) formed into a
random layer and then compacted in some way.
Nonwoven media are usually fabricated in fairly
thick layers and therefore function as depth l-
ters (Nonwoven Fabrics).
Papers made of cellulose bers are available
in a bewildering range of grades for use in lab-
oratories, small plants, and daily life (Paper
andPulp). Heavier papers andpaperboards (2 6
mm thick) can also be fabricated from mixed
materials, such as cellulose plus diatomaceous
earth or activated carbon. These lter materials
meet a wide range of requirements and are em-
ployed in large quantities for the clarication of
beer, wine, or juices, and in the dyes and coat-
ings, food, and cosmetics industries. By virtue of
the rawmaterials and fabrication processes, one-
64 Filtration
time use of the lter medium is often economic;
this feature is especially desirable in sterile l-
tration.
Conventional felts of relatively short natural
and manmade bers are usually not adequate to
the ltration and service loads in plant equip-
ment. Felts that incorporate woven-fabric rein-
forcement are often preferred for this reason.
Some loss in terms of the basic advantage of felt
must be accepted when these composite materi-
als are used. As a rule, needled felts have lower
resistances than woven fabrics. They have now
become the most commonly employed nonwo-
vens (Felts).
Packed Beds and Precoat Layers. Loose
beds of granular materials can also be used
in clarifying and some cake ltrations. For eco-
nomic reasons, cheap natural products such as
sand, gravel, or coal are used, but it is desirable
to remove the excessively ne-grained fractions
from them by washing. Common particle sizes
are 3 50 mmfor gravel, 0.35 0.45 mmfor ne
sand, 0.45 0.55 mm for medium sands, and
over 0.55 mm for coarse sands. Other candidate
materials include minerals (quartz, limestone),
coke, anthracite, slag, wood pulp, glass beads,
wadding, or bers.
The hydraulic properties are calculated with
the relations developed in the theory section,
in particular Darcys law, Section 2.1.1, or if
the appropriate material data are known the
Kozeny equation (66).
Rigid Porous Materials. Porous materials,
which are fundamentally similar in structure,
are suitable for any kind of ltration. Their
mechanical stability allows their use as lter me-
dia without supporting structures (up to pres-
sures of 2 MPa and more). They are preferably
employed as lter cartridges or in the form of
sheets and more recently, honeycomb and other
three-dimensional shapes. Because they are gen-
erally too valuable to use just once, and be-
cause they can withstand vigorous backwash-
ing without special preparation, solids should
be collected mainly by surface ltration. Deep-
bed ltration is recommended only when the
particles are very small in comparison with the
pores and are collected by an adsorptive mecha-
nism. In the intermediate region, solid particles
block the pores too quickly, and backwashing
may not adequately remove them. Materials for
porous lters include minerals, diatomaceous
earth, porcelain, clay, coal, graphite, corundum,
rubber, plastics, and metals, including highly re-
sistant nickel chromium molybdenumalloys.
The most important criterion for selection is
chemical and thermal stability.
Porous materials made from equal-sized par-
ticles are distinguished by uniform porosity. In
order to reduce the blocking tendency, layers
withdifferingparticle sizes canbe superimposed
with the ne-pored layer toward the lter cake;
thus only the topmost layer need be cleaned, and
the lter offers better service qualities.
Porcelain lters are available with pore radii
from 0.08 to 6 m and porosities of 30 70 %.
Sintered metal lters range from 1 to over 100
m in pore size and from 30 to 60 % in poros-
ity. Plastic lter media with about the same pore
spectra can easily be tted to the equipment.
Their relatively poor temperature stability, how-
ever, restricts their use. Besides poly(vinyl chlo-
ride) and polyethylene, porous masses can also
be fabricated frompolytetrauoroethylene, with
its excellent stability.
Membranes. At present a great number of
distinct membranes are on the market, not just
for ultraltration and reverse osmosis, but also
for microltration ( Membranes and Mem-
brane Separation Processes). They are made of
plastics and are suitable for clarifying as well
as pure surface ltration. If very small pores are
needed, membranes are assembledfromtwolay-
ers, a coarse-pored supporting layer (with, e.g.,
15 20 m pore openings) topped with a lter
layer (0.2 m pores). All membranes have a rel-
atively narrow range of pore sizes.
A special form of membrane ltration em-
ploys hollow bers made of, e.g., polyamides,
with diameters of 50 m and wall thicknesses
of 13 m. The solution being separated is intro-
ducedintothe hollowbers under anappropriate
pressure. Pure liquid escapes through the wall of
the ber, while the dissolved material remains
behind and can easily be carried out of the ber.
11. Filter Aids
Some suspensions contain solids that quickly ll
or block the openings in the lter medium. This
Filtration 65
group includes suspensions substantially made
up of relatively ne, soft (especially gelatinous),
compressible solids. Such feeds can still be l-
tered well if the blocking of the lter medium is
prevented by the inclusion of inert, readily l-
terable granular materials. These lter aids form
a layer permeable to the ltrate, and at the same
time they carry out the functions of a loose lter
cake. The particles being retained are deposited
on the lter aid. This process must still leave
sufciently free openings through which the l-
trate can pass. The main candidates for lter aids
are thus highly porous materials. Diatomaceous
earths are very economical examples; their ne-
ness is characterized by 10 %>40 m at the
high end and between 1 and 7 m at the low
end. Also used are perlites and, for special jobs,
Fullers earths, powdered glass, coal prepara-
tions, cellulose bers, wood pulp, paper stock,
bagasse (a sugar cane residue), talc, or plastics,
although these products are less effective than
diatomaceous earths and perlites. The require-
ments for a good lter aid are uniform quality,
correct particle size and shape, and the highest
possible wet volume (specic volume in the wet
condition, usually greater than the wet volume
per mass of lter aid).
In diatomaceous earths, particles consist-
ing of needle-shaped diatoms have proved ad-
vantageous; disk-shaped particles lter poorly
and produce too dense a cake. Unsuitable con-
stituents such as clay, or sand must be removed
from the raw materials, because they hinder l-
tration by blocking the pores.
A synthetic lter aid is manufactured from
volcanic rocks. The raw material, bound with
some water, is ground, then suddenly heated to
a high temperature; the grains inate to form
spheres. Regrinding yields half-shell-shaped
fragments, whose large surface area makes them
suitable as a lter aid once the nes are screened
out. Their action in ltration is somewhat worse
than that of diatomaceous earth, with a porosity
20 40 % lower.
In precoating, a layer of lter aid several mil-
limeters thick is deposited on the lter medium
before ltration proper. After the ltration is
completed, the lter aid is often discarded along
with the collected solids.
Several steps take advantage of the effect of
the loosened lter cake. First, a precoat some
1 2 mmthick, containing 200 800 g/m
2
of l-
ter aid, is ltered onto the medium (base layer)
so that a clear ltrate can be obtained right from
the start of ltration. Filter aid is also added to
the slurry continuously during ltration. In batch
operation, this is done by stirring in the slurry
tank; in continuous ltration, dry or wet lter
aid can be metered into the slurry delivery line.
No generally valid rules exist for the proper ra-
tio of lter aid to solids concentration. A variety
of empirical values are available for some pro-
cesses and solids types. If there is too little lter
aid, the ltration goes badly; if too much, the
ltrate ow rate drops off too sharply.
The main applications for ltration with lter
aids are the precoat technique for poorly lter-
able slimes of chemical and mineral products
and the clarifying ltration of beverages (beer,
wine, juices) containing soft, gelatinous impuri-
ties, as well as the clarication of gelatins, cane
juice, or edible oils.
12. References
1. H. Darcy: Les fontaines publiques de la ville
de Dijon, Paris 1856.
2. VDI-Richtlinie Filtrierbarkeit von
Suspensionen VDI 2762, 1997.
3. L. Svarovsky: Solid Liquid Separation,
Butterworths, London 1977, p. 175.
4. M. Tiller, T. Cleveland, R. Lu, Ind. Eng. Chem.
Res. 38 (1999) 590 595.
5. P. H. Hermans, H. L. Bred ee, Rec. Trav. Chim.
54 (1935) 680 700.
6. A. Rushton, A. S. Ward, R. G. Holdich:
Solid Liquid Filtration and Separation
Technology, VCH Verlagsgesellschaft,
Weinheim 1996, pp. 64 ff.
7. K. Luckert, Wiss. Z. Techn. Univ. Magdeburg
36 (1992) no. 5/6, 74 80.
8. P. L. Boucher, J. Proc. Inst. Civ. Eng.
(1946 1947) 415 445.
9. A. Rushton, A. S. Ward, R. G. Holdich:
Solid Liquid Filtration and Separation
Technology, VCH Verlagsgesellschaft,
Weinheim 1996, pp. 177 ff.
10. K. J. Ives, in A. Rushton (ed.): Mathematical
Models and Design Methods in Solid-Liquid
Separation, NATO ASI series E No. 88,
Martinus Nijhoff, Dordrecht 1985, pp. 90 ff.
11. A. Rushton, A. S. Ward, R. G. Holdich:
Solid Liquid Filtration and Separation
Technology, VCH Verlagsgesellschaft,
Weinheim 1996, p. 198.
66 Filtration
12. G. Dierckx in: La Filtration Industrielle des
Liquides, Tome III, Soci et e Belge de
Filtration, Li` ege 1978, pp. 295 314.
13. A. Palinski, W. Uhl, R. Gimbel, Vom Wasser
89 (1997) 175 189.
14. J. Altmann, S. Ripperger, J. Membrane Sci.
124 (1997) 119 128.
15. S. Ripperger, J. Altmann, Filtrieren Separieren
11 (1997) no. 3, 109 113.
16. W. Bender, Chem. Ing. Tech. 55 (1983) no. 11,
823 829.
17. K.-H. Steiner, Chem. Ing. Tech. 61 (1989) no.
1, 1 8.
18. H. Anlauf, Fortschrittsber. VDI, Reihe 3, 114
(1986).
19. H. Schubert: Kapillarit at in por osen
Feststoffsystemen, Springer Verlag,
Heidelberg 1982, p. 184.
20. H. Anlauf, Maschinenmarkt 95 (1989) no. 8,
26 30.
21. R. J. Wakeman: Cake Dewatering, in L.
Svarovsky (ed.): Solid Liquid Separation,
Butterworths, London 1977, p. 300.
22. R. J. Wakeman, Filtr. Sep. 16 (1979) no. 6,
655 669.
23. A. Rushton, A. S. Ward, R. G. Holdich:
Solid Liquid Filtration and Separation
Technology, VCH Verlagsgesellschaft,
Weinheim 1996, pp. 348 ff.
24. D. Redeker, K.-H. Steiner, U. Esser, Chem.
Ing. Tech. 55 (1983) no. 11, 829 839.
25. I. Nicolaou, W. Stahl, Aufbereit. Tech. 33
(1992) no. 6, 328 338.
26. J. Walker, Spektrum Wissensch. 12 (1986)
188 197.
27. W. Tiedemann, Fortschrittsber. VDI, Reihe 3,
453 (1996).
28. M. Shirato, T. Murase, E. Iritani, S. Nakatsuka,
Filtr. Sep. 24 (1987) no. 2, 115 119.
29. A. Rushton, A. S. Ward, R. G. Holdich:
Solid Liquid Filtration and Separation
Technology, VCH Verlagsgesellschaft,
Weinheim 1996, pp. 455 ff.
30. D. Leclerc, S. Rebouillat, in A. Rushton (ed.):
Mathematical Models and Design Methods in
Solid Liquid Separation, NATO ASI series E
No. 88, Martinus Nijhoff, Dordrecht 1985, pp.
356 ff.
31. J. Kozeny, Ber. Math.-naturwiss. Abt. Akad.
Wien. (1927) 271 306.
32. A. E. Scheidegger: The Physics of Flow
Through Porous Media, 3rd ed., University of
Toronto Press, Toronto 1974, pp. 137 ff.
33. H. Rumpf, A. R. Gupte, Chem. Ing. Tech. 43
(1971) no. 6, 367 375.
34. J. Gregory in K. J. Ives (ed.): The Scientic
Basis of Flocculation, Sijdhoff & Noordhoff
Alphen aan de Rijn 1978, p. 91.
35. H.-J. Jacobasch, P. Weidenhammer, Chem.
Ing. Tech. 68 (1996) 1590 1594.
36. D. Houi, R. Lenormand, in F. Family, D. P.
Landau (eds.): Kinetics of Aggregation and
Gelation, Elsevier, Amsterdam 1984.
37. P. Schmitz, B. Wandelt, D. Houi, M.
Hildenbrand, J. Membrane. Sci. 84 (1993)
171 183.
38. W. H oinger, C. St ockelmayer, A. Hackl, Filtr.
Sep. (1994) Dec., 807 811.
39. W. H oinger, C. St ockelmayer, Staub
Reinhalt. Luft 55 (1995) 423 428.
40. W.-M. Lu, C. C. Lai, K.-J. Hwang, Sep.
Technol. 5 (1995) 45 53.
41. Filtrierbarkeit von Suspensionen,
VDI-Richtlinie 2762 (1997).
42. H. Anlauf, Filtrieren Separieren 8 (1994) no.
2, 63 66, 69 70.
43. H. Anlauf, Filtrieren Separieren 8 (1994) no.
3, 116 118, 121 126.
44. W. G osele, Filtrieren Separieren 9 (1995) no.
1, 14 22.
45. F. M. Tiller, C. S. Yeh in H. S. Muralidhara
(ed.): Advances in Solid Liquid Separation,
Batelle Press, Columbus OH 1986, pp. 1 37.