1 s2.0 004313549290155W Main

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Waz. Res. Vol. 26. No. 12. pp. 1571-1.575, 1992 0043-1354/92 $5.00+ 0.

00
Printed in Great Britain.All rightsreserved Copyright '~ 1992Pcrgarnon Press Lid

SUSPENDED PARTICLE SIZE DISTRIBUTION AND THE


PERFORMANCE OF DEEP BED FILTERS
R. I. MACKIE~) and P~NBI BAI
Department of Civil Engineering, The University, Dundee DDI 4HN, Scotland

(First received November 1991; accepted in ret'ised form May 1992)

AbstractaThe paper examines the importance of suspended particle size distribution in deep bed filtration.
Expariments were conducted with suspensions of different size distributions. The changes in total
concentration, concentrations of particles of different sizes and the headloss were observed. The
suspension with the largest mean particle size gave the greatest removal. It was also observed that the
size distribution changed with depth and time. and this was shown to be crucial in understanding filter
behaviour. It was concluded that in modelling filter performance it is not sufficient to use the mean
characteristics of the influent suspension. Some evidence was found for the interaction between particles
of different sizes, and the particle size distribution also affects the distribution of deposit throughout the
bed.

Key words---deep hod filtration, water clarification, particle size

INTRODUCTION particle size. For instance, they found that the removal
of large particles stopped while the removal of small
In deep bed filtration suspended particles are removed ones was still increasing. They also carried out some
by passing the fluid through a granular bed. The most theoretical work that indicated that there is interaction
common use of the process is in final removal of between particles of different sizes in the removal
particles in water treatment. The process has received process.
much attention over recent years (e.g. Ives, 1970; There is, therefore, substantial evidence that par-
Tien and Payatakes, 1979) but is still far from being ticle size distribution may be important in deep bed
fully understood. In particular, while it has long been filtration, but there is a scarcity of experimental data,
recognized that particle size is important (Ison and and the evidence so far is to some extent contradictory.
Ives, 1969) little attention has been paid to particle This paper describes some recent experimental results
size distribution. Most research has either dealt with on the effect of particle size distribution. The work
monodisperse suspensions, or has characterized the used polydisperse suspensions and suspensions with
suspension by the mean particle size of the influent. different size distributions were prepared using various
For instance, Yao et al. (1971) showed that there is pretreatment methods.
a critical particle size of about i F m for which the
removal efficiency is a minimum. For smaller panicles EXPERIMENTAL METHOD
diffusion plays an increasing role, and for larger par-
Filter, media and packing
ticles gravity and interception become more important.
O'Melia and All (1978), using monodispcrse suspen- The filter consisted of a 138 mm internal diameter perspex
column with brass inlet and outlet units. The whole filter
sions, showed that the influence of size continued was fitted on a vibrating table, this table being used for
throughout the filtration process, and that headloss compaction. Ballotini glass beads were used as the filter
was higher for suspensions of smaller particles. media, these are spherical and the size can be closely
Recently, some attention has been paid to bimodal controlled. The media used in the experiments reported here
was 0.600-0.710 mm diameter. The bed was prepared by
suspensions. Vigneswaran et al. 0985, 1990) found filling the filter with water. The media was immersed in
that the presence of larger particles tended to increase water and fed via a hydraulic raiser in the filter. When every
the removal efficiency of smaller particles. However, third of the media had been added the filter was vibrated
Darby and Lawler 0990) stated that particles of one for one minute. This method was found to give the most
size apparently preferred to be captured by particles uniform filter bed. The porosity of the clean bed was 0.371.
of a similar size. Tobiason et al. (1990) found that Suspension preparation
the removal efficiency of the smallest particles was The suspension used in this was a PVC powder (Corvic
reduced by the presence of larger particles, and the P72/757, supplied free ofcharge by ICI, U.K.). The particles
presence of smaller particles improved the removal of were spherical and their size ranged from less than 0.5
to about 15/~m. Three types of suspension were prepared,
larger ones. Mackie et al. (1987, 1989) used poly- the object being to vary the size distribution.
disperse suspensions and showed that the qualitative Type A. This was the standard suspension. 22.5 g of PVC
as well as quantitative behaviour was dependent upon powder was mixed with 300 ml water and then sheared using

157l
1572 R. I. MACKIEand l ~ s s t BAI

a heavy duty emulsifier. After being immersed in an ultra-


sonic bath for I min the suspension was added to the main 0.9-
storage tanks. The process was repeated as often as necessary
to prepare sufficient suspension of the required concentration. 0.?"
i 0.8-
Type B. The suspension was prepared as for Type A O.e"
and was then allowed to settle for 12 h in 2251. tanks. The 0.5"
supernatant was then used as the influent for the filtration 0.4-
experiment. The size range varied from less than 0.5 to 0.3-
0.2-
about l0/am.
0.1"
Type C. Again the suspension was prepared as for
Type A. Then the suspension was filtered through a filter
bed consisting of 0.420-0.500mm glass beads, and bed Time (h)
depth 400 ram, i.e. the suspension was prefiltered, the object Fig. 2. Total concentration, 25 mm. i . A; +, B; ,, C.
being to remove as many of the larger particles as possible.
Prefiltering took place at 4.8 m/h. The first 10 rain effluent
was discharged, and the next 50 rain was collected for use. and total concentration data. A 30/am tube was used with
This process was repeated several times until sufficient the Coulter Counter, thus enabling particles in the range 0.6
suspension was prepared. The size ranged from less than 0.5 to about 12/am to be analysed.
to about 6/am.
Particle size analysis was carried out using a Coulter Headloss
Counter. Although the instrument was capable of giving 14 All the manometer points were set at the same depth as the
size channels of data, for the purpose of analysis the par- sampling points with a 27.5 mm centre to centre separation.
ticles were lumped into four size groups in order to reduce The manometer points were connected to glass manometer
statistical errors. The size groups were: tubes.
Group Sizerange (~m)
RESULTS
I 0.63-1.26
2 1.26-2.52
3 2.52-5.04 The three types of suspension were filtered under
4 5.04 - I 0 . 0 8 identical conditions. Figures 2 and 3 show the change
in concentration with time at depths of 25 and 75 min.
Figure I shows typical size distributions of the three types
of influent. The inlet concentrations varied between 90 and
I l0 mg/I. in order to remove the effects of differences
Filtration run in inlet concentration the times have been been
During the course of the experiment the suspension was normalized to an influent concentration of 100 rag/l,
stored in 225 I. tanks, the suspension was stirred in order to
e.g. 1 h at I I 0 m g / I is equivalent to I.I h at 100 rag/I,
prevent settling. The filter was fed via a constant head tank,
and the flow rate was maintained constant during the course I h at 90 mg/I is equivalent to 0.9 h at 100 mg/I. It is
of an experiment by means of flow controllers. clear from the figures that the total removal is for
the most part greatest for suspension A, though in the
Sampling
later stages the removal for suspension A starts to
This is perhaps the most difficult aspect of filtration
experiments. Sampling points were placed at 25 mm above decrease sooner, and at a depth o f 75 mm suspension
the filter bed (for influent sampling), and at 25, 75, 125, 175, B and C are still increasing. This general behaviour
225, 275 and 375 mm below the top of the bed. The sampling can be explained by the difference in the particle size
points were placed in a helical pattern around the filter so distribution (and hence the mean size) of suspension A
that every sampling point was 90" further round than the
and the other two suspensions. It is interesting to note
previous one. Each sampling point had an internal diameter
of 3 ram, protruded 25 mm into the bed and the ends were that although the minimum concentration reached at
covered with a 300 x 300/am gauze. 3 mm internal diameter 25 mm bed depth is dependent upon the suspension
plastic tubing fed the suspension to test tubes via a peristaltic type, at 75 mm there appears to be no significant
pump, the latter being used to maintain a constant flow rate. difference between the minima for suspensions A
Sample analysis and B. Due to the practical difficulties involved in
The samples were immediately analysed on a Coulter preparing and using suspension C (namely the time
Counter (model TAII) to obtain particle size distribution involved) the experiments were not continued to the

j0
point where a minimum was reached at 75 mm.

50
45

O.q

0.'
15
1~ 0.:
O 1
~ 2 3 4
0I
Time(h)
Fig. I. Influent size distribution. I]. A: II. B; I~. C. Fig. 3. Total concentration, 75 ram. I1, A: +, B; ,, C.
Suspended particle size in filtration 1573

1
0.9-
0.8
0.7
I 0.6
0.5.
0eJ
i 0.4 I .4"1
0.3.
0.2" 0.21
0.1.
00 ot
3500
Time(h) Removal g/m s
Fig. 4. Total concentration. 375 ram. II. A; +, B; *, C. Fig. 7. Size Group 2, 375 mm. i , A; +, B; *, C.

in it, this happens because by this time the top


part of the filter has stopped removing the larger
particles.
From the above it is apparent that the change in
size distribution with time and depth is an important
factor. However, it has been suggested by various
workers that there is also interaction between particles
of different sizes. The above results neither prove nor
disprove this. In order to decide whether this is
I 2 3 4 happening or not one needs to look at the behaviour
SizeOroup
of particles within each size group. Figures 6-8 show
Fig. 5. Size distribution, 75 mm, time 3 b. ~ , A, i , B; ea C.
the change in relative concentration (in volume terms)
against total amount of suspension removed for size
As mentioned above the generally greater removal groups I, 2 and 3 at a depth of 375 ram. The total
in the top bed layer for suspension A can be explained removal is the total amount of deposit per unit area
by the greater mean particle size of the influent. of the filter, this means that performance of the filter
However, one feature of filter bchaviour that does not for equal degrees of overall clogging can be compared.
appear to have reccivecl much attention is the change The filter appears to be more effective at removing
in suspended particle size distribution with depth and particles for suspension A than it is for the other two
time. Figures 2 and 3 showed the change in concen- suspensions. Eventually the effectiveness converges,
tration relative to the inlet. Figure 4 shows the change though the difference persists longer for the smaller
at 375 mm relative to the concentration at 75 mm. Up particles (group I). The difference is less pronounced
until about 5 h there is no great difference between for sampling points nearer the top of the filter, but is
the different suspensions, and indeed for a period still there. All the experiments were repeated and the
suspension C is experiencing greater removal in this same trends were observed, but further investigations
layer. The reason becomes apparent by looking at the are being carried out. The results for size group 4 are
size distribution at, for instance, time 3 h (Fig. 5). At not shown as the numbers present in suspension C are
time 3 h and depth 75 mm suspension A is actually a so small as to be meaningless.
finer suspension than the others, and suspension C Suspension size distribution also appears to affect
contains more particles of size 3 than the others, headloss. Figure 9 shows headloss vs removal at
hence the observed behaviour. After about 5.5 h the 25 mm. The graph exhibits a clear dependency upon
removal for suspension A is much greater than for suspension type, with the finer suspension giving the
the others. This again can be explained by the greater headloss. However, if the same graph at
change in size distribution. At time 5.5 h suspension 375 mm (Fig. 10) is observed there is no clear differ-
A again has a significant number of size 4 particles ence between the curves.

I
0.9
I 0.8 I 0.8
0.7
O.e. 0.6
0.5
0.4 0.4
~ 0.3"
0.2- 0.2
0.|"
oi ,~0 1000 1500 2000 2500 3000 5O0 1000 1500 2000 ~ o ~o 35OO
RemovalWma Removeig/m a
Fig. 6. Size Group !, 375 ram. i , A; + , B; ,, C. Fig. 8. Size Group 3, 375 mm. II, A; +, B; *, C.

WR 26~12--B
1574 R. I. MAcKx[ and l~Nst BAt

1200 for suspension A, and 0.929 and 0.937 for B and C.


1000 These results cannot be directly transferred to con-
ditions within a filter, but they do indicate that size
i 8OO
distribution is likely to affect the self porosity of the

l
(CO
deposit, and hence filter performance. This may
4OO explain the headloss results at 25 mm (Fig. 9). Again
200 it must be remembered that the size distribution
changvs with depth and time, so the sedimentation
0
200 400 600 800 1000 1200 1400 1600 tests do not indicate that the deposit below the top
Remo~d g/m 2
layer of the filter will have a lower porosity for
Fig. 9. Headloss at 25 ram. I , A; +, B; *, C. suspension A. For the first 25 mm the effect of deposit
distribution will probably be much less pronounced

__//
1800- and deposit morphology will be more important,
1600- hence the higher headloss for suspension C.
1400-
As well as affecting the porosity of the deposit the
i 1200-
I000- size distribution may affect the microscopic nature of
800. the deposit. This possibility was discussed by Mackie
800- (1989). Briefly the presence of larger particles in the
400- suspension may produce a "rougher" appearance to
200-
the deposit, leading to a greater removal efficiency for
all particles.
Romowd g/m z
Fig. 10. Headloss at 375 ram. II, A; +, B; *, C. CONCLUSIONS
These results indicate that size distribution plays an
Why might size distribution affect the results? The important and complex role in deep bed filtration.
size distribution of a suspension may have two effects: Much more work needs to be done, but some
(i) it might affect the overall distribution of deposit; conclusions can be drawn.
and (ii) it might affect the morphology of the deposit.
(I) The size distribution of a suspension changes
Deposit distribution with depth and time. This means that at any
one time different parts of the filter are treating
Figure I! shows the deposit distribution when
different suspensions, and that the suspension
about 1650g/m 2 had been removed by the filter.
treated at any one depth will change with
(The actual amounts for suspensions A, B and C were
time. Most mathematical modelling work in
1546, 1645 and 1780g/m 2, respectively.) The more
filtration has treated the suspension as having
large the particles in the suspension the more the
a constant composition. These models have
deposit was concentrated near the top of the filter.
had only limited success and the work de-
Since the relationship between headloss and deposit
scribed herein shows that an accurate model
is non-linear (e.g. see Fig. 9) this factor would cause
must take account of changes in suspension
coarser suspensions to give a greater increase in
characteristics.
headloss for the whole filter than fine suspensions,
(2) The size distribution of a suspension may
counteracting the effect shown in Fig. 9.
affect the deposit morphology and the deposit
Deposit morphology distribution. This can affect the performance
of a filter, both in terms of removal efficiency
Sedimentation tests were carried out on the back-
and headloss. The work described indicates
washed deposits for the clogged filters for the three
that the removal of smaller particles is
suspension types in order to estimate the porosity of
enhanced by the presence of larger particles
the deposits. The results gave a porosity of 0.898
in the suspension. For the top layer of the
filter (the first 25 mm) the finer suspension
1
0.8 led to the greatest increase in headloss, but

S
I 0Jl the headloss over the whole bed was much
0+7 less dependent upon size distribution.
O.O
O.S (3) This work was concerned with a polydisperse
I OA suspension and demonstrates the complex
0.3 nature of filtration. For flocculent suspensions
02
0.1 the situation will be even more complicated.
O'
50 100 150 20O 250 300 35O 4O0 Acknowledgements--Part of this work is funded by the
(ram)
Science and Engineering Research Council, U.K. The
Fig. 11. Distribution of deposit at 1650g/m:. 1 , A; +, B; authors would also like to thank ICI for kindly supplying
'lb.C. the PVC powder.
Suspended particle size in filtration 1575

REFERENCES
Tien C. and Payatakes A. C. (1979) Advances in deep-bed
filtration. A.I.Ch.E.JI 25, 737-759.
Darby J. L. and Lawler D. F. (1990) Ripening in depth Tobiason J. E., Johnson G. S. and Westerhoff P. K. (1990)
filtration: effect of particle size on removal and headloss. Particle size and filter performance: model studies. In
Enrir. Sci. Technol. 24, 1069-1079. Natn. Conf. Erie,it. Engng, Proceedings of the 1990 Special
lson C. R. and Ires K. J. (1969) Removal mechanisms in Conference, Arlington Va, U.S.A., pp. 733-739.
deep-bed filtration. Chem. Engng Sci. 24, 717. Vigneswaran S. and Ben Aim R. (1985) The influence of
Ires K. J. (1970) Rapid filtration. War. Res. 4° 201-223. suspended particle size distribution in deep bed filtration.
Mackie R. I. (1989) Rapid gravity filtration--towads a A.LCh.E.JI 31, 321-324.
deeper understanding. Filth Sepn 26, 32-36. Vigneswaran S., Chang J. S. and Janssens J. G. (1990)
Mackie R. I., Homer R. M. W. and Jarvis R. J. (1987) Experimental investigation of size distribution of sus-
Dynamic modelling of deep bed filtration. A.I.Ch.E.J133, pended particles in granular bed filtration. War. Res. 24,
1761-1775. 927-930.
O'Melia C. R. and Ali W. (1978) The role of retained Yao K. M., Habibian M. T. and O'Melia C. R. (1971)
particles in deep bed filtration. Prog. War. Technol. 10, Water and wastewater filtration: concepts and application.
167-182. Envir. Sci. Technol. 5, 1105-1 ! 12.

You might also like