Drinking Water technology

8/17/2019 Drinking Water technology

http://slidepdf.com/reader/full/drinking-water-technology 1/201





Drinking water

Water treatment technology

prof. ir. J.C. van Dijk

Delft University of Technology



Further information about this and other publications can be obtained at:

Delft University of Technology

Faculty of Civil Engineering,

Section of Sanitary Engineering,

Stevinweg 1

2628 CN

Delft

tel: +31-15-2785440

fax: +31-15-2787966

English translation and editing: Adele Sanders, Delft EdiTS

© TU Delft 2007



PREFACE

Contents

Water treatment schemes 11

Coagulation and occulation 37

Sedimentation 51

Flotation 67

Filtration 81

Adsorption 103

Disinfection 115

Aeration and gas stripping 133

Softening 153

Micro- and ultraltration 173

Nanoltration and reverse osmosis 189

Laboratory experiments 203



6

CONTENTS



7

CONTENTS

Detailed contents

Water treatment schemes 11

Framework, contents, study goals 12

1. Introduction 13

2. Groundwater 14

3. Riverbank groundwater 20

4. Surface water with direct treatment 2222

5. Surface water with inltration 30

Coagulation and occulation 37


1. Introduction 39

2. Coagulation 40

3. Flocculation 45

Sedimentation 51


1. Introduction 53

2. Theory 533. Inueances on settling in a horizontal ow tank 58

4. Practice 62

5. Settling tank alternatives 64

Flotation 67


1. Introduction 69

2. Principle 69

3. Theory 71

4. Practice 76

Filtration 81


1. Introduction 83

2. Principles 83

3. Theory 87

4. Practice 915. Alternative applications of ltration 95



8

CONTENTS

Adsorption 103


1. Introduction 105

2. Theory 107

3. Practice 110Practice 110110

Disinfection 115


1. Introduction 117

2. Purpose of disinfection 117

3. Disinfection kinetics 125

4. Disinfection methods 129

Further reading 133

Aeration and gas stripping 135


1. Introduction 137

2. Theory of gas transfer 138

3. Practice 144

Softening 153


1. Introduction 155

2. Principle 155

2. Theory 159

3. Practice 164

Micro- and ultrafltration 173


1. Introduction 175

2. Principle 177

3. Theory 179

4. Practice 184

5. Operation 187



9

CONTENTS

Nanofltration and reverse osmosis 189


1. Introduction 191

2. Principle 191

3. Theory 193

4. Practice 198





Water treatment

schemes

WA T

E R T R E A T M E

N T

WATER TREATMENT



Framework

This module represents a short introduction to treatment schemes used for the production of drinking

water.

Contents

This module has the following contents:

1. Introduction

2. Groundwater

2.1 Types of groundwater

2.2 Aerobic groundwater (phreatic)

2.3 Slightly anaerobic groundwater

2.4 Deep anaerobic groundwater

3. Riverbank groundwater

3.1 Types of riverland groundwater 3.2 Riverland groundwater

3.3 Riverbank ltrate

4. Surface water with direct treatment

4.1 Historical developments

4.2 Contemporary treatment

4.3 Future treatment

5. Surface water with inltration

5.1 Surface water with open inltration

5.2 Surface water with deep inltration

5.3 Pre-treatment

5.4 Inltration

5.5 Final treatment of inltrated water

Study goals

After having studied this module, you will be able to:

• understand treatment schemes for groundwater and surface water

12

WATER TREATMENT SCHEMES WATER TREATMENT



In public water supply, ve components may be

distinguished (Figure 1):

- abstraction

- treatment

- transport

- storage

- distribution.

Raw water can be abstracted from groundwater

or surface water. For almost every type of water, a

corresponding treatment is necessary before it can

be supplied as drinking water. During transport,

storage and distribution, the quality of the drinking

water must not deteriorate below the established

standards.

The purpose of placing a low-lying, clear water

reservoir after the treatment step is to adjust the

differences between the treatment plant and the

1. Introduction

For personal hygiene and other domestic activi-

ties (washing, cleaning, toilet ushing, etc.), it is

important to have sufcient water available. In

addition, for consumption, water of good quality

is essential.

More than 99% of Dutch households receive their

water via a public water supply.

This water is hygienically reliable, clear, good tast-

ing and of a pleasant temperature. The water is

distributed through piped networks under sufcient

pressure for even the most remote houses. High

quality drinking water is supplied into people’shomes at a very “low” price.

The Dutch water supply system is one of the best-

known systems in the world.

Figure 1 - Drinking water supply of a city

intake structure with

raw water pumpstreatmentplant

water reservoir

with transport

pumps and transportmains

water reservoir

with distribution

pumps

distribution area

WATER TREATMENT

13

WATER TREATMENT SCHEMES



transport pumps.

A distribution reservoir at the edge of the city levels

off consumption variations between day and night,

resulting in a constant ow through the treatment

plant and the transport main, and in minimal en-

ergy consumption.

The drinking water is distributed continuously un-

der sufcient pressure (>20 meters water column

(200 kPa) above street level). This pressure also

prevents inltration of the groundwater into the

distribution system, which would result in quality

deterioration.

Because drinking water quality is only required for

human consumption (5 % of the total production),separate water supplies with different qualities

could be considered. The distribution network,

however, is the most expensive element of the

public water supply. The costs are up to 50 to 70%

of the total water price. The savings on treatment

costs do not compensate for the extra costs for

supply and distribution and the risks to public

health (by cross-connections).

Domestic water consumption in the Netherlands

is still increasing, despite efforts to encourage

consumers to save water (Figure 2).

In addition, the standards for drinking water have

become more rigorous and knowledge about

contamination and its effects on public health is

increasing.

Important developments are the discovery of harm-

ful byproducts that are formed during the chemical

disinfection of drinking water.

Parallel to these developments, the discovery of

pesticides in raw water and drinking water (the

Bentazon-affair at Amsterdam Water Supply in

1987) has led to changes in opinions about drink-

ing water production.

Additionally, many water companies have set

guidelines for the maximum hardness of drink-

ing water. Hard water can cause lime deposits in

warm water installations and reduce the forma-

tion of foam from soap. Hence, energy and soap

consumption increase.

Also, more rigorous environmental standards have

led to better treatment of surface water before it

is inltrated.Finally, backwash water for the lters is more often

treated and recirculated.

The above implies that, in the near future, the

infrastructure must be expanded and improved.

For civil engineers this creates a great chal-

lenge.

The design of the components of a drinking water

supply system requires not only knowledge of

sanitary engineering, but also knowledge of other

disciplines as well.

Knowledge of water management is necessary

when determining the source location and for

minimizing the consequences of the abstraction

on the environment and other activities (e.g.,

agriculture, navigation).

In addition, there is a need for uid mechanics and

structural engineering to determine the dimen-

sions of pipes, pumps, treatment installations and

reservoirs.

Finally, knowledge of chemical engineering andbiotechnology is needed to design and optimize

treatment processes.

2. Groundwater Groundwater

2.1 Types of groundwater

Groundwater has a near-constant quality. Per

location, however, large differences in water com-

position can be found.Figure 2 - Drinking water production in the Nether-

lands

0

1820 1870 1920 1970 2020

d r i n k i n g w a t e r p r o d u c t i o n ( m l n m

3 / y )

3

2

1

14




This composition is a result of the natural

environment from which the groundwater is

abstracted, and the route that the water has

followed to get there.

Three types of groundwater can be roughly

distinguished with respect to the treatment in

drinking water production:

- aerobic groundwater (phreatic)

- slightly anaerobic groundwater

- deep anaerobic groundwater

The above list implies that, for the treatment of

groundwater, the level of oxygen (aerobic, slightly

or deep anaerobic) is very important.The redox potential is a good indicator for this,

but this potential is seldom measured in practice.

To what type a certain groundwater belongs can

be determined from the concentrations of oxygen,

iron, and methane.

The three types of groundwater will be further

discussed separately, on both their typical

characteristics and their treatment schemes.

After this the different treatment processes will

be described.

2.2 Aerobic groundwater (phreatic)

Phreatic groundwater has an open groundwater

table and is, consequently, connected to the

atmosphere. When the organic matter content

of the soil is limited, the water does not lose its

oxygen (i.e., become anaerobic). As a result, no

anaerobic reactions (e.g., iron dissolution) occur

in the soil.

In special cases aerobic groundwater meets the

requirements for drinking water.

In the Netherlands a couple of groundwater

abstraction facilities are located on the Veluwe

from which the abstracted water is directly

distributed as drinking water. Normally, some

treatment is necessary or desired.

Despite the fact that we are dealing with aerobic

(i.e., containing oxygen) groundwater, the rst

treatment step is an aeration phase. Because of

this aeration phase, the concentration of oxygen is

increased further and the concentration of carbon

dioxide is decreased. If the water complies with

the legal standards after this treatment step, thenthe water can be distributed.

In the case of aerobic phreatic groundwater, only

the parameters pH, Ca, SI and HCO3- have to be

taken into account. The other parameters generally

comply with the legal requirements.

Therefore, the treatment scheme of phreatic

aerobic groundwater includes, in addition to a

possible aeration, conditioning (Figure 3).

Aggressive water

When aerobic groundwater is abstracted from

sandy soils (no calcium in the underground), the

groundwater is often aggressive to limestone.

Because of a number of breakdown processes,

carbon dioxide is present in groundwater, and,

because the calcium is missing, the concentration

of carbon dioxide is higher than the equilibrium

concentration of carbon dioxide. The value of the

saturation index, SI, is smaller than 0. To make

distribution of this water possible, the saturationindex has to be increased.

The SI is increased by aerating the water, which

removes carbon dioxide. Then the SI may meet

the requirements, but the requirements for pH and

HCO3

- buffering are often not met because the

concentration of HCO3- is too low. When limestone

(marble) ltration is applied, the requirements for SI,

pH and HCO3- buffering are met. During limestone

ltration, the aggressive water is ltered through a

lter bed consisting of marble grains (limestone)

Figure 3 - Treatment of phreatic aerobic groundwa-

ter

WATER TREATMENT

15




(Figure 4). Because the water is aggressive, it

dissolves the marble grains. After some time, the

lter bed has to be relled with new grains.

Hard water

Aerobic groundwater, if abstracted from soils

rich in calcium (for example, limestone area of

Zuid-Limburg), is often very hard (>3 mmol/l).

Because of the biological processes in the soil,

the concentration of CO2 can result in a substantial

dissolution of limestone, forming Ca2+ and HCO3-

in the water. The abstracted water, therefore, will

be hard.

Groundwater will sometimes be in equilibri-

um regarding calcium carbonate (limestone).

Water that is supersaturated with respect to

calcium carbonate cannot be found in nature;

because of the long residence time, a possible

supersaturation would already have disappeared

due to precipitation. When this water is pumped up

and comes in contact with air, the carbon dioxide

disappears from the water. The carbon dioxide

concentration is, after all, larger than the saturationconcentration of carbon dioxide in water being in

equilibrium with air. Because of the removal of

carbon dioxide from the water, the water becomes

supersaturated with respect to calcium carbonate

(SI > 0).

To prevent limestone precipitation in a distribution

network or in consumers’ washing machines and

heaters, and to satisfy the recommendation of

a maximum hardness of 1.5 mmol/l, the water

is softened. This softening occurs by dosing

chemicals (NaOH or Ca(OH)2) into the water in

cylindrical reactors with upward ow (Figure 5).

These reactors contain small sand grains, which

are used as crystallization nuclei on which the

CaCO3 precipitates.

The softening installation should be followed by

granular media ltration, because possible post-

precipitation might occur. After all, the time the

water stays in the pellet reactor is short (a couple

of minutes), and for the complete process of

chemical softening more time is needed. When,

after the softening, a granular media ltration

phase is executed, post-precipitation takes

place in the lter bed. If this ltration phase isn’tprovided, then the precipitation will take place

in the distribution network or in the consumers’

household machines.

Alternatively, acid neutralization can be applied.

Example

As an example of the change in water quality,

the Hoenderloo pumping station on the Veluwe is

described. Treatment at the Hoenderloo pumping

station consists of aeration/gas transfer followed

by limestone ltration. The values of the different

parameters in Table 1 are the annual averages.

Figure 4 - Aeration above limestone lter

Figure 5 - Pellet reactor for softening water in

Meersen (Limburg)

16




The pH of the treated water is higher than the raw

water, because the water is aerated (removal of

CO2) and because the water is ltered through a

limestone lter (decrease in the CO2 concentration,

increase in the HCO3- and the Ca2+ concentration).

The SI increases under the inuence of the lower

concentration of CO2, and the higher HCO

3-, and

the Ca2+ concentrations; the water becomes less

aggressive with respect to calcium carbonate.

Because of the use of limestone ltration, the

HCO3

- and the Ca2+ concentrations will increase.

More ions will get into the water, as a result ofwhich the conductivity (EC) will increase.

We can calculate the increase in the HCO3

-

concentration when we assume that all produced

HCO3- comes from the limestone. For every formed

mmol/l Ca2+, 2 mmol/l HCO3- are produced. At this

pumping station the Ca2+ -concentration increases

with 0.3475 mmol/l because of the limestone

ltration, and the concentration of HCO3- has to be

increased by 2 · 0.3475 = 0.695 mmol/l. There was

0.34 mmol/l HCO3- present in the raw water and

thus, there has to be 0.695 + 0.34 = 1.035 mmol/l

HCO3- in the treated water. This corresponds to

63.1 mg/l.

2.3 Slightly anaerobic groundwater

Slightly anaerobic groundwater is found when the

groundwater is located under a conning layer,

and is characterized by the lack of oxygen and the

presence of ammonium, iron and manganese.

The treatment of slightly anaerobic groundwater

often consists of aeration followed by submerged

granular media ltration (Figures 6 and 7).

Aeration is necessary for the addition of oxygen

and the removal of carbon dioxide. The oxygen is

used for the oxidation of Fe2+ to Fe3+ (a chemical

process), and it is also needed for the oxidation

of NH4+ to NO

3- and of Mn2+ to MnO

2.

Aeration is followed by submerged sand ltration.

In the lter the oxidized ferric iron reacts with OH+-

Table 1 - Quality data of the raw and treated water at

the Hoenderloo pumping station (Gelder-

land)

Parameter Unit Raw water Clear water

Temperature °C 9.6 10

pH - 6.1 7.8EGV mS/m 9.3 14.3

SI - -3.4 -0.3

Turbidity FTU - < 0.1

Na+ mg/l 8.1 7.9

K+ mg/l 1 1

Ca2+ mg/l 8.6 22.5

Mg2+ mg/l 1.6 1.6

Cl- mg/l 12 12

HCO3- mg/l 21 63

SO42- mg/l 9 10

NO3- mg/l 2.7 2.7

O2

mg/l 4.2 8

CH4 mg/l - -CO

2 mg/l 31 2

Fe2+ mg/l 0.06 0.03

Mn2+ mg/l 0.02 < 0.01

NH4+ mg/l < 0.04 < 0.04

DOC mg/l < 0.2 < 0.2

E.coli n/100 ml 0 0

Bentazon µg/l - -

Chloroform µg/l - -

Bromate µg/l - -

Figure 6 - Treatment of slightly anaerobic groundwa-

ter

Figure 7 - Treatment of slightly anaerobic groundwa-

ter

WATER TREATMENT

17




ions and is transformed into Fe(OH)3-ocs, which

are ltered in the sand bed (a physical process).

Manganese undergoes a partly chemical and

partly biological transformation, while ammonium

is biologically transformed. The transformation

of ammonium is accomplished by the bacteria

Nitrosomonas and Nitrobacter . During this

transformation a lot of oxygen is used; per mg/l

ammonium, the oxygen consumed is 3.55 mg/l.

Also, a lot of nitrate is formed; per mg/l ammonium,

3.44 mg/l nitrate is produced.

As a result of the biological transformation of

ammonium and manganese and the physical

removal of the iron hydroxide flocs, the porevolume between the sand grains decreases,

because the pores are lled by either bacteria

or by ocs and deposits. The result of this is the

increase in the hydraulic resistance of the water

when owing through the lter bed. When this

resistance becomes too large, the lter should be

backwashed.

Example

As an example of the change in water quality of

slightly anaerobic water, the pumping station at the

Zutphenseweg will be described. The treatment

at the Zutphenseweg pumping station consists of

aeration/gas transfer followed by sand ltration

and a second aeration. The values of the different

parameters in Table 2 are the average values over

a year. As a result of aeration the concentration

of CO2 will decrease and the pH of the water will

increase. Because of aeration the concentration of

oxygen will increase to a value near the saturation

value (ca. 10 mg/l). The concentration of Fe2+,

Mn2+ and NH4+ will decrease due to the inuence

of oxidation and biological transformations.The nitrate content will increase, because the

ammonium is transformed into nitrate. Since the

decrease in ammonium is approximately 0.8 mg/l,

the nitrate content should increase circa 2.7 mg/l.

The oxygen consumption is approximately 2.8

mg/l. To get a high oxygen content, post-aeration

is used.

Table 2 - Quality data of the raw and treated water atZutphenseweg pumping station (Overijssel)


Temperature °C 13.1 13.1

pH - 7.7 7.9

EGV mS/m 58 58

SI - -0.1 0.1


Na+ mg/l 75 75

K+ mg/l 6.7 6.7

Ca2+ mg/l 47 46

Mg2+ mg/l 7.8 8

Cl- mg/l 108 110

HCO3- mg/l 185 177

SO42- mg/l < 1 < 1

NO3- mg/l < 0.1 2.8

O2

mg/l 0.4 9.5

CH4

mg/l - -

CO2

mg/l 7 4

Fe2+ mg/l 0.39 0.03

Mn2+ mg/l 0.03 < 0.01

NH4+ mg/l 0.82 < 0.04

DOC mg/l 2 1,7

E.coli n/100 ml 0 0

Bentazon µg/l - -


Bromate µg/l - - Figure 8 - Treatment of deep anaerobic groundwater

18




2.4 Deep anaerobic groundwater

Deep anaerobic groundwater is found when the

water is abstracted under a conning layer andno oxygen is present in the water. Furthermore,

there is no nitrate present and organic material

is broken down with sulfate as an oxidant. Iron,

manganese and especially ammonium are present

in high concentrations, while hydrogen sulde and

methane are also present in the groundwater.

During the removal of ammonium, a lot of oxygen

is used. When the ammonium content is larger

than 3 mg/l, the amount of oxygen necessary for

the removal of the ammonium is greater than the

total amount of oxygen, which can be dissolved

in water (saturation concentration). To prevent

anaerobic conditions in the last filter, double

submerged ltration or dry ltration followed by

submerged ltration is used during groundwater

treatment with a high amount of ammonium.

Dry ltration is followed by submerged ltration

because in a dry lter, the breakthrough of particles

may occur. When these materials pass through thedry lter, they are ltered in the submerged lter

and do not show up in the drinking water.

An aeration phase is present before every ltration

step, so the oxygen concentration is high before

the water enters the lter and the carbon dioxide

is removed (Figures 8 and 9). A dry lter is a lter

lled with sand grains with a diameter between 0.8

and 4 mm. A layer of water is not present in the

lter, like with submerged ltration. In the dry lter

the water ows down past the grains, at the same

time air is owing with the water. The oxygen in the

air replenishes the oxygen in the water, which is

used by bacteria. In this way more than 3 mg/l ofammonium can be transformed without anaerobic

results in the lter.

Example

As an example of the change in water quality of

deep anaerobic groundwater, the St. Jansklooster

pumping station is described. The treatment at this

pumping station consists of aeration, dry ltration,

groundwater pre-filter post-filterclear waterreservoir water tower consumers

Figure 9 - Treatment of groundwater with double aeration/ltration


St. Jansklooster pumping station (Overijs-sel)



pH - 6.9 7.6

EGV mS/m 51 48

SI - -0.4 0.2


Na+ mg/l 23 21

K+ mg/l 3 3

Ca2+ mg/l 82 77

Mg2+ mg/l 5.2 6.3

Cl- mg/l 41 41

HCO3- mg/l 267 241

SO42- mg/l 18 21

NO3- mg/l 0.07 1.6

O2

mg/l 0 10.7

CH4

mg/l 2 < 0.05

CO2

mg/l 63 11

Fe2+ mg/l 8.8 0.04

Mn2+ mg/l 0.3 < 0.01

NH4+ mg/l 2.2 < 0.01

DOC mg/l 7 6

E.coli n/100 ml 0 0

Bentazon µg/l - -


Bromate µg/l - -

WATER TREATMENT

19




aeration and submerged ltration. The values of

the different parameters in Table 3 are the average

values over a year.

As a result of the aeration phases, the amount

of carbon dioxide will decrease and the pH will

increase. Furthermore, the amount of oxygen will

increase. The concentration of Fe2+, Mn2+ and NH4+

will decrease because of chemical and biological

transformations; the amount of nitrate, on the other

hand, will increase. The concentration of nitrate

increases less than the theoretical calculation.

3. Riverbank groundwater

3.1 Types of riverbank groundwater

Riverbank ltration is groundwater abstracted

directly adjacent to surface water, usually from

a river. The abstraction takes place in such a

way that the abstracted water consists mostly of

surface water. This surface water is inltrated into

the soil via the riverbank or the river bottom. In

this way, a mixture of surface water and natural

groundwater is abstracted.

The residence time of the inltrated surface water

in the soil can be several years. In this case we call

it riverbank groundwater. This groundwater has the

characteristics of groundwater, but the chemical

composition also reveals surface water. In the

Netherlands such abstractions are found along

the Lek and the IJssel. The distance between

abstraction wells and the river vary between 200

and 1,000 m.

In Germany most abstraction wells are placed

much closer to the river (Figure 10). Residence

times of several weeks are common. It is then

called riverbank filtrate. It is clear that the

existence of surface water in such cases is easier

to recognize. A well-dened boundary between

riverbank groundwater and riverbank filtrate

doesn’t exist.

3.2 Riverbank groundwater

The treatment of riverbank groundwater has many

similarities to the treatment of slightly anaerobic

groundwater. Riverbank groundwater is, for themost part, “natural” groundwater. The other part is

surface water that has some of the characteristics

Figure 10 - Riverbank groundwater well in Germany Figure 11 - Treatment of riverbank groundwater

20




of groundwater due to a long residence time inthe soil.

The treatment scheme for riverbank groundwater

is shown in Figure 11. Depending on the soil

composition, higher concentrations of iron,

ammonium, manganese and methane can be

found. Furthermore, the hardness can be fairly

high because of inltration of river water. Due

to high concentrations of ammonium, which are

biologically transformed to nitrate, a lack of oxygen

can occur in the treatment; therefore, an extra dry

ltration stage is often included (Figure 12).

Activated carbon ltration is also used for the

treatment of riverbank groundwater because of

taste problems and problems with pesticides.

Because part of the water is surface water, it also

contains substances associated with surface

water.

For riverbank groundwater, UV-disinfection is oftenapplied as the last disinfection stage, especially if

activated carbon ltration is used in the treatment.

In the activated carbon lters, microorganisms

grow due to the breakdown of organic material;

these can subsequently end up in the water.

With UV-disinfection the microorganisms are killed,

without the formation of disinfection by-products

which are typical for chemical disinfection.

Example

As an example of riverbank groundwater, the

Nieuw-Lekkerland pumping station of Hydron Zuid-

Holland is described. Table 4 shows that the water

contains a high concentration of ammonium and

that there are pesticides present in the water.

Hence, the treatment scheme is as follows:

aeration, dry filtration, aeration, submerged

filtration, activated carbon filtration, and UV-

disinfection.

Chloride can become a problem for riverbank

groundwater treatment plants along the Rhine. In

the treatment, chloride isn’t removed. When the

concentration in the Rhine is too high, then the

standard for chloride may be exceeded.

Oxygen increases because of the aeration steps.

Manganese and iron decrease because of the

combination of aeration and ltration. Ammonium

decreases because of transformation in the dry

and submerged lters. Because of that, the nitrate

Figure 12 - Aeration over a dry lter in Zwijndrecht (Zuid-

Holland)

Table 4 - Quality data of raw and treated water at

Nieuw-Lekkerland pumping station (Zuid-

Holland)

Parameter Unit Raw water Clear water Temperature °C 12 12

pH - 7.3 7.4

EC mS/m 78.4 77

SI - -0.2 -0.1


Na+ mg/l 69 70

K+ mg/l 4 4

Ca2+ mg/l 84 84

Mg2+ mg/l 12 12

Cl- mg/l 128 135

HCO3- mg/l 223 187

SO42- mg/l 55 59NO

3- mg/l < 0.1 2.3

O2

mg/l 0.8 5.7

CH4

mg/l 1 < 0.05

CO2

mg/l 20 14

Fe2+ mg/l 3.8 0.02

Mn2+ mg/l 0.9 < 0.01

NH4+ mg/l 3 < 0.03

DOC mg/l 3 2.5

E.coli n/100 ml 0 0

Bentazon µg/l 0.32 < 0.05


Bromate µg/l - -

WATER TREATMENT

21




content increases. Because of activated carbon

ltration, the Bentazon content decreases.

There are no E.coli in the raw water because the

raw water passes through the soil. During UV-

disinfection, possible organisms are killed that

grow in the activated carbon lter.

3.3 Riverbank ltrate

The treatment of riverbank ltrate doesn’t show

many differences from the treatment of riverbank

groundwater. Only in this case, the share of

surface water is larger, which makes activated

carbon ltration and post-UV-disinfection moreimportant.

Dosing with ozone is applied in a number of cases

for riverbank ltrate, oxidizing the micropollut-

ants.

Dosing with ozone is, in this case, not necessary

for disinfection. When the water is passing through

the soil, all (harmful) bacteria are removed, even

with a relatively short residence time.

For soil passage the ltration of microorganisms

is the most important removal mechanism. For

this, reference is made to the good microbiological

water quality, obtained with slow sand ltration in

surface water treatment. In a slow sand lter, good

disinfection is obtained with a residence time of

only 0.5 - 1.0 days.

4. Surface water with direct treat-Surface water with direct treat-ment

From a global point of view, the direct treatment

of surface water is the most applied method for

drinking water production. This is mainly because

large cities have developed along river banks,

making surface water directly available.

In order to be suitable for drinking water,

suspended solids must be removed together with

pathogenic bacteria. Over the years the removal

of micropollutants has become necessary as well,

together with the construction of storage basins,

to used when the concentration of micropollutants

is too high. Micropollutants often originate from

human activities upstream.

The next section will describe the historical

development of the direct treatment of surface

water. Then, the contemporary treatment schemes

will be described, followed by a description of

future treatment schemes. Finally, the individual

treatment processes in those schemes will be

explained.

4.1 Historical developments

Throughout the ages, because of the increasing

quantitative demand (due to population growthand consumption growth) and the increasing

qualitative demand (because of worse sources

and more stringent quality legislation), direct

treatment methods for drinking water production

have changed drastically.

Traditionally, direct treatment was performed by

clarication in large sedimentation basins and

subsequent slow sand ltration. Characteristic

of this procedure was the enormous spatial

demand and the labor intensive operation (manual

removing of the “Schmutzdecke” from the slow

sand lter).

By adding rapid ltration, the load on the slow sand

ltration was decreased, making an increased

production capacity with traditional means

possible. To guarantee the bacteriological quality

of the drinking water, a safety chlorination was

applied as a nal step in the treatment process.

This caused a small amount of chlorine to bepresent in the water at the customers’ taps.

In time, production needed to be increased further.

This caused the rapid ltration system to be heavily

loaded, resulting in run times that were too short

between backwashing. The problem was solved

by adding a occulant before sedimentation, thus

increasing the effectiveness of the sedimentation

step.

22




When production demands increased further, the

surface area of the slow sand ltration installation

became the bottleneck. The slow sand ltration

not only removed suspended solids, but removed

(pathogenic) bacteria as well. Slow sand ltration

was increasingly replaced by chemical disinfection

(e.g., break-point chlorination). Break-point

chlorination oxidizes ammonium (NH4

+) to nitrogen

(N2) as well.

Increased river contamination necessitated the

construction of reservoirs to be able to stop

the direct intake of river water. Additionally,

micropollutants needed to be removed by activated

carbon (i.e., dosing of powdered activated carbon,PAC). This traditional treatment process (Figure

13) is still widely applied around the world.

The reservoirs were shallow basins at rst. In

these shallow reservoirs however, a considerable

algal population can develop during spring and

summer.

The first step in the treatment process is the

application of microstrainers because of this algal

bloom. Algae are quite difcult to remove by way of

sedimentation, which is, in fact, only possible when

using very high doses of occulants. Additionally,

algae can cause taste and odor problems.

Since the 1970s, mainly deep reservoirs have been

used. In these deep reservoirs, algae growth can

be quite well controlled, making the microstrainers

obsolete. Using deep reservoirs with a very long

residence time will yield a considerable amount of

self-purication in the basins as well.

Water from such reservoirs is typied by a low

concentration of suspended solids (<5 mg/l), few

algae, and a low ammonium concentration. With

this water quality, sedimentation is sometimes

unnecessary, and a very low dose of occulantfollowed by rapid sand ltration can be used.

4.2 Contemporary treatment

Problems with traditional treatment

Contemporary treatment originated from the

chlorination issue and the increased river water

pollution. In 1974 J. Rook, of the Rotterdam Water

Company, discovered harmful by-products from

the chlorination process (disinfectant by-products).

These are mainly trihalomethanes (THMs), from

which chloroform (CHCl3) is produced at the

highest level. THMs are created by the reaction of

chlorine with humic acids present in the water, and

are harmful to human health. The Dutch Decree

on Water Supply sets a standard of 25 µg/l (sum)

for THMs. Chlorination may lead to exceedence

of this standard; but without sufcient chlorination,

the disinfection would be inadequate, a worse

condition from the point of view of public health.This caused the Dutch Decree to temporarily allow

a THM value to 100 µg/l (until January 1, 2006).

In 1987 the insecticide Bentazon was found

in Amsterdam’s drinking water. Like many

micropollutants, this insecticide proved to be

insufciently removed in the treatment process,

even in the case of articial inltration into sand

dunes. Other insecticides, like Atrazin and Diuron,

have also been shown to pass through a traditional

treatment process. Because activated carbonFigure 13 - Tratitional treatment scheme for direct treat-

ment of surface water

WATER TREATMENT

23




ltration does remove these pollutants sufciently,

it has become a typical step in any contemporary

treatment process.

In 1993, a severe Cryptosporidium outbreak

occurred in Milwaukee, Wisconsin (USA), resulting

in 400,000 ill people and 100 deaths. Chlorination

did not prove to be a sufcient barrier to cysts

like Cryptosporidium and Giardia. Using higher

doses and longer contact times will produce

trihalomethanes (THMs). Alternatives to this are

stronger disinfectants like chlorine dioxide (ClO2)

or ozone (O3).

Chlorine dioxide also produces by-products, but

less than when using chlorine or hypochlorite.Disinfection using ozone can produce bromate,

which is also harmful to public health. However,

because sufcient disinfection is essential for

the drinking water supply, the Dutch Decree on

Water Supply allows an increase in the maximum

bromate concentration from 1.0 µg/l to 5.0 µg/l

when using ozone disinfection.

Characteristics of temporary direct treatment

Characteristics of the current treatment of surface

water for production of drinking water are:

- storage reservoirs with a retention time of 1 - 3

months, making an intake stop possible in case

of severe river contamination, and with a depth

of over 20 meters to control algae growth

- process reservoirs with a retention time of about

1 month and a depth of over 20 meters, leading

to signicant self-purication (sedimentation of

suspended solids, ammonium oxidation) while

still keeping algae growth under control

- removal of suspended solids by coagulation(adding flocculants), flocculation and floc

removal by ltration, possibly preceded by

sedimentation or otation

- primary disinfection using a minimal amount of

chlorine or ozone

- removal of micropollutants by activated carbon

ltration

- secondary disinfection using a minimal amount

of chlorine or chlorine dioxide

Example of contemporary direct treatment

(chlorine and activated carbon ltration)

An example of current direct treatment can be

found at the Berenplaat production plant (Figures

14 and 15).

At this site drinking water is produced from

Meuse water, which has rst been stored in the

Biesbosch storage reservoirs. At the Berenplaat

plant, microstrainers form the rst step in the

treatment scheme. This process was selected

because, previously, the water, stored in a shallow

basin, led to algae growth. In the current scheme

the microstrainers could have been omitted, but,

actually, they have been left in service.

To disinfect the water, hypochlorite is added (about

1 mg/l as Cl2). This needs to contact the water for

half an hour, which happens in a tank with a canal

labyrinth. After this phase a occulant is added (ca.

5 mg/l Fe3+ in the form of FeCl3) for the coagulation

Figure 14 - Contemporary direct treatment of surface

water (Berenplaat, before 2006)

24




of suspended solids, and then lime is added to

correct the pH value of the water (ca. 6 mg/l in

the form of CaO). When necessary, a occulant

aid (ca. 1 mg/l in the form of Wispro in winter) and

powdered activated carbon (ca. 7.5 mg/l in case of

severe pollution) are added. The added chemicals

are mixed with the water using mechanical stirrers

for rapid mixing.

The adding of FeCl3 is for removing suspended

solids that remain in the water after the storage

reservoirs. The Fe3+ together with the lime OH-

form small Fe(OH)3

ocs around the particles.

Mechanical stirrers cause turbulence in the water,

and the ocs collide and grow (occulation). Aocculant aid can accelerate this process.

Because the ocs are heavier than water, they

can be removed by sedimentation. This is done

in the oc-blanket clarier (oc removal). The total

retention time in the oc-blanket clarier is about

one hour at a sedimentation rate of not more than

4.8 m/h. The settled ocs (sludge) are drained

into a very large sedimentation basin, where they

accumulate at the bottom.

To remove the remaining turbidity, taste, odor,

and micropollutants, the water is treated using

activated carbon lters. The lters consist of a

layer of granular activated carbon at a height of 1.1

meters, applied over a supportive gravel layer. The

ltration rate is no more than 9.4 m/h, equivalent to

an approximate retention time (empty bed contact

time) of 7 minutes minimum.

Because the lters will slowly clog, they need to

be backwashed with clean water in an upwarddirection every few days. Because carbon

activity decreases over time, the carbon must be

reactivated every 1 - 1.5 years.

Cascades (ve steps with a total height of about

2 meters) bring oxygen into the water; before that

Figure 15 - Drinking water production at the Berenplaat production plant (Zuid-Holland)

WATER TREATMENT

25




water is pumped into the clear water reservoirs.

Aeration is included in the treatment process in

order to add oxygen which could be low in the raw

water because of biological processes. Chlorine

dosing causes the biological activity during the

treatment process to be minimal.

Hypochlorite is added before the clear water

reservoirs (ca. 0.5 mg/l as Cl2) to prevent regrowth

during transportation.

The data shown in Table 5 indicate the water

quality of the untreated and treated water. The

raw water has a high pH value, caused by sodiumhydroxide softening in the Biesbosch reservoirs.

By forming Fe(OH)3 the pH value is reduced, and

by adding lime it is raised again to the desired

level. The suspended solids are mainly removed

in the oc-blanket clariers and during (activated

carbon) ltration. Chlorination causes an increased

chloroform concentration and a reduced E.coli

number.

Example of contemporary direct treatment

(ozone with activated carbon ltration)

Another example of contemporary treatment is

found at the Kralingen production plant (Figures

16 and 17). Here, drinking water is also produced

from Meuse water from the Biesbosch reservoirs.

At the Kralingen plant a occulant is added rst (ca.4 mg/l Fe3+ in the form of FeCl

3), before the water

goes through a static mixer. This causes small

Fe(OH)3 ocs to form and to include pollutants

from the water. If necessary, another occulant aid

(ca. 1 mg/l in the form of Wispro, during winter)

is added. In four serial occulation compartments

having a total retention time of at least 20 minutes,

slowly rotating mixers cause the ocs to grow. The

mixing decreases in intensity in each consecutive

compartment in order to prevent ocs from being

destroyed. The ocs are removed in a lamella

Table 5 - Quality data of the raw and clear water at the

Berenplaat drinking water production plant

(Zuid-Holland)



pH - 9 8.1EC mS/m 51 54

SI - 0.9 0.1

Turbidity FTU 2 0.1

Na+ mg/l 46 49

K+ mg/l 6 6

Ca2+ mg/l 51 54

Mg2+ mg/l 8 8

Cl- mg/l 72 74

HCO3- mg/l 87 95

SO42- mg/l 64 83

NO3- mg/l 3 3

O2

mg/l 11.1 10.8

CH4 mg/l - -CO

2 mg/l 0.3 1.3

Fe2+ mg/l - -

Mn2+ mg/l - -

NH4+ mg/l - -

DOC mg/l 3.6 2.6

E.coli n/100 ml 100 0


Chloroform µg/l 0 38

Bromate µg/l < 2 2.0

Figure 16 - Contemporary direct treatment of surface

water with ozonation

26




separator where they settle between ascending

plates. This arrangement creates an enormous

settling surface in a relatively small area. Particles

with a sedimentation rate of over 1.2 m/h are all

separated in this installation. The settled ocs slide

down over the plates into a sludge thickener, which

is equipped with stirrers. The thickened sludge is

pumped to sludge-drying beds for dewatering.

After the occulation and oc removal, sulfuric

acid is added first to lower the pH, because

ozone is more effective at low pH values. Ozone

is produced locally from liquid oxygen. In ozone

generators the oxygen is exposed to high electric

voltages, thus creating ozone. By means of adiffuser, the ozone is injected into the water (ca.

1.2 - 2.0 mg/l in the form of O3). The ozone spreads

through the water in the form of ne dissolving

bubbles, being active there during a contact period

of 8 - 10 minutes. The ozone gas that is released at

the water’s surface is destroyed thermally. Ozone

kills bacteria and viruses, destroys micropollutants,

and improves the taste of the water.

To remove the remaining turbidity, the water is

treated in a dual-layer sand lter. For an effective

performance of this lter, rst an extra occulant is

added (ca. 0.5 mg/l Fe3+ in the form of FeCl3).

The sand lters have a surface area of 9 by 4

m and consist of a sand layer of 0.7 m and an

anthracite layer of 0.8 m. Below these layers there

is a gravel support layer. The ltration rate is a

maximum of 20 m/h. Because the lters clog, they

are backwashed daily with air (max. 80 Nm/h) and

water (max. 45 m/h) in an upward direction.

Subsequently, the ltered water is treated with

activated carbon for an approximate contact

period (empty bed contact time) of 10 minutes. Theremaining micropollutants and the taste and odor

compounds are removed. Because the activated

carbon activity decreases in time, it needs to be

reactivated every 1 - 2 years. Also, every two to

three weeks the lters need to be backwashed

in order to remove suspended solids. After the

activated carbon treatment, sodium hydroxide is

added to correct the pH value.

Figure 17 - Kralingen drinking water production plant

WATER TREATMENT

27




To make sure that microbiological regrowth does

not occur during distribution, hypochlorite is added

(ca. 0.3 mg/l in the form of Cl2).

Table 6 shows the water quality of both the raw and

treated water. The raw water has a high pH value,

caused by the softening with sodium hydroxide

in the Biesbosch reservoirs. Due to formation of

Fe(OH)3 and sulfuric acid, the pH value is reduced,

and by adding sodium hydroxide it is increased

again to the normal value. The suspended solids

are mainly removed in the lamella separators and

the dual-layer lters. Adding ozone results in anincreased bromate content and a reduced E.coli

number.

Mainly, the concentrations of DOC and Bentazon

are reduced during the activated carbon ltration.

With ozone, the increased retention time, and the

greater biological activity, DOC removal is better at

Kralingen than at the Berenplaat production plant.

The low chlorine dosing results in a small increase

in chloroform in the water.

4.3 Future treatment

Problems of contemporary treatment

Contemporary treatment techniques still face some

problems, such as the by-products (THMs and

bromates) that are formed during disinfection and

oxidation, due to the discovery of new emerging

micropollutants, and the required prevention of

Legionella.

The effective removal of Cryptosporidium and

Giardia requires high dose ozone. The tightened

regulations regarding bromate make this more

difcult. Besides, new and difcult to remove polar

micropollutants have been discovered. These

compounds may require an oxidation process

with high doses, which will again give rise to the

formation of undesirable by-products.

Also, the increase of hormones in surface water

(e.g., estrogen) and materials which act as endo-

crine disruptors and lead to hormonal deviations

will be important in future drinking water production

from surface water. Finally, the Legionella issue willrequire an improved water quality in order to reduce

Legionella growth in the distribution network. This

will require a further reduction in the amount of

assimilable organic matter (AOC) in the water.

The above developments require a renewed

orientation of the integral setup of treatment

schemes for the direct production of drinking water

from surface water. It may be that biological and

physical processes will increasingly take over the

role of the chemical processes for disinfection and

oxidation (Figure 18).

Table 6 - Quality data of the raw and clear water of the

Kralingen production plant (Zuid-Holland)



pH - 9 8.2

EC mS/m 51 55SI - 0.9 0.1

Turbidity FTU 2 0.05

Na+ mg/l 46 52

K+ mg/l 6 6

Ca2+ mg/l 51 51

Mg2+ mg/l 8 8

Cl- mg/l 72 73

HCO3- mg/l 87 94

SO42- mg/l 64 85

NO3- mg/l 3 3

O2

mg/l 11.1 10.2

CH4

mg/l - -

CO2 mg/l 0.3 0.9Fe2+ mg/l - -

Mn2+ mg/l - -

NH4+ mg/l - -

DOC mg/l 3.6 1.9

E.coli n/100 ml 100 0


Chloroform µg/l 0 1.8

Bromate µg/l < 2.0 3.9

Figure 18 - Future treatment of surface water: physical

or chemical

28




Biological processesIn biological processes, many pollutants are

assimilated by biomass and removed in this way.

Also, biological processes will result in reduced

amounts of organic matter (DOC, AOC, etc.).

The treated water should be biologically stable,

so that the biological activity in the distribution

network will be low and residual disinfection will

be unnecessary.

The biological treatment processes that are

currently considered for large-scale applications

are:

- biologically activated carbon ltration

- slow sand ltration

For Amsterdam’s water supply some steps in

this direction have been taken in recent years,

and some aspects of it are currently operational.

Further optimization of the contemporary treatment

processes is being researched.

Physical processesPhysical processes currently being considered for

large-scale applications are:

- UV disinfection (Figure 19)

- membrane ltration

By exposing the water to UV radiation, the DNA

structure of organisms is destroyed, thereby

stopping growth. It has proved very effective to

combine UV disinfection with hydrogen peroxide

as a strong oxidant. Both processes have not

shown any harmful side-effects to date. An

Figure 19 - UV disinfection (right) followed by activated

carbon ltration (left) at Berenplaat produc -

tion plant (Zuid-Holland)

Figure 20 - Direct treatment of surface water at Andijk

production plant (2005)

Figure 21 - UV / H 2 O

2 (Andijk)

WATER TREATMENT

29




example of such a system is the Noord-Holland

(Andijk) water supply (Figures 20 and 21).

With membrane ltration, the water is pressurized

through a membrane. These membranes are

available in several different pore sizes (Figure

22).

Ultra- and microltration mainly retain the coarser

pollutants, like suspended solids, cysts and

bacteria. Nanoltration also retains divalent ions

(Ca2+, SO4

2- etc.), most larger organic compounds

(humic acids), and most micropollutants. Here,

cysts, bacteria and viruses are entirely ltered

out. Reverse osmosis increases the ltration to

monovalent ions and almost any micropollutant.

There are some objections to the application of

membrane ltration:

- risk of membrane defects and thus incomplete

disinfection

- disposal of concentrate

- high costs of construction and operation

Recently, a treatment plant based on membrane

ltration (ultraltration followed by reverse osmosis,

Figures 23 and 24) was started up in Noord-Holland

Filtrationmethod

Particles

Molecularweight

Size (μm)

acids

viruses

humic acids

bacteria

algae

sand

clay silt

cysts

0.001 0.01 0.1 1.0 10 100

100 200 1,000 10,000 20,000 100,000

nanofiltration

ultrafiltration

microfiltration

1,000

conventional filtration

reverse osmosis

Figure 22 - Application elds for membrane ltration

Figure 23 - Membrane ltration plant in Heemskerek (Noord-Holland)

30




(Wijk aan Zee). The produced water is mixed with

drinking water from a system of articially inltrated

surface water.

5. Surface water with inltration

5.1 Surface water with open inltration

The source of inltration water is surface water.

However, the disadvantages of surface water

include a temperature and salinity vary throughout

the year, contamination from pathogenic micro-organisms, and the possibility that, even after

treatment, the growth and settling of particles in

the distribution network may occur.

By inltrating the surface water into the ground, its

quality is improved. This means that the pathoge-

nic microorganisms are degraded and that the

water is in a better biological and chemical state,

causing no settling and no regrowth to occur.

Also, the temperature changes are levelled. Both

the temperature and the salinity will be more or

less constant.

Water can be inltrated into freatic groundwater.

When the supply of treated surface water is

obstructed somehow (e.g., accident, network

repair), it is possible to continue the abstraction for

some time. During this time the groundwater level

decreases, but this is, to some extent, acceptable

without damaging the natural biology.

Inltration projects cover large areas that also needto be protected, because they mainly deal with

large amounts of water. However, as the inltration

area is always developed as a natural area, it will

have a high recreational value.

Figure 25 shows an example of an intake stop

because of contamination of the surface water by

insecticides. Because the contamination occurred

in winter, an intake stop of some weeks could be

taken without damaging the environment in the

inltration area.

Figure 24 - Direct treatment of surface water at Heems-

kerk production plant

Figure 25 - Pollution at the intake point of the WRK

Nieuwegein

10 15 20 25 30

November 2001

5 10 15 20 25 30

December 2001

4 9 14 19 240.00

0.10

0.20

0.30

0.40

0.50

0.60

January 2002

c o n c e n t r a t i o n

( μ g / l )

isoproturon chloride proturon norm limit

intake stop intake stop

WATER TREATMENT

31




5.2 Surface water with deep inltration

Expansion of natural inltration into the dunes is

not always possible because of environmental

aspects. A water company using water from the

dunes has, nevertheless, two possibilities to

increase capacity.

First, the pre-treated surface water can be puried

directly (see direct treatment), after which the

water is mixed with the inltrated water.

Second, the company may apply deep inltration.

Deep inltration inltrates the water into a conned

aquifer. Because of the enclosing clay layers,

there is almost no exchange with the freatic water

above, so the inltrated water does not inuencethe ecosystem.

aWhen using deep inltration, care has to be

taken during the preliminary treatment. The

fewer particles that are present in the water, the

smaller the chance that the inltration wells will

clog. Storage in deep inltration is limited. When

extracting water without supplying the necessary

water for a long period, the chance of salt water

intrusion exists.

Deep inltration is not pursued only in the dunes.

Also in other parts of the Netherlands, deep

inltration is used for the production of drinking

water (Figure 26). The requirements are that the

soil is sufciently permeable and that there are

conning layers in the underground.

5.3 Pre-treatment

To make the surface water suitable for transport

and inltration, suspended particles need to beremoved rst.

The standard pre-treatment of surface water

consists of occulation followed by oc removal

and rapid ltration (Figure 27).

Flocculation is achieved by adding a occulant,

which removes the negative charge of the colloid

particles, thereby making occulation possible.

These flocs can be removed by means of

sedimentation in large ponds or in compact lamella

separators, or by means of otation.

Not all ocs are removed during sedimentation.

Small ocs remain suspended and need to be

removed by means of rapid ltration.

Figure 26 - Deep inltration at Someren (Noord Bra-

bant)

Figure 27 - Preliminary treatment of surface water for

inltration

32




In the ocs, many other materials are removed

like heavy metals (being positively charged and

adsorbed to the ocs), and microorganisms. As

a occulant, mostly a trivalent metal salt, like iron

chloride (FeCl3), iron chloride sulfate, or aluminum

sulfate, is used.

At the WRK Nieuwegein production plant (NV

Watertransportmaatschappij Rijn-Kennemerland),

occulation is accomplished with iron chloride,

followed by settling in a large sedimentation tank,

and rapid ltration (Figure 27 and 28, Table 7).

The quality of surface water varies throughout the

year. For example, the turbidity at the Nieuwegein

site varies between 5.5 and 25.5 FTU and the

temperature between 2 and 23°C. This inuences

the settling behavior, the ltration, and the biological

processes.

During the treatment process the composition of

the water changes. The quantity of suspended

solids, the turbidity, the amount of heavy metals,

like cadmium and nickel, and the colony count

decrease.

By adding iron chloride, the chloride concentration

of the water rises. On average 3 mg/l Fe3+

areadded, implying an increase in the chloride

concentration of 5.7 mg/l. The ferric ions form a

compound together with the hydroxide ions, thus

removing hydroxide ions and reducing the pH

value.


Nieuwegein (Utrecht)



pH - 8 7.8

EC mS/m 80 80SI - 0.4 0.2

Turbidity FTU 10.4 0.2

Na+ mg/l 80 81

K+ mg/l 6 6

Ca2+ mg/l 81 81

Mg2+ mg/l 11 11

Cl- mg/l 149 155

HCO3- mg/l 157 156

SO42- mg/l 66 67

NO3- mg/l 4 4

O2

mg/l 9.2 7.3

CH4

mg/l - -

CO2 mg/l 2.6 4.4Fe2+ mg/l - -

Mn2+ mg/l - -

NH4+ mg/l - -

DOC mg/l 3.9 3

E.coli n/100 ml 5,000 50

Bentazon µg/l 0.2 0.2


Bromate µg/l < 2.0 < 2.0

Figure 28 - Open sedimentation pond at Nieuwegein

(Utrecht)

Figure 29 - Treatment scheme of inltration water where

organic micropollutnats are removed

WATER TREATMENT

33




The quality standards for inltration water have

been made stricter.

Originally, preliminary treatment was performed

in order to prevent contamination of the transport

pipelines and the clogging of the infiltration

ponds. Nowadays, another requirement is that

no elements foreign to the inltration environment

accumulate (like organic micropollutants). The

standard pre-treatment process for infiltration

water does not remove organic micropollutants

like insecticides. This requires the preliminary

treatment process to be expanded with activated

carbon ltration (Figure 29).

At the WRK Andijk site, the preliminary treatment

process consists of a reservoir, occulation, ocremoval (lamella settling), rapid filtration and

activated carbon ltraton. Iron chloride sulfate

(FeClSO4) is used as a occulant.

Growth of algea is common in the IJsselmeer.

Therefore, the amount of organic matter (DOC)

and the turbidity are high (Table 8). To reduce the

algae, the occulant needs to be added in relatively

large amounts, 20 mg/l Fe3+ on average. This

causes the pH value of the water to decrease. To

increase the pH value, lime (Ca(OH)2 is added.

Because WRK Andijk water is now used for both

deep infiltration and membrane filtration, the

requirements regarding turbidity and the clogging

capacity, expressed as MFI (membrane fouling

index), have been increased. Those demands may

be met by the current treatment process, given that

the process is well-controlled.

At WRK Andijk no inexpensive iron chloride nor

caustic soda is added, because the IJsselmeer

contains large concentrations of chloride and

sodium, which should not be further increased.

5.4 Inltration

Pre-treatment is sited at the intake point of the

surface water. This site is rather remote from the

inltration area. Inltration, therefore, requires long

transport pipes.

The pre-treated surface water infiltrates into

the soil, which results in a quality improvement,

including the levelling of concentrations.

A retention time of two months is deemed enough

to make the water reliable, from a microbiological

point of view.

The composition of the inltrated water is different

from the composition of the original dune water.

Foreign water is inltrated into the dunes. This will

cause nutrients to enter the normally poor sand

soil, thus changing the vegetation. In this way

inltration can affect natural areas.

After being abstracted out, the dune water is

transported through pipes to the nal treatment

plant.

5.5 Final treatment of inltrated water

After being abstracted, the water requires a nal

treatment.

The soil passage in the dunes removes micro-

organisms and, at the same time, iron, manganese,

and ammonium ions from the soil are dissolved.

These ions need to be removed from the water.

In case the water was not treated with activated

carbon before inltration, organic micropollutants

Table 8 - Quality data of the raw and treated water at Andijk (Noord-Holland)


Temperature °C 11 11.1

pH - 7.8 7.8

EC mS/m 95 80

SI - 0.1 0.2

Turbidity FTU 8 0.2

Na+ mg/l 113 81

K+ mg/l 9 6

Ca2+ mg/l 70 81

Mg2+ mg/l 14.5 11

Cl- mg/l 150 160

HCO3- mg/l 138 115

SO42- mg/l 37 67

NO3- mg/l 4 4

O2

mg/l 10.1 9.5

CH4

mg/l - -

CO2

mg/l 2.6 1.3

Fe2+ mg/l - -

Mn2+ mg/l - -

NH4+ mg/l - -

DOC mg/l 8 3.3

E.coli n/100 ml 5,000 50



Bromate µg/l < 2.0 < 2.0

34






reactors are used after aeration and before rapid

ltration, in order to remove the carry-over.

When they have not been removed during

preliminary treatment, organic micropollutants

are removed by means of powdered carbon

or activated carbon filtration. The addition of

powdered carbon is done before weir aeration,

because that process will sufficiently mix the

powder with the water. When activated carbonltration is used, it is done after rapid ltration.

To improve the removal of organic micropollutants,

ozonation can be used. Ozonation oxidizes

organic macro-molecules into smaller organic

molecules, which can be removed more easily

during activated carbon ltraton. A disadvantage

of the combination of ozonation and activated

carbon ltration is that there will be a high rate of

biological activity in the activated carbon lters.

This may cause microorganisms to be present

in the water, which will require disinfection after

activated carbon ltration. This disinfection is done

in the slow sand lters.

All in all this makes for quite an extensive treatment

process, especially considering the fact that the

water has been pre-treated before inltration and

transported over large distances. Therefore, the

price of drinking water prepared from inltration

water is higher than that of water produced from

groundwater.

Figure 31 - Inltration area and post-treatment plant near Scheveningen

36




coagulantdosing

Coagulation

and occulation

WA T

E R T R E A T M E

N T

WATER TREATMENT

+

humic acid(- charge)

Fe(OH)2+

(+ charge)

f

e

d

c

b

a

a raw water feedb stirring mechanismc blending space

d floc blanket

e clear water exitf floc exit



Framework

This module represents coagulation and occulation.

Contents


1. Introduction

2. Coagulation

2.1 Theory of coagulation

2.2 Coagulation in practice

3. Flocculation

3.1 Theory of oc formation

3.2 Floc formation in practice

38

COAGULATION AND FLOCCULATION WATER TREATMENT



1 Introduction

In surface water different compounds are present

that must be removed if drinking water is to be pro-

duced. The compounds can be subdivided into:

- suspended solids

- colloidal solids

- dissolved solids.

Suspended solids have a diameter larger than 10-6

m, colloidal solids between 10-9 and 10-6 m and

dissolved solids smaller than 10-9 m.

Particles with a diameter larger than 10-5 m, and

a specic density larger than 2,000 kg/m3

willsettle in water. Smaller particles will also settle,

but more slowly.

In Table 1 the settling time of particles with a den-

sity of 2,650 kg/m3 (e.g., sand) is given.

To be removed, particles that are smaller than

10-5 m must be made larger or heavier. The latter

is impossible and, therefore, removal is only pos-

sible by increasing the particle size.

During the coagulation process, coagulants are

added to the water to aid in oc formation. These

ocs are precipitates in water, wherein small par -

ticles are incorporated.

To express the concentration of compounds in

water, sum parameters are used. The most im-

portant sum parameters for surface water are

“suspended solids” concentration (dry weight),

“turbidity,” “natural organic matter” (expressed in

TOC/DOC) and “color.”

“Suspended solids” concentration and turbidity

(Figure 1) are caused by colloidal particles (order

of magnitude 0.1 - 10 µm). Colloidal particles are

negatively charged and repulse each other.In the tropics high concentrations of suspended

solids can occur and rivers can be become “mud

ows” (Figure 2).

Color (Figure 3) is caused by humic substances

(order of magnitude 0.01 µm). The charge of humic

substances (and thus the removal) is dependent

upon the pH of the water.

In Table 2 the water quality data from the surface

water of several rivers in the Netherlands and in

tropical countries are given. The high values of

organic matter and color in the Drentsche Aa are

caused by peat-containing soils (with high organic

matter content) that the river crosses.

2 Coagulation

The coagulation process is the dosing of a co-

Figure 2 - Rivers in the tropics sometimes have high

suspended solids contents

Diameter

(m)

Types of particles Settling time

over 30 cm

10-2 gravel 0.3 sec

10-3 coarse sand 3 sec

10-4 ne sand 38 sec

10-5 silt 33 min

10-6 bacteria 35 hours

10-7 clay 230 days

10-8 colloids 63 years

Table 1 - Settling time of particles with a density of

2,650 kg/m3

Figure 1 - Turbidity in surface water

39

WATER TREATMENT COAGULATION AND FLOCCULATION



agulant in water, resulting in the destabilization

of negatively charged particles.

2.1 Theory of coagulation

Coagulants

To remove particles present in water, the particles

must be incorporated into ocs. These ocs will

be formed after dosing coagulant.

In the Netherlands iron chloride (FeCl3) is fre-

quently used as the coagulant. Alternatively, alu-

minum sulfate (Al2(SO

4)

3) can be applied.

Iron

Iron chloride is easy to dissolve in water; the solu-

bility product (Ksp

) is 27.9 mol4·l-4. Consequently,

162 mg FeCl3 can be dissolved in one liter of

water, resulting in 55.8 mg/l Fe3+ and 106.5 mg/l

Cl-.

In addition to other ions, the ions Cl-, SO4

2-, Na+,

Ca2+, H3O+- and OH- are dissolved in water.

The OH- ions play an important role in coagula-

tion. Fe3+

- and OH-

ions precipitate, because the

solubility product of iron hydroxide is low. Since

Ksp Fe(OH)3

= 1 · 10-38 mol4·l-4, only 7.8·10-10 mol/l

Fe3+ and 2.34·10-9 mol/l OH--ions can be present

in water.

When the concentration of these ions is higher,

they will precipitate into Fe(OH)3-ocs.

When the pH of the surface water is known ,

the concentration of iron (Fe3+) ions can be

calculated using the solubility product of

iron hydroxide and the ion product of water:

+ − −

+ − −

+ ⋅ → = ⋅

⋅ + = ⋅

3 38

3 sp

14

2 3 w

Fe 3 OH Fe(OH) K 1 10

2 H O H O OH K 1 10

Rewriting the water equilibrium results in the fol-

lowing equation:

− +

+

⋅= ⋅ = ⋅ ⇒ =

-1414 - -

w 3

3

1 10K 1 10 [H O ] [OH ] [OH ]

[H O ]

Combining the equation mentioned above with

the solubility product of iron hydroxide gives:

+

++ +

−

+ +

⋅ = ⋅ ⇒

= ⋅ = ⋅ ⋅⋅

= ⋅ + ⋅ = − ⋅

3 - 3 -38

33 -38 4 33

314 3

3 4

3

[Fe ] [OH ] 1 10[H O ]

[Fe ] 1 10 1 10 [H O ](1 10 )

log[Fe ] log(1 10 ) 3 log[H O ] 4 3 pH

In addition to iron hydroxide the following hydro-

lyses products of Fe3+ are also formed:

Fe(OH)2+, Fe(OH)2

+, Fe(OH)4-.

In Table 3 the solubility constants of different reac-

tions are given. From here, after some calculation,

Figure 4 can be constructed.

River Suspended solids

(mg/l)

Turbidity

(NTK)

Color

(mg Pt/l)

DOC

(mg/l)

Rhine 9 - 53 5.5 - 22.5 9 - 17 3.1 - 6

Meuse 4 - 31 2.2 - 27 10 - 22 3.4 - 5.4

Biesbosch reservoirs 1.5 - 9 0.9 - 5.6 6 - 12 3.2 - 4.0

IJsselmeer 4 - 115 2.5 - 4.0 10 - 30 5 - 13.3

Drentsche Aa 2 - 20 3.4 - 39 10 - 100 4.8 - 14.9

Tropical river 10,000 5,000 1,000 500

Drinking water < 0.05 < 0.1 < 20 1

Table 2 - Water quality data of several rivers

Figure 3 - Color in water

40






trostatic coagulation does not play an important

role in water treatment.

Adsorptive coagulation

In adsorptive coagulation, particles are adsorbed

to the positively charged hydrolyses products

FeOH2+ and FeOH2+.

These products mainly occur at low pH (Figure 4).

The optimal pH-range for adsorptive coagulation

with iron salts is between 6 and 8; the optimalpH-range with aluminum salts is narrower and is

about 7.

Characteristics of adsorptive coagulation are that

dosing is proportional to the removal of organic

matter and that restabilization can occur after an

overdose of coagulant. After an overdose, the

colloids will be positively charged and repulsion

of the particles will take place.

Adsorptive coagulation is a rapid process. Within

one second, positively charged hydrolyses prod-

ucts are formed and are adsorbed to the negatively

charged particles.

Precipitation coagulation

In precipitation coagulation, or sweep coagula-

tion, colloids are incorporated into neutral (iron)

hydroxide ocs. This mechanism occurs mainly

in waters with low suspended solids content (10

mg/l). In order to form hydroxide flocs, more

coagulant must be dosed than is necessary foradsorptive coagulation.

2.2 Coagulation in practice

Jar testThe coagulation process can be researched by

executing jar tests. In this test the coagulation and

oc formation process is simulated.

The jar-test apparatus consists of 6 jars lled

with water (Figure 9). To each jar a certain dose

of coagulant is added. After rapid mixing, a slow

stirring, and a settling phase, the water turbidity

is measured.

By modifying the process conditions (dosage, pH,

occulation time, settling time, stirring energy for

mixing and/or occulation), the optimal conditions

can be determined.

Mechanisms

The coagulation mechanisms discussed above

occur in practice in parallel. This can be illustrated

by discussing the results of several jar-test experi-

ments.

Figure 10 - Results of jar-test experiment with varying

coagulant dosing

0

1

2

3

4

5

0 4 8 12 16 20 24

dosage Fe (mg/l)

t u r b i d i t y ( N

T U

Figure 9 - Jar-test apparatus

Figure 8 - Mechanism of precipitation coagulation

humic acid colloid Fe(OH)3-floc

+ +

Fe3+ + H2O Fe(OH)n+

Fe(OH)3-floc

< 1 sec 1-7 sec

Fe FeFeFeFe OHOHOHOHOH OH

OH

OH

OH

OH

OH

OH

OH OH

OH

Figure 7 - Mechanism of adsorptive coagulation

+

humic acid(- charge)

Fe(OH)2+

(+ charge)

42




In Figure 10 the results of a jar-test experimentof Biesbosch water is shown. Biesbosch water

originates from the river Meuse and is collected in

reservoirs. Due to the long retention times (about

6 months) in the reservoirs, the suspended solids

concentration of Biesbosch water is low, about 5

mg/l.

It can be concluded from the gure that turbidity

decreases with an increased coagulant dosing.

The lowest turbidity is attained when about 12 mg/l

iron chloride is dosed. With a higher dosage the

turbidity does not increase and thus restabilization

does not occur.

In Figure 12 a coagulant dose of 12 mg/l and a

varying pH is represented. The turbidity increases

with a decreasing pH (pH<7).

The predominant coagulation mechanism of Bies-

bosch water is precipitation coagulation.

In the province of Zeeland in the Netherlands, the

drinking water company takes its water in from apolder as its source for the water treatment. Polder

water has a high content of organic matter (like

humic acids).

In Figure 13 the results are represented for jar-test

experiments in which the coagulant dose varied

with pH. At pH between 6 and 7 the lowest turbidity

is found. At higher pH the turbidity is higher.

The prevailing mechanism is thus adsorptive

coagulation.

An evident difference between the adsorptive

and precipitation mechanisms is encountered

during the coagulation of water from the Rhine.

The water transport company Rijn-Kennemerland

abstracts raw water from the river at Nieuwegein

(WRK I-II) and from the IJsselmeer (lake) at Andijk

(WRK III).

Although both water sources originate in the river

Rhine, the coagulation mechanisms differ strongly

Figure 11 - Inuence of coagulant dose (left: high dose,

right: low dose)

Figure 13 - Results of jar-test experiment of “polder

water” with varying pH

0

0.5

1

1.5

2

6 7 8 9pH

t u r b

i d i t y ( N T U )

0

0.25

0.5

0.75

1

1.25

r e s t a l u m i n u m

( m g / l )

turbidity

rest aluminum

Figure 12 - Results of jar-test experiment with varying

pH

0

0.5

1

1.5

6 7 8 9

pH

t u r b i d i t y

( N T U

)

dose = 12 mg/l

Figure 14 - Coagulation of Rhine water

0

5

10

15

20

25

30

35

0 10 20 30

dose (mg/l)

t u r b i d i t y

( N T U )

WRK I-IIWRK III

43




(Figure 14). The river water has a higher turbidity

than the lake water; the lake water has a higherhumic acid content than the river water due to

algae bloom in summer.

During coagulation of lake water, restabilization

can occur and the prevailing mechanism is adsorp-

tive coagulation.

Restabilization is not detected in the coagulating

river water and, therefore, the prevailing mecha-

nism is precipitation coagulation.

Mixing

Rapid mixing after coagulant dosing is an impor-

tant design parameter. The coagulant must be

uniformly mixed with the raw water. In case mixing

is poor, local under- and overdosing occurs, result-

ing in poor performance of the process.

The parameter expressing mixing intensity is

called the velocity gradient or G-value.

The velocity gradient is dened as follows:

=µ ⋅

c

PG

V

in which:

Gc = velocity gradient for rapid mixing (s-1)

P = dissipated power (W)

µ = dynamic water viscosity (N·s/m2)

V = volume of mixing tank (m3)

The inuence of the velocity gradient can be de-

termined by jar-test experiments (Figure 15).

When the velocity gradient is low (less intensive

mixing), the residual turbidity will be higher than in

situations where the velocity gradient is high.

In practice, the recommended G-value for rapidmixing is 1500 s-1 , at a minimum.

Two different mixing systems can be applied:

- mechanical mixing

- static mixing

In the rst system mechanical mixers dissipate the

power in the raw water, while in the second system

gravity forces cause the mixing effect. Here, the

dissipated power is a consequence of the head

loss over the mixing tank:

= ρ ⋅ ⋅ ⋅ ∆P g Q H

Figure 15 - Rest turbidity at different Gc -values

t u r b u d i t y

( N T U )

Gc-value (s-1

)

5

4

3

2

1

0100 500 1000 2000

Figure 16 - Mechanical mixers

floc aid dosage

floc aid dosage

floc dosage

floc dosage

coagulantdosing

Figure 17 - Cascade mixer

44




in which:

ρ = density of water (kg/m3)

g = gravity constant (m/s2)

Q = ow (m3/s)

∆H = head loss over mixing tank (m)

The equation for the velocity gradient for static

mixers can be written as:

ρ ⋅ ⋅ ∆=

µ ⋅ τc

c

g HG

in which:

τc = residence time in the mixing zone (s)

The most frequently applied static mixer is the

cascade. Water falls over a weir into a receiving

body. In the turbulent space that is caused by the

falling water, coagulant is dosed.

3 Flocculation

3.1 Theory of oc formation

After coagulation and the resulting destabiliza-

tion of particles, the particles must collide. The

collision of particles can take place under natural

circumstances (perikinetic oc formation) or by

dissipation of mixing energy (orthokinetic oc

formation).

Perikinetic oc formation

During perikinetic oc formation, particles collide

as a result of Brownian motion. Von Smoluchowskidescribed the decrease in the total number of

spherical particles as a function of time with the

following equation:

⋅ ⋅− = α ⋅ ⋅

⋅ µ

2dn 4 k Tn

dt 3

in which:

n = total number of particles per unit water vol-

ume (m-3)

α = collision efciency (-)

K = Boltzmann constant (J·K-1)

T = absolute temperature (K)

Not every collision will result in attachment and

therefore the collision efciency is incorporated

into the equation.

From experiments it can be concluded that periki-

netic oc formation is a fast process but results in

poor settling characteristics of the formed ocs.

Orthokinetic oc formation

By mixing, the collision frequency of the particles

is articially increased. The decrease in the total

number of particles as a function of time is de-scribed by the following equation:

− = ⋅α ⋅ ⋅ ⋅ ⋅3

1 2 v

dn 4n n R G

dt 3

in which:

Gv = velocity gradient for oc formation (s-1)

R = collision radius (m)

n1 = number of particles with diameter d

1 (-)

n2 = number of particles with diameter d

2 (-)

The collision radius is dened by 0.5·(d1+d

2).

Assuming that all particles have the same diam-

eter, the equation can be rewritten as:

⋅ α ⋅ ⋅ ⋅− =

2 3

v4 n d Gdn

dt 3

For spherical particles the volumetric concentra-

tion is described as:

= ⋅ π ⋅ ⋅ 3

v

1c n d

6

Deriving n·d3 and substituting it:

⋅ α ⋅ ⋅ ⋅− =

πv v8 n c Gdn

dt

resulting in a solution for plug ow:

45




− ⋅ ⋅ ⋅= a v vk c G t

o

ne

n

and for completely mixed systems:

=+ ⋅ ⋅ ⋅o a v v

n 1

n 1 k c G t

With these equations it can be calculated that or -

thokinetic oc formation of particles with a diameter

of 1 µm only takes place when velocity gradients

higher than 10 s-1 are applied. Otherwise, periki-

netic oc formation is predominant.

3.2 Floc formation in practice

Parameters that are important to the design of a

oc formation installation are the following:

- residence time T

- residence time distribution

- velocity gradient for oc formation Gv

- oc volume concentration cv.

Residence time

Time is needed for the formation of removable

ocs. The applied residence time varies between

500 and 3600 sec. On average the residence time

for oc formation is about 30 minutes.

To determine the required residence time, jar-testexperiments are carried out.

Residence time distribution

When water ows through a tank, the residence

time of every droplet is different. For some droplets

the residence time is longer and for others shorter

than the average. The consequence is that, in

practice, the floc formation process performs

worse than can be expected, based on theory.

In order to approach the perfect plug ow where

every droplet has the same residence time, criteria

are developed for the design of a oc formation

chamber. A plug ow can be approached when the

ratio between the length and width of a tank is at

least equal to 3.

Mixers in oc formation chambers take care of the

dispersion of energy and collision of the particles.

It is, however, important that the mixers be in line

with the ow direction (Figure 18).

Figure 19 - Plug ow mixing systems for oc formation,

mixer in line with water ow

top view

side view side view

top view

Figure 20 - Mixing device

Figure 18 - Mixers in line and perpendicular to the owdirection of the water

flow parallel to stirring axis: no short circuit flow

flow perpendicular to stirring axis: short circuit flow

if flow speed = 0.03 m/s, tip speed = 1 m/sthen water speed -0.97 to 1.03 m/s

46




If the mixers are placed perpendicular to the ow

direction, some water droplets are accelerated

and other are slowed down, resulting in a larger

residence time distribution.

When the axis of the mixer is in line with the ow,

the inuence is limited.

In two different oc formation systems the length/

width ratio of 3 and the direction of the mixers are

considered:

- horizontal, long and narrow (Figure 19 right).

- vertical, deep and narrow chambers (Figure 19

left).

Velocity gradient After coagulation the colloids and humic acid are

destabilized and many small particles are present

in the water.

Mixers that are placed in the oc formation cham-

bers dissipate energy in the water (Figure 20),

resulting in the collision of neutral particles and

the formation of ocs.

The degree of energy dissipation is expressed,

like for coagulation, in the velocity gradient. The

velocity gradient is mainly created by mixers.

Alternatively, hydraulic oc formation can be ap-

plied where the head loss between two chambers

delivers the energy for the formation of ocs. The

drawback of hydraulic occulation is the uneven

energy input.

The velocity gradient for oc formation is ex-

pressed in the parameter Gv and is dened by:

=µ ⋅

v

PG

V

The energy dissipation from the mixers can be

calculated with the following equation:

= ρ ⋅ π ⋅ − ⋅ ⋅ ⋅ ⋅ −∑43 3 3 4

w 2 d blade u iP (1 k ) N (c L (r r ))

in which:

k2 = constant≈ 0.25 (-)

N = rotation speed (rpm)

Cd = constant ≈ 1.50 (-)

Lblade

= length of mixer blade (m)

r u = distance from exterior of mixing blade to

axis (m)

r i = distance from interior of mixing blade to

axis (m)

According to the formula for the dissipation en-

ergy from the mixer, the rotation speed is the only

operation parameter. The other parameters are

already determined during the design process.

The velocity gradient in operation can thus be

calculated by:

= ⋅ 3

vG const. N

Figure 21 - Tip velocity

Figure 22 - Floc formation installation WRK I/II

Figure 23 - Floc formation installation WRK III, division

in different compartments

47




The calculated velocity gradient is the average inthe oc formation chamber.

The velocity of the mixing blade in the oc for -

mation chamber depends on the radius and the

rotation speed. The velocity is greatest at the tip

of the mixing blade and is called the tip velocity

(Figure 21) and can be calculated by:

= ⋅ π⋅ ⋅tipv 2 r N

The higher the rotation speed, the higher the tip

velocity. When the tip velocity is higher than 1

m/s, formed ocs are broken up.

When the rotation speed of a mixer is known,

the maximum radius of a mixing blade can be

determined. A rotation speed of 4 rotations per

minute allows a maximum radius of 2.4 meters.

During coagulation small neutral particles are

formed and grow, after collision, into removable

ocs.To increase the collision frequency of the particles,

a high mixing intensity must be applied. The small

particles collide and larger particles are formed,

but in the mean time, the risk of oc break-up

increases as a result of uid shear.

Therefore, the oc formation chamber is divided

into several compartments (Figure 22 and 23) with

decreasing velocity gradients (Figure 24).

In the rst compartment the velocity gradient will

be high (about 100 s-1) and in the last compart-

ment the velocity gradient will be low (about 5

s-1). The optimal operation of the mixers must be

determined empirically.

The flow opening between the compartments

must be large enough to avoid local energy dis-

sipation, as is the case in hydraulic occulation

(Figure 25).

Floc volume concentration

Figure 27 - Mixing device oc blanket installation at

Berenplaat

Figure 26 - Floc blanket installation at Berenplaat

f

e

d

c

b

a

a raw water feedb stirring mechanismc blending spaced floc blankete clear water exitf floc exit

Figure 25 - Hydraulic oc formation

Figure 24 - Inuence of G-value on oc formation in

different compartments

5,5,5,5 20,20,10,10 40,20,10,5 80,40,10,50

1

2

3

4

5

G -value (s-1)

t u r b i d i t y

( N T U )

v

48




Figure 28 - Mixing device ow blanket installation at

Berenplaat

Further reading

Water treatment: Principles and design, MWH

(2005), (ISBN 0 471 11018 3) (1948pgs)

Unit processes in drinking water treatment,

W. Masschelein (1992), (ISBN 0 8247 8678

5) (635 pgs)

•

•

49




h= 0.75 m

h= 1.5 m

h= 2.25 m

h= 3.0 m

h/t [m/h]

c u m u l a t i v e f r e q

u e n c y d i s t r i b u t i o n [ % ]100

80

60

40

20

01 2 3 4 50

Sedimentation WA T

E R T R E A T M E

N T

WATER TREATMENT

1

2

3

4

5

t = 0 t = t = 2

tube

constantwater temperature

sample ofthe solution

silt

trap

100

80

60

40

20

00 0.5 1 1.5 2 2.5 3

distance under water surface [m]

s u s p e n d e d s o

l i d s c o n t e n t [ % ] time [s]

600

5400

9001200

1800

2700

7200

3600



Framework

This module represents sedimentation.

Contents


1. Introduction

2. Theory

2.1 Sedimentation of discrete particles

2.2 Horizontal ow settling tanks in practice

2.3 Settling efciency of a suspension

3. Inuences on settling in a horizontal ow tank

3.1 Inuence of turbulence

3.2 Inuence of stability

3.3 Inuence of bottom scour

3.3 Inuence of occulant settling

4. Practice

4.1 Determination of the dimensions of an ideal settling tank

4.2 Inlet constructions

4.3 Outlet constructions

5. Settling tank alternatives

5.1 Vertical ow settling tank

5.2 Floc blanket clarier

5.3 Tray settling tanks

5.4 Tilted plate settling

52

SEDIMENTATION WATER TREATMENT



1 Introduction

Sedimentation is a treatment process in which

suspended particles, like ocs, sand and clay are

re-moved from the water.

Sedimentation can take place naturally in reser -

voirs or in compact settling installations.

Examples of settling installations are the horizon-

tal ow settling tanks, the tilted plate settlers and

the oc blanket installations.

Sedimentation is frequently used in surface water

treatment to avoid rapid clogging of sand lters

after coagulation and oc formation (Figure 1).

Sedimentation is applied in groundwater treat-ment installations for backwash water treatment.

In horizontal ow settling tanks (Figure 2) water

is uniformly distributed over the cross-sectional

area of the tank in the inlet zone.

A stable, non-turbulent, ow in the settling zone

takes care of the settling of suspended matter in

the settling zone.

The sludge accumulates on the bottom or is con-

tinuously removed.

In the outlet zone the settled sludge must be pre-

vented from being re-suspended and washed out

with the efuent.

Sedimentation occurs because of the difference

in density between suspended particles and wa-

ter.

The following factors inuence the sedimentation

process: density and size of suspended particles,water temperature, turbulence, stability of ow,

bottom scour and occulation:

- density the greater the density of the par-

ticles, the faster the particles set-

tle

- size the larger the particles are, the

faster they settle

- temperature the lower the temperature of the

water is, the higher the viscosity,

so the slower the particles settle

- turbulence the more turbulent the ow is, the

slower the particles settle

- stability instability can result in a short-cir -

cuit ow, inuencing the settling

of particles

- bottom scour during bottom scour, settled

particles are re-suspended and

washed out with the efuent

- occulation occulation results in larger parti-

cles, increasing the settling veloc-

ity.

2 Theory

2.1 Sedimentation of discrete particles

Discrete particles do not change their size, shape

or weight during the settling process (and thus do

not form aggregates).

A discrete particle in a uid will settle under the

inuence of gravity. The particle will accelerate

Reservoir

Fe (III)

by sedimentation

Cl2/ClO

2

Floc formation

Floc removal

Ozonation

Filtration

Activated carbon filtration

Clear water reservoir

Figure 1 - Process scheme of a surface water treat-

ment plant

sedimentation zone L

Q

Q

Q

Q

B

H

V0

V0

inlet

zone

outlet

zone

slib zone

Figure 2 - Horizontal ow settling tank

53

WATER TREATMENT SEDIMENTATION



until the frictional drag force of the uid equals

the value of the gravitational force, after which

the vertical (settling) velocity of the particle will be

constant (Figure 3).

The upward directed force on the particle, caused

by the frictional drag of the uid, can be calcu-

lated by:

2wup D sF c v A

2

ρ= ⋅ ⋅ ⋅

in which:

Fup = upward directed force by friction [N]

cD = drag coefcient [-]

ρw = density of water [kg/m3]

vs = settling velocity [m/s]

A = projected area of the particle [m2]

The downward directed force, caused by the dif -

ference in density between the particle and the

water, can be calculated by:

( )down s wF g V= ρ − ρ ⋅ ⋅

in which:

Fdown = downward directed ow by gravity [N]

ρs = specic density of particle [kg/m3]

g = gravity constant [m/s2]

V = volume of particle [m3]

Equality of both forces, assuming a spherical par -

ticle, gives as the settling velocity:

s ws

D w

4v g d

3 c

ρ − ρ= ⋅ ⋅ ⋅

⋅ ρ

in which:

d = diameter of spherical particle [m]

The settling velocity is thus dependent on:

- density of particle and uid

- diameter (size) of particle

- ow pattern around particle.

The ow pattern around the particle is incorporat-

ed in the drag coefcient. The value of the drag

coefcient is not constant, but depends on the

magnitude of the Reynolds number for settling.

For spherical particles the Reynolds number is

given by:

sv dRe

⋅=

ν

in which:

ν = kinematic viscosity [m2/s]

In drinking water treatment practice, laminar set-

tling normally occurs. The Reynolds number for

laminar settling of spheres is Re<1, resulting inthe following relationship between the Reynolds

number and the drag coefcient:

Substitution of this relationship in the equation for

the settling velocity gives the Stokes’ equation:

2s ws

w

1 gv d

18 v

ρ − ρ= ⋅ ⋅ ⋅

ρ

The settling velocity is thus dependent on the vis-

cosity of the uid and also the temperature.

Fup [N]

Fdown [N]

sedimentation speed

Vs [m/s]

Figure 3 - Forces on a settling particle

1,000

100

10

1

0.10.1 1 10 100 1,000 10,000 100,000

Reynolds number [-]

r e s i s t a n c e

c o e f f i c i e n t c D

observed relationship

cD=Re

+Re

+ 0.3424 3

cD=Rena

Figure 4 - Relationship between Reynolds number

and drag coefcient

54




The relationship between kinematic viscosity and

temperature is:

( )6

1.5497 10vT 42.5

−⋅=+

in which:

T = temperature [oC]

When the Reynolds number Re > 1600, settling

is turbulent and when 1<Re<1600, settling is in

transition between laminar and turbulent.

In Figure 4 the relationship between the drag co-

efcient and the Reynolds number is represent-

ed.

In Figure 5 the settling velocity as a function of

particle size and density is shown.

2.2 Horizontal ow settling tanks

in practice

In practice, settling occurs in owing water. An

ideal horizontal ow settling tank has the follow-

ing characteristics:

- at the inlet the suspension has a uniform com-

position over the cross-section of the tank

- the horizontal velocity vo is the same in all

parts of the tank

- a particle that reaches the bottom is denitive-

ly removed from the process.

The ow velocity in a horizontal settling tank is:

oQ

vB H

=⋅

in which:

vo = horizontal ow velocity [m/h]

Q = ow [m3/h]

B = width of the tank [m]

H = height of the tank [m]

The surface loading of a settling tank is deter -

mined by:

Qq

B L=

⋅

in which:q = surface loading [m3/(m2•h)]

L = length of the tank [m]

In Figure 6 the trajectory of a particle is repre-

sented. After t1 the water leaves the tank and

after t2 the particle is settled. The particles will

settle, therefore, when t2 <t1.

The velocity of the particle is divided into horizon-

tal and vertical components and the settling times

can be written as:

2 1s 0 s

H L H B H Lt t

v v v Q

⋅ ⋅≤ ⇒ ≤ ⇒ ≤

ss

1 1v q

v q⇒ ≤ ⇒ ≥

In special cases, when the settling velocity equals

the surface loading, the particle reaches the end

of the tank. This settling velocity is called the criti-

cal velocity vso.It can be concluded that a particle will only be

removed if the settling velocity is greater than or

equal to the critical settling velocity (Figure 7).

Figure 5 - Settling velocity of discrete spherical parti-

cles

T=10oc10,000

100

1

0.01

0.00010.0001 0.001 0.01 0.1 1 10 100

s e t t l i n g v e l o c i t y [ m m / s ]

diameter [mm]

5 0 0 0

5 0 0

5 0

5 1

ρs - ρw =

Figure 6 - Settling in a horizontal ow settling tank

H, t2 v0

L, t1

q

55




After determining the settling velocity of a particle

during a settling test, the surface loading and thus

the dimensions of the tank can be determined.

It is remarkable that, in theory, settling in a hori-

zontal ow settling tank is only determined by the

ow and the surface area of the tank and is inde-

pendent on the height of the tank.

The fraction of the particles that settle in case vs

< vso is (Figure 7):

s s

so so

v T vh

H v T v

⋅= =

⋅

in which:

T = residence time of water in the settling tank[s]

The residence time of water in the settling tank is

expressed as T and equals t1 from Figure 6.

2.3 Settling efciency of a suspension

In a suspension the fraction of particles with a

settling velocity higher than the surface loading

settle completely. The fraction with a lower set-

tling velocity settles partly. The efciency is deter -mined from the cumulative frequency distribution

of settling velocities obtained from a settling test.

The settling test is executed in a cylindrical con-

tainer (column) lled with a homogeneous sam-

ple of the suspension to be tested (Figure 8). At

different time intervals samples are taken at dif -

ferent depths and analyzed for suspended solids,

turbidity or any other index that can be reduced

by settling. The depth is measured with the water

surface as reference. In Table 1 the analyses of a

settling column test at depth h=1.0 m are repre-

sented (Figure 8).

Figure 7 - Settling of a suspension in a horizontal ow

1

2

3

4

5

t = 0 t = t = 2

tube

constantwater temperature

sample ofthe solution

silt

trap

Figure 8 - Settling column and representation of different settling velocities

vs > vso - all particles settle completely

vs = vso - all particles settle completely

vs < vso - some of the particles settle completely

v0

vs

v0H

vs

v0h

L

vs

56




In Figure 9 the cumulative frequency distribution

of the settling velocities is represented. The ratio

of sampling depth and time is given as a func-

tion of the relative solids concentration. The sol-

ids with the lowest settling velocity determine the

residence time of a settling system.

The particles with a settling velocity higher than

the critical settling velocity vso are removed com-

pletely. This is represented in Figure 9 by the redarrow. Expressing the relative solids concentra-

tion for a settling velocity of vso as po, the rst part

of the settling efciency is:

1 or 1 p= −

in which:

r 1 = part of the efciency caused by complete

settling [-]

po = relative solids concentration at surface

loading so [%]

From the particles with a lower settling velocity

than vso, only the particles that enter the tank at a

reduced height will be removed.

From the fraction of particles dp with settling ve-

locity vs, only the fraction h/H or vs/vso will be

removed. This part of the efciency (partial re-

moval) can be described by:

o op ps

2 s

so so0 0

v 1r dp v dp

v v

= =∫ ∫

in which:

r 2 = part of the efciency caused by partial set-

tling [-]

The efciency caused by partial settling is rep-

resented by the blue surface in Figure 9 divided

by the critical settling velocity. Graphically, this

part of the total efciency can be determined as

shown in Figure 10.

p [ % ] 100

80

60

40

20

00 0.5 1 1.5

po

vso vs [10-3 m/s]

vs dp∫po

0

1

vso

equal-sizedsurfaces

Figure 10 - Efciency of partial settling

p [ % ] 100

80

60

40

20

00 0.5 1 1.5

po

vs dp po

0

dp

vso vs [10-3 m/s]

1-po

Figure 9 - Cumulative frequency distribution of settling

velocities

Table 1 - Particle concentration and relative particle concentration from a settling test at a depth of h = 1.0 m

t (s) 0 666 900 1800 2700 3600 5400 7200

c (ppm) 86 84 79 57 41 29 7 3

p=c/co (%) 100 98 92 66 48 34 8 4

57




The equation of the total settling efciency be-

comes:

( )op

o sso 0

1r 1 p v dp

v= − + ∫

For different values of vso the efciency is calcu-

lated and the results are represented in Figure

11.

It can be concluded that with increasing surface

loading of the settling tank (by increasing ow),

the settling efciency decreases.

3 Infuences on settling in a hori-

zontal fow tank

In the preceding paragraphs an ideal ow and

discrete settling were assumed.

In practice, however, the ideal situation does not

exist and the efciency is inuenced by:

- turbulence of ow

- instability of ow- bottom scour

- occulation.

3.1 Inuence of turbulence

In laminar ow in a horizontal ow tank, a particle

follows a straight line.

In turbulent ow, eddies will transport particles in

a random direction, inuencing the settling of the

particles (some settle faster and others slower)

(Figure 12).

With the Reynolds number the ow characteris-

tics can be determined:

- laminar ow: Re < 2000

- turbulent ow: Re > 2000.

The Reynolds number for ow in a tank can be

calculated with:

ov RRe

⋅=

ν

in which:

R = hydraulic radius of a settling tank [m]

The hydraulic radius of a rectangular tank can be

calculated with:

B HR

B 2 H

⋅=

+ ⋅

With the expression vo=Q/(B•H) the Reynolds

0 1 2 3 4 50

20

40

60

80

100

vso [m/h]

e i c i e n c y r %

Figure 11 - Removal efciency in a horizontal ow set-

tling tank

Figure 12 - Inuence of turbulence on settling in a hori -

zontal ow settling tank

2 3 4 6 8 2 3 4 6 8 2 3 4 6 81.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

00.001 0.01 0.1 1

2.0

1.5

1.2

1.11.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

V s

V so

V s

V o

e f f i c i e n c y

[ - ]

Figure 13 - Inuence of turbulence on the efciency of

settling

58




number can be rewritten as:

Q 1Re

B 2 H= ⋅

ν + ⋅

In Figure 13 the settling efciency for turbulent

ow is represented as a function of vs/vso and

vs/vo.

In practice turbulence is not always a disadvan-

tage because, in general, occulant settling oc-

curs (section 3.4). Turbulence increases the col-

lision frequency of particles, thus increasing the

efciency of the occulant settling.

3.2 Inuence of stability

Flow is called stable when short circuiting does

not occur.

In Figure 14 an example of a short-circuit ow

caused by wind effects is illustrated. The wind

creates a dead zone (or eddy) in the corner of

the settling tank. The water ow can then ow,

locally, in the opposite direction from the general

ow through the tank.

Stability of ow is characterized by the Camp

number cp:

2o

p

vc

g R=

⋅

Substituting the equations for vo and R for a rec-

tangular tank, the Camp number becomes:

2

p 3 3

Q B 2 Hc g B H

+ ⋅= ⋅ ⋅

cp > 1• 10-5 stable ow

cp < 1• 10-5 unstable ow

In Figure 15 the minimal residence time (Ti) and

the average residence time (Ta) of water droplets

are represented in comparison with the theoreti-

cal residence time (To) for different values of the

Camp number.

From Figure 15 it can be concluded that the lower

the Camp number is (and thus more short-circuit

ow occurs), the shorter the minimal and average

residence times become. This is due to the de-

crease in the effective cross-section of the set-

tling tank and, therefore, to an increase in owvelocity.

The efciency of a settling tank, therefore, will be

lower than is the case in a stable ow condition.

3.3 Inuence of bottom scour

In theory, a particle is removed from the water

when it reaches the bottom of the settling tank.

In practice, however, it is possible that resuspen-

sion of already settled particles occurs.

In Figure 16 the forces on particles at the bottom

of the tank are shown.

The shear force of water on a spherical particle

is:

lt = × r ×

2w scv

8

down

Figure 14 - Short-circuit ow caused by wind

1.0

0.8

0.6

0.4

0.2

010

-710

-610

-510

-410

-310

-210

-1

Cp

T0

T

Ta

T0

Ti

T0

Figure 15 - Short-circuit ow

59




in which:

τ = hydraulic shear [N/m2]

λ = hydraulic friction factor (λ = 0.03) [-]

vsc = critical scour velocity [m/s]

The shear force of particles at the bottom (me-

chanical friction) is proportional to the submerged

weight of the sludge layer:

f = β . (ρs - ρw ) . g . d

in which:

f = mechanical shear [N/m2]

β = mechanical shear factor (b = 0.05) [-]

In equilibrium the hydraulic shear equals the me-

chanical shear and the critical scour velocity can

be calculated:

s wsc

w

40v g d

3

ρ − ρ= ⋅ ⋅ ⋅

ρ

When the ow velocity in a settling tank is lower

than the scour velocity, bottom scour will not oc-

cur:

v0 <= vsc no bottom scour

Given the surface loading, the width and depth of

a settling tank can be determined based on this

criterion.

3.4 Inuence of occulant settling

During settling, aggregates are formed as a result

of collisions between particles, and settling ve-

locities will increase. This phenomenon is called

occulant settling (Figure 17).

In Table 2 the results of a settling test of a oc-

culant suspension are shown..

In Figure18 the cumulative frequency distribu-

tion of settling velocities is given at different tank

depths. From the fact that the distributions differ

d

f

N

v0

sediment

bottom of tank

Figure 16 - Bottom scour

h = 0.075 m h = 1.5 m h = 2.25m h = 3.0 m

t = 0 s 100 100 100 100

t = 600 s 93 96 98 99

t = 1200 s 81 86 88.5 89.5

t = 1800 s 70.5 77.5 81 83

t = 2700 s 28 38 46.5 53

t = 3600 s 13.5 22 31 40

t = 5400 s 3 8 13.5 20

t = 7200 s 1.5 3 6 9.5

Table 2 - Relative particle concentration from a settling test

t=0 t= t=2 t=3

Figure 17 - Flocculant settling

60






over the height of the tank, it can be concluded

that occulant settling occurs.

From Figure 19 it can also be concluded that the

efciency increases with the increasing depth.For occulant settling, in contrast to discrete set-

tling, the height of the tank is of importance to the

settling efciency.

4 Practice

4.1 Determination of the dimensions of

an ideal settling tank

In ideal settling tanks the ow is stable (cp > 10-5)

without turbulence (Re < 2000).

At a temperature of 10oC these conditions are

met with a horizontal ow velocity and a hydrau-

lic radius of:

vo = 6.4•10-3 m/s

R < 0.41 m

Tanks that meet these conditions are short, wide

and shallow or long, narrow and deep (Figure

20).

These constructions, however, are expensive

due to the amount of space they occupy.

In practice, a tank will be a compromise between

the Reynolds and Camp numbers, on the one

hand, and the construction costs, on the other,

limiting the length/width/depth ratios.

4.2 Inlet constructions

In the preceding paragraphs it was assumed that

the water is uniformly distributed over the cross-

section of the tank, but in practice this assump-

tion is not totally accurate.

For an even distribution of the water over the

width (and depth) of the tank, inlet constructions

are introduced.In Figure 21 an example of an inlet construction

is represented. The inlet velocity is reduced by

introducing several inlet channels, followed by

a diffuser wall that distributes the water over the

entire cross-section of the tank.

A diffuser wall (Figure 22) has openings to dis-

tribute the water over the width (and depth) of the

tank. At the end of the wall, the ow velocity in

the inlet channel is zero and so is the velocity

head. The head loss caused by friction, however,

is lower than the decrease in velocity head, re-

sulting in an increase in the piezometric level.

The water level at the end of the inlet channel

is, thus, higher than the level at the beginning.

The result is that at the end of the inlet channel

more water enters the tank than at the beginning

of the inlet channel. To avoid this uneven distribu-

tion, the head loss over the openings in the dif -

fuser wall must be larger than the difference in

piezometric level induced by the decrease in ow

V o

V o

≈0.82

≈0.41≈

Figure 20 - Settling tanks with laminar and stable

ows

Q

Q4

diffuser wall

sedimentation zone

Figure 21 - Inlet construction

h

v02

2g

Q

0

z h z+

Q

vvi

e ner gy line

Figure 22 - Diffuser wall

62




velocity.

In practices, more as well as alternative inlet con-

structions exist, like the Clifford and the Stuttgar -

ter inlet (Figure 23).

4.3 Outlet constructions

The outlet construction is situated at the end of

the settling tank and generally consists of an

overow weir.

At the outlet construction, re-suspension of set-

tled solids must be prevented and the ow ve-

locity in an upward direction will thus be limited

(Figure 24).

The ow velocity in an upward direction is:

H so1 Q

v v5 B H

= ⋅ <⋅

in which:

vH =outow velocity in an upward direction [m/s]

Resulting in:

L5

H<

Most horizontal ow settling tank have an L/H>5

and thus:

soQ

5 H vn B

< ⋅ ⋅⋅

Therefore, the length of the overow weir must

be several times the width of the tank.

To create sufcient length for the overow weir,

several troughs are placed parallel to each

other(Figure 25).

4.4 Sludge zone and removal

In the sludge zone the solids are accumulated.

The removal of the sludge can be done hydrauli-

cally and mechanically.

Hydraulic sludge removal is done at regular in-

tervals by dewatering the tank and ushing the

sludge with pressured water (from hydrants) to a

hopper at the bottom of the tank from where it is

removed by gravity or by pumping.

Figure 23 - Clifford and Stuttgarter inlet

Clifford inlet Stuttgarter inlet

H Vo

L

Vso VH

Figure 24 - Upward velocity to overow weir

Figure 25 - Overow weir for efuent discharge

63




Mechanical sludge removal is frequently applied

when sludge volumes are large or the sludge is

unstable, resulting in anaerobic decomposition

during storage in the sludge zone.

Mechanical sludge removal consists of scrapers

that transport the sludge to a hopper in the mid-

dle of a round settling tank or near the inlet of a

rectangular tank. From the hopper, the sludge is

removed.

5 Settling tank alternatives

5.1 Vertical ow settling tank

In vertical ow settling tanks the inlet of the wa-ter to be treated is situated at the bottom of the

tank and the water ows in an upward direction

(Figure 26).

The ow velocity equals, in this case, the surface

loading:

o o

Qv s

B L= =

⋅

The result is that only particles with a settling ve-

locity higher than the upow velocity will settle

and others will be washed out:

s >= s0 settles completely

s < s0 does not settle

The settling efciency is entirely determined by

the particles that settle completely (see Figure

9):

or 1 p= −

The settling efciency of discrete particles in ver -

tical ow settling tanks is lower than in horizontal

ow tanks, and vertical ow tanks are therefore

not used for discrete, totally occulated, suspen-

sions.

In the case of occulant settling, vertical ow

tanks are used (e.g., in the form of oc blanketclariers).

5.2 Floc blanket clarier

The oc blanket clarier consists of a (conical)

vertical ow tank (Figure 27).

Coagulant is dosed at the inlet of the clarier and

oc formation occurs in the installation. Small,

light ocs with a settling velocity lower than the

upow velocity are transported with the water

ow in an upward direction and collide with larg-

er, heavier ocs. After attachment, the settling ve-

locity increases until they reach the bottom of the

Q

H

V0

Vs

Figure 26 - Vertical ow settling

Q/A

sludgedischarge

floc blanket

sludgedischarge

A

heavy sludgedisposal possibility

Q

Figure 27 - Floc blanket installation

64






o ow t

s ' sH cos w

+= ⋅

⋅ α −

In Table 3 the design parameters of Figure 31

that are applied in practice are given.

The angle of the plates in co-current systems can

be gentler than in counter-current systems with-

out deteriorating the sludge removal.

Substituting the values of Table 3 into the equa-

tions for surface loading for both co-current and

counter-current systems results in:

oo

ss '

20≈

The space occupied by tilted plate settling tanks

is thus a factor 20 smaller than is needed for hori-

zontal ow tanks.

Both the Camp number and the Reynolds numberdepend on the hydraulic radius and the horizontal

ow velocity.

In Figure 32 the stability boundary, cp > 10-5, and

the turbulence boundary, Re < 2000, are given.

In addition, the combinations of hydraulic radius

and horizontal ow velocity of horizontal ow and

tilted plate tanks applied in practice are shown.

From the graph it can be derived that the ow

in horizontal ow tanks is turbulent and in somecases instable (and short-circuit ow can occur).

The ow in tilted plate tanks, however, is favora-

ble. The Reynolds number is always smaller than

2000, resulting in laminar ow; and the Camp

number is always higher than 10-5, resulting in a

stable ow without short-circuiting.

Design parameters Value

counter-current 550 - 600

co-current 300 - 400

H 1 - 3 m

w 3.4 - 8 cm

t 5 mm

Table 3 - Design parameters

100.000

10.000

1.000

0.100

0.010

0.0010.0001 0.001 0.01 10.1 10

horizontal flowsedimentation tanks

tilted platesedimentation

C p >

1 0 - 5

v s < 0 . 0

5 m / s

R e < 2 , 0 0 0

h y d r a u l i c r a d i u s [ m ]

horizontal flow speed v0 [m/s]

Figure 32 - Hydraulic conditions for optimal settling

q

H

Vo

V‘so

L

t

w

w sin

L cos

Figure 31 - Flow through a tilted plate settler

Further reading

Sedimentation and otation, TU-Delft (2004)

Water treatment: Principles and design,

MWH (2005), (ISBN 0 471 11018 3)

(1948 pgs)

•

•

66




0.01 0.1 1 10 100 1,000

1

0.1

0.01

0.001

floc size (µm)

c o l l i s i o n

p r o b a b i l i t y

( - )

flocs: 100-1,000 µmtherefore interception

T = 10 oCdb = 40 µmrd = 1,003 kg/m3

η η η

Flotation WA T

E R T R E A T M E

N T

WATER TREATMENT

= floc = air

t= t 2t 3t 4t t= t 2t 3t 4t

A

BC

D

transport mechanisms:

A = diffusion

B = interception

C = inertia

D = sedimentation

covered path

flow path

floc

air bubble



Framework

This module explains otation

Contents


1. Introduction

2. Principle

3. Theory

3.1 Saturation unit for the supply of air

3.2 Efciency of the bubble lter in the ltration zone

3.3 Collision probability between bubbles and ocs in the ltration zone

3.4 Determination of the surface loading of the separation zone

4. Practice

4.1 Design parameters 4.2 Saturation

4.3 Flotation tank

68

FLOTATION WATER TREATMENT



1 Introduction

In the past, otation was mainly used as a oc

removal process in the Scandinavian countries

and the UK.

Zevenbergen (Brabant Water) was the rst Dutch

treatment plant to make use of that process (1979).

Other treatment plants where otation is applied

are Braakman (Evides), Elsbeekweg Enschede

(Vitens) and Scheveningen (Dune Water Company

South Holland).

Flotation is applied to remove ocs during surface

water treatment and is preceded by oc formation

(occulation).In the otation process very small air bubbles are

used to “air-lift” the ocs to the water’s surface.

The number and size of the air bubbles is the key

factor for the upward velocity of the ocs, and thus

for the separation efciency of the process.

The upward velocity of the ocs is much higher

than the sedimentation rate of these ocs. There-

fore the surface load of otation is also much

higher, than the surface load for the sedimentation

process.

Flotation is often selected as the oc-removal

process for conditions which are less favorable

for sedimentation:

- low temperature of the water, resulting in a

reduced sedimentation rate due to increased

viscosity

- high algae content in the water, resulting in a

reduced sedimentation rate (algae might even

oat due to the release of oxygen during the

night).The air dosing level is an operation parameter

which can be brought into compliance with varying

operational conditions (temperature, viscosity, oc

size , oc density).

2 Principle

The principle of otation is shown in Figure 1.

The occulated raw water is distributed at the bot-

tom of the otation tank and ows in an upward

direction over the bafe.

At the same time, a water ow with supersaturatedair is supplied through nozzles. This water ow is

called the saturation or recirculation ow. Due to

the pressure drop in the supersaturated water at

the nozzles, small air bubbles form.

The rising velocity of the air bubbles is greater than

the water velocity, so the air bubbles collide with

the ocs. Air pockets form beneath the ocs and

the density of the aggregates decreases below the

water density. As a consequence, the aggregates

will oat on the water’s surface.

Filtration zone

Schematically, the process in the zone before the

above-mentioned bafe can be considered as a

ltration process.

In Figure 2 a oc and an air bubble are represented

in subsequent periods of time. It can be seen that

the air bubbles rise faster than the ocs. Assum-

ing the air bubbles are xed (point of reference),

the water moves with the ocs downwards and

the ocs are ltered from the water by the airbubbles.

Separation zone

Figure 1 - Principles of oc formation and otation

filtration zone

emical dosingd mixing

flocculation flotation

sludgedisposal

outlet

separation zone

float layer

baffle weirwater

air

saturation unit

air dosing

6-8% recirculation

Figure 2 - Filtration principle during otation

= floc = air

t= t 2t 3t 4t t= t 2t 3t 4t

69

WATER TREATMENT FLOTATION



Removal of the oating ocs takes place in the

separation zone.

The ocs and the air bubbles form a oat layertogether at the water’s surface (Figure 3).

The oat layer is transported by the water ow to

a weir and is drained. To accelerate the removal

process of the oat layer, rotors can be placed at

the sludge overow.

The treated water ows under the sludge removal

device and over a weir to the outlet of the otation

system (Figure 1).

In some installations the weir of the oat layer

removal device is exible. When the weir is high,

the oat layer is thin and there is a risk that water

will ow into the sludge removal system. When

the oat layer is thick, the risk is that air bubbles

will escape from the oat layer and the ocs will

start to settle.

The height of the weirs (water and sludge) is of

primary importance to the performance of the

otation system.

Saturation unitThe saturation water is made by the saturation

unit (Figure 4).

The saturation unit is supplied by a water ow and

an air ow.

The water ow is about 6% - 8% of the total water

ow through the treatment plant and is abstracted

downstream of the otation process. The water is

pumped into a pressure vessel, at a pressure of

4 to 8 bar.

Air is supplied to the pressure vessel via an air

compressor. Because of the high pressure, more

air can be dissolved than is possible under atmos-

pheric circumstances.

The supersaturated water ow is transported from

the saturation unit to the ltration zone and

Figure 4 - Saturation unit

Figure 3 - Float layer in separation zone

Figure 5 - Size and distance between bubbles and

ocs

filtration zone

vs

vo

separation zone

overflow

bubbles

40 mm161 mm

200 mmHOH

70




inserted through the distribution nozzles.

Specic parameters

To give an idea of the order of magnitude of the airbubbles and ocs, some values that are encoun-

tered in practice are represented in Table 1.

Depending on the type of nozzle, the median

diameter of an air bubble lies between 30 and

40 µm.

In Figure 5 some specic parameters determined

at an air dose of 8 l/m3 are represented.

From Figure 5 and Table 1 it can be concluded that

collision between air bubble and oc is inevitable,

because the distance between two air bubbles is

smaller than the diameter of a oc.

3 Theory

3.1 Saturation unit for the supply of air

The amount of air that can be dissolved in a certain

volume of water depends on the pressure and the

water temperature and can be calculated with

Henry’s Law:

s H a H

MW pc k c k

R T

×= × = ×

×

in which:

cs = saturation concentration of gas in water

(g/m3)

kH = distribution coefcient (-)

ca = concentration of gas in air (g/m3)

MW = molecular weight of gas (g/mol)

p = partial pressure of gas in air (Pa)

R = universal gas constant =8.3142 (J/(oK.mol))

T = (air) temperature (K)

When the concentration of gas in water is calcu-

lated with Henry’s Law, the total air pressure can

be used instead of the partial pressure. Then, the

concentration ca represents the specic density of

air at the prevailing temperature and pressure.

In Table 2 the distribution coefcient for different

air temperatures is represented.

By increasing the pressure, the amount of gas that

can be dissolved in a volume of water increases

proportionally, as is shown in Figure 6, where the

saturation concentration is given as a function of

pressure (atmospheric pressure at sea level is

101,325 kPa).

The gas exchange between water and air is more

extensively explained in the module on aeration

and gas stripping.

Air bubbles Flocs

parameter value unit parameter value unit

diameter 10 - 100 µm diameter 100 - 200 µm

density (1.5 - 3.0) .1011 bubbles/m3 oc density (2.5 - 19) .107 ocs/m3

distance between

bubbles150 - 188 µm

distance between

ocs3600 - 3700 µm

air dosage 5 - 10 l/m3 density 1003 - 1006 kg/m3

particle

concentration10 - 25 g/m3

Table 1 - Specic parameters of air bubbles and ocs

300

0

50

100

150

200

250

0 200 400 600 800

s o l u b i l i t y ( m g /

l )

pressure (kPa)

1 mg/l = 0.78 l/m3

air with extra nitrogen, 20 oC

air with extra nitrogen, 0 oC

air, 20 oC

air, 0 oC

Figure 6 - Solubility of air in water

71




3.2 Efciency of the bubble lter in the

ltration zone

The efciency of a bubble lter in the ltration zone

is equal to the proportion of ocs that collide with

the bubble lter.

This portion of ocs can be derived from the mass

balance, assuming a permanent attachment with

one or more air bubbles.

Kinetics equationIn Figure 7 a plug ow is represented together

with the principle of collision between ocs and

air bubbles (see also Figure 2).

From the mass balance for a unit element dH (Fig-

ure 7), the kinetic equation for the collision of ocs

and air bubbles in a plug ow can be derived.

The mass balance for a unit element dH is:

dd d

dNQ N dt Q N dH dt

dH

⋅ = ⋅ + +

db T d b bN N dV A dH⋅ η ⋅ ⋅ ⋅ ⋅

in which:

Nd = oc density (ocs/m3)

αdb

= collision coefcient between air bubble and

oc (-)

ηT = collision frequency between air bubble and

oc (-)

Nb = air bubble density (air bubbles/m3)

dV = volume of unit element dH (m3)

Ab = projected area of an air bubble

(m2

/air bubble)

The second part of the right half of the equation

represents the number of ocs that collide and at-

tach to an air bubble in the unit element dH.

The number of ocs depends on the collision

frequency, the collision efciency, the number of

ocs and bubbles in the unit element and the size

of the projected collision area of the bubble.

The collision efciency is determined by pre-treat-

ment of the water and is negatively inuenced

by turbulence in the ltration zone. The collision

frequency is elaborated on in section 3.3.

Rearranging the mass balance leads to:

ddb T d b b

dNN N A

dH= −α ⋅ η ⋅ ⋅ ⋅

with

dbdH v dt= ⋅

in which:

vdb

= approaching velocity between air bubble and

oc (m/s)

The kinetics equation for collision between air

bubbles and ocs becomes:

ddb T d b b db

dNN N A v

dt= −α ⋅ η ⋅ ⋅ ⋅ ⋅

Because the air bubbles rise much faster than the

Table 2 - k H -values at different temperatures and the molecular weight of air

Concentration of

gas in air T = 0oC T = 10oC T = 20oC T = 30oC MW [g/mol]

79% N2, 21% O

20.0288 0.0234 0.0200 0.0179 28.84

Nd + dHdNd

dH

Q

Q

dH

Figure 7 - Mass balance during ltration of the ocs

72




ocs, it is assumed that the approaching velocity

is equal to the bubbles’ rising velocity.

The rising velocity of an air bubble under laminar

ow conditions can be calculated with Stokes’

Law:

2b

b

g d1v

18

⋅= ⋅

ν

in which:

vb = rising velocity of air bubbles (m/s)

db = diameter of air bubbles (m)

ν = kinematic viscosity (m2/s)

The air bubble density is equal to the air dosagedivided by the volume of the air bubble:

bb

3b

N1

d6

ϕ=

⋅ π ⋅

in which:

ϕb = air dosage (m3 air/m3 water)

and Ab is:

2

b b

1 A d

4= ⋅ π ⋅

Substituting vdb

with vb and inserting N

b and A

b

gives the following kinetics equation:

d db T b bd

dN d g1N

dt 12

v0

α ⋅ η ⋅ ⋅ ϕ ⋅= − ⋅ ⋅

ν

This equation is a first-order reaction and isequivalent to dc/dt = -kc in ltration and activated

carbon ltration.

Efciency

Under the assumption that the ltration zone can

be schematized by a plug ow, the equation men-

tioned above can be integrated with the following

boundary conditions:

- at t = 0: Nd = N

d,i

- at t = t : Nd = N

d,e

in which:

Nd,i = oc density of the inuent (ocs/m3)

Nd,e = oc density of the efuent (ocs/m3)

After integration, the efciency of ltration can be

expressed as:

db T b b

0

d g1

12d,i d,e v

N NR 1 e

ν

æ ö÷ç ÷ç ÷ç a ×h × ×j × ×t÷ç ÷ç- × ÷ç ÷÷ç ÷ç ÷ç ÷÷çè ø-= = -

in which:

R = efciency of a bubble lter (-)

τ = residence time in the ltration zone (s)

3.3 Collision probability between bub-

bles and ocs in the ltration zone

In the equation of the efciency of ltration, the

collision probability between air bubbles and ocs

is incorporated.

The collision probability is the ratio between the ac-

tual number of collisions and the possible number

of collisions between air bubbles and ocs.

It is assumed that the oc is situated in a water

column above the air bubble with a surface area

(perpendicular to the rising direction) equal to the

projected area of the air bubble.

The four different removal mechanisms of ltration

can be used to describe the collision between air

bubble and oc:

- diffusion ηD, collision caused by Brownian mo-

tion of mainly small ocs and particles

- interception ηI, collision because the oc trajec-

tory approaches the air bubble and interception

of the oc by the air bubble is possible- sedimentation η

S, collision caused by large

and heavy ocs that deviate from the original

trajectory

- inertia ηTA

, collision caused by mainly large and

heavy ocs that deviate from the curvature of

the original trajectory.

For these transport mechanisms (Figure 8), the

collision probability η or the Single-Collector Col-

lision Efciency (SCCE) can be quantied:

73




222

33b

D

w d b

k T 1 16.18

g d d

⋅η = ⋅ ⋅ ⋅

⋅ ρ

2

dI

b

d3

2 d

η = ⋅

2

d w dS

w b

d

d

ρ − ρη = ⋅

ρ

2d b d

TA

w

g d d

324

⋅ ρ ⋅ ⋅η =

⋅ ν ⋅ ρ

in which:

kb = Boltzmann constant = 1.38.10-23 (J/oK)

T = absolute water temperature (oK)

dd = diameter of the ocs (m)

ρw = density of water (kg/m3)

ρd = density of ocs (kg/m3)

The total collision probability ηT is equal to the

sum of the separate collision probabilities of each

transport mechanism:

T D I S Tη = η + η + η + η

In Figure 9 the collision probabilities are repre-

sented as a function of the oc diameter, and the

following can therefore be concluded:

- the diffusion mechanism is predominant for

ocs smaller than 1 µm- the interception mechanism is predominant for

ocs larger than 5 µm.

From the equation of collision probability for in-

terception and Figure 9, it can be concluded that

with an air bubble of 40 µm diameter, the collision

probability ηT =1, if the oc diameter is larger than

32 µm. After this manner of oc formation, the oc

size is between 100 and 1,000 µm.

Consequently, in practice, the interception mecha-

nism predominates and the collision probability is

equal to 1.

The progress of the total collision probability in

Figure 9 is equivalent to the curves of the collision

probability in the ltration and the oc formation

theories.

Similar to ltration, the collision probability is mini-

mal for particle sizes of about 1 µm.

The main difference between the processes,

however, is that the collision probability in otationprocesses is an order of magnitude larger than in

ltration and, therefore, the ocs will collide more

easily during otation.

3.4 Determination of the surface loading

of the separation zone

Removal of air bubble-oc aggregates takes place

in the separation zone.

All aggregates are removed if the time needed for

the otation of an aggregate at the bottom of the

0.01 0.1 1 10 100 1,000

1

0.1

0.01

0.001

floc size (µm)

c o l l i s i o n

p r o b a

b i l i t y

( - )

flocs: 100-1,000 µmtherefore interception

T = 10 oCdb = 40 µmrd = 1,003 kg/m3

ηT

ηIηs

ηTA ηD

Figure 9 - Collision probability

A

BC

D

transport mechanisms:

A = diffusion

B = interception

C = inertia

D = sedimentation

covered path

flow path

floc

air bubble

Figure 8 - Transport mechanisms

74




separation zone is less than the residence time

in the separation zone (the opposite of discrete

settling):

b

st

t 1t

>

in which:

tb = residence time in the separation zone (s)

tst = time an aggregate needs to reach the water

surface (s)

Assuming a plug ow in the separation zone:

st so 1v v1 m

> × -

in which:

vst

= rising velocity of the air bubble-oc

aggregate (m/s)

vso

= surface loading in the separation zone

(m3/(m2.s))

m = fraction of dead space (eddies) in the sepa-

ration zone (-)

It can be concluded that the maximum surface

loading is determined by the rising velocity of the

aggregates in the separation zone.

Rising velocity of the air bubble-oc aggre-

gates

Assuming a laminar ow and spherical aggregates,

the rising velocity can be calculated with Stokes’

Law:

2w ast a

w

1 gv d18

ρ − ρ= ⋅ ⋅ ⋅ ν ρ

in which:

ρa = density of the aggregate (kg/m3)

da = diameter of the aggregate (m)

The density and the diameter of the aggregate can

be determined with the following equation:

a b d

1

1 1

βρ = ⋅ ρ + ⋅ ρ β + β +

3a dd 1 d= + β ⋅

in which:

β = volume ratio between air bubbles and ocs in

the aggregate (-)

In Figure 10 the inuence of the volume ratio on

the diameter, density and rising velocity of the

aggregate is represented.

In calculating the diameter of the aggregate da, the

volume of the air bubbles is assumed to be divided

over the entire surface of the oc. It is practically

physically impossible for more than one bubble

layer to exist. For a oc diameter of 200 µm, this

results in a maximum volume ratio of 1.

The volume ratio between the total volumes of

inserted air and ocs is about 500. Thus, only 0.2%

of the total inserted air is effectively used during

the oc removal process.

Figure 11 - Rising velocity for different β

dd = 100 µm

dd = 200 µmdd = 500 µm

volume ratio of air bubble/floc (-)

r i s i n g v e l o c i t y ( m / h )

0 0.5 1 1.5 2 2.50

100

200

300

400

500

600

Figure 10 - Air bubble-oc aggregate parameters

diameter of air bubblesdiameter of floc = 200 mm

= 40 mm

b

ra

vst

0.008

1.00

995.1

0.43

0.08

1.03

928.8

6

0.8

1.21

557.8

44.7

8.0

2.08

112.5

152 m/s

kg/m3

dd

da

75




The applied (high) air dosage, however, is neces-

sary for a sufciently high efciency of the bubblelter in the ltration zone.

In Figure 11 the rising velocity is represented as a

function of volume ratio for ocs with a diameter

of 100 µm, 200 µm and 500 µm.

Because the volume ratio between air bubbles

and ocs is (physically) restricted (βmax » 1), an

increase in the rising velocity vst can only be real-

ized by an increase in oc size.

4 Practice

4.1 Process parameters

Air dosage

The relationship between air dosage and efciency

of the bubble lter is exponential, as seen in Fig-

ure 12.

In addition it can be concluded that at lower water

temperatures, higher air dosages are required toobtain the same efciency.

The air dosage is an operation parameter for the

efciency of the bubble lter.

In Figure 12 the total removal efciency, measured

in practice, is represented as a function of the air

dosage. The total efciency consists of the ef-

ciency in the ltration zone and the efciency in

the separation zone.

The total efciency approaches the value 0.9

– 0.95 and not 1.

This can be due to the occurrence of short-circuit

ows in the ltration zone, resulting in a decrease

in the efciency in the ltration zone or to the lim-

ited inuence of the air dosage on the efciency

in the separation zone (less than 1% of the air is

effectively used for separation), and separation in

practice is not optimal.

Figure 13 - Inuence of contact time in theory and practice (total efciency)

1.2

1

0.8

0.6

0.4

0.2

00 60 120 180 240 300

theory

e f f i c i e n c y o f t h e b u b b l e f i l t e r ( - )

contact time (s)

T = 2oC T = 10oC T = 20oC

jb = 5 l/m3

= 0.5

db

dd

rdadb

= 40 mm= 100 mm

= 1,003 kg/m3

1.2

1

0.8

0.6

0.4

0.2

00 60 120 180 240 300

practice

t o t a l r e m o v a l e f f i c i e n c y ( - )

contact time (s)

T = 3oC T = 10oC T = 16oC

Figure 12 - Inuence of air dosage in theory and practice (total efciency)

1.1

1

0.9

0.8

0.7

0.6

0.51 3 5 7 9 11

theory practice

e f f i c i e n c y o f t h e b u b

b l e f i l t e r ( - )

air dosage (l/m3)

1.1

1

0.9

0.8

0.7

0.6

0.51 3 5 7 9 11

t o t a l r e m o v a l e

f f i c i e n c y ( - )

air dosage (l/m3)

T = 2oC T = 10oC T = 20oC without sludge layer with sludge layer

Tdb

dd

rdαdb

= 1.6 min= 40 µm= 100 µm= 1,003 kg/m3

= 0.5

76




Contact time

Theoretically, the contact time should be longerthan 90 seconds, as seen in Figure 13.

In practice, contact times of 54 to 126 seconds

are applied.

In this gure, the results of measurements on full-

scale plants, where the contact times are varied,

are represented.

In general, the theory is conrmed, but an ef-

ciency of 1 is not reached, even with innite

contact times.

The reason is that in cases where the contact times

are increased, the water ow must be decreased

(for the same otation tank). At low water ows,

plug ow no longer occurs and the efciency will

be lower than 1.

Temperature

The water temperature determines the viscosity

of the water, inuencing both ltration and sepa-

ration.

From calculations it can be derived that the airdosage at 5 ºC must be a factor 1.6 to 1.7 higher

than at 20 ºC to obtain the same efciency (Figure

12).

It can also be derived that the rising velocity of

an aggregate in the separation zone is 1.6 times

lower at 2 ºC than at 20 ºC.

Therefore, the efciency in the separation zone

will be lower at lower temperatures.

In Figure 14 the progress of the residual iron con-

centration in a otation system is represented. The

temperature effect is obvious.

4.2 Saturation

Saturation unit

In practice, two types of saturation units are ap-

plied.

The saturation units can be of a packed column

type, similar to the tower aerator systems, or

of a venturi type, similar to the venturi aeration

systems.

The saturation units that make use of a packed

column, similar to the tower aerator system, or the

venturi aeration units.

Both systems are schematically represented in

Figure 15.

For the design and functioning of the saturation

units, reference is made to the module on aeration

and gas stripping.

Nozzles

In Figure 16 the principle of a nozzle that is applied

time (month)

e f f l u e n t i r o n c o n t e n t ( m g / l )

0

0.1

0.2

0.30.4

0.5

0.6

0.7

0.8

jan feb mar apr may jun jul aug sep oct nov dec

Figure 14 - Seasonal inuence by temperature varia-

tion Figure 15 - Saturation units

packed column venturi

Figure 16 - Nozzle

77




in the ltration zone for the distribution of satura-

tion water is shown.

In practice Bete, AKA and WRC nozzles are ap-

plied.

In Figure 17 the median bubble diameter for the

different nozzles is given as a function of saturation

pressure of the recirculation ow.

Based on the required median bubble diameter, a

nozzle and a saturation pressure can be chosen.

The median bubble diameter is, for most of the

nozzles, between 30 µm and 40 µm.

4.3 Flotation tank

In Figure 18 a otation tank with a ltration or

contact zone, separation zone and bafe is rep-

resented. These will be discussed in the following

paragraphs.

Filtration zone

The ltration zone (Vc in Figure 18) is designed

based on a contact time longer than 90 seconds.To obtain a plug ow in the ltration zone, the

length/width ratio must be higher than 5 (long

and narrow).

A column reactor would thus be suitable for the

ltration zone.

In existing otation tanks, the saturation water is

released into the ltration zone, resulting in local

(near the nozzles) velocity gradients of 20-30 m/s,

turbulence and break-up of ocs and bubble-oc

aggregates.

Therefore, strong (and thus small) ocs must be

formed during oc formation to resist the turbulent

ows.

Moreover, this turbulence can be minimized by the

arrangement of the nozzles. An example of this is

the application of a separate nozzle zone outside

the main water ow.

Separation zone

The separation zone (Va in Figure 18) is designed

based on surface loading, which is determined

from the rising velocity of the bubble-oc aggre-

gates and takes the dead zones Vl into account.

In practice, surface loadings of 10 to 25 m3/(m2.

h) are applied.

The residence times in the separation zone vary

between 5 and 10 minutes.

The height H must be about 2 meters to avoid

Figure 17 - Performance of three different nozzles

Bete

AKA WRC

3 4 5 6 7 8

50

40

30

20

m e d i a n b u

b b l e s i z e

( µ m )

saturation pressure (bar)

Figure 18 - Design of otation tank

L

Vc VaH

B

Q

VI

baffle

overflow

Figure 19 - Division of ltration and separation zone

78




large ow gradients and the zone must be long

and narrow to approach a plug ow.Finally, the water ow must be uniformly distributed

and collected over the width of the separation zone

to limit the fraction of dead zone.

Division between ltration zone and separation

zone

The division between the ltration zone and sepa-

ration zone is achieved with a bafe or an overow

(Figure 19).

Comparing the overow structure with the bafe,

no difference is observed in efuent quality and

residence time. In both cases the residence time

is about 70 % of the gross residence time (Figure

18):

bruto

br c a l

Q Q Q Q Qt

V L B H V V V= = = + +

⋅ ⋅

The ow conditions with an overow are favorable

and less sludge is deposited. It can therefore be

concluded that an overow is preferred abovethe bafe.

Different types of overow structures are repre-

sented in Figure 20.

Figure 20 - Design of otation tank with different over -

ow structures

A

B

C

1.8 m

1.8 m

1.8 m

3.6 m

7.1 m

7.1 m

0.5 m

T = 0.1oC

T = 15oC

flotation time (min)

e f f l u e n t

i r o n c o n t e n t ( m g / l )

00

0.2

0.4

0.6

0.8

1

1.2

1.4

10 20 30 40 50

CB A

Advanced literature

• Flotatie: Theorie en praktijk (Dutch), G.J.

Schers MSc thesis TU-Delft (1991)

• Summary: H2O (in Dutch), 5th Gothenburg

Symposium (1992)

Further reading

J. Haarhof •

79




80




Granular

ltration

WA T E R T R E

A T M E N T

WATER TREATMENT



Framework

This module explains ltration.

Contents


1. Introduction

2 Principle

2.1 Filtration mechanisms

2.2 Filtration, column tests

3 Theory

3.1 Filtration

3.2 Backwashing

4 Practice

4.1 Backwashing 4.2 Filter bottom

4.3 Filter material

4.4 Filter troubles

5 Alternative applications for ltration

5.1 Multiple layer ltration

5.2 Pressure ltration

5.3 Upow ltration

5.4 Limestone ltration

5.5 Continuous ltration

5.6 Dry ltration

5.7 Slow sand ltration

82

FILTRATION WATER TREATMENT



1 Introduction

In general, ltration is a process where water ows

through a permeable layer, either a membrane,

lter paper, a sieve, a porous medium or such.

In water treatment, granular ltration is a process

where water ows through granular material (often

sand) while suspended solids (sand, clay, iron

and aluminum ocs) are retained, substances

are biochemically decomposed and pathogenic

microorganisms (bacteria, viruses, protozoa) are

removed.

The suspended solids slowly ll the pores, result-

ing in an increase in hydraulic resistance.The suspended solids are removed by periodically

cleaning the lter beds. This prevents the resis-

tance from becoming too high or the break through

of suspended solids.

Filters are also used for chemical and biological

reactions. This is mainly of importance for the

treatment of groundwater where the oxidation of

iron, manganese, ammonium and, in case of poor

gas stripping, methane takes place.

The removal of pathogenic microorganisms is of

importance for surface water treatment, and the

efciency is approximately 90 to 99%. The removal

of pathogenic microorganisms occurs by decay

and retention on the (sand) grains.

The most common application of ltration is rapid

sand ltration (Figure 1).

Rapid sand ltration consists of a bed with a coarse

granular medium (0.8-1.2 mm) and supernatant

water. The ltration velocities (between 5 and 20

m/h) are controlled by varying the supernatantwater level (inlet-controlled) or by operating a valve

at the outlet pipe (outlet-controlled).

Due to clogging, maximum resistance is reached,

and the lter bed must be cleaned by backwashing.

During backwashing the lter bed is expanded, and

the accumulated suspended solids are removed.

The backwash water is drained through a central

trough to a waste receptacle. The backwash fre-

quency is repeated every few days.

Rapid sand lters are present in nearly every water

treatment plant.

Surface water treatment uses the lters after oc

formation and removal to get rid of the remaining

ocs and pathogens and to decompose ammo-

nium.In groundwater treatment, the lters are usually

placed after aeration to remove iron ocs, man-

ganese and ammonium. With softening, lters

are often placed after pellet reactors to remove

the ‘carry-over’.

2 Principle

2.1 Filtration mechanisms

When water ows through the lter bed, sus-

pended and colloidal particles are retained by the

lter material.

Particles that are larger than the pores in the lter

bed will remain on the bed (Figure 2). With rapid

ltration this does not occur often, because the

larger particles (iron or aluminum ocs) are already

removed in the preceding oc removal process

(sedimentation or otation).

If smaller lter material is used, the pores are also

smaller and the screening process results in theso-called cake ltration. The cake will also retain

5 m

v = 10 m/h

10 m

1,0 m

0,3 m

1,5 m

0,5 m backwash gutter

Figure 1 - Principle of rapid sand ltration(side and front view respectively)

83

FILTRATIONWATER TREATMENT



small particles, and treatment occurs mainly in the

top layer of the lter.

The disadvantage of cake ltration is that with

high concentrations of suspended and colloidal

particles rapid clogging of the lter bed occurs.

During rapid ltration the removal of suspended

and colloidal particles usually occurs inside the

lter bed.

The clogging will thus be spread over the entire

height of the lter bed.

The suspended and colloidal particles are trans-

ported to the lter material in different ways (Figure

3).

Generally, the particle follows the trajectory of the

water that ows through the lter bed. This trajec-

tory follows the complicated pore structure of the

bed. When the trajectory curves, a heavy particle

can be transported to the lter material due to in-ertia. If the trajectory approaches the lter grains,

then particles can also be intercepted. Heavy

particles are especially subject to sedimentation,

lighter particles to diffusion. Due to these mecha-

nisms, the particle can switch to other trajectories

that ow nearer to a grain or can collide directly

with a grain, and it remains at the surface or on

the grain.

When the suspended and colloidal particles col-

lide with the lter grains, attachment could take

place.

There are two types of forces that result in attrac-

tion and repulsion of the particles.

The VanderWaals forces ensure that two bodies

are attracted.

Electrostatic forces can have an attracting or

repulsing effect, depending on the charge of theparticles.

In general the lter material (sand) and the sus-

pended and colloidal particles have a negative

charge and repulsion takes place.

Attachment of the particles depends on the mag-

nitude of both opposing forces. If the particles

are destabilized by the addition of trivalent iron

or aluminum salts, attachment will be easier than

without destabilization.

In addition to physical processes to remove sus-

pended and colloidal solids as described above,

chemical and biological processes occur in the

lter bed.

From groundwater, iron(II) and manganese must

be removed by oxidation. By adding oxygen in the

preceding aeration step, iron(II) will be transformed

into iron(III) and iron ocs will be formed. The iron

ocs are removed by the same mechanisms as

described for the removal of suspended and col-loidal particles.

Manganese is transformed to manganese oxide in

clay particle 20 μm

grain diameter = 400 μm

pore diameter 62 μm

Asterionella 30 μm

Bacillus 2 μm

Al or Fe floc 10 μm

Figure 2 - Principle of screening

sedimentation interception diffusion inertia turbulence

Figure 3 - Transport of impurities towards the grain

84




the presence of previously deposited manganese

oxide (catalytic process).

Consequently, it can take several months before

manganese removal in the lter bed is initialized.

Therefore, measures have to be taken to avoid the

total removal of manganese oxide during back-

washing to keep the oxidation process alive.

There are also indications that manganese is

oxidized in the presence of bacteria, suggestinga biological process.

Other biological processes in the lter bed are

the decomposition of methane, ammonium, and

(biodegradable) organic matter.

The decomposition of methane in the lter bed has

to be avoided, because it results in an uninhibited

growth of bacteria, which can lead to clogging

and breakthrough. Methane, therefore, must be

removed early in the process.

Ammonium is transformed into nitrate in two

steps.

A group of nitriers (e.g., nitrosomonas) take care

of the transformation of ammonium into nitrite;

another group of nitrifiers (e.g., nitrobacteria)

transform nitrite into nitrate.

The nitriers are situated on the surface of the

lter material and for their growth they use energy

that is produced during the transformation of am-

monium or nitrite. The amount of ammonium that

can be transformed depends on the growth rateof the bacteria, the size of the bacteria population,

and the amount of ammonium that is transported

to the bacteria (by diffusion).

In the beginning the growth rate of the bacteria

is optimal and uninhibited. The population dur -

ing this lag phase is small and little ammonium

is transformed. After some time the population

will grow and nally stabilize (growth is equal to

decay). At that time the maximum ammonium

removal occurs.

For iron and manganese removal, the oxygen

consumption is limited.

Iron is normally present in concentrations lower

than 10 mg/l and manganese concentrations are

seldom above 1 mg/l. The oxygen concentrations

that are needed for these reactions are, after aera-

tion, dissolved in water (approximately 10 mg/l).

For biological processes the oxygen consumption

is considerably higher. When low concentrationsof ammonium and/or methane are present in the

water (few mg/l), the amount of oxygen that can be

dissolved in water under atmospheric conditions is

insufcient to complete these reactions. A single

ltration step is no longer sufcient. Furthermore,

high bacteria concentrations in the lter can in-

crease the risk for the growth of Aeromonas.

2.2 Filtration, column tests

It is important to learn from the experiences of

other treatment plants that have had to deal with

Chemical and biological decompositions in the lter bed

4 Fe2+ + O2 + 2 H20 + 8 HCO3- → 4 Fe(OH)3 + 8 CO2

2 Mn2+ + O2 + 6 H2O → 2 MnO2 + 4 H3O+

NH4+ + 2 O2 + 2 H3O+ → NO3- + 3 H2O

L

H

LHdgrain

v

= 1 m= 1.5 m= 1.0 - 1.6 mm= 5 m/h

V

A

B

C

A: supernatant waterB: filter bedC: filter bottom

D

Figure 4 - Filtration, experimental setup

85




similar water to determine the dimensions of a

lter.

Further information can be obtained with a test

lter (Figure 4).

In the test setup, the optimal combination of the

following design parameters should be found:

- grain diameter of lter material

- ltration velocity

- height of the lter bed- height of the supernatant water.

The optimal combination leads to a lter that is

cheap, always satises the required efuent quality

and has a reasonable lter run time. In addition,

during the lter run, the suspended solids should

be divided over the lter bed height to avoid cake

ltration.

The lter surface area has to be as small as pos-

sible to reduce investment costs.

Consequently, the ltration velocity has to be

high. The higher the ltration velocity, the sooner

the efuent quality will deteriorate during the lter

run. This can be compensated for by increasing

the lter bed height or by choosing lter material

with a smaller grain size.

A higher filter bed, however, means a higher

lter construction and, thus, higher construction

costs.Filter material with a smaller grain size will clog

faster, shorter lter runs will occur and operational

costs will be increased.

In the graphs of Figure 5 the efuent quality and the

lter resistance are represented for different lter

materials and for different lter bed heights.

In general it is assumed that the efuent quality has

to remain under a determined efuent guideline.

The lter run time during which the efuent quality

satises the guideline is called Tq.

s u s p e n d e d s o l i d s c o n t e n t c

( g / m 3 )

0

0.5

1

1.5

2

Tq (hours)

cO

= 15 g/m3

L = 0.8m

0 24 48 72 96

0

0.5

1

1.5

2

0 24 48 72 96

0

0.5

1

1.5

2

0 24 48 72 96

0

0.5

1

1.5

2

0 24 48 72 96

s u s p e n d e d s o l i d s c o n t e n t c ( g / m

3 )

Tq (hours) Tq (hours)

Tq (hours)

f i l t e r r e s i s t a n c

e H

( m )

f i l t e r r e s i s t a n c e H (

m )

v = 2x10-3 m/s

T = 10 OCd = 0.7 mm

d = 0.8 mm

d = 0.9 mm

d = 1.0 mm

d = 1.2 mm

d = 1.5 mm

Figure 5 - Results of different lter runs to obtain an optimally functioning lter

86




Normally, lters are backwashed after run time Tr

, when a predetermined maximum resistance is

reached.

To prevent the water quality from deteriorating

before the maximum resistance is reached, the

lter design should fulll the condition: Tr <Tq.

The aforementioned design parameters determine

the values of both Tq and Tr (Figure 5).

In practice some restrictions are given to this op-

timization process.

Normally, safety margins are introduced to main-

tain the quality of drinking water above all suspi-

cion and lter run times of 1 to 2 days are used.

In addition, a lter plant is always designed to take

future developments into account. Consequently,most of the time a lter is operated below its ca-

pacity and far from the optimal situation.

3 Theory

3.1 Filtration

Without making use of the ltration theory, a long

series of ltration experiments would be necessary

to come to an optimal solution for an installation,

in practice.

This is problematic, because the raw water quality

varies during the year and experiments would take

at least a year to complete.

The ltration theory makes it possible to quanti-

tatively predict the effects of changes in design

parameters, based on the results of a reduced

number of ltration experiments.

Efuent quality

During the ltration process suspended and colloi-

dal solids accumulate on the grains. Consequently,

the concentration of suspended and colloidal solids

decreases with the increasing lter bed depth.

In addition, the pore volume will be reduced in time

due to the accumulation of suspended and colloi-

dal solids, and the grain size of the lter material

will be increased.

With a constant ltration rate (supercial velocity),

the pore velocity will increase as lter clogging

proceeds.

The equation that is formulated for ltration:

∂ ∂= − ⋅ − λ ⋅ ⋅

∂ ∂c c

u u ct y

together with the mass balance:

∂σ ∂= − ⋅

∂ ∂c

vt y

in which:

c = concentration of suspended and colloidal

solids (g/m3)

y = depth of the lter bed (m)

v = ltration rate (m/s)

p = porosity (%)

u = pore velocity (=v/p) (m/s)

λ = ltration coefcient (m-1)

σ = accumulated solids (g/m3)

In the stationary situation the following is valid:

∂=

∂c

0t

therefore the kinetics equation is transformed

into:

∂= −λ ⋅

∂c

cy

To solve the system of equations the value of theFigure 6 - Reduction of pore volume as a result of

accumulated solids

87




ltration coefcient λ must be known.

However, λ depends on different factors, such as

the ltration velocity, viscosity, grain size, quality of

the raw water, and the clogging of the bed.

After start-up of the ltration process, the ltration

coefcient will initially increase because of bet-

ter attachment characteristics on the preloaded

material.

Due to pore clogging, the pore velocity increases

and fewer solids will accumulate, expressed by a

lower ltration coefcient λ.

When the solids are retained in the top layer of the

lter bed, lower layers will take over until the lter

is saturated and the lter breaks through.

The clean bed ltration coefcient λ0 and the rela-

tionship between λ and σ have to be determined

in practice (through column experiments).

Several researchers have found empirical relation-

ships. Well-known relationships are those of Lerk

and Maroudas.

Lerk:

λ =

⋅ ν ⋅

10 3

k

v d

Maroudas:

σλ = λ ⋅ − ⋅

ρ ⋅ 0 2

d 0

1 kp

in which:

d = grain size (m)

p0 = initial porosity (%)

k1, k2 = constantsν = kinematic viscosity (m2/s)

The ratio between the accumulated solids σ and

the density is the reduction in pore volume (σv)

σ= σ

ρ v

d

in which:

ρd = density of the ocs (kg/m3)

σv = volume concentration in pores (m3/m3)

The value of the constant k1 is often assumed to be

9 ·10-18 and the constant k2 is the reciprocal value

of the maximum pore lling n (0<n<1).

Notice that in the case of Madouras it is assumed

that the ltration coefcient decreases linearly as

clogging increases. Although this is a simplica-

tion, with this assumption the system of equations

can be solved.

With the boundary conditions y = 0, c = c0 and the

initial condition t = 0, σv = 0 and:

⋅ ⋅ λα =

⋅ ρ ⋅0 0

d 0

v c

n p

The solution becomes:

- general solution:

α⋅

λ ⋅ α⋅= ⋅

+ −0

t

0 t t

ec c

e e 1

- efuent quality (y=L):

α⋅

λ ⋅ α⋅

= ⋅+ −0

t

e 0 t t

ec c

e e 1

and:

α⋅

λ ⋅ α⋅

−σ = ⋅ ⋅

+ −0

t

v 0 y t

e 1n p

e e 1

Filter resistance

During ltration, pore clogging increases and,

therefore, so does resistance in the lter bed.When the lter reaches the maximum available

, the lter needs to be backwashed to avoid a

decrease in the ltration velocity. The maximum

available head loss is the difference between

the supernatant water level and the head of the

outowing water, minus the clean bed resistance

and head loss caused by lter bottoms, pipes and

valves (Figure 7).

The clean bed resistance (H0) can be derived

from the equation of a ow through a pipe (pore)

88




and can be described with the Carman-Kozeney

equation:

( )−ν= = ⋅ ⋅ ⋅

2

000 3 2

0 0

1 pH vI 180

L g p d

in which:

I0 = initial resistance gradient (-)

This equation (the linear relationship between

velocity and resistance) is only valid when:

⋅= ⋅ <

ν0

0

v d1Re 5

p

When clogging occurs, the resistance formula

changes to:

= ⋅

− σ

2

00

0 v

pI I

p

in which:

I = resistance gradient (-)

The solids accumulation in the pores σv is known

along the height and thus the resistance gradi-

ent can be calculated over the height of the lter

bed.

By integrating the gradient, the total resistance

over the lter bed can be calculated.

As presented in Figure 7 the largest resistance is

built up in the upper layers of the lter bed, where

most of the solids accumulate.

In the lower layers the resistance gradient is almost

equal to the clean bed gradient.

In time the resistance in the upper layers will

increase.

A pressure drop in the lter bed below atmospheric

(negative pressure) must be avoided. In such a

case, dissolved gases will come out of solution

and, then, released gas bubbles will disturb the

lter bed.

Accumulated gas bubbles hinder downward water

movement, increase lter resistance and end lter

runs prematurely.

Negative pressure can be avoided by maintaining

a high supernatant water level and shortening

lter runs. This can be achieved by increasing the

height of the outow weir.

3.2 Backwashing

After a certain operation period the pores in a lter

bed are lled with accumulated suspended solids.

The porosity has decreased from p0 to p, whichresults in a higher resistance and/or a poor efuent

Figure 7 - Progress of the lter bed resistance in time,

the so-called Lindquist diagram

Figure 8 - Filtration, backwashing with water and air

89




quality. Rapid lters are cleaned by backwashing

with clean water (ltrate).

During backwashing, the water ows in an upward

direction through the lter. The water scours the

lter grains, erodes the accumulated solids from

the lter material, expands the lter bed, and trans-

ports the solids towards the backwash troughs.

The shearing (scouring) force of the water is equal

to the mass of grains under water:

( ) ( )πτ ⋅ π ⋅ = ⋅ ⋅ ρ − ρ ⇒ τ = ⋅ ρ − ρ2 3

f w f w

dd d

6 6

in which:

τ = shearing force (kg/m2)

ρf = mass density of the lter material (kg/m3)

ρw = mass density of the water (kg/m3)

The larger the diameter of the grains, the larger

the shearing forces.

From practice, it is known that with grain diameters

smaller than 0.8 mm backwashing is difcult.

Therefore, a combination of water and air is used.

By using air a more turbulent situation is created

which facilitates the removal of the solids from

the pores.

Hydraulics of back washing

Bed expansion is an important parameter for the

design of a backwash facility:

−= e 0

0

L LE

L

in which:

E = bed expansion (-)

L0 = initial height of the lter bed (m)

Le = height of the expanded lter bed (m)

The applied bed expansion depends on the diam-

eter of the lter material.

When the lter material has a diameter of 0.8 mm

an expansion of 15 to 20% is used, while a diam-

eter of 1.2 mm requires an expansion of 10%.

During backwashing, the amount of lter mate-

rial remains constant (with a well-designed lter

no loss of lter material occurs). When the initial

porosity (p0), the height of the lter bed during

ltration, and the height during backwashing are

known, the porosity during expansion can be

calculated:

+− ⋅ = − ⋅ ⇒ =

+0

0 0 e e e

p E(1 p ) L (1 p ) L p

1 E

in which:

pe = porosity of expanded bed

A backwash rate of 40 m/h through a lter bed with

a porosity of 40%, a grain diameter of 1 mm anda temperature of 10 °C gives a Reynolds number

of 14.1. Thus, the water ow during backwashing

is no longer laminar, but situated in the transition

zone between laminar and turbulent, and the Kar -

man-Kozeney equation is not valid.

From experiments, an empirical equation for

the resistance during backwashing has been

derived:

−ν= ⋅ ⋅ ⋅ ⋅

1,80,8 1,2

e e3 1,8

e

(1 p )vH 130 L

g p d

p0

L0

Le

pe

low backwash velocity

no expansion

high backwash velocity

expansion

Figure 9 - A non-expanded and expanded lter bed

90






After the ltrate drainage is blocked (valve B), the

backwash process is started. That is when valves

C and D are opened. The lter is backwashed for

a certain period of time with water and air. When

the bed is sufciently clean, the supply of water

needed for backwashing is stopped and the wash

water drain is closed by valve D.

By opening valve E the supernatant water ltrates

through the bed. Then valve A is opened. Since

the water that leaves the lter during the ripening

period is generally of poor quality, this water is

drained into a waste receptacle.

After the ripening period, valve E is closed and B

is opened.

The total time that a lter is not in operation dueto the backwash procedure varies from 30 to 60

minutes. The backwashing process itself will last

approximately 20 minutes.

When the lter run time and the backwash time

are known, the net production through the lter

can be calculated.

Over a short period of time a high wash water ow

is needed. There are two possibilities to supply

these ows:

- backwash pumps

- elevated water reservoir.

When using backwash pumps, water is obtained

directly from the ltrate reservoirs.

During a short period, high energy consumption

takes place. This is expensive. The pipes that

transport the wash water from the ltrate reser -

voirs to the lters are the largest pipes in a water

treatment plant.

Assuming that the maximum permitted velocity in

a pipe is 1 m/s, it can be calculated that the diam-eter of the backwash supply pipe to a lter with a

surface area of 80 m2 and a backwash rate of 50

m/h is almost 1.2 meters.

The diameter of the supply pipe for raw water is

much smaller. When the ltration rate is 5 m/h, this

diameter is 0.35 meter.

When an elevated water reservoir is used, back-

wash pumps with a lower capacity (10 to 20% of

the backwash capacity) can be applied. These

backwash pumps continuously ll the reservoir.

A disadvantage, however, is that a separate el-

A lter with a surface area of 80 m2, a lter

run time of 72 hours, and a filtration rate of5 m/h is backwashed for 20 minutes with a

backwash rate of 50 m/h. In addition, the lter

is not operating for 20 minutes due to drain-

age of the supernatant water, and also due to

lter the waste.

The ltrate production is:

(72-40/60)*5*80=28.533 m3.

The quantity of ltrate used for backwashing

is 0.333*50*80=1.333 m3, that is a water loss

of 5%.

Figure 12 - Discharge pipes for backwash water are the

largest pipes in a lter installation

H3

H2

H1

Vmax

pmin

pmax

from clear well to waste

pmax

pmin

Vmax

H1

H2

H3

= maximum water level in reservoir

= minimum water level in reservoir

= capacity of reservoir

= headloss of filter bed

= headloss of filter bottom

= pipeline losses

Figure 13 - Elevated water reservoir as backwash sys-

tem

92




evated water reservoir needs to be built, thereby

increasing the investment costs.

The required pressure for backwash water is

around 2-5 mWc. In addition to the static eleva-

tion height, most of the pressure losses are due

to the resistance resulting from the ow of water

through the lter bottom, the lter bed, and the

piping system.

After passing the lter bed, the wash water is

drained through a system of troughs.

The troughs are designed to limit the (horizontal)

distance the water must travel after leaving the

lter bed.

The moment the water leaves the lter bed, the

wash water velocity decreases by a factor of 2.5

and settling in the supernatant water can occur.

The right conguration of backwash water troughs

is found by optimization. The more troughs, the

higher the investment costs, but the lower the

wash water loss is.

In practice, it turns out that troughs on the sidesand troughs at the front side are satisfactory and

cheap.

In large lters, a “water sweep” is applied to re-

duce the volume of backwash water. Raw water is

supplied from the sides onto the lter bed. Thus,

the water ows from the sides to a central trough,resulting in a stable ow without short-circuits and

eddies.

In the past wash water was drained to waste water

ponds, the solids would settle and the supernatant

water was drained to the surface water or sewer -

age system.

Because of stricter regulations concerning dis-

charge to surface waters, soil protection and

groundwater protection measures, backwash

water ponds are not used anymore. These days,

backwash water is transported to backwash water

treatment installations.

In such installations, a coagulant is added to the

backwash water, resulting in oc formation incor -

porating the solids. The ocs are removed after -

wards by tilted-plate separators and rapid sand

lters. After UV-disinfection the treated water can

be recycled into the main treatment process.

An alternative backwash water treatment plant

consists of micro-/ultraltration.

Figure 15 - A central backwash water trough

Figure 14 - A central backwash trough

raw water feed

water sweep water sweep

Figure 16 - Water sweep

Figure 17 - Nozzle

93




4.2 Filter bottom

Water that passes the lter bed is drained through

nozzles to the ltrate reservoir.Nozzles are synthetic tubes that are incorporated

into the construction of the lter bottom.

To avoid loss of lter sand, perforated heads are

placed onto the tubes (Figure 17 and 18).

Frequently, a number of support layers of coarse

lter material are placed between the lter bottom

and the lter bed to enable larger slot sizes in the

nozzle, to avoid clogging the nozzles.

In addition to draining the ltered water, nozzles

also have a function in the supply of backwash

water and air.

The most important aspect is a uniform distribu-

tion of water and air over the lter bed, which can

be achieved by introducing a considerable lter

bottom resistance (0.5-2 m).

4.3 Filter material

Not all granular material can be used as lter

material.The material should have certain characteristics,

such as:

- resistant to abrasion (wear)

- free of impurities

- uniform grain-size distribution.

Typically, river sand is applied as lter material.

River sand has a great variety of grain sizes and

must therefore be sieved before application. The

uniformity of the lter material can be expressed

in the uniformity coefcient, dened as:

= 60

10

dU

d

in which:U = uniformity coefcient (-)

d10 = size of sieves that let pass 10% of the

sand mixture (mm)

d60 = size of sieves that let pass 60% of the

sand mixture (mm)

If the uniformity coefcient equals 1, the material

is uniform. A higher coefcient indicates a larger

variety in the grain sizes (Figure 19).

For rapid ltration the value of the uniformity co-

efcient should be between 1.3 and 1.5 to avoid

stratication of the lter bed during backwashing.

A lower value of the coefcient is possible, but this

results in higher sieving costs and provides little

additional advantage.

Other lter materials are given in Table 2.

Filter material with a low density is used when the

diameter of the lter material should be large and

the backwash rate is limited.

The heavier lter materials are used during upowltration to avoid premature expansion of the lter

bed.

Filter material Specic density

Plastic grains 1,050-1,300

Pumice 1,200

Anthracite 1,400-1,600

Sand 2,600

Garnet 3,500-4,300

Magnetite 4,900-5,200

Table 2 - Density of different lter materials

Figure 18 - Nozzle bottom

100

80

60

40

20

0.2 0.5 1.0 5.0

passingpart(%)

sieve opening (mm)

d10 d60

Figure 19 - Sieve curve of lter sand

94




4.4 Filter troubles

In spite of rapid ltration being a simple process,

many problems can occur.

The choice of the lter material and the design of

the lter bottom are crucial.

When the lter material is badly sieved and thus

not uniform, stratication will occur during the

backwash process. Hence, the lighter, smaller

grains will collect at the top of the lter bed, and

the heavier, larger grains will settle on the bottom

of it.

During ltration all suspended solids will accumu-

late in the ne upper layer. This phenomenon is

called surface or cake ltration. The cake is difcultto remove during regular backwash procedures.

Cracks will be formed in the cake, and preferential

ows will occur. In addition, mud balls will form and

settle on the lter bottom, clogging the nozzles.

When a stratied bed is backwashed at low veloci-

ties, only the upper layer will be expanded (with

the small grain sizes). The lower layers will not or

will hardly be expanded and accumulated solids

will not be removed. A faster backwash rate will

result in a washing out of the upper layers.

A non-uniform ow during backwashing (by a poor

design or clogging) can lead to preferential ows

in and disturbance of the lter bed.

This can result in total mixing of the lter material,

and support layers that must be situated at the bot-

tom can be found at the surface. This phenomenon

is called sand boil.

The ltrate and the wash water must be able to

pass through the lter bottom, but lter material

must be retained.Because the grain size of the lter material is about

1 mm, a small crack in the bottom is large enough

for the grains to pass through it and for the lter to

become a huge sandglass (Figure 20)

These cracks can be caused by damage in

the nozzle or inaccurate sealing of the bottom

plates.

5 Alternative applications of ltra-

tion

5.1 Multi-layer fltration

Multiple layer ltration consists of a lter bed with

Figure 20 - Result of a poorly designed lter bottom

L = 0.50 m; d = 1.0 - 1.4 mm

L = 0.75 m; d = 1.4 - 2.0 mm

L = 0.75 m; d = 2.0 - 2.8 mm

L = 0.30 m; d = 8 - 11 mm

L = 0.15 m; d = 32 - 45 mm

A :

B :

C :

D :

E :

A

B

C

DE

Figure 21 - Multiple layer ltration (upow lter)

95




various layers of different grain sizes. The water

passes the coarser grains rst, resulting is a ta-

pered ltering.

In upow lters, the coarse grains are at the bot-

tom (Figure 21).

In downow lters, the grain size decreases in a

downward direction.

The size and the density are chosen so that the

settling velocity of the material in the bed increases

in a downward direction and mixing between the

two layers during backwashing does not occur.

Usually sand is used as lter material in combina-

tion with either:

- a lter layer with a large grain size and a low

density on top of the sand (anthracite)

- a lter layer with a small grain size and a high

density below the sand layer (garnet).

Multiple layer ltration has an advantage that the

larger solids are retained in the top layer of the

bed and the smaller ones in the lower parts of

the bed.

Consequently, the increase in lter resistance is

spread over the entire height of the bed, resulting

in an extended lter run time.

In Figure 22 the progress of the water quality and

the resistance are represented for a single layer

and a double layer lter bed.

Using the same ltration velocity the resistance of

a double layer bed is lower than of a single layer

bed and the efuent water quality is better.

5.2 Pressure fltration

Pressure lters are based on the same principles

as gravity rapid lters. The only difference is that

the lter bed with the supporting lter bottom andthe supernatant raw water are encased in a wa-

ter-tight steel cylinder. This gives a closed system

in which the water to be treated can be forced

through the lter bed under pressure.

On one hand, this high pressure allows a large lter

resistance without the danger of negative heads;

while on the other hand, ltered water pumps are

no longer required and the lter can be placed at

any random level. Hence, the hydraulic head does

not have to be considered.

d = 1.0 m

d = 0.7 m

pressure distribution

heightfilterbed

t = tt = 0

Figure 23 - Resistance build-up for a multiple layer l -

ter

2.0

1.5

1.0

0.5

0

0 0.5 1.0 1.5

suspendedsolidcontent(g/m

3 )

Time (105 sec)

2.0

1.5

1.0

0.5

0

0 0.5 1.0 1.5

resistance(m)

Time (105 sec)

co

v

T

= 15 g/m3

= 3 mm/s

= 10oC

L = 1.1 m sand, d = 0.8 mm

L = 1.0 m athracite, d = 1.0 mm

Figure 22 - Difference between single and multiple layer ltration for equal bed heights

96




In addition, the application of a large lter re-

sistance permits the use of high ltration rates

through lter beds with long lter run times (Tr ).

The ltration rates normally vary from 7 to 20 m/h,

while values of 55 m/h are no exception. The sur -

face area of a pressure lter can thus be small.

The contact time between the water to be treated

and the ltering material becomes a limiting factor,

requiring a higher lter bed (3 m).

In practice, pressure lters are hardly used in drink-

ing water treatment because regular inspections

are difcult and the systems are rather sensitive.

Pressure lters are widely used in industrial water

supply.

The diameter of the steel cylinders are at the most

5 m. Hence, the capacity of the lter is 1000 m3/h

at its maximum.

When larger capacities are required, a horizontal

pressure lter can be applied. This is a pressure

lter with a width of 4 to 5 meters and an unlimited

length and, therefore, large lter surface areas can

be obtained. In practice this length has a maximum

of 15 meters.

The height above the lter is determined by the

distance between the drainage troughs and the

lter bed. This distance varies between 0.4 and

0.6 meters.

5.3 Upow fltration

The longest lter runs and the best water quality

are obtained when water passes a coarse fraction

rst and then a ner fraction of the lter material.

In upow ltration the coarse material is situated

at the bottom and the ne material at the top.

During backwashing and ltration, the lter bed is

conserved as a result of (natural) stratication.

The elevation height of the water is equal to the

wash

water

supply

filtrate

discharge

raw water supply

wash water

discharge

Figure 24 - Principle of pressure ltration

well with submersible pump

rapid pressure filter

elevated reservoir

to distribution system

Figure 25 - The ltered water ows without a pumping

phase towards the next treatment step

Figure 26 - Cylindrical steel pressure lters

97




hydrostatic water pressure plus the resistance due

to ow and clogging. This resistance is the largest

in the bottom of the bed (y=L).

Fluidization of sand with a density of 2600 kg/m3

and a porosity of 40% will occur when the resis-

tance is approximately equal to the height of the

lter bed.

If uidization occurs too quickly, a higher lter

bed or lter material with a higher density can be

applied.

In addition to the aforementioned advantage of

good efuent quality, upow ltration has several

disadvantages:

- wash water and ltrate are drained with the

same trough. This increases the risk of con-

tamination of the ltrate

- uidization of the top layer of the lter bed

can occur, resulting in a washing out of lter

material, diminishing the lter bed height and

lowering the removal efciency

- raw water is uniformly distributed by nozzles in

the bottom of the lter. The nozzles can become

clogged by impurities in the raw water,resulting

in extra resistance and a non-uniform distribu-

tion of water over the lter bed.

5.4 Limestone fltration

Limestone lters are lled with grains of calcium

carbonate or half-burned dolomite. When ag-

gressive water (with high levels of carbonic acid)

passes these lters, the concentration of carbonic

acid will decrease and the levels of hydrogen car -

bonate and pH will increase.

Water that is not in (calcium-carbonic acid) equilib-

rium dissolves limestone grains (calcium carbon-ate) according to the reaction:

CaCO3 + CO

2 + H

2O ←→ Ca2+ + 2.HCO

3-

Because the limestone grains are dissolved, they

need to be replenished regularly.

Normally, replenishing is executed when 10% of

the limestone is used. If limestone ltration is used

in groundwater treatment after aeration, ferric and

manganese removal and nitrication can occur in

the lter.

r e s i s t a n c e ( m )

L

c0

T

d

v

= 0.75 m

= 15 g/m3

= 10 oC

= 0.615 - 0.710 - 0.804

= 2 mm/s

filter resistance

suspended solids content

2

1.5

1

0.5

00 0.5 1 1.5 2

time (105 sec)

Figure 27 - Changes in quality and resistance of anupow lter

water 0.95 m

sand 1.25 m

gravel 0.65 mbeams to avoid

clogging of bottom

construction

Figure 28 - Principle of uplow ltration

h

L

A B C

soil pressure

water pressure, t = t

water pressure, t = 0

water pressure, v = 0

D E

AC

AB

BC

DE

= soil pressure

= water pressure

= grain pressure

= H

Figure 29 - Pressure distribution in an upow lter

98




5.5 Continuous fltration

In a continuous lter, sand is re-circulated and puri-

ed by a pump centrally placed in the lter. Hence,

the backwash process can be continuous.

In a continuous lter, the water ows in an upward

direction, and the transport of sand occurs in a

downward direction (Figure 30). The inlet of raw

water (with impurities) is at the bottom of the lter.

At the top of the lter the water ows over a weir

and is transported to other treatment processes.

The sand retains the impurities. By means of a

sand pump (mammoth pump), the lowest layers

of sand are removed from the lter and brought

to a sand washer that is situated above the lter.The sand washer removes the impurities from the

sand and the clean sand is supplied on top of the

continuous lter.

Due to the continuous removal of impurities, the

quality of the ltrate, the bed resistance and the

pressure distribution in the lter are constant and

time independent. The ltration rates of a continu-

ous lter vary from 14 to 18 m/h.

The advantages of a continuous lter compared

to a rapid sand lter are:

- continuous ltration process

- continuous wash water ow

- less accumulation of sludge.

The disadvantages of a continuous lter compared

to a rapid sand lter are:

- large wash water ows

- sand wash installation is in direct contact with

ltrate, resulting in contamination risks.

5.6 Dry fltration

Dry ltration is used when the water contains a

high ammonia concentration. Therefore, dry ltra-tion is only applied in river bank and groundwater

treatment.

The oxidation of ammonia into nitrate requires

large amounts of oxygen: 3.55 mg/l O2 per mg/l

NH4

+. The oxygen concentration of water is ap-

proximately 10 mg/l. Hence, in water with ammonia

concentrations larger than 2.5 mg/l, nitrication will

be incomplete.

Dry ltration has no supernatant water level. The

water to be treated ows in a downward direction

through a bed of granular material, accompanied

by a downward or upward ow of air of about

the same magnitude. A continuous gas trans-

fer between air and water will take place. The

oxygen consumed during the treatment can be

replenished directly by the accompanying air. The

formed carbon dioxide is removed from the water.

The pores are only partially lled with water and,

thus, the velocity of the water through the pores is

greater than in rapid sand (wet) ltration. The ow

conditions through the pores are turbulent ,therebypromoting the hydrodynamic transport of impurities

from the owing interstitial water to the lter grain

surfaces where they attach.

The ltered water collects below the lter bottom

and ows via gravity to the next treatment pro-

cess. From the ltrate chamber, air is continuously

pumped by a ventilator maintaining a (forced) si-

multaneous ow of air through the lter bed (Figure

31). When, in addition to oxygen transfer, the dry

lter is also used for gas stripping, a counter-cur -

filtratewash

water

sand sand

water water

water

Figure 30 - Schematic representation of a continuous

lter

99




rent ow of water and air can be used in the lter

(Figure 32).

The water to be treated is distributed over the full

area of the dry lter bed as evenly as possible with

the help of spray nozzles.

Spraying has two objectives:

- gas transfer (addition of oxygen and removal

of methane and carbonic acid)

- uniform distribution of the water over the l-

ter.

A dry lter does not only remove ammonia, but

also iron and manganese. In the top layer of the

lter bed (depth of 0.5 to 1.5 m) iron removal takes

place. After completion of this process, manga-

nese and ammonia removal occurs more or less

simultaneously. Dry lters are often followed by

rapid lters. The reason is that in a dry lter bacte-ria form. The rapid sand (wet lter) forms a barrier

against the breakthrough of these bacteria.

5.7 Slow sand fltration

When the most important objective of a lter is to

remove bacteria and viruses and the lter is an

alternative for chemical disinfection, slow sand

lters are suitable.

The lter material has a small grain size (e.g., 0.2

to 0.6 mm) and the ltration rate is below 1 m/h.

For treatment of the same water ow, a larger

ltration surface area is needed than that used for

rapid lters. This is illustrated in the aerial picture

of the treatment plant at Leiduin (Figure 33).

Filtration occurs mainly in the top layer of a slow

sand ltration, where a biologically active “Sch-

mutzdecke” is formed.

air and water

air

raw water

used air

filtrate

Figure 31 - Schematic representation of a co-currentlter

air and water

raw water

air

filtrate

used air

Figure 32 - Schematic representation of a counter-cur -

rent dry lter

Figure 33 - Difference in surface area between rapid

ltration(orange) and slow sand ltration

red)

100






102




Adsorption WA T

E R T R E A T M E

N T

WATER TREATMENT

pump

pump a t r a z i n e (

m g / l )

2

1.5

1

0.5

00 10,000 20,000 30,000

bed volumes (m3/m3)

influent

effluent

macro pore

micro pores

pesticides

meso pore



Framework

This module explains adsorption.

Contents


1. Introduction

2. Theory

2.1 Equilibrium

2.2 Kinetics

2.3 Mass balance

2.4 Solutions for the basic equations

3. Practice

3.1 Pseudo-moving-bed ltration

3.2 Pressure ltration 3.3 Powdered activated carbon

104

WATER TREATMENT ADSORPTION



1 Introduction

Water contains dissolved organic matter that can-

not be removed with oc formation, oc removal

or sand ltration. These dissolved organic com-

pounds are:

- odor-, taste- and color-producing compounds

- organic micropollutants (pesticides, hydrocar -

bon compounds)

Activated carbon adsorbs (part of) the organic

matter and is mainly used to treat drinking water

produced from surface water.

In the past, drin�ing water produced from surfacerinking water produced from surface

water would pass through the following steps: ocformation, oc removal (sedimentation and ltra-

tion) and disinfection with chlorine.

This was sufcient to comply with the drin�ing

water guidelines for turbidity, odor, taste and hygi-

enic reliability.

In 1987 Amsterdam Water Supply discovered

the presence of pesticides (Bentazon) in drinking

water. Due to this discovery, the traditional treat-

ment of surface water was no longer satisfac-

tory and an extension with activated carbon was

required.

In addition, chlorine can react with organic matter

(precursor), and trihalomethanes (THMs) can be

formed. These THMs are toxic.

To reduce the concentration, the formed THMs can

be removed with activated carbon. The problem,

however, is that carbon is rapidly saturated with

THMs and has to be regenerated frequently. It is

preferable to prevent the THMs from being formed.This can be done by reducing the concentration of

the precursor before adding chlorine.

Precursors cannot be measured directly, but the

concentration of organic matter, expressed as TOC

or DOC, is an indication.

The best way to prevent THM formation is to avoid

the dosing of chlorine, but this requires another

treatment setup.

Activated carbon is a substance with a high carbon

concentration (e.g., pit-coal, turf).

Under high temperatures this material becomes

carbonated, meaning that the carbon partly trans-

forms into carbon monoxide and water. This is how

the carbon gets its open structure (Figure 1).

The internal surface area of the activated carbon is

several times larger than the external surface area.

macro pore

micro pores

pesticides

macro pores > 25 nm

meso pores 1 - 25 nm

micro pores < 1 nm

meso pore

Figure 1 - The open structure of activated carbon

The activated carbon lters of the drin�ing water

production plant at Kralingen have the objective

to improve the taste of the water, to reduce the

regrowth of bacteria in the piped networ�, and

to remove toxic substances from the water.

The carbon lters are placed after the oc for -

mation, sedimentation and rapid sand ltration

to avoid rapid clogging.

The installation is based on pressure lters, so

the construction height can be limited and only

an extra pumping phase (middle pressure) is

needed.

The installation has the following characteris-tics:

Filter bed height: 4 m

Filtration surface area: 28 m2

Number of lters: 12

Contact time (EBCT): 12 min

Type of carbon: Norit ROW 0.95

Regeneration frequency: 1.5 year

105

ADSORPTIONWATER TREATMENT



Hence, a majority of the adsorbed substances is

adsorbed inside the carbon.

The dissolved organic matter can be removed

from the water by ltration through a bed of acti-

vated carbon.

Organic matter diffuses from the water phase to

the surface of the carbon grains. The organic com-

pounds are further transported into the carbon to

be attached in the pores.

The adsorption of organic matter is not nite.

There is equilibrium between the concentration of

dissolved compounds in water and the quantity of

substances that are adsorbed onto the carbon.

When different kinds of organic compoundsare present in the water, competition will occur.

Compounds that have been well-adsorbed will

occupy adsorption places that cannot be used

by compounds that are less adsorbable. Large

organic molecules can also bloc� micro pores,

thus preventing the smaller organic molecules

from entering these micro pores.

After some time the activated carbon is saturated

with adsorbed organic matter and the carbon

needs to be cleaned. This is done by removing

the carbon from the installation and heating it to

1000 oC.

This regeneration process has to be carried out

once every couple of years.

Activated carbon lters operate in the same way

as rapid sand lters. Mostly downward ow, open

lters are applied to prevent ne carbon grains

from washing out (Figure 2).

Since the contact time is the most important param-eter for good removal, the lter is often designed

with high beds to reduce the construction surface

area. Therefore, activated carbon lters cannot be

operated under gravity, ma�ing an extra pumping

phase necessary. After passing the lter bed, the

water reaches the bottom construction of the lter

and is then collected and transported to the clear

water tank or the next treatment step.

When the lter is clogged with suspended matter

or biomass and the resistance is high, the lter

bed is backwashed.

The backwash water is drained by troughs that are

placed above the lters.

Clogging of the lter by suspended matter can

lead to a higher regeneration frequency. Therefore,

activated carbon lters are usually placed after

oc formation, oc removal and rapid sand ltra-

tion (Figure 3).

A

A cross-section A-A

Figure 2 - Longitudinal and cross-section of an acti-

vated carbon ltration installation

clear water reservoir

activated carbon filtration

rapid filtration

floc removal

reservoir

ozonation

floc formation

Cl2/ClO2

Chlorine dosage for transport = 0.3 mg/l

activated carbon filtration:- removal of organic matter

- removal of pesticides

ozonation:- desinfection

- oxidation of organic matter dosage = 2-3,5 mg/l O3

reservoir:- storage

- leveling off - auto purification

Fe(III)

Figure 3 - Treatment process at Kralingen (Evides)

Figure 4 - Activated carbon ltration at Andijk (treatment

of surface water)

106




After activated carbon ltration, disinfection ta�es

place to prevent biological growth in the piped

network. This process can consist of slow sand

ltration, UV-disinfection or dosing chlorine or

chlorine dioxide.

2 Theory

2.1 Equilibrium

Similar to aeration, equilibrium is established dur-dur-

ing adsorption..

The maximum loading (loading capacity qmax)

depends on the concentration of adsorbable mat-

ter in the bulk liquid (water). The higher this con-centration, the higher the loading capacity is.

The relationship between the loading capacity and

the concentration of adsorbable matter in the bulk

liquid is called the adsorption-isotherm.

The best known is the Freundlich isotherm:

n

max s

xq K c

m= = ⋅

in which:

qmax = loading capacity (g/kg)

cs = equilibrium concentration (g/m3)

x = adsorbed amount of compound (g)

m = mass of activated carbon (kg)

K = Freundlich constant ((g/kg).(m3/g)n)

n = Freundlich constant (-)

The Freundlich constants K and n are inuenced

by water temperature, pH, type of carbon and the

concentration of other organic compounds.

Using laboratory experiments the Freundlich con-

stants can be determined for a single substance

with a specic type of activated carbon.

In Figure 5 the results of a laboratory experiment

are represented.

When these graphs are plotted on a logarithmic

scale (Figure 6), the Freundlich constant K can

be determined from the intersection of the graph

with the y-axis. The slope of the line is equal to

the Freundlich constant n.

The higher the K-value, the better the adsorp-

tion.

In Table 1 the values of the constants K and n are

given for some known substances.

From the structure formula of a substance, the

adsorbability can be derived. In general, non-

polar substances are better adsorbed than polar

substances. Substances with double bonds will

be better adsorbed than substances with single

bonds.

phenol

chloro phenol

dichloro phenol

trichloro phenol

350

300

250

200

150

100

50

00 10 20

cequilibrium (mg/m3)

q ( m g / k g )

Figure 5 - Loading capacity as a function of equilibrium

concentration

phenol

chloro phenol

dichloro phenol

trichloro phenol

150

100

50

5 10 30

cequilibrium (mg/m3)

0

0

q(mg/kg)

Figure 6 - Logarithm representation of the Freundlich

isotherm

107




2.2 Kinetics

The kinetics equation (equation of motion) for

activated carbon ltration is as follows:

2 0 s

dc dcu k (c c )

dt dy= − − ⋅ −

in which:

k2 = mass transfer coefcient (d-1)

c0 = initial concentration of organic compound

(mg/l)

u = pore velocity of the water (m/s)

cs = equilibrium concentration of organic

compound linked to a certain loading of the

activated carbon (mg/l)

The kinetics equation consists of a convection term

with which transport of the compound through the

lter bed can be described::

dcu

dy

and a removal term:

2 0 sk (c c )−

The rate of mass transfer is similar to aeration pro-

portional to the difference between the prevailing

concentration and the equilibrium concentration.

The equilibrium concentration depends on the

loading and is determined by the Freundlich iso-

therm.

The lower the loading of the carbon, the lower is

the equilibrium concentration and the higher is the

mass transfer rate.

The mass transfer coefcient depends on the

compound to be adsorbed and the type of carbon

(including the grain size).

In addition, the mass transfer coefcient can be

inuenced by the velocity of the water passing

the carbon grains. The higher the velocity of thewater, the better the mass transfer is between

liquid and carbon.

2.3 Mass balance

In Figure 7 the activated carbon lter is schema-

tized as a cube, in which:

Q = ow (m3/h)

B = width of lter (m)

L = length of lter (m)

Compound K.

((g/kg).

(m3 /g)n)

n (-)

alkanes

CH3Cl 6.2 0.80

CH2Cl

212.7 12.7

CH2Br 44.4 0.81

CHCl3 (chloroform) 95.5 0.67

CHBr 3 (bromoform) 929 0.66

CH2Cl - CH

2Cl (DCEA) 129 0.53

CH2Br - CH

2Br (EDB) 888 0.47

CH2Cl - CHCl - CH

3 (1,2 DCP) 313 0.59

CH2Br - CHBr - CH

2Cl(DBCP) 6910 0.60

alkenes

CCl2 = CHCl (TCE) 2000 0.48

CCl2

= CCl2

(PCE) 4050 0.52

pesticides - organochlorides

Dieldrin 17884 0.51

Lindane (HCH) 15000 0.43

Heptachlor 16196 0.92

Alachlor 81700 0.26

pesticides - organitrogenes

atrazine 38700 0.29

simazine 31300 0.23

pesticides - fenolderivates

dinoseb 30400 0.28

PCP 42600 0.34pesticides - fenoxycarbonicacid

2,4 D 10442 0.27

2,4,5 TP 15392 0.38

aromates

C6H

6(benzene) 1260 0.53

C6H

5Cl 9170 0.35

CH5CH

3 (toluene) 5010 0.43

C6H

5NO

2 (nitrobenzene) 3488 0.43

C6H

5COOH 2802 0.42

C6H

5OH (phenol) 503 0.54

C10H8 (naftalene) 7260 0.42

Table 1 - Freundlich constants K and n of several

substances

108




dy = height of lter (m)

c = concentration of organic compound (g/m3)

An organic compound with a concentration c0

enters the system with a ow Q and leaves the

system with a concentration c1. The difference in

concentration between in- and outow is adsorbed

on the activated carbon and increases the loading

of the carbon.

The continuity equation or mass balance is:

dq v dc

dt dy= −

ρ

in which:

v = ltration rate = Q/BL (m/h)

q = loading (g/g)ρ = density of the carbon (g/m3)


The system of equations is non-linear and cannot

be solved analytically.

When a stationary situation is assumed, and

when the inuent concentration and the ow are

assumed to be constant, the efuent concentra-

tion of the activated carbon lter can be calculated

using the Bohart-Adams equation. This equation

is derived from the mass balance and the kinet-

ics equation.

0 02

e

c BV c1 exp � EBCT 1c q

⋅ = + ⋅ ⋅ − ⋅ ρ

VEBCT

Q=

Q T TBV

V EBCT

⋅= =

in which:

EBCT = empty bed contact time (h)

BV = ltered water per bed volume (m3/m3)

T = lter run time (h)

V = volume lter (m3)

Figure 8 shows the progress of the organic com-

pound concentration in the activated carbon l-

tration.

Water with a concentration of organic compound

c0 is supplied. Since, in the beginning, the carbon

is not yet loaded, the efuent concentration of the

organic compound drops to zero.

After some time, the loading of the carbon

increases, the available adsorption places are

lled, and brea�through of the organic compound

in the efuent occurs.

BV=0

∆ t

2∆ t

3∆ t

H

co

Figure 8 - Progress of the concentration in time and

height

Figure 7 - Schematic representation of activated car-

bon lter

109




Finally, the activated carbon is saturated without

any removal of organic matter.

In Figure 9 the efuent concentration is plotted

against the lter run time (expressed in number

of bed volumes). This curve is called the break-

through curve.

In the beginning the efuent concentration is 0.

After time the efuent concentration increases

until the activated carbon is saturated and the

efuent concentration is equal to the inuent con-

centration.

When the efuent concentration of the activated

carbon lter no longer meets the standards, the

lter must be regenerated.

The run time of activated carbon lters depends

on the objective.

The brea�through of THMs occurs relatively fast

(15,000 BV); for the removal of taste substances,

however, longer run times can be applied (50,000BV) without intermediate regeneration (Figure

9).

Contact time

The correlation between contact time and lter run

time depends on the adsorption characteristics of

the compound to be removed.

In general the lter run time increases exponen-

tially with increasing contact time (Figure 10).

Hence, per cubic meter of activated carbon, a

larger volume of water can be treated before

regeneration is necessary. Application of a shorter contact time indicates that

a smaller volume of activated carbon is needed.

This leads to lower investment costs.

The regeneration costs, on the other hand, will

increase.

An economical optimum depends on the adsorp-

tion behavior of the compound to be removed.

3 Practice

To solve the above-given basic equation, a sta-

tionary situation is assumed in which the inuent

concentration is considered to be constant.

In reality, the inuent concentration is not constant

but varies. Hence, the brea�through curve will not

describe a perfect “S”-form (Figure 11).

If the inuent concentration is high, relatively large

amounts of organic matter are adsorbed because

300

f i l t e r r u n t i m e ( d a y s )

contact time (min)

250

200

150

100

50

00 10 20 30 40

Figure 10 - Relationship between contact time and lter

run time THM Bentazon taste

0 20,000 40,000 60,000

0.25

0.5

0.75

1

0

bed volumes (m3 /m3)

c e

/ c 0 (

- )

Figure 9 - Breakthrough curve

a t r a z i n e (

m g / l )

2

1.5

1

0.5

00 10,000 20,000 30,000

bed volumes (m3/m3)

influent

effluent

Figure 11 - Breakthrough curve measured in practice

110




of a large driving force. When, afterwards, a low

inuent concentration occurs and the loading of thecarbon is already high, the saturation concentra-

tion can be almost equal to the inuent concentra-

tion and adsorption is limited.

In extreme cases even desorption (higher concen-

trations of organic compounds in efuent than in

inuent) can occur.

This happens, for example, after terminating trans-

port chlorination. Before termination, THMs were

removed in the lters and accumulated in the acti-

vated carbon. After termination, THMs no longer

form and the inuent concentration is nil (Figure

12). In the efuent of the activated carbon lters,

THMs are still present because of desorption.

3.1 Pseudo-moving-bed fltration

A special application for activated carbon ltration

is the “pseudo-moving-bed” system (Figure 13),

where two lters are placed in a series.

After brea�through occurs in the rst lter, the sec-

ond lter ta�es care of the polishing, and the efu-

ent of the second lter still meets the guidelines.

The moment the second lter brea�s through, the

rst lter is regenerated and then connected after

the second lter.

The cleanest lter is thus always the last one

(Figure 14).

The advantage of this setup is that the storage

capacity of the lters is better used, resulting in

longer run times.

Disadvantages of this setup are the high hydraulic

loadings of the lters and the complex system of

pipes and valves.

3.2 Pressure fltration

c h l o r o f o r m C

H C l 3 ( m g / l )

0

0 20,000 30,000

bed volumes (m3/m3)

influent

effluent

10,000

10

20

30

40

50

transport chlorination stopped

Figure 12 - Occurrence of desorption after termination

of transport chlorination

t = 0 t = ∆T

filter 1 is regenerated

t = 2∆T t = 3∆T

filter 2 is regenerated

Figure 14 - Breakthrough curves for pseudo-moving-bedactivated carbon ltration

Figure 15 - Steel pressure lters with activated carbon

pump

pump

Figure 13 - Principle of pseudo-moving-bed ltration

111




When activated carbon lters are placed in pres-

sure vessels, an extra pumping phase can be

avoided (Figure 15). The pressure is then sufcient

to lead the water through the activated carbon

lters to the clear water tan�s.

3.3 Biological activated carbon fltra-

tion

Biological activity occurs in all carbon filters.

Bacteria that grow on the carbon will decompose

organic matter.

Even with pre-chlorination bacteria can grow,

because the residual chlorine is adsorbed by the

activated carbon.

When organic matter is pre-oxidized by ozone (a

strong oxidizer), biological activity is stimulated,

resulting in increased lter run times and increased

removal of organic matter. This is called “biologi-

cally activated carbon ltration” (Figure 16).

Because of the increased biological activity on

the carbon, the organic macropollutants (DOC)

will occupy fewer adsorption places in the car-

bon. Hence, more space is left for the (persistent)

organic micropollutants. Some persistent organic

micropollutants like pesticides can even be (par-

tially) biologically decomposed after ozonation.

In some seasons (e.g., summer), the biological

activity will be greater than in other seasons (e.g.,

winter).

Therefore, the adsorption process will be the domi-

nant process in winter. The organic matter that is

adsorbed in winter will be partially decomposed

biologically in summer.

This phenomenon is called bio-regeneration.

With biologically activated carbon ltration, quic�ly

degradable matter is formed that will stimulate

bacterial growth.

Breakthrough of this degradable matter (expressed

as Assimilable Organic Carbon(AOC)) must be

avoided to prevent the regrowth of bacteria in the

piped network.

In addition, bacteria can be eroded from the car -

bon and enter the efuent, thus increasing thecolony counts.

A disinfection step with UV-radiation or chlorine

will then be necessary.

3.4 Powdered activated carbon

In addition to granular activated carbon ltration

(GAC), powdered activated carbon (PAC) dosing

can be applied.

With powdered activated carbon, small carbon

particles (1µm) are added to water.

These carbon particles are so small that during

transport they do not settle.

When the water is in contact with the carbon par-

ticles (after some time), equilibrium between the

organic matter in the water and on the powdered

carbon will be established.

The particles will be removed afterwards by a sand

ltration step.

For a powdered activated carbon dosing, a mass

balance can be set up, schematically represented

in Figure 17 :

0 0 e ec V q m c V q m⋅ + ⋅ = ⋅ + ⋅

0 ee

e

c cmq 0 W

V q

−= ⇒ = =

in which:

m = mass of activated carbon (g)

r u n t i m e ( y e a r s )

00 100

contact time (minutes)

with ozone

without ozone

guideline

2

1.5

0.5

50

1

Figure 16 - Inuence of pre-ozonation on required con-

112




A powdered activated carbon dosage has theactivated carbon dosage has thecarbon dosage has the

advantage over granular activated carbon ltra-tion that there are limited investment costs: there

is no need for a ltration installation (with an extra

pumping phase); a dosing unit for the powdered

activated carbon together with a mixing tank iscarbon together with a mixing tank is

sufcient.

The removal of pesticides with powdered activatedactivated

carbon, however, is limited and the rapid lters are

quickly clogged. This results in a large backwash

water loss.

Granular activated carbon ltration has a more

intensive water/carbon contact, thus pesticides

(and THMs) are removed more efciently.

V = volume lter (m3)

W = dosing of activated carbon (g/m3)

Starting with the Freundlich isotherm, the efu-

ent concentration of an organic compound can

be calculated, or the powdered activated carbon

dosage can be calculated for a determined efu-

ent concentration.

It is assumed that equilibrium occurs and the car-

bon is maximally loaded.The loading capacity is determined by the pre-

vailing concentration in the reactor, and in a

completely mixed system this equals the efuent

concentration.

For two ideal mixers placed in a series, the loading

capacity of the powdered activated carbon in the

second tan� is lower than in the rst tan� because

the concentration in the second tank is lower than

in the rst tan� (Figure 18).

V c0

q0, m

ce

qe, m

Figure 17 - Mass balance of powderd carbon dosage

c1 c0

c1 c0c2

q1

q2

q1

Ideal mixer, 1 - step

2 - step

c1 - c0 = W · q1

c1 - c0 = -W1·q1 -W·q1

(mass balance)

Figure 18 - One and two mixers in series

Figure 19 - Powdered activated carbon dosing unit at

Scheveningen

Further reading

• Water treatment: Principles and design, MWH

(2005), (ISBN 0 471 11018 3) (1948 pgs)

• Unit processes in drin�ing water treatment, W.

Masschelein (1992), (ISBN 0 8247 8678 5)

(635 pgs)

113




114




Combined chlorine Free chlorine Chlorine dioxide Ozone UV Light

Required

Ct or lt

0.01

0.10

1.0

10

100

1.000

10.000 C. Parvum

C. Parvum

C. Parvum

C. Parvum C. Parvum

Giardia

Legionella

Legionella

Legionella

Legionella

Mycobacterium

fortuitumM. fortuitum

M. fortuitum

M .

f o r t u i t u m

Poliovirus

Poliovirus

Poliovirus

Poliovirus

Poliovirus

Adenovirus

Adenovirus Adenovirus

E. coli

E. coli

E. coli

E. coli

Calicivirus

Calicivirus

Calicivirus

Microsporidium

Giardia Giardia

Giardia Giardia

Legionella

PneumophilaMicrosporidium

Microsporidium

Adenovirus

Calicivirus

E. Coli

Adenovirus

ReovirusMS-2

CalicivirusRotavirus

Hepatitis A

Disinfection WA T

E R T R E A T M E

N T

WATER TREATMENT



Framework

This module will describe the aspects of water disinfection. For this, the purpose of disinfection will be

given, the kinetics, and the practical application.

The content of this module is abstracted from Alternative Disinfectants and Oxidants Guidance Manual

(EPA 1999) and Water treatment: Principles and design (MWH 2005).

Contents


1. Introduction

2. Purpose of disinfection

2.1 Diseases and drinking water

2.2 Pathogens of primary concern

2.3 Recent waterborne outbreaks 2.4 Mechanism of pathogen inactivation

2.5 Other uses of disinfectants in water treatment

2.6 Current practice of disinfection (and oxidation)

2.7 Disinfection byproducts

3. Disinfection kinetics

3.1 Chick’s Law

3.2 Chick-Watson model

3.3 Other models

3.4 C t -values

4. Disinfection methods

4.1 Chlorine

4.2 Ozone

4.3 UV radiation

4.4 Chlorine dioxide

4.5 Other methods

Further reading

116

DISINFECTOIN WATER TREATMENT



produced primarily as a result of chlorination

- organic oxidation byproducts such as alde-

hydes, ketones, assimilable organic carbon

(AOC), and biodegradable organic carbon

(BDOC) that are associated primarily with

strong oxidants such as ozone, chlorine, and

advanced oxidation

- inorganics such as chlorate and chlorite as-

sociated with chlorine dioxide, and bromate

that is associated with ozone, and has also

been found when chlorine dioxide is exposed

to sunlight.

The type and amount of DBPs produced dur-

ing treatment depends largely on the type of

disinfectant, water quality, treatment sequences,contact time, and environmental factors such as

temperature and pH.

When considering the use of alternative disinfec-

tants, systems should ensure that the inactivation

of pathogenic organisms is not compromised.

Pathogens pose an immediate critical public health

threat due to the risk of an acute disease outbreak.

Although most identied public health risks associ-

ated with DBPs are chronic, long-term risks, many

systems will be able to lower DBP levels without

compromising microbial protection.

In this module the purpose of disinfection is pre-

sented rst. Thereafter, the DBPs are discussed,

since they play an important role in the selection

of the disinfection method.

After this, disinfection kinetics are presented.

Finally, an overview is given of the different dis-

infection methods, in which the pros and cons of

the major methods are provided.

2 Purpose of disinfection

2.1 Diseases and drinking water

Although the epidemiological relationship between

water and disease had been suggested as early

as the 1850s, it was not until the development

of the germ theory of disease by Pasteur in the

mid-1880s that water as a carrier of disease-

producing organisms was understood.

1 Introduction

The most important use of disinfectants in water

treatment is to limit waterborne diseases and inac-

tivate pathogenic organisms in water supplies.

The rst use of disinfection as a continuous pro-

cess in water treatment took place in a small town

in Belgium in the early 1900s (White, 1992), where

chlorine was used as the disinfecting reagent.

Since the introduction of ltration and disinfection

at water treatment plants, waterborne diseases,

such as typhoid and cholera, have been virtually

eliminated. For example, in Niagara Falls, NY,

USA, between 1911 and 1915, the number of ty-

phoid cases dropped from 185 deaths per 100,000people to nearly zero following the introduction of

ltration and chlorination (White, 1986).

For nearly a century, chlorine gas or chlorine re-

agents (hypochlorite, etc.) were, by far, the most

commonly used disinfectant chemicals for drinking

water production

In 1974, researchers in the Netherlands and the

United States demonstrated that trihalomethanes

(THMs) were being formed as a result of drink-

ing water chlorination (Rook, 1974; Bellar et al.,

1974).

THMs form when chlorine or bromide reacts with

organic compounds in the water. THMs and other

disinfection byproducts (DBPs) have been shown

to be carcinogenic, mutagenic, etc. These health

risks may be small but need to be taken seriously,need to be taken seriously,,

when you consider the large population being

exposed.

As a result of DBP concerns from chlorine, the wa-ter treatment industry has placed more emphasis

on the use of disinfectants other than chlorine.

Some of these alternative disinfectants, however,

have also been found to produce DBPs as a re-

sult of either reactions between disinfectants and

compounds in the water or as a natural decaying

process of the disinfectant itself (McGuire et al.,

1990; Legube et al., 1989).

These DBPs include:

- halogenated organics, such as THMs, halo-

acetic acids, haloketones, and others that are

WATER TREATMENT

117

DISINFECTION



In the 1880s, while London was experiencing the

“Broad Street Well” cholera epidemic, Dr. John

Snow conducted his now famous epidemiological

study. Dr. Snow concluded that the well had

become contaminated by a visitor with the

disease who had arrived in the vicinity.

Cholera was one of the rst diseases to be

recognized as capable of being waterborne.

Also, this incident was probably the rst reported

disease epidemic attributed to the direct recycling

of non-disinfected water.

Now, over 100 years later, the list of potential

waterborne diseases due to pathogens is

considerably longer, and includes bacterial,

parasitic, and viral microorganisms, as shown in

Tables 1, 2 and 3, respectively.

A major cause for the number of disease

outbreaks in potable water is contamination of the

distribution system from cross-connections and

back siphoning with non-potable water. However,

outbreaks resulting from distribution system

contamination are usually quickly contained and

result in relatively few illnesses compared to the

many cases of illness per incident when there is

contamination of the source water or a breakdown

in the treatment system.

When considering the number of cases, the major

causes of disease outbreaks are source water

contamination and treatment deciencies (White,

1992). For example, in 1993 a Cryptosporidiosis

outbreak affected over 400,000 people in

Milwaukee, Wisconsin (USA). The outbreak was

associated with deterioration in the raw water

Causative agent Disease Symptoms

Salmonella typhosa Typhoid fever Headache, neasea, loss of appetite, constipation or diarrhea,

insomnia, sore throat, bronchitis, abdominal pain, nose

bleeding, shivering and increasing fever, rosy spots on trunk.

Incubation period: 7 - 14 days.

S. paratyphi

S. schottinulleri

S. hirschfeldi C.

Paratyphoid fever General infection characterized by continued fever, diarrhea

disturbances, sometimes rosy spots on trunk. Incubation

period: 1 - 7 days.

Shigella fexneri

Sh. dysenteriae

Sh. sonnei

Sh. paradysinteriae

Bacillary dysentery Acute onset with diarrhea, fever, tenesmus and stool fre-

quently containing mucus and blood. Incubation period: 1 - 7

days.

Vibrio comma

V. Cholerae

Cholera Diarrhea, vomiting, rice water stools, thirst, pain, coma.

Incubation period: a few hours to 5 days.

Pasteurellla tularensis Tularemia Sudden onset with pains and fever; prostration. Incubation

period: 1 - 10 days.

Brucella melitensis Brucellosis (undulant fever) Irregular fever, sweating, chi lls, pain in muscles.

Pseudomonas pseudomallei Melioidosis Acute diarrhea, vomiting, high fever, delerium, mania.

Leptospira icterohaemorrhagiae

(spirochaetales)

Leptospirosis (Well’s

disease)

Fevers, rigors, headaches, nausea, muscular pains, vomit-

ing, thirst, prostration and jaundice may occur.

Enteropathogenic E. coli Gastroenteritis Water diarrhea, nausea, prostration and dehydration.

Table 1 - Waterborne diseases from bacteria

Causitive agent Disease Symptoms

Ascario lumricoidis (round worm) Ascariasis Vomiting, live worms in feces.

Cryptosporidium muris

Cryptosporidium parvum

Cryptosporidiosis Acute diarrhea, abdominal pain, vomitin, and low-grade

fever. Can be life-threatening in immunodecient patients.

Entamoeba histolytica Amebiasis Diarrhea alternating with constipation, chronic dysentery with

mucus and blood.

Giardia lamblia Giardiasis Intermittent diarrhea.

Naegleria gruberi Amoebid menigoecephalitis Death.

Schistosoma mansoni Schistosomiasis Liver and bladder infection.

Taenia saginata (beef tapeworm) Taeniasis Abdominal pain, digestive disturbances, loss of weight.

Table 2 - Waterborne diseases from Parasites (Protozoa)

118




quality and a simultaneous decrease in the

effectiveness of the coagulation-ltration process

(Kramer et al., 1996; MacKenzie et al., 1994).

Historically, about 46 percent of the outbreaks

in public water systems are found to be related

to deciencies in source water and treatment

systems, with 92 percent of the causes of illness

due to these two particular problems.

All natural waters support biological communities.

Because some microorganisms can be

responsible for public health problems, the

biological characteristics of the source water are

one of the most important parameters in water

treatment.

In addition to public health problems, microbiology

Causative agent Disease Symptoms

Enterovirus Polio (3) Muscular paralysis

Aseptic meningitis

Febrille episode

Destruction of motor neurons

Inammation of meninges from virus

Viremia and viral multiplication

Enterovirus Echo (34) Aseptic meningitisMuscular paralysis

Guillain-Barre’s Syndrome1

Exanthem

Respiratory diseases

Diarrhea

Epidemic myalgia

Pericardits and myocarditis

Hepatitis

Inammation of meninges from virusDestruction of motor neurons


Dilation and rupture of blood vessels

Viral invasion of parechymiatous of respiratory tracts and second-

ary inammatory responses intestinal infections

Not well known

Viral invasion of cells with secondary infammatory responses

Invasion of parencheyma cells

Enterovirus Coxsackie (>24) Herpengina2 Viral invasion of mucosa with secondary inammation

Enterovirus A Aculte lymphatic pharyngitis

Aseptic meningitis

Muscular paralysis

Hand-foot-mouth disease3

Respiratory disease

Infantile diarrhea

Hepatitis

Pericarditis and myocarditis

Sore throat, pharyngeal lesions



Viral invasions of skin cells of hands-feet-mouthViral invasion of parenchymiatous of respiratory tracts and

secondary infammatory responses

Viral invasion of cells of mucosa

Viral invasion of parenchyma cells

Viral invasion of cells with secondary inammatory responses

Enterovirus B (6) Pleurodynia4

Aseptic meningitis

Muscular paralysis

Meningoencephalitis

Pericarditis, endocarditis,

myocarditis

Respiratory disease

Hepatitis or Rash

Spontaneous abortion

Insulin-dependent diabetes

Congenital heart anomalies

Viral invasion of muscle cells



Viral invasion of cells

Viral invasion of cells with secondary inammatory responses

Viral invasion of parenchymiatous of respiratory tracts and

secondary inammatory responses

Invasion of parenchyma cellsViral invasion of vascular cells

Viral invasion of insulin-producing cells

Viral invasion muscle cells

Reovirus (6) Not well known Not well known

Adenovirus (31) Respiratory diseases

Acute conjunctivitis

Acute appendicitis

Intussusception

Subacute thyroiditis

Sarcoma in hamsters

Viral invasion of parenchymiatous of respiratory tracts and

secondary inammatory responses

Viral invasion of cells and secondary inammatory responses

Viral invasion of mucosa cells

Viral invasion of lymph nodes

Viral invasion of parenchyma cells

Sarcoma in hamsters

Hepatitis (>2) Infectious hepatitis

Serum hepatitis

Down’s syndrome

Invasion of parenchyma cells

Invasion of parenchyma cells

Invasion of cells

Table 3 - Waterborne diseases from Human Enteric Viruses

WATER TREATMENT

119

DISINFECTION



can also affect the physical and chemical water

quality and treatment plant operation.

2.2 Pathogens of primary concern

Table 4 shows the attributes of three groups of

pathogens of concern in water treatment, namely

bacteria, viruses, and protozoa.

Bacteria

Bacteria are single-celled organisms typically

ranging in size from 0.1 to 10 µm.

Shape, components, size, and the manner in which

they grow can characterize the physical structure

of the bacterial cell.

Most bacteria can be grouped by shape into four

general categories: spheroid, rod, curved rod or

spiral, and lamentous.

Cocci, or spherical bacteria, are approximately 1

to 3 µm in diameter.

Bacilli (rod-shaped bacteria) vary in size and range

from 0.3 to 1.5 µm in width (or diameter) and from

1.0 to 10.0 µm in length.µm in length.m in length.Vibrios, or curved rod-shaped bacteria, typically

vary in size from 0.6 to 1.0 µm in width (or diam-

eter) and from 2 to 6 µm in length.

Spirilla (spiral bacteria) can be found in lengths up

to 50 µm, whereas lamentous bacteria can occur

in lengths in excess of 100 µm.

Viruses

Viruses are microorganisms composed of the

genetic material deoxyribonucleic acid (DNA) or ri-

bonucleic acid (RNA) and a protective protein coat

(single-, double-, or partially double-stranded).

All viruses are obligate parasites, unable to carryout any form of metabolism and are completely

dependent upon host cells for replication.

Viruses are typically 0.01 to 0.1 µm in size and

are very species specic with respect to infection,

typically attacking only one type of host.

Although the principal modes of transmission for

the hepatitis B virus and poliovirus are through

food, personal contact, or exchange of body u-

ids, these viruses can also be transmitted through

potable water.

Some viruses, such as the retroviruses (including

the HIV group), appear to be too fragile for water

transmission to be a signicant danger to public

health (Riggs, 1989).

Protozoa

Protozoa are single-cell eucaryotic microorgan-

isms without cell walls that utilize bacteria and

other organisms for food.

Most protozoa are free-living in nature and can beencountered in water; however, several species

are parasitic and live on or in host organisms.

Host organisms can vary from primitive organisms

such as algae to highly complex organisms such

as human beings.

Several species of protozoa known to utilize hu-

man beings as hosts are shown in Table 5.

2.3 Recent waterborne outbreaks

Within the past 40 years, several pathogenic

Organism Size

(µm)

Mobility Point(s) of origin Resistance to disinfection

Removal by

sedimentation,

coagulation and

ltration

Bacteria 0.1 - 10 Motile,Nonmotile

Humans and animals,water and contami-

nated food

Type specic - bacterial spores typicallyhave the highest resistance whereas veg-

etative bacteria have the lowest resistance

Good, 2 to 3 - logremoval

Viruses 0.01 - 0.1 Nonmotile Humans and animals,

polluted water, and

contaminated food

Generally more resistant than vegetative

bacteria

Poor, 1 to 3 - log

removal

Protozoa 1 - 20 Motile,

Nonmotile

Humans and animals,

sewage, decaying

vegetation, and water

More resistant than viruses or vegetative

bacteria

Good, 2 to 3 - log

removal

Table 4 - Attributes of the three waterborne pathogens of concern in water treatment

120




agents never before associated with documented

waterborne outbreaks have appeared in the drink-

ing water industry.

Enteropathogenic E. coli and Giardia lamblia were

rst identied as the etiological agents responsible

for waterborne outbreaks in the 1960s.

The rst recorded Cryptosporidium infection in hu-

mans occurred in the mid-1970s. Also during that

time was the rst recorded outbreak of pneumonia

caused by Legionella pneumophila (Centers for

Disease Control, 1989; Witherell et al., 1988).

Recently, there have been numerous documented

waterborne disease outbreaks that have beencaused by E. coli, Giardia lamblia, Cryptospo-

ridium, and Legionella pneumophila.

E-coli

The rst documented case of waterborne disease

outbreaks associated with enteropathogenic E. coli

occurred in the 1960s in the United States.

Various serotypes of E. coli have been implicated

as the etiological agent responsible for disease in

newborn infants, usually the result of cross-con-

tamination in nurseries.

Now, there have been several well-documented

outbreaks of E. coli (serotypes 0111:B4 and 0124:

B27) associated with adult waterborne disease

(AWWA, 1990, and Craun, 1981).

In 1975, the etiologic agent of a large outbreak at

Crater Lake National Park was E. coli serotype 06:

H16 (Craun, 1981).

Giardia lamblia

Similar to E. coli , Giardia lamblia was rst identi-ed in the 1960s to be associated with waterborne

outbreaks in the United States.

Giardia lamblia is a agellated protozoan that is re-

sponsible for Giardiasis, a disease that can range

from being mildly to extremely debilitating.

Giardia is currently one of the most commonly

identied pathogens responsible for waterborne

disease outbreaks.

The life cycle of Giardia includes a cyst stage when

the organism remains dormant and is extremely

resilient (i.e., the cyst can survive some extreme

environmental conditions).

Once ingested by a warm-blooded animal, the life

cycle of Giardia continues with excystation.

The cysts are relatively large (8-14 µm) and can

be removed effectively by ltration using diatoma-

ceous earth, granular media, or membranes.

Giardiasis can be acquired by ingesting viable

cysts from food or water or by direct contact with

fecal material.

In addition to humans, wild and domestic animals

have been implicated as hosts.

Between 1972 and 1981, 50 waterborne outbreaks

of Giardiasis occurred with about 20,000 reportedcases (Craun and Jakubowski, 1986).

Currently, no simple and reliable method exists to

assay Giardia cysts in water samples.

Microscopic methods for detection and enumera-

tion are tedious and require examiner skill and

patience. Giardia cysts are relatively resistant to

chlorine, especially at higher pH levels and low

temperatures.

Cryptosporidium

Cryptosporidium is a protozoan similar to Giardia.

It forms resilient oocysts as part of its life cycle.

The oocysts are smaller than Giardia cysts, typi-

cally about 4-6 µm in diameter. These oocysts can

survive under adverse conditions until ingested by

a warm-blooded animal, and then continue with

excystation.

Due to the increase in the number of outbreaks

of Cryptosporidiosis, a tremendous amount of

research has focused on Cryptosporidium withinthe last 10 years.

Medical interest has increased because of its oc-

currence as a life-threatening infection to individu-

als with depressed immune systems.

As previously mentioned, in 1993, the largest doc-

umented waterborne disease outbreak in United

States history occurred in Milwaukee and was

determined to be caused by Cryptosporidium.

An estimated 403,000 people became ill, 4,400

people were hospitalized, and 100 people died.

WATER TREATMENT

121

DISINFECTION



The outbreak was associated with deterioration

of the raw water quality and a simultaneous de-

crease in effectiveness of the coagulation-ltration

process, which led to an increase in the turbidity

of treated water and the inadequate removal of

Cryptosporidium oocysts.

Legionella pneumophila

An outbreak of pneumonia occurred in 1976 at the

annual convention of the Pennsylvania American

Legion. A total of 221 people were affected by the

outbreak, and 35 of those aficted died.

The cause of the pneumonia was not determined

immediately, despite an intense investigation by

the Centers for Disease Control. Six months afterthe incident, microbiologists were able to isolate

a bacterium from the autopsy lung tissue of one

of the Legionnaires.

The bacterium responsible for the outbreak was

found to be distinct from other known bacterium

and was named Legionella pneumophila (Witherell

et al., 1988).

Following the discovery of this organism, other

Legionella-like organisms were discovered. All

together, 26 species of Legionella have been

identied, and seven are etiologic agents for Le-

gionnaires’ disease (AWWA, 1990).

Legionnaires’ disease does not appear to be trans-

ferred from person-to-person. Epidemiological

studies have shown that the disease enters the

body through the respiratory system.

Legionella can be inhaled via water particles less

than 5µm in size from facilities such as cooling tow-

ers, hospital hot water systems, and recreational

whirlpools (Witherell et al., 1988).

2.4 Mechanisms of pathogen inactiva-

tion

The three primary mechanisms of pathogen inac-

tivation are to:

- destroy or impair cellular structural organiza-

tion by attacking major cell constituents, such

as destroying the cell wall or impairing the

functions of semi-permeable membranes

- interfere with energy-yielding metabolism

through enzyme substrates in combination with

prosthetic groups of enzymes, thus rendering

the enzymes non-functional

- interfere with biosynthesis and growth by pre-

venting synthesis of normal proteins, nucleic

acids, coenzymes, or the cell wall.

Depending on the disinfectant and microorganism

type, combinations of these mechanisms can also

be responsible for pathogen inactivation.

In water treatment, it is believed that the primary

factors controlling disinfection efciency are:

(1) the ability of the disinfectant to oxidize or rup-

ture the cell wall.

(2) the ability of the dis infectant to di ffuseinto the cell and interfere with cellular activ-

ity (Montgomery, 1985).

2.5 Other uses of disinfectants in water

treatment

Disinfectants are used for more than just disinfec-

tion in drinking water treatment.

While inactivation of pathogenic organisms is a

primary function, disinfectants are also used as

oxidants in drinking water treatment for several

other functions:

- control of nuisance Asiatic clams and zebra

mussels

- prevention of algal growth in sedimentation

basins and lters

- removal of taste and odors through chemi-

cal oxidation

- improvement of coagulation and ltration ef -

ciency

- oxidation of iron and manganese- removal of color

- prevention of regrowth in the distribution sys-

tem and maintenance of biological stability.

2.6 Current practice of disinfection (and

oxidation)

USA

In the USA, most water treatment plants disinfect

water prior to distribution.

122




The 1995 Community Water Systems Survey

(USEPA, 1997a) reported that 81 percent of all

community water systems provide some form

of treatment on all or a portion of their water

sources.

The survey also found that virtually all surface

water systems provide some treatment of their

water.

Of those systems reporting no treatment, 80

percent rely on groundwater as their only water

source.

The most commonly used disinfectants/oxidants

are chlorine, chlorine dioxide, chloramines, ozone,

and potassium permanganate.Table 5 displays a breakdown of the chemical

usage based on the survey’s data. Note that the

table shows the percentages of systems using the

particular chemical as a disinfectant or in some

other role. The table shows the predominance of

chlorine in surface and groundwater disinfection

treatment systems with more than 60 percent of

the treatment systems using chlorine as a disin-

fectant/oxidant.

Potassium permanganate, on the other hand, is

used by many systems, but its application is pri-

marily for oxidation rather than for disinfection.

Permanganate will have some benecial impact

on disinfection since it is a strong oxidant that

will reduce the chemical demand for the ultimate

disinfection chemical.

Chloramine is used by some systems and is more

frequently used as a post-treatment disinfectant.

In the USA, the most common uses for ozone are

for oxidation of iron and manganese and for taste

and odor control.

Twenty-four of the 158 ozone facilities used GAC

following ozonation.

In addition to 158 operating ozone facilities in the

USA in 1997, 19 facilities were under constructionand another 30 under design.

In May 1998, 264 drinking water plants in the

United States were using ozone.

Europe

In the Netherlands, as well as in most other West-

ern European countries, the practice regarding

disinfection and oxidation is completely different

from what happens in the USA.

In Europe, disinfection of groundwater is seldom

applied. The water is abstracted by hygienic

means (closed wells, etc.), and the treatment and

storage facilities are covered and protected. Oxida-

tion of iron, ammonia and manganese is, in nearly

every case, performed by oxygen (after aeration)

instead of by chemical oxidants.

Since 2006, chlorination is no longer applied to

surface water in the Netherlands, as mandated

by the drinking water regulations. For primary

disinfection in direct treatment systems (without

inltration or river bank inltration), UV is used,either by itself or in combination with peroxide.

Sometimes, ozone is used.

Whenever post-disinfection occurs, in most cases

chlorine dioxide is applied.

Gaseous chlorine is rarely used in Western Eu-

rope, in keeping with safety regulations.

2.7 Disinfection byproducts

Table 6 is a list of disinfection residuals and dis-

Treatment Ground-

water

Surface

water

Number of systems 31,579 3,347

Pre-disinfection 1% 4%

Primary disinfection/oxidation 66% 90%Chlorine 64% 64%

Chlorine dioxide 0% 6%

Chloramines 0% 3%

Ozone 0% 1%

KMnO4

2% 16%

Post-disinfection 32% 80%

Chlorine 31% 68%

Chlorine dioxide 0% 2%

Chloramines 0% 8%

Post-disinfection combinations 0% 3%

Table 5 - Disinfection practice (USA)

WATER TREATMENT

123

DISINFECTION



infection byproducts (DBP) that may be of health

concern.

Formation of DBPs

Halogenated organic byproducts are formed when

natural organic matter (NOM) reacts with free

chlorine or free bromine.

Free chlorine can be introduced to water directly as

a primary or secondary disinfectant, with chlorine

dioxide, or with chloramines.

Free bromine results from the oxidation of the

bromide ion in source water.

Factors affecting formation of halogenated DBPs

include the type and concentration of natural or-

ganic matter, oxidant type and dose, time, bromide

ion concentration, pH, organic nitrogen concentra-

tion, and temperature.

Organic nitrogen significantly influences the

formation of nitrogen containing DBPs such as

the haloacetonitriles, halopicrins, and cyanogen

halides (Reckhow et al., 1990; Hoigné and Bader,

1988).

The parameter TOX represents the concentration

of total organic halides in a water sample (calcu-

lated as chloride). In general, less than 50 percent

of the TOX content has been identied, despite evi-

dence that several of these unknown halogenated

byproducts of water chlorination may be harmful

to humans (Singer and Chang, 1989).

Non-halogenated DBPs are also formed when

strong oxidants react with organic compounds

found in water.

Ozone and peroxone oxidation of organics leads to

the production of aldehydes, aldo- and keto-acids,

organic acids, and, when bromide ion is present,

brominated organics (Singer, 1992).

Many oxidation byproducts are biodegradable and

appear as biodegradable dissolved organic carbon

(BDOC) and assimilable organic carbon (AOC) in

treated water.

Bromide ion plays a key role in DBP formation.

Ozone or free chlorine oxidizes bromide ion to

hypobromate ion/hypobromous acid, which sub-

sequently forms brominated DBPs.

Brominated organic byproducts include com-

pounds such as bromoform, brominated acetic

acids and acetonitriles, bromopicrin, and cyanogen

bromide. Only about one third of the bromide ionsincorporated into byproducts has been identied.

DBP precursors

Numerous researchers have documented that

NOM is the principal precursor of organic DBP

formation (Stevens et al., 1976; Babcock and

Singer 1979; Christman et al., 1983).

Chlorine reacts with NOM to produce a variety of

DBPs, including THMs, haloacetic acids (HAAs),

and others.

Ozone reacts with NOM to produce aldehydes,

Chemical Carcinogen

Disinfection residuals

Free chlorine

Monochloramine

(Ammonia)

Hydrogen peroxide

Chlorine peroxide

Inorganic byproducts

Chlorate

Chlorite

Bromate +

Iodate

Organic oxidation byproducts

Aldehydes +

Carboxylic acids

Assimilable Organic Carbon (AOC)

Nitrosoamines

Halogenated organic byproducts +

Trihalomethanes (THM) +

Haloacetic acids (HAA) ?

Haloacetonitriles

Haloketones +

Chlorophenols

Chloropicrin ?

Chloral hydrate

Cyanogen chloride

N-Organochloramines

MX

Table 6 - Chemicals with health risks related to dis-

infection

124




organic acids, and aldo- and keto-acids; many of

these are produced by chlorine as well (Singer

and Harrington, 1993).

Natural waters contain mixtures of both humic

and nonhumic organic substances. NOM can be

subdivided into a hydrophobic fraction composed

of primarily humic material, and a hydrophilic frac-

tion composed of primarily fulvic material.

The type and concentration of NOM are often as-

sessed using surrogate measures.

Although surrogate parameters have limitations,

they are used because they may be measured

more easily, rapidly, and inexpensively than the

parameter of interest, often allowing on-line moni-toring of the operation and performance of water

treatment plants.

Surrogates used to assess NOM include:

- Total and dissolved organic carbon (TOC and

DOC)

- Specic ultraviolet light absorbance (SUVA),

which is the absorbance at 254 nm wavelength

(UV-254) divided by DOC (SUVA = (UV-254/

DOC)*100, in L/mg-m)

- THM formation potential (THMFP) -- a test

measuring the quantity of THMs formed with a

high dosage of free chlorine and a long reaction

time

- TTHM Simulated Distribution System (SDS)-- a

test to predict the TTHM concentration at some

selected point in a given distribution system,

where the conditions of the chlorination test

simulate the distribution system at the point

desired.

On average, about 90 percent of the TOC is dis-solved.

DOC is dened as the TOC able to pass through

a 0.45 µm lter.

UV absorbance is a good technique for assessing

the presence of DOC because DOC primarily con-

sists of humic substances, which contain aromatic

structures that absorb light in the UV spectrum.

Oxidation of DOC reduces the UV absorbance of

the water due to oxidation of some of the organic

bonds that absorb UV absorbance.

Complete mineralization of organic compounds

to carbon dioxide usually does not occur under

water treatment conditions; therefore, the overall

TOC concentration is usually constant.

3 Disinfection kinetics

3.1 Chick’s Law

In 1908 Ms. Harriet Chick found that her disinfec-

tion experiments could best be described by a

rst-order reaction:

dNk N

dt= − ⋅

or:

oln(N/N ) k t= − ⋅

in which:

N = concentration of organism [- / m3]

NO = initial concentration of organism [- / m3]

t = time [s]

k = rate constant [1/s]

The rate constant k differs per disinfectant, dis-

infectant concentration, organism and tempera-

ture.

The rate of inactivation depends upon such factors

as the penetration of the cell wall, and the time

needed to penetrate vital centers. Each species

of microorganism, therefore, will have a different

sensitivity to each disinfectant.

According to this relationship, known as Chick’s

Law, you can achieve a doubling of the log-removal

by providing for a contact time twice as long,

assuming a constant disinfectant concentration(Figure 1).

It should be noted that Chick’s Law resembles the

formula for natural decay. Disinfection increases

the decay constant k.

A complete inactivation of the microorganism is

not feasible according to this model.

Efciency of disinfection

The efciency of disinfection is reported in terms

WATER TREATMENT

125

DISINFECTION





dose indicates a value n<1 (0.8-0.9). For higher

inactivation, the required C t -value is more than

the model assumes.

3.3 Other models

Alternative models have been developed over the

years to get better ts between model and data.

Rennecker-Mariñas model

For inactivation of oocysts and endospores, often

a certain lag concentration is observed. Below this

lag concentration of disinfectant, no disinfection

is obtained.

This phenomenon is incorporated in the Renneck-

er-Mariñas model, which uses a “net disinfectantconcentration” in the Chick-Watson model:

actual lagC C C= −

Collins-Selleck model

Collins and Selleck developed a model to describe

the inactivation of coliform organisms in waste-

water disinfection. They observed that increased

C t -values were required for very large inactivation

(log 4 to 6). This is probably due to the encapsula-

tion of a small part of these organisms, making it

less approachable for disinfectants.

Hom-Haas model

In the Hom-Haas model, Cp tq is used instead

of C t.

With this extension, a better t can be obtained, but

more empirical constants should be determined for

different conditions.

3.4 C t -values

In most cases the C t -value is used as the basis

for disinfection.

This approach is also used for disinfection with UV

radiation, for which the C t -value is modied into

UV light intensity (mW/cm2) multiplied by the time

of exposure (s), giving the dose (mJ/cm2).

For many pathogens and disinfectants, information

can be found on C t -values and inactivation.

The US EPA began the practice of specifying C t

-values that must be met as a way of regulating

the control of pathogens within the Surface Water

Treatment Regulations.

At present, they have published tables for criti-

cal pathogens (e.g. Giardia, Cryptosporidium,

viruses) for all relevant disinfection methods, and

different log inactivation credits, at different water

temperatures and pH.

An impression of the required C t -values for differ-

ent disinfectant methods is shown in Figure 4.

Notice that Cryptosporidium Parvum and Giardia

are difcult to inactivate with chemical disinfectants

(high C t required) and easy to inactivate with UVradiation (low I t).

The opposite is true for viruses.

The required C t -values for chemical disinfectants

show large variations (range 103 – 106).

For UV radiation, this variation is much smaller

(range 102).

Declining concentration

Chemical disinfectants are oxidants reacting with

components in the water. Therefore, the concen-

tration of the disinfectant declines in time.

Additionally, the disinfectant/oxidant might de-

compose. Because ozone naturally decomposes

so fast, this is an important consideration for the

disinfection process.

To calculate the disinfection credits based on C t

-values, reductions in the disinfectant concentra-

tion should be taken into account.

Temperature

At lower temperatures, disinfection requires higher

C t -values for the same inactivation.

At 1 to 5 oC, the required C t -value might be some

5 – 10 times higher than the C t -value at 25 oC.

Short-circuiting

The C t -values are based on batch lab experi-

ments in which the concentration and residence

time are controlled.

In practice, disinfection is applied in full-scale

WATER TREATMENT

127

DISINFECTION



contact tanks, having non-ideal residence times

(residence time distribution, short-circuiting).

The Chick-Watson model can be used to deter-

mine the effect of a non-ideal ow in a disinfection

reactor.

As an example we compare the disinfection ef-

ciency of a full plug ow reactor with a reactor inwhich half the ow has a residence time of 60%,

and the other half a residence time of 140%. As-

sume that the plug ow reactor has an inactivation

of log 4. The water at low residence time has an

inactivation of (4*0.6=) 2.4, while the water in the

high residence time has an inactivation of (4*1.4=)

5.6. The total inactivation is (-log(102.4+105.6)/2 =)

2.8.

This shows that short-circuiting has a substantial

negative effect on the efciency of disinfection,

particularly when a high inactivation is required.

The t10

concept

Because of the effect of short-circuiting, the deten-

tion time in the US EPA regulations are dened as

being the detention time in which 10% of the ow

has passed the contactor (t10).

In poorly designed contactors, this greatly reduces

the disinfection credits.

In order to improve these reactors, better plug

ow conditions can be achieved by proper ow

splitting, bafing, and/or by designing long and

narrow contactors.

Bypassing

Calculations can be made on the effect of bypass-

ing, which occurs, for instance, when part of the

ow does not receive any disinfectant. Bad mixing

can occur when there is an uneven distribution of

Figure 4 - Disinfection requirements for 99% inactivation (min mg/l or mJ/cm2 )

Combined chlorine Free chlorine Chlorine dioxide Ozone UV LightRequired

Ct or lt

0.01

0.10

1.0

10

100

1.000

10.000 C. Parvum

C. Parvum

C. Parvum

C. Parvum C. Parvum

Giardia

Legionella

Legionella

Legionella

Legionella

Mycobacterium

fortuitumM. fortuitum

M. fortuitum

M .

f o r t u i t u m

Poliovirus

Poliovirus

Poliovirus

Poliovirus

Poliovirus

Adenovirus

Adenovirus Adenovirus

E. coli

E. coli

E. coli

E. coli

Calicivirus

Calicivirus

Calicivirus

Microsporidium

Giardia Giardia

Giardia Giardia

Legionella

PneumophilaMicrosporidium

Microsporidium

Adenovirus

Calicivirus

E. Coli

Adenovirus

ReovirusMS-2

CalicivirusRotavirus

Hepatitis A

128




the ow.

If 1% of the ow is bypassing a log 4 disinfection

reactor, the overall efciency can be calculated

as being log 2.

(influent, and bypass 10,000 organisms, dis-

infected main stream 1 organism, overall ef-

fect 10,000*0.01+1*0.99=100.99 organisms or

log(100.99/10,000) = -1.996).

This example shows the dramatic reduction in

disinfection caused by bypassing, in particular at

a high disinfection requirement.

4 Disinfection methods

In the following section the advantages and dis-

advantages of different disinfection methods for

drinking water are presented.

Because of the wide variation of system sizes,

water quality, and dosages applied, some of these

advantages and disadvantages may not apply to

all systems.

4.1 Chlorine

Advantages

- Oxidizes soluble iron, manganese, and sul-

des

- Enhances color removal

- Enhances taste and odor

- May enhance coagulation and ltration of par -

ticulate contaminants

- Is an effective biocide- Is the easiest and least expensive disinfection

method, regardless of system size

- Is the most widely used disinfection method,

and, therefore, the best known

- Is available as calcium and sodium hypochlo-

rite. Use of these solutions is more advanta-

geous for smaller systems than chlorine gas

because they are easier to use, are safer, and

need less equipment compared to chlorine

gas

- Provides a residual

Disadvantages

- May cause a deterioration in coagulation/ltra-

tion of dissolved organic substances

- Forms halogen-substituted byproducts

- Finished water could have taste and odor prob-

lems, depending on water quality and dosage

- Chlorine gas is a hazardous corrosive gas

- Special leak containment and scrubber facilities

could be required for chlorine gas

- Typically, sodium and calcium hypochlorite are

more expensive than chlorine gas

- Sodium hypochlorite degrades over time and

with exposure to light

- Sodium hypochlorite is a corrosive chemical- Calcium hypochlorite must be stored in a cool,

dry place because of its reaction with moisture

and heat

- A precipitate may form in a calcium hypochlorite

solution because of impurities, therefore, an

antiscalant chemical may be needed

- Higher concentrations of hypochlorite solutions

are unstable and will produce chlorate as a

byproduct

- Is less effective at high pH

- Forms oxygenated byproducts that are biode-

gradable and which can enhance subsequent

biological growth if the chlorine residual is not

maintained.

- Release of constituents bound in the distribu-

tion system (e.g., arsenic) by changing the

redox state

Generation

Chlorination may be performed using chlorine gas

or other chlorinated compounds that may be inliquid or solid form.

Chlorine gas can be generated by a number of

processes including the electrolysis of alkaline

brine or hydrochloric acid, the reaction between

sodium chloride and nitric acid, or the oxidation of

hydrochloric acid.

Since chlorine is a stable compound, chlorine

gas, sodium hypochlorite, and calcium hypochlo-

rite are typically produced off-site by a chemical

manufacturer.

WATER TREATMENT

129

DISINFECTION



Primary uses

The primary use of chlorination is disinfection.

Chlorine also serves as an oxidizing agent for

taste and odor control, preventing algal growths,

maintaining clear lter media, removing iron and

manganese, destroying hydrogen sulde, remov-

ing color, maintaining the water quality at the dis-

tribution systems, and improving coagulation.

Inactivation efciency

The general order of increasing chlorine disin-

fection difculty is bacteria, viruses, and then

protozoa.

Chlorine is an extremely effective disinfectant for

inactivating bacteria and a highly effective viricide.However, chlorine is less effective against Giardia

cysts. Cryptosporidium oocysts are highly resis-

tant to chlorine.

Byproduct formation

When added to the water, free chlorine reacts with

NOM and bromide to form DBPs, primarily THMs,

some haloacetic acids (HAAs), and others.

Point of application

Chlorine can be applied at different points: in the

raw water storage, pre-coagulation/post-raw wa-

ter storage, pre-sedimentation/ post-coagulation,

post-sedimentation/pre-filtration, post-filtration

(disinfection), or in the distribution system.

Special considerations

Because chlorine is such a strong oxidant and

extremely corrosive, special storage and handling

considerations should be considered in the plan-

ning of a water treatment plant. Additionally, health concerns associated with

the handling and use of chlorine is an important

consideration.

4.2 Ozone

Advantages

- Ozone is more effective than chlorine, chlora-

mines, and chlorine dioxide for inactivation of

viruses, Cryptosporidium, and Giardia.

- Ozone oxidizes iron, manganese, and sul-

des.

- Ozone can sometimes enhance the clarication

process and turbidity removal.

- Ozone controls color, taste, and odors.

- One of the most efcient chemical disinfectants,

ozone requires a very short contact time.

- In the absence of bromide, halogen-substitutes

DBPs are not formed.

- Upon decomposition, the only residual is dis-

solved oxygen.

- Biocidal activity is not inuenced by pH.

Disadvantages

- DBPs are formed, particularly by bromate andbromine-substituted DBPs, in the presence of

bromide, aldehydes, ketones, etc.

- The initial cost of ozonation equipment is

high.

- The generation of ozone requires high energy

and should be generated on-site.

- Ozone is highly corrosive and toxic.

- Biologically activated lters are needed for

removing assimilable organic carbon and bio-

degradable DBPs.

- Ozone decays rapidly at high pH and warm

temperatures.

- Ozone provides no residual.

- Ozone requires higher level of maintenance

and operator skill.

Generation

Because of its instability, ozone should be gener-

ated at the point of use.

Ozone can be generated from oxygen present

in air or high purity oxygen. The feed gas sourceshould be clean and dry, with a maximum dew

point of -60 0C.

Ozone generation consumes power at a rate of 8

to 7 kWh/kg O3. On-site generation saves a lot

of storage space.

Primary uses

Primary uses include primary disinfection and

chemical oxidation. As an oxidizing agent, ozone

can be used to increase the biodegradability

of organic compounds destroys taste and odor

130




control, and reduce levels of chlorination DBP

precursors.

Ozone should not be used for secondary dis-

infection because it is highly reactive and does

not maintain an appreciable residual level for the

length of time desired in the distribution system.


Ozone is one of the most potent and effective

germicide used in water treatment. It is effective

against bacteria, viruses, and protozoan cysts. In-

activation efciency for bacteria and viruses is not

affected by pH; at pH levels between 6 and 9.

As water temperature increases, ozone disinfec-

tion efciency increases.

Byproduct formation

Ozone itself does not form halogenated DBPs;

however, if bromide ion is present in the raw water

or if chlorine is added as a secondary disinfectant,

halogenated DBPs, including bromate ion may be

formed.

Other ozonation byproducts include organic acids

and aldehydes.

Limitations

Ozone generation is a relatively complex process.

Storage of LOX (if oxygen is to be the feed gas) is

subject to building and re codes.

Points of application

For primary disinfection, ozone addition should be

prior to bioltration/ltration and after sedimenta-

tion.

For oxidation, ozone addition can be prior to co-

agulation/sedimentation or ltration depending onthe constituents to be oxidized.

Safety considerations

Ozone is a toxic gas and the ozone production and

application facilities should be designed to gener-

ate, apply, and control this gas, so as to protect

plant personnel. Ambient ozone levels in plant

facilities should be monitored continuously.

4.3 UV radiation

Generation

Low pressure and medium pressure UV lamps

are available.

Primary uses

Primary physical disinfectant; requires secondary

chemical disinfectant for residual in distribution

system.


This method is very effective against bacteria

and viruses at low dosages (5-25 mW•s/cm2 for

2-log removal and 90-140 mW•s/cm2 for 4-log

removal).

Much higher dosage required for Cryptosporidiumand Giardia (100-8,000 mW•s/cm2 for 2-log re-

moval)

Byproduct formation

Minimal disinfection byproducts produced.

Limitations

Limited experience and data with large ows.

Water with high concentrations of iron, calcium,

turbidity, and phenols may not be applicable to

UV disinfection.


It is preferable to apply UV radiation prior to the

distribution system.


Extremely high UV dosages for Cryptosporidium

and Giardia may make surface water treatment

impractical.

4.4 Chlorine dioxide

Advantages

- Chlorine dioxide is more effective than chlorine

and chloramines for inactivation of viruses,

Cryptosporidium, and Giardia.

- Chlorine dioxide oxidizes iron, manganese, and

suldes.

- Chlorine dioxide may enhance the clarication

process.

WATER TREATMENT

131

DISINFECTION



- Taste and odors resulting from algae and

decaying vegetation, as well as phenolic com-

pounds, are controlled by chlorine dioxide.

- Under proper generation conditions (i.e., no

excess chlorine), halogen-substituted DBPs

are not formed.

- Chlorine dioxide is easy to generate.

- Biocidal properties are not inuenced by pH.

- Chlorine dioxide provides residuals.

Disadvantages

- The chlorine dioxide process forms the specic

byproducts chlorite and chlorate.

- Generator efciency and optimization difculty

can cause excess chlorine to be fed at theapplication point, which can potentially form

halogen-substitute DBPs.

- Costs associated with training, sampling, and

laboratory testing for chlorite and chlorate are

high.

- Equipment is typically rented, and the cost of

the sodium chlorite is high.

- Measuring chlorine dioxide gas is explosive,

so it must be generated on-site.

- Chlorine dioxide gas is explosive, so it must be

generated and measured on-site.

- Chlorine dioxide decomposes in sunlight.

- Can lead to production noxious odors in some

systems.

Generation

Chlorine dioxide must be generated on-site. In

most potable water applications, chlorine dioxide

is generated as needed and directly educed from

or injected into a diluting stream.

Generators are available that utilize sodium chlo-rite and a variety of feedstocks such as Cl2 gas,

sodium hypochlorite, and sulfuric or hydrochloric

acid.

Small samples of generated solutions, up to 1 per-

cent (10 g/l) chlorine dioxide can be safely stored

if the solution is protected from light, chilled (<5oC), and has no unventilated headspace.

Primary uses

Chlorine dioxide is utilized as a primary or sec-

ondary disinfectant for taste and odor control,

TTHM/HAA reduction, Fe and Mn control, color

removal, sulde and phenol destruction, and Zebra

mussel control.


Chlorine dioxide rapidly inactivates most microor-

ganisms over a wide pH range. It is more effective

than chlorine (for pathogens other than viruses)

and is not pH dependent between pH 5-10, but is

less effective than ozone.

Byproducts formation

When added to water, chlorine dioxide reacts

with many organic and inorganic compounds.

The reactions produce chlorite and chlorate as

end-products (compounds that are suspected ofcausing hemolytic anemia and other health ef-

fects). Chlorate ion is formed predominantly in

downstream reactions between residual chlorite

and free chlorine when used as the distribution

system disinfectant.

Chlorine dioxide does not produce THMs. The use

of chlorine dioxide aids in reducing the formation

of TTHMs and HAAs by oxidizing precursors, and

by allowing the point of chlorination to be moved

farther downstream in the plant after coagulation,

sedimentation, and ltration have reduced the

quantity of NOM.


In conventional treatment plants, chlorine dioxide

used for oxidation is fed either in the raw water

or in the sedimentation basins, or following sedi-

mentation.

To limit the oxidant demand, and therefore the

chlorine dioxide dose and formation of chlorite,

it is common to add chlorine dioxide followingsedimentation.

Concerns about possible taste and odor com-

plaints have limited the use of chlorine dioxide to

provide a disinfectant residual in the distribution

system. Consequently, public water suppliers

who are considering the use of chlorine dioxide

for oxidation and primary disinfectant applications

may want to consider chloramines for secondary

disinfection.


132




An oxidant demand study should be completed

to determine an approximate chlorine dioxide

dosage to obtain the required C t -value as a

disinfectant.

In addition to the toxic effects of chlorine, chlorine

dioxide gas is explosive at levels > 10% in air. The

chlorine dioxide dosage cannot exceed 1.4 mg/l

so as to limit the total combined concentration of

ClO2, ClO2-, ClO3

-, to a maximum of 1.0 mg/l.

Under the proposed DBP regulations, the MRDL

for chlorine dioxide is 0.8 mg/l and the MCL for

chlorite is 1.0 mg/l. Regulations concerning the

use of chlorine dioxide vary from state to state.

4.5 Other methods

Alternative disinfection methods are used during

large scale water treatment for drinking water

production:

- Hydrogen peroxide / Ozone (Peroxone)

- Hydrogen peroxide / UV

- Potassium permanganate

- Chloramines.

For a description of these systems, reference is

made to the literature.

Further reading


(2005), (ISBN 0 471 11018 3) (1948 pgs)

• Unit processes in drinking water treatment, W.

Masschelein 1992 (ISBN 0 8247 8678 5) (635

pgs)

• Water quality and treatment, AWWA 1999

(ISBN 0 07 001659 3) (1233 pgs)

• Water treatment and pathogen control, WHO

2004 (ISBN 92 4 156255 2) (139 pgs)

• Assessing microbial safety of drinking water,

WHO 2003 (ISBN 1 84339 036 1) (297 pgs)

• Water disinfection, CEPIS-PAHO/WHO 2003

(208 pgs) (for small water systems)

WATER TREATMENT

133

DISINFECTION



134




Aeration and

gas stripping

WA T

E R T R E A T M E

N T

WATER TREATMENT

Qw, cw,0

Qa,ca,0

Qa, ca,e

Qw, cw,e

k5k4 k3

k2

k1

0.001 0.01 0.1 1 10

1

0.8

0.6

0.4

0.2

0

k1

k2

k3

k4

k5

RQ

k 2tk DT

K

( - )

===

1.610.03910oC



Framework

This module explains aeration and gas stripping.

Contents


1. Introduction

2. Theory of gas transfer

2.1 Equilibrium

2.2 Kinetics

2.3 Mass balance


3. Practice

3.1 Cascade

3.2 Tower aerator 3.3 Plate aerator

3.4 Spray aerator

3.5 Alternative aeration systems

AERATION AND GAS TRANSFER WATER TREATMENT

136



the water falls over a weir into a lower placed

trough. When the falling stream enters the water

body, air is entrapped in the form of bubbles, pro-

viding for a mixture of water and air in which gastransfer will occur.

The tower aerator (Figure 2) consists of a cylin-

drical vessel of steel or synthetic material that is

lled with packing material, usually consisting of

elements of synthetic material. Water falls down

and air is blown in a co-current or counter-current

direction.

A plate aerator (Figure 3) is a horizontal perfo-

rated plate. Water ows over the plate and air is

blown through the orices, creating a bubble bed

of air and water above the plate.

Sprayers (Figure 4) are typically used because

of their simple implementation in existing treat-

ment plants. By spraying, a contact surface be-

tween the air and water is created for the gas

exchange.

2 Theory of gas transfer

1 Introduction

Aeration (gas addition) and gas stripping (gas re-

moval) are normally the rst treatment steps dur -

ing the production of drinking water from ground-

water or riverbank water. This articially induced

gas transfer aims at the addition of oxygen (O2)

and the removal of carbon dioxide (CO2), meth-

ane (CH4), hydrogen sulde (H2S), and other

volatile organic compounds (for example 1.2 Di-

chloropropane (1.2 DCP), Trichloroethene (TRI),

Tetrachloroethene (PER) and Trichloromethane

(chloroform)).

Gas transfer is seldom applied in the treatment ofsurface water because surface water has been

in contact with air for a prolonged period. Conse-

quently, surface water contains sufcient oxygen,

and other gases, like methane and hydrogen sul-

de, are absent.

The addition of oxygen is required for the oxida-

tion of bivalent iron (Fe2+), manganese (Mn2+)

and ammonium (NH4+). These substances are

present in dissolved form in groundwater. Due to

chemical and biological oxidation, the substan-

ces can be removed by following a ltration step.

This will be discussed in the chapter on granular

ltration.

Reducing the carbon dioxide concentration leads

to a rise in pH and a reduction of aggressive car-

bon dioxide that is able to disintegrate (concrete)

pipes.

Methane should be removed because its pres-

ence has negative inuences on the ltration pro-

cesses.

Hydrogen sulde has an annoying odor (rottingeggs) and therefore needs to be removed from

the water.

Volatile organic compounds are usually toxic;

some of them are even carcinogenic. Obviously,

these compounds are not allowed in drinking wa-

ter.

To achieve gas transfer a number of systems

have been developed over the years.

One of the oldest systems is the cascade (Figure

1). The water falls in several steps. In each step,

Figure 1 - Cascade aeration

Figure 2 - Tower aeration

WATER TREATMENT AERATION AND GAS TRANSFER

137



2.1 Equilibrium

Henry’s law

Water contains dissolved gases. In a closed ves-

sel containing both gas (e.g., air) and water, the

concentration of a volatile component in the gas-

phase will be in equilibrium with the concentra-

tion in the waterphase, according to Henry’s law.

The equilibrium concentration can be calculated

using the following form of Henry’s law:

= ⋅w H gc k c

in which:

cw = equilibrium concentration of a gas in water

[g/m3]

kH = Henry’s constant or distribution coefcient

[-]

cg = concentration of the gas in air [g/m3]

The distribution coefcient kH depends on the

type of gas, and the temperature.

In addition, pollution and impurities in the water

inuence the equilibrium concentration. This is-

sue will not be discussed here.

In literature, many different forms of Henry’s law

are found.

Often partial pressure is used in stead of the gas

concentration in air, and/or molar concentration in

the water in stead of weight concentration. Con-

sequently this results in a different unit for the dis-

tribution coefcient, or Henry’s law constant (ie.

[mol/(m3 Pa)] or [mol/l/atm]).

For gas stripping, often the volatility is given instead of the solubility of a gas. In this case, the

distribution coefcient is inverted (gas/water, in

stead of water/gas).

Distribution coefcient

In Table 1 for a number of gases a list of values

is given of the distribution coefcient at different

water temperatures, (intermediate values can be

obtained with linear interpolation).

In the table it is shown that nitrogen, oxygen and

methane have low kH-values. This means that

these gases hardly dissolve in water and they

can, therefore, be easily removed.

The other gases have high kH –values and dis-

solve easily, which makes it difcult to remove

them from the water or easy to add them to wa-

ter.

Gas concentration in air

The gas concentration in the air cg must be known

before the equilibrium (or saturation) concentra-tion can be calculated. This concentration can be

determined using the universal gas law:

p V n R T⋅ = ⋅ ⋅

in which:

p = partial pressure of gas in gas phase [Pa]

V = total gas volume [m3]

n = number of moles of a gas [mol]

R = universal gas constant = 8.3142 [J/(K.mol)]

T = (air) temperature [K]

Figure 4 - Spray aeration

Figure 3 - Plate aeration


138



The gas concentration can be calculated by multi-

plying the molar gas concentration in air [mol/m3]

with the molecule weight of the considered gas:

gn p

c MW MWv R T

= ⋅ = ⋅

⋅

in which:

MW = molecular weight of a gas [g/mol]

Partial pressure

The partial pressure of a certain gas is propor-

tional to the volume fraction of that gas in air:

o f p p V= ⋅

in which:

po = standard pressure at sea level (=101,325)

[Pa]

Vf = volume fraction [-]

In Table 2 the volume fractions of different gases

that occur in air are given.

These values are valid for dry air with a stand-

ard pressure of 101,325 Pa. With these volume

fractions the partial pressures of all gases in air

can be calculated. Gases that do not occur in air

have a partial pressure equal to zero and thus a

cg equal to zero and also a cw equal to zero (for

example, methane).

In Figure 5 the equilibrium (saturation) concentra-

tion of oxygen is given as a function of water tem-

perature. With an increase in water temperature,

the saturation concentration decreases because

less oxygen can be dissolved in warm water.

The saturation concentration cw is linearly depen-

dent on pressure. The saturation concentration

for oxygen at the standard pressure of 101,325

Pa is 11.3 g/m3.

At a height of 8,000 meters (for example, Mount

Everest), the air pressure is only 10,000 Pa which

means that the saturation concentration for oxy-

gen is 1.1 g/m3.

In the sea at a depth of 100 meters below sea

level, the pressure is 1,100,000 Pa. This results

in a saturation concentration for oxygen of 113

g/m3.

2.2 Kinetics

As soon as water and air are in contact, gas

Gas

Distribution coefcient (kH)

T = 0 oC T = 10 oC T = 20 oCMolecular weight

(MW) [g/mol]

Nitrogen (N2) 0.023 0.019 0.016 28

Oxygen (O2) 0.049 0.041 0.033 32

Methane (CH4) 0.055 0.043 0.034 16

Carbon dioxide (CO2) 1.71 1.23 0.942 44

Hydrogen sulde (H2S) 4.69 3.65 2.87 34

Tetrachloroethelene (C2HCl

4) -1 3.20 1.21 167

Tetrachloroethene (C2HCl

3) -1 3.90 2.43 131.5

Chloroform (CHCl3) -1 9.0 7.87 119.5

Ammonia (NH3) 5000 2900 1800 17

1 These substances are still in the liquid phase at a temperature of 00C and therefore the kH is not known

Table 1 - Distribution coefcient for gases and the molecule weight

Gas

Volume

fraction1

[%]

Saturation

concentration2

cw

[g/m3]

Nitrogen (N2) 78.084 17.9

Oxygen (O2) 20.948 11.3

Argon (Ar) 0.934 -

Carbon dioxide (CO2) 0.032 0.79

Other gases 0.02 -

1 In dry air at a standard pressure of 101,325 Pa

2 Water and air temperature of 10 0C

Table 2 - Volume fractions of gases


139



molecules will be exchanged continuously. The

direction of the net gas transport depends on thegas concentration in the water (cw) and the equi-

librium concentration ce.

In Figure 6 the gas concentration in the water at

time t=0 is smaller than the equilibrium concen-

tration. This means that more gas can be dis-

solved in the water than is present at time t=0.

A net gas transport from air to water occurs, as

indicated by the arrow in the gure. The net gas

transport continues until time t=innite and the

gas concentration in the water is equal to the

equilibrium (or saturation) concentration. Then,

the gas transport from water to air and vice versa

are equal. Hence, no net gas transport occurs

and the gas concentration in the water and air do

not change. In that case, a dynamic equilibrium

is established.

The velocity of gas transfer is determined by the

kinetic equation:

w2 s w

dck (c c )

dt= ⋅ −

in which:

cw = concentration of a gas in water [g/m3]

k2 = gas transfer coefcient [s-1]

The time-dependent gas concentration change

in water is represented by the term dcw/dt. The

changes in concentration are determined by the

magnitude of the gas transfer coefcient k2 andthe driving force (cs – cw).

The gas transfer coefcient k2 is a device-depen-

dent parameter. The larger the contact surface

area between the air and water and the renewal

of this surface area, the better the gas transfer

and the higher the gas transfer coefcient.

The driving force is dened by the amount of gas

that can maximally be dissolved in a volume of

water, the saturation concentration cs, and the

amount of gas that is present in a volume of wa-

ter, the concentration cw. The larger the driving

force, the faster the gas transfer.

The increase in the oxygen concentration in time

is shown in Figure 7 for a constant cs (10 mg/l)

and an initial oxygen concentration of 0 mg/l. In

the beginning, when the difference between the

cs and the cw is the largest, the gas transfer oc-

curs at maximum velocity. As time passes, the

gas concentration in water increases and thedriving force decreases, which gradually results

in a lower gas transfer rate. For t=innite the oxy-

gen concentration in water equals the saturation

concentration cs.

For a batch reactor the differential equation can

be solved by integration, with cw=cw,0 at time t=0,

taking into account that cs is constant:

2( k t)

w s s w,0

c c (c c ) e − ⋅= − − ⋅

water temperature (oC)

c w

( g / m 3 )

15.0

12.5

10.0

7.5

5.0

2.5

00 5 10 15 20 25 30 35

air: 21% oxygenpressure: 101325 Pa

Figure 5 - Saturation concentration of oxygen as a

function of the water temperature

cg

air interface water

t=infinite

c1

c0

t=2

t=1

t=0

c o n c e n t r a t i o n

Figure 6 - Gas transport from air to water


140



or:

2( k t)s w

s w,0

c ce

c c

− ⋅−

=

−

2.3 Mass balance

In the preceding paragraph it is assumed that the

oxygen concentration in air is constant. This is

a simplication that is not always applicable. For

situations in which the gas concentration chang-

es in air are important, a mass balance needs to

be formulated.

In Figure 8 a mass balance for a gas transfer sys-

tem is schematically presented.

A water ow (Qw), with a gas concentration in the

water phase (cw,0), and an airow (Qa), with a

gas concentration (ca,0), enter the system. The

same water ow (Qw), with a gas concentration

in the water phase (cw,e), and the same airow(Qa), with a gas concentration (ca,e), leave the

system.

For the gas transfer system, the law of continuity

is valid: the total amount of gas that enters and

leaves the system must be equal and a mass bal-

ance can be set up:

w w,0 a a,0 w w,e a a,eQ c Q c Q c Q c⋅ + ⋅ = ⋅ + ⋅

By using the mass balance, the gas concentra-

tions in the air and water are linked and can be

applied in the gas transfer equations presented

below.

The RQ is the relationship between the airow

and the water ow. Using the mass balance RQ,that relationship can be dened as follows:

w,e w,0a

w a,0 a,e

c cQRQ

Q c c

−

= =

−


For gas transfer systems three equations are de-

rived:

- equilibrium equation

- kinetic equation

- mass balance

With these equations it is possible to calculate

the changes in the gas concentrations in water

and air.

Combining the equilibrium equation and the

mass balance results in two equations with two

unknown variables, cw and ca. With different ini-tial conditions, different solutions for these equa-

tions can be obtained.

In the following section a number of equations

are presented that form the basis for the calcula-

tion of gas concentrations in water for different

gas transfer systems.

If the variation in the gas concentration in the air

cannot be neglected, the mass balance needs

to be taken into account. The efciency of a gas

transfer system can be calculated by dividing the

Qw, cw,0

Qa,ca,0

Qa, ca,e

Qw, cw,e

Figure 8 - Gas transfer system with in- and outow of

water and air

0

4

8

12

0 200 400 600 800 1000 1200 1400

time (s)

c o n c e n t r a t i o n O 2

( g / m 3 )

saturation concentration

driving force

co = 0 g/m3

cs = 10 g/m3

k 2 = 0.00193 s-1

Figure 7 - Oxygen concentration in water as a func-

tion of contact time


141



realized gas transfer by the maximum achievable

gas transfer:

w,e w,0

s w,0

c cK

c c

−

=

−

The following basic systems can be distin-

guished:

- plug ow with a constant gas concentration in

air

- complete mixed system with a constant gas

concentration in air

- plug ow, co-current ow and a variable gas


- plug ow, counter-current ow and a variablegas concentration in air

- complete mixed system with a variable gas


Plug ow with a constant gas concentration

in air

A characteristic of a plug ow is that the water is

supposed to ow as a “frozen volume” through

the gas transfer system. Thus, all water particles

in the system will have the same retention time.

The efciency equation, then, can be written into

the following equation:

− ⋅= − 2( k t )

K 1 e

An example of a plug ow where the gas con-

centration in air and thus cs is supposed to be

constant is a falling droplet from a spray aerator

into a large open space. The change in the gas

concentration in air as a result of gas transfer can

then be neglected.

Complete mixed system with a constant gas

concentration in the air

The opposite of a plug ow is a complete mixed

system. In such a gas transfer system the water

drops are mixed extensively. Consequently, the

retention time of the water drops is variable. Some

water drops leave the system directly (short-cir-

cuit ow) and others stay for a prolonged period

of time in the system (eddy formation). The ef-

ciency is calculated with:

⋅

=

+

2

2 1

k t

1K

1

Plug ow, co-current ow and a variable gas


The equation for co-current ow can be found

with the following initial conditions:

cw = cw,0 at time t=0;

ca = ca,0 at time t=0

The following solution can be derived:

d2

d

k ( k t( 1 ) )

RQ

3 k

RQ

1 eK

1

− ⋅ +

−

=

+

Plug ow with counter-current ow and vari-

able gas concentration in the air

The equation for counter-current ow can be

found with the following initial conditions:

cw = cw,e at time t=te ;

ca = ca,e at time t=te.


− ⋅ ⋅ −

− ⋅ ⋅ −

−=

− ⋅

d2

d2

d

k( k t (1 ))

RQ

4 k( k t (1 ))k RQ

RQ

1 eK

1 e

System RQ

Application

drinking

water

Application

wastewater

Cascade 0.4 O2, CH

4-

Tower aerator 5-100 CO2

CHCl3

Plate aerator 20-60 CH4, CO

2, O

2-

Spray aerator 0.5 O2, CO

2-

Deep well aerator 0.1-0.4 O2

O2

Cone aerator >5 - O2

Table 3 - Air/water ratio for different gas transfer sys-tems and the gases that can be removed

by the system


142



Complete mixed system with variable gas



⋅

=

+ +d

2

5 k1

k t RQ

1K1

In Figure 9 the efciencies for oxygen (kH = 0.039

at T=10°C) for the 5 basic equations are plotted

against the RQ with a k2t of 1.61.

The lines for K1 and K2 are obviously constant,

because, in this case, RQ is not of importance.

The lines for K3, K4 and K5 climb at increasing

values of RQ. When RQ approaches innity, the

lines for the different plug ow systems K1, K3 and K4 and for the mixed systems K2 and K5 co-

incide.

It can be concluded that a counter-current ow

reactor has a higher efciency than a co-current

ow reactor, and plug ow reactors have a higher

efciency than a complete mixed system.

The RQ is an important factor for the gas transfer

systems.

During the design of a gas transfer system, the

RQ value must be chosen. This depends on the

required efciency and the type of gas that needs

to be removed (Example 1).

The example to the right shows that the RQ nec-

essary for a 90% removal efciency of chloro-

form is 200 times greater than the value of RQ

for methane. This means that for the same water

ow the airow through the system and the ca-

pacity of the ventilator must each be at least 200

times greater.

A general rule that is applicable for the inuence

of the type of gas on the efciency is: the higher

the value of kH, the more air is needed for re-

moval, resulting in an increased RQ. Different

gas transfer systems have different characteris-

tics with respect to RQ.

A cascade, for example, has an RQ of approxi-

mately 0.4 and is therefore suitable for the re-

moval of methane and the addition of oxygen, but

is not used for the removal of chloroform.

Tower aerators are operated under different RQ

values and can be used for gases that are either

easy or difcult to remove, like tetra- and trichlo-

roethene.

Deep well aerators have the same characteristics

as cascades.

Example 1: The effect of RQ on the ef-

ciency

Calculate for a gas transfer system, that can

be represented by a complete mixed system,

the RQ that is necessary for a gas removal

efciency of 90% for methane, carbon dioxide

and chloroform. Assume that the contact time

in the reactor is innite and that the water tem-

perature is 100C. The efciency for a complete

mixed system can be calculated with the fol-

lowing equation:

⋅

=

+ +d

2

5 k1

k t RQ

1K

1

The contact time is innite, so 1/k2t = 0. The

above equation can be simplied as:

=

+d

5 k

RQ

1K

1

Gas Efficiency

[%]K5

[-] KD [-] RQ

Methane 90 0.90 0.043 0.39

Carbon dioxide 90 0.90 1.23 11.1

Chloroform 90 0.90 9.62 86.6

k5k4k3

k2

k1

0.001 0.01 0.1 1 10

1

0.8

0.6

0.4

0.2

0

k1

k2

k3

k4

k5

RQ

k 2t

k DT

K

( - )

=

=

=

1.61

0.039

10oC

Figure 9 - Efciencies of the different basic equations


143



3 Practice

3.1 Cascade

The water in a cascade is falling onto several

steps. Each step contains an overow weir and a

receiving gutter. When water passes over a weir,

an interface between air and water is created.

When the jet submerges into the receiving body

of water, signicant amounts of air are entrained.

The entrained air is then dispersed in the form of

bubbles throughout the receiving body of water,

which leads to an excessive transfer of gases.

The gas transfer takes place at the interface be-

tween the water and the air bubbles (Figure 10).

Because the amount of air that is entrained is lim-

ited, the RQ is also limited. According to practical

measurements and model investigations, the RQ

of cascades is approximately 0.4.

The energy consumption of a cascade is 10-30

Wh/m3.

Efciency An estimate of the efciency for a cascade can

be made, assuming that there is a relationship

between the measured fall height and the ef-

ciency. The efciency of a cascade depends on

the fall height of each cascade step and the num-

ber of steps:

w, e w, 0 n

s w, 0

c cK 1 ( 1 k)

c c

−

= = − −

−

in which:

k = efciency for each step [-]

n = number of steps

In Table 4 the efciency is given for oxygen, car -

bon dioxide and methane as a function of the fall

height of a step. With the data from Table 4 and

the equation mentioned above, the efciency of a

cascade with n steps can be calculated.

In practice, the total fall height of all the cascadesteps together varies between 2 and 7 meters.

From Table 4 it can be seen that oxygen and

methane efciencies increase with an increase in

fall height, but that the carbon dioxide efciency

remains constant. This is a result of the low RQ

value for cascades. Carbon dioxide removal re-

quires a higher value of RQ. The interface be-

tween air and water gets saturated rapidly with

carbon dioxide, regardless of the retention time

of air bubbles in the water, which is dependent

on the fall height. The greater the fall height, the

deeper the penetration in the trough, and the lon-

ger the retention time.

Weir loading

Weir loading is the amount of water per meter per

hour that ows over the weir.

The weir loading can be calculated by dividing

the ow by the net weir length (Figure 11):

=w

wnett

Qq

L

in which:

Figure 10 - Scheme of a cascade

K [%] h = 0.2 h = 0.4 h = 0.6 h = 0.8 h = 1.0 h = 1.2

O2 14 25 36 46 51 55

CO2 14 14 15 15 15 15

CH4 14 27 37 48 56 62

Table 4 - Efcency coefcient k of different gases as a function of the weir height


144



qw = weir loading [m3/m•h]

Lnett = total weir length [m]

From various experiments it can be concluded

that the efciency of a cascade is almost inde-

pendent of the weir loading. The advantage of

this is that the gas transfer is still satisfactory at

production ows that are lower than the design

ow.

With cascades the weir loading is generally be-

tween 50 and 100 m3/(m•h).

Trough depth

The trough depth of a cascade is chosen in sucha way that the falling water jet will not reach the

bottom. Air bubbles are dragged to a maximum

depth and this results in a maximum contact or

retention time and a maximum gas transfer time.

As a rule of thumb, the tray depth must be more

than two-thirds of the fall height.

Trough width

The trough width must be large enough to receive

the falling water jet (Figure 12).

The fall time of the water jet can be calculated

with the following equation:

= ⋅ ⋅21

h g t2

or

⋅=

2 ht

g

The distance x can be calculated when the water

velocity vo is known. To calculate the velocity, the

equation of the complete overow is used:

23 w

net

Qd

g L

=

×

and

wo

net

Qv

L d=

×

in which:

Qw =discharge [m3

/s]d = thickness of the falling water jet [m]

vo = velocity of the falling water jet [m/s]

The distance can be calculated with the equa-

tion:

= ⋅ox v t

With the distance x the trough width can be cal-

culated.

As a rule of thumb, the trough width is at least

twice the distance x:

= ⋅B 2 x

It is obvious that the trough width must be calcu-

X

h

H

B

Figure 12 - Scheme of the width of a cascade trough

80 mm 80 mm 80 mm+ + +(...) = Lnet

40 mm

Lgross

Figure 11 - Weir loading of a cascade aerator


145



lated using the maximum ow that is discharged

over the weir.

Congurations

The cascade troughs can be placed in two differ-

ent ways. They can be placed next to each other

or on top of each other (Figure 13).

Placing them next to each other is advantageous

because it looks attractive.

The advantage of putting them on top of each

other is that less space is used. The disadvan-

tage, however, is that this makes maintenance

more difcult.

3.2 Tower aerator

A tower aerator consists of a cylinder of steel or

synthetic material that is lled with a packing me-

dium.

Packing media can consist of stacked slats or

tubes, or specially designed packing material like

the Pall-ring and the Berl-saddle.

In the top section of the tower the water is divided

over the packing medium and ows down over

the medium surface. As a result of the ow of wa-

ter over the packing medium, a large contact sur -

face between the air and water is created for gas

transfer. In addition, the water falls in drops from

one packing element to the other, continuously

forming new drops thus renewing the air-water

interface.

The air can be renewed by natural ventilation or

with the help of a ventilator. In case a ventilator

is used, the air can have a co- or counter-current

ow in the tower. In Figure 14 a tower aerator

with counter-current ow is represented.

In Figure 15 different types of packing material

are represented. The packing material can be

produced from synthetic material, metal, carbon

or ceramic material.

The dimensions of the individual pieces vary from

6 mm to 75 mm. In practice, installations used

for purifying drinking water use mostly synthetic

packing material with a dimension of 25-50 mm.

Figure 13 - Cascades beside each other and on top

of each other

A

B

C

D

E

A influentB packing materialC air supplyD effluentE air discharge

Figure 14 - Representation of a counter-current tower

aerator


146



Surface loading

The surface loading (ow divided by surface

area) that in practice is used in tower aerators is40 to 100 m3/(m2•h).

The applied packing height, that determines the

retention time of the water in the tower aerator,

varies between 3 and 5 meters.

Efciency

With tower aerators, removal efciencies can be

as high as 95%.

The applied RQ depends on the gases that need

to be removed.

In Figure 16 the results of a pilot experiment us-

ing a tower aerator are represented.

It can be concluded that the efciency hardly

changes when the surface loading is increased.

This is considered remarkable. In most gas trans-

fer systems ,a larger ow results in a greater ow

rate, resulting in a shorter retention time for the

water, and a lower efciency.

This insensitivity to the surface loading with atower cascade can be explained by the fact that

the retention time in a tower aerator is practically

independent of the water ow. The water falls un-

der the inuence of gravity, so the retention time

is mainly determined by the type of packing ma-

terial used and the height of the bed. It is indiffer-

ent if more or less water falls through the tower

because the retention time remains unchanged.

In Figure 17 more results from the removal ef-

ciency experiments are given.

For all points in the graph, with the combina-

tion of packing height and RQ, an efciency of

99% is reached. From this graph it can be con-

cluded that, at a certain point, an increasing RQ

value does not lead to a reduction of the packing

height. At that point the amount of air is not de-

cisive but the minimum necessary retention time

for removal of 99% is reached.

Clogging

A disadvantage of the tower aerator is that the

system is sensitive to clogging. If iron (Fe2+) is

present in groundwater, it will oxidize in the tower

aerator (Fe3+) and remain on the packing material

(Fe(OH)3). Because the oxidized iron inuences

the gas transfer negatively, it will be necessary to

back ush the tower aerator. Water with a high

velocity, or a combination of water and air, is thenushed through the tower aerator, removing the

iron contamination from the packing material. In

addition to ushing, it will be necessary to pe-

riodically clean periodically the packing mate-

rial chemically. In this case, the packing material

must be removed from the tower aerator.

Co- or counter-current ow

A tower aerator can be operated in both co-cur-

rent ow and counter-current ow (Figure 18).

80

85

90

95

100

0 20 40 60 80 100

e f f i c

i e n c y ( % )

RQ (-)

18 m3 /(m2*h)36 m3 /(m2*h)

trichloro ethene

packing material: hy-pack steel 30mmheight packing material 3mtemperature: 11 oC

54 m3 /(m2*h)72 m3 /(m2*h)

Figure 16 - Removal efciency of a tower aerator as a

function of RQ at different surface loadings

Figure 15 - Different types of packing material


147



In the paragraph on theory it was explained that

counter-current ow results in a higher efciency

than co-current ow. Still, co-current ow is ap-

plied. The reasons for this are:

- to avoid high carbon dioxide removals which

will cause limestone scaling. Using a co-cur -

rent aerator with low values of RQ, the addi-

tion of oxygen and the removal of methane

are sufcient while carbon dioxide removal

will be limited.

- to apply needed high surface loadings. Us-

ing counter-current ow, “ooding” can occur.

This means that a water layer is created in the

column because of the buoyancy of air, which

can even result in the tower aerator lling up

with water.

3.3 Plate aerator

A plate aerator consists of a horizontal perforated

plate. Water ows over the plate and air is blown

through its orices, creating a bubble bed of air

and water above the plate (Figure 19).

This results in intense contact between the air

and the water.

The combination of horizontal water ow and ver -

tical airow (i.e., the ows are perpendicular), is

called cross-ow aeration.

The height of the bubble bed is determined by

adjusting the height of the weir at the end of the

plate.

The diameter of the holes in the perforated plate

is usually 1-1.5 mm. The open surface area var-

ies from 1.5 % to 3% of the total plate surface

area.

The energy consumption of a plate aerator is 30-

40 Wh/m3.

Due to the reduced construction height and head

loss, this technique offers good possibilities for in-

corporating it in existing treatment plants. Some-

times it is possible to place the plate aerators inthe lter building directly above the lters.

Efciency

The efciency of plate aerators is mainly deter -

mined by the applied RQ and the retention time

of the water on the plate. There is no analytical

equation for calculating the efciency, unlike the

co- and counter-current ows.

In practice, the applied RQs vary from 20 to 60

and the applied surface loading varies from 30 to

0

5

10

15

0 10 20 30 40

h e i g h t p a c k i n g m a t e r i a l ( m )

RQ (-)

18 m3 /(m2*h)36 m3 /(m2*h)

trichloro ethene

packing material: hy-pack steel 30mmefficiency: 99%temperature: 11 oC

54 m3 /(m2*h)72 m3 /(m2*h)

Figure 17 - Required packing height and RQ to achieve

an efciency of 99% at different surface

loadings

co-current flowcounter-current flow

airwater water

air

Figure 18 - Design alternatives for tower aerators

air

water

Figure 19 - Representation of a plate aerator


148



tween air and water is saturated. Because the

droplet remains intact during the fall, the interface

is not renewed and the gas transfer stops.

Energy consumption

Spray aerators need a certain pressure to guar-

antee an equally distributed spray. For sprayers

that produce ne droplets (mist), the pressure is

the greatest, about a 10-meter water column.

The energy consumption of these high pressure

spray aerators is, therefore, the largest.

Clogging

A disadvantage of sprayers is their high sensitiv-

ity to clogging.

Alternatives in practice

40 m3/(m2.h).

Clogging

Plate aerators are sensitive to clogging because

of the small orices in the plate. Iron deposits

found on the plate can block the orices and af -

fect the ow through the plate.

Short-circuit ows can occur, inuencing nega-

tively the gas transfer.

Depending on the iron loading, the plate has to

be cleaned once a month or once every other

month. It might also be necessary to clean the

plate chemically once or twice a year.

3.4 Spray aerator

Spray aerators divide water into small droplets,

which results in a large air-water interface (Figure

20). The energy consumption of spray aerators is

10-50 Wh/m3, depending on the type of aerator.

An advantage of spray aerators is the ease of in-

corporation into existing installations. The spray

aerators can be placed directly above the lters.

Efciency

When the air is intensively renewed, the efcien-

cy of spray aerators can be calculated with the

following equation:

− ⋅

− ⋅= − = −

22

2h( k )

g( k t)K 1 e 1 e

The efciency for the addition of oxygen can vary

from 65 to 80%, for the carbon dioxide removal

the efciency varies from 60 to 80%.

In Figure 21 the efciency of the Dresden-nozzle

for carbon dioxide removal as a function of the

fall height is shown.

It is remarkable that after a certain fall height the

efciency remains more or less constant. The

reason is that after some time the interface be-

Figure 20 - Spraying small droplets of water

2

1

0

0 0.25 0.5 0.75 1

K CO2 [-]

h [ m ]

Figure 21 - Efciency Dresden-nozzle as a function of

the fall height


149



Spray aerators can be divided into two groups:

upward- and downward-directed spray aerators. An example of the rst type is the ‘Amsterdam’

spray aerator (Figure 22). In this type of spray

aerator, two jets are directed perpendicular to

each other, dispersing the water. This results in

many droplets in the air. During the fall of the wa-

ter droplets, the gas transfer takes place.

An example of the second type of sprayer is the

Dresden sprayer (Figure 23), or the plate spray-

er. Here, the water ows through a plastic tube

and strikes a disc (plate), shaping the water like

an umbrella, and eventually disintegrating into

droplets.

3.5 Alternative aeration systems

Vacuum gas transfer system

A vacuum gas transfer system is usually execut-

ed as a tower aerator lled with a packing mate-

rial in which the pressure is lowered by a vacuum

pump (Figure 24).

Due to the vacuum pump, gas is removed from

the tower, resulting in lower gas concentrations

and a decreased pressure there. Because the

gas concentrations in the tower are lower than in

the atmosphere, the saturation concentrations in

the tower are also lower. Because of the low sat-

uration concentrations, it is possible to remove

higher levels of gas from the water than is pos-

sible under atmospheric conditions. This makes

a vacuum gas transfer system ideal for remov-

ing dissolved nitrogen and oxygen from the water

and is frequently applied before the denitricationprocess.

The efciency of the vacuum gas transfer system

depends on the vacuum pressure that is main-

tained in the tower. In the absence of an air ow,

the RQ equals zero. Since oxygen is not brought

into the system, oxidation of iron cannot occur.

This allows the water to be pumped to the next

treatment process, contrary to a cascade. In

a cascade oxidation of iron does occur, which,

when the water is pumped to the next treatment

Figure 22 - Amsterdam sprayer

Figure 23 - Dresden sprayer

A

B

D

E

A influent

B packing materialC air supply

D effluent

E air discharge

Epump

pump

Figure 24 - Representation of a vacuum liquid-gas

exchange


150



process, causes the iron ocs to break up making

them harder to remove in the lter.

Like the tower aeration system, the vacuum sys-

tem is not very sensitive to surface loading. The

applied surface loading varies from 50 to 100

m3/(m2.h).

A great disadvantage of the vacuum gas transfer

system is its high energy consumption, requiring

approximately 1,600 Wh/m3 to maintain it.

Deep-well aerator

Water ows through the deep well, entraining airby a venturi (Figure 25 right), or air is supplied at

the bottom of the well (Figure 25 left).

Due to the high water pressure at the bottom

of the well, an increase in air pressure is estab-

lished, which results in a higher oxygen concen-

tration. With a higher saturation concentration,

more oxygen can be dissolved into the water

than at atmospheric conditions.

Deep well aerators are mainly used in the treat-

ment of wastewater, because the oxygen con-

sumption of wastewater is normally high.

The advantage of a deep well aerator is that large

amounts of water can be treated against relative-

ly low energy costs. The energy consumption for

the deep well aerator is approximately 5 Wh/m3.

Venturi aerator

The venturi aerator consists of a tube with a re-duced cross-sectional area, where the increased

water velocity occurs. At the place where the

water velocity is the highest (through orices

in the tube), air is entrained. Due to the strong

turbulence, an intensive mixing of the entrained

air with the water leads to the dispersion of ne

bubbles.

Since the amount of air that can be entrained

is relatively small, the RQ of a venturi aerator is

rather small, varying between 0.2 to 0.4.

The efciency for oxygen addition ranges from 80

to 95%.

The advantage of the venturi aerator is that it

requires little space and the system is not ex-

pensive. A disadvantage is that only limited ow

variations can be allowed for an optimal effect.

The energy consumption is approximately 20-30

Wh/m3.

Bubble aeration

The transfer of gas by means of a bubble aera-

tor is accomplished by injecting compressed

air through orices of various sizes into the

water(Figure 27). Air is distributed by perforated

pipes at the bottom of a tank. During the rise of

the formed bubbles, gas transfer takes place.

This system is mainly used in wastewater treat-

ment. The principle of gas transfer by bubble

aeration is the same as in cascades.

Cone aerator

A cone aerator is used as a gas transfer systemfor the treatment of wastewater.

The cone aerator consists of a large rotating

h

H

2 rows of air pipes

inflowing

water outflowing water

supply of

compressed

air

discharge

aerated

water

supply of

raw water

discharge

of sludge

Figure 25 - Design alternatives for a deep-well aerator

air filter

air supply

raw water supply

aerated water evacuation

Figure 26 - Representation of a venturi aerator


151



blade in the form of a cone, situated in a basin

on the water’s surface(Figure 28). Through the

blade, water is abstracted from underneath thecone and sprayed laterally over the water’s sur-

face. Because water droplets are formed and air

is entrained, gas transfer can be achieved.

As a result of the suction of water from under-

neath and the horizontal distribution of the water,

a circular ow is created and the water in the ba-

sin is aerated.

Figure 27 - Bubble aeration system Figure 28 - Cone aerator

Further reading


(2005), (ISBN 0 471 11018 3) (1948 pgs)

• Modellering van intensieve gasuitwisselings-gasuitwisselings-

systemen (in Dutch), A.W.C. van de Helm (MSc

thesis)


152



Softening WA T E R T R E

A T M E N T

WATER TREATMENT

diameter 0.5 - 4 m

h e i g h t + / - 6 m e t e r s

A

B

C

E

D

supply of hard watersupply of lyeperiodic dosing of sand grains (0.1-0.4 mm)forming pelletsoutlet for softened water

periodic outlet of pellets (2 mm)

A BCDE

F

in rest

in progress

F

hardness reduction [Ca2+] (mmol/l)

0

6

0

H C O 3 - r a w

w a t e r ( m m o l / l )

5

1

2

3

4

5

1 2 3 4

NaOH

Na2CO3

Ca(OH)2NaOH

Na2CO3

Na2CO3

groundwater NaOH

surface water NaOH

groundwater Ca(OH)2

surface water Na2CO3



Framework

This module explains softening.

Contents


1 Introduction

2 Principle

2.1 Why softening?

2.2 Water quality and softening

2.3 Softening processes

2.4 Pellet reactor

2.6 Softening in a treatment plant

3 Theory

3.1 Equilibrium 3.2 Kinetics

3.3 Mass balance

3.4 Hydraulics

3.6 Inuence of parameters

4 Practice

4.1 Split treatment

4.2 Choice of chemicals

4.3 Construction alternative for reactors

4.4 Seeding material

4.5 Pellet storage

154

WATER TREATMENT SOFTENING



1 Introduction

Groundwater normally remains in the subsoil for

many years before it is pumped up or ows out into

the surface water. Due to the long residence time in

the subsoil, groundwater is in chemical equilibrium

(i.e., calcium carbonate equilibrium).

Groundwater comes in contact with the atmos-

phere when it is pumped up or discharged into

surface water. When carbon dioxide disappears

from the water, it is not in calcium carbonate equi-

librium anymore.

Also, when water is heated the equilibrium is

changing, the Ca2+ and HCO3- - ions will precipi-

tate in the form of calcium carbonate (CaCO3).Especially high concentrations of Ca2+ and HCO3

-

- ions will lead to inconveniences for the customers

because of the calcium carbonate scaling (e.g.,

deposits in water boilers).

To prevent precipitation of calcium carbonate at

the customers’ taps, calcium ions are partially

removed from the water by drinking water com-

panies. This is called softening.

2 Principle

2.1 Why softening?

The nancial benets of softening are greater than

the costs.

Amsterdam Water Supply calculated that the

benets of softened water for one single household

comes to a saving of approximately 45 euros a

year (mainly as a result of their decreased use of

detergent, less maintenance on washing machines

and boilers, and lower energy costs), whereas thesoftening costs for a household are approximately

10 euros a year.

In addition to decreasing the hardness of water,

another important reason for softening is the

reduced release of heavy metals. Other reasons

why softening is used are given in Table 1.

2.2 Water quality and softening

The hardness of water is classied from very soft

to very hard (Table 2).

A number of water quality parameters are inu-

enced as a result of the softening process. For

these parameters, standards are included in

the Dutch National Drinking Water Standards(Waterleidingbesluit). In addition to these standards,

guideline values were developed by VEWIN.

Public health

- decreased release of heavy metals from dis-

tribution network

- no use of household softening devices

Ethics

- prevention of stains

- user’s comfort

Environment

- reduction of heavy metals in sludge WWTP

- reduction in use of detergent and decreased

phosphate content in wastewater

- reduction of concentrate discharge of

household softening devices

Economy

- reduction in usage of detergent

- reduction of scaling and corrosion of house-

hold equipment

- reduction of energy consumption of heating

devices

- reduction in damage to clothes

Table 1- Reasons why softening is applied

Table 2 - Classication of hardness

unit very soft soft fairly soft fairly hard hard very hard

mmol/l <0.5 0.5 - 1.0 1.0 - 1.8 1.8 - 2.5 2.5 - 5.0 > 5.0

eq/m3 < 1 1 - 2 2 - 3.5 3.5 - 5 5 - 10 > 10

*D < 3 3 - 6 6 -10 10 -15 15 - 25 > 25

155

SOFTENINGWATER TREATMENT



The produced water always needs to comply with

the standards. The guideline is a target that the

water companies set themselves.

The most important water quality parameters

inuenced by softening are acidity, hardness, bi

carbonate, sodium, and the solubility potential for

metals like copper and lead.

Acidity (pH)- standard = 7.0 < pH < 9.5

- guideline = 8.0 < pH < 8.3

Directly after the softening process, acidity of the

water is higher than the above-mentioned guide-

line. By means of pH-correction (acid dosing), pH

is decreased to the desired value.

Hardness

- standard = 1.5 mmol/l at the minimum

- guideline = 1.5 < hardness < 2.5 mmol/l

Hardness is dened as the sum of the concentra-

tion of dissolved calcium and magnesium ions. The

hardness is sometimes expressed in °D (German

degrees), or equivalents, per m3. Table 2 shows

the conversion of mmol/l to German degrees and

its equivalents per m3.

Bicarbonate concentration

- standard = no standard

- guideline > 2 mmol/l

The bicarbonate concentration should be higher

than 2.0 mmol/l, resulting in water with sufcient

buffering capacity (pH stability);

Sodium concentration

- standard = 120 mg/l

- guideline = as low as possible

Because sodium inuences blood pressure and

therefore, indirectly, heart and vascular diseases,

the concentration should not be higher than 120

mg/l.

Solubility potential

Softening also works to reduce the solubility of

metals from pipe material. For drinking water the

most important metals are lead (Pb) and cop-

per (Cu), because these metals have a health

impact.

- standard Cu2+ < 3 mg/l

- guideline Cu2+ < 2 mg/l

- standard Pb2+ < 0.2 mg/l

- guideline Pb2+ < 0.01 mg/l

The values for copper and lead solubility are deter-mined by a pipe test. The pipe test is performed

in stagnant water and takes 16 hours. Empirical

relationships have been derived to give a rapid

indication about the release:

2

max 4Cu 0.52 TAC -1.37 pH 2 SO 10.2− = ⋅ ⋅ + ⋅ +

maxPb -141 pH 12 T 1135= ⋅ + ⋅ +

Figure 1- Scaled heating element

156




in which:

Cumax = copper dissolving capacity (mg/l)

TAC = Total Anorganic Carbon (mmol/l)

Pbmax

= lead dissolving capacity (mg/l)

SO42- = sulfate concentration (mmol/l)

T = temperature (oC)

The TAC concentration can be inuenced by the

softening process (mainly caused by the decrease

in HCO3-).

2.3 Softening processes

The hardness of raw water in the Netherlands varies

between 0.5 and 5 mmol/l.

Groundwater extracted from calcareous subsoils,

especially, can have a high degree of hardness.

Water extracted from deep sand layers of the

Veluwe is, on the other hand, fairly soft (approxi-

mately 0.5 mmol/l).

The hardness of surface water is normally from

2.0 to a maximum of 3.0 mmol/l.

The hardness of water can be decreased by

means of different processes:

- dosing of a base

- ion exchange

- membrane ltration

Dosing of base (NaOH, Ca(OH)2 or Na2CO3 )

Because of a shift in the calcium carbonic acid

equilibrium, spontaneous crystallization occurs.

By dosing the base in a reactor with seeding

grains, crystallization will occur on the surface

of the seeding grains, forming limestone pellets.

This process is called softening in a pellet reactor

and will be described further in section 2.4. The

process of softening by means of a pellet reactorwas developed in the early 70s by Amsterdam

Water Supply.

Ion exchange

The ions (for this, calcium and/or magnesium)

are exchanged with other ions (sodium is used

the most).

Membrane fltration

Depending on the type of membrane, the hardness

is partly (nanoltration) or fully (reverse osmosis)

removed.

2.4 Pellet reactor

The principle of the pellet reactor is shown in

Figure 3.

The pellet reactor consists of a cylindrical vessel

partially lled with seeding material. The diameter

of the seeding material is approximately 0.2 - 0.6

mm and it has a large crystallized surface.Water is pumped in an upward direction through

the reactor at a velocity varying between 60 and

100 m/h. At these velocities the sand bed is in a

uidized condition.

Raw water and chemical (base) are injected into

the bottom of the reactor by separate nozzles.

Water and chemicals are well-distributed over the

cross-section of the reactor (plug ow) once suf-

cient ow resistance is realized over the nozzles.

Figure 2 - Pellet reactor used for softening of drinking

water

157




The process conditions (such as chemical doses

and ow velocity) need to be selected so that

the solubility product of calcium carbonate is

exceeded. As a result, calcium carbonate will be

formed (quick reaction) and precipitate onto the

seeding material.

The seeding grains’ diameter will increase as a

result of the calcium carbonate deposit (forma-

tion of pellets). The pellets will become heavier

and settle to the bottom of the reactor. Finally,

the pellets (at a diameter of 1.0 - 1.2 mm) will be(at a diameter of 1.0 - 1.2 mm) will bewill be

removed from the reactor and new seeding grains

will be brought in.

The pellets can be reused in the industry.

Softening with a pellet reactor does not generate

waste products.

2.5 Softening in a treatment plant

To incorporate a softening installation (includ-

ing post treatment) into an existing groundwatertreatment process, the following possibilities are

considered:

- softening of raw water

- softening of aerated water

- softening after rapid ltration.

Softening of raw water is done directly after it is

pumped up.

If iron and manganese are present in dissolved

form in the water (anaerobic water), these sub-

stances will be trapped in the CaCO3 grains.

The advantage of this is that the loading on the

sand lters is reduced.

A disadvantage is that the CaCO3 grains become

less pure, affecting the growth of crystals, result-

ing in uffy pellets.

Another disadvantage is that the base dosage

is high due to the high concentration of carbon

dioxide in raw water. Before the softening reaction

starts, the carbon dioxide needs to be convertedthe carbon dioxide needs to be converted

to HCO3

-

and CO3

2-

..

When softening takes place after an aeration

phase, a lower chemical dose will be sufcient,

because some of the carbon dioxide is removed

during aeration.

An additional (possible) cost advantage of soften-

ing (aerated) raw water is that, in many cases,

existing lters that have been used for iron and

manganese removal uptill this point, can also be

applied as ‘carry-over’ lters.

diameter 0.5 - 4 m

h e i g h t + / - 6 m e t e r s

A

B

C

E

D

supply of hard watersupply of lyeperiodic dosing of sand grains (0.1-0.4 mm)forming pellets

outlet for softened waterperiodic outlet of pellets (2 mm)

A BCD

EF

in rest

in progress

F

Figure 3 - Schematic representation of a pellet reactor

Figure 4 - Limestone pellets

158




When softening after ltration is applied, the pur -

est pellets are formed. Iron and manganese are

removed by the lters.

A disadvantage is that after softening a new

(expensive) rapid ltration step must be added to

the process to remove the ‘carry-over.’

3 Theory

3.1 Equilibrium

The calcium carbonic acid equilibrium deter-

mines whether calcium carbonate precipitates.

For an extensive explanation on this subject,

you are referred to lecture notes from the course‘Introduction in Sanitary Engineering (CT3420),’

specically the chapter on water quality. Only

the most important formulas are mentioned

here(answers for T = 10 ºC) :

[ ]3 3 7

1

2

H O HCOK 3.44 10

CO

+ −

− ⋅ = = ⋅

2

3 3 11

2

3

H O COK 3.25 10

HCO

+ −

−

−

⋅ = = ⋅

2 2 9

s 3K Ca CO 4.4 10+ − − = ⋅ = ⋅

5s 1a

2

K KK 4.6 10

K

−⋅= = ⋅

( )s

3 2 s

SI pH pH

2 log HCO pK pK log 2−

= −

= − ⋅ + + +

( )2

3 2 2 3 aCaCO CO H O Ca 2 HCO K+ −+ + ↔ + ⋅

In groundwater abstracted in South Limburg, a

high concentration of calcium ions is present.

Given that there is no CO32- in the water (pH is 6),

calcium does not precipitate but remains in dis-

solved form in the water, resulting in water with a

high degree of hardness.

By softening, the pH of the water is increased as

a result of dosing a base. When caustic soda is

used, the following reactions will occur:

NaOH

OH-

+ HCO3

-

CO3

2-+ Ca

NaOH + Ca2

++ HCO

3

-

Na+

+ OH-

CO3

2-+ H

2O

CaCO3

CaCO3

+ H2

O + Na+

→

→

→

→

The above-mentioned reactions are irreversible;

in reality, equilibrium will be set.

Also for the bases Ca(OH)2 and Na2CO3, similar

reaction equations can be formulated.

By dosing a base, the carbonic acid equilibriumshifts to the left, forming calcium carbonate. The

Saturation Index exceeds 1. At a similar SI, crys-

tallization of calcium carbonate occurs, forming

a deposit on the seeding grains present in the

reactors.

3.2 Kinetics

Experimental research shows that the kinetic

equation for precipitation of calcium carbonate can

be described with the following equation:

( )2

2 2

t 3 s

d Ca- k S Ca CO -K

dt

+

+ − = ⋅ ⋅ ⋅

in which:

kt = reaction constant (..)

S = specic area (..)

(…) = supersaturation or driving force

The reaction constant kt is a function of tempera-

ture and is given by the next equation:

Kt = 0.0255 . 1.053(T-20)

The specic area in uidized reactors is dened

as:

( )1 pS 6

d

−= ⋅

159




in which:

p = porosity (-)

d = diameter of the pellets (m)

A smaller diameter of pellets results in a larger

specic area and, thus, a faster softening reaction.

A smaller porosity results in a higher specic area

and a faster reaction.

Supersaturation is the chemical driving force for

the crystallization reaction. The higher this driving

force, the faster the reaction proceeds.

3.3 Mass balanceIn a pellet reactor calcium carbonate forms a

deposit on the seeding grains added to the reac-

tors. A decrease in calcium concentration results

in an increase in pellet diameter. This increase is a

function of the calcium concentration decrease:

( )d f c∆ = ∆

The total equation becomes:

( ) [ ] [ ]( )3 3

k 2 1 p 1 2N d d Ca Ca M Q

6π⋅ ⋅ − ⋅ ρ = − ⋅ ⋅

in which:

Ca1 = calcium concentration before reaction

(mol/m3)

Ca2 = calcium concentration after reaction

(mol/m3)

M = molecular weight of calcium carbonate

(100 g/mol)

Q = ow (m3

/s)Nk = number of pellets in the reactor per time

unit (-)

d1 = diameter of seeding material (m)

d2 = diameter of pellets when they are re-

moved from the reactor (m)

ρp = calcium carbonate density (=2840) (kg/m3)

The seeding material with a small diameter will be

located at the top of the reactor. Slowly, calcium

carbonate starts to deposit on the seeding mate-

rial, and the pellets grow and settle.

Eventually, the pellets are located at the bottom

of the reactor and are discharged.

3.4 Hydraulics

To design a pellet reactor, one must understand

the hydraulics of a uidized bed.

With the hydraulic formulas, the porosity and

height of the expanded(fluidized) bed can be

determined.

The hydraulics of pellet reactors are the same as

backushing rapid lters.

Water ows in an upward direction through the

bottom of the reactor and, because of the highvelocity, the bed uidizes and expands. In sand

ltration the expansion will extend a maximum of

20%; in the softening process the expansion can

reach 200%.

The maximum resistance is given by the weight

of the grains under water, or:

( ) p w

max

w

H 1 p Lρ − ρ

= − ⋅ ⋅

ρin which:

Hmax = maximum resistance (m)

ρw = density water (kg/m3)

ρp = density pellets (kg/m3)

The velocity at maximum resistance is called

vmin

.

At a higher velocity than vmin, the resistance

remains constant and the bed expands.

The expansion can be calculated with the equa-

tion:

e o

o e

L 1 pE

L 1 p

−= =

−

in which:

Le = height of expanded bed (m)

Lo = height of xed bed (m)

pe = porosity of expanded bed (-)

po = porosity of xed bed (-)

160




The porosity of an expanded bed at a certain

upward velocity is calculated using:

( )

3 0.8 1.2e w

0.8 1.83p w

e

p v v

130 g d1 p

ρ

= ⋅ ⋅ ⋅ρ − ρ−

The height of the expanded bed can be calcu-

lated when the upward velocity and the porosity

are known:

oe o

e

1 pL L

1 p

−= ⋅

−

In Figures 5 through 7 the inuence of somehydraulic parameters as a function of upward

velocity is given.

A particle with a diameter of 0.3 mm is seeding

material; a particle with a diameter of 1.5 mm is

the discharged pellet.

Figure 5 implies that with an increasing upward

velocity, porosity in the reactor increases.

Besides that, it is obvious that with a larger diam-

eter (for example, caused by deposits of calcium

carbonate on seeding material forming pellets) the

porosity decreases.

The specific surface area for crystallization

decreases in the reactor from top to bottom. A

direct consequence of the increase in porosity

at a higher upward velocity is the greater bed

expansion.

The specic area decreases at a higher upward

velocity due to higher porosity.

For particles with a diameter of 0.3 mm, the

increase in the expansion at higher upward veloci-

ties is relatively large.

These small particles can be ushed out.

0.35

0.45

0.55

0.65

0.75

0.85

0.95

50 60 70 80 90 100 110 120

P o r o s i t y ( - )

velocity (m/h)

d = 0.3 mm

d = 0.6 mm

d = 1.0 mm

d = 1.5 mm

Figure 5 - Porosity as a function of pellet diameter and

upward velocity

Figure 7 - Specic area as a function of pellet diameter

and upward velocity

1500

2000

2500

3000

3500

4000

50 60 70 80 90 100 110 120

s p e c i f i c s u r f a c e a r e a S ( m 2 / m 3 )

velocity (m/h)

d = 0.3 mm

d = 0.6 mm

d = 1.0 mm

d = 1.5 mm

0.8

1.6

2.4

3.2

4

50 60 70 80 90 100 110 120

b e

d

e x p a n s i o n

E (

- )

velocity (m/h)d = 0.3 mm

d = 0.6 mmd = 1.0 mm

d = 1.5 mm

Figure 6 - Bed expansion as a function of pellet dia-meter and upward velocity

161




3.5 Inuence of parameters

With the basic equations for softening reactions

and the hydraulic equations for an expanded bed,

it is possible to design a pellet reactor and show

the values of some characteristic parameters as

a function of height.

Because the solution of the equation is difcult

(defining the reactor in terms of a number of

discrete intervals, dx, and solving the equation

per dx), computer programs are used to design

installations.

Table 3 shows some input data of a reference

calculation, Figure 8 shows the results of a calcu-

lation graphically.The most important parameter for the design of a

pellet reactor is the height of the expanded bed.

This determines the height of the pellet reactor

and the building where the pellet reactor will be

placed.

Using the input data from Table 3 it follows that,

after calculating, the height of the expanded bed

is 5.43 m.

Table 4 shows the consequence of varying input

parameters. The table still includes an unknown

parameter, dCa. The computer program calculates

a theoretical base dosage to reach the efuent

concentration Ca2. This value is only reached when

the reactor is innitely high, when the equilibrium

is completely reached. However, an innitely high

reactor is neither practical nor feasible; therefore, a

supersaturation of calcium carbonate is accepted,

and leads to a lower reactor. To reach the efu-

ent concentration Ca2, a higher dosing of a base

should take place. In practice a value of dCa to

0.10 mmol/l is acceptable.

Not only the expanded bed height changes with

changes in one of the input data, but other para-

meters also change.

Table 5 gives an overview of the inuences of

varying one input data point.

Figure 9 gives some parameters as a function of

the level in the reactor.

L

pd

d2

d1

v

Ca1

Ca2

Ca1

Ca2

Figure 8 - Resistance as a function of upward velocity

and grain diameter

Raw water

composition

Ca1

TAC

HCO3-

T

(mmol/l)

(mmol/l)

(mmol/l)

(oC)

3.5

5.0

4.25

10

Softened water

composition

Ca2

dCa

(mmol/l)

(mmol/l)

1.5

0.06

Pellet reactor

characteristics

v

d1

ρ1

d2

ρp

(m/h)

(mm)

(kg/m3)

(mm)

(kg/m3)

80

0.3

2650

1.0

2840

Table 3 - Softening with caustic soda in pellet reac -

tor Table 4 - Consequences of variation in input data on

expanded bed height, dosing caustic soda

Referenceconditions

Variableconditions

Expan-ded bed

height Le

(m)

T = 100C T = 5 0C 6.73

v = 80 m/h v = 120 m/h 10.9

d2= 1.0 mm d

2= 0.75 mm 5.39

d1= 0.3 mm d

1 = 0.2 mm 5.58

ρo = 2650 kg/m3 ρ

p =4200 kg/m3 4.55

dCa = 0.06 mmol/l dCa = 0.10 mmol/l 2.78

Ca2 = 1.50 mmol/l Ca2 = 1.0 mmol/l 3.12

162




Ca : driving force calcium concentration = softening (supersaturation largest at the bottom of the

reactor (see SI): thus most removal at the bottom of reactor);

d : as a result of the growing of pellets, stratication will take place, because the pellets with large

diameters will settle to the bottom;p : porosity increases because the diameter decreases in height (opposite of d);

S : specic area has a maximum at a certain height. Below this height S decreases because then

the pellet diameter is decisive. Above this height S decreases because porosity is more deci-

sive;

pH : highest at the bottom of the reactor due to chemical dosage at the bottom;

SI : see Ca ( as a result of chemical dosage the supersaturation increases).

Ca (mmol/l)

0

6

0 5

1

2

3

4

5

1 2 3 4

L e

( m )

0

6

0 1.50

1

2

3

4

5

0.50 1.00

d (mm)

L e

( m )

p (-)

0

6

0.5 1.0

1

2

3

4

5

0.6 0.7 0.8 0.9

L e

( m )

0

6

1500

L e

( m )

2000

1

2

3

4

5

1800 2100 2400 2700

S (m2/m3)

0

6

0 3.00

SI

1

2

3

4

5

0.50 1.00 1.50 2.00 2.50

L e

( m )

0

6

7 11

1

2

3

4

5

8 9 10

pH

L e

( m )

Figure 9 - Some characteristic softening parameters as a function of height

163




4 Practice

4.1 Split treatment

When only a part of the water ow is softened, it

is called split treatment.

Figure 10 shows the principle. One part of the

water passes through the softening installation

(and obtains a lower hardness) and one part does

not pass the softening installation (and has the

same hardness as the raw water).

Afterwards, the two ows will mix, resulting in an

overall hardness of 1.5 mmol/l.

Split treatment has a number of advantages. One

is that the consumption of chemicals is lower. In

raw water an amount of carbon dioxide exists

which needs to be converted into carbonate.

In the case of split treatment, only carbon dioxide

needs to be converted in one part of the ow; in the

by-pass, no dosage of chemicals takes place.

When softened water is mixed with raw water

which bypassed the softening installation, the

water has a lower supersaturation after mixing

(principle of Tillmans curve).

Another advantage of split treatment is that the

investment costs will be lower, because fewer

reactors need to be built.

The choice for split treatment is signicant for large

design capacities (i.e., larger savings).

Split treatment will only be used when the con-centration of magnesium in the raw water is not

too high.

The maximum softening depth is down to a calcium

concentration of 0.5 mmol/l. When the magnesium

concentration is high (about 1 mmol/l), the bypass

percentage approaches 0% to reach a total hard-

ness of 1.5 mmol/l.

4.2. Choice of chemicals

Selection of the proper chemical (caustic soda,

lime or soda) is determined by the raw water

composition and desired quality after softening.

In case several chemicals are applicable, aspects

of operational management will also become

important.

Previous standards were given that must be com-

plied with.

These standards determine, to a considerable

extent, what base can be used for softening.

Table 6 shows the change in water quality for themost important parameters when bases are dosed

to water (on the basis of irreversible reactions).

The water quality after softening can be easily

determined and conclusions can be drawn as to

whether the water meets the standards. In real-

ity, equilibrium reactions occur but change the

concentrations only slightly. However, for a rst

estimate, the values in Table 6 give an indication

of the changes.

THsplit treatment

THend

THraw water

(1-R) QQ

R Q

Figure 10 - Principle of split treatment

Table 5 - Inuence of changes in input data on para-

meters

increase of inuence on

Le

dos Nk

S G SImax

T << < - > >> <

v >> - - << >> -

d2

> - << << >> -

d1

<> - >> > > -

p1

< - >> > > -

Ca1

<< > - - - >

Ca2

<> << << - - <<

Ca(OH)2

>> - - - - <<

164




in bicarbonate concentration in relation to hard-

ness reduction.

With the application of caustic soda, for every

mmol/l calcium reduction the hydrogen carbonate

concentration decreases with 1 mmol/l (line 1:1).

With lime per mmol/l hardness reduction, the

hydrogen carbonate concentration decreases with

2 mmol/l (line 1:2).

Using the previously mentioned computer pro-

gram, the exact water quality after softening with

caustic soda or lime can be calculated. Figure 12

shows the progress for both the bases of some

characteristic softening parameters as a functionof height in the reactor. Distinct differences can be

noticed. In case different chemicals can be used, a

choice will need to be made that takes other quality

parameters (such as Cu solubility, TAC) and opera-

tional management aspects into account.

Caustic soda

Caustic soda is provided as a 50% solution (= 50%

caustic soda and 50% water) by a tanker truck.

Since a 50% caustic soda solution crystallizes

at a temperature lower than 12°C, caustic soda

is diluted to a 25% solution using demineralized

water. This 25% solution only crystallizes at tem-

peratures lower than -18°C.

The caustic soda is pumped out of the tanker in a

storage tank truck.

It should be noted that with the above-mentioned

reactions, the ‘consumption’ of bicarbonate (HCO3

-

) differs.

With the application of caustic soda (NaOH), 1

mmol/l HCO3- is used for the removal of 1 mmol/l

calcium.

With lime (Ca(OH)2) that is 2 mmol/l and a dosing

of soda ash (Na2CO

3), no HCO3

- is used.

The extent to which HCO3

- is still present in water,,

after removal of the desired concentration of cal-

cium, is of importance regarding the bufferingis of importance regarding the buffering

capacity of water and several other water quality

parameters (copper resolution, corrosion index,

etc.).

In addition, it should be mentioned that with the

application of lime, twice the amount of calcium

carbonate is formed than with the application of

caustic soda or soda ash (thus more pellets).

When the sodium and calcium concentrations of

raw water are high, it will not be possible to softenthis water with caustic soda, because the sodium

standard of 120 mg/l will be exceeded. When raw

water has a low bicarbonate concentration, soften-

ing with lime will also not be possible.

In the Netherlands several softening installations

have been built in which different bases are used

for softening.

Figure 11 shows what base is used in Dutch prac-

tice. The gure uses lines to indicate the decrease

hardness reduction [Ca2+] (mmol/l)

0

6

0

H C O 3

- r a w

w a t e r ( m m o l / l )

5

1

2

3

4

5

1 2 3 4

NaOH

Na2CO3

Ca(OH)2NaOH

Na2CO3

Na2CO3

groundwater NaOH

surface water NaOH

groundwater Ca(OH)2surface water Na2CO3

Figure 11 - Application of base as a softening chemi -

cal in Dutch practice

NaOH Ca(OH)2

Na2CO

3

neutralization

CO2 -1 -2 -1

HCO3- 1 2 2

Ca2+ 0 1 0

Na+ 1 0 2

softening

CO2

0 0 0

HCO3- -1 -2 0

Ca2+ -1 -1 -1

Na+ 1 0 2

Table 6 - Change in water composition (mmol/l) per

mmol/l dosage of chemicals

165




Ca (mmol/l)

0

6

0

L ( m )

5

1

2

3

4

5

1 2 3 40

6

0

L ( m )

1.50

1

2

3

4

5

0.50 1.00

d (mm)

p (-)

0

6

0.5

L ( m )

1.0

1

2

3

4

5

0.6 0.7 0.8 0.90

6

1500

L ( m )

3000

1

2

3

4

5

1800 2100 2400 2700

S (m2/m3)

0

6

7

L ( m )

11

1

2

3

4

5

8 9 10

pH (-)

0

6

0

L ( m )

3.00

SI

1

2

3

4

5

0.50 1.00 1.50 2.00 2.50

Ca : decrease in calcium concentration is slower with the application of lime than with caustic

soda.

d : softening with lime is well-distributed over the reactor height.

p : porosity is directly dependent on the diameter of the particles.

S : specic area has a maximum at a certain height. With lime the maximum area is at a higher

point.

pH : the pH of water increases initially with dosing lime as a result of dissolving the Ca(OH)2

particles (lime dissolves faster than softening takes place).

SI : see Ca (as a result of chemical dosage, the supersaturation increases)

Figure 12 - Softening with caustic soda (red) or lime (blue)

NaOH

Ca(OH)2

166




Often, more storage tanks are present.

Safety regulations prescribe that the storage tanks

be placed apart from the softening installation in

a reservoir with a volume of at least one storage

tank plus 10% of the total storage capacity.

When caustic soda is diluted with partially sof-

tened water, calcium carbonate will be formed

and needs to be removed from the storage tanks.

Calcium carbonate will not precipitate in the stor-

age tanks when demineralized water is used for

dissolution. The demineralized water is prepared

with ion exchangers.

From the storage tanks caustic soda is pumped

into a dosing installation. This dosing installationcan be one single dosing pump per pellet reac-

tor or a caustic soda ring pipe. Caustic soda is

injected using a nozzle at the bottom of the pellet

reactor.

A nozzle is a specially designed dosing element

for the equal distribution of the chemical. Figure

13 shows the Amsterdam nozzle, Figure 14 shows

the working mechanism.

Because water and caustic soda ow through the

Amsterdam nozzle, a false bottom design is uti-

lized in the reactor. Under this bottom, the water is

present. Between the bottom plates caustic soda

is present, and above the bottom plates the actual

reactor starts.

Water to be softened ows through the nozzle into

the reactor. At the same time, but through another

channel, a concentration of caustic soda ows

through the same nozzle into the reactor. When

the outow velocity of the water is sufcient (1-9

m/h), a good mixing of the chemical and water

takes place.

Figure 13 - The Amsterdam nozzle

lye chamber

in false floor

influent

lye

water v=1.9 m/s

detail injector (35 per m2)

Figure 14 - Schematic representation of an Amsterdam

nozzle

Figure 15 - Dosing system with separate caustic soda

and water dosing nozzles in the bottom

167




In addition to the Amsterdam nozzle, many more

dosing systems exist. There are systems with

separate dosing points in the bottom of the reac-tor, with an inow of water by separate water

nozzles.

An equal distribution of chemical (calcium or

caustic soda) and water over the bottom should

be taken into account in the design.

Especially for dosing caustic soda, sufcient dosing

points need to be realized. For every m2 of reactor

bottom, 30-40 nozzles need to be present.

Dosing on one point in the reactor, like a tangential

inlet, is exclusively possible with lime, because the

softening reaction is much slower.

Lime

Dosage of lime is, as far as operational manage-

ment is concerned (necessary installation + main-

tenance activities), more complicated than dosing

caustic soda.

Lime is a suspension that is less soluble than

caustic soda (solubility 1.7 g/l) and it needs to be

produced on location.There are several options for the production of

lime:

- quick lime (installation: dosing + hydration

installation, dosing + solution installation, dos-

ing installation)

- hydrated lime (installation: dosing + solution

installation, dosing installation)

- stable lime water (installation: dosing installa-

tion).

The ow of lime water can be up to 10% of the

total ow through the reactor.

Lime water in powdered form is supplied by tanker

trucks and stored in silos. From the lime silos,

powdered lime is transported to a production tank

(underneath the silo). In the production tank, lime

water is produced in the desired concentration.

Lime water dosage can take place with one dosing

pump per reactor or by using a ring pipe.

In Figure 17 a lime water dosing nozzle is shown,

and in Figure 18 the supply pipes of the lime dos-

ing nozzles are represented.

Equal resistance in every pipe is important. If that

is not the case an unequal distribution of chemicals

Figure 16 - Dosing system with separate caustic soda

and water dosing nozzles in the bottom

Figure 17 - Lime water dosing nozzle

Figure 18 - Supply pipes of the lime dosing nozzles

168




takes place at the bottom of the pellet reactor and

the softening process will be affected.

Another aspect of lime as a chemical is that

water, after leaving the softening reactor, has

an increased suspended matter concentration.

The deposit content is called carry-over and is a

result of:

- contamination of calcium hydroxide

- CaCO3 from calcium hydroxide preparation

- undissolved calcium hydroxide particles

- homogeneous nucleation (spontaneous pre-

cipitation not on seeding material) and pellet

erosion.

4.3 Construction alternatives for reac-

tors

Different types of reactors can be used for soften-

ing. There are cylindrical reactors and reactors with

varying diameters over height. In the Netherlands

,mainly cylindrical reactors are used. Two Dutch

versions are briey discussed:

- cylindrical reactor with at bottom (Amsterdam

reactor)

- cylindrical reactor with conical bottom part and

tangential inlet.

Due to the cylindrical form of the Amsterdam reac-

tor, homogeneous uidization occurs; mixing in

horizontal directions hardly occurs.

Pellet reactors are discharged several times a day.

To remove limestone grains several discharge

points are installed in the bottom. The discharge

of grains cannot take place in one central place,

otherwise a cone shape occurs in the reactor.During the discharge the dosing of caustic soda

is stopped to prevent loss of caustic soda in the

reactor.

Seeding material is brought in at about 1 m abovet about 1 m above

the bottom of the pellet reactor.

In the cylindrical reactor with a tangential inlet,

water is brought in at the bottom of the reactor

(mixing compartment) using a bafe to direct the

water ow. In the mixing compartment, the chemi-

cal is dosed and mixing takes place.

Limestone grains are discharged through a point in

the mixing compartment. At about 1 m above the

mixing compartment, seeding material is brought

into the pellet reactor.

Softened water leaves the pellet reactor through

an overow weir.

The function of the overow weir is to provide a

uniform abstraction of softened water from the

reactor by avoiding preferential ow paths. For a

uniform abstraction, a notched weir can be used

(Figure 22).

To prevent seeding material from ushing out,

the pellet reactor can have a widened upper part.

In this widened upper part, the upward velocity

Figure 19 - Different types of reactors

Figure 20 - Tangential ow of raw water

Figure 21 - Several pipes at the bottom of a pellet reac -

tor

169




will decrease and the seeding material will settle

back.

Several pipes are present at the bottom of the

reactor:

- a supply pipe for raw, not softened, water

- a supply pipe for dosing the chemical

- a supply pipe for seeding material

- a drain pipe for pellets.

All these pipes need to be well arranged in the

treatment building. Typically, pipes of different

colors are chosen to prevent mistakes.

The upper part of the softening installation needsto be constructed in such a manner that, with a

possible unequal inow of water, the water will

leave the softening reactor equally. Therefore, a

notched weir construction can be used as shown

in Figure 22.

4.4 Seeding material

Seeding sand storage takes place in a silo.

Seeding sand is dosed from the silo (by a vibrat-

ing gully, for example) into a seeding sand washer,

where small particles can be washed out.

For the benet of disinfection, it is also possible

to dose caustic soda into the washed seeding

sand.

The seeding material is usually disinfected to be

sure no bacteriological contamination of water

will occur. This disinfection of seeding material

takes place with caustic soda or chlorine bleach-

ing lye.

To prevent seeding material from washing out of

the reactor and affecting the next treatment proc-

esses with a ne fraction of seeding material, the

seeding material is washed before it enters the

reactor. Here, seeding material is brought into a

silo with a washing velocity higher than the velocity

in the pellet reactor. In this way, the nest fraction

of seeding material is removed.

Figure 22 - Notched water weir Figure 23 - Seeding sand storage silo with sand washer

underneath

170




The use of seeding sand is minimal.

The storage capacity is usually sufcient for a

number of months.

4.5 Pellet storage

Pellet storage can be managed in silos and con-

tainers.

The size of the pellet storage is dependent on

the pellet production and frequency of pellet col-

lection.

With the application of lime, twice the amount

of pellets are produced than are produced with

caustic soda.

The volume of a pellet silo is usually equal to theamount of pellets produced in one week.

The pellet silos are equipped with a drainage

system to drain water that comes with the pellet

discharge.

The pellets shown in Figure 24 consist of 99.5%

calcium carbonate. These pellets are brown in

color. The reason for this brown coloring is the

presence of only 0.5% iron in the pellets.

Figure 24 - Pellet storage silo

Further reading

• Unit processes in drinking water treatment,W.

Masschelein (1992), (ISBN 0 8247 8678 5) (635

pgs)

• Het kalkkoolzuurevenwicht opnieuw bezien

DHV (1983), (118 pgs)

171




172




dissolved substances

water

colloïds

concentrate

membrane

permeate

Micro- and

ultraltration

WA T

E R T R E A T M E

N T

WATER TREATMENT

12

surface water drinking water

pre-treatment filtration



Framework

This module explains micro- and ultraltration.

Contents


1. Introduction

2. Principle

2.1 Membrane material

2.2 Membrane module

2.3 Dead-end ltration mode

2.4 Inside-out ltration

3 Theory

3.1 Mass balance

3.2 Kinetics 3.3 Membrane fouling

3.4 Cleaning

4 Practice

4.1 Module design

4.2 Choosing a module design

5 Operation

5.1 Constant pressure or constant ux mode

5.2 Cross-ow ltration

5.3 Fouling prevention

174

MICRO- AND ULTRAFILTRATION WATER TREATMENT



1 Introduction

Membrane ltration is a treatment process based

on the physical separation of compounds from

the water phase with the use of a semi-permeable

membrane. Until recently membrane ltration was

regarded as a futuristic, expensive and complica-

ted treatment process. Because of the develop-

ment of the technique during the past years, the

process can be regarded as proven technology.

The quality of the permeate of a membrane ltra-

tion installation is excellent.

The costs of membrane ltration have strongly

decreased over the past ten years because of the

decreased costs of membrane elements.

Membrane ltration can be divided into two catego-

ries based on the pore sizes of the membrane:

- micro- and ultraltration (MF and UF) remove

colloidal substances and microorganisms

- nanoltration and reverse osmosis (NF and

RO) remove colloidal substances and microor -

ganisms but also dissolved substances like

micropollutants and ions.

Micro- and ultraltration remove substances from

the water phase by a sieve mechanism.

In Figure 1 an overview is given of the different l-

tration processes and the sizes of the compounds

removed. Also, an indication of the applied pres-

sure needed for the ltration process is given.

Microltration removes bacteria and the larger

viruses (down to a size of 0.05 µm).

Ultraltration also removes bacteria, but because

of the smaller pore size all the larger viruses

are removed. Also, all the colloidal particles are

removed by UF as long as the membrane is not

damaged.

The removal of suspended solids (measured as a

percentage of the feed concentration) of MF and

UF is at least 99%.

The removal of microorganisms is referred to in

log units. A removal of one log unit corresponds

to a 90% removal. The removal of 4 log units cor -

responds to a 99.99% removal.

In Table 1 the log removal capacity of MF and UF

is shown for different microorganisms.

The so-called molecular weight cut-off (MWCO)

can also be used as an indication of the ability

of membranes to reject compounds. MWCO is

dened as the molecular weight of spherical mol-

Figure 1 - Overview of different ltration processes and sizes of compounds removed

approximatemolecularweight

relativesize of materialsin water

treatment

size, µm0.001 0.01 0.1 1.0 10 100 1,000

100 200 1000 10,000 20,000 100,000 500,000

viruses bacils

dissolved salts algae

metal ions humic acids cysten sand

clay

ED and EDR

reverse osmosis

nanofiltration

ultrafiltration

microfiltration

conventional filtration processes

metal ionsarsenicnitratenitrite

cyanide

dissolved saltscalciumsaltssulfate saltsmagnesium saltsaluminum salts

virusescontagious

hepatitis

humic acidstrihalomethane

precursors

bacilssalmonellashigellavibrio cholerae

cystenprotozoagiardiacryptosporidium

silt ∆ P (bar)

0.01

0.05

0.1

5

30

175

WATER TREATMENT MICRO- AND ULTRAFILTRATION



ecules which are 90% rejected by the membrane’s

pores. The unit of MWCO is Dalton (1 Dalton is

the mass of one hydrogen atom = 1.66x10-27kg).

The MWCO for MF/UF is in the range of 10,000

to 500,000 Dalton (10 to 500 kD).

MF/UF for drinking water

In drinking water treatment, UF can be used in

different stages of the process:

- as a pre-treatment of surface water before inl-

tration in the dunes or as pre-treatment before

NF/RO ltration

- as treatment of backwash water from rapid

sand ltration

- treatment of surface water as the rst step indrinking water production.

Drinking water can be produced from surface water

with either a direct or an indirect process.

An indirect treatment is dened as a process dur -

ing which the water spends a certain residence

time in the sub-surface. The soil passage guar -

antees the bacteriological quality of the produced

drinking water.

With direct treatment (no soil passage), the bac-

teriological quality must be guaranteed by several

disinfection steps in the treatment process.

With a direct as well as an indirect treatment of

surface water, MF and UF can be used as the rst

step in the treatment process.

The goal of the pre-treatment is to remove sus-

pended solids, heavy metals, bacteria and viruses

in order to prevent pollution of the dunes, or to

prevent clogging of the NF/RO membranes. In

some cases, the MF/UF installation is preceded

by a conventional coagulation/occulation/oc

removal treatment process in order to reduce the

risk of membrane fouling. Because of the improved

membranes and the improved possibilities of

fouling control, only an inline coagulation in front

of the membranes will remain in the future as a

pre-treatment for MF/UF.

MF/UF for backwash water

In groundwater, high concentrations of ions (Fe2+,

Mn2+, NH4

+) are present as a result of the long

residence time in the sub-surface. These ions have

to be removed in order to produce drinking water.

Figure 2 - American advertising brochure for ultral -

tration

ParticleParticle size

(µm)

Log-elimination MF

(pore size 0.2 µm)

Log-elimination UF

(pore size 0.01 µm)

Protozoa

- Giardia Lamblia 5-12 6 6

- Cryptosporidium Parvum 4-7 6 6

Bacteria

- E.coli 0.5 - 2 5 5

- Pseudomonas 0.5 - 1.5 5 5

Viruses

- Enterovirus 0.02 0 4

- MS2 - virus 0.025 0 4

Table 1 - Log-removal capacity of MF and UF for different microorganisms

176




The treatment steps used are aeration and rapid

sand ltration. The backwash water of the rapid

sand lters is loaded with high concentrations of

iron hydroxide and biomass.

The backwash water can be concentrated by

ultraltration. The permeate of ultraltration can

then be used directly as drinking water or it can be

treated further in the existing groundwater treat-

ment process. In this way a signicant amount of

valuable water is saved.

2 Principle

The membrane is the barrier responsible for

the separation of compounds out of the water

phase.

The membrane is semi-permeable. The pore size

determines the removal of different compounds.The removed compounds remain at the raw wa-

terside of the membrane and accumulate on the

membrane.

Three water streams can be distinguished:

- the dirty water or raw water is called feed

water

- the water passing the membrane is called

the permeate or product water. This water is

particle free

- the water with the rejected particles is called

concentrate or retentate.

2.1 Membrane material

Most of the membranes used are synthetic mem-

branes made of organic polymers (also called

polymeric membranes).

The thermal, chemical and mechanical proper-

ties of the polymer determine the properties of

the material.

There are several techniques to produce mem-

brane materials. The production of membranes,

however, will not be discussed here.

2.2 Membrane module

If a membrane was produced as a single, at hori-

zontal plate, a very large area is needed for the

water production resulting in very high investment

costs. Therefore, membranes are purchased as

Figure 3 - Possibilities for the use of MF and UF for drinking water production

12

surface water drinking water

pre-treatment filtration

2. ultrafiltration of surface water

- ultrafiltration as a barrier for bacils and viruses

- change of filter phases

- adaptation of treatment neccesary

1. ultrafiltration of drinking water

- ultrafiltration as a barrier for bacils and viruses

- already a high quality of the raw water before ultrafiltration

- high flux ultrafiltration possible

dissolved substances

water

colloïds

concentrate

membrane

permeate

Figure 4 - Membrane and the different ows

177




compact modules with as much ltration area as

possible.

Different module designs are possible. It is pos-

sible to compare modules based on the specicsurface of the modules.

The specic surface is dened as:

memspec

2module

A n d L A

1VD

4

× π × ×= =

× π ×

in which:

Aspec = specic surface area

Amem = membrane areaV

module = volume module (m3)

n = number of membranes in module (-)

d = diameter of membrane (m)

L = length of membrane (m)

D = diameter of membrane module (m)

The aim of a good design is to create a large

membrane area in a small volume.

Most modules are cylindrical. The length of a

cylinder varies from 1 to 6 meters. The diam-

eter of the cylinder varies from 1 to 12 inches (1

inch=0.0254 m).

2.3 Dead-end fltration mode

In dead-end ltration, all the feed water is through

the membrane. The suspended solids remain

on the feed side of the membrane. As a conse-

quence, the resistance of permeation will increase

in time.

The water flux decreases if the pressure is

constant, or the pressure increases if the ux is

constant.

Periodically the membrane has to be cleaned in

order to reduce the resistance of permeation.

To clean the membrane different methods are

used, which are described further on.

The period of permeation is called ltration time.

A ltration run is the ltration time together with the

cleaning time (also called ltration cycle).

2.4 Inside-out fltration

In a conguration with inside-out ltration, feed

water enters the inside of the capillaries or tubular

membranes. The water is pushed from the inside

to the outside of the membrane. Permeate is col-

lected outside the membrane and transported to

the permeate tube.

3 Theory

3.1 Mass balance

For dead-end ltration the following mass balance

Figure 5 - Membranes put together in modules Figure 6 - Principle of dead-end ltration modules

time

flux

resistance

permeateinlet

filtration

178




can be dened:

f pQ Q=

in which:

Qf = feed ow (m3/h)

Qp = permeate ow (m3/h)

For a ltration run, the mass balance is:

f p bwQ Q Q= +

in which:

Qbw

= backwash ow (m3/h)

Recovery

The recovery is the amount of permeate divided

by the amount of feed water used.With dead-end ltration the recovery is, of course,

100% during the ltration time. All the feed water

is recovered as permeate during this period.

But for a ltration run (ltration and cleaning), the

recovery is less than 100% because the perme-

ate is used for backwashing the membranes.

The recovery is now dened as:

p bw

p

V V

Vγ

-=

in which:

γ = recovery (-)

Vp = volume of produced permeate (m3)

Vbw = volume used for backwash (m3)

In order to achieve a high recovery (>90%),

the ltration period should be extended and the

backwash should be carried out with a minimum

amount of permeate.

3.2 Kinetics

The most important process parameter in MF- and

UF installations is ux.

Flux is dened as the water ow through a square

meter of membrane surface.

= =ν ⋅mem tot

Q TMPJ

A R

in which:

J = ux (m3/(m2.s))

Q = volume ow (m3/h)

Amem

= membrane surface area (m2)

TMP = trans membrane pressure (Pa)

ν = dynamic viscosity (Pa/s)

Rtot

= total resistance (m)

Water passes through the membrane under the

influence of pressure. The pressure difference

across the membrane is called Trans Membrane

Pressure (TMP).The temperature of the water inuences the ux

at a certain TMP. Each degree of temperature

(ºC) increase gives 3% more flux at the same

pressure. When the temperature of the water

changes (e.g., with surface water or wastewater),

the ux has to be normalized by:

1,5

ref cor measured 1,5

measured

(42.5 T )J J

(42.5 T )

+= ×

+

Figure 7 - Principle of inside-out ltration

feed

permeate

179




in which:

Jcor

= ux corrected for temperature

(l/(m2.h))

Jmeasured

= ux measured at temperature T

(l/(m2.h))

Tref = reference temperature (°C)

Tmeasured = measured temperature (°C)

In order to compare uxes of different installa-

tions, the ux is also normalized for the applied

pressure (TMP).

Because the ux is linear, depending on the pres-

sure, the normalized ux is:

r ef norm cor

measured

PJ J

P= ×

in which:

Jnorm

= normalized ux (l/(m2.h))

Pref

= reference pressure (bar)

Pmeasured

= actual pressure (bar)

Trans membrane pressure

The trans membrane pressure (TMP) is the pres-

sure difference between permeate and the feed

side of the membrane expressed in bar (Figure

8).

hydr

f perm

PTMP P P

2

∆= − −

in which:

Pf = feed pressure (Pa)

Phydr

= hydraulic pressure loss (Pa)

Pperm

= permeate pressure (Pa)

The hydraulic pressure loss in an ultraltration

module is small and can be ignored.

The permeate pressure needed to transport the

permeate is rather small (0.1 bar).

The pressure on the feed side of the MF/UF mem-

brane is typically 0.5 bar.

3.3 Membrane fouling

During ltration the resistance increases as a result

of fouling of the membrane surface. The resis-

tance increases because the pores in the mem-

brane are blocked and because caked suspended

matter is built up on the membrane surface. This

resistance increase is referred to as fouling.

The denition of fouling, given by the IUPAC, is: the

deposition of suspended or dissolved substances

on the membrane surface or in front of the pores

or in the membrane pores.

From this denition it is clear that fouling can be

subdivided into different mechanisms. In Figure 9

different resistances are dened:

- membrane resistance

- pore blocking

- adsorption in the pores

- cake resistance

- high concentration of dissolved substancesnear the surface.

The sum of all resistances is the total resistance

(Rtot

). Due to the accumulation of solids on and

in the membrane during dead-end ltration, the

total resistance increases with time. If the Rtot

-time

relation is known, the ux of an installation can be

calculated. Prediction and minimization of the total

resistance is an important research topic.

Figure 8 - Pressure difference between permeate and

feed size

Pperm

P f P c

0.25c

2c

c

ccf hydr

Re0.316λ

Lvrd2

λPPDP

−

⋅=

⋅⋅⋅=−=

permhydr

f permcf P

2

PPP

2

PPTMD −

∆−=−

+=

dead-end filtration

⋅

180




Membrane resistance

As a new membrane is permeated with deminer -

alized water, the measured resistance is only the

membrane resistance. There are no particles in the

water to block the pores or to form a cake layer.

The ux measured as a function of pressure gives a

linear relation. From this the membrane resistance

can be calculated. The membrane resistance can

also be calculated using the theory of water ow

through a packed bed (Hagen-Poiseuille):

τm 2

pore

8 lR

p d

× ×=

×

in which:R

m = membrane resistance (m)

p = porosity of the membrane (-)

dpore

= diameter of a pore (m)

τ = tortuosity of the pores (-)

l = thickness of the membrane (m)

The Rm of MF/UF-membranes is in the range of

1011-1014 m-1.

Sometimes the permeability of the membrane is

used rather than the membrane resistance.

The permeability constant K is dened as:

m

1K

R=

One of the goals of membrane manufacturers is

to produce membranes with a high permeability

together with a high rejection of the target com-

pounds.

Adsorption, pore blocking and cake forma-

tion

Accumulation of compounds on the membrane

surface and in the pores is a consequence of the

rejection of these compounds by the membrane.

With synthetic water (made from demineralized

water with added compounds), the different

mechanisms can be distinguished.

Because a range of compounds are present in thefeed water, adsorption, pore blocking and cake

formation will occur at the same time, and it is not

possible to distinguish the different mechanisms.

Therefore, the theoretical approach behind these

resistances is presented together.

Filtration model

The cake formation model is based on the as-

sumption that the feed water has a constant

concentration of particles with a constant size

and shape. The cake resistance is calculated

from the specic cake resistance (the specic

cake resistance is constant because the particle

concentration in the feed is constant) multiplied by

the cake thickness:

c c cR l r = ×

in which:

Rc

= cake resistance (m)

lc = thickness of the cake layer (m)r

c = specic cake resistance (m-2)

The Kozeny-Carmen relation gives the specic

cake resistance:

2

c 2 3

s

(1 )r 180

d

ε

ε

-= ×

×

in which:

ε = porosity of the cake layer (-)

ds = diameter of a particle (m)Figure 9 - Resistance processes

181




The thickness of a cake layer is given by:

ρ ε

sc

s mem

ml

(1 ) A=

× - ×

in which:

ms = cake mass (kg)

ρs = density of the particles (kg/m3)

The mass of a cake layer is difcult to measure.

The thickness of the cake layer also depends on

the TMP. The thickness of the cake layer is in the

range of several micrometers, depending on the

rejected compounds.

3.4 Cleaning

As a result of the dead-end mode, the membrane

has to be cleaned often in order to remove the

rejected compounds. The cleaning intervals can

be constant in time or can be determined by a

maximum pressure.

If possible, cleaning of membranes should be

avoided because during the cleaning no permeate

is produced. Also, permeate and energy are used

for the cleaning. With specic cleanings chemicals

are also used.

Different methods or a combination of methods can

be used to clean a membrane module:to clean a membrane module::

- forward ush (FF)

- back ush (BF)

- air ush (AF)

- chemical enhanced ush (CEF) or enhanced

back ush (EBF)- cleaning in place (CIP) or chemical soaking

After a cleaning the clean water resistance (CWR)

is measured in order to measure the effect of

the chemical cleaning. The CWR is obtained by

measuring the ux of demineralized water at a

certain pressure. By comparing the CWR of a

cleaned membrane with the CWR of the unused

membrane, the cleaning can be judged.

The Reynolds number is an indication of the

turbulence of the ow. If the Reynolds number

is smaller than 2300, the ow is laminar and the

shear at the membrane wall is low. If the Reynolds

number exceeds 2300, then the ow is turbulent

and accumulated compounds may be removed

from the membrane surface.

0 h0

h

v d ReRe v

d

ν

ν

× ×= Þ =

in which:

Re = Reynolds number (-)

v0 = cross-ow rate (m/s)

dh = hydraulic diameter (m)

With tubular or capillary membranes, the hy-

draulic diameter is equal to the diameter of the

membrane.

Forward ush

Particles and compounds on the membrane sur -

face can be removed with a forward ush. The

forward ush is a turbulent cross-ow along the

feed side of the membrane surface (Figure 10).

This is the opposite of the ltration mode where

the ow is through the membrane (ow direction

perpendicular to the membrane surface).

In Table 2 velocities are shown where a turbulent

ow at 10oC is obtained with different, commer -

cially available membrane sizes. Also, the needed

pressure difference is calculated.

From this table it is clear that with the smaller

diameters, high cross-ow rates are needed to

obtain turbulent ow. This velocity is many timeshigher compared to the velocity during dead-end

ltration. For a forward ush, feed water can be

used to obtain a high recovery.

Figure 10 - Principle of forward ush

flush water

forward flush

feed

182




Back ush

The back ush or backwash resembles the back-

wash of a rapid sand lter in the conventional

treatment. The ltration direction is reversed so

the ltration is now outside in (Figure 11). Perme-

ate is used for the backwash in order to keep the

permeate side of the membrane free of particles.With permeate the dirt is removed from the pores

and from the membrane surface. The backwash

ux is 2 to 2.5 times the ux during ltration.

After removing the particles from the pores and

the membrane surface, the particles and the cake

have to be transported out of the module. Because

the amount of permeate used for a backwash is

limited (because of the recovery), the transport

of dirt may be insufcient. A combined back ush

and forward ush can be used to overcome this

problem. First, a back ush is used to clean the

pores and to lift the cake. Then, a forward ush is

used to transport the dirt out of the module.

With the backwash, the recovery of the system

decreases because permeate is used to remove

the accumulated compounds.

Air/water ush

An air/water ush can be used to clean the mem-

brane wall from adhering fouling. The air/waterush is commercialized as AirFlush and is actually

a forward ush with a combination of air and water

(Figure 12). The air is used to create a turbulent

ow in the membrane under process conditions

where no turbulence is attained with the water

ow.

The cleaning efciency depends on the kind of two-

phase ow obtained in the membranes (Figure 13).

If the water/air ratio is high, only small air bubbles

Figure 11 - Back ush schedule

flush waterback flush

product

inlet

back flush with forward flush

product

flush water

d (mm)

Rear ow at the

end of a module

(m/s)

Required time

for ushing a

module (s)

5.2 0.05 19

1.5 0.19 5

1.0 0.28 4

0.7 0.40 3

Table 3 - Cross-ow rate at the rear end of a 1 meter

module with a back ush ux of 250 l/(m 2 •h)

water

air

Figure 12 - Principle of air ush

Diameter

(mm)

Cross-ow rate

(m/s)

Pressure difference

(Pa)

5.2 0.58 1473

1.5 2.01 61,370

1.0 3.01 207,120

0.7 4.3 603,850

Table 2 - Needed cross-ow rate in order to get turbu-

lence (L = 1 m)

183




are present in the water and the turbulence is only

slightly enhanced. When the water/air ratio is too

small, the air ows through the middle of the mem-

brane (chimney effect) and the cleaning effect is

low. The best cleaning is obtained with bullet-like

air bubbles (Figure 13).

Chemical cleaning

If forward ush, back ush and air ush are not

enough to clean the membrane, a chemical clean-

ing (often called enhanced back ush or chemical

back ush) can also decrease the clean water

resistance. There is, however, a cost factor toconsider with the use of chemicals.

This kind of cleaning means that the module is

soaked with a solution of hypochloric acid, hydro-

gen chloride, hydrogen peroxide or a specially

developed mixture of chemicals. After the soak-

ing, a backwash or a forward ush removes the

dissolved dirt. The main drawback of chemical

cleaning is that the membranes age because of

the chemicals, and the lifetime of the membranes,

therefore will be shortened. Also, the chemicals

are a cost factor and with a chemical cleaning a

chemical waste stream should be discharged.

Besides the periodical chemical cleaning which

is part of the automated process control of the

installation, a more intensive chemical cleaning

might also be necessary. The so-called “clean-

ing in place” (CIP) can last from a few hours to

several days and is typically not automated. If the

CIP is not able to clean the membranes, they are

replaced by new ones.

4 Practice

4.1 Module designThere are several module concepts. In a module

design, a large surface area (a high specic sur -

face area results in low investment costs) is com-

bined with a low fouling behavior (clogging results

in high cleaning and replacement costs).

The membrane manufacturers commercialize

several module designs. Two different systems can

be distinguished: the tubular-shaped membranes

and the at sheet membranes. These two systems

are described further below in Table 4.

Tubular-shaped membranes

In a module, one to several thousand membranes

are combined. The rear ends of the membranes

are glued in the module (this is called potting) in

order to x the membranes in the module (Figure

14). The potting is 2 to 3 centimeters thick.

Tubular membranes

Tubular membranes have a diameter of more

than 5 mm.Tubular membranes are not self-supporting. The

membrane material is not strong enough to resist

the pressure during ltration and especially not

during the backwash (the backwash pressure is

Figure 13 - Different types of air bubbles in waterwith

the type in the middle providing the best

clearing

Tubular- shaped mem-

branes

Flat sheet membranes

Tubular membranes Plate membranes

Capillary membranes Cushion membranes

Hollow ber membranes Spiral-wound membranes

Table 4 - Subdivision of different MF/UF systems

184




higher and is outside-in).

The membrane is therefore xed on a support

layer.

The specic surface area of a module is low (about

400 m2/m3) because of the large diameter of the

tube. Because of this low specic surface area and

because the membranes are built with two layers

(membrane and support layer), the investmenthe investment

costs of these membranes are high..

The benet of a large diameter is that the mem-

branes are not very sensitive to fouling.

The application of these membranes is in water

environments with a high load of suspended solids

(backwash water from rapid sand ltration or other

wastewater) or in industrial locations.

Tubular membranes can be cleaned well. Because

of their large diameter, there are low cross-ow

rates required for turbulent conditions. A forwardush can clean the membrane surface because

of the turbulent conditions. Also, the tubular mem-

branes can be backwashed (also called back

ush).

Capillary membranes

Capillary membranes have a tubular shape and a

diameter between 0.5 and 5.0 mm.

The capillary membranes are self-supporting, so

they are strong enough to resist the pressure du-

ring ltration and backwash.

With the smaller diameter of the capillary mem-

branes, the specic surface area of a module is

large (about 2000 m2

/m3

). This inuences the in-vestment costs, which are low compared to tubu-

lar membranes.

The capillary membranes are more sensitive to

fouling because of the small diameter.

Capillary membranes can be backwashed but the

forward ush is less effective because the cross-

ow is only turbulent at very high velocities. There-

fore, the forward ush is used to transport the dirt

after a backwash rather than to remove the fouling

from the surface.

Hollow-ber membranes

The diameter of a hollow-ber membrane is only

about 100 micrometers.

Hollow-ber membranes resemble the diameter of

a human hair. Because of the small diameter, the

specic surface area of a module is very high (up

to 100,000 m2/m3), but at a large risk for clogging

is high.

Hollow-ber membranes are not backwashed.That is why these membranes are only used

with reverse osmosis and not for micro- or

ultraltration.

Flat sheet membranes

Spacers separate at sheet membranes from each

other. Spacers and membranes are put together

alternately.

Plate membranes

Figure 14 - Tubular membranes

Figure 15 - Capillary membranes

185




Plate membranes are alternately piled together

with spacers (membrane, feed spacer, membrane,

permeate spacer, membrane, etc.). The feed

spacer is also used to create turbulence in the

feed channel to prevent fouling.

In a module a large number of membranes are

put together, but the specic surface area remains

rather low (about 100-400 m2/m3), resulting in high

investment costs.

The sealing of the membranes in the modules is

also a weak point in this membrane design.

Plate membranes are seldom used in drinking

water production or wastewater treatment.

Cushion membranes

A modification of the plate membrane is the

cushion membrane. A spacer is placed between

two membranes; the edges of the membranes

are glued together resulting in a cushion shape.

A permeate tube is xed through the membrane

and the spacer.

Feed water is forced outside-in through the mem-

branes and is collected on the inside of the cushion

and transported through the permeate tube.The specic surface area of a cushion module is

100 to 400 m2/m3, depending on the distance be-

tween cushions. The distance can be adapted to

the quality of the feed water.

A cushion module can be cleaned with both a

forward ush and with a back ush.

Spiral-wound membranes

In spiral-wound membranes several at sheet

membranes are wound around a central perme-

ate tube. The distance between two membranes

is small (0.25 to 1.0 mm), and membrane clogging

is a serious problem in the feed spacer.

Spiral-wound membranes are not backwashed.

This module design, therefore, is not used in MF/

UF, but only in NF/RO.

4.2 Choosing a module design

spacer

membrane

support plate

membrane

spacer

Figure 16 - Plate membranes

permeate

transportmembrane

carrier plate

feed

Figure 17 - Cushion module

Tubular Capillary Hollow ber Plate Cushion Spiral wound

diameter feed-

ing channel

(mm)

5-25 0.5 - 5.0 0.1 - 0.5 1 - 3 1 - 3 0.25 - 1.0

inuentoutside - in

inside - out

inside - out

outside - ininside - out outside - in outside - in outside - in

cleaning pos-

sibilitygood good not not not not

specic area

(m2/m3)< 80 < 800 < 1000 100 - 400 1000

constipation

sensitivity low high high low low high

Table 5 - Overview of different membrane congurations

186




The choice of a module design will be determined

by economical reasons.

There is a difference between investment costs

and exploitation costs. The investment costs are

minimal with modules having the highest specic

surface area and low module costs. The exploita-

tion costs are minimal at low energy costs and a

high fouling resistance.

Depending on the type of feed water, an eco-

nomical conguration can be found. In many cases

comprehensive research is needed in order to nd

an optimal conguration.

In Table 5 an overview is given of different mem-brane congurations and the main criteria for an

optimal choice of a membrane design.

5 Operation

5.1 Constant pressure or constant ux

mode

Dead-end ltration can be performed in two modes:

with a constant ux or with a constant pressure.

With a constant ux mode the pressure is increas-

ing in time.

In constant pressure mode the ux is decreasing

in time (Figure 18).

Constant pressure mode is not preferred because

water production is not constant. It is better to in-

crease the pressure during permeation to keep

the ux (and the production) constant.

The backwash can be started either at a constant

time or if a maximum pressure is reached. Theltration time in dead-end mode depends on the

suspended solid concentration, usually 15 to 20

minutes. Cleaning lasts several seconds to one

minute.

Depending on the type of cleaning, feed water (for -

ward ush) or permeate (backwash) is used. The

pressure during a cleaning is in the range of 0 to

1 bar. For the treatment of surface water, a ux of

70 l/(m2.h) is used. Backwash water of rapid sand

ltration is treated with a ux of 120 l/(m2.h).

5.2 Cross-ow fltration

For water with a high suspended solids concentra-

tion, often cross-ow ltration is used.

With cross- ow ltration the majority of the water(90%) ows across the membrane and a small

part permeates through the membrane (10%).

The cross-ow rate is high because the cake layer

thickness can be minimized, but the permeate

production is low. Particles on the membrane are

removed by the high cross-ow rate and, therefore,

removed from the module.

The drawback of a cross-ow mode is that it uses

more energy compared to the dead-end mode.

This energy is used to pump 90% of the feed water

across the membrane. The energy consumption of

a cross-ow system is about 5 kWh/m3 permeate.

For dead-end ltration the energy consumption is

only 0.1 to 0.2 kWh/m3 permeate.

The typical ux-time diagram for cross-ow ltra-

tion is drawn in Figure 20. The ux decreases as

a function of time which is a result of the cake

build-up and the pore blocking. Because of the

high cross-ow rate, the cake layer thickness isconstant after a while and the ux does not de-

time

constant flux constant pressure

time

flux

TMD

Figure 18 - Constant pressure versus constant ux

mode

Figure 19 - Flows in cross-ow ltration

permeate

concentrate

feed

187




crease as fast as in the dead-end mode.

Critical ux is the ux achieved at a certain cross-

ow rate. At this rate the cake layer has a certain

thickness. If the cross-ow rate is increased, the

cake layer decreases as a result of the high shearstresses and the ux increases (Figure 21). The

increase in ux is rather small. Above a certain

cross-ow rate the ux will become constant.

Membrane systems with cross-ow mode are also

cleaned. Backwash and chemical cleaning are

used in the same way as in a dead-end system.

5.3 Fouling prevention

In order to protect the pores of the membranes

from blocking iron or aluminum, coagulation can

be used. Coagulant dosing is used to make larger

particles incapable of penitrating the membrane

and can, therefore, be removed more easily.

In Figure 22 the ux decrease is shown (constant

pressure mode) for two UF modules. One is fed

with coagulated water and the other with non-coagulated water. Because the smaller particles

are captured in the iron ocs, the production is

higher in the module with coagulant compared to

the module without coagulant.

Figure 21 - Flux at different cross-ow veloci -

time (min)

f l u x ( l / m 2 h )

0 50 100 150 200 250

140

120

100

80

60

4020

0

vcr= 2.4 m/svcr= 1 m/s

19-02 26-02 5-03 12-03 19-03 26-03 2-04 9-04 16-04 23-04 30-04 7-05

date

0

200

400

600

800

1000

1200

f l u x ( l / m 2

h a t 1 0 0 C a n d 1 b a r )

no flocculation aid added FeCl3 added

Figure 22 - Flux decline with and without FeCl 3-dosing

fouling

concentrationpolarization

time

f l u x

Figure 20 - Flux decline with cross-ow ltration

Further reading


(2005), (ISBN 0 471 11018 3) (1948 pgs)

188




hyperfiltration element

RO-product

product spacer

membrane

glue seam

feed spacer

Nanoltration

and reverse

osmosis

WA T

E R T R E A T M E

N T

WATER TREATMENT

Qf , p

f , c

f

QP

, pP

, cP

QC

, pC

, cC

membrane

concentratefeed

permeate

M em b r a an

Permeaat

c P

δ

J·c P

δ



Framework

This module examines nanoltration and reverse osmosis.

Contents


1. Introduction

2. Principle

2.1 (Reverse) osmosis

2.2 Fouling of membranes

2.3 Membrane conguration

2.4 Feed, permeate and concentrate

2.5 Cross-ow operation

3. Theory

3.1 Mass balance

3.2 Kinetics

3.3 Concentration polarization

4. Practice

4.1 Nanoltration

4.1 Christmas tree conguration

4.2 Cleaning

4.3 Field installations

190

NANOFILTRATION AND REVERSE OSMOSIS WATER TREATMENT



1 Introduction

Reverse osmosis is one of the membrane ltration

processes. The process is used to remove salts

and organic micropollutants from water.

Because reverse osmosis is able to remove very

small particles from water, fouling of the membrane

can easily occur. Reverse osmosis is therefore al-

ways preceded by a pre-treatment step to remove

particulate matter. This pre-treatment can be a

conventional pre-treatment (coagulation, occula-

tion, sedimentation, ltration) or an ultraltration

pre-treatment.

In reverse osmosis almost all dissolved particlespresent in water will be retained, so the produced

ow (permeate) has a low mineral content. There-

fore, the permeate is sometimes conditioned (lime-

stone ltration or aeration) to correct the pH and

the aggressiveness of the permeate.

In nanofiltration almost all divalent ions are

retained; the monovalent ions are only partly

retained.

2 Principle

2.1 (Reverse) osmosis

Osmosis is a natural process of ow through a

semi-permeable membrane. When pure water of

the same temperature is present on both sides of a

membrane and the pressure on both sides is also

equal, no water will ow through the membrane.

However, when the salt on one side is dissolved

into the water, a ow through the membrane from

the pure water to the water containing salts will

occur (Figure 1, left and middle). Nature tries to

equalize concentration differences.

When pressure is applied on the side where the

salts are added, a new equilibrium will develope.

The extra pressure will result in a ow of water

through the membrane, but the salts do not ow

through.

This phenomenon is called reverse osmosis (Fig-

ure 1, right).

The driving force for reverse osmosis is the applied

pressure minus the osmotic pressure.The energy consumption of reverse osmosis is

directly related to the salts concentration, since

a higher salt concentration has a higher osmotic

pressure.

2.2 Fouling of membrane

The fouling of a reverse osmosis membrane is

almost inevitable.

Particulate matter will be retained and is an ideal

nutrient for biomass, resulting in biofouling.

Another important fouling process is scaling, the

formation of salt precipitates.

Both fouling processes (scaling and biofouling)

should be avoided as much as possible to ef-

ciently operate reverse osmosis.

osmoticpressure

reverse

osmoticpressure

semi-permeablemembrane



purewater

saltsolution

purewater

saltsolution

purewater

saltsolution

Figure 1 - Principle of osmosis and reverse osmosis

191

WATER TREATMENT NANOFILTRATION AND REVERSE OSMOSIS



2.3 Membrane confguration

The application of large at membranes is not

practical, because a large footprint is needed to

obtain the necessary permeate production. There-

fore, a system is used with a high specic surface

(membrane area per volume).

Spiral-wound membranes

Almost all reverse osmosis membranes are of the

spiral-wound conguration.

Water is fed from one side into a module. Via

spacers (supporting layers between membrane

sheets), the water is distributed over a membrane

element. An element is a number of membrane

sheets twisted around a central permeate collect-ing tube (Figures 2 and 3).

The length of a membrane element is normally

one meter, so one person can replace it from the

installation. After passing one element the water

ows to a second element.

To withstand the high operating pressures, a pres-

sure vessel (membrane module) is used. It is not

economically feasible to have a pressure vessel

for every element and, therefore, six elements

are generally placed in one membrane module

(Figure 4).

Spiral-wound membranes have a large specic

area (1000 m2/m3).

A disadvantage of spiral-wound membranes is that

rapid fouling of the spacer channels with particu-

late matter can occur.

Reverse osmosis membranes cannot be hydrauli-

cally cleaned like ultraltration membranes and

fouling of the membranes should therefore be

avoided.

2.4 Feed, permeate and concentrate

In membrane ltration processes, three different

types of ow are distinguished.

The feed ow is separated by the membrane into a

permeate (or product) ow and into a concentrate

(or retentate) ow.

The salt concentration in the permeate ow is lower

than the salt concentration in the feed ow.

In the concentrate ow the salt concentration is

higher than in the feed ow.

Figure 2 - Open spiral-wound membrane Figure 4 - Membrane modules

Figure 3 - Principle of spiral-wound membranes

hyperfiltration element

RO-product

product spacer

membrane

glue seam

feed spacer

192




It is not possible to have an unlimited concentra-

tion of salts in the concentrate ow, because atcertain salt concentrations precipitation of salts

will occur.

2.5 Cross-ow operation

Reverse osmosis modules are always operated in

cross-ow mode (Figure 5).

This means that only a small part of the feed ow

is produced as permeate (between 1 and 10% per

element), while most of the feed water ows along

the membrane surface and exits the membrane

element as concentrate.

Because of this large concentrate ow, the velocity

in the membrane channels is high and the build up

of a laminar boundary layer is disturbed.

3 Theory

3.1 Mass balance

The water mass balance for a membrane elementis given by:

Q Q Qf c p= +

in which:

Qf = feed ow (m3/h)

Qc = concentrate ow (m3/h)

Qp = permeate ow (m3/h)

Also, the dissolved material of mass balance (Fig-

ure 6) can be derived by:

Q c Q c Q cf f c c p p= +

in which:

cf = concentration of dissolved material in feedwater (g/m3)

cc = concentration of dissolved material in con-

centrate (g/m3)

cp = concentration of dissolved material in per-

meate (g/m3)

Recovery

The recovery indicates the overall production of

the system.

It is the relationship between permeate and feed

ow:

p

f

Q100%

Qγ =

in which:

γ = recovery (%)

A recovery of 80% means that 80% of the feed

ow is produced as permeate.

This also means that the concentration of saltsin the concentrate is 5 times higher than the con-

centration in the feed ow, assuming that all salts

are retained.

The recovery of one element is between 1 and

10%, therefore more elements should be placed in

a series to obtain the desired recovery of 80%.

For sea water desalination, the maximum achiev-

able recovery is about 50%.

This recovery is limited by the possibility of scaling,

Figure 6 - Mass balance

Q f , p f , c f

QP

, pP

, cP

Q C , p C , c C

membrane

concentratefeed

permeate

Figure 5 - Cross-ow operation

≈

≈

≈

≈

≈

≈

0.10·Q

Q

0.90·Q

cover open cover shut pump

193




caused by high salt concentrations.

For groundwater, however, recoveries up to 95%

can be obtained.

Rejection

Rejection indicates the amount of material rejected

by a membrane.

Rejection is calculated by:

Re = -1c

c

p

f

in which:

R = rejection (-)

3.2 Kinetics

Flux

The ux is the permeate ow through one square

meter of membrane surface or:

TMPJ

K=

μ

in which:

J = volumetric ux (m/s)

K = membrane resistance coefcient (m-1)

μ = dynamic viscosity of water (Ns/m2)

TMP = transmembrane pressure (Pa)

The volumetric ux is often expressed as a “sur-

face load” (ow per area (l/h/m2)).

Transmembrane pressure

Water does not automatically ow through a mem-brane. The membrane has a resistance against

ltration and this resistance has to be overcome

by a pressure.

The net pressure difference over a membrane

is called the transmembrane pressure (TMPnet)

and acts as the driving force for a membrane

process.

The SI-unit for pressure is (Pa), however, in mem-

brane ltration processes, the more common (bar)

is used. One bar is equal to 105 Pa.

TMPnet is given by:

hydr

net f p

PTMP P P P

2

∆= ∆ − ∆π = − − − ∆π

in which:

Pf = pressure of feed (Pa))

ΔPhydr

= hydraulic pressure loss (Pa)

PP = pressure of permeate (Pa)

Δπ = osmotic pressure difference (Pa)

The hydraulic pressure loss is the difference be-

tween the pressure of the feed and concentrate,

or:

hydr f cP P P∆ = −

in which:

Pc = pressure of concentrate (Pa)

The TMPnet is dependent on place and time.

As can be seen in the TMP equation, these place

and time dependent effects are averaged.

Depending on the concentration of dissolved

material, the feed pressure for reverse osmosis

is between 15 and 70 bar.

The pressure in permeate is often or almost 0 bar.

The reason for this is the almost atmospheric con-

ditions for permeate outow.

Hydraulic pressure loss

Hydraulic losses occur in the water moving from

feed (inlet) to concentrate (outlet) as a result of

wall friction. Because of this wall friction, Pc willalways be smaller than P

f .

The friction loss in spiral-wound membranes can

be calculated by:

2hydr

H

dP v

dx 2 d

λρ=

⋅

in which:

λ = friction factor (-)

194




dH

= hydraulic diameter (m)

v = liquid velocity (m/s)

This friction loss is shown in Figure 7.

The friction factor λ for spiral-wound membranes

is given by:

0.36.23Re (100 Re 1000)−λ = < <

in which:

Re = Reynolds number (-)

For capillary membranes the following relation-

ship is used:

64 (Re 2000)Re

λ = ≤

0.250.316Re (Re 2000)−λ = >

At smaller diameters of the membrane channels,

the Reynolds number decreases and the friction

factor λ increases.

In spiral-wound membranes the membrane chan-

nels are rectangular and there are spacers present. A spacer is a special layer resulting in more tur-

bulence in the membrane channel and therefore

creates a ow of feed water to the membrane sur-

face.

The hydraulic diameter is dependent on the height

of the spacer. In most spiral-wound membranes,

a value of 0.9 mm for the hydraulic diameter is

common.

Hydraulic line

In Figure 8 the hydraulic line in an RO-module

containing one single element is depicted.

The storage tank with (1) feed water is open. After

the storage tank the hydraulic line decreases

slightly because of hydraulic losses in the feeding

pipeline.

By means of a pump, the water is pressurized; a

large increase in the pressure level is observed.

In the membrane module, a further hydraulic loss

occurs.

A valve is placed in the concentrate pipeline. This

valve regulats the driving force (TMP). A large

pressure drop takes place across this valve.

The concentrate ows into a second storage tank

(2).

The permeate, about 10% of the feed ow, ows

to tank 3.

From the permeate tank we calculate back. The

permeate transported to the tank encounters

hydraulic headlosses. A line has been drawn from

tank 3 to the membrane module.

Q

0.90 · Q

permeate

feed side module

concentrate

TMD

0.10 · Q

permeate side

cover open cover shut pump

1

2

3

Figure 8 - Hydraulic line at permeate side (lightblue line) and feed/concentrate side

(dark blue line)

)Pa(p

L

)m(x

2v

2

1

d

L⋅⋅⋅⋅ ρλ

Figure 7 - Hydraulic pressure loss

195




Osmotic pressure

Osmotic pressure is a uid property dependent on

salt concentration and temperature and independ-

ent of the presence of a membrane.

The osmotic pressure is calculated by:

i i

i

R T c z

M

⋅ ⋅ ⋅π = ∑

in which:

π = osmotic pressure (Pa)

R = gas constant (J/K.mol)

T = temperature (K)

ci = concentration ion (g/m3)

Mi = molecular weight ion (g/mol)zi = valence ion (-)

Valence is determined by the ion. Sodium has a

valence of 1 (Na+, z =1), chloride also (Cl-, z=1),

while carbonate has a valence of 2 (CO32-, z =2).

To calculate the osmotic pressure, it is sufcient

to take into account the most important in water

dissolved ions. These are HCO3

-, SO4

2-, Cl-, Na+,

Ca2+ and Mg2+.

Osmotic pressure difference

The osmotic pressure difference over a membrane

is given by:

f c

p2

π + π

∆π = − π

in which:

Δπ = osmotic pressure difference (Pa)

πf = osmotic pressure of feed (Pa)

πc

= osmotic pressure of concentrate (Pa)

πp

= osmotic pressure of permeate (Pa)

The pressure difference is averaged to be inde-

pendent of the position in the membrane and, thus,

there is no dependency of π on the position.

Example 2

In water from the IJsselmeer (18o), the following

ions are present at the given concentrations:

[HCO3

-] 135 g/m3 M = 61.0 g/mol

[SO42-] 63 g/m3 M = 96.1 g/mol

[Cl-] 95 g/m3 M = 35.5 g/mol

[Na+] 52 g/m3 M = 23.0 g/mol

[Ca2+] 60 g/m3 M = 40.1 g/mol

[Mg2+] 11 g/m3 M = 24.3 g/mol

R J K mol= ×8 314. /

Calculate the osmotic pressure of the

IJsselmeer water.

RTc zi i

Mi

π = ∑

8.314 (273 18)= ⋅ + ⋅

135.1 63.2 95.1 52.1 60.2 11.2

61 96 36 23 40 24

+ + + + +

50.3 10 Pa 0.3 bar = ⋅ =

By comparison, the osmotic pressure of brack-ish groundwater (2000 mg/l NaCl) is 1.7 x 10

5

Pa (= 1.7 bar), the osmotic pressure of sea

water (35.000 mg/l NaCl) is 30 x 105 Pa (= 30

bar).

Example 1

In a spiral-wound reverse osmosis membrane

module, six elements, each with a length of 1

m, are placed.

Calculate the hydraulic pressure loss per

module (v=0.25 m/s (average), dH=0.9 mm,

water temperature is 20oC).

Answer:

T= 20oC, so ν=1.0x10-6

3

6

0.25 0.90 10Re 225

1 10

−

−

⋅ ⋅= =

⋅

0.36.23 225 1.23−λ = ⋅ =

2

hydr 3

1P 1.23 1000 0.25

2 0.9 10−∆ = ⋅ ⋅ ⋅

⋅ ⋅

= =42603 0 43 Pa bar .

196




Example 3

Why is the osmotic pressure in the concentrate

higher than in the feed?

Answer

The feed is separated into permeate and

concentrate ows. The concentrate ow

contains the same amount of salts as the feed

ow, however, they are dissolved in less water.

A higher salt concentration means a higher

osmotic pressure.

Because the concentration of salts in the permeate

is very low, the osmotic pressure in the permeate

is almost always neglected.

On the other hand, the osmotic pressure of theconcentrate is higher than the osmotic pressure

of the feed.

The following equation is valid:

c f

1

1π = π

− γ

Combining this with what we saw before of the

osmotic pressure difference over a membrane,

we see:

f

2

2 (1 )

− γ∆π = π ⋅

⋅ − γ

3.3 Concentration polarization

During ltration a concentration build-up of the

retained material will occur in the boundary layer

close to the membrane.

This effect is called concentration polarization and

results in an initial rapid decline in ux.However, this decline will not continue in time, like

in the case of fouling (Figure 9). Concentration

polarization is reversible and will disappear as the

driving force becomes zero.

The concentration polarization can be limited by

disturbance of the boundary layer, for example, by

enhancement of the velocity along the membrane

surface.

The relationship between concentration close to

the membrane surface and in the feed (Figure 10)

is represented by the concentration polarization

factor β which is given by:

m p

v p

c c Jexp

c c D

− δβ = =

−

in which:

β = concentration polarization factor (-)

cm = concentration at membrane surface (mg/l)

cp = concentration in permeate (mg/l)

cf = concentration in feed (mg/l)

J = permeate ux (m3/m2.s)

δ = thickness of boundary layer (m)

D = diffusion coefcient (m2/s)

Because cp << cf < cm, cp can be neglected, and

when coefcient k is taken for the mass transfer,

the following relation can be used:

Dk =

δ

in which:

k = mass transfer coefcient (m/s)

Figure 9 - Concentration polarization and fouling in

time in a cross-ow operation

flux concentrationpolarization

fouling

time

Figure 10 - Concentration polarization

M em b r a an

Permeaat

c P

δ

J·c P

δ

197




Then β can be rewritten to:

m

f

c Jexp

c kβ = =

Concentration polarization results in a higher os-

motic pressure difference across the membrane.

Scaling

Scaling can occur when the transport of salts, as

a result of convection to the membrane, is larger

than the transport of salts from the membrane by

diffusion.

Scaling is the precipitation of inorganic saltscaused by exceeding the solubility product (super

saturation).

Whether scaling occurs depends on numerous

factors, like pH, temperature and the presence

of other ions.

Super saturation is defined by the saturation

index SI:

SIIP

KSP

= log

in which:

SI = saturation index (-)

KSP

= solubility product salt (mol/l)

IP = ion product (mol/l)

The solubility product KSP is temperature depend-

ent.

The value of the ion product of a salt is determined

by the ion strength, pH and the ion afnity.

Scaling can be prevented by the dosing of acids or

anti-scalants, by removal of seeding material and

by not exceeding the solubility product.

Limiting the concentration polarization layer by

increasing the cross-ow velocity helps to prevent

exceeding the solubility product. However, this

results in a higher energy consumption.

4 Practice

4.1 Nanofltration

It is not always necessary to remove all dissolved

ions. For example, when water has to be softened

nanoltration will be sufcient.

Nanoltration removes divalent ions (like Ca2+ ,

Mg2+ and SO4

2-), while monovalent ions are not

rejected.

Nanoltration membranes have larger pores than

reverse osmosis membranes, resulting in a lower

resistance for ltration and also lower operational

pressures (2 - 10 bar).

The pores of nanoltration are smaller than ultra-

ltration pores.

Nanoltration membrane modules can be con-

structed as spiral-wound membranes, and now

as capillary membranes as well.

4.2 Christmas tree confguration

To obtain a high recovery, several modules are

placed in a series in an RO/NF-membrane ltra-

tion installation to concentrate the concentrate

even further.

This in-series placement of membranes is called

staging. Normally, two to three stages are used.

The osmotic pressure in the rst stage will always

be lower than the osmotic pressure in the second

stage; the osmotic pressure in the second stage

will always be lower than in the third stage.

It is clear that when scaling occurs this will be in

the stage where the concentrations are highest.

To prevent scaling, the cross-flow velocity inthe last stage should be higher than in the rst

stages. Therefore, a Christmas tree conguration

is often used. The number of modules in a stage

decreases when the stage number increases. So,

for example, in the rst stage there are three mod-

ules, in the second stage there are two modules,

and in the third stage there is only one module

(Figure 11).

198




4.3 Cleaning

To prevent a ux decrease (or increase in TMP) inspiral-wound NF/RO-systems, different techniques

can be used:

- dosing of acids and anti-scalants

- chemical cleaning

- increasing the cross-ow velocity by recircula-

tion.

Some anti-scalants have biofouling (growth of mi-

croorganisms in a membrane module, resulting in

ux decrease) as their side effect, especially when

the anti-scalants are not 100% pure and contain

some organic material.

When the ux at a certain standard TMP becomes

too low, the membrane is cleaned chemically.

Depending on the type of fouling (biofouling, scal-

ing or particulate fouling), a certain chemical will

be added.

After soaking, the chemicals are ushed from themodule and ltration can start again.

“Preventing is better than curing.” Therefore, it is

necessary to have a high cross-ow velocity to

limit the concentration polarization layer.

However, larger cross-ow velocities result in more

energy consumption.

To overcome the high energy consumption, recir-

culation of the concentrate can take place.

A special conguration for this is semi-dead-end

nanoltration. The installation is operated in a

dead-end conguration, but the concentrate is

continuously recirculated (Figure 12). After some

time the concentrate is disposed of. In this way

the energy consumption is limited.

4.4 Field installations

Heemskerk, PWN North-Holland

Water from the IJsselmeer is conventionally pre-treated by coagulation, sedimentation, ltration

and activated carbon ltration, and then trans-

ported over 70 km to Heemskerk. Here, a large

surface water membrane treatment plant has been

built with a capacity of 3000 m3/h.

The water is rst treated by ultraltration to remove

suspended material, bacteria and viruses. The

permeate of the ultraltration is feed water for the

reverse osmosis installation.

This RO installation consists of two stages. In the

rst stage, 24 modules are placed; in the secondFigure 12 - Semi-dead-end operation

≈

≈

≈

≈

cover open pump

0.10·Q

Q

0.90·Q

Figure 11 - Christmas tree conguration

feed

module

concentrate

product

199




stage only 12 modules are present (2:1 staging).The permeate of the RO is conditioned by pH-cor-

rection and, after mixing with water treated in the

dunes is transported to the customers.

This mixing with dune water takes place because

permeate from an RO is low on necessary minerals

for humans and dehydrates the human body.

Schiermonnikoog, Vitens

On the island of Schiermonnikoog, anaerobic

groundwater is treated to produce drinking water

by means of nanoltration.

Nanoltration is used because the groundwater

has a high color content and hardness level.

The groundwater is treated while it is still anaero-

bic, because iron and manganese are still present

in dissolved form.

If oxygen were present, iron and manganese would

directly precipitate and form ocs that would foul

the installation.

After the NF the water is aerated and treated by

slow sand ltration before it is distributed.

Industry

There are many industrial applications of NF/RO

in the Netherlands.

Small scale laundries, slaughterhouses and green

houses use NF/RO installations.

On a larger scale, chemical industries (DSM in

Geleen (2000 m3/h) or Heineken in Zoeterwoude

(500 m3/h)) use NF/RO membranes.

Figure 13 - 3D-engineering Heemskerk

Figure 14 - Membrane installation at Heemskerk

Figure 15 - Anaerobic NF-installation

200




In waste watertreatment NF/RO is not yet used.

Figure 16 - NF at Schiermonnikoog

Further reading


(2005), (ISBN 0 471 11018 3) (1948 pgs)

201





Documents

Drinking Water technology