Assignment 5 Solutions: Filtration

Note: Many of the approaches taken are somewhat arbitrary and there would be wide range for suitable
approaches with different answers. Below is just one of what could be a suitable solution.

1. a) Calculate the gross specific area (m 2/m3) in a sand filter with sand nominal diameter 0.4 mm
and porosity 0.4.
b) What is the relationship between interstitial velocity, superficial velocity and porosity?
c) Sand with an effective size of 0.45 mm and uniformity coefficient of 1.35 is required.
Available sand has an effective size of 0.55 mm and uniformity coefficient of 2.55. Which fraction
of the sand would be usable and which fractions smaller and larger than required need to be
discarded?

Solution

a) Assume the shape coefficient is 0.9 (water-worn sand), the gross specific area for a fully
A 6 6
packed bed would be   3
 16666.67m 2 / m 3
 d 0.9  0.4  10
However, since the void spaces are not filled with sand, the real area available will have to be
multiplied by 1- = 0.6 and the available surface area will be 10,000 m 2/m3.
v
b) vi  s where vi  the interstitial velocity; v s  the superficial velocity,   porosity

This follows from Q remaining constant and equal to velocity x available area.

d 60
c) d10 = 0.45mm; uc   d 60  uc  d10  1.35  0.45  0.61mm
d10

'
d 60 ' '
0.55mm uc   d 60  uc '  d10  2.55  0.55  1.40mm
'
d10’ = '
d10
Use a log-probability plot of the desired sand grading and the stock sand to determine
Pst10 = 5.5 and Pst60 = 13

Puse + Pf + Pc = 100

Puse = 2(Pst60 – Pst10) = 2(13 –5.5) = 15%

Pf = Pst10 - 0.1Puse = Pst10 - 0.2(Pst60 – Pst10) = 5.5 – 0.2(13 – 5.5) = 4%

Pc = 100 – Pf – Puse = 100 – 4.0 – 15 = 81%

Sand with a size smaller than 0.41 and larger than 0.7 mm needs to be discarded.

1
 This represents a substantial loss and helps explain the cost of filter sand. The discarded fractions can still be
used for other purposes such as in mortars and concrete, though.

2. A dual-media rapid filter has 0.4m of sand and 0.4m of anthracite. Sand effective size 0.48mm,
sphericity 0.9 and porosity 0.4. Anthracite effective size 0.9mm, sphericity 0.75 and porosity 0.48.
Uniformity coefficient for both is 1.5. Divide each medium in three layers on a mass basis for
calculations. Operating temperature will be 10C.
a) Calculate the initial headloss at 8 m/h
b) Calculate the minimum fluidization velocity for the bed.
c) Calculate the backwash velocity that will expand the largest layer of sand by 15%. Find the
total bed expansion at this rate.
d) What will the total headloss be through the expanded bed under conditions as in (c)?

Solution:
Given
Sand
Depth = 0.4 m
Effective size, ES = 0.48 mm
Sphericity = 0.9
Porosity = 0.4
Uniformity coefficient, UC = 1.5

Anthracite
Depth = 0.4 m
Effective size, ES = 0.9 mm
Sphericity = 0.75
Porosity = 0.48
Uniformity coefficient, UC = 1.5

Temperature = 10°C

2
h kV (1  f ) 2 A
a) = and this applies to all layers
L gf 3 2

Where
h = head loss in depth of bed, L
g = acceleration of gravity
k = dimensionless Kozeny constant with a value of about 5 for most filtering
conditions
A = grain surface area
 = grain volume

2
 f = porosity
V = superficial velocity about the bed = flow rate/bed area
μ = absolute viscosity of fluid
ρ = mass density of fluid

A 6

 d
Where
 = sphericity
d = grain diameter

A/ = 6/(0.9d) = 6.67 d-1

μ/ρ = υ = 1.306x10-6 m2/s at 10°C

h/L = 5 * (8 m/h * 1 h/3600 s)(1.306x10 -6 m2/s)(1 – 0.4)2(6.67/d)2/[(9.81 m/s2)(0.4)3] =

3.70x10-7/d2

Mid-diameter Layer depth

Medium Size* (mm) h/L (m) h (m)
d16.7 0.52 1.37 0.133 0.18
Sand d50.0 0.67 0.82 0.133 0.11
d83.3 0.87 0.49 0.133 0.07
d16.7 0.98 0.39 0.133 0.05
Anthracite d50.0 1.27 0.23 0.133 0.03
d83.3 1.64 0.14 0.133 0.02
Total 0.46
*Neglects the effect of intermixing between adjacent layers.

Anthracite
b) Sand d10 = 0.9 mm
d10 = 0.48 mm d60 = 1.5 * d10 = 1.5 * (0.9 mm) = 1.35 mm d 90 =
d60 = 1.5 * d10 = 1.5 * (0.48 mm) = 0.72 mm d10(101.67 log UC) =
(0.9 mm)(101.67 log 1.5) = 1.77 mm
d90 = d10(101.67 log UC)
= (0.48 mm)(101.67 log 1.5) = 0.94 mm

Vmf = μ/(ρdeq) * [(33.72 + 0.0408 Ga)0.5 – 33.7]

3
 Where
Vmf = minimum fluidization velocity
μ = absolute viscosity of fluid
ρ = mass density of fluid
deq = grain diameter of sphere of equal volume
Ga = Galileo number (dimensionless)

Ga = deq3 ρ(ρs – ρ)g/μ2

Where
ρs = mass density of the grains
g = acceleration of gravity

To apply these equations to a real bed with grains graded in size, it is necessary to
calculated Vmf for the coarser grains in the bed to ensure that the entire bed is fluidized.
The d90 sieve size is a practical diameter choice in this calculation.

At 10°C
μ = 1.307x10-3 N·m/s2
ρ = 999.7 kg/m3
μ/ρ = υ = 1.306x10-6 m2/s
Sand
deq (use d90) = 0.94 mm
ρs = 2650 kg/m3 (Table 8.2 in AWWA p.8.10)

Ga = (0.94x10-3 m)3 * 999.7(2650 - 999.7) * (9.81 m/s2)/(1.307x10-3)2 = 7869

Vmf = (1.306x10-6 m2/s)/(0.94x10-3 m) * [(33.72 + 0.0408 * 7869)0.5 – 33.7] =
0.0062 m/s = 22.3 m/h

Anthracite
deq (use d90) = 1.77 mm
ρs = 1600 kg/m3 (estimated from Table 8.2 in AWWA p.8.10)

Ga = (1.77x10-3 m)3 * 999.7(1600 - 999.7) * (9.81 m/s2)/(1.307x10-3)2 = 19111

Vmf = (1.306x10-6 m2/s)/(1.77x10-3 m) * [(33.72 + 0.0408 * 19111)0.5 – 33.7] =
0.0074 m/s = 26.64 m/h
The minimum fluidization velocity for anthracite, being the largest, will govern.

c) Given for sand layer

ε0 = 0.4

4
 L/L0 = 1.15 for largest layer
ψ = 0.9

L/L0 = (1 – ε0)/(1 – ε)
Where
L/L0 = ratio of expanded depth (L) to fixed-bed depth (L 0)
ε0 = fixed loose bed porosity
ε = expanded bed porosity

ε = 1 – (1 – ε0)/(L/L0) = 1 – (1 – 0.4)/1.15 = 0.48

Re1 = Vρ/[Sv(1 – ε)μ]

Where
Re1 = modified Reynolds number
V = superficial backwash velocity
ρ = mass density of fluid
Sv = specific surface of grains = 6/(ψdeq)
ε = expanded bed porosity
μ = absolute viscosity of fluid

μ/ρ = υ = 1.306x10-6 m2/s at 10°C

Use deq = d90.

Sv = 6/(ψdeq) = 6/(0.9 * 0.94x10-3 m) = 7092 m-1
Re1 = V/[(7092 m-1)*(1 – 0.48)*(1.306x10-6 m2/s) = 206.9 V

A1 = ε3/(1 – ε)2 * ρ(ρs – ρ)g/(Sv3μ2)

A1 = 0.483/(1 – 0.48)2 * 999.7(2650 - 999.7) * (9.81 m/s2)/[(7092 m-1)3(1.307x10-3)2] = 10.7

log A1 = 0.56543 + 1.09348 log Re1 + 0.17971 (log Re1)2 – 0.00392 (log Re1)4 – 1.5 (log ψ)2

log A1 = 1.028
log Re1 = log (206.9 V)
1.5 (log ψ) = 0.00314
2

1.028 = 0.56543 + 1.09348 log (206.9 V) + 0.17971 log (206.9 V) 2 – 0.00392 log (206.9 V) 4 –
0.00314
Backwash velocity, V = 0.0121 m/s = 43.5 m/h
Given for anthracite layer
ε0 = 0.48

5
 ψ = 0.75
V = 0.0121 m/s

Use deq = d90.

Sv = 6/(ψdeq) = 6/(0.75 * 1.77x10-3 m) = 4520 m-1

Re1 = (0.0121 m/s)/[(4520 m-1)*(1 – ε)*(1.306x10-6 m2/s) = 2.05/(1 – ε)

A1 = ε3/(1 – ε)2 * 999.7(1600 - 999.7) * (9.81 m/s2)/[(4520 m-1)3(1.307x10-3)2] = 37.3 ε 3/(1 – ε)2

log [37.3 ε 3/(1 – ε)2] = 0.56543 + 1.09348 log [2.05/(1 – ε)] + 0.17971 (log [2.05/(1 – ε)]) 2
– 0.00392 (log [2.05/(1 – ε)])4 – 1.5 (log 0.75)2

ε = 0.50 L/L0 = (1 – 0.48)/(1 – 0.5) = 1.04

Total expansion = 0.15 + 0.04 = 0.19, or 19% based on d 90 size of sand and anthracite.
The expansion is larger when calculated for each layer and summed (table below).

Mid-diameter
Medium Size (mm) ε Sv Re1 A1 L/L0 L
d16.7 0.52 0.62 12821 1.89 7.2 1.57 0.21
Sand d50.0 0.67 0.55 9950 2.09 8.1 1.34 0.18
d83.3 0.87 0.49 7663 2.38 9.5 1.18 0.16
d16.7 0.98 0.65 8163 3.24 14.0 1.48 0.20
Anthracite d50.0 1.27 0.58 6299 3.54 15.7 1.25 0.17
d83.3 1.64 0.52 4878 3.97 18.3 1.09 0.14
Total L 1.05
Total expansion = (1.05 m)/(0.8 m) = 1.32, or 32%.

d) The pressure drop after fluidization is equal to the buoyant weight of the grains.

Δp = hρg = L(ρs – ρ)g(1 – ε)

h = L(ρs – ρ)(1 – ε)/ρ
Where
h = head loss in depth of bed, L
ρ = mass density of fluid
ρs = mass density of the grains
g = acceleration of gravity
ε = porosity

6
 h = [(0.4 m)(2650 - 999.7)(1 – 0.4) + (0.4 m)(1600 - 999.7)(1 – 0.48)]/999.7

h = 0.52 m
3. The City of Ratlerville is considering a water treatment plant to treat a eutrophic supply. Specify
the major design quantities and dimensions for a slow sand filter to treat the requirements for 5000
people with a demand of 300L/cap/d to the Director of Public Works there. Outline the strategy for
operation, the expected period between scrapings, anticipated medium replacement, etc.

Solution:
According to data in AWWA
The most common filtration rate is 0.07~0.12m/h
Typical cycle length is 1~ 6 months.
Initial bed depth: 0.4~1.52m (most common 0.9m)
Sand effective sizes: 0.15~0.4mm (most common 0.3mm)
Sand uniformity coefficient: 1.5~3.6 (most common 2)

Choose a filtration rate 0.1m/h – this should be adequate for eutrophic water.
Required minimum filtration area:
Q 5000  0.3m 3 / d
A   625m 2
V 1  24h / d
It is better to use more than one filter and provide for standby capacity during cleaning, say use
three filters of 625/3 = 208m 2 each and add one more of these for spare. Choose the bed depth =
0.9m to provide enough depth to allow multiple scraping before resanding.

In summer, algal blooms can be expected in the source. Therefore, microscreening and or
prechlorination may be needed to prevent algae from blocking the filter and the sand scraping is
more frequently needed.

Cycle length is decided according to the head loss, water quality, and filtration rate. Typical cycle
time is 1~6 month and it is shorter due to algae in summer. The expected period between scraping
can be expected to be 4 to 8 weeks and the anticipated medium replacement will be when half of
the bed has been removed, i.e. about every 2 – 3 y.

4. Three sand filters are to be designed to operate at a constant head of 1.5 m to supply a total of
12000 m3/d. Filter sand of 0.8 mm will be used, 1.2 m deep, upflow operation. Pilot scale tests on
the same water under the same head showed that the throughput rate halved within 16 hours of
operation. Downtime is 45 minutes while draining and then backwashing the filter with sand filtered
water for 20 minutes at an approach velocity of 30 m/h. Design the sand filters to minimize size.
Frequency of backwashing is left to the designer. Make assumptions where required using typical
values encountered in text or literature.
Calculate the throughput rate at the start of operation using the Kozeny equation:

7
h = (180/g)(/)V((1-f)2/f3) (L/(2d2))
Solution:

Calculate the throughput rate at start of operation using the Kozeny’s equation:

h = (180/g)(/)V((1-f)2/f3) (L/(2d2))

Given: Grain diameter, d = 0.8 mm = 0.0008 m

Bed thickness, L = 1.2 m
Head loss through filter = 1.5 m

Assume: Viscosity,  = 10-3 Ns/m2

Density,  = 1000 kg/m3
Porosity, f = 0.4
Shape coefficient,  = 0.7 (assume angular sand)

1.5 = (180/9.81)(10-3 /1000)V((1-0.4)2/0.43) (1.2/(0.72.0.00082)) = 395V

Loading at start of operation, V = 0.0038 m/s = 13.7 m/h

From tests, throughput is halved at 16 hours of operation.

i.e. V = 13.7/2 = 6.85 m/h

From h = V(a + bV) for t = 0 (at start of operation),

h = Va
h = 1.5 m & V = 13.7 m/h
a = 0.109 b = (13.7-6.85)/(0-16) = -0.428

13.
7
Approach
velocity 6.8
(m/h) 5 3.
9

Time (h)
1 22. 2
6 9 4

It is not essential to calculate these constants, just to work with linear decreases.

8
Assume a 24 hour cycle but operate filter for 22.9 h (to allow for backwash):

From the linear relationship, 16((13.7-v)/6.85) = 22.9,

where v is the velocity reached before backwash. v = 3.9 m/h

Volume of water filtered can be calculated from the average speed of filtration times period of operation. For a
linear decrease in rate, the average speed would be (13.7 + 3.9)/2 m/h or 8.8 m/h. Maintaining that average for
22.9h, would result in a production of 22.9 x 8.8 m 3/m2 = 201.5 m3/m2 of filter area per 24h production period.

Q = 12000 m3/d, therefore, the area to pass this quantity = 12000/201.5 = 59.5 m 2

Water losses during a cycle, including draining and backwashing

Draining will result in a loss of all water in the sand, that above and below it. Assume a combined level of water
of 0.5m above and below the sand, i.e. 30m 3.
Water in sand = 1.2m x 59.5 = 90 m 3 - volume of sand, i.e. 1.2 x 59.5 x 0.4 = 47m3. Total loss during drainage =
77 m3. The filter has to be filled up again before actual backwashing occurs, requiring the 77 m 3 out of stored
water. If drained again after backwashing, another 77 m 3 would be required.

Backwashing at 30m/h for 20 min, will require 59.5 x 30 x 20/60 = 595 m 3

Total water losses during a 24h period then amount to 672 (or 750) m 3. This would require another 672/201.5 =
3.3 m2 of filter area, ie 5.5%. Since this additional 5.5% will itself need backwash water, we need to add another
5.5% of the 3.3m2 or 3.5 m2 total.

Filter area required would then be 63m 2, say 4 filters of 16 m2 each.

At this stage, it might be advisable to specify 5 filters of 16m 2 each for standby capacity during backwash.

In order to minimize size, one would have to recalculate, using shorter operational periods (with higher average
flows, but larger % losses due to more frequent backwashing and water requirements) and longer operational
periods.

The easiest way to optimize this is to maximize the cumulative flow through 1 m 2 over a 24h period. For
instance, backwashing every 17h, would require 12% less filter area. It is possible, however, that the optimum
occurs at backwash frequencies shorter or longer than 17h. If the differences are small, it is easier to work on a
24h cycle for the benefit of the operators.

Note: It can be shown that the backwashing velocity required for a bed expansion of 30% is
closer to 30m/h. This would be done by using the analysis of the forces acting on a particle
corrected for hindered settling.

9
10