104 Solid/Liquid Separation: Scale-up of Industrial Equipment
Area, a
I 1^
*e» £e
/,£
::: M
1
When the grains are in equilibrium:
Upward force = Downward force
Upward force = Pressure difference x Area
= pgha
where h is the loss of water head.
Downward force = Weight of the grains in water
= al,{\ -£e)(Ps-p)g
Equating the upward and downward forces gives:
pgha = al,(l-£,)(p,-p)g
Rearranging gives an expression for the head loss gradient in a
fluidised bed:
(4)
/„ P
The head loss, or hydraulic gradient is also given by the Kozeny-
Carman equation (3) with kc = 5:
h _ V ju{l-^,y 36
(5)
S P < i//'d
3 • Deep bed filters 105
Equating equations (4) and (5) gives:
36
2j2
(6)
P g P y/'d
which can be rearranged to give an expression for the minimum
fluidising velocity, v^f. At the point of incipient fluidisation, that is, as
the bed just starts to expand the voidage/porosity and the bed depth are
equal to those for the fixed bed, so we have an expression from which
we can calculate the minimum fluidisation velocity:
g {Ps-p)s'y/^d^
(7)
180 ^ (1-.)
From equation (7) it is clear that the minimum fluidisation velocity
depends on media grain size, media density, media voidage and packing,
and temperature (as it affects the water viscosity). Changes in any of
these parameters affect bed expansion and fluidisation conditions.
If we examine a granular bed subjected to water upflow then we see
that as we increase the flow rate (velocity) the head loss increases
linearly with velocity while the bed remains fixed, that is, Darcy's Law
is obeyed.
Once we have reached v^f, then the bed starts to expand upwards with
the flow and the head loss becomes constant. The head loss is constant
as there is no more loss of energy as drag on the grains. All the energy
is used up in supporting the grains, that is, balancing their weight.
Typical fluidisation curves are shown in Figure 3.14. This provides a
simple, experimental method, using a laboratory column, for deter-
Q.
0
•o
•o
0
00
Vmf Vmf Water velocity
Figure 3 J 4 Typical fluidisation curves for an ideal monosize medium. Graded media results in a
gradual change in gradient on both graphs
106 Solid/Liquid Separation: Scale-up of Industrial Equipment
mining v^f and the expansion characteristics for a particular granular
material under specific conditions.
3.5.2.2 Calculating head loss
To begin with, consider the simplest case of a single media filter.
Equation (3) gives the head loss gradient in the fluidised bed; if the
backwash head loss is to be calculated then /^ and s^ must be
eliminated from the equation as they are not known and not easily
measured.
During backwashing of a fluidised bed the total volume of solids is
unchanged, so that:
Substitute this result into equation (4) to obtain:
h = ^l^^)i^iZfli (8)
P
For a typical filter sand the voidage, s = 0.4, and the density, p^ = 2650
kg m~^; putting these values into the equation gives:
, 0.6x1650, ,
1000
where the head loss (h) and the bed depth (/) are measured in metres
(m) of water.
This is very useful to remember for sand filters, but it should be
remembered that this only applies to sand media where the grain
density is 2650 kg m~^. For other media materials different values of £
and Ps will need to be used in equation (8).
Research has shown that water only backwash is a weak cleaning pro-
cess as grain collisions are minimal in a liquid fluidised bed. The film
of water flowing around the sand grains is thought to protect them
from collisions as it thins (Amirtharajah, 1978). It is often inadequate
for preventing mud ball formation and other filter problems if the
deposits in the filter are so inclined. A minimum expansion is required
in order to flush out the deposits and get the whole bed mobile.
Increasing the bed expansion has not been shown to improve media
cleaning.
3 • Deep bed filters 107
3.5.2.3 Use of air scour
When air scour precedes the water wash, it is intended to break up and
detach deposits so that they are ready to be flushed out by the fluidising
water.
Research has shown that whilst air scour causes a lot of agitation to
grains in the top few centimetres of the bed, there is very little
agitation deep down (Amirtharajah, 1984). The agitation is caused by
bubbles erupting at the surface of the media. If a filter bed retains most
of its deposits at the surface then air scour alone is adequate. If,
however, it acts as a deep bed filter as intended, then deeper deposits
may not be dislodged. Air scour followed by a water wash is common
in the UK.
3.5.2.4 Combined air and water wash
A combined wash means that the air and the water flow simultaneously
up through the bed. At particular combinations of air and water flow
rates for a given medium a phenomenon known as collapse-pulsing
(Amirtharajah, 1993) is observed. Later research has shown this to be at
flow rates around the onset of three phase fluidisation (Hemmings and
Fitzpatrick, 1997). What this means in practice is that fluidised bed
conditions are achieved for a much lower water rate - typically less than
0.5 v^f. In addition, during collapse-pulsing there is a very high degree
of bed agitation, as pockets of air form and collapse within the media.
Table 3.1 illustrates typical minimum fluidisation velocities, velocities
for 20% expansion (V200/J and collapse-pulsing air and water rates for
some common filter media. These values were obtained experimentally
in the laboratory for water temperatures in the range 11-14^C.
Collapse-pulsing conditions, obtained experimentally, are illustrated in
Figure 3.15.
Table 3.1 Backwash parameters for selected media.
Media Vmf ^200/0 Collapse Va Pulsing (C-P)v^ Attrition at C-P
(m h-1) (m h-^) (m h-^) (m h-^) (0/0 loss)
F400 GAC 10 17 36 4 1.7-5.3
F300 GAC 16 31 36 10 4.5 - 7.3
R0W.8 GAC 22 41 36 9 2.1 -3.5
Anthracite No.2 28 50 30 21 0.5 - 3.2
Garnet (1.4-2.36) 110 170 80 80 not measured
Sand (0.5-1 mm) 18 45 40 15 not measured
Sand (1-2 mm) 52 89 24 43 0.1 -2.3
108 Solid/Liquid Separation: Scale-up of Industrial Equipment
10 20 50 60
Air Rate m/h
Figure 3.1 5 Air and water flow rates giving collapse-pulsing conditions
for a range of typical UK filter media (Humby and Fitzpatrick, 1996)
3.5.2.5 Cleaning mechanisms
When a filter is being backwashed, the mechanisms that actually get
the filter grains clean include fluid shear forces, grain collisions, abrasion
between grains, and forces associated with the air/water interface.
Cleaning mechanisms have been discussed further by Fitzpatrick (1993).
3.5.2.6 Media attrition and loss during backwash
Media loss may result from inappropriate wash rates for the type of
media in use. It is usually due to the lighter grains being carried out
over the wash water weir with the backwash water. Lower density
materials are washed out more easily, despite lower backwash rates.
This is due to their density being closer to that of water and may be
exacerbated by attachment of small air bubbles.
Attrition of more friable filter media may result from a vigorous
backwash, particularly if air scour is used. Table 3.1 shows the loss due
to attrition of various filter media after 100 hours of combined air and
water backwash. More details can be found in Humby and Fitzpatrick
(1996). Further work on backwash flow rates and media attrition for
biological aerated filter (BAF) media used in sewage treatment has
been reported by Kent et al (1996).
3.5.2.7 Backwashing of dual and triple media filters
Dual and triple media filters are usually designed so that they remain
stratified after backwashing. This is not usually a problem for anthracite
3 • Deep bed filters 109
and sand. If garnet or ilmenite is used as a third layer the fine garnet or
ilmenite may mix with the sand. This may not be detrimental to filter
performance but further research is required on this subject. Research
has been done by Brown et al (1996) on the mixing of the fine garnet
and sand layers. Mixing definitely occurs between the sand and garnet
combinations found in practice at treatment works.
3.6 Filter media
3.6.1 Media types
The most commonly used filter medium is silica sand. As dual and
triple media filters were developed then anthracite and garnet or
ilmenite were used in addition to sand. More recently granular
activated carbon (GAC) has been used in combination with sand to
replace anthracite. GAC performs the dual roles of filtration of
particulates and adsorption of organics. Crushed recycled glass is now
being marketed for filtration. This is a waste product resulting from
glass recycling at bottle banks. It has been shown to have similar
filtration performance to sand but its grains have more angular shapes
than most filter sands. Crushed recycled glass presents a potentially
viable alternative to sand in filtration but further work is required to
assess the potential leaching of metals that may be present in the glass
before it will be approved for use in drinking water treatment. Also,
different sources of glass may have different densities which affects
their performance during fluidisation.
Other media being used in drinking water filtration include crushed
pumice and granular ferric hydroxide. For wastewater filtration a wider
range of materials can be used, including Lytag, pulverised fuel ash
(PFA), china clay products and a variety of plastic beads. Such media
may be used in biological aerated filters (BAFs) as well as in
conventional filters. Further details can be found in Kent et al (2000).
3.6.2 Media testing
Although the ultimate test of the suitability of any granular filter media
will be its use in full-scale or model deep bed filters, a preliminary
assessment can be made by examining certain properties of the media.
No British Standard exists but the basis of a specification has been
published which allows assessment of media other than sand, for
example: coal, pumice, plastics, ceramics, garnet, ilmenite, alumina
