Filterability and Sludge Concentration in

Membrane Bioreactors

Copyright © 2011 by Maria Lousada Ferreira Printed: Gildeprint- Enschede- The Netherlands ISBN/EAN: 978-90-8957-023-9 September, 2011

Printed and distributed by: Maria Lousada Ferreira E-mail: m.lousadaferreira@tudelft.nl All rights reserved. No part of the material protected by the copyright may be reproduced

or utilized in any form or by any means, electronic, mechanical, incluing photocopying,

recording or by any information storage and retrieval system, without written permission

from the publisher.

Filterability and Sludge Concentration in

Membrane Bioreactors

Proefschrift

ter verkrijging van de graad van doctor aan de Technische Universiteit Delft

op gezag van de Rector Magnificus prof. ir. K.C.A.M. Luyben, voorzitter van het College voor Promoties,

in het openbaar te verdedigen op donderdag 15 september 2011 om 12.30 uur

door Maria da Conceição LOUSADA-FERREIRA

Environmental Engineer, Branch Environment, Universidade Nova de Lisboa (New University of Lisbon)

Geboren te Portugal.

Dit proefschrift is goedgekeurd door de promotor: Prof. Ir. J. H. J. M. van der Graaf Samenstelling promotiecommissie: Rector Magnificus, voorzitter Prof. Ir. J. H. J. M. van der Graaf, Technische Universiteit Delft, Promotor Prof. Dr. Ir. J. B. van Lier, Technische Universiteit Delft Prof. Dr. S. J. Judd, Cranfield University, United Kingdom Prof. Dr. T. Melin, Rheinisch-Westfälische Technische Hochschule Aachen, Germany Prof. Dr. T.O. Leiknes, Norwegian University of Science and Technology, Norway Prof. Dr. C. Cabassud, Institut National des Sciences Appliquées Toulouse, France Ir. J. W. Mulder, Evides Waterbedrijf, The Netherlands Prof. Dr. Ir. F. H. L. R. Clemens, Technische Universiteit Delft, reservelid Dit proefschrift is tot stand gekomen met financiële ondersteuning van: Project EUROMBRA, European Comission, 6Th Framework Programme, Contract nº 018480.

V

Acknowledgements

My first “Thank you” goes to my promoter Prof. Jaap van der Graaf. Independently from the scientific background or personal experience, Prof. Jaap van der Graaf knows how to listen and evaluate the usefulness. His comments or decisions always made me think and, in the end, the lessons I learned were not only scientific but also professional and personal. I am indeed very grateful for the opportunity to work with someone with such a vast experience. In the TU Delft I had the good fortune of working with excellent scientists but also excellent people. It is really not possible to name them all…Prof. Jules van Lier, thank you very much for keeping your door open for me; Jaap de Koning, thanks for always keeping your calm and finding solutions; Jasper, thanks for the good advices and for making time to listen; Mieke, thanks for teaching me the Dutch ways; Tonny and Patrick, what would I do without your help? Also this PhD would have been a lot more difficult without the support of my former university, University of Beira Interior, Portugal. A special thanks to Prof. Cavaleiro and Prof. Antonio Albuquerque. Thank you for supporting my decisions and understating my choices. The EUROMBRA project opened the doors of Europe for me. I was lucky to contact with probably the most enthusiastic and brilliant MBR people Europe has. To all the responsible of the MBR installations I visited, thank you for allowing our visits and commenting on our reports. To the Wastewater Treatment plant operators, thanks a lot for all your support. The meetings of the project MBR2 and the Wastewater Treatment Plant meetings at Heenvliet MBR were always a source of stress but also a very powerful incentive. To all the participants, for the use of their “grey matter’’ to the benefit of my work, I thank you. I could not end this paragraph without a particular thanks to the Heenvliet crew…Thank you Jan Willem Mulder, Olaf Duin and Merle de Kreuk for allowing me to work at Heenvliet and for always being critical towards my work. Thank you Wilfred and André, especially for the good spirits and the countless cups of coffee during the morning… Thanking colleagues and friends is complicated. Some of my colleagues are also my friends so no clear distinction exits anymore. Stefan and Adrien, many times you cracked my head completely but this PhD would have been a lot more difficult without you both; Viviane, thank you for being there when I most needed a friend; Pawel, thanks for listening; Arie, thanks for your reliability; Diana e Francisco, my former students, working with you was great and reminded me how rewarding teaching can be; Bertha, thanks for your friendship and for all the help in practical matters; Sofia, Edi, Nesen, Carmen, Bruno and Maarten, thanks for always being there. My move to the Netherlands widened my professional word but most of all transformed my world of affections to an immense place. As a good friend of mine once said, a PhD is like a marathon and in my marathon I had many colleagues and friends cheering, giving a push, throwing a bottle of water along the way. Because of my move to the Netherlands they are now Portuguese, Dutch, Italians, Turkish, German, Spanish, Chinese… I am also grateful to many others than the ones I mentioned.

VI

I wouldn’t be writing these acknowledgments, or be the person I am today, without my parents. Their support to this PhD was, if possible, even more than unconditional. Their wishes and expectations were always in second place, compared to my happiness. There are no words to tell my parents how grateful I am for all their support and how proud I am to be their daughter. Ana, Helder, Madalena and Francisco, it doesn’t matter how far we are from each other. In my heart, my house will always be next to you. Your presence and support makes my life richer and much, much easier. The most rewarding surprise about my move to the Netherlands, was finding you, Rick. We can indeed dream of the lottery and sometimes get it. You and your warm family, Coby, Leendert, Mieke and Roland, welcomed me and made me one of your own. With you I learned that the joys of life, and its most important values, are independent from cultural backgrounds. I was not expecting to find another family, I already had one! Life is really full of surprises…

Maria Lousada-Ferreira Delft, 25 of August, 2011

VII

Summary

Membrane Bioreactors

A membrane bioreactor (MBR) is a wastewater treatment system that combines an activated sludge process with a filtration process, relying on membrane technology to separate activated sludge from effluent. The main advantages of MBR technology are its small layout and the production of a high-quality and largely disinfected effluent, further referred to as permeate. The main disadvantage is the high capital and operational costs. The high operational costs are especially due to the need to control membrane fouling, understood as the accumulation of materials on the membrane, hampering successful continuous filtration. Aim and Approach of the Thesis

In the early stages of MBR technology, Mixed Liquor Suspended Solids (MLSS) concentration was understood as the major fouling parameter. However, research work results regarding the effect of MLSS concentration on membrane filtration are not consistent, as some authors report increasing fouling with increasing MLSS concentrations, while others see no effect or present opposite results. Nowadays, MLSS concentration alone is considered a poor indicator of membrane fouling, but the correlation between MLSS concentration and fouling propensity has not yet been clarified. Consequently, optimal MLSS concentrations for MBR design and operation are not defined. The main aim of this thesis is to explore the correlation between MLSS concentration and fouling propensity and contribute to the definition of optimal MLSS concentrations for the design and operation of municipal MBRs. Moreover, this thesis provides additional information on MBR technology, including assessment of membrane integrity measured through the permeate quality. The approach of the research work described in this thesis consisted of firstly characterizing the existing membrane tank activated sludge quality in 4 full-scale and 4 pilot-scale municipal MBRs, scattered across Europe and, secondly, changing the MLSS concentration of the activated sludge by diluting, with permeate, and concentrating, by changing the permeate flow/return flow ratio in a full-scale MBR. The MBR activated sludge quality was analyzed through the following parameters: filterability, meaning the capacity of the activated sludge to be filtered through a membrane, measured through the Delft Filtration Characterization method (DFCm); MLSS concentration; Soluble Microbial Products (SMP) concentration; viscosity; particle counting in the range of 2-100 µm and particle counting in the range of 0.4-5 µm. Results and conclusions

Our results showed that the effect of MLSS concentration on MBR activated sludge filterability depends on the applied MLSS concentration range, which can be explained through the MBR activated sludge structure. MBR activated sludge with good to moderate filtration quality and high MLSS concentration, i.e. exceeding a critical MLSS value, is capable of retaining fouling particles in its matrix, so an increase in MLSS concentration leads to improved filterability. In contrast, activated sludge with low MLSS concentration, i.e. below a critical MLSS concentration, is unable to retain fouling

VIII

particles in its matrix. Therefore, in membrane tank sludge with low MLSS concentration, a decrease of the sludge MLSS concentration will lead to filterability improvement. The structure of bulk MBR activated sludge with high MLSS concentrations and good to moderate filtration quality is probably composed of a network of particles linked by SMP and/or bound Extra Polymeric Substances (EPS), acting as a fine mesh. In contrast, the structure of activated sludge with low MLSS concentration is likely to be composed of fragmented particles with unlinked SMP and/or bound EPS. The abovementioned hypothesis explains the apparently contradictory results provided by research studies concerning the influence of MLSS on MBR filtration. In this research, a critical MLSS concentration of 10.5 g/L was obtained. However, since the abovementioned result was defined based on diluting and concentrating experiments with activated sludge collected mainly from one full-scale MBR installation, and always relying on filterability measured according to the DFCm, the absolute value of 10.5 g/L should be used as a reference but not extrapolated, without further research, to all MBR installations handling municipal wastewater. The results presented in this thesis showed that filterability improves with floc size. Nevertheless, in the membrane tank of one MBR installation, variations in floc size are limited probably due to aeration restrictions. Our research showed that improved MBR design and operation control, in municipal MBRs, would be achieved by placing the membranes in aeration tanks, intended for carbonaceous and nutrient removal, and operating at low MLSS concentrations, as defined in this research. Moreover, low return ratios between separate membrane tanks and carbonaceous and nutrient removal tanks are preferable for MBR installations operating at high MLSS concentration, as defined in this research. This research also showed that there are always particles, in the size range of 2-100 µm, in permeate produced in MBRs applying Ultrafiltration (UF) membranes. A physical, chemical or biochemical process occurs constantly, either in the membrane pores or in the permeate side of the membrane, enabling particles to aggregate. A methodology to assess membrane integrity and excessive growth or accumulation of particles in the permeate collection system was proposed. Recommendations for further research

The recommendations for further research proposed in this thesis are as follows: to investigate filterability in the various tanks of one MBR installation; to investigate the effect of submicron particles on activated sludge filterability; to evaluate reliable methods for SMP quantification; to investigate the critical MLSS concentration in other full-scale MBRs; to investigate the structure of activated sludge with high and low MLSS concentration, as defined in this research, through morphological parameters, capable of preserving the activated sludge structure; to optimize aeration ports, aeration scouring and return ratio in MBRs with a separate membrane tank, applying high MLSS concentrations, as defined in this research; and to investigate the nature of permeate particles.

IX

Samenvatting

Membraanbioreactoren

Een membraan bioreactor (MBR) is een afvalwaterzuivering systeem dat een actief slib proces met een filtratie proces combineert, en vervolgens gebruik maakt van membraantechnologie om het actief slib te scheiden van het afvalwater. De belangrijkste voordelen van MBR-technologie zijn de kleine omvang en de productie van een kwalitatief hoogwaardig en grotendeels gedesinfecteerd effluent, verder aangeduid als permeaat. Het belangrijkste nadelen zijn de hoge investerings- en operationele kosten. De hoge operationele kosten zijn vooral te wijten aan de noodzaak om de vervuiling van de membranen tegen te gaan, hiermee wordt de accumulatie van materialen op het membraan bedoeld, aangezien dit een goede doorstroming belemmert. Doel en aanpak van het proefschrift

In de vroege stadia van de MBR-technologie werd het droge stofgehalte aangeduid als de belangrijkste vervuilingsparameter. Echter, onderzoeksresultaten met betrekking tot het effect van het droge stofgehalte op membraanfiltratie zijn niet consistent, sommige auteurs melden bijvoorbeeld toenemende aangroei met toenemend droge stofgehalte, terwijl anderen geen effect zien of zelfs een negatieve relatie rapporteren. Tegenwoordig word het droge stofgehalte beschouwd als een slechte indicator voor membraanvervuiling, maar de correlatie tussen het droge stofgehalte en membraan-vervuiling is nog niet opgehelderd. Hierdoor zijn optimale droge stofgehalten voor MBR ontwerp en gebruik niet gedefinieerd. Het voornaamste doel van dit proefschrift is de correlatie tussen het droge stofgehalte en de vervuiling te verkennen en een definitie van het optimale droge stofgehalte voor het ontwerp en gebruik van de gemeentelijke MBR’s te ontwikkelen. Bovendien levert dit proefschrift extra informatie over MBR-technologie, inclusief de beoordeling van de integriteit van het membraan, gemeten door middel van een beoordeling van de permeaat kwaliteit. De aanpak van het onderzoek beschreven in dit proefschrift bestaat uit, in de eerste plaats, het karakteriseren van de kwaliteit van het actief slib in de membraantank van vier gemeentelijke MBR’s op full-scale en vier op pilot-scale, verspreid over Europa en ten tweede, het veranderen van het droge stofgehalte van het actief slib door het te verdunnen met permeaat, en te concentreren, door het veranderen van de permeaatstroom / terugloop verhouding in een full-scale MBR. De kwaliteit van het MBR actief slib werd geanalyseerd door middel van de volgende parameters: filtreerbaarheid, hetgeen de mogelijkheid van het actief slib om te worden gefilterd door een membraan inhoud, gemeten via de Delftse Filtratie Karakterisering methode (DFCm), droge stofgehalte, concentratie opgeloste microbiële producten (SMP); viscositeit; aantal deeltjes in het bereik van 2-100 µm en het aantal deeltjes in het bereik van 0,4 tot 5 µm.

X

Resultaten en conclusies

Onze resultaten toonden aan dat het effect van het droge stofgehalte op de filtreerbaarheid van het MBR actief slib afhankelijk is van de gebruikte concentratie droge stof, wat kan worden verklaard door de structuur van het MBR actief slib. MBR actief slib met een goede tot matige filtratie kwaliteit en een hoog droge stofgehalte, dat wil zeggen hoger dan een kritische waarde voor de concentratie droge stof, is in staat om vervuilende deeltjes in haar matrix op te nemen, dus een toename van de concentratie droge stof leidt tot een verbeterde filtreerbaarheid. Dit in tegenstelling tot actief slib met een laag droge stofgehalte, dat wil zeggen onder een kritische waarde voor de concentratie droge stof, dat niet in staat is om vervuiling tegen te houden in de matrix, hier leidt een afname van de concentratie droge stof tot een verbeterde filtreerbaarheid. De structuur van bulk MBR actief slib met een hoge concentratie droge stof en een goede tot matige filtreerbaarheid is waarschijnlijk samengesteld uit een netwerk van deeltjes die met elkaar verbonden zijn door SMP en/of gebonden Extra Polymere Stoffen (EPS), als een fijn gaas. In tegenstelling tot de structuur van actief slib met een lage concentratie droge stof die waarschijnlijk bestaat uit gefragmenteerde deeltjes met ongebonden SMP- en/of EPS. De bovengenoemde hypothese verklaart de schijnbaar tegenstrijdige uitslagen van onderzoeken met betrekking tot de invloed van het droge stofgehalte op MBR filtratie. In dit onderzoek is een kritische waarde van 10,5 g / L verkregen voor de concentratie droge stof. Aangezien het bovengenoemde resultaat is verkregen door middel van het verdunnen en concentreren van actief slib, voornamelijk verzameld in een enkele full-scale MBR-installaties, en de filtreerbaarheid altijd is gemeten volgens de DFC methode, moet de waarde van 10,5 g / L worden gebruikt als een referentiewaarde, en kan dit niet zonder verder onderzoek worden geëxtrapoleerd naar andere MBR installaties voor de behandeling van stedelijk afvalwater. De resultaten gepresenteerd in dit proefschrift laten zien dat filtreerbaarheid verbetert met toenemende vlokgrootte, hoewel in de membraantank van een MBR installatie de variaties in vlokgrootte beperkt zijn, waarschijnlijk door beperkingen in de beluchtingsmethoden. Ons onderzoek toont aan dat een beter MBR ontwerp en operationele controle, in de gemeentelijke MBR, zou kunnen worden bereikt door het plaatsen van de membranen in beluchtingstanks, bestemd voor de verwijdering van organische stoffen en nutriënten, en door te werken met lage concentraties droge stof, zoals gedefinieerd in dit onderzoek. Bovendien is een lage terugstroomverhouding tussen afzonderlijke membraantanks en verwijderingstanks voor organische stof en nutriënten preferabel voor MBR-installaties die opereren bij een hoge concentratie droge stof, zoals gedefinieerd in dit onderzoek. Dit onderzoek toont ook aan dat er altijd deeltjes zijn in een grootteorde van 2-100 µm in het permeaat dat geproduceerd wordt in MBR’s die gebruik maken van ultrafiltratie (UF) membranen. Een fysisch, chemisch of biochemisch proces verloopt continu, hetzij in de membraanporiën of aan de permeaatkant van het membraan, waardoor deeltjes in staat zijn te aggregeren. Ook wordt een methodiek om membraanintegriteit, overmatige groei of ophoping van deeltjes in het permeaat netwerk te evalueren voorgesteld.

XI

Aanbevelingen voor verder onderzoek

De aanbevelingen voor verder onderzoek voorgesteld in dit proefschrift zijn als volgt: De filtreerbaarheid onderzoeken in de verschillende tanks van een MBR-installatie. Het effect van submicron deeltjes op de filtreerbaarheid van het actief slib te onderzoeken. Betrouwbare methoden voor SMP kwantificering te evalueren. Het kritische droge stofgehalte te onderzoeken voor andere full-scale MBR’s. De structuur van actief slib met een hoog of laag droge stofgehalte te onderzoeken, zoals gedefinieerd in dit onderzoek, door middel van morfologische parameters, die in staat zijn de structuur van het actief slib te behouden. Het optimaliseren van de poorten van het beluchtingsvat, beluchtingsreiniging en terugstroomverhoudingen in MBR’s met een aparte membraantank. Het toepassen van hoge concentraties droge stof, zoals gedefinieerd in dit onderzoek. En de aard van de deeltjes die in het permeaat doordringen te onderzoeken.

XII

XIII

Table of contents

Acknowledgements V Summary VII Samenvatting IX 1. Introduction 1

1.1 Background 1 1.2 Research Problem 4 1.3 Aim of the Thesis 6 1.4 Thesis Outline 8

2. Fundamentals of Membrane Bioreactors and Activated Sludge Quality 11

2.1 Introduction 11 2.2 Membrane Bioreactor Technology 11 2.3 Membrane Bioreactor Activated Sludge Quality 21 2.4 Impact of Mixed Liquor Suspended Solids Concentration on Membrane

Filtration 27

2.5 Permeate Quality 31 2.6 Conclusions 34

3. Framework and Clarification of Concepts 35

3.1 Introduction 35 3.2 The Delft Filtration Characterization Method 35 3.3 The Delft Filtration Characterization Method in Practice 38 3.4 Qualifying Membrane Bioreactor Activated Sludge 40 3.5 Analyzing Activated Sludge Characteristics by Diluting and

Concentrating 41

3.6 Assessing Membrane Bioreactor Integrity through Permeate Quality 43 3.7 Conclusions 44

4. Materials and Methods 47

4.1 Filterability 47 4.2 Mixed Liquor Suspended Solids 51 4.3 Soluble Microbial Products 51 4.4 Viscosity 52 4.5 Particle Counting in the Range of 2-100 µm 52 4.6 Particle Counting in the Range of 0.4-5 µm 57

XIV

5. Activated Sludge Quality in European Membrane Bioreactors - On site Measurements

61

5.1 Introduction 61 5.2 Methodology 62 5.3 Filterability Results and Discussion 64 5.4 Return Ratio Effect on Membrane Tank Filterability 67 5.5 Particle Counting Results in the Range of 2-100 µm and Discussion 68 5.6 Relationship between Filterability and Particle Counting in the Range

of 2-100 µm 74

5.7 Conclusions 75 6. Diluting Membrane Bioreactor Activated Sludge 77

6.1 Introduction 77 6.2 Methodology 79 6.3 Characterization of Membrane Bioreactor Activated Sludge: Results

and Discussion 81

6.4 Comparison between Membrane and Aeration Tank Activated Sludge 94 6.5 Comparison between Membrane Tank Activated Sludge from Different

Membrane Bioreactors 95

6.6 Results Using a Flat Sheet Membrane Installation 98 6.7 Comparison between Permeate, Tap Water and Conventional Activated

Sludge Effluent as Dilution Media 99

6.8 Comparison between Constant and Variable Cross-Flow Velocity 100 6.9 General Discussion 102 6.10 Conclusions 109

7. Concentrating Membrane Bioreactor Activated Sludge 111

7.1 Introduction 111 7.2 Methodology 112 7.3 Background Information about the Heenvliet Membrane Bioreactor 113 7.4 Results and Discussion 119 7.5 Link between Concentrating and Diluting Results 133 7.6 Conclusions 134

8. Assessing Membrane Integrity through Particle Counting 137

8.1 Introduction 137 8.2 Methodology 138 8.3 Results 143 8.4 Discussion 147 8.5 Conclusions 152

XV

9. Technological and Scientific Relevance 155 9.1 Introduction 155 9.2 Effect of Mixed Liquor Suspended Solids Concentration on

Filterability 155

9.3 Effect of Floc Size on Filterability 159 9.4 Effect of Soluble Microbial Products Concentration on Filterability 161 9.5 Effect of Submicron Particles on Filterability 163 9.6 Effect of Return Ratio on Membrane Tank Filterability 164 9.7 Structure of Activated Sludge with High and Low Mixed Liquor

Suspended Solids Concentration 165

9.8 Critical Mixed Liquor Suspended Solids Concentration 169 9.9 Consequences of the Applied Mixed Liquor Suspended Solids

Concentration Range 171

9.10 Suspended Particles in the Membrane Bioreactor Permeate 173 9.11 Methodology for Verifying Membrane Integrity 174 9.12 Recommendations for Further Research 177 9.13 Conclusions 179

Bibliography XVII Curriculum Vitae XXV Annexes

A- Relevant Operational Parameters in Submerged Membrane Bioreactors

XXIX

B- Flat Sheet Membrane Unit XXXI C- Results of Diluting Experiments XXXII

XVI

1

1. Introduction

1.1 Background

There is no doubt that water scarcity is one of the major challenges of our world. According to the World Health Organization (WHO 2009), water scarcity affects one in three people on every continent of the world. However, water scarcity issues occur worldwide, but are geographically unevenly distributed. Within the European Union, southern countries are usually more affected by water shortage, because they suffer more draughts and higher temperatures. In addition, during the summer months, agricultural and touristic water demand increases sharply, precisely when water resources become scarce. Figure 1.1 shows the Water Exploitation Index (WEI) in European countries. The WEI measures the amount of water used compared to the available long-term fresh water resources by country or region. An index of 20% or more indicates water scarcity, while a percentage exceeding 40% indicates a severe problem (EC 2010a).

Figure 1.1 - Water exploitation index (percentage) in the 1990s and latest available year (in brackets) (EC 2010b). As Figure 1.1 shows, countries such as Cyprus, Belgium, Spain, Italy and Malta are the most affected by water scarcity, but the reasons may differ within each country. Belgium, for instance, has water scarcity problems due to the abstraction of water for energy purposes, i.e. cooling water. Furthermore, the WEI by country, shown in Figure 1.1, does not reflect regional and seasonal variations. In Portugal the WEI by river basin is

2

extremely variable: Sado river basin, 132%, Lis river basin, 82%, Minho river basin 1%, Lima river basin, 5%; however, the average WEI by country is 15% (EC 2010a). To deal with water scarcity, the European Union Commission identified, in 2007, an initial set of policies intended to move the European Union towards a water-efficient and water-saving economy (EC 2010b). At the same time, there is widespread concern about biodiversity loss and the need to take effective actions to reduce it. Biodiversity can be defined as the number and variety of living organisms, from microorganisms to whales or giant sequoia trees (EC 2010b). It is widely understood that ecosystem productivity decreases as species diversity declines, leading to socioeconomic losses. Figure 1.2 shows the causes of biodiversity loss.

Figure 1.2 - Causes of biodiversity loss (EC 2010b).

Nature 2000 is a network of natural habitat and species sites, based on two European Union Directives1, aimed at preserving biodiversity in the European context. The Nature 2000 network has the particularity of designating natural areas where ongoing human activity is also present. In 2009, 17.9% of the EU member states area was designated as part of the Nature 2000 network (EC 2010b). In the Netherlands, in particular, according to the Netherlands Environmental Assessment Agency (PBL 2003), the land area occupied by Nature 2000 sites is about 10% of the territory; however, the protected areas are mainly constituted by the major water bodies, such as the surface area of inland waters and the North Sea, in a total of about 2,900 Km2. European legislation on water quality is based on the Water Framework Directive2 (WFD), aimed at protecting and restoring clean water across Europe. Under the WFD, all water an wetland habitats belonging to Nature 2000 sites must achieve “good ecological status’’ by 2015. To achieve the abovementioned goals, the water cycle should be optimal, i.e. all wastewater should undergo a certain level of treatment. In certain locations, this implies the need for advanced treatment followed by reuse. Membrane bioreactors (MBRs) constitute a compact wastewater treatment system that produces a high quality and largely disinfected effluent (Metcalf & Eddy 2003). MBR technology effluent is, therefore, especially suitable for reuse purposes or for discharging in environmentally sensitive water bodies. In particular, the reuse potential of MBR

1 Council Directive 92/43/EEC and Council Directive 79/409/EEC. 2 Directive 2000/60/EC.

3

effluent has been a powerful driver for many industries to classify MBR as the “Best available technology” (Lesjean et al. 2008). MBR is a compact wastewater treatment system in which sludge and clear water are separated by membrane filtration. Figure 1.3 shows a general layout of an MBR system.

Pre-treatment

Membrane tank

influent effluent

Sludge flow

Bioreactor

Figure 1.3 - Layout of a membrane bioreactor system.

In a MBR system, the pre-treatment step usually consists of sieves to remove coarse and fine materials. The bioreactor provides removal of carbonaceous substances and can be designed to remove phosphorus and nitrogen to concentration levels below the discharge limits. The sludge and clear water are separated by membrane filtration. The membranes are usually submerged in the activated sludge of the bioreactor. The effluent resulting from membrane filtration, henceforth referred to as permeate, has no suspended solids and is largely disinfected, complying with the standards of tertiary treatment. The most widely used wastewater treatment concept is the activated sludge system. The activated sludge system has several possible process configurations. A general layout of a conventional activated sludge (CAS) system is shown in Figure 1.4.

Figure 1.4 - Layout of a conventional activated sludge system. In a CAS system, the pre-treatment step is usually designed to remove coarse material, sand and fat, while the primary clarifier removes about 60% of suspended solids. In CAS systems applying extended aeration the primary clarifier is not required. The bioreactor and secondary clarifier are usually designed to remove about 90% of suspended solids and remaining organic material. Bioreactors especially designed to remove not only carbonaceous material, but also phosphorous and nitrogen below the discharge limits, can

Pre-treatment

Primary clarifier

Bioreactor Secondary clarifier

influent effluent

Sludge flow

4

also perform tertiary treatment. Tertiary treatment can also be achieved by additional treatment processes, such as sand filtration and disinfection, coupled with an activated sludge process ensuring secondary treatment. In Europe, about 800 MBR installations were in operation by the end of 2008, among which 37 with a capacity greater than 5,000 m3/d (Lesjean et al. 2008; Huisjes et al. 2009). A stringent legislation on effluent discharge and local water scarcity issues are powerful drivers for the implementation of MBR technology. However, although MBRs are considered a proven technology in industry, the situation is quite different in municipal wastewater treatment, mainly due to its costs, in particular its energy comsumption. Overall, municipal MBRs are still considered an expensive technology. Significant steps towards the reduction of MBR costs have been accomplished in the last decades. Over the last 20 years, the cost of the membranes and whole MBR processes have decreased drastically due to improved design and extended membrane lifetime (Judd 2006). Recent reports on the economic aspects of full-scale MBRs indicate that MBR life-cycle costs are competitive against CAS, when the conventional technology is upgraded to achieve similar effluent quality (Brepols et al. 2009). However, increasing membrane lifetime and energy efficiency, particularly due to improved design and operation, are still key factors in the reduction of MBR technology costs.

1.2 Research Problem

One of the main advantages of MBR technology is the possibility to work at high Mixed Liquor Suspended Solids (MLSS) concentrations, which allows designing tanks with smaller volume than in CAS systems. In contrast, one of the main disadvantages is the need to control membrane fouling, which increases operating costs. Membrane fouling can be defined as the deposition and accumulation of microorganisms, colloids, solutes and cell debris inside and around the membrane (Meng et al. 2009). In the early stages of MBR technology, MLSS concentration was considered one of the possible fouling parameters. In the early 1990s, an MBR was understood as a compact process where carbonaceous and nutrient removal and the membrane separation step would all occur in the same bioreactor. MLSS concentration was therefore relatively constant and a concentration range between 4 and 40 g/L was applied. It was generally accepted that increasing MLSS concentrations would lead to increased fouling. Figure 1.5 shows the layout of the MBR process during the 1990s and the current decade. At present, the layout illustrating the MBR process in the current decade is more frequently applied however the 90’s MBR layout is still in use depending on the influent characteristics, discharge limits or other relevant criteria.

5

influent permeate

membrane

MBR

permeate

influent

MBR

MBRs in the ’90s

MBRs in the current decade

Figure 1.5 - Layout of the MBR process during the 1990s and the current decade.

At present, an MBR is composed of several tanks, designed to remove target components, and the membranes are usually located in a separate tank. In the tanks intended for the removal of target components, a complex biology is stimulated for the purpose of removing carbonaceous materials, phosphorus and nitrogen. The membranes are usually located in the separate tank and the applied MLSS concentration in the membrane tanks varies between 8 and 18 g/L. However, between the various tanks composing an MBR, MLSS concentration also changes, leading to a dynamic MLSS concentration. It is, therefore, logical that the effect of a dynamic MLSS concentration on membrane filtration differs from the effect of a constant MLSS concentration. Nowadays, it is generally accepted that MLSS concentration alone is a poor indicator of biomass fouling propensity (Jefferson et al. 2004). However, it is generally understood that biological flocs, i.e. the suspended fraction of the activated sludge, play a key role in the fouling layers built up around the membranes (Le-Clech et al. 2006). In short, MLSS concentration is considered to be a poor indicator of fouling propensity but, nevertheless, the suspended fraction of the sludge does play a role in the build-up of the cake layer. Questions such as how is fouling propensity related to the suspended fraction of the sludge and what is the role of MLSS concentration in MBR filtration, remain unanswered so far. It was demonstrated that operating at MLSS concentrations lower than 15 g/L could significantly improve the oxygen transfer efficiency, quantified as the α factor (Germain et al. 2005). Therefore, due to the required energy input, MLSS concentrations below 15 g/L are considered preferable for MBR design and operation (van Nieuwenhuijzen et al. 2008). Nevertheless, because the correlation between MLSS concentration and fouling is not fully understood, optimal MLSS concentrations for MBR design and operation are not completely defined.

6

Regardless of the impact of MLSS concentration on membrane filtration, defining optimal MLSS concentrations for design and operation of MBRs is critical because it defines the MBR tank volumes and, consequently, contributes to optimizing investment and energy costs.

1.3 Aim of the Thesis

The main aim of this thesis is to help clarify the dynamic relationship between the suspended fraction of the activated sludge and fouling propensity. The suspended fraction of the MBR activated sludge can be characterized through several parameters, starting with MLSS concentration. Parameters such as viscosity and floc size are also directly related to the suspended fraction of the sludge. Therefore, in this research, besides MLSS concentration, viscosity and floc size activated sludge characteristics were also analyzed. The fouling propensity of the activated sludge was characterized through filterability, i.e. the potential of the activated sludge to cause fouling, as measured by the Delft Filtration method (DFCm) (Evenblij et al. 2005). There are no universally agreed constituents to which fouling can be primarily attributed (Judd 2008). Several constituents, such as Soluble Microbial Products (SMP) and submicron particles, are identified. In this research, the impact of SMP concentration and submicron particles on the fouling propensity of MBR activated sludge was also analyzed. The aim is to help clarify the relationship between fouling and the aforementioned fouling constituents. To understand the relationship between the suspended fraction of the sludge and the fouling propensity, it is important to first characterize and compare the filtration characteristics of the activated sludge produced by different MBR installations. To allow an unequivocal comparison, the Delft University of Technology (TU Delft) developed the DFCm. The DFCm comprises the Delft Filtration Characterization installation (DFCi) and a measuring and cleaning protocol (Evenblij et al. 2005). Activated sludge samples obtained from different MBR installations are filtered in the DFCi following a specific operation and cleaning protocol. Therefore, differences in results can be directly attributed to the activated sludge quality. The development of the DFCm and the participation of TU Delft in the European projects EUROMBRA and MBR-Train, enabled an activated sludge characterization tour through 4 full-scale and 4 pilot-scale MBRs, scattered across Europe. The aim of the aforementioned tour was to characterize the existing MBR activated sludge in terms of fouling propensity and, in the specific case of this thesis, its relationship with the suspended fraction of the sludge. To further study the effect of MLSS concentration on membrane filtration, an approach of diluting and concentrating the activated sludge, followed by the relevant measurements, was adopted in this thesis. MBR activated sludge was diluted with permeate to obtain sludge with decreasing MLSS concentrations. The opposite procedure, i.e. concentrating the MBR activated sludge, was performed at a full-scale MBR, using the operational approach described below.

7

The main factors affecting MLSS concentration in a membrane tank can be identified through a mass-balance calculation, under steady-state conditions. An MBR with a separate membrane tank, with completely mixed conditions, where the sludge is wasted from the carbonaceous and nutrient removal tanks, was used for the following calculations. A negligible MLSS concentration growth in the membrane tank was assumed.

[ ] ( ) [ ] [ ]BT R R PBT MT PMLSS Q Q MLSS Q MLSS Q+ = + (Equation 1.1)

Where:

[MLSS]BT - MLSS concentration in carbonaceous and nutrient removal tanks. QBT - Flow from carbonaceous and nutrient removal tanks to the membrane tank; QR - Return flow from the membrane tank to carbonaceous and nutrient removal

tanks; [MLSS]MT - MLSS concentration in the membrane tank; [MLSS]P - MLSS concentration in the permeate; QP - Permeate flow from the membrane tank.

Assuming that QBT is equal to QP and that [MLSS]P is zero, we can calculate the mass balance as a function of [MLSS]MT as follows:

[ ] [ ] 1 BT

MT BTR

QMLSS MLSS

Q

= +

(Equation 1.2)

Equation 1.2 shows that the MLSS concentration in the membrane tanks is a function of the MLSS concentration in the carbonaceous and nutrient removal tanks and of the ratio between the flow entering the membrane tank, equal to the permeate flow, and the return flow. If there is no permeate extraction, the MLSS concentration in the membrane tanks will be equal to the MLSS concentration in the carbonaceous and nutrient removal tanks. In this research, concentrating, i.e. obtaining increased MLSS concentrations in the activated sludge, was performed by increasing the ratio permeate flow/ return flow in a full-scale MBR. Due to a construction failure in a membrane module of a full-scale MBR, a permeate monitoring campaign was launched. Opportunity and curiosity lead us to repeat it in all the MBRs visited across Europe. MBR technology can remove protozoan cysts, oocysts, helminthes ova, bacteria and viruses (Metcalf & Eddy 2003). However, microorganism removal is only effective if the membranes are intact. The aim of the aforementioned monitoring campaign was to assess membrane integrity in MBRs.

8

1.4 Thesis Outline

Figure 1.6 shows the outline of the thesis and the main subjects in each chapter or group of chapters.

Figure 1.6 Outline of the thesis.

The remaining chapters of this thesis are structured as follows: Chapter 2 presents a condensed review of the state of the art of MBR technology. The review focuses on MBR activated sludge quality. As an introduction, the fundamentals of MBR technology are described. This is followed by a description of how to measure MBR activated sludge quality and of the known facts concerning bulk MBR activated sludge quality. The available knowledge on the impact of MLSS concentration on membrane filtration is presented and discussed. The chapter ends with a description of MBR permeate quality. The relevant parameters for MBR operation are described in Annex A.

9

Chapter 3 presents the framework and clarification of concepts concerning this thesis. The main findings obtained by former PhD students of the Department of Sanitary Engineering of TU Delft involved in MBR research are summarized. The qualification of MBR activated sludge, diluting and concentrating MBR activated sludge and assessment of membrane integrity, as it is performed in this thesis, are defined and the corresponding approach clarified. Chapter 4 describes the materials and methods used in this thesis, as well as the processing of results. Chapter 5 presents the activated sludge quality results obtained from 4 full-scale and 4 pilot-scale municipal MBRs scattered across Europe. Special emphasis was placed on the return ratio effect on membrane tank filterability, as well as the relationship between filterability and the suspended fraction of the sludge, measured by particle counting in the range of 2-100 µm. Chapter 6 describes the experiments aimed at studying the effect of MLSS concentration on membrane filtration, which involve decreasing the MLSS concentration in the MBR activated sludge by diluting with permeate. In addition, diluting experiments conducted to check our own results are also described. The latter experiments evaluate differences between activated sludge within one MBR and from different MBRs and the effect of membrane configuration, dilution media and hydraulic regime on our own results. Chapter 7 describes the experiments aimed at studying the effect of MLSS concentration on membrane filtration, which involve increasing the MLSS concentration in the MBR activated sludge through a concentrating operation. Activated sludge with increased MLSS concentrations was obtained in a full-scale membrane tank by increasing the permeate flow/return flow ratio. In addition, Chapter 7 discusses the link between concentrating and diluting experiments by describing the results obtained when concentrated sludge was further diluted with permeate. Chapter 8 presents the assessment of MBR membrane integrity, at 4 full-scale and 4 pilot-scale municipal MBRs scattered across Europe, through particle counting in the range of 2-100 µm. Results from particle counting in the range of 2-100 µm in demineralized water, process water and CAS effluent samples, are also described to frame the permeate results. Chapter 9 presents the general discussion and main conclusions of this thesis. Emphasis was placed on the impact of MLSS concentration, floc size, SMP concentrations, submicron particles and return ratio on activated sludge filterability. Operational consequences of the obtained conclusions are discussed, such as the existence of a critical MLSS concentration, consequences of the applied MLSS concentration range and the possible need for a separate membrane tank. In addition, the existence of particles in the MBR permeate is discussed and a methodology for checking membrane integrity proposed. Recommendations for further research and conclusions are presented in the last two sections of Chapter 9.

10

11

2. Fundamentals of Membrane Bioreactors and Activated Sludge Quality

2.1 Introduction

This chapter provides a short review of the state of the art of MBR technology. The fundamentals of MBR technology are described, as well as its main advantages and disadvantages. The focus is on municipal submerged MBRs, since they are the scope of application of this thesis. The main bottleneck of MBR technology, fouling, is defined and current knowledge of the phenomenon presented. This review focuses on membrane activated sludge quality. The available parameters for qualifying bulk MBR activated sludge are described and discussed. Current knowledge of MBR activated sludge quality is presented with special emphasis on the effect of MLSS concentration on membrane filtration. The available knowledge of MBR permeate quality is also presented and discussed. To facilitate the reading of this chapter, a list of relevant parameters for MBR operation is provided in Annex A.

2.2 Membrane Bioreactor Technology

A MBR process couples a biological treatment with a membrane separation step. A membrane can be defined as a material that allows certain components to pass through it more easily than others (Judd 2006). The effluent obtained through membrane separation is usually referred to as permeate. The MBR process evolved from the Conventional Activated Sludge (CAS) process. Figure 2.1 shows a simplified layout of a CAS and a MBR process operated under similar Sludge Retention Times (SRT).

Pre-treatment Secondary clarifier

Biological treatment

Excess sludge

Pre-treatmentBiological treatment

Membrane operation

Return sludge

Return sludge

Excess sludge

Figure 2.1 - Simplified layout of a Conventional Activated Sludge process (top) and a

membrane bioreactor process (bottom).

12

Figure 2.1 clearly shows that in the MBR process, the separation of clear water and sludge occurs through membranes, while in the CAS process, it relies on clarifiers. MBR technology background

The MBR process was introduced to the market in the late 1960s, when ultrafiltration (UF) and microfiltration (MF) membranes became commercially available. The process was invented by Dorr-Oliver Inc. as an application for ship-board sewage treatment. The design combined an activated sludge bioreactor with a cross-flow membrane filtration loop. A layout of a side-stream MBR process configuration is shown in Figure 2.2 (a). The original process was very expensive, mainly due to the price of the membranes and the rapid loss of performance resulting from fouling. To reduce fouling, the activated sludge was pumped at high cross-flow velocity, with a significant energy consumption of 10 kWh/m3 of produced permeate (Le-Clech et al. 2006). The high cost of the original process restricted its application to niche areas such as isolated trailer parks and ski resorts. Nevertheless, the coupling of the two processes, activated sludge treatment and membrane filtration, was established and the early MBR process was marketed with some success until the early 1990s (Judd 2006).

Figure 2.2 - MBR process configurations: (a) side-stream (b) submerged (Judd 2008).

In 1989, Yamamoto presented a new MBR design where membranes were submerged in the bioreactor (Yamamoto et al. 1989). An example of a submerged MBR process configuration is shown in Figure 2.2 (b). As mentioned above, the early MBR systems had a side-stream design, meaning that the membranes were located in an external loop to the activated sludge tank. In the design presented by Yamamoto, the membranes were submerged in the activated sludge tanks where the static pressure, caused by the activated sludge liquid on top of the membranes, contributed to the extraction of permeate. The water flow through the membranes, per membrane area, is further referred to as flux. In this new generation of MBRs, modest fluxes were applied, around 25% less than in the early MBR processes, and air was used to control fouling. Therefore, this new design was much cheaper than the early MBR process. The new MBR design, together with decreasing membrane costs, stimulated an exponential increase of MBR applications since the mid-1990s.

13

The MBR design, particularly its operational parameters, has changed considerably since the early MBR process. Initially, MBRs operated under a Sludge Retention Time (SRT) of about 100 days and a Mixed liquor Suspended Solids (MLSS) concentration as high as 30 g/L. At present, the tendency is to apply a lower SRT, around 20 d, and MLSS concentrations between 10 and 15 g/L. With these options, fouling was decreased and membrane cleaning simplified. The commercial options currently available include side-stream and submerged MBRs, both using air to clean the membranes by scouring. In the recent MBR processes, energy consumption is typically below 1 kWh/m3 of produced permeate (Le-Clech et al. 2006). The energy consumption of MBRs varies between 0.5 and 2.5 kWh/m3 of produced permeate (van Nieuwenhuijzen et al. 2008). At present, MBR design still raises questions and challenges, such as whether a separate membrane tank is actually needed. In an MBR, the removal of carbonaceous material and nutrients and the membrane separation step can all occur in the same tank. One example of this design is the Nordkanal MBR in Germany, as shown in the top part of Figure 2.3. Another possibility is the location of the membranes in a separate tank, preceded by several others where the removal of carbonaceous material and nutrients occurs. One example of this design is the Heenvliet MBR in the Netherlands, as shown in the bottom part of Figure 2.3.

Anoxic tank Aerobic/membrane tank

Anaerobic tank

Anoxic tank

Aerobic tank

Membrane tank

Figure 2.3 - Layout of possible MBR tanks sequences.

There is no clear answer for whether a separate membrane tank is more or less effective. Overall, a successful MBR design and operation is understood as an MBR that provides successful treatment and removes the target contaminants at the lowest overall cost (Judd 2008). Nowadays, the optimal operational parameters for MBR design are also controversial. The application of lab- and pilot-scale results to full-scale MBR design and operation is restricted, as described by Kraume et al. (2009). MBR operation should occur at high MLSS concentrations, in order to reduce the size of the bioreactors and the volume of wasted sludge, but higher MLSS concentrations require more airflow and increase the risk of membrane clogging. Moreover, higher fluxes should be applied to reduce the membrane area, but this requires increased cleaning frequency (Judd 2008). Overall, optimal operating conditions increase the MBR costs and, therefore, do not contribute to

14

successful treatment at the lowest possible cost. The use of anaerobic MBRs are currently good options to reduce energy demands in MBRs (Judd 2008). Membrane technology

The membrane technologies applied in the municipal sector are mainly pressure driven. According to Metcalf & Eddy (2003), pressure-driven membranes can be classified in the following operations according to the nominal size of the separation achieved:

- reverse osmosis (RO): separation achieved by differing solubility and diffusion rates of water and solutes in water (<2 nm);

- nanofiltration (NF): separation achieved through a combination of charge rejection, solubility-diffusion and sieving through micropores (<2 nm);

- ultrafiltration (UF): separation by sieving through mesopores (2-50 nm); - microfiltration (MF): separation of suspended solids from water through

macropores (>50 nm). Figure 2.4 shows the components that each pressure-driven membrane operation can remove.

Figure 2.4 - Components removed by each pressure-driven membrane operation (Judd

2006).

All these processes produce a purified permeate product and concentrated waste. The rejected constituents tend to accumulate on the membrane surface, reducing the flux of water through the membrane at a given Trans-Membrane Pressure (TMP) or vice versa. These phenomena are referred to as fouling. The definition of fouling, adopted in this thesis, is specified in section “Fouling”.

15

Process configuration

MBRs can have different membrane process configurations, namely extractive, diffusive or rejection systems (Judd 2008). In extractive MBRs, the membrane is used to extract specific components from the bioreactor; in diffusive MBRs, the membrane is used to introduce gas into the bioreactor. Extractive and diffusive MBRs are not yet available on the market for municipal wastewater. The commercially available MBRs are biomass rejection MBRs, where biomass is retained in the bioreactor while clarified water goes through the membrane. Rejection MBRs combine the biological treatment of wastewater with MF and UF membranes that can be placed either inside or outside the bioreactor. Figure 2.2 above shows the possible process configurations for rejection MBRs, namely side-stream and submerged. For municipal wastewater treatment, the current trend in MBRs tends to favor submerged over side-stream applications (Le-Clech et al. 2006). The side-stream configuration can work with higher fluxes and has greater hydrodynamic control, but generally provides lower permeability (Judd 2005). In submerged systems, the membranes can be configured as vertical flat sheets or plates, vertical or horizontal hollow fibers, out-to-in filtration, or tubes, in-to-out filtration (Le-Clech et al. 2006). The flow layout of the aforementioned membrane configurations is shown in Figure 2.5.

Figure 2.5 - Schematic flows through flat sheet (a), hollow fiber (b) and tube (c) membranes (Adapted from Evenblij (2006)).

Membranes configured as tubes are more common in side-stream applications. Submerged systems are usually either flat sheet or hollow fiber membranes. Hollow fiber membranes are cheaper, but require more frequent cleaning. Flat sheet membranes are 20-25% more expensive than hollow fiber (Le-Clech et al. 2006), but tend to run at higher permeability and are simpler to operate (Judd 2005). In submerged MBRs, the membranes are mounted in modules. The modules are composed of membranes, support structures, inlets and outlets. Pumps, placed in the clean water side of the membranes draw the water through the membrane while solids are retained in the bioreactor (Metcalf & Eddy 2003). To keep the biomass in suspension,

a) c) b)

16

continuously scour the membranes and provide dissolved oxygen (DO) to the biomass, whenever carbonaceous removal is occurring in the membrane tanks, compressed air is introduced into the process by a distribution manifold at the base of the membrane module. These MBRs work in semi-dead-end operation mode, also referred to as cross-flow operation, meaning that, for a single passage of activated sludge across the membrane, only a fraction is converted into permeate. Opposite to the semi-dead-end operation mode, also referred to as cross-flow operation mode, is the dead-end mode, where all the activated sludge reaching the membrane is converted into permeate (Judd 2006). Advantages vs. disadvantages of MBR technology

The advantages and disadvantages of MBR technology are presented in Table II-I. Table II-I Advantages and disadvantages of MBR technology.

Advantages Disadvantages Production of high-quality effluent and largely disinfected permeate

High capital costs

Small footprint Need to control membrane fouling Operation at high Mixed Liquor Suspended Solids concentrations

High energy costs

Higher volumetric loading rates and shorter Hydraulic Retention Times

Potential high cost of periodic membrane replacement (limited data on membrane lifetime)

Longer Sludge Retention Times (SRT), thus, less sludge production

Operation at low Dissolved Oxygen concentrations with potential for simultaneous nitrification - de-nitrification in long SRT designs

(Adapted from Metcalf &Eddy (2003)) Mixed Liquor Suspended Solids (MLSS) concentration is a key operational parameter for membrane bioreactor (MBR) technology. The possibility to use high concentrations of solids contributes to a small footprint, stated as one of the main advantages of membrane bioreactor technology. However, the use of MBR technologies is still limited, especially due to the difficulty of controlling the fouling of the membrane surface. Fouling

The definitions and subdivisions of fouling vary acoording to the author. In this thesis, the following definitions were adopted. Fouling can be defined as the processes leading to deterioration of flux due to the surface or internal blockage of the membranes (Judd 2006). Fouling should be distinguished from the more serious problem of clogging. Clogging is the blockage of the channels between the membranes and/or aerator ports by sludge solids (Judd 2005; Judd 2008). The clogging of the aerator ports is usually minimized by regular flushing (Judd 2005).

17

In this thesis, fouling was subdivided into removable, irremovable and irreversible fouling, as follows. Removable fouling is the one that can be removed through physical cleaning, while irremovable fouling is only removed by chemical cleaning. Removable fouling is caused by loosely attached fouling material, usually attributed to the formation of the cake layer, while irremovable fouling is caused by pore blocking and strongly attached material. Irreversible fouling is a permanent fouling that cannot be removed (Meng et al. 2009). The three major factors affecting fouling are biomass and feed characteristics, membrane operation and membrane module characteristics (Judd 2006; Meng et al. 2007). Figure 2.4 shows the factors affecting fouling in submerged MBRs as presented by Le-Clech et

al. (2006).

Figure 2.4 - Factors affecting fouling in submerged MBRs (Le-Clech et al. 2006).

The correct interpretation of Figure 2.4 is that all parameters involved in designing and operating MBRs have an impact on fouling. The impact can be either direct or indirect. A direct impact on fouling can be attributed to the parameters that determine the quantity or quality of the materials inside or around the membrane, while membrane filtration is taking place. An indirect impact on fouling can be attributed to the parameters that determine the characteristics of the bulk activated sludge, such as temperature. Lower permeability was reported during winter time (Wedi 2006). However, temperature affects the biochemical and physical characteristics of the activated sludge directly and the resulting membrane fouling indirectly. Nevertheless, the same parameter can have both a direct and indirect effect on fouling. Fouling is the consequence of the interactions between all the parameters, mentioned in Figure 2.4, which makes it extremely difficult to describe and predict. From a mechanistic point of view, fouling has been associated with three mechanisms, namely: adsorption, pore blocking and cake filtration. If foulants are smaller than or similar in size to the membrane pores, such as colloids and solutes, they can be absorbed to the pore wall or pore blocking may occur; if foulants are much larger than the membrane pores, such as sludge flocs and colloids, they tend to form a cake layer on the

18

membrane surface (Meng et al. 2009). Figure 2.5 shows the three aforementioned mechanisms.

(a) (b) (c)

Figure 2.5 - Fouling mechanisms: adsorption (a), pore blocking (b) and cake filtration (c). Once the cake layer is formed on the membrane surface, its permeability can be affected by flux, electrostatic interactions and particle size (Le-Clech et al. 2006). The hydrodynamic conditions, near the membrane, play a decisive role in the control of removable fouling. In cross-flow filtration, it is generally understood that a turbulent regime is preferable to a laminar regime to achieve successful process operation (Rosenberger et al. 2002c).

The cake layer formation, linked to removable fouling, was considered the major contributor to membrane fouling (Meng et al. 2009). The physical cause of cake filtration is associated with the resistance caused by a boundary layer around the membrane itself. This boundary layer is composed of particles of different sizes but always larger than the pore size of the membrane. A cake layer is a porous medium with a complex system of interconnected particle voids where the flow direction may even be opposite to that of the pressure gradient (Yang et al. 2007). In the cake layer and in the bulk activated sludge, macromolecules are understood as particles that link all particles together (Le-Clech et al. 2006). In literature, these macromolecules are named as Extracellular Polymeric Substances (EPS). The term EPS includes all sorts of macromolecules, such as polysaccharides, proteins, nucleic acids, lipids and other polymeric substances that compose microorganisms. EPS can be divided into bound EPS: materials integrated in biological flocs, and soluble EPS: materials not integrated in biological flocs but present in the activated sludge supernatant. The latter are also referred to as Soluble Microbial Products (SMP). EPS was identified as a very important fouling parameter (Cho et al. 2002). Bound EPS was found to be crucial in determining the specific cake resistance (Cho et al. 2005a). Other scientists focused on SMP as a major fouling parameter, specifically in the form of polysaccharides (Lesjean et al. 2005; Rosenberger et al. 2005). SRT is known to influence EPS in its bound and soluble form: Ng et al. (2006) recorded an increase in bound EPS and SMP with decreasing SRT. Massé et al. (2006) reported accumulation of soluble and colloidal proteins and polysaccharides, which were progressively degraded as SRT increased. Other factors such as the type of wastewater, sludge loading rate, MLSS

19

concentration, mechanical stress and unsteady states were found to be capable of influencing SMP concentrations (Drews et al. 2006). Opposite results have recently been published concerning the effect of SMP concentrations on fouling behavior. Lyko et al. (2008), in a long-term monitoring study on a full-scale MBR, concluded that SMP concentrations are inappropriate indicators of sludge filterability. Geilvoet (2010) reached similar conclusions. As a justification for these results, several authors propose that the influence of SMP concentrations on fouling propensity is only verified under low SRT conditions (Drews et al. 2008; Kimura et al. 2009). Drews et al. (2008) proposes an SRT less than 8 days and Kimura et al. (2009) less than 17 days, to obtain a relationship between SMP concentrations and fouling. Noteworthy to mention that, the applied SRT contributes to define the biological characteristics in each system. Therefore, in process operation of a full-scale MBR installation SRT is not easily varied. The dissolved organic matter was also identified as a possible fouling component. Cicek et al. (2003) concluded that the permeate flux was affected by the concentration of soluble organic compounds. Miyoshi et al. (2009) separated the reversible and irreversible fractions of fouling and concluded that the reversible fouling was most likely caused by the dissolved organic matter in the activated sludge. Wu et al. (2009) analyzed the correlations between filterability and major factors affecting membrane fouling and conclude that the soluble organic matter and polysaccharide concentration were the most significant factors affecting membrane filterability. Recently, a link was identified between filterability and submicron particles (Leiknes et

al. 2006; Geilvoet et al. 2007a). According to the latter authors, worse filterability is related to a higher amount of particles in the submicron range, i.e. smaller than 0.5 µm. Ivanovic et al. (2008) concluded that the relevant particle size range for membrane filtration is less than 0.5 µm, in particular, less than 0.1 µm. Furthermore, Geilvoet (2010) showed that the volume of submicron particles, defined as the accumulated volume of particles in the size range of 0.4 to 1 µm, is a good indicator of sludge filterability. Figure 2.6 shows the relationship between filterability and the volume of submicron particles, as obtained by Geilvoet (2010).

20

Figure 2.6 - Filterability vs. submicron particle volume (Geilvoet 2010).

The bound EPS, SMP concentrations, dissolved organic matter and submicron particles were identified as major fouling components. However, there is no universally agreed constituent of the mixed liquor to which fouling can be primarily attributed (Judd 2008). Overall, the cake layer imposes additional resistance on the membrane and hence has a negative effect on filtration. In contrast, the cake layer has also been seen as a pre-filter or a protective barrier to the membrane, thus with a positive effect on filtration (Lee et al. 2001). The extent of the negative or positive effect of the cake layer on membrane filtration has been evaluated through its properties such as specific cake resistance and compressibility. The specific cake resistance depends on the density, particle diameter and porosity of the cake layer. In a constant flux filtration experiment, the specific cake resistance will increase during filtration (Geilvoet 2010). Compressibility expresses the compression potential of the cake layer and varies between 0 and 1 for incompressible and highly compressible cake layers, respectively (Belfort et al. 1994). Compressibility can vary considerably according to the nature of the bound or soluble extracellular components (Le-Clech et al. 2006). Given that fouling is one of the major drawbacks of MBR technology, several cleaning and mitigation techniques have been developed. In submerged MBRs, membrane fouling is usually reduced by coarse bubble aeration (Judd 2008). Aeration, to supply DO to the biomass, is obtained through fine bubble aerators to maximize the mass transfer between the gas and liquid phase (Judd 2006). However, this strategy has severe cost implications. Aeration is the main power requirement in an MBR (Germain et al. 2005) and energy demand increases roughly linearly with increasing air flow rate (Judd 2005). Fouling removal is carried out by physical and chemical cleaning. Physical cleaning includes membrane relaxation and backwashing. Membrane relaxation can be defined as ceasing the permeation whilst continuing to scour the membrane with air bubles (Judd 2006). Chemical cleaning can be performed as enhanced backwash, maintenance cleaning or intensive cleaning (Le-Clech et al. 2006). Chemical cleaning requires

21

downtime and intensive cleaning produces chemical waste (Judd 2005). Furthermore, several chemical cleaning strategies have been developed, but the science of chemical cleaning in MBRs requires further research (Judd 2008). Limitation of fouling has also been achieved by optimizing the membrane properties and operating conditions and by modifying the biomass characteristics through the addition of coagulant/flocculent or adsorbent agents such as power activated carbon (Koseoglu et al. 2008; Iversen et al. 2009b; Remy et al. 2009). Lee et al. (2003) showed that larger flocs produce cake layers of lower specific resistance as opposed to smaller flocs. The work of Cho et al. (2005c) confirms the previous findings, because it showed an improvement of membrane permeability with coagulated flocs of lower fractal dimension, which tend to have higher porosity and aggregate more loosely. Membrane fouling is expected to be more severe with hydrophobic than hydrophilic membranes (Madaeni et al. 1999). However, after the cake layer is formed on the membrane, the membrane chemical characteristics are expected to become secondary to those of the sludge materials covering the membrane (Le-Clech et al. 2006).

2.3 Membrane Bioreactor Activated Sludge Quality

The membrane bioreactor activated sludge quality can either refer to the quality of the bulk activated sludge or to the quality of the activated sludge in direct contact with the membrane that is either adsorbed into the membrane pore walls or onto the membrane surface. In this research, we focused on the quality of the bulk activated sludge.

How to measure MBR activated sludge quality

In order to obtain reliable measurements, activated sludge samples should be collected after the MBR activated sludge is acclimatized and a steady state is reached (Cicek et al. 1998; Le-Clech et al. 2003). Le-Clech et al. (2003) reported a period of at least two SRT to allow proper biomass acclimatization. Membrane bioreactor activated sludge quality can be measured either through bulk parameters or through specific components. One example of a bulk parameter is filterability and one example of specific components is the MLSS concentration. The disadvantage of measuring bulk parameters is that the effect of specific components cannot be distinguished and the disadvantage of measuring specific components parameters is that synergetic effects are neglected. For a more complete characterization of the activated sludge quality, both types of parameters, bulk and specific components, should be used. The suspended fraction of the activated sludge, known as MLSS, is a physical property described by the “Standard Methods for the Examination of Water and Wastewater” (APHA 2000) as the portion of total solids retained by a filter under specific conditions. Several other physical and chemical parameters, such as polysaccharide or protein concentrations, are also used to characterize activated sludge fractions.

22

APHA (2000) describes the following bulk parameters that could be used to measure MBR activated quality:

- Oxygen Consumption Rate: determination of the oxygen consumption rate of a sample of biological suspension such as activated sludge; when used as a routine test, it will indicate changes in operating conditions.

- Settled Sludge Volume: determination of the settled volume of an activated sludge sample for a given time interval; the 30 minutes settled sludge volume is used to determine the Sludge Volume Index (SVI); useful in routine monitoring of biological suspensions.

- SVI: determination of the volume occupied by 1 g of suspension after 30 minutes of settling; useful in routine process control.

- Capillary Suction Time (CST): determination of the rate of water released from activated sludge; useful for assisting sludge dewatering and conditioning processes and evaluating coagulation effects on the rate of water release from sludge.

- Time-to-Filter (TTF): determines the time required to produce 100 mL of filtrate; useful for sludge dewatering and conditioning processes.

The aforementioned parameters are particularly appropriate for conventional activated sludge processes (APHA 2000). They can be performed on MBR activated sludge, but their usefulness is debatable, as explained below. The most recent MBRs are designed for long SRT, i.e. higher than 20 d, and low Food to Microorganisms (F/M) ratios. Consequently, it is expected that the oxygen consumption of the biomass will be mainly due to endogenous respiration, i.e. the main consumption of oxygen will result from the degradation of the decaying biomass. Therefore, the obtained results are site-specific and their interpretation is only possible when they are indeed used as a routine procedure in a pilot- or full-scale MBR. The Settled Sludge Volume and the more specific SVI parameters mainly qualify the activated sludge in terms of its settling properties, while in MBRs, the separation of the sludge and clean water does not occur through settling but through membrane separation. The aforementioned parameters do give indications concerning the flocculation status of the activated sludge. However, practice shows that membrane tank MBR sludge hardly settles, so to measure the SVI, the activated sludge needs to be diluted, what will change the flocculation conditions of the sludge. The determination of dilution SVI is only defined in national standard methods, such as the Dutch and German standards, but not in international standards. The CST and TTF are useful parameters to determine sludge dewatering and conditioning, thus, to handle the sludge flow of the MBR but not the liquid flow. All the aforementioned parameters have limitations when applied to MBR activated sludge. Due to this fact, the EUROMBRA project produced the D3-Deliverable where the useful parameters to characterize bulk MBR activated sludge were described (UNITN 2006). In this report, the following parameters were proposed:

23

- MLSS: quantifies the amount of suspended solids within the activated sludge. - Mixed Liquor Volatile Suspended Solids (MLVSS): quantifies the volatile

fraction of the MLSS; gives an indication of the organic matter content of the sludge.

- Specific Resistance to Filtration (SRF): quantifies the average cake resistance to filtration per mass unit through an unstirred dead-end filtration test.

- Viscosity: quantifies the rate of displacement of a fluid with a given shear force. - Particle size distribution: quantifies the amount of particles according to their size

in a specific size range. In addition, the D3-Deliverable (UNITN 2006) also proposed Oxygen Consumption Rate, SVI and CST as useful parameters to characterize MBR activated sludge. The parameters list proposed by the EUROMBRA project has only one method that quantifies MBR activated sludge through a filtration step, the SRF. MBR activated sludge, contrary to CAS activated sludge, will be separated from clarified water through membrane filtration. Therefore, including a filtration step in the determination of the MBR activated sludge quality can provide extra information about the filtration operation. However, while in municipal MBRs membranes work in semi-dead-end mode, in SRF determination sludge is filtered in a dead-end mode. Several partners of the EUROMBRA and AMADEUS projects have developed specific methods and corresponding parameters to qualify MBR activated sludge using a membrane filtration step. The methods and corresponding parameters are listed as follows:

- Cranfield University developed the flux step method (Le-Clech et al. 2003), which measures the critical flux (Jc) that is the flux above which permeability decreases and severe fouling occurs. There is no standard set-up and protocol defined for this method. The Jc measurement is an ex situ filtration measurement.

- TU Delft developed a method, the Delft Filtration Characterization method (DFCm) (Evenblij et al. 2005), to measure activated sludge filterability, using the Delft Filtration Characterization installation (DFCi), working in cross-flow mode and operated under set parameters. The DFCm is an ex situ filtration measurement.

- The “Vlaamse Instelling voor Technologish Onderzoek” (VITO) developed the MBR-VITO Fouling Measurement method (VFMm) (Huyskens et al. 2008). This method comprises a specific installation that works in cross-flow mode, driven by aeration. The measurement has a specific procedure, is performed at constant pressure and provides permeability results, later converted to fouling resistances of reversible and irreversible fouling. The VFMm can be applied as ex situ and in

situ filtration measurement. - TU Berlin and “Kompetenzzentrum Wasser Berlin” (KWB) developed the Berlin

Filtration method (BFM) (De la Torre et al. 2009). The BMF measures Jc in a specific installation that works in cross-flow mode, driven by aeration. The BMF is an in situ filtration measurement.

24

What is the bulk MBR activated sludge quality? The MLSS concentration in an MBR is controlled by the applied SRT that defines the amount of sludge regularly withdrawn from the MBR. In full-scale municipal MBRs, MLSS operational concentrations vary between 8 and 18 g/L (Judd 2008). MLSS concentrations between 10 and 15 g/L seem to be practical with regard to sludge production and excess sludge discharge, although in the USA there is a recent trend towards designing MBR installations to operate with MLSS concentrations lower than 10 g/L (van Nieuwenhuijzen et al. 2008). Currently, maximum MLSS concentrations are limited by the oxygen transfer to the biomass, explained as follows. As mentioned earlier, aeration has a triple function in submerged MBRs: it maintains the solids in suspension, provides DO to the biomass and suppresses fouling. According to the MBR design, i.e. with or without a separate membrane tank, the aeration functions may be split or not. When the MBR has a separate membrane tank, carbonaceous removal mainly occurs in the carbonaceous removal tanks; when the membranes are submerged in the carbonaceous removal tanks, all aeration functions have to be fulfilled in the same tank. The oxygen transfer to the biomass is usually characterized by the overall mass transfer coefficient KLa and the α correction factor, i.e. the ratio between the activated sludge KLa and clean water KLa (Germain et al. 2005). The latter authors described several studies where an exponential relationship between the α-factor and the MLSS concentration of the activated sludge was found, as shown in Figure 2.7.

Figure 2.7- Alpha factor vs. MLSS concentration (Germain et al. 2005).

Germain et al. (2005) concluded that operating at MLSS concentrations below 15 g/L could significantly improve the oxygen transfer to the biomass. Therefore, in order to improve the oxygen transfer to the biomass, in MBRs with a separate membrane tank, a maximum MLSS concentration of 15 g/L should be applied in the carbonaceous removal tanks, while in MBRs without a separate membrane tank, the referred maximum concentration should be applied in the membrane/carbonaceous removal tanks. MBR activated sludge is a non-Newtonian fluid, meaning that activated sludge does not have a constant viscosity, in particular MBR activated sludge has a decreasing apparent

25

viscosity with increasing applied shear rate (Rosenberger et al. 2002c; Hasar et al. 2004; Meng et al. 2007). The behavior of MBR viscosity has also been referred to as pseudo-plastic, i.e. the particles tend to flocculate in a large network that, when disrupted, by increasing the applied shear rate, results in a decrease in viscosity (Rosenberger et al. 2002c). Several models were proposed where the apparent viscosity of the MBR activated sludge was calculated as a function of MLSS concentration, shear rate and temperature (Rosenberger et al. 2002c; Laera et al. 2007; Yang et al. 2009). The models proposed by the abovementioned authors are presented in Table II-II.

Table II-II Models for determining the viscosity of MBR activated sludge at 20ºC.

Reference

Equation

Rosenberger et al. (2002c) 0.41 0.372 0.23MLSS MLSSC C

eη γ× − ×= ×

Laera et al. (2007) 0.494 0.6310.882 0.05MLSS MLSSC C

eη γ× − ×= ×

Yang et al. (2009) 1.359 0.80732.36MLSS

Cη γ −= × ×

Key: η- apparent viscosity of activated sludge [mP s]; CMLSS- MLSS concentration [g/L];γ - shear rate [s-1].

In a lower extent than MLSS concentration, shear rate and temperature, factors such as mechanical stress and SMP concentrations were also reported to affect viscosity. (Rosenberger et al. 2002c). The MBR floc size is measured through particle counting in the suspended fraction of the sludge. A bimodal particle size distribution, for MBR activated sludge, with one peak in the colloidal fraction and another peak in the suspended particle range, has been reported (Wisniewski et al. 1998; Leiknes et al. 2006; Sombatsompop et al. 2006). Lawler (1997) mathematically defined particle size distributions and concluded that a variable β, i.e. the slope of the normalized particle counting data, when apparent, should be between set limits. The possible particle size distributions defined by Lawler (1997) are presented in Figure 2.8.

26

Figure 2.8 - Particle size distribution graphs (Lawler 1997).

Ceronio et al. (2005) analyzed the β value in extensive sets of data and concluded that, for full-range particle size characterization, a variable β was fundamentally more correct, but a constant β could also be accurate in certain size range fractions of the particle size distribution. In any case, sizing particles does have limitations. Particles have multiple shapes and, depending on the angle of observation, the diameter of a particle can vary at least by a factor of 10 (Boller et al. 2011). Nevertheless, particle counting is defined as standard method (APHA 2000), but the limitations of each applied method should be identified and discussed. Advantages and disadvantages of each particle counting method are available in literature (Lawler 1997; Hargesheimer et al. 1998; APHA 2000). The MBR floc size was reported to be smaller than flocs in CAS systems, with a size range between 10 and 40 µm (Bae et al. 2005). According to Gregory (1998) increased aeration is likely to lead to restructuring and densification of flocs. Other research works provided different results: Huang et al. (2001) measured MBR floc sizes between 15 and 48 µ; Chang et al. (1999) measured the size of the MBR floc as 131 µm; and Massé et al. (2006) reported MBR floc sizes from 120-220 µm to 70-100 µm. According to Germain et al. (2005) particle sizes differ from one MBR to another because of the specific characteristics of aeration, biomass and feed. Regardless of the absolute size measured,

27

all of the reported MBR floc sizes are considerably larger than the nominal pore size of UF membranes, so they are not expected to directly block the membrane pores. Nevertheless, regardless of its size, it is understood that biological flocs play a key role in the formation of the cake layer on the membrane surface (Le-Clech et al. 2006). The size of the particles is affected by aeration and return ratio. The airflow has a dual effect on the activated sludge, mainly through the available DO concentration and mixing intensity, both of them with consequences on the size of the particles (Germain et al. 2005). Abbassi et al. (1999) worked on the quantification of the effects of DO and mixing intensity to minimize sludge production and found that 10% of the floc break-up was attributed to the mixing intensity, while 12% was attributed to increased DO concentration. Ji et al. (2006) provided similar conclusions, i.e. the increase in aeration lead to floc breakage. Therefore, an increase in DO concentration and mixing intensity, caused by a higher airflow, reduces the particle size by floc break-up. Wisniewski et al.

(1998) studied the effects of return ratio on particle size in a side-stream MBR and found that activated sludge flocs were broken mainly by the return pumps; without return flow, only 15% of the flocs were smaller than 100 µm, while with return flow, 98% was smaller than 100 µm. Stricot et al. (2010) investigated the influence of hydrodynamics on floc structure and concluded that strong shear stresses lead to floc breakage. Thus, return flow, as well as an increase in DO concentration and mixing intensity, reduces the sludge floc size.

2.4 Impact of Mixed Liquor Suspended Solids Concentration on Membrane Filtration

When the new generation of MBRs was presented by Yamamoto et al. (1989), an MLSS concentration threshold of 40 g/L was proposed. MLSS concentrations above the aforementioned concentration would lead to an abrupt flux decline. However, which would be the most suitable MLSS value, or range of values, for MBR design was still to be defined. Several attempts have been made to clarify the impact of MLSS on membrane filtration. Table II-III presents results provided by several authors concerning the impact of MLSS on MBR fouling.

28

Table II-III Impact of MLSS concentration shift on MBR fouling MLSS Shift (g/L) Fouling parameters References Fouling increase 4→18 Cake and pore blocking

resistance: 10→120 1011/m (Meng et al. 2007)

12→18 Normalized permeability decreased 10%

(Trussell et al. 2007)

3→10 Increased fouling potential (Psoch et al. 2006) 0.09→3.7 Cake resistance: 21→54 1011/m (Chang et al. 2005) 2.4→9.6 Total resistance: 9→22 1011/m (Fang et al. 2005) 7→18 Critical flux: 47→36 L/m2.h (Han et al. 2005) 2.1→15.4 Limiting flux:105→50 L/m2.h (Cicek et al. 1998) Fouling decrease 3.5→10 Critical flux:>80→<60 L/m2.h (Defrance et al. 1999) No (or little effect) 4.4→11.6 No impact between 4-8 g/L;

Slightly less fouling for 12 g/L (Le-Clech et al. 2003)

4→15.1 Critical flux: 25→22 L/m2.h (Bouhabila et al. 1998) 1→20 Equilibrium flux: 2.5→1.2

L/m2.d (Harada et al. 1994)

(Adapted from Judd (2006)) Table II-III shows that literature reports on the effect of MLSS on membrane filtration are not consistent and sometimes contradictory. Some scientists report increased fouling with increasing MLSS concentrations, while others see no effect or present opposite results. Meng et al. (2007) observed that membrane fouling increased exponentially with increasing MLSS concentrations. Aeration intensity had minor effects on membrane fouling when MLSS concentration was below 10 g/L, but significantly mitigated membrane fouling when MLSS was above 10 g/L. When MLSS concentrations were between 8 and 12 g/L, there was a dramatic increase in viscosity that would worsen the filtration process. When sludge viscosity was higher than 3 mPa.s, then cross-flow velocity would decrease sharply. The authors concluded that MLSS concentrations in MBRs should be kept below 12 g/L. Viscosity is closely related to MLSS concentration (Rosenberger et al. 2002c; Hasar et al. 2004; Laera et al. 2007; Meng et al. 2007; Yang et al. 2009). A critical value of MLSS concentration, around 10 g/L, in terms of its effect on viscosity was identified; with MLSS concentration under the critical value, viscosity remains low and increases slowly with MLSS concentration; above the MLSS critical value, there is an exponential increase in viscosity with raising MLSS concentrations (Itonaga et al. 2004). In a reported work, about full-scale MBR activated sludge characterization, the critical MLSS concentration, in terms of its effect on viscosity, was identified as being 14 g/L (Reid et

29

al. 2008). High viscosities are known to reduce the mass transfer of oxygen, thus affecting the DO concentration in the system (Germain et al. 2005). According to Germain et al. (2005), high MLSS concentrations decrease the oxygen transfer, but the importance of the oxygen transfer drops when MLSS concentration is increased depends on the characteristics of the system. In the aforementioned review paper, several studies showed an exponential decreasing relationship between α-factor and increasing MLSS concentration. Germain et al. (2005) concluded that by operating at MLSS concentrations lower than 15 g/L, oxygen transfer could be significantly improved. Aeration efficiency, and the consequently necessary energy input, seems to be the limiting factor for a maximum MLSS concentration of around 15 g/L (van Nieuwenhuijzen et al. 2008). Opposite conclusions were obtained by other authors. Le-Clech et al. (2003) saw no effect on fouling for MLSS concentrations between 4 and 8 g/L, but for MLSS concentrations of 12 g/L, the Jc increased, indicating that a protective layer may have formed on the membrane. According to the authors, MLSS concentrations were mainly affecting the build-up of pressure around the membrane. With higher MLSS concentrations, the pressure on the membrane was stronger, but Jc was indeed improved. In a review paper about fouling in submerged membrane bioreactors, a decrease in fouling at high MLSS concentrations coinciding with a decrease in EPS and SMP levels was reported (Judd 2005). The results obtained are shown in Figure 2.9.

Figure 2.9 - Carbohydrate and protein fractions of EPS (EPSc and EPSp, respectively)

and SMP (SMPc) extracted from a single bioreactor operating at increasing MLSS concentrations (Judd (2005), referring to Le-Clech, P. (2003) Process configuration and

fouling in membrane bioreactors, PhD thesis, Cranfield University, UK). The results shown in Figure 2.5 demonstrate that in certain MBRs and operating conditions, increasing the MLSS concentration, i.e. concentrating, can bring about an improvement in the activated sludge characteristics. The absence of a clear correlation between MLSS concentration and fouling parameters leads to the conclusion that MLSS alone is a poor indicator of biomass fouling propensity

30

(Jefferson et al. 2004). This conclusion implies that MLSS should not be seen as the only indicator of fouling propensity, but it does not exclude it as a fouling parameter. This reasoning is supported by several empirical mathematical relationships developed to quantify total resistance (Lee et al. 2002), rate of filtration resistance (Liu et al. 2003), specific cake resistance (Cho et al. 2005b) and fouling resistance (Meng et al. 2006) using MLSS concentration as one of the parameters. It can be concluded that MLSS concentration alone does not explain fouling, but it does affect it. The activated sludge mixed liquor can be fractionated into three components, according to their physical size, namely: suspended solids, colloids and solutes. According to Metcalf & Eddy (2003), suspended solids have sizes of up to 1 µm, colloids are smaller than 1 µm and larger than 0.01 µm, and solutes are smaller than 0.01 µm. However, according to the author, the size range of the different sludge fractions can vary.

Several studies were carried out to identify the relative contribution of the different fractions of biomass to MBR fouling. Figure 2.10 shows the results obtained by several researchers.

Figure 2.10 - Relative contributions of the different biomass fractions to MBR fouling

(Judd 2006). As Figure 2.10 shows, the results are extremely variable, meaning that all biomass fractions were already identified as being the main fouling component. The differences in results were attributed to different biomass characteristics and operating conditions in each research work.

Itonaga et al. (2004) evaluated the contribution of the various biomass fractions to the overall fouling resistance in two MBR pilots. In the MBR pilot with a specific pre-treatment, consisting of coagulation and sedimentation, the soluble fraction was the main contributor to fouling, where as in the MBR pilot without the aforementioned pre-treatment, it was the colloidal fraction. Min et al. (2008) verified that, under low DO conditions, worse filterability was measured, due to the presence of a larger amount of high molecular weight compounds in the suspended fraction of the sludge. Bae et al. (2005) observed that the fouling contribution of each biomass fraction varied with

31

permeability and hydrodynamic conditions and appeared to be dependent on particle size. In the aforementioned study, the main activated sludge fraction contributing to the fouling process consisted of suspended solids, but the MBR wastewater was fed with synthetic substrate, which is easily biodegradable. In these conditions, biomass is likely to produce less waste material and, therefore, lower amounts of colloids and solutes will be available in the bioreactors. Overall, there is a tendency to consider colloids and solutes as the main fouling activated sludge fractions (Le-Clech et al. 2006). However, the same authors who remark this also state that biological flocs, i.e. the suspended fraction of the activated sludge, play a key role in the formation of the cake layer. MLSS concentration is the result of operational choices, mainly the applied SRT. On the other hand, from a physical point of view, the existing MLSS concentration affects viscosity, oxygen transfer in MBRs and particle size. Le-Clech et al. (2003) measured a reduction in floc size from 200 to 50 µm when MLSS concentration was shifted from 3.5 to 10 g/L. The complexity of the relationship between MLSS and membrane filtration is probably due to the fact that MLSS concentration has both a direct and indirect effect on membrane filtration. The direct effect can be defined as the deposition of suspended particles on the membrane surface. This direct effect occurs, in cross-filtration mode, with particles larger than the membrane pore size but small enough to avoid being dragged by the shear forces near the membrane. The indirect effect results from the contribution of MLSS concentration to the overall quality of the bulk activated sludge in an MBR. The suspended material in the bulk MBR activated sludge is aggregated in flocs that retain colloids and dissolved components. The impact of MLSS concentration on membrane filtration is complex because the direct and indirect effects of MLSS concentration on membrane filtration can hardly be dissociated.

2.5 Permeate Quality

In literature, the permeate quality of an MBR can be addressed directly or indirectly. Many pilot- or lab-scale studies are conducted to evaluate the treatability of a specific influent and/or membrane performance, therefore, removal efficiencies of Chemical Oxygen Demand (COD), Biochemical Oxygen Demand (BOD) and nutrients are usually provided (van Nieuwenhuijzen et al. 2008). The removal efficiency rates provide direct information about the performance of the MBR, but only show indirectly the permeate quality. Direct values of MBR permeate quality are available mainly when MBR is coupled with another treatment, usually RO, for reuse purposes. It is generally accepted that an MBR produces a rather disinfected effluent. MBR technology achieves log removals of total coliforms of more than 6.6, making it suitable for the post-treatment of effluents discharged in sensitive water bodies (van Nieuwenhuijzen et al. 2008). However, the removal of microorganisms is only effective if the membranes are intact. Membrane integrity can be tested by either direct or indirect measurements. Direct tests measure changes in pressure, air flow or sound, whose magnitude is a direct function of membrane breaches, while indirect methods rely on

32

water quality parameters, such as turbidity and particle counts, analyzed in the permeate, as a surrogate measure of integrity loss (WEF 2006; Guo et al. 2010). In general, water treatment membrane suppliers have their own integrity test procedures (Oxtoby 2003; Guo et al. 2010). The pressure decay test is the most frequently used direct method for membrane integrity monitoring, commonly used in water treatment plants (Oxtoby 2003; WEF 2006). However, some drinking water experts argue that pressure decay tests are limited to the detection of membrane defects around 3 µm in diameter (Lebleu et al. 2009) and can provide false-negative results (Guo et al. 2010). Direct integrity tests, particularly pressure decay tests, are not practical for wastewater treatment plants, since the level of detection is far beyond that which is required to maintain the effluent quality requirements and in-tank testing is not feasible (WEF 2006). Therefore, the primary means of determining membrane integrity in a MBR is by turbidity of the permeate (WEF 2006). Turbidity monitoring is less expensive than particle counting but offers lower sensitivity, i.e. even permeate with turbidity below 0.1 NTU can have significant bacterial contamination (Guo et al. 2010). Particle counting has widespread use in the water treatment industry and is more sensitive than particle and turbidity monitoring (Guo et al. 2010). Additionally, new indirect membrane integrity tests, applying membrane-based sensors, have also been described (Phattaranawik et al. 2008). One widely recognized characteristic of MBR permeate is the absence of suspended solids (Metcalf & Eddy 2003). On the other hand, MBR technology does not seem to remove endocrine disruptive compounds and micropollutants at the desired levels (van Nieuwenhuijzen et al. 2008). MBR permeate quality is also considered extremely stable. Organic shock loads, with a COD of 4000mg/L in the MBR influent, do not cause significant changes in the permeate quality (Rosenberger et al. 2002b). Direct results of MBR permeate quality were reported by Remy et al. (2009). In this research, COD values between 36 and 30 mg/L were measured in the permeate of two pilot-scale MBRs fed with municipal wastewater. In addition, SMP, namely polysaccharide and protein concentrations, were 3.7 to 7.4 mg/L and 3.4 to 4.3 mg/L, respectively. SMP are, by definition, soluble products, therefore, smaller than 0.01 µm according to mixed liquor size fractionation proposed by Metcalf & Eddy (2003). In the research described by Remy et al. (2009), the nominal pore size of the membranes is 0.2 µm, thus, it is not surprising that particles smaller than the nominal pore size of the membrane are measured in the MBR permeate. The values of MBR permeate quality provided by scientists working on coupling MBR with RO are presented in Table II-IV.

33

Table II-IV Permeate quality of municipal submerged MBRs. References

Qin et al.

(2006) Gur-Reznik

et al. (2008) pH 6.2-6.4 - Total Suspended Solids mg/L - - Turbidity NTU - <1 Conductivity µS/cm - - Total Organic Carbon mg/L 4.9-5.1 9.6-12.6 Chemical Oxygen Demand mg/L - 19.6-65.6 Biochemical Oxygen Demand mg/L - 0-8.8 T-Phosphorous mg/L - - PO4 mg/L - 6-8 T-Nitrogen mg/L - - NH4-N mg/L 0.05-0.62 Undetectable

NO3 mg/L 17.6-22.8 6.5-10.5 Total coliforms CFU/100 mL - - Faecal coliforms CFU/100 mL - -

The values presented in Table II-IV confirm that the MBR permeate is almost free of suspended solids and is a rather disinfected effluent. However, the values of Total Organic Carbon (TOC) and BOD shown in Table II-IV indicate that some organic material is found in the permeate composition. In addition, values of COD, coupled with BOD results, and nutrient concentrations show that inorganic material is also present in the MBR permeate. The aforementioned concentrations are most likely caused by dissolved material. A size analysis of the MBR permeate can be performed through particle counting. However, very few particle counting results in MBR permeate are available. Poyatos et al.

(2010) reported the existence of particles, in the range of 2-5 µm, in the permeate of a pilot MBR fed with municipal wastewater. The existence of particles with sizes between 1 and 32 µm and their aggregation in drinking water distribution systems was observed (Verberk et al. 2007). A slow deposition of particles followed by growth and re-suspension, therefore, a biochemical and physical process, caused discoloration in drinking water distribution systems (Vreeburg et al. 2008). Poyatos et al. (2010) reported that an increased number of particles in the range of 2-5 µm were obtained immediately after chemical cleaning. Fenu et al. (2009) measured scaling on the permeate side of the membrane with precipitation of soluble salts such as CaCO3, Fe(OH)2 and Ca3(PO4)2 during chemical cleaning, due to increase of pH resulting from CO2 release. These results indicate that inorganic particles, present in the MBR permeate, can be attributed to the chemical by-products produced in the system when chemicals are back-flushed through the permeate side of the membrane. This procedure is current practice in hollow fiber membranes. Changes in pH, occurring when chemicals are introduced, can lead to precipitation of soluble salts in the MBR permeate. Nevertheless, the reported results on MBR permeate may provide an explanation for the existence of inorganic particles but do not inform about possible organic components.

34

2.6 Conclusions

A review on MBR technology fundamentals was presented to introduce the main concepts applied to municipal submerged MBRs. At present, MBR design still raises challenges and questions, such as what is the most effective MBR tank sequence and which operational parameters should be selected to obtain an optimal design that provides successful treatment at a minimum cost. The available parameters to quantify MBR activated sludge were presented and discussed, as well as the current knowledge about MBR activated sludge quality, particularly MLSS concentration, viscosity, floc size and particle size distribution. The effect of MLSS concentration on membrane filtration still raises controversy. Optimal MLSS concentrations for MBR design and operation vary between authors and locations. The effect of MLSS concentration on fouling, one of the main disadvantages of MBR technology is also not fully understood. Furthermore, MLSS concentrations are known to affect viscosity and oxygen transfer, limiting the operating MLSS concentrations. The impact of MLSS concentration on membrane filtration has a complex interpretation, with both direct and indirect effects. It can be concluded that MLSS concentration alone does not explain fouling, but it certainly affects the process. MBR permeate quality was described. MBR permeate is a rather disinfected effluent relatively free of suspended solids. However, it also contains organic and inorganic material, most likely in the dissolved form.

35

3. Framework and Clarification of Concepts

3.1 Introduction

Between 2006 and 2010, three PhD theses concerning filterability measurements in MBR activated sludge were successfully defended at the Water Management Department of the Delft University of Technology (TU Delft). The concept of this thesis was mainly based on the conclusions provided by Evenblij (2006) and Geilvoet (2010). The research work here presented was developed in parallel with the research work of Adrien Moreau, described in his thesis (Moreau 2010). Between 2006 and 2010, the Water Management Department of the Delft University of Technology was involved in two European projects, EUROMBRA and MBR-Train, dealing with membrane bioreactor technology3. These projects sponsored the PhD of Adrien Moreau and of the present author and enabled the creation of an MBR network of contacts that allowed the performance of several on site measurement campaigns in MBR installations scattered across Europe. The first two sections of this chapter summarize the main conclusions described in the three aforementioned theses, Evenblij (2006), Geilvoet (2010) and Moreau (2010), which are relevant for the present thesis. The subsequent three sections describe the drivers and clarify the concepts of the research work described in this thesis.

3.2 The Delft Filtration Characterization Method

The Delft Filtration Characterization method (DFCm) was first described in 2005 (Evenblij et al. 2005). The DFCm is presented in detail in Chapter 4. In short, the DFCm comprises a Delft Filtration Characterization installation (DFCi) and a measuring protocol. The protocol includes the determination of membrane resistance, short-term activated sludge filtration and membrane cleaning. The main outcome of the method is the filterability of the activated sludge, obtained through specific membrane and operational settings. By applying the DFCm, differences in results can be directly attributed to the activated sludge characteristics. Herman Evenblij’s thesis described the DFCm in detail and addressed key issues arising out of the method itself. The DFCm was also assessed throughout the work of Stefan Geilvoet and Adrien Moreau. The main questions and answers are summarized below. Can the DFCm quantify filterability and discriminate between different MBRs?

Yes, as proven by the successful application of the method on activated sludge from: 3 MBR pilots (Evenblij 2006), 2 full-scale MBRs (Geilvoet 2010) and 7 pilot-scale MBR and 3 full-scale MBRs (Moreau 2010). The DFCm enabled the qualification of activated sludge filtration quality through filterability.

3 http://www.mbr-network.eu/

36

Can filterability be manipulated and successfully measured?

Yes, as proven by substrate addition tests (Evenblij 2006) and stress experiments (Geilvoet 2010). The addition of easily degradable substrate, the reduction of dissolved oxygen, the increase of shear stress and temperature reduction had a negative effect on filterability, which was measured with precision through the DFCm.

What is the hydraulic regime in the DFCi when the operating conditions defined in the

DFCm are applied?

When water is flowing through the membrane tube, at the standard cross-flow velocity of 1 m/s defined in the DFCm, the regime is turbulent (Geilvoet 2010). When MBR activated sludge is flowing through the membrane tube, an analysis based on Darcy-Weisbach equations and measured pressure loss through the membrane tube showed that, for activated sludge with MLSS concentrations above 5 g/L, the regime is laminar (Geilvoet 2010). Stefan Geilvoet also fitted the cake layer filtration theory to the DFCm output. Thus, the compressibility and specific cake resistance, coupled with the concentration of particles in the membrane cake layer, could be obtained from the DFCm output (Geilvoet 2010). In the latter thesis, the referred concentration of particles is understood as concentration of fouling particles retained in the cake layer. Figure 3.1 presents examples of the DFCm output, fitted to the cake layer filtration theory, as proposed by Geilvoet (2010).

( )11

11 ssR iR c Vα −−∆ = ⋅ ⋅

Where: ∆R- Total cake layer resistance increase [m-1] αR - Specific cake resistance [m.kg-1] ci - Concentration of fouling particles [kg.m-3] s - Compressibility coefficient [-] V- Specific permeate production [L.m-2]

Figure 3.1 - Examples of the DFCm output and mathematical fit to the cake layer filtration theory (Geilvoet 2010).

Figure 3.1 shows filtration results, from different activated sludge samples, represented in logarithmic scale to allow a better understanding of the αR.ci product and compressibility coefficient contribution to the total cake layer resistance increase. Applying the cake layer filtration theory allowed us to fit a power law equation to each activated sludge filtration experiment and obtain the following 3 parameters to characterize the corresponding cake layer:

37

- ∆R20: total cake layer filtration resistance obtained when 20 L of permeate per m2 of membrane area are produced.

- αR.ci: product of the specific cake layer resistance by the concentration of fouling particles in the membrane cake layer.

- s: compressibility coefficient expressed as the compression potential of the cake layer.

How representative is the DFCm of MBR filtration behavior?

Overall, the design and operation of the DFCi does not fully reproduce the filtration process in MBR installations (Geilvoet 2010). However, this was not the aim of the DFCm. The aim was to define a method capable of qualifying MBR activated sludge. In the short-term sludge filtration test from the DFCm, the dominant fouling is removable (Geilvoet 2010). In the DFCm output, the fouling is quantified as additional resistance therefore mainly attributed to the removable fouling. The irreversible fouling is specifically related to the membrane history of the DFCi. In MBR installations, membrane performance, which expresses the consequences of fouling in membrane filtration, is usually assessed through permeability. The filterability results provided by the DFCm were compared with permeability data and it was concluded that there is a statistical correlation between filterability and permeability in hollow fiber membranes (Moreau 2010). These results can be explained through the cleaning strategy applied to each membrane configuration. The cleaning strategy is relevant because it contributes to define which type of fouling will prevail on the membrane. Can filterability be correlated with soluble microbial products or other parameters?

Filterability had some correlation with SMP concentrations when the activated sludge was manipulated in a lab environment (Evenblij 2006; Geilvoet 2010). When the DFCm was applied to monitor filterability in pilot- or full-scale MBR installations there was no correlation between filterability and SMP concentrations (Evenblij 2006; Geilvoet 2010; Moreau 2010). The volume of submicron particles in the free water, i.e. the volume of particles with a diameter between 0.4 and 1 µm, is a better indicator of filterability than SMP concentrations (Geilvoet 2010). A lower volume of submicron particles is related to a better filterability. How can the DFCm be used to obtain its maximum performance?

The DFCm should be used as a practical oriented tool placed at on site MBR installations (Geilvoet 2010). Together with operational results, in particular permeability readings, the obtained filterability enables the identification of activated sludge quality problems in an early stage and prevent extra costs to the MBR owner due, for instance, to increased aeration or cleaning procedures. The DFCm can also be used as a lab-scale research tool to manipulate and control filterability (Geilvoet 2010).

38

3.3 The Delft Filtration Characterization Method in Practice

Applications of the DFCm were performed and described by the three former PhD students: Herman Evenblij, Stefan Geilvoet and Adrien Moreau. One of the main characteristics of the DFCm application is that real wastewater sludge, collected from MBR installations, was always used. Intensive measurement campaigns and lab-scale experiments were described by Stefan Geilvoet and Adrien Moreau in their respective theses. Moreover these authors compared the DFCm with other filterability measurement methods through the respective advantages and disadvantages. At present, one clear advantage of the DFCm is the abundant gathered know-how. Stefan Geilvoet presented filterability results from two full-scale MBR plants: Varsseveld (Rijn & Ijseel Water Board, in the Netherlands) and Heenvliet (Hollandse Delta Water Board, in the Netherlands). The obtained results helped clarify DFCm answers. In addition, other conclusions were obtained, namely that MBR activated sludge filterability is closely related to flocculation and deflocculation mechanisms. The MBR process control should aim at optimal flocculation in the membrane tanks (Geilvoet 2010). Stefan Geilvoet also described lab-scale tests intended to manipulate activated sludge characteristics, in particular, MBR activated sludge filterability through DO. This author showed that MBR sludge filterability would decrease without aeration, as excepted. However, the recovery of filterability, after a period of approximately 4 days under anaerobic conditions, was 12 times faster than the decay rate. Figure 3.2 shows the results obtained by Stefan Geilvoet in the aforementioned stress experiments.

Figure 3.2 - Filterability variation over time (Geilvoet 2010).

Adrien Moreau presented filterability results obtained from an extensive on site measuring campaign that covered 7 pilot-scale MBRs and 3 full-scale MBRs scattered across Europe, in winter and summer periods. The conclusions obtained contributed to clarify the DFCm. Other conclusions are summarized as follows:

- The feed water quality, namely its toxicity, was found to be critical for filterability and operation of MBR applications. Reduced organic loadings were found to be beneficial to filterability.

39

- There is a seasonal variation of filterability linked to temperature values. The effect of temperature on apparent viscosity was found to be insignificant and did not explain the filterability variations. Therefore, the filterability variations are most likely explained by variations in the influent composition and flocculating conditions.

- Filterability results are scale-dependent, i.e. filterability in MBR pilot plants is more homogeneous and less extreme than in full-scale MBRs. Membrane configurations do not have significant impact on activated sludge filterability. The studied membrane configurations were hollow fiber, flat sheet and multi-tube membranes.

Two specific parts of Adrien Moreau thesis are especially significant for this thesis. More detailed explanations and relevant conclusions from the aforementioned parts are presented below. The MBR pilot at Cranfield University, UK, is composed of an aeration tank followed by three air lift side-stream modules, meaning that the sludge flow, through the side-stream module, is provided by aeration, quantified as Specific Aeration Demand (SAD). In the experiments described by Moreau (2010) at the referred MBR pilot, the SAD was decreased, leading to significant filterability improvements, in the flat sheet and hollow fiber configurations. Moreover, filterability could be clearly related to an increase in MLSS concentrations. Figure 3.3 shows the filterability vs. MLSS concentrations obtained in the aforementioned experiments.

Figure 3.3 - Filterability vs. MLSS concentration in membrane tanks with flat sheet (FS),

hollow fiber (HF) and multi-tube (MT) membranes (Moreau 2010). The improvement of filterability, coupled with increasing MLSS concentrations, represented in Figure 3.3, was explained as a case of sludge thickening. The choice of words was justified by the fact that filtration and gravity thickening are governed by identical and general force balances (Moreau 2010). According to this author, sludge thickening enabled the retention of fouling particles, resulting in improved filterability.

40

The DFCm output, obtained in the field campaigns at 5 pilot-scale MBRs and 3 full-scale MBRs, was used to obtain information about the cake layer built up, namely through compressibility, specific cake resistance and concentration of fouling particles on the membrane cake layer. Figure 3.4 presents the aforementioned results.

a) b)

Figure 3.4 - Specific cake resistance x fouling particles concentration (αR.ci) (a) and compressibility coefficient (s) (b) according to MLSS concentration (Moreau 2010).

As Figure 3.4 shows, maximum specific cake resistance x fouling particles concentration and compressibility coefficient results were found for MBR activated sludge samples with a MLSS concentration of around 10 g/L (Moreau 2010).

3.4 Qualifying Membrane Bioreactor Activated Sludge

MBR activated sludge is usually understood as only a more concentrated activated sludge than that of Conventional Activated Sludge (CAS) systems. According to the author of this thesis, this is only partially true. The same influent, sewage, is being treated by both technologies, MBR and CAS processes. However, MLSS concentrations in MBRs are over 2 to 3 times higher than in CAS installations. Furthermore, in an MBR installation, particles are retained in the bioreactor, as opposed to CAS installations, where small particles are washed out through the clarifier. The complete retention of microorganisms, contrary to CAS systems, also indicates that microbial communities may differ between CAS and MBR systems. Differences in MLSS concentrations, particles and microorganisms retained in the bioreactors are likely to lead to activated sludge with different characteristics in CAS and MBR systems. At the beginning of this research work, a choice was made to study MBR activated sludge characteristics by applying the DFCm. In addition, the participation of TU Delft in European projects enabled an extensive European on site measuring campaign in full-scale and pilot-scale MBR installations. Several PhD and MSc students were involved in the aforementioned campaign and individual outcomes had to be provided. The choice of the author of this thesis was to study the relationship between filterability and the

41

suspended fraction of the sludge by applying the DFCm, coupled with particle counting in the range of 2-100 µm. The research questions motivating the aforementioned choice were the following:

- What is the activated sludge quality, namely filterability and floc size, in MBR

installations?

- Is filterability related to the suspended fraction of the sludge and, if so, how?

- Is filterability changing along the various tanks of one MBR and, if so, why and

how?

This thesis was developed at a Faculty of Civil Engineering and, consequently, the approach is eminently physical. Therefore, the parameters used to qualify MBR activated sludge during the European on site measuring campaign were filterability, measured through the DFCm, and particle counting in the range of 2-100 µm. MLSS concentration results were provided by the visited WWTP staff. In a lab environment, or when the MBR installation was within a travelling distance of 1 hour from the TU Delft water lab, the following parameters were also measured: MLSS concentration, SMP concentrations, viscosity and particle counting in the range of 0.4-5 µm. Further details on materials and methods are provided in Chapter 4. The physical approach of this thesis does not imply that the qualification of activated sludge is complete from a physical point of view. On the contrary, biochemical and microbiological information would be extremely valuable to fully qualify MBR activated sludge. The discussion of the measured physical parameters will obviously be mainly physical, but biochemical processes or microbiological characteristics will be mentioned whenever they are likely to have an effect on the obtained results.

3.5 Analyzing Activated Sludge Characteristics by Diluting and Concentrating

As mentioned earlier, at the beginning of this research work, a choice was made to study MBR activated sludge characteristics by applying the DFCm. Over time, it became clear that studying activated sludge characteristics would have to start by analyzing the role of MLSS concentration in fouling propensity. MLSS concentrations are decisive for determining the volume of the MBR tanks and are one of the most used parameters to control the process. Furthermore, in the current MBR installations, MLSS concentration is a dynamic parameter and its impact on membrane filtration is not completely understood. This awareness lead to the formulation of the main research question of this thesis:

- What is the role of MLSS concentration in MBR filtration?

The main research question gave rise to the following research questions:

- What is the optimal MLSS concentration or MLSS concentration range for MBR

filtration?

42

- What is the influence of indentified fouling particles, such as SMP and submicron

particles, on MBR filtration?

- What is the optimal layout, implying the possible existence of a separate

membrane tank, for an MBR installation?

To study the role of MLSS concentrations in MBR filtration, the obvious operation to manipulate MLSS concentration in the activated sludge is to either reduce it, by diluting, or increase it, by concentrating. Therefore, a choice was made to study activated sludge characteristics by diluting and concentrating activated sludge. Both operations would be performed with real wastewater sludge and the dilution samples would be produced with permeate, to preserve the MBR sludge characteristics. Figure 3.5 illustrates the concept of diluting and concentrating activated sludge in order to study the role of MLSS concentration in MBR filtration.

Figure 3.5 - Diluting and concentrating activated sludge to study the effect of MLSS concentration on MBR filtration.

The diluting experiments could easily be performed on site or in a lab environment because they only required samples of activated sludge and permeate from MBR installations. The advantage of performing these experiments in a lab environment is that a greater number of activated sludge parameters can be measured. The complete methodology of the diluting experiments is described in section 6.2. The diluting tests were commenced at an early stage of this research work, in 2007. The opposite of diluting is concentrating, but the concentrating experiments required a different approach from the diluting tests. In order to increase the MLSS concentration from 8 to 15 g/L, the DFCi takes 5 hours and the same sludge sample is submitted to several consecutive cycles. The DFCi is not suitable for long-term tests and the activated sludge characteristics, namely its structure, would be affected. Therefore, performing concentrating experiments in a lab environment using the DFCi was not an option. During the European on site measuring campaign, activated sludge samples were collected from all the tanks of the MBRs to verify if, why and how filterability changed

MLSS [g/L]

Poor quality

Moderate quality

∆R20 [x1012m-1]

Good quality

Diluting

Concentrating

43

within one MBR. Our results showed that, in some MBRs, a better filterability, coupled with a higher MLSS concentration, is found in the membrane tanks, when compared with the activated sludge filterability of carbonaceous and nutrient removal tanks in the same MBR installation. The results are presented and discussed in Chapter 5. These results provided a solution for the concentrating experiments explained below. In an MBR with heterogeneous filterability between tanks, the activated sludge quality would be monitored when the MLSS concentration in the membrane tanks increased. It was decided to perform these experiments in a full-scale MBR, so the obtained results could be used for full-scale design and optimization. The disadvantage was that concurrent phenomena were taking place, so the effect of each factor, influencing activated sludge quality, is hardly isolated. The concentrating tests were performed during 2009 at the Heenvliet MBR (Hollandse Delta Water Board, in the Netherlands), located at a travelling distance of less than 1 hour from the TU Delft water lab. In this thesis, the activated sludge samples with increased MLSS concentrations were collected from the bulk activated sludge. The activated sludge samples were then submitted to several measurements such as filterability and MLSS concentration. Given that these parameters are measured in the bulk MBR activated sludge, the increase in MLSS concentration is referred to herein as a concentrating operation.

3.6 Assessing Membrane Bioreactor Integrity through Permeate Quality

The performance of an MBR is usually perceived as either the result of what is occurring on or inside the membrane, or the permeate quality obtained after the filtration step. Permeate quality is usually assessed by determining physical-chemical parameters that enable removal rate calculations and, therefore, an assessment of what is occurring on or inside the membrane. In MBR technology, membranes are perceived as size-excluding barriers that will effectively retain all particles larger than its nominal pore size. In 2006/2007, the Heenvliet MBR staff detected filamentous bacteria in the MBR permeate and requested support from TU Delft to assess the situation. Details about this incident are provided in Chapter 8. To evaluate the extent of the problem, regular on site measurements, of particle counting in the range of 2-100 in the MBR permeate, were initiated. The results were totally unexpected. Curiosity, and the availability of the aforementioned equipment, motivated the continuation of the permeate measurements in all visited European MBR installations. Therefore, in this thesis, MBR membrane integrity was evaluated through the permeate quality by measuring the number of particles in the range of 2-100 µm. The results, discussion and conclusions of these measurements are provided in Chapter 8.

44

3.7 Conclusions

The conclusions obtained in Chapter 3 can be summarized as follows: - The DFCm was developed. The main characteristics of the method are:

o the DFCm can quantify filterability and discriminate between different MBRs;

o filterability can be manipulated and successfully measured; o the hydraulic regime, when MBR activated sludge is filtered, is laminar if

the MLSS concentration in the sample exceeds 5 g/L; o the DFCm does not exactly reproduce the filtration process of MBR

installations; o in the sludge filtration step of the DFCm, the dominant fouling is

removable; o there is a statistical correlation between filterability and permeability in

hollow fiber membranes due to their frequent cleaning strategy; o filterability has some correlations with SMP concentrations when the

activated sludge is manipulated in a lab environment, but not when full- or pilot-scale activated sludge is monitored;

o the volume of submicron particles is a better indicator of filterability than SMP concentrations;

o the DFCm is a practical oriented tool that can be applied on site or used to manipulate and control filterability in a lab environment.

- The DFCm was applied in several on site campaigns at MBRs scattered across Europe. The main outcomes of these campaigns, so far, were the following:

o MBR process control should aim at optimal flocculation in the membrane tanks;

o feed water quality is crucial for MBR operation; o there is a seasonal variation of filterability linked to temperature; o filterability in MBR pilots is more homogeneous and less extreme than in

full-scale MBRs; o membrane configurations do not have significant impact on the

filterability of activated sludge. - Furthermore, the application of the DFCm in on site campaigns provided two

results that are particularly relevant to this thesis, namely: o Filterability improved with increasing MLSS concentrations. o A larger cake layer and maximum compressibility in the DFCi membrane

were found at an MLSS concentration of around 10 g/L, when analyzing data obtained from 3 full-scale and 5 pilot-scale MBRs.

- The research work described in this thesis will focus on MBR activated sludge characteristics. The main research question will be:

o What is the role of MLSS concentration in MBR filtration?

- In this research, the following key questions will be addressed:

o What is the activated sludge quality, namely filterability and floc size, in

MBR installations?

45

o Is filterability related to the suspended fraction of the sludge and, if so,

how?

o Is filterability changing along the various tanks of one MBR and, if so,

why and how?

o What is the optimal MLSS concentration or MLSS concentration range for

MBR filtration?

o What is the effect of identified fouling particles, such as SMP and

submicron particles, on MBR filtration?

o What is the optimal layout, implying the possible existence of a separate

membrane tank, in an MBR installation?

- Owing to the presence of unusual particles in the MBR permeate of a full-scale MBR, a particular assessment of MBR membrane integrity was performed. Curiosity and opportunity led us to repeat it in all visited MBR installations. The results are presented in this thesis as additional information about MBR technology.

46

47

4. Materials and Methods

This chapter presents the materials and methods used in the experiments described in this thesis. Table IV-I shows the parameters measured in the different types of experiments described in this thesis. The types of experiments are identified according to chapter number and corresponding main subject.

Table IV-I Parameters measured in each type of experiment. 5- Monitoring

Activated Sludge

6- Diluting Activated

Sludge

7- Concentrating Activated

Sludge

8- Assessing Permeate Quality

Filterability Yes Yes*** Yes No MLSS No ** Yes Yes No SMP No Yes Yes No Viscosity No Yes Yes No P.C. * 2-100 µm

Yes Yes Yes Yes

P.C.* 0.4-5 µm

No Yes Yes No

Key: * P.C.: Particle counting **MLSS concentration values provided by the WWTP staff. *** Filterability was measured according to the DFCm and exceptionally using a flat sheet

membrane unit (FSu) (Atasoy 2008). The FSu is described in Annex B.

In the research described in this thesis, one activated sludge or permeate sample was used for the determination of only one parameter. This procedure was adopted to avoid interference of the measuring methods on the activated sludge or permeate characteristics.

4.1 Filterability

A filtration apparatus, the Delft Filtration Characterization installation (DFCi), and a measuring protocol, the Delft Filtration Characterization method (DFCm), were defined (Evenblij et al. 2005). The theory connected to this filtration procedure is based on Darcy’s law:

p t

TMPJ

Rη=

[L/m2.h](Equation 4.1)

Under a specific set of hydraulic conditions, Trans Membrane Pressure (TMP), permeate viscosity (ηp) and total resistance (Rt) define the flux (J). The TMP, ηp and Rt parameters are influenced by feed characteristics, operating conditions and membrane characteristics. When the operating conditions and membrane characteristics are constant, filtration behavior can be related exclusively to the feed water characteristics, i.e. the quality of the MBR activated sludge (Evenblij 2006).

48

The Delft Filtration Characterization Installation (DFCi)

The layout of the DFCi is shown in Figure 4.1. The main components of the DFCi are: a single-tube membrane; the equipment for activated sludge recirculation and permeate extraction; the data acquisition system and the equipment for membrane cleaning.

Mass balance

pHO2 T

P3

P2

P1 F

NaOCl

Demi-water

Activated sludge

Back flush pump

Permeate

UF

Me

mb

ran

e

Sewer

Sewer

Air

Centrifugal pump

Peristaltic pump

Damper

Computer

Peristaltic pump

Figure 4.1 - Layout of the Delft Filtration Characterization Installation (DFCi).

The membrane of the DFCi is a single-tube X-Flow membrane with an internal diameter of 8 mm and a nominal pore size of 0.03 µm, thus providing a UF operation. The membrane is 95 cm long, corresponding to a total membrane area of 0.024 m2. The membrane material is polyvinylidene difluoride (PVDF) and permeate is extracted from inside to outside the membrane. The activated sludge is driven along the membrane through a Watson-Marlow 700 series peristaltic displacement pump in a continuous loop. The pulse of the pump is flattened by a damper. The pump speed is adjustable and the flow is measured through a Krohne IFC 010K electromagnetic flow meter. While the activated sludge is circulating, permeate is extracted with a Watson-Marlow 520 series peristaltic pump by creating an underpressure on the permeate side of the membrane. Pump speed is adjustable and permeate mass production is measured by a Mettler Toledo mass scale. During activated sludge filtration, DO concentration, pH and temperature are continuously measured in the activated sludge vessel. Membrane resistance is measured through pressure sensors placed on the feed, concentrate and permeate. The sensor readings and the permeate mass production are sent to a computer every 12 seconds. The computer is equipped with a TestPoint program capable of calculating, displaying and storing the relevant filtration parameters. Each activated sludge filtration test is followed by a cleaning procedure. Cleaning can be performed using centrifugal pumps to either forward flush or back flush water through

49

the membrane. Furthermore, a small peristaltic pump is used to soak the feed side with cleaning chemicals. DFCm measuring protocol

The measuring protocol defined by Evenblij et al. (2005) consists of three basic steps: membrane resistance determination, activated sludge filtration and membrane cleaning. These steps are described as follows:

1. Membrane resistance determination: a. Clean water circulation (at a cross-flow velocity of 1 m/s); b. Permeate extraction (at a flux of 80 L/m2.h); c. Determination of filtration resistance (0.4x1012 ±0.1x1012m-1); d. Chemical cleaning procedure when membrane resistance exceeds

0.5x1012m-1. 2. Activated sludge filtration:

a. Activated sludge recirculation (at a cross-flow velocity of 1 m/s); b. Permeate extraction (at a flux of 80 L/m2.h) until a permeate production of

25 L/m2 or TMP value of 0.6 bar is reached; c. Redefinition of the flux in case of extremely bad or good filterability.

3. Membrane cleaning: a. Forward flush with demineralized water (at a cross-flow velocity higher

than 5 m/s) b. Chemical cleaning by soaking the membrane with NaOCl (at 1500 ppm)

for at least 15 m. DFCi output and processing of results

The aforementioned membrane resistance determination step is performed to assess the actual state of the membrane. If filtration resistance is between 0.3 to 0.5x1012m-1, no further action is required. If filtration resistance exceeds 0.5x1012m-1, then chemical cleaning must be performed to restore the membrane filtration characteristics. As mentioned earlier, the membrane resistance determination step is intended to evaluate the membrane filtration state, thus the obtained results are no longer used. The main step of the measuring protocol is the activated sludge filtration step. As mentioned earlier, several activated sludge parameters, namely DO concentration, pH and temperature, and process parameters, namely TMP, flux and cross-flow velocity, are continuously monitored during the activated sludge filtration step. In the output, these parameters are plotted against the specific permeate production volume [L/m2]. The resulting figures are used to control the development of the activated sludge filtration step online. Furthermore, total resistance (Rt) is calculated according to Darcy’s law (Equation 4.1). The calculation of Rt is preceded by a flux and permeate viscosity temperature correction. It is assumed that Rt is the result of membrane resistance (Rm) plus the resistance imposed by the cake layers built up on the membrane during sludge filtration, referred to as fouling resistance or added resistance (∆R). In the calculation of ∆R , Rm is assumed as the initial value of resistance, i.e. the first obtained value of ∆R in the activated sludge

50

filtration step. The main output of the DFCi consists of a graph that plots added resistance (∆R), caused by cake layer filtration, according to specific permeate production (Vs), as exemplified by Figure 4.2.

y = 0,0058x1,2813

R2 = 0,9353

0,01

0,1

1

10

1 10 100Vs [L m

-2]

∆R

[*1

01

2m

-1]

Figure 4.2 - Added resistance according to permeate volume production per m2 of

membrane. The calculated values of added resistance are then used to establish a fouling tendency through a power law equation, according to Geilvoet (2010). An example is also shown in Figure 4.2. The obtained mathematical expression is then used to calculate the ∆R20 parameter, αR*ci product and s coefficient. The latter results are obtained by fitting the cake layer filtration theory to the DFCm output, as presented by Geilvoet (2010) and shown in Figure 4.3.

b

sR aV∆ =

DFCm output

Cake layer filtration theory

( )11

11 ssR i sR c Vα −−∆ = ⋅ ⋅

=>

20 .20bR a∆ =

1 1

1

bb s

s b

−= <=> =

−

( )1

11 . ss

R i R ic a c aα α −−⋅ = <=> =

Key: ∆R- Total cake layer resistance [m-1]; a,b- absolute values obtained in each filtration experiment; Vs- Specific permeate production [L.m-2]; αR - Specific cake resistance [m.kg-1]; ci - Concentration of fouling particles [kg.m-3]; s - Compressibility coefficient [-]; ∆R20- Total cake layer resistance obtained after 20 L of permeate per m2 membrane area are produced [m-1];

Figure 4.3- Processing of DFCm output.

51

A classification linking the assessed ∆R20 and MBR activated sludge filterability was defined by Geilvoet (2010) and is shown in Table IV-II.

Table IV-II ∆R20 values and corresponding MBR activated sludge filterability (for standard DFCm measuring protocol).

∆R20[×10 12m-1]

Good < 0.1

Moderate 0.1-1

Poor >1

The accuracy of the DFCm is in the range of 0.05, therefore, the correlation factor represented in Figure 4.2 by R2 is lower for samples with good filterability, i.e. with ∆R20 below 0.1. Consequently, according to Geilvoet (2010), the compressibility coefficient is only reliable when the samples have moderate to poor filtration quality. In activated sludge with good filtration quality and a compressibility factor of 0, the resistance increases linearly with the specific permeate production (Geilvoet 2010). Collection of activated sludge samples

The activated sludge samples were preferably collected in the upper decks of the MBR installations, from central areas of the MBR tanks. Concerning in particular membrane tanks, the samples were collected from the activated sludge layers on top of the submerged membranes.

4.2 Mixed Liquor Suspended Solids

MLSS concentration was determined according to the NEN6621 standard, as follows: a 5892 Schleicher & Schuell paper filter, with a diameter of 70 mm, is wetted with demineralized water, dried and weighed. A known volume of activated sludge, usually 50 mL, is filtered through the filter in dead-end mode. The filter is subsequently dried in an oven at 105ºC, for at least three hours, and then weighed. MLSS concentration is calculated as the difference between the two aforementioned weighting values, divided by the volume of activated sludge, placed on the paper filter.

4.3 Soluble Microbial Products

In this research, SMP concentrations, namely protein and polysaccharide concentrations, were measured in the MBR activated sludge and permeate of the DFCi. The SMP concentrations potentially retained in the DFCi membrane are obtained by calculating the difference between MBR activated sludge and permeate concentrations. The activated sludge samples are prepared for SMP determination by filtration through a 5892 Schleicher & Schuell paper filter, with a pore size between 7 and 12 µm. The

52

resulting water phase is used to determine protein and polysaccharide concentrations according to the methods described below. The permeate samples are not submitted to filtration. The permeate samples are used to determine protein and polysaccharide concentration according to the following methods. Protein concentration is determined according to the method defined by te Poele (2005). This method is based on the colorimetric method proposed by Lowry et al. (1951). Bovine Serum Albumin (BSA), in a concentration between 5 and 50 mg/L is used for calibration. Protein concentration is obtained by measuring absorption at 750 nm, in a 4 cm glass cuvette. Polysaccharide concentration is determined according to the method defined by te Poele (2005). This method is based on the colorimetric method proposed by Dubois et al. (1956). Glucose (J.T. Baker), in a concentration between 2.5 and 25 mg/L, is used for calibration. Polysaccharide concentration is obtained by measuring absorption at 487 nm, in a 4 cm glass cuvette.

4.4 Viscosity

The apparent viscosity of MBR activated sludge is measured directly in an Anton Paar type rotational rheometer. The rheometer is coupled to a hardware-driven unit, an air pressure system, a water bath controller and a computer. The air pressure system provides dry pressurized air for the air bearing. The water bath ensures that the activated sludge samples are always measured at the constant temperature of 20ºC. The computer is equipped with US200 software that allows controlling the rheological measurements. The measurements are performed as follows: a MBR activated sludge sample of 100 mL is placed on the rheometer. An aluminum cylinder, original concentric geometry Z2, is submerged in the activated sludge sample and rotated at several rotation speeds that create fixed shear rates, between 5-1000 s-1. The rheometer measures the torque, from which the shear stress is calculated. The apparent viscosity of the sample is calculated by dividing the applied shear rate by the resulting shear stress.

4.5 Particle Counting in the Range of 2-100 µm

In this research, grab samples of MBR activated sludge samples, CAS effluent, permeate, process water and demineralized water are submitted to particle counting in the range of 2-100 µm. The MBR activated sludge samples are submitted to the sample preparation procedure described below. The CAS effluent, permeate, process water and demineralized water are measured directly after collection, without dilution.

53

Sample preparation

The MBR sludge samples are diluted with permeate by a factor of 100 and sieved, with a 100 µm sieve, before their measurement. The MBRs studied in this research were designed for Mixed Liquor Suspended Solids (MLSS) concentrations between 8 and 18 g/L in the membrane tanks. The number of particles in the membrane tank samples exceeds 16,000 particles per mL, revealing the need to dilute the samples prior to measurement. Figure 4.4 a) presents the total number of counted particles in one sample submitted to different dilution factors. In the remainder of our research, the factor 100 was selected as the optimal dilution factor, because it results in the highest number of counted particles (Figure 4.4 a). Figure 4.4 b) presents the number of particles in one sample, prepared with sieves of different size meshes. Figure 4.4 b) shows that the number of particles between 2 and 15 µm increases when the sample is filtered using sieves with a small mesh size. Apparently, sieving samples with mesh sizes less than 100 µm disrupts the particles and increases the number of smaller particles. However, for practical purposes, i.e. to avoid equipment contamination, particle counting samples should only contain particles smaller than 100 µm. Therefore, sieving was adopted in sample preparation, using the largest possible mesh size of 100 µm.

11000

11200

11400

11600

11800

12000

12200

12400

1-50 1-75 1-100 1-125

Cu

mu

lative n

um

be

r o

f p

art

icle

s

Dilution factor

0

10000

20000

30000

40000

50000

60000

0 10 20 30 40

Nu

mb

er o

f p

art

icle

s/m

L

Particles size [µm]

Sieve 32 micron meter

Sieve 63 micron meter

Sieve 100 micron meter

a) b) Figure 4.4 - Total number of counted particles (a) and number of particles/mL per particle

size (b). Equipment and methods

The samples are measured in a Met One particle counter, in the size range of 2-100 µm. Particles are counted in fixed-size increments of 0.5 µm. The particle counter is equipped with an LB1020 sensor. The operating principle of this sensor is the light-blockage method, with a coincidence loss lower than 10% at 16,000/mL. In light-blockage particle counters, a focused beam of light shines from one side of the measuring zone to a photovoltaic cell on the other side (APHA 2000). The particles pass

54

though the measuring zone at a standard flow of 100 mL/min. Light blockage, caused by a particle passing through the measuring zone, brings about a change in the voltage of the photovoltaic cell. The signal is used to determine the size of the particle. The schematic layout of the particle counting equipment, in the range of 2-100 µm, used in this research is presented in Figure 4.5.

Particle counter

Pump

Flow Meter

Sample Outlet

P C

Sample Inlet

Figure 4.5 - Schematic layout of the particle counting equipment, in the range of 2-100

µm. The particle counter is coupled to a computer equipped with WGS Software. The initial output consists of the number of particles/mL, between 2 and 100 µm, in the aforementioned size increments. The particle counter is cleaned with demineralized water, flowing through the equipment for at least 10 minutes, after each day of measurements. On each day of measurements, the drift of the particle counter is controlled by measuring the total number of particles, in demineralized water samples, before and after the measurements are performed. The particle counter was calibrated in April 2008 using spherical particles manufactured for that purpose, namely monodispersed polymer microspheres. In the zero count test, 2 particles/mL were counted, out of an allowable maximum of 5 particles/mL. Processing of results

In this research, the particle counting data are mainly presented by the particle number distribution, i.e. number of particles/mL per particle size and normalized particle counting data. The normalization of particle counting data consists of dividing the concentration of particles, in a given size rage, by the size interval and presenting the particle size in logarithmic scale. This calculation prevents the creation of apparent peaks or valleys in a particle size distribution (APHA 2000). An example of the particle number distribution and normalized data is presented in Figure 4.6 for activated sludge and demineralized water samples.

55

0

5

10

15

20

25

0 20 40 60 80 100

Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[#

*103

/mL

]

Demineralized water

Activated sludge

-2

0

2

4

6

1 1,2 1,4 1,6 1,8 2

Log (Particle size) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

10 100

Particle size[µm]

De-mineralized water

Activated sludge

Lower end

[------------------------------------]Upper end

[----------------------]

a) b) Figure 4.6 - Number of particles/mL per particle size (a) and normalized particle counting

data (b) from activated sludge and demineralized water samples. Given that particle counting generates significant amounts of data, a reduction technique must be used to effectively compare different samples. Two reduction techniques are usually adopted: the representation of data in terms of total count and the use of power-law coefficients. The first technique does not provide information about how the particles are distributed over the different size ranges and the second technique assumes that the particle distribution follows a power-law function. In other words, the latter technique assumes that a particle size distribution has a constant decrease in particles with increasing particle size, graphically represented by a constant slope, also referred to as β. A particle size distribution can be presented as a power-law function as follows:

log ( ) log log( )p pn d A dβ= − (Equation 4.2)

Where: dp - particle diameter; ( )pn d - derivative, at any point, of the cumulative number distribution;

β, log A - coefficients of the power-law function; β - slope of the power-law function.

The coefficient β can either be constant or variable. Lawler (1997) mathematically defined particle counting distributions and stated that a variable β, if apparent, is less than 1 at the lower end of the size range (Figure 4.6 b)) and higher than 4 at the upper end of the size range (Figure 4.6 b)). The demineralized water sample shown in Figure 4.6 b) is an example of a particle size distribution where a variable β is not apparent, so β is constant. On the other hand, the activated sludge sample shown in Figure 4.6 b) is an example where a variable β is apparent. In this research, the reliability of the obtained particle counting data is verified by comparing them with the theoretical particle size distributions defined by Lawler (1997), shown in Figure 2.8. It was observed that for undiluted samples, such as CAS effluent, permeate, process water and demineralized water, particles in the range of 2-3.5 µm were not successfully counted by our equipment. Therefore, for undiluted samples, the reliable

56

range of particle counting results in the range of 2-100µm was defined as 3.5-100 µm. For diluted samples, such as activated sludge, the particles between 2 and 10 µm were not successfully counted. Thus, the reliable range for activated sludge particle counting results in the range of 2-100 µm was defined as 10-100 µm. In parts of this research, the particle counting data, in the range of 10-100 µm, are used to calculate the average floc size, as follows: the shape of the particles is assumed to be spherical, the volume of one spherical particle calculated and multiplied by the corresponding number of particles/mL, leading to a volume distribution, shown in Figure 4.7. The average floc size of one sample was defined as the particle size where the maximum volume of particles was obtained, circled in Figure 4.7.

Figure 4.7 - Volume (1 ppb= 1000 µm3/mL) per particle size from a sludge sample.

In parts of this research, the number and volume distributions are used to calculate the total number and total volume, defined as the integral, i.e. the area below the number and volume distributions. Figure 4.8 shows examples of the total number and total volume for an activated sludge sample, as defined in this research.

0

5

10

15

20

25

10 20 30 40 50 60 70 80 90 100

Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[#

*103

/mL

]

Total number

0

50

100

150

200

10 20 30 40 50 60 70 80 90 100

Particle size [µm]

Vo

lum

e o

f P

art

icle

s [

*10

3p

pb

]

Total volume

a) b) Figure 4.8 – Total number (a) and total volume (b) from a sludge sample.

0

50

100

150

200

0 20 40 60 80 100

Particle size [µm]

Volu

me o

f P

artic

les [*1

03ppb]

57

4.6 Particle Counting in the Range of 0.4-5 µm

Sample preparation

The MBR activated sludge samples submitted to particle counting in the range of 0.4-5 µm are grab samples. The sludge samples are previously filtered and diluted, before being submitted to particle counting measurements in the range of 0.4-5 µm, in accordance with the procedure described below. MBR activated sludge is filtered through a 5892 Schleicher & Schuell paper filter, with pore size between 7 and 12 µm. The filtration step is required to avoid blockage of the particle counter measuring channels, therefore, to increase the reliability of the obtained results and protect the equipment. The free water, obtained through the aforementioned filtration step, is then diluted, by a factor of 100, with demineralized water, immediately before being submitted to particle counting measurements. The dilution step is required to avoid concentration errors, i.e. when more than one particle passes through measuring channels at the same time (Hargesheimer et al. 1998). The preparation of the samples, namely filtration and dilution, will change the characteristics of the samples, so the obtained conclusions, concerning absolute numbers, should be cautious. However, comparisons between particle counting results can provide valuable information because all samples were subjected to the same operations and, therefore, changed in the same way. Particle counter, in the range of 0.4-5 µm

The particle counter, in the range of 0.4-5 µm, used in this research is a HIAC ChemShield equipped with a MicroCount 100S sensor in a sapphire cell. The operating principle of the sensor is light scattering, with a coincidence loss lower than 10% at the concentration limit of 100,000 particles/mL. The abovementioned particle counter is a flow instrument, where the direct path of the light beam is blocked by a particle when it flows through the measuring zone, inside the fluid. The light scattered over a fixed range of angles is collected and measured. Particle size is determined from the scattering angle and intensity according to the principles of Fraunhofer and/or Mie (APHA 2000). The particles are counted in size increments of: 0.1 µm in the range of 0.4-1.5 µm, 0.5 µm in the range of 1.5-2 µm, and 1 µm in the range of 2.0-5.0 µm. The flow of the particle counter is fixed at 100 mL/min and is set in the particle counter itself. The schematic layout of the particle counting equipment, in the range of 0.4-5µm, used in this research is presented in Figure 4.9.

58

P C

Sample Outlet Sample Inlet

Sample

Pump

Particle counter

Flow Meter

Demineralized Lab network

Figure 4.9 – Schematic layout of the particle counting equipment, in the range of 0.4-5

µm. The flow through the particle counter is an extension of the demineralized water network of the TU Delft water lab. The free water of the MBR activated sludge sample is fed into the demineralized water network by a peristaltic pump, set to a flow of 1 mL/min. It is assumed that the potential effect of demineralized water on the particles, as a result of osmotic pressure differences, is negligible due to short contact time. After flowing through the particle counter, the sample is discharged to the sewer. The particle counter is connected to a computer equipped with Particle Vision Online software. The initial output consists of the number of particles/mL, between 0.4-5 µm, in the aforementioned size increments. The particle counter is cleaned with demineralized water, flowing through the equipment for at least 30 minutes, after each day of measurements. On each day of measurements, the drift of the particle counter is controlled by measuring the total number of particles, in demineralized water samples, before and after the measurements are performed. The particle counter was calibrated in April 2009 using spherical particles manufactured for that purpose. The state of particle counter, before calibration, was evaluated as reliable. Processing of results

The processing of the particle counting results, in the range of 0.4-5 µm, followed a procedure similar to that described for particle counting in the range of 2-100 µm. The processing of the particle counting results followed the indications provided by APHA (2000) for graphical representation, namely: use of absolute instead of relative numbers; data normalization by dividing the concentration of particles in a given size range by the size interval and presenting the particle size in logarithmic scale. The particle counting results in the range of 0.4-5 µm are presented by the particle number distribution, i.e. number of particles/mL per particle size and normalized particle counting data.

59

The β value, i.e. the slope of normalized particle counting data, is analyzed and compared with the theoretical particle size distributions defined by Lawler (1997). It was observed that particles in the range of 0.4-0.5 µm were not successfully counted by our equipment. Furthermore, it was found that the total number of particles in the range of 1-5 µm constitutes 1% of the total number of particles in the range of 0.4-5 µm. Therefore, in this research, the reliable and significant range for the particle counting results in the range of 0.4-5 µm it was defined as 0.5-1 µm.

60

61

5. Activated Sludge Quality in European Membrane Bioreactors - On site Measurements

5.1. Introduction

The main aim of this thesis is to clarify the relationship between the suspended fraction of the activated sludge and fouling propensity. This chapter describes the experiments intended to characterize the MBR activated sludge in terms of fouling propensity and its relationship with the suspended fraction of the sludge. The limitations of lab- and pilot-scale MBR experiments, for full-scale MBR design and operation, were described by Kraume et al. (2009). According to these authors, a greater number of publications on full-scale results and their comparison with lab-scale results are highly desirable. In the research described in this chapter, the focus is on the results obtained in full-scale MBRs. Nevertheless, MBR pilot-scale results are also presented and discussed. As mentioned in Chapter 2, a link was identified between filterability and submicron particles (Leiknes et al. 2006; Geilvoet et al. 2007a). Worse filterability is linked to a higher number of particles in the submicron range. In addition, several authors report that MBR activated sludge has a bimodal distribution, with one peak in the colloidal range and another peak in the suspended particle range (Wisniewski et al. 1998; Leiknes et al. 2006; Masse et al. 2006; Sombatsompop et al. 2006). Geilvoet et al. (2007a) showed that filterability deterioration is accompanied by release of submicron particles. These submicron particles can derive from microbiological activity, reacting to the imposed stressful conditions, and/or from deflocculation of larger particles. If the release of submicron particles is caused by deflocculation of micron particles, then this may be reflected in the particle size distribution of the total sludge suspended fraction. Particle counting of suspended particles should be performed on site in order to preserve the sample characteristics. Therefore, a portable device is required. The D3-Deliverable from the EUROMBRA project (UNITN 2006) advises MBR activated sludge characterization in the range of 0.04 to 2000 µm through a laser diffraction method. However, equipment with these characteristics is not portable. Consequently, the choice made was to count particles in the range of 2-100 µm through a light-blockage method. The downside of this particle counting method is that light-blockage particle counters are not accurate to measure flocculated samples (Hargesheimer et al. 1998). The development of the DFCm (Evenblij et al. 2005) and the participation in European projects, namely EUROMBRA and MBR-Train, enabled the launch of a European on site measurement campaign. Therefore, in the described research, we studied the possible correlation between suspended particles and MBR sludge filterability. All measurements were performed on site. The DFCi and a particle counter were operated in 4 full-scale MBR Wastewater Treatment Plants (WWTP) and 4 pilot-scale MBRs.

62

5.2 Methodology

The DFCi and a particle counter, in the measuring range of 2 to 100 µm, were placed in the following full-scale MBRs: Schilde, Monheim, Nordkanal and Heenvliet. The on site campaigns were performed during at least two full weeks: one in summer and one in winter. The collected activated sludge samples were submitted directly to filterability and particle counting measurements. Table V-I presents the main characteristics of the WWTPs.

Table V-I Characteristics of full-scale MBRs. WWTP Schilde

Belgium Monheim

Germany Nordkanal

Germany Heenvliet

Netherlands Population Equivalent

- 28,000 9,700 80,000 13,000

Wastewater treatment

- CAS+MBR MBR MBR CAS+MBR

Location of the membranes

- Submerged in separate tanks

Submerged in separate tanks

Submerged in nitrification

tanks

Submerged in separate tanks

Membrane Supplier

- Zenon (Hollow Fiber)

Zenon (Hollow Fiber)

Zenon (Hollow Fiber)

Toray (Flat sheet)

Membrane pore size

µm 0.04 0.04 0.04 0.08

Total membrane area

m2 10,560 12,320 84,480 4,110

Flux (design) L/m2.h 33.6 23.4 27 24.3

Flux (average operation)

L/m2.h 33 8.1 27 18.2

MLSS g/L 11 11 12 12

Return ratio * - 5.8 10.7 to 53.3 4 ** 1.5

Sludge age d 21 30 27 20

Cleaning - Mechanical and chemical (once a week)

Mechanical and chemical (twice a year)

Mechanical and chemical (once a week)

Mechanical and chemical (twice a year)

Key: * Return ratio from membrane tanks to carbonaceous or nutrient removal tanks; ** At the Norkanal MBR, the return line is between the nitrification/membrane tank and de-nitrification tank.

The DFCi and particle counter, in the measuring range of 2 to 100 µm, were also placed in the following 4 pilot-scale MBRs: TUB/KWB (Berlin), NTNU (Trondheim), UNITN (Trento) and EAWAG (Dübendorf). At these locations, the on site campaign was performed during one week. As in the full-scale campaign, the collected activated sludge samples were submitted directly to filterability and particle counting measurements. Table V-II presents the main characteristics of the MBR pilot plants. All the MBR pilots investigated in this research use real wastewater and are connected to real sewer systems.

63

Table V-II Characteristics of pilot-scale MBRs. WWTP TUB/KWB

Germany NTNU

Norway UNITN

Italy EAWAG

Switzerland Population Equivalent

- 250 8 200 100

Wastewater treatment

- MBR **BF-MBR MBR MBR

Location of the membranes

- Submerged in separate tanks

Submerged in separate

tanks

Submerged in separate

tanks

Submerged in separate tanks

Membrane Supplier

- A3 (Flat Sheet)

Zenon (Hollow Fiber)

Zenon (Hollow Fiber) Eidos

Zenon, Puron (Hollow Fiber),

Kubota (Flat Sheet)

Membrane pore size

µm 0.2 0.04 0.04;0.1 0.04;0.04;0.4

Total membrane area

m2 75 1.86 100 46;30;40

Flux (design) L/m2.h 16-19 35 10.5-15 10;12;16

Flux (average operation)

m3/h 0.33-0.54 0.031 1.5-2.2 1.4-2

MLSS g/L 14.5-16.4 0.14 9.6-10.2 8.98-12.5; 8.2-11; 6.1-13.6

Return ratio *

- 1.7-2.7 - 5.9-11.1 2

Sludge age d 25 - 23.8-25.6 13

Cleaning Mechanical and chemical

(Once a month)

Mechanical and chemical

Mechanical and

chemical

Mechanical and chemical (Twice a

month)

Key: * return ratio from membrane tanks to carbonaceous or nutrient removal tanks; ** BioFilm reactor combined with membrane separation (BF-MBR).

The materials and methods required for the filterability and particle counting measurements, in the range of 2-100 µm, are described in Chapter 4. The research work presented in this Chapter was developed simultaneously with the research work of Adrien Moreau, described in his thesis (Moreau 2010). This author presented filterability results, measured according to the DFCm, from the pilot- and full-scale MBR installations referred to in this Chapter, with the exception of the Heenvliet MBR, collected during two full weeks. Given that the results presented by Moreau (2010) constitute a larger set of filterability results than those presented in this Chapter and the method applied to measure filterability was the same, the results and conclusions obtained by the abovementioned author will be referred to in section 5.3.

64

5.3 Filterability Results and Discussion

Full-scale

The filterability results provided by samples collected from the membrane tanks of the full-scale plants are presented in Figure 5.1. Average values are represented by columns and standard deviation by error bars in each column. Results are identified by location, month and year of collection. The boxes on top of each column contain the number of samples collected in each campaign.

0

1

2

3

4

5

6

Schilde

(2/2007)

Schilde

(4/2008)

Schilde

(8/2008)

Monheim

(3/2007)

Monheim

(9/2008)

Nordkanal

(11/2008)

Nordkanal

(7/2007)

Heenvliet

(3/2008)

Heenvliet

(6/2008)

∆R

20[x

10

12m

-1]

5

5

9 12 14 16 15 4 2

Figure 5.1- Filterability of the activated sludge from full-scale MBR membrane tanks.

With one exception, for all campaigns, the average values of ∆R20 ranged between 0.01 and 1 x1012m-1 of added resistance, i.e. the resistance built up on the membrane during the sludge filtration step, excluding the resistance imposed by the membrane itself. Following the classification proposed by Geilvoet (2010), MBR activated sludge filterability varied between good and moderate. This quality range is in line with previously presented results (Geilvoet et al. 2007b). In Figure 5.1, it is possible to observe that there are differences in filterability between winter and summer campaigns. During winter campaigns, the measured filterability is moderate to poor, whereas in summer, it can be classified as good. Also Wedi (2006) reported a decrease in membrane permeability during winter. Table V-III presents the temperature of the activated sludge in each campaign, identified by location, month and year of collection.

65

Table V-III Temperature of the activated sludge from full-scale MBRs. Temperature [ºC]

Minimum Maximum

Schilde (2/2007) 10.5 10.8

Schilde (4/2008) 11.9 12.5

Schilde (8/2008) 16.5 17.2

Monheim (3/2007) 8.4 9.2

Monheim (9/2008) 17.5 18.3

Nordkanal (11/2008) 9.5 12.7

Nordkanal (7/2007) 17.8 19.5

Heenvliet (3/2008) 9.4 10.4

Heenvliet (6/2008) 17.8 20.3 Table V-III shows that, in our research, the activated sludge temperature during winter campaigns was aproximatly 10ºC lower than in summer. Temperature affects biochemical and physical aspects (Metcalf & Eddy 2003). Miyoshi et al. (2009) studied the seasonal variation of fouling and measured an accumulation of dissolved organic matter, during winter time, in the mixed liquor suspension, showing that conversion rates were lower in winter than in summer. In the full-scale MBRs studied in this research, biomass is suspended and assembled in flocs. Considering a decrease in microbial growth and conversion rates during the winter period, a temperature impact on the flocculation processes and resulting flocs can also be expected. In addition, a decrease in temperature will increase the sludge viscosity, decreasing aeration efficiency. Submerged membranes rely on aeration to remove clogging. If clogging increases due to inefficient aeration, total membrane permeability will consequently decrease. Higher viscosity and poor flocculation, as a consequence of lower temperature, may be the direct factors responsible for lower filterability during winter periods. Figure 5.1 shows that the activated sludge filterability at the Schilde MBR during the April 2008 campaign was poor. In this period, a severe decrease in permeability was measured by Schilde WWTP staff. The results from this particular campaign show that the DFCm is capable of qualifying MBR activated sludge filterability when filtration performance is extremely poor. Pilot-scale

Figure 5.2 presents the filterability results provided by samples collected from the membrane tanks of the pilot-scale MBRs. The average values are represented by columns and standard deviation by error bars in each column. The results are identified by location, month and year of collection. The boxes on top of each column contain the number of samples collected in each campaign.

66

0

1

2

3

4

5

6

TUB/KWB

(6/2007)

NTNU

(10/2007)

UNITN

(11/2007)

EAWAG

(11/2007)

∆R

20[x

10

12m

-1]

5 22 8 36

Figure 5.2- Filterability of the activated sludge from the membrane tanks of pilot-scale

MBRs. For all the on site campaigns represented in Figure 5.2, the ∆R20 average results are approximately between 0.03 and 1x1012m-1 of added resistance. According to the classification proposed by Geilvoet (2010), activated sludge filterability varied between good and moderate. Table V-IV presents the temperature of the activated sludge in each pilot-scale campaign, identified by location, month and year of collection.

Table V-IV Temperature of the activated sludge from pilot-scale MBRs. Temperature [ºC]

Maximum Minimum

TUB/KWB (6/2007) 20.8 24

NTNU (10/2007) 12.1 15.3

UNITN (11/2007) 12 13.4

EAWAG (11/2007) 15.6 16.1 As Table I-IV shows the campaigns at the MBR pilot-scale installations were not carried out both during summer and winter periods. Therefore, no further conclusions regarding seasonal variations of filterability in pilot MBRs can be drawn. Filterability results of pilot- and full-scale MBRs collected during summer and winter periods, were presented by Moreau (2010). The author concluded that filterability in MBR pilots is more homogenous and less extreme than in full-scale MBRs. Kraume et al. (2009) described the differences between pilot- and full-scale MBR plants, mentioning that feed and temperature vary much more in full-scale than in pilot-scale plants. The smaller variation in the inflow and temperature may contribute to explain why filterability in pilot plants is more homogenous and less extreme than in full-scale installations.

67

Full-scale and pilot-scale

Moreau (2010) reported αR.ci and s coefficient values obtained in all the locations analyzed in this research, with the exception of the Heenvliet MBR. As above mentioned, in the latter research work, all on site campaigns were conducted for at least two full weeks, one during summer and one during winter. Moreau (2010) showed a significant correlation, i.e. with a Pearson coefficient of 0.97, between the ∆R20 parameter and the αR.ci product. The compressibility factor s was not statistically correlated to the ∆R20 results. According to Moreau (2010), sludge with moderate to good filterability and a ∆R20 below 0.5*1012m-1, tended to have an s coefficient of zero, while sludge with moderate to poor filterability and ∆R20 above 0.5*1012m-1, tended to have an s coefficient above zero, thus forming a compressible cake layer.

5.4 Return ratio Effect on Membrane Tank Filterability

Filterability results assessed with samples collected from all tanks of the full-scale MBRs are shown in Figures 5.3 and 5.4. Each MBR has its own design, generally comprising several tanks where carbon and nutrient removal takes place, followed by membrane filtration. The tanks where the membranes are located are indicated as membrane tanks. The functionality of the other MBR tanks is designated according to the amount of oxygen provided and the presence of ammonia. Figures 5.3 and 5.4 show two typical results found during the campaigns: either a homogenous filterability in all tanks (Figure 5.3) or a heterogeneous filterability between tanks with the best results in the membrane tanks (Figure 5.4).

0,00

0,05

0,10

0,15

0,20

Anoxic

Area

Aerobic

Area

Membrane

Tank 1

Membrane

Tank 2

Membrane

Tank 3

Membrane

Tank 4

∆R

20 [

*10

12m

-1]

Anoxic Aerobic

Area Area

Membrane Tank 1

Membrane Tank 3

Membrane Tank 2

Membrane Tank 4

a) b) Figure 5.3 - Homogenous filterability in all MBR tanks (a) and simplified scheme of the corresponding MBR, Monheim MBR (b).

68

0,00

0,05

0,10

0,15

0,20

Anaerobic

Tank

Anoxic

Tank

Aerobic

Tank

Aeration

Tank

Membrane

Tank 2

∆R

20 [

*10

12m

-1]

Anaerobic tank

Anoxic tank

Membrane Tank 1

Aerobic tank

Aeration tank

Membrane Tank 2

a) b) Figure 5.4 - Heterogeneous filterability in all MBR tanks (a) and simplified scheme of the corresponding MBR, Heenvliet MBR (b).

Homogenous and heterogeneous filterability between MBR tanks can be explained by the return ratios. If the return rate is high, a homogenous filterability is observed in all tanks of the MBR. A heterogeneous filterability occurs when the return rate is low. In the latter situation, the lowest ∆R20 values are found in the membrane tanks (Figure 5.4). The increase in activated sludge MLSS concentration prevailing in the membrane tanks due to the extraction of permeate and low return ratio, may enable the activated sludge to retain fouling particles. Interestingly, Wisniewski et al. (1998) observed a decrease in permeate flux in activated sludge MBRs operated with high return ratios. Apparently, in accordance with our current results, full-scale MBR activated sludge filterability can be improved by applying low return ratios.

5.5 Particle Counting Results in the Range of 2-100 µm and Discussion

Full-scale

The particle counting results from samples collected from the membrane tanks of the four full-scale MBRs are shown in Figure 5.5. The results are identified by location, month and year of collection and number of samples measured in each campaign. The presented results are average values for each campaign.

69

0

0,2

0,4

0,6

0,8

1

0 20 40 60 80 100

Particle size [µm]

Rel.

cum

ul.

num

ber [#

/mL]

Schilde 4/2008 (4 samples)

Schilde 8/2008 (6 samples)

Monheim 9/2008 (12 samples)

Nordkanal 7/2007 (5 samples)

Heenvliet 3/2008 (8 samples)

Heenvliet 6/2008 (1 sample)

Figure 5.5 - Relative cumulative number of particles/mL per particle size in the activated

sludge from the membrane tanks of full-scale MBRs. Figure 5.5 shows that, in all samples, 85 to 95% of the particles are smaller than 45 µm. In addition, the results of total number of particles (results not shown) have a standard deviation of 7.7%. MLSS concentrations from the samples presented in Figure 5.5 are between 11 and 12 g/L, which can account for a deviation of 8% in the total number of counted particles. Therefore, particle counting results of MBR sludge samples obtained through light-blockage methods are reproducible, contrary to the statements of Hargesheimer et al. (1998). The observed reproducibility of our MBR particle counting results using the light-blockage method can be explained through floc density and compactness. Gregory (1998) mentions that it is possible to produce highly compact flocs. Furthermore, agitation is likely to lead to restructuring and densification of flocs. The membrane tanks of full-scale MBRs are extremely agitated due to intensive aeration. Aggregates have extended possibilities of colliding and finding the best fit, resulting in a denser and more compact floc. Figure 5.6 shows the normalization of the average particle size distributions. Table V-V presents the average values of β at the lower and upper ends of the size range, shown in Figure 5.6, according to the origin of the sample.

70

0

1

2

3

4

5

1 1,2 1,4 1,6 1,8 2

Log (dp) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

10 100

Particle size [µm]

Schilde 4/2008

Schilde 8/2008

Monheim 9/2008

Nordkanal 7/2007

Heenvliet 3/2008

Heenvliet 6/2008

Lower end

[-----------------------------]Upper end

[------------] Figure 5.6 - Normalized particle counting data of full-scale MBR membrane tank

activated sludge.

Table V-V Average values of β, at the lower and upper ends of the size range, in each campaign at full-scale MBRs.

Average β

Lower end Upper end

Schilde (4/2008) -0.77 7.65

Schilde (8/2008) -0.57 7.77

Monheim (9/2008) -0.18 6.51

Nordkanal (7/2007) -0.52 7.30

Heenvliet (3/2008) -0.60 6.86

Heenvliet (6/2008) -0.43 6.60

Average -0.51 7.12

Stand. Deviation (%) 38 8 Figure 5.6 shows that, in each sample, there are several β values. The β values at the lower and upper end of the power-law distribution, presented in Table V-V, comply with the limits described by Lawler (1997). Table V-V shows that the β values at the upper end of the size range have a standard deviation of 8%, which can be attributed to the differences in MLSS concentrations in the samples. The β values at the lower end of the size range have a standard deviation of 38%. Considering that particle counting data between 2 and 10 µm are not reliable, as reported in Chapter 4, it is very well possible that the ranges at the lower end of the size range are inaccurate. In contrast, the β values at the upper end are stable in all MBRs, indicating that there are similarities in the sludge suspended fraction of the different MBRs. In Table V-VI presents the average sludge floc size, in the range of 2-100 µm, in the membrane tanks. As mentioned in Chapter 4, the average floc size was defined as the particle size where the maximum volume of particles was obtained. These results show that MBRs, with homogeneous filterability in all MBR tanks, such as Monheim, are

71

characterized by membrane tank flocs that are smaller than MBRs with heterogeneous filterability, such as Heenvliet. Apparently, particles in the range of 10-100 µm are being affected by the return ratio. Floc breakage due to return flow was also reported by Wisniewski et al. (1998).

Table V-VI Floc size in the membrane tanks of the full-scale MBRs, in the range of 2-100 µm.

Number of samples

Standard deviation on floc

size

Average floc size [µm]

Schilde (4/2008) 6 2.8 36

Schilde (8/2008) 6 1.6 37

Monheim (9/2008) 12 1.3 36

Nordkanal (7/2007) 5 5.5 49

Heenvliet (3/2008) 8 2.6 44

Heenvliet (6/2008) 1 - 46

Pilot-scale

The particle counting results of samples collected from the membrane tanks of the four pilot-scale MBRs are shown in Figure 5.7. The results are identified by location, month and year of collection and number of samples measured in each campaign. The presented results are average values for each campaign.

0

0,2

0,4

0,6

0,8

1

0 10 20 30 40 50 60 70 80 90 100

Particle size [µm]

Rel.

cum

ul.

Num

ber [#

/mL]

TUB/KWB 6/2007 (4 samples)

NTNU 10/2007 (3 samples)

UNITN 11/2007 (9 samples)

EAWAG 11/2007 (18 samples)

Figure 5.7 - Relative cumulative number of particles/mL per particle size in the activated

sludge from the membrane tanks of pilot-scale MBRs.

Figure 5.7 shows that the particle size distributions are very different between MBR pilots. At the NTNU pilot, 80% of the particles are smaller than 19 µm, at the UNITN and EAWAG pilots, 80% of the particles are smaller than 25 µm, and at the TUB/KWB pilot, 80% of the samples are smaller than 36 µm. Furthermore, the results of the total number of particles (results not shown) have a standard deviation of 42%. In the MBR

72

pilots the amount of smaller particles, between 10 and 30 µm, is higher than in full-scale MBRs and the particle size distributions are more scattered (Figure 5.5). A higher amount of smaller particles in the NTNU particle size distribution was expected. The BF-MBR at the NTNU is a combination of a moving bed biofilm reactor with a membrane separation step, meaning that the biomass in the aeration tank is attached to a physical support that is constantly moving due to the introduction of air (Leiknes et al. 2007). The amount of suspended solids reaching the membrane tank is therefore very low, with MLSS concentrations of 0.14 g/L (Table V-II). In the remaining MBR pilots, MLSS concentrations vary between 6.1 and 16.4 g/L (Table V-II). Considering that in the full-scale MBRs, MLSS concentrations vary between 11 and 12 g/L (Table V-I), we expected to find more scattering in the particle size distribution results of the MBR pilots than in the full-scale MBR results. Figure 5.8 presents the normalization of the average particle size distributions. Figure 5.8 shows that the value of log A (Equation 4.2) varies between 3.8 and 4.6, while in the full-scale results it varies between 4.3 and 4.4 (Figure 5.6), confirming the variability of the particle size distribution results of the MBR pilots. Among all MBR pilot results, the particle size distribution obtained at the TUB/KWB MBR is the most similar to the full-scale results (Figure 5.6). Noteworthy to point out that the TUB/KWB MBR was being operated with similar goals to full-scale MBRs, namely stable and continuous operation.

0

1

2

3

4

5

1 1,1 1,2 1,3 1,4 1,5 1,6 1,7 1,8 1,9 2

Log (dp) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

10 100Particle size [µm]

TUB/KWB 6/2007 (4 samples)

NTNU 10/2007 (3 samples)

UNITN 11/2007 (9 samples)

EAWAG 11/2007 (18 samples)

Lower end

[------------------------------]Upper end

[------------] Figure 5.8 - Normalized particle counting data of pilot-scale MBR membrane tank

activated sludge.

Table V-VII presents the β values at the lower and upper ends of the size range shown in Figure 5.8. The values comply with the limits defined by Lawler (1997), i.e. less than 1 at the lower end of the size range and greater than 4 at the upper end of the size range.

73

Table V-VII Average values of β at the lower and upper ends of the size range, in each campaign at MBR pilots.

Average β

Lower end Upper end

TUB/KWB (6/2007) -0.74 8.16

NTNU (10/2007) -0.27 6.73

UNITN (11/2007) -0.44 8.76

EAWAG (11/2007) -0.5 8.03

Average -0.48 7.92

Stand. Deviation (%) 35 9.1

Table V-VIII presents the average sludge floc size, in the range of 2-100 µm, in the membrane tanks of the MBR pilots studied in this research. As mentioned earlier, the average sludge floc was defined as the particle size where the maximum volume of particles is obtained. The average floc size at the TUB/KWB MBR is in the size range of full-scale MBR flocs (Table V-VI). The remaining MBR pilots produce smaller flocs than full-scale applications (Table V-VI). Table V-VIII Floc size in the membrane tanks of the pilot-scale MBRs, in the range of 2-

100 µm. Number of

samples Standard

deviation on floc size

Average floc size [µm]

TUB/KWB (6/2007) 4 1.2 42

NTNU (10/2007) 3 1.8 29

UNITN (11/2007) 9 2.6 25

EAWAG (11/2007) 18 3.3 33 The particle counting results in the range of 2-100 µm provide opposite conclusions to filterability, meaning that MBR pilot particle counting results are more heterogeneous than full-scale results. In section 5.3, we referred to the conclusion presented by Moreau (2010), stating that filterability in MBR pilots was less extreme and more homogenous than in full-scale applications, probably due to smaller variations in feed and temperature (Kraume et al. 2009). Several reasons can explain the heterogeneity of the particle counting results in the range of 2-100 µm in MBR pilots. Firstly, the MLSS range applied in the MBR pilots was 0.14 to 16.4 g/L, while in the full-scale MBRs it was 11-12 g/L (Table V-II and Table V-I, respectively). Particle counting results in the size range of 10-100 µm represent a fraction of the MLSS concentration results. Nonetheless, higher variability is expected in the MBR pilot results. Secondly, the particle size is extremely dependent on the aeration characteristics (Germain et al. 2005), which also define the hydrodynamics of the membrane tanks. Higher variability in the hydrodynamics of the pilot-scale MBRs, when compared to full-scale MBRs, may contribute to the heterogeneity obtained in the particle

74

counting results. Thirdly, differences in the biomass of pilot- and full-scale MBRs may also be responsible for the difference between pilot- and full-scale results. Kraume et al. (2009) reported higher SMP concentrations, in the form of polysaccharides, in pilot-scale MBRs than in full-scale MBRs. Additionaly, Drews et al. (2007) showed that SMP loading rate has the strongest influence on SMP elimination in MBRs. Considering that SMP concentrations derive from microbiological activity, differences in SMP concentrations are likely to indicate differences in microbiological activity and/or in feed characteristics.

5.6 Relationship between Filterability and Particle Counting in the Range of 2-100 µm

Full-scale and pilot-scale

Figure 5.9 shows the relationship between filterability and floc size at all the locations visited in this research. At the full-scale MBRs, the average floc size varies between 36 and 49 µm, but filterability is mainly below 0.5x1012m-1, with the exception of the Schilde 4/2008 campaign, the results of which are circled in Figure 5.9. At the pilot-scale MBRs, the average floc size varies between 25 and 42 µm and the filterability values are mainly below 1x1012m-1 of added resistance. In order to establish a possible tendency, a wide range of floc size and filterability data is required, so the results from pilot- and full-scale MBRs were coupled.

0

1

2

3

4

5

6

0 20 40 60 80 100

Floc size [µm]

∆R

20 [

x1

01

2m

-1]

Schilde 8/2008

Monheim 9/2008

Nordkanal 7/2007

Heenvliet 3/2008

Heenvliet 6/2008

TUB/KWB 6/2007

NTNU 10/2007

UNITN 11/2007

EAWAG 11/2007

Schilde 4/2008

Figure 5.9 - Filterability, represented by ∆R20 vs. average floc size.

Apparently, according to Figure 5.9, there is a relationship between filterability and floc size in MBR activated sludge. Good filterability is obtained when flocs are approximately 40 µm in size and poor filterability is obtained when flocs are approximately 20 µm in size. The results presented in Figure 5.9 also show that filterability does not only depend on floc size. In both Schilde campaigns, where filterability varied between poor (∆R20>1) and moderate (0.1<∆R20<1), the floc sizes were similar. The raw particle counting data from both Schilde campaigns (data not shown) also show that, despite the differences in filterability, particle counting in the range of 10-100 µm had similar distributions. In

75

addition, also the flocs at the Monheim and Schilde MBRs had similar sizes. However, activated sludge filterability was good at the Monheim MBR (∆R20<0.1) and poor (∆R20>1) to moderate (0.1<∆R20<1) at the Schilde MBR. Therefore, floc size is not the only factor affecting filterability. The results shown in Figure 5.9 indicate that there is a relationship between the suspended fraction of the activated sludge, characterized by floc size, and filterability. However, this relationship is only apparent when a wide range of filterability and floc size results is available. As mentioned earlier, floc size is extremely dependent on the aeration characteristics (Germain et al. 2005). In a single MBR, and particularly full-scale MBRs, aeration has a restricted operating range, so variations on floc size are likely to be limited.

5.7 Conclusions

Filterability and particle counting measurements where applied to MBR activated sludge at 4 full-scale and 4 pilot-scale municipal MBRs. We initially postulated that filterability changes could be measured in the particle size range of 2-100 µm. The deflocculation of micron particles would produce submicron particles, subsequently causing a worsening of sludge filterability. The conclusions are summarized as follows:

- In 88% of the full-scale membrane tank samples, filterability varied between good and moderate (0<∆R20<1).

- In full-scale MBRs, filterability is worse in winter campaigns, most likely due to the temperature impact on viscosity, thus decreasing aeration efficiency, and flocculation.

- In full-scale MBRs, filterability between the tanks of one MBR can be homogenous or heterogeneous, according to the return ratio. Apparently, if the return ratio is low, the best filterability is measured in the membrane tank.

- Particle counting results in the range of 2-100 µm of MBR activated sludge are reproducible using light-blockage particle counters.

- MBR activated sludge filterability improves with the activated sludge floc size. This result supports our initial hypothesis that differences in filterability could be measured in the suspended fraction of the sludge.

- A relationship between filterability and floc size is only apparent when a wide range of filterability and floc size results are available. In a single MBR, probably due to aeration restrictions, the relationship between filterability and floc size may not be visible.

- Filterability is affected by other factors besides floc size, such as microorganism’s kinetics and system stability. Even in the same MBR installation, activated sludge with similar floc sizes can have different filtration qualities.

76

77

6. Diluting Membrane Bioreactor Activated Sludge

6.1 Introduction

The main aim of this thesis is to study the role of MLSS concentration in MBR filtration. This chapter describes the experiments intended to manipulate MBR activated sludge, namely MLSS concentration, by diluting it with permeate. Literature reports on the effect of MLSS concentrations on membrane filtration are contradictory, as described in Chapter 2. Some authors report an increase in fouling with increasing MLSS concentration (Meng et al. 2007); other authors report no effect of suspended solids on the critical flux (Le-Clech et al. 2003) and others a decrease in critical flux for activated sludge samples with higher MLSS concentration (Defrance et al. 1999). To explain the abovementioned contradiction, the following hypothesis is proposed. In MBR activated sludge with low MLSS concentration, i.e. below a certain MLSS concentration, fouling particles are available in the free water of the bulk activated sludge. In the low concentration range, as the MLSS concentration increases, so does the amount of free fouling particles, responsible for an increasing fouling propensity. In activated sludge with high MLSS concentration, i.e. above a certain MLSS concentration, and moderate to good filtration quality, fouling particles become entrapped in the sludge matrix. In these circumstances, increasing MLSS concentrations will lead to an increasing entrapment of fouling particles, resulting in a particular structure responsible for an improvement in filterability. The MLSS concentration above which the entrapment of particles results in improved filterability is further referred to as critical MLSS concentration. Figure 6.1 clarifies the definitions used in the above hypothesis and, hence, in this research. Figure 6.2 represents the structure of activated sludge with low and high MLSS concentration, as proposed in the above hypothesis, produced in one MBR installation.

Filterability

Poor

MLSS [g/L]

Low MLSS High MLSS

Moderate

Good

Critical MLSS concentration

Figure 6.1- Schematics of concepts.

78

Figure 6.2 – MBR activated sludge structure with low MLSS concentration (left) and

high MLSS concentration (right), in one MBR installation.

In this research, wastewater was collected from membrane tanks of full-scale MBRs and submitted to filterability measurements, according to the DFCm, and other relevant measurements. The original sludge is then diluted with permeate to obtain suspensions with different MLSS concentrations, submitted to the following measurements: filterability, MLSS concentration, SMP concentrations, sludge viscosity, particle counting in the range of 0.4-5µm and particle counting in the range of 2-100 µm. In addition, several diluting experiments were carried out to assess our own results. With this aim in mind, the following diluting experiments were performed:

- diluting activated sludge from different tanks within the same MBR installation; - diluting activated sludge from membrane tanks of different MBR installations; - diluting MBR activated sludge and measuring filterability in a flat sheet

membrane installation; - diluting MBR activated sludge with permeate, tap water and CAS effluent; - diluting MBR activated sludge and measuring filterability using constant and

variable cross-flow velocity. The specific purposes of the above comparison tests are, respectively: to evaluate differences in activated sludge from one MBR; to evaluate differences in activated sludge from different MBRs; to evaluate the effect of membrane configuration on our own results; to evaluate the effect of the dilution media on our own results; to evaluate the effect of the hydraulic regime on our own results.

79

6.2 Methodology

The diluting experiments were organized in 44 sets performed between January 2007 and February 2009. For each set, a minimum of three filtration measurements are performed: one for the original sample and the other two for dilution samples. Identical sludge samples, 30 L each, from full-scale MBRs are collected. On one of them, hence referred to as the original sample, filterability is measured without dilution. To the remaining samples different quantities of permeate are added in order to obtain activated sludge suspensions with lower MLSS concentration, also referred to as dilution samples. The dilution samples are then submitted to filtration measurements. The original and dilution samples are also submitted to MLSS concentration, particle counting in the range of 2-100 µm and of 0.4-5 µm, viscosity and SMP measurements, depending on the vicinity and availability of the required equipment. Details of the materials and methods used in this research are provided in Chapter 4. The general purpose of each set of experiments is presented in Table VI-I. Each set is identified by number and date of the experiments.

Table VI-I Description of the diluting sets. Set # to set # Date # to date # General purpose

1 to 22

31-1-07 to 23-1-09 To characterize MBR activated sludge with

low and high MLSS concentrations

23 to 28

10-6-08 to 23-1-09 To compare activated sludge from MBR

membrane and aeration tanks

29 a 32

16-4-08 to 12-2-09 To compare membrane tank activated

sludge from different MBRs

- *

28-4-08 to 2-6-08 To compare filterability results obtained in

the DFCi and in a flat sheet membrane installation

33 a 38

29-1-08 to 7-2-08 To compare permeate, tap water and CAS

effluent as dilution media

39 to 44

8-1-09 to 23-1-09 To compare results using constant and

variable cross-flow velocity Key: * These experiments are unpublished work of Elif Atasoy, who kindly allowed us to include them

in this thesis.

Most of the MBR activated sludge samples were collected from the membrane tanks of the Heenvliet MBR (Hollandse Delta Water Board). For sets 23 to 28, shown in Table V-I, half of the samples were collected from the aeration tank of the Heenvliet MBR, described as follows. The Heenvliet MBR is composed of several tanks for carbonaceous and nutrient removal (Figure 7.1). At the Heenvliet MBR, before entering the membrane tank, the activated sludge is submitted to extra aeration in a tank known as aeration tank. One of the main differences between the membrane and aeration tank activated sludge is the MLSS concentration. Therefore, activated sludge samples were collected simultaneously from both tanks and submitted to

80

diluting experiments, following the aforementioned methodology. The specific purpose of these tests is to evaluate possible differences in activated sludge of one MBR. To evaluate differences in activated sludge from different MBR installations, samples were collected from membrane tanks of the Heenvliet MBR, Schilde MBR (AQUAFIN), Ootmarsum MBR (Regge and Dinkel) and Varsseveld MBR (Rijn and Ijssel). Details of these MBR installations are provided in Table VI-II. Table VI-II Characteristics of full-scale MBRs. WWTP Heenvliet

Netherlands Schilde

Belgium Ootmarsum

Netherlands Varsseveld

Netherlands Population Equivalent

- 13,000 28,000 18,500 23,150

Wastewater treatment

- CAS+MBR CAS+MBR CAS+MBR MBR

Location of the membranes

- Submerged in separate tanks

Submerged in separate tanks

Side-stream membranes

Submerged in separate tanks

Membrane Supplier

- Toray (Flat sheet)

Zenon (Hollow Fiber)

Norit (Multi-tube)

Zenon (Hollow Fiber)

Membrane pore size

µm 0.08 0.04 0.03 0.04

Total membrane area

m2 4,110 10,560 2,436 20,160

Flux (design) L/m2.h 24.3 33.6 55 37.5

Flux (average operation)

L/m2.h 18.2 33 45 25

MLSS g/L 12 11 9.2 10.2

Return ratio * - 1.5 5.8 - -

Sludge age d 20 21 42 33

Cleaning - Mechanical and chemical (twice a year)

Mechanical and chemical (once a week)

Mechanical and chemical (once every 2

months)

Mechanical and chemical (once a week)

Key: * return ratio from membrane tanks to carbonaceous or nutrient removal tanks.

The activated sludge collected in the different MBR installations was submitted to diluting experiments, as described above. To evaluate the effect of membrane configuration on our own results, diluting experiments were performed, following the methodology described above, by measuring the filterability in a flat sheet membrane installation (Atasoy 2008). Details of this installation are provided in Annex B. The activated sludge samples were collected from the membrane tanks of the Heenvliet MBR. The impact of the dilution media on our own results was evaluated as follows: MBR activated sludge samples were collected at the Heenvliet MBR. After the measurements of the original samples, dilution samples were prepared using permeate, tap water and CAS effluent as dilution media, collected from the aforementioned WWTP. The

81

methodology applied followed the steps of all other diluting experiments, described above. The effect of the hydraulic regime on our own results was evaluated using the following approach: Geilvoet (2010) showed that, in the sludge filtration step of the DFCm, a laminar flow is obtained for activated sludge samples with MLSS concentration exceeding 5 g/L. Rosenberger et al. (2002c) proposed a minimum cross-flow velocity of 2.9 m/s, for MBR activated sludge with MLSS concentration of 20 g/L, to obtain a minimum Reynolds number of 5,000 in tubular membranes. Therefore, in order to assure that a turbulent regime would be created in the DFCi, the following cross-flow velocities were selected for the original, dilution 1 and dilution 2 samples: 2.5, 2.2 and 1.8 m/s, respectively. MBR activated sludge was collected from the Heenvliet MBR and diluted with permeate. The same MBR activated sludge samples, original and dilutions, were used to measure filterability using constant cross-flow velocity, at 1 m/s, and variable cross-flow velocity, at the previously mentioned velocities, varying by type of sample, namely, original, dilution 1 and dilution 2.

6.3 Characterization of Membrane Bioreactor Activated Sludge: Results and Discussion

Filterability

The filterability results obtained in the sets intended to characterize MBR activated sludge are summarized in Figure 6.3. Figure 6.3 shows the average results obtained from 21 sets of experiments, therefore 63 sludge filtration tests. Among the total 22 sets of experiments, intended to characterize MBR activated sludge, there was only one set that did not match the results presented in Figure 6.3. In Figure 6.3, the original sample results are represented by larger dots. The arrows indicate the sequence of the filterability measurements performed. The analysis of the filterability results behavior within each set showed that the filterability results should be split into two groups, as presented in Figure 6.3. The first group, “High MLSS”, shows the results obtained when the original samples had MLSS concentration exceeding 10.5 g/L. The second group, “Low MLSS”, shows the results obtained when the original samples had MLSS concentration equal to or lower than 10 g/L. In this research, the terms low and high MLSS concentration always assume 10.5 g/L as the critical MLSS concentration.

82

0,0

0,4

0,8

1,2

1,6

2,0

0 5 10 15 20

MLSS[g/L]

∆R

20 [*

10

12m

-1]

High MLSS

Low MLSS

Figure 6.3 - Average filterability vs. MLSS concentration.

The results presented in Figure 6.3 show that the filterability of the dilution samples varies according to the MLSS concentration in the original sample. MBR activated sludge with low MLSS concentration produces dilutions with increased filterability. The opposite occurs for original samples with high MLSS concentration. MBR activated sludge with high MLSS concentration produces dilution samples with worse filterability. Figure 6.4 presents the filterability results by set, according to sample MLSS concentration, for the aforementioned “Low MLSS” group.

Figure 6.4- Filterability vs. MLSS concentration by set.

Figure 6.4 shows that the filterability of the original samples varied between 1.62 and 0.34 x1012m-1. According to the classification proposed by Geilvoet (2010), filtration quality varied between poor and moderate. No original samples with good filtration quality, i. e. with ∆R20 below 0.1 x1012m-1, and MLSS concentration equal to or lower than 10 g/L were obtained in this research. However, an increase in filterability in activated sludge with lower MLSS concentration corresponds to the theoretically expected results. MLSS concentration has a direct effect on viscosity: lower MLSS concentration results in lower viscosity (Rosenberger et al. 2002c; Hasar et al. 2004).

0,0

0,4

0,8

1,2

1,6

2,0

0 5 10 15MLSS [g/L]

∆R

20 [*

10

12m

-1]

Set 1

Set 2

Set 3

Set 20

83

Consequently, suspensions with lower viscosity will more easily create turbulent regimes, which are considered preferable for process operation in cross-flow filtration membranes (Rosenberger et al. 2002c). In standard operation conditions of the DFCi, activated sludge with MLSS concentration below 5 g/L will create turbulent regimes, while sludge with MLSS concentration above 5 g/L will create laminar regimes (Geilvoet 2010). Nevertheless, for activated sludge with MLSS concentrations between 10.5 g/L and 5 g/L, a decreasing viscosity is most likely to lead to improved filterability. Therefore, it can be assumed that an increase in filterability by diluting occurs in MBR activated sludge samples with low MLSS concentration, regardless of the filtration quality of the original sludge. Figure 6.5 presents filterability vs. MLSS concentration, by set, in all the sets belonging to the aforementioned “High MLSS” group. Figures 6.5 a) and b) present different ranges of ∆R20 values.

a) b)

Figure 6.5 - Filterability vs. MLSS concentration by set. Figure 6.5 shows that the filterability of the original samples varied between 0.52 and 0.01. According to the classification proposed by Geilvoet (2010), the filtration quality of the original samples varied between moderate and good. All original samples represented in Figure 6.5 produced dilutions with decreased filterability. An increase in viscosity in activated sludge with higher MLSS concentration should lead to decreased filterability. However, as previously reported, while some authors measured increasing fouling with increasing MLSS concentrations (Meng et al. 2007), others reported a lower critical flux with higher MLSS concentrations (Defrance et al. 1999). These apparently contradictory results and our own results, shown in Figure 6.5, can be explained through our hypothesis, as follows. Activated sludge with high MLSS concentration and moderate to good filterability retains fouling particles in its matrix, resulting in increasing filterability with increasingly higher MLSS concentrations; in contrast, activated sludge with high MLSS concentration but poor filtration quality does not retain fouling particles, resulting in decreasing filterability with increasingly high MLSS concentrations.

0,0

0,1

0,2

0,3

0,4

0,5

0 5 10 15 20MLSS [g/L]

∆R

20

[*1

01

2m

-1]

Set 5

Set 6

Set 10

Set 11

Set 12

Set 13

Set 14

Set 15

Set 16

Set 17

Set 180,0

0,4

0,8

1,2

1,6

2,0

0 5 10 15 20MLSS [g/L]

∆R

20

[*1

01

2m

-1]

Set 7

Set 8

Set 9

Set 19

Set 21

Set 22

84

Figure 6.6 shows the ∆R20 results according to the respective αR*ci product and s coefficient, obtained in sets 1 to 22. The results are identified by the MLSS concentration of the original sample in each set, namely high or low MLSS, as defined previously.

Figure 6.6 - ∆R20 vs. αR*ci product and s coefficient in the diluting experiments.

Figure 6.6 shows that the ∆R20 value is linearly related to the αR*ci product, i. e. the specific resistance by the concentration of fouling particles. A linear regression between the two aforementioned parameters provides a R2 fitting of 0.99. This result is also obtained when results from “Low MLSS” or “High MLSS” sets are considered separately. In contrast, the results of the s coefficient are not statistically related to ∆R20. The majority of the s coefficient results are between 0 and 0.2, indicating that the cake layer is hardly compressible. In “Low MLSS” sets, 17% of the samples provide an s coefficient of 0, while in the “High MLSS” sets, 28% of the samples present an s coefficient equal to 0, indicating that an incompressible cake layer is more frequently obtained in “High MLSS” sets. Figure 6.7 represents the relationship between the MLSS concentration of the samples and the resulting αR*ci product and s coefficient. The results are identified by the MLSS concentration of the original sample in each set, as defined previously.

0,0

0,4

0,8

1,2

1,6

2,0

0 20 40 60 80 100αR*ci [*10

-3m

2]

∆R

20 [*1

01

2m

-1]

0,0 0,2 0,4 0,6 0,8 1,0s [-]

aR*ci (High MLSS)

aR*ci (Low MLSS)

s (High MLSS)

s (Low MLSS)

αR*ci

αR*ci

85

0

20

40

60

80

100

0 5 10 15 20

MLSS [g/L]

αR

*ci [

*10

-3m

2]

High MLSS

Low MLSS

0,0

0,1

0,2

0,3

0,4

0 5 10 15 20

MLSS [g/L]

s [

-]

High MLSS

Low MLSS

a) b) Figure 6.7 - αR*ci product (a) and s coefficient (b) vs. MLSS concentration.

Figure 6.7 a) shows that, in this research, there is a clear difference between samples with high and low MLSS concentration. On average, the αR*ci product in activated sludge samples with high MLSS concentration is 13.6 x10-3m2 while in samples with low MLSS concentration is 21.6 x10-3m2. Therefore, activated sludge with high MLSS concentration and moderate to good filtration quality produces a cake layer with a lower specific resistance by the concentration of fouling particles. Figure 6.7 b) shows that, on average, activated sludge samples with high MLSS concentrations have an s value of 0.05, while sludge samples with low MLSS concentrations have an s value of 0.1, indicating that the cake layer produced by these samples is more compressible than the cake layer produced by samples with high MLSS concentration. On average, in activated sludge with low MLSS concentration, the cake layer has a higher specific resistance by the concentration of fouling particles and higher compressibility. These results show that the cake layers produced by samples with high and low MLSS concentration have different characteristics. Furthermore, the specific resistance by the concentration of fouling particles and compressibility are lower in samples with high MLSS concentration and moderate to good filtration quality. During the filterability measurements, several parameters are measured and calculated. One of these parameters is the longitudinal pressure loss along the tube, referred to as ∆P. An analysis based on Darcy-Weisbach equations and measured ∆P showed that, in the DFCm, a laminar regime is obtained for samples with MLSS concentration exceeding 5 g/L (Geilvoet 2010). Therefore, a transition from turbulent to laminar regime does not explain the decrease in filterability of dilution samples produced from original “high MLSS” samples because such decrease occurs in samples with MLSS concentration exceeding 5 g/L, thus when the regime is constantly laminar. The abovementioned ∆P parameter is understood as a rough shear stress indicator. Within each set of experiments, ∆P varied to a maximum of 0.02 bar, which represents 10 to 16% of the total measured ∆P. Furthermore, higher ∆P values are obtained in the samples with higher MLSS concentration, within each set (results not shown). The viscosity of the sludge increases with increased MLSS concentration (Rosenberger et al. 2002a), thus, at

86

a constant flow velocity, also the shear stress increases by filtrating activated sludge with higher MLSS concentration. Therefore, it is likely that ∆P variation can be attributed to an increased shear stress resulting from the higher MLSS concentration in the sludge. Soluble Microbial Products

SMP have been identified as the main fouling particles in MBR activated sludge, (Lesjean et al. 2005; Rosenberger et al. 2005). The main constituents of SMP are considered to be proteins and polysaccharides. In our research, SMP concentrations are slightly higher in samples with MLSS concentration exceeding 10.5 g/L, as follows: the average concentration of proteins and polysaccharides is 15.9 and 10.8 mg/L, in samples with MLSS concentration exceeding 10.5 g/L; the average concentration of proteins and polysaccharides is 12 and 8.6 mg/L, in samples with MLSS concentration equal to or lower than 10 g/L. Figure 6.8 represents filterability vs. protein and polysaccharide concentrations in the bulk activated sludge and retained in the DFCi membrane. It is noteworthy to point out that, in this research, the applied methodology affected the obtained SMP results because the activated sludge was diluted with permeate, which also contains SMP.

0,0

0,4

0,8

1,2

1,6

2,0

0 10 20 30 40

Proteins [mg/L]

∆R

20 [

*10

12m

-1]

Bulk

Retained

0,0

0,4

0,8

1,2

1,6

2,0

0 10 20 30 40

Polysaccharides [mg/L]

∆R

20 [

*10

12m

-1]

Bulk

Retained

a) b)

Figure 6.8 – Filterability vs. protein a) and polysaccharide b) concentrations in bulk activated sludge and retained on DFCi membrane.

Figure 6.8 indicates that, in this research, filterability and SMP results are not related. Maximum protein and polysaccharide concentrations are obtained in samples with moderate filtration quality and average values in samples with poor filtration quality. Figure 6.9 represents the protein and polysaccharide concentrations obtained in the bulk activated sludge vs. the SMP concentrations retained in the DFCi membrane.

87

0

10

20

30

40

0 10 20 30 40

Proteins bulk [mg/L]

Pro

tein

s r

eta

ine

d [

mg

/L]

0

10

20

30

40

0 10 20 30 40

Polysaccharides bulk [mg/L]

Po

lys

ac

ch

ari

de

s r

eta

ine

d [

mg

/L]

a) b) Figure 6.9 – Protein a) and polysaccharide b) concentrations retained on the DFCi membrane vs. protein and polysaccharide concentrations in bulk activated sludge.

Figure 6.9 shows that there is indeed a relationship between SMP concentrations in the bulk activated sludge and retained in the DFCi membrane. A linear regression between the aforementioned parameters provides a R2 fitting of 0.67 for protein concentrations and of 0.99 for polysaccharide concentrations, demonstrating a stronger relationship for polysaccharides than for proteins. On average, 46% of the proteins are retained, while 70 % of the polysaccharides remain on the membrane. It is noteworthy to point out that the definition of retained SMP, in cross-flow operation mode, refers to SMP potentially retained in the membrane, i.e. SMP that would be indeed retained if it was a dead-end operation mode.

Viscosity

Figure 6.10 represents the apparent viscosity results according to the applied shear rate, for all measured samples in sets 1 to 22.

0

50

100

150

200

0 20 40 60 80 100

Shear rate [s-1

]

Ap

pa

ren

t v

ico

sity

[m

Pa

*s]

2,200 g/L 2,620 g/L

0,00 g/L 3,663 g/L

3,670 g/L 4,235 g/L

4,608 g/L 5,338 g/L

5,965 g/L 6,300 g/L

6,820 g/L 7,258 g/L

7,535 g/L 7,535 g/L

7,760 g/L 8,455 g/L

8,490 g/L 9,215 g/L

10,060 g/L 10,410 g/L

10,660 g/L 10,735 g/L

12,290 g/L 12,960 g/L

14,010 g/L 14,075 g/L

14,325 g/L 15,098 g/L

15,150 g/L 15,670 g/L

16,125 g/L 17,030 g/L Figure 6.10 - Apparent viscosity vs. applied shear rate for MBR activated sludge

samples identified by their MLSS concentration.

88

Figure 6.10 shows that the MBR activated sludge behaved as a non-Newtonian fluid, with a decreasing apparent viscosity along with an increasing applied shear rate, as previously reported (Rosenberger et al. 2002c; Hasar et al. 2004; Meng et al. 2007). These authors also report a strong relationship between MLSS concentration and viscosity, which can also be seen in Figure 6.10: samples with higher MLSS concentrations generally present a higher apparent viscosity. Several authors describe the apparent viscosity of MBR activated sludge through shear rate and MLSS concentration, at a temperature of 20ºC (Rosenberger et al. 2002c; Laera

et al. 2007; Meng et al. 2007; Yang et al. 2009). Rosenberger et al. (2002c) measured viscosity in original sludge samples collected in 9 MBRs; Laera et al. (2007) measured viscosity in pre-sieved sludge samples from one bench-scale MBR; Yang et al. (2009) measured viscosity of sludge collected in one MBR and obtained sludge with different MLSS concentrations by diluting with effluent or by dewatering the sludge. Figure 6.11 presents a comparison between the viscosity results obtained in this research and calculated by applying the equations proposed by these authors, as shown in Table II-II, for a shear rate of 100 s-1 and at a temperature of 20ºC.

0

10

20

30

40

0 5 10 15 20

MLSS [g/L]

Appare

nt vis

cosity

[m

Pa*s

]

Yang et al. (2009)

Laera et al. (2007)

This research

Rosenberger et al. (2002)

Figure 6.11 - Apparent viscosity vs. MLSS concentration of MBR activated sludge.

Figure 6.11 clearly shows that the obtained and estimated viscosity results of samples with low MLSS concentration, i.e. less than 10.5 g/L, are less scattered than those of samples with high MLSS concentration, i.e. above 10.5 g/L. The viscosity in samples with low MLSS concentration varies between 0.8 and 16 mPa.s, while in samples with high MLSS concentration, it varies between 3.4 and 24.5 mPa.s. Both the viscosity results obtained in this research and the estimated values from reported models show that viscosity is much more variable in the samples with high MLSS concentration than in the samples with low MLSS concentration. In this research, the viscosity results in samples with low MLSS concentration had a standard deviation of 3.6, while the standard deviation in viscosity results of samples with high MLSS concentration was 6.5. Figure 6.11 also shows that viscosity depends on other factors besides temperature, shear rate and MLSS concentration. The apparent viscosity measured in this research is between the results of the models proposed by Laera et al. (2007) and Rosenberger et al.

89

(2002c). While the model proposed by Laera et al. (2007) underestimates the value of viscosity, the models proposed by Rosenberger et al. (2002c) and Yang et al. (2009) overestimate it. If the abovementioned factors were enough to describe viscosity, the obtained values in this research would match the estimated values. Furthermore, the unaccounted factors affecting viscosity have greater impact on samples with high MLSS concentration than on those with low MLSS concentration, because the variability of the viscosity results is higher in samples with high MLSS concentration. The relationship between apparent viscosity, at a shear rate of 100 s-1, MLSS concentration and filterability is shown in Figure 6.12.

0

5

10

15

20

0 5 10 15 20 25 30

Apparent viscosity [mPa*s]

ML

SS

[g

/L]

0,01

0,1

1

10

∆R

20 [

*10

12m

-1]

MLSS

Dr20

Figure 6.12 - Apparent viscosity results vs. MLSS concentration and filterability.

If all the results presented in Figure 6.12 are taken into account, no statistical relationship is visible between filterability and apparent viscosity. The Pearson coefficient result between the aforementioned parameters is -0.15. The maximum R2 fitting obtained, particularly when multiple 1st to 4th order logarithmic and exponential regressions are applied, is 0.08. As Figure 6.12 shows, there is a tendency for values of viscosity exceeding 15 mPa.s to be measured when MLSS concentration is high and filterability is moderate to good. In contrast, values of apparent viscosity lower than 15 mPa.s, are mainly measured in samples with low MLSS concentration and moderate to poor filterability. Particle counting in the range of 0.4-5 µm

Figure 6.13 and 6.14 represent the particle counting distributions and normalized data, in the range of 0.5-1 µm, for all measured samples in sets 1 to 22. The results are presented as average numbers of the “High MLSS” and “Low MLSS” groups, as described above, and identified by type of sample, namely: original, dilution 1 and dilution 2. The number of samples used to calculate each average is indicated in brackets in the key.

90

0

500

1000

1500

2000

2500

3000

0,5 0,6 0,7 0,8 0,9 1,0

Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[#

*103

/mL

]

Original Low MLSS (1 sample)

Dilution 1 Low MLSS (1 sample)

Dilution 2 Low MLSS (1 sample)

0

500

1000

1500

2000

2500

3000

0,5 0,6 0,7 0,8 0,9 1,0

Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[#

*103

/mL

]

Original High MLSS (9 samples)

Dilution 1 High MLSS (9 samples)

Dilution 2 High MLSS (9 samples)

a) b) Figure 6.13 - Average number of particles/mL per particle size, for “Low MLSS” sets

(a) and for “High MLSS” sets (b) by type of sample.

4

5

6

7

8

-0,25 -0,2 -0,15 -0,1 -0,05 0

Log (Particle size) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

0,56 0,66 0,76 0,86 0,96

Particle size [µm]

Original Low MLSS (1 sample)

Dilution 1 Low MLSS (1 sample)

Dilution 2 Low MLSS (1 sample)

4

5

6

7

8

-0,25 -0,2 -0,15 -0,1 -0,05 0

Log (Particle size) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

0,56 0,66 0,76 0,86 0,96

Particle size [µm]

Original High MLSS (9 samples)

Dilution 1 High MLSS (9 samples)

Dilution 2 High MLSS (9 samples)

a) b) Figure 6.14 - Normalized particle counting data for “Low MLSS” sets (a) and for “High

MLSS” sets (b) by type of sample.

Figure 6.13 shows that the amount of particles in the “Low MLSS” group is significantly higher than in the “High MLSS” group. The normalized data presented in Figure 6.14 has a fundamental difference between the “Low MLSS” and “High MLSS” groups. In the “High MLSS” samples, the absolute value of β, i.e. the normalized distribution slope, increases from Original to Dilution 1 sample and so forth, as opposed to “Low MLSS” samples. However, since group “Low MLSS” is composed of only one set, no conclusions should be drawn at this point. The number of particles/mL per particle size and normalized data in the “High MLSS” group, presented in Figures 6.13 b) and 6.14 b), respectively, show that the original sample has a higher amount of particles than dilution 1 and so forth. A reduction in the number of particles between the original, dilution 1 and dilution 2 samples was expected due to the dilution operation. Shifting of the particle size distribution, i.e. when a higher

91

number of particles/mL is counted in the smaller size ranges while a lower number of particles/mL is counted in the larger size ranges, shows breakage or deflocculation of the activated sludge particles. In figures presenting normalized data, breakage or deflocculation of the sludgewould be represented by unparallel lines. The described result was not verified in any dilution samples of “High MLSS” sets in the range of 0.4-5 µm. Therefore, it can be concluded that there is a reduction in the number of particles/mL, from original to dilution 1 samples and from dilution 1 to dilution 2 samples, due to the diluting operation itself, but no deflocculation was visible in the range of 0.4-5 µm. Geilvoet (2010) showed that the volume of submicron particles, i.e. the accumulated volume of particles in the range of 0.4-1 µm (assuming that all counted particles have a spherical shape), is a better indicator of filterability than SMP concentrations. Figure 6.15 shows the average filterability and average volume of submicron particles, according to the average MLSS concentration in the samples. The number of sets used to calculate each average is indicated in brackets in the key. The arrows show the sequence of the measurements.

Figure 6.15 - Average filterability and average volume of submicron particles vs. average

MLSS concentration of MBR activated sludge.

As mentioned earlier, the original sludge of the “High MLSS” group produces dilutions with decreased filterability. Figure 6.15 shows that, in terms of the volume of submicron particles, the dilutions of the “High MLSS” group have a lower volume of submicron particles than the original samples. A reduction in the volume of submicron particles can be explained by the dilution procedure, but this does not explain the decreased filterability in dilution samples of the “High MLSS” group. Figure 6.15 also shows that dilutions of the “Low MLSS” group do not show a decrease in the volume of submicron particles. In the aforementioned case, the original and dilution samples have similar values of submicron particles. However, the result is not conclusive because it is provided by only one set.

0,01

0,1

1

10

0 5 10 15 20

MLSS [g/L]

∆R

20[*

10

12m

-1]

0

0,2

0,4

0,6

0,8

1

Vo

lum

e s

ub

mic

ron

[*1

06 µ

m3/m

L]

DR20 High MLSS(9 sets)

DR20 Low MLSS(1 set)

V.sub micron LowMLSS (1 set)

V. sub micronHigh MLSS (9sets)

92

Particle counting results in the range of 2-100 µm

Figure 6.16 presents the average particle counting distributions, in the range of 10-100 µm, for “Low MLSS” and “High MLSS” groups, by type of sample, namely: original, dilution 1 and dilution 2. The number of samples used to calculate each average is indicated in brackets in the key.

0

5

10

15

20

25

0 20 40 60 80 100

Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[*1

03/m

L]

Original High MLSS (13 samples)

Dilution 1 High MLSS (13 samples)

Dilution 2 High MLSS (13 samples)

a) b) Figure 6.16 - Average number of particles/mL per particle size, for “Low MLSS” sets

(a) and for “High MLSS” sets (b) by type of sample.

Figure 6.16 shows that, in the suspended fraction of the sludge, i.e. the fraction comprised of particles with sizes larger than 1 µm, according to the definition proposed by Metcalf & Eddy (2003), there are no apparent differences in the results of “High MLSS” and “Low MLSS” groups. In both groups, there are similar amounts of particles/mL and particle size distributions between types of samples. In both groups shown in Figure 6.16, two types of effects are visible: the effect of dilution, represented by full arrows, and the effect of deflocculation, represented by dashed arrows. A dilution operation, as mentioned earlier, causes a reduction in the number of counted particles in all size ranges. In Figures 6.16 a) and b), this effect is visible in the 25-100 µm size range in both dilution 1 and dilution 2 samples. As mentioned earlier, a deflocculation process causes higher counts in the smaller size ranges, while lower numbers are counted in the larger size ranges. In Figures 6.16 a) and b), higher numbers of particles/mL are counted in the 10-25 µm size range, while lower numbers are obtained in the 25-100 µm size range, in both dilution 1 and dilution 2 samples. A dilution operation and a deflocculation process occur in both the “High MLSS” sets and the “Low MLSS” sets when samples are diluted, resulting in similar particle size distributions, by type of sample. Figure 6.17, presents the average normalized particle counting data, in the range of 10-100 µm, for samples grouped according to the MLSS concentration in the original sample, i.e. “High MLSS” and “Low MLSS”, as described earlier, by type of sample, namely: original, dilution 1 and dilution 2. The number of samples used to calculate each average is indicated in brackets in the key.

0

5

10

15

20

25

0 20 40 60 80 100

Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[*1

03/m

L]

Original Low MLSS (1 sample- set 20)

Dilution 1 Low MLSS (1 sample-set 20)

Dilution 2 Low MLSS (1 sample-set 20)

93

1

2

3

4

5

1 1,2 1,4 1,6 1,8 2

Log (dp) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/mL

/ µm

]

10 100

Particle size [µm]

Original Low MLSS (1 sample)

Dilution 1 Low MLSS (1 sample)

Dilution 2 Low MLSS (1 sample)

Lower end

[-----------------------------]

Upper end

[------------]

1

2

3

4

5

1 1,2 1,4 1,6 1,8 2

Log (dp) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/mL

/ µm

]

10 100

Particle size [µm]

Original High MLSS (13 samples)

Dilution 1 High MLSS (13 samples)

Dilution 2 High MLSS (13 samples)

Lower end

[-----------------------------]

Upper end

[------------]

a) b) Figure 6.17 - Normalized particle counting data for “Low MLSS” sets (a) and for “High

MLSS” sets (b) by type of sample.

Figure 6.17 confirms the similarity, in the suspended fraction of the sludge, between types of samples (i.e. original, dilution 1 and dilution 2) belonging to the two groups under analysis. Furthermore, dilution, represented by the splitting of the original and dilution sample distributions, and deflocculation, represented by a higher number of particles at the lower end of the size range, coupled with a lower number of particles at the upper end of the size range, are visible when comparing the dilution and original samples in both groups: “High MLSS” and “Low MLSS”.

94

6.4 Comparison between Membrane and Aeration Tank Activated Sludge

Filterability according to the MLSS concentration in the MBR activated sludge, collected simultaneously in the membrane tank and aeration tank of the Heenvliet MBR, is shown in Figure 6.18. The sets were grouped in pairs, one collected from the aeration tank and the other from the membrane tank. The original samples are represented by larger dots and the arrows indicate the sequence of the dilutions performed.

Figure 6.18 - Filterability vs. MLSS concentration of samples collected from the aeration

tank (white dots) and membrane tank (black dots) of the Heenvliet MBR. Figure 6.18 shows that in this MBR, specifically the Heenvliet MBR, there are significant variations in filterability. The filterability of original samples collected from the membrane tank is always higher than that of original samples collected from the aeration tank. It can be concluded that in the membrane tanks of the Heenvliet MBR, the reverse of a diluting operation is occurring, i.e. the membrane tank is concentrating, resulting in a higher MLSS concentration and better filterability. The results presented in Figure 6.18 confirm the reliability of our diluting experiments. The filterability of dilution samples, produced with membrane tank sludge, corresponds to the filterability of original samples collected from the aeration tank.

The results of αR*ci product, s coefficient, SMP concentrations, viscosity, particle counting in the range of 0.4-5 µm and 2-100 µm provided similar conclusions to those obtained in sets 1 to 22, performed to characterize MBR activated sludge with low and high MLSS concentrations. Therefore, the aforementioned results are presented in Annex C and will be discussed in section 6.9.

0,0

0,4

0,8

1,2

1,6

2,0

0 5 10 15 20

MLSS [g/L]

∆R

20[*

10

12m

-1]

Set 23+24

Set 25+26

Set 27+28

95

6.5 Comparison between Membrane Tank Activated Sludge from Different Membrane Bioreactors

Filterability

The filterability vs. the MLSS concentration in activated sludge collected at the Schilde, Ootmarsum and Varsseveld MBRs are presented, by set, in Figure 6.19. The arrows indicate the sequence of the diluting experiments performed. The original sludge is represented by larger dots.

a) b) Figure 6.19 - Filterability vs. MLSS concentration performed with activated sludge

from the Schilde MBR (a), Ootmarsum and Varseeveld MBRs (b).

Figure 6.19 a) shows that the MLSS concentration in the original sample collected at the Schilde MBR exceeded 10.5 g/L but there was no decrease in filterability in the dilution samples, as obtained in the other previous 21 sets. It is noteworthy to point out that, when the aforementioned sludge was collected, an unusual growth of filamentous bacteria, namely Nocardia amarae, was occurring at the Schilde MBR (Moreau 2010). Our own results show that the filtration quality of the original sample is poor. Apparently, a decrease in the filterability of dilution samples collected from original sludge with high MLSS concentration does not occur when the sludge has already a very poor filtration quality. In the sets presented in Figure 6.19 b), the original samples show poor filtration quality and low MLSS concentration. In these sets, the dilution samples have better filtration quality than the original samples as in 7 previous sets of data. It can be concluded that an increase in filterability by diluting occurs in sludge with low MLSS concentration, regardless of the sampled MBR. The results of αR*ci product, s coefficient, SMP concentrations and particle counting in the range of 0.4-5 µm provided no further conclusions and are therefore presented in Annex C and will be discussed in section 6.9.

0

1

2

3

4

5

0 5 10 15

MLSS [g/L]

∆R

20[*

10

12m

-1]

0

1

2

3

4

5

0 2 4 6 8 10

MLSS [g/L]

∆R

20[*

10

12m

-1]

Set 30(Ootmarsum)

Set 31(Varseeveld)

Set 32(Varseeveld)

96

Viscosity

The MBR activated sludge, in the sets collected from the Ootmarsum and Varsseveld MBRs, behaved as a non-Newtonian fluid, with decreasing apparent viscosity along with increasing applied shear rate (results not shown). These results are in line with published results (Rosenberger et al. 2002c; Hasar et al. 2004; Meng et al. 2007) and with our own results for sets 1 to 28. The relationship between apparent viscosity, at a shear rate of 100 s-1, MLSS concentration and filterability, according to the origin of the sample, is represented in Figure 6.20. Figure 6.20 also presents linear regressions and the corresponding R2 between apparent viscosity and MLSS concentration values and between apparent viscosity and ∆R20 values.

y = 1,2191x + 0,8966

R2 = 0,8191

y = 0,9638x + 1,2058

R2 = 0,9956

0

5

10

0 5 10 15Apparent viscosity [mPa*s]

ML

SS

[g

/L]

MLSS (Ootmarsum)

MLSS (Vaarseveld)

a) b)

Figure 6.20 - Apparent viscosity results vs. MLSS concentration (a) and filterability (b) in samples obtained from the Ootmarsum and Varseeveld MBRs.

In all the samples represented in Figure 6.20 a), apparent viscosity is lower than 15 mPa.s. Furthermore, the MLSS concentration is low, i.e. equal to or lower than 10 g/L. As mentioned earlier, for sets 1 to 22, viscosity of sludge samples with low MLSS concentration is strongly influenced by MLSS concentration, temperature and shear rate. This conclusion is supported by the strong correlation between apparent viscosity and MLSS concentration, represented in Figure 6.20 a). The data presented in Figure 6.20 b) also show a strong correlation between apparent viscosity and filterability, as opposed to the results obtained in sets 1 to 28. The activated sludge quality in sets 1 to 28 ranged from poor to good. The sets presented in Figure 6.20 b) refer to activated sludge mainly with poor filtration quality. Apparently, the effect of viscosity on filterability depends on the activated sludge quality, thus suggesting a second order relationship between viscosity and filterability.

y = 0,6525x + 0,2505

R2 = 0,9209

y = 0,2963x - 0,2866

R2 = 0,9936

0,1

1

10

0 5 10 15

Apparent viscosity [mPa*s]

∆R

20 [

*10

12m

-1]

DR20 Ootmarsum

DR20 Vaarseveld

∆R20 ∆R20

97

Particle counting in the range of 2-100 µm

Figures 6.21 and 6.22 present the average particle counting distribution and normalized particle counting data, respectively, in the range of 2-100 µm, for the set of samples collected from the Schilde MBR.

0

5

10

15

20

25

30

0 20 40 60 80 100

Particle size [µm]

Nu

mb

er

of P

art

icle

s [#

*103

/mL

]

Original (Schilde MBR)

Dilution 1 (Schilde MBR)

Dilution 2 (Schilde MBR)

Figure 6.21 - Number of particles/mL per particle size, for samples collected from the

Schilde MBR.

1

2

3

4

5

1 1,2 1,4 1,6 1,8 2

Log (dp) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

10 100

Particle size [µm]

Original (Schilde MBR)

Dilution 1 (Schilde MBR)

Dilution 2 (Schilde MBR)

Lower end

[----------------------------]

Upper end

[------------] Figure 6.22 - Normalized particle counting data for samples collected from the Schilde

MBR.

Figure 6.21 shows that the dilution effect, previously described for the particle size range of 2-100 µm, is minimized. In both Figures 6.20 and 6.21, a detachment between the particle size distributions of the original and dilution samples is visible, due to the dilution operation itself, but not between the two dilution samples. Furthermore, no deflocculation occurred, which can be seen at the lower end of the size range in Figure 6.22, where the normalized distribution of all samples is overlapping. The deflocculation effect in the particle size distribution could have been avoided due to the presence of filamentous bacteria, as mentioned earlier.

98

6.6 Results Using a Flat Sheet Membrane Installation

The filterability according to the MLSS concentration in the samples, obtained when measuring the filterability in a flat sheet unit (FSu), is presented in Figure 6.23, by set. The sets are identified by date and the original samples are represented by larger dots.

0,0

0,4

0,8

1,2

1,6

2,0

0 5 10 15 20

MLSS [g/L]

∆R

20 [*

10

12m

-1]

02.06.08

30.05.08

28.04.08

Figure 6.23 - Filterability vs. MLSS concentration in MBR activated sludge.

In the sets represented in Figure 6.23, MLSS concentration exceeded 10.5 g/L in all original samples. Figure 6.23 shows that when the filterability is measured in the FSu, there is also a decrease in filterability in the dilution samples. The magnitude of the decrease is smaller than that measured in the tubular membrane installation, used in the DFCi. However, while in the DFCi sludge filtration is performed at a constant flux of 80 L/m2.h, in the FSu the constant flux was 40 L/m2.h (Annex B). The difference in flux is probably responsible for the lower decrease in filterability measured in the FSu. Nevertheless, MBR activated sludge samples with high MLSS concentration also produce dilutions with decreased filterability, regardless of the membrane configuration used to measure filterability. The results presented in Figure 6.23 seem to indicate that activated sludge with poor filtration quality, i.e. ∆R20 greater than 1x1012m-1 of added resistance, and high MLSS concentration also produces dilutions with decreased filterability. However, this reasoning is not correct. The relationship between ∆R20 and filtration quality was established for the DFCm, i.e. for a particular setup and operation procedure. The FSu has different setup and operational parameters. Therefore, the classification proposed by Geilvoet (2010) must be verified and most probably modified before being applied to the FSu results.

99

6.7 Comparison between Permeate, Tap Water and Conventional Activated Sludge Effluent as Dilution Media

The filterability according to the MLSS concentration in the activated sludge, for sets using permeate, tap water and CAS effluent as dilution media is presented in Figure 6.24, by set. The original samples are represented by larger dots. The arrows indicate the sequence of the diluting experiments.

Figure 6.24 - Filterability vs. MLSS concentration of MBR activated sludge.

The results in Figure 6.24 should be analyzed two by two, that is set 33 and set 34 and so forth, because the same original sludge was used to dilute with permeate and tap water or CAS effluent. The results from sets 33/34, presented in Figure 6.24, show that both permeate and tap water produce dilutions with increased filterability, when the MLSS concentration in the original sample is low. However, in set 34, prepared with tap water, the second dilution has similar filterability to the first, opposed to set 33, prepared with permeate, were dilution 2 has better filterability than dilution 1. The ionic strength of tap water is likely to interfere with the activated sludge. The results from sets 35/36 and sets 37/38 support the previous reasoning. CAS effluent is likely to have similar ionic strength to permeate, therefore both permeate and CAS effluent, when used to dilute original samples with high MLSS concentration, produces dilutions with decreased filterability. Thus, the results shown in Figure 6.24 demonstrate that permeate is likely to be the most suitable dilution media. However, a reduced number of experiments were performed therefore our results are not conclusive. Nevertheless, interference from the dilution media in the obtained filterability results is not expected. The results of αR*ci product, s coefficient and SMP concentrations provided no further conclusions and are therefore presented in Annex C and further discussed in section 6.9.

0,0

0,2

0,4

0,6

0,8

1,0

0 5 10 15 20

MLSS [g/L]

∆R

20[*

10

12m

-1]

Set 35(permeate)Set 36 (CASeffluent)Set 37(permeate)Set 38 (CASeffluent)Set 33(permeate)Set 34 (tapwater)

100

6.8 Comparison between Constant and Variable Cross-Flow Velocity

Figure 6.25 presents the filterability and MLSS concentration of samples with low MLSS concentration, i.e. in this research, equal to or less than 10 g/L, submitted to sludge filtration using either a constant cross-flow velocity of 1 m/s, or a variable cross-flow velocity. The arrows indicate the sequence of the diluting experiments. It is noteworthy to clarify that filterability results obtained with variable cross-flow velocity should not be compared within one set, because the results were not obtained under the same operating conditions. The cross-flow velocity was 2.5, 2.2 and 1.8 m/s for original, dilution 1 and dilution 2 samples, respectively, as indicated in Figure 6.25.

a) b)

Figure 6.25 - Filterability vs. MLSS concentration of sludge with low MLSS concentration in the original sample (larger dots), using constant and variable cross-

flow velocity.

Figure 6.25 shows that the hydraulic regime created in the tubular membrane is affecting our results. This situation is not unexpected. In cross flow-filtration it is understood that a turbulent regime is preferable to a laminar regime in order to achieve successful process operation (Rosenberger et al. 2002c). The results presented in Figure 6.25 show that filterability obtained under turbulent conditions, i.e. variable cross-flow velocity, is always better than filterability obtained under laminar conditions, i.e. constant cross-flow velocity. Therefore, our results confirm that a turbulent regime is preferable to a laminar regime for process operation of cross-flow filtration membranes. The results presented in Figure 6.25 also show that a turbulent regime is very effective in eliminating the removable fouling measured through the DFCm. MBR activated sludge with poor filtration quality, measured under laminar conditions, has much better filterability, when measured under turbulent conditions. However, Figure 6.25 also shows that the recovery of filterability, meaning a reduction in the ∆R20 value due to turbulent conditions, is lower if the filtration quality of the sludge is moderate, when measured under laminar conditions.

0,0

0,5

1,0

1,5

2,0

0 5 10 15

MLSS [g/L]

∆R

20[*

10

12m

-1]

Set 39(constant)

Set 40(variable)

2.2 m/s

1.8 m/s 2.5 m/s

0,0

0,5

1,0

1,5

2,0

0 5 10 15

MLSS [g/L]

∆R

20[*

10

12m

-1]

Set 43(constant)

Set 44(variable)

2.2 m/s

1.8 m/s 2.5 m/s

101

Figure 6.26 shows the filterability and MLSS concentration of MBR activated sludge samples, with high MLSS concentration in the original sample, i.e. in this research, exceeding 10.5 g/L, submitted to sludge filtration using either a constant cross-flow velocity or a variable cross-flow velocity. The variable cross-flow velocity was 2.5, 2.2 and 1.8 m/s for original, dilution 1 and dilution 2 samples, respectively, as indicated in Figure 6.26. The arrow indicates the sequence of the diluting experiments.

Figure 6.26- Filterability vs. MLSS concentration of sludge, with high MLSS

concentration in the original sample (larger dots), using constant and variable cross-flow velocity.

Figure 6.26 shows that the cross-flow velocity of 2.2 m/s, used to determine the filterability of the first dilution under turbulent conditions, is not enough to eliminate the decrease in filterability obtained in the first dilution sample of set 41. Furthermore, Figure 6.26 confirms that the recovery of filterability, meaning a reduction in the ∆R20 value due to turbulent conditions, is dependent on the filtration quality of the sludge, determined under laminar conditions. The results of αR*ci product and s coefficient provided no further conclusions and are therefore presented in Annex C and further discussed in section 6.9.

0,0

0,5

1,0

1,5

2,0

0 5 10 15 20

MLSS [g/L]

∆R

20[*

10

12m

-1]

Set 41(constant)

Set 42(variable)

1.8 m/s

2.2 m/s

2.5 m/s

102

6.9 General Discussion

Initially, we postulated that activated sludge with low MLSS concentration would present an increasing fouling propensity with increasing MLSS concentration. On the contrary, activated sludge with high MLSS concentration and moderate to good filtration quality would retain fouling particles in its matrix due to a particular structure, resulting in decreased fouling potential with increasing MLSS concentration. A discussion including the results obtained to characterize MBR activated sludge and the results obtained to evaluate our own results is presented next. Filterability

The results obtained in this study showed that MBR activated sludge with low MLSS concentration, in this research equal to or less than 10 g/L, and MBR activated sludge with high MLSS concentration, in this case exceeding 10.5 g/L, do not provide the same results when diluted. MBR activated sludge samples with low MLSS concentration provides dilutions with increased filterability, while sludge with high MLSS concentration results in dilutions with decreased filterability. The increase in filterability by diluting occurs in activated sludge samples with low MLSS concentration, regardless of the filtration quality of the original sludge. This conclusion is supported by the data obtained from the activated sludge of the Heenvliet, Ootmarsum and Varseeveld MBRs. In contrast, the decrease in filterability by diluting activated sludge samples with high MLSS concentration occurs when filtration quality is moderate to good, but not when the filtration quality of the sludge is poor. This conclusion is based on the results of the sets of activated sludge collected from the Heenvliet and Schilde MBRs. The aforementioned results show that the role of MLSS concentration in MBR filtration is dependent on the activated sludge filtration quality. In all the results obtained in this research, a linear relationship between αR*ci and ∆R20 parameter was obtained: a minimum R2 fitting of 0.96 was obtained in the sets aiming to compare activated sludge from different MBRs; a maximum R2 fitting of 0.99 was obtained in the sets aiming to compare results using constant and cross-flow velocity. The s coefficient results do not present a statistical correlation with the ∆R20 parameter. In this research, the s coefficient results mainly vary between 0 and 0.2, indicating that the produced cake layer is hardly compressible. These results show that the obtained ∆R20 values are mainly dependent on the αR*ci product. This conclusion is in line with the statements of Geilvoet (2010), particularly that the total cake layer resistance is mainly determined by the concentration of fouling particles in the cake layer. Figure 6.27 shows all the αR*ci and s coefficient values obtained in this research, according to the MLSS concentration of the samples. Results are identified by the MLSS concentration of the original sample in each set.

103

0

20

40

60

80

100

0 5 10 15 20

MLSS [g/L]

αR

*ci [

*10

-3m

2]

High MLSS

Low MLSS

0,0

0,1

0,2

0,3

0,4

0 5 10 15 20

MLSS [g/L]

s [

-]

High MLSS

Low MLSS

a) b) Figure 6.27 - αR*ci product (a) and s coefficient (b) vs. MLSS concentration.

The results shown in Figure 6.27 confirm that, in this research, there are differences in the cake layer produced when filtrating samples with high and low MLSS concentrations, as previously obtained in the sets intended to characterize MBR activated sludge quality. In this research, sludge with high MLSS concentration and moderate to good filtration quality produces a cake layer with lower αR*ci product, i.e. specific cake resistance by the concentration of fouling particles and lower s coefficient values. Average values of αR*ci and s coefficient for activated sludge with MLSS concentration above 10.5 g/L and sludge with MLSS concentration equal to or below 10 g/L are 18 [*10-3m2] and 0.06 opposite to 32 [*10-3m2] and 0.08, respectively. Therefore, considering that cake layer resistance is mainly determined by the concentration of fouling particles (Geilvoet 2010), the above results show that in activated sludge with high MLSS concentration and moderate to good filtration quality, the cake layer has a lower concentration of fouling particles and is less compressible, than in activated sludge with low MLSS concentration. Our explanation for the aforementioned results is related to the structure of the activated sludge, i.e. the ability or inability to retain fouling particles in the sludge matrix. Sludge with high MLSS concentrations and moderate to good filtration quality is probably able to retain fouling particles, contrary to sludge with low MLSS concentrations. Therefore, SMP concentration, viscosity, particle counting in the range of 0.4-5 µm and particle counting in the range of 2-100 µm were performed to analyze possible differences in the structure of activated sludge with low and high MLSS concentration. Soluble Microbial Products

Table VI-III presents the average values of protein and polysaccharide concentrations, in the bulk activated sludge, for samples with high and low MLSS concentration according to the purpose of the diluting experiments.

104

Table VI-III Average protein and polysaccharide concentrations of activated sludge with high and low MLSS concentration in diluting experiments.

Proteins Polysaccharides High MLSS Low MLSS High MLSS Low MLSS General purpose

[mg/L] [#] *

[mg/L] [#] *

[mg/L] [#] *

[mg/L] [#] *

To characterize MBR activated sludge

15.9 24 12 30 10.8 25 8.6 32

To compare activated sludge from MBR membrane and aeration tanks

22.3 5 15 11 15 5 14.7 11

To compare membrane tank activated sludge from different MBRs

9.4 1 14.4 11 8.5 1 9.8 11

To compare permeate, tap water and CAS effluent as dilution media

17.6 7 14.7 5 17.3 7 11.7 5

To compare results using constant and variable cross-flow velocity

28.8 2 17.6 6 26 2 23.6 6

Key: * Number of results used to calculate the average concentration.

Table VI-III shows that the SMP results from MBR activated sludge samples with high MLSS concentrations are higher than SMP results from activated sludge with low MLSS concentration. The only exception is obtained in the samples intended to compare activated sludge from different MBRs but, in these sets, only 1 sample with high MLSS concentration was measured, so the aforementioned result is not considered significant. Figure 6.28 represents filterability vs. protein and polysaccharide concentrations, in the bulk activated sludge and retained in the DFCi membrane, in all the samples measured in this research.

0,001

0,01

0,1

1

10

0 10 20 30 40

Proteins [mg/L]

∆R

20[x

10

12m

-1]

Bulk

Retained

0,001

0,01

0,1

1

10

0 10 20 30 40 50

Polysaccharides [mg/L]

∆R

20[x

10

12m

-1]

Bulk

Retained

a) b)

Figure 6.28 - Filterability vs. protein (a) and polysaccharide (b) concentrations in the bulk activated sludge and retained in the DFCi membrane.

105

Figure 6.28 shows that, as concluded for the sets intended to characterize MBR activated sludge, filterability and SMP results are not directly related. Table VI-IV represents statistical results, namely the Pearson coefficient and R2 fitting of several regressions, between the ∆R20 parameter and protein and polysaccharide concentrations in the bulk activated sludge and retained in the DFCi membrane.

Table VI-IV Statistical correlation results between filterability and protein or polysaccharide concentrations

Proteins Polysaccharides ∆R20 vs. Bulk Retained Bulk Retained

Pearson coefficient 0.20 0.20 0.16 0.15 1st order regression (R2) 0.04 0.04 0.03 0.02 2nd order regression (R2) 0.09 0.18 0.07 0.07 3rd order regression (R2) 0.14 0.19 0.22 0.22 4th order regression (R2) 0.16 0.19 0.22 0.22 Logarithmic regression (R2) 0.03 0.04 0.04 0.04 Exponential regression (R2) 0.03 0.04 0.04 0.04

As Table VI-IV shows, the highest values obtained are 0.2, for the Pearson coefficient, and 0.22, for the R2 fitting. Therefore, in this research, no direct correlation was found between the ∆R20 parameter and SMP results. However, on average, a direct relationship between filterability and SMP results is not obtained because the applied SMP methods are quantifying all the SMP present in the samples, including relevant and irrelevant SMP for membrane filtration. Several authors suggest that SMP concentration affects fouling only under low SRT conditions (Drews et al. 2008; Kimura et al. 2009). The value of low SRT varies from author to author, namely, for Drews et al. (2008), it is 8 days and for Kimura et al. (2009), it is 17 days. Nevertheless, the MBRs studied in this research apply SRT values higher than the aforementioned low SRT definitions (Table VI-II). Therefore, it is also possible that a link between filterability and SMP concentration is not obtained in our research because high SRT values are applied in the analyzed MBRs. Figure 6.29 represents the protein and polysaccharide concentrations obtained in the bulk activated sludge vs. the SMP concentrations retained in the DFCi membrane.

106

0

10

20

30

40

0 10 20 30 40

Proteins bulk [mg/L]

Pro

tein

s r

eta

ine

d [

mg

/L]

0

10

20

30

40

50

0 10 20 30 40 50

Polysaccharides bulk [mg/L]

Po

ly.

reta

ine

d [

mg

/L]

a) b)

Figure 6.29 – Protein (a) and polysaccharide (b) concentrations retained on the DFCi membrane vs. protein and polysaccharide concentrations in the bulk activated sludge.

Figure 6.29 shows that SMP of the bulk activated sludge is retained in the DFCi membrane. A linear regression between the aforementioned parameters provides a R2 fitting of 0.74 for proteins and 0.99 for polysaccharides. On average, 49% of the proteins are retained while 77% of the polysaccharides remain on the membrane. Once more it is noteworthy to point out, that the definition of retained SMP, in cross-flow operation mode, refers to SMP potentially retained in the membrane. Viscosity

The viscosity measurement results showed differences between MBR activated sludge with low MLSS concentrations and MBR activated sludge with high MLSS concentrations, as described earlier. A comparison between measured and estimated apparent viscosity values, calculated on the basis of models that define apparent viscosity through temperature, shear rate and MLSS concentration (Rosenberger et al. 2002c; Laera et al. 2007; Yang et al. 2009), showed that estimated viscosity is twice as accurate in samples with MLSS concentration equal to or lower than 10 g/L as in samples with MLSS concentration exceeding 10.5 g/L. In addition, the results obtained from the Ootmarsum and Varsseveld MBRs showed that apparent viscosity can have a strong statistical relationship with MLSS concentration and filterability, which occurs for poor quality activated sludge with low MLSS concentration. It can be concluded that other factors, besides temperature, shear rate and MLSS concentration, have a greater impact on viscosity when the samples have high MLSS concentration that when the activated sludge has low MLSS concentration. Rosenberger et al. (2002c) stated that mechanical stress, prior to measurement, and SMP concentrations affected the apparent viscosity results. In the sets intended to characterize MBR activated sludge, the sludge was collected at one MBR installation, the Heenvliet MBR, so differences in mechanical stress should not affect the obtained results. However, in the aforementioned sets of data, SMP concentrations in samples with high MLSS concentration are higher than in samples with low MLSS concentration. Therefore, SMP concentrations, or the structure of the sludge, may be affecting the viscosity results. If so,

107

and taking into account our previous conclusions, SMP concentrations or sludge structure may have a stronger impact on apparent viscosity in samples with high MLSS concentration than in samples with low MLSS concentration. Itonaga et al. (2004) identified 10 g/L as a critical MLSS concentration, above which viscosity increases sharply. However, in the aforementioned paper, there is no reference to the applied shear rate to measure viscosity. Other authors reported critical MLSS values concerning the effect on viscosity, namely 14 g/L for an applied shear stress of 5s1 (Reid et al. 2008). Figure 6.10 shows that the apparent viscosity measured for shear rates of 5 s-1 shows differences between samples that are not sustainable at shear rates of 50 s-1. In this research, the viscosity results were mainly analyzed at a shear rate of 100 s-1. Furthermore, a critical MLSS concentration above which viscosity would increase sharply was not obtained. Particle counting in the range of 0.4-5 µm

The particle counting results in the range of 0.5-1 µm showed that the number of particles decreases when the samples are diluted. The results are consistent by set, meaning that original samples have more particles than dilution 1 samples and so forth. A reduction in the number of particles due to the dilution operation itself was expected. It occurs in sets belonging to both “Low MLSS” and “High MLSS” groups. The normalization of the particle counting data in the range of 0.4-5 µm showed a fundamental difference between the “Low MLSS” and “High MLSS” sets, namely, the value of β, i.e. the normalized distribution slope, increases from Original to Dilution samples in “High MLSS” sets, as opposed to “Low MLSS” sets. Table VI-V shows the average values of β obtained in the “Low MLSS” and “High MLSS” sets, according to the purpose of the diluting experiments and per type of sample. Table VI-V Average β values in “Low MLSS” and “High MLSS” sets, according to the

purpose of the diluting experiments and per type of sample. High MLSS Low MLSS β value [#] * β value [#] *

To characterize MBR activated sludge Original 8.4 9 13.7 1 Dilution 1 9.1 9 12.5 1 Dilution 2 12.8 9 10.1 1 To compare activated sludge from MBR membrane and aeration tanks Original 10.6 2 9.5 1 Dilution 1 10.5 2 9.3 1 Dilution 2 10.7 2 9.2 1 To compare membrane tank activated sludge from different MBRs Original - 10.5 3 Dilution 1 - 10.5 3 Dilution 2 - 9.3 3

Key: * Number of results used to calculate the average value.

A reduction of β from original to dilution samples is a consequence of the reduction in the number of particles due to the dilution operation itself. As Table VI-V shows, the aforementioned result is clearly obtained in the “Low MLSS” sets. However, an

108

increased value of β, from Original to Dilution samples, can be an indication of a deflocculation process leading to higher particle counts, smaller than the analyzed particles size range. Therefore, deflocculation may occur in particles smaller than 0.5 µm. As mentioned earlier, a link between filterability and submicron particles was identified (Leiknes et al. 2006; Geilvoet et al. 2007a). Worse filterability was related to a higher amount of submicron particles. Furthermore, the volume of submicron particles, defined as the accumulated volume of particles in the range of 0.4-1 µm, was proposed as a suitable indicator of activated sludge filterability (Geilvoet 2010). It was therefore excepted that, since samples with high MLSS concentration provide dilutions with worse filterability, dilution samples from the “High MLSS” group would contain a higher amount of submicron particles than original samples. However, the obtained results do not confirm this hypothesis. The volume of submicron particles, as defined by Geilvoet (2010), in dilution samples of the “High MLSS” group is lower than in original samples. For “High MLSS” sets, a decrease of about 40% in the volume of submicron particles due to the dilution operation itself can be predicted. However, in sets 1 to 22, the measured decrease was of about 20%. The difference between predicted and measured results is probably due to the particles in the permeate used to dilute the original samples. In principle, it would be expected that permeate from UF membranes would not contain particles in the range of 0.4-5 µm. However, our own research, described in Chapter 8, shows that there are always particles between 3.5 and 35 µm in the permeate of UF membranes. Therefore, it can be predicted that the permeate may contain about 100,000 particles/mL in the range of 0.4-5 µm during stable MBR operation and, in that case, the difference between predicted and measured volume of submicron particles is probably due to permeate particles. Ivanovic et al. (2008) concluded that the relevant particle size range for membrane filtration is below 0.5 µm, in particular below 0.1 µm. Geilvoet (2010) stresses the importance of particles smaller than 0.5 µm to determine the volume of the cake layer. Our own results indicate that a shift in the particle size distribution, capable of demonstrating deflocculation, is likely to occur for size ranges below 0.5 µm. Therefore, it is likely that in this research, deflocculation was not measured in the dilution samples of the “High MLSS” group because the relevant submicron size range is below 0.5 µm. Particle counting in the range of 2-100 µm

The particle counting results in the range of 2-100 µm showed dilution and deflocculation effects in the sets of both “Low MLSS” and “High MLSS” groups. The only exception was the set with sludge from the Schilde MBR, but at the time of collection, an unusual growth of filamentous bacteria was reported (Moreau 2010). It can be concluded that, through particle counting in the range of 2-100 µm, no difference was detected between the sets of the “Low MLSS” and “High MLSS” groups.

109

Effect of membrane configuration and dilution media

Our results showed that the filterability results for the “High MLSS” group are independent of the membrane configuration and are likely to be independent from the dilution media. A decrease in the filterability of dilution samples in the “High MLSS” group was obtained in a flat sheet membrane installation, similar to that obtained in the tubular membrane installation used in this research. The sets prepared with permeate, tap water and CAS effluent as dilution media demonstrated that diluting with media of ionic strength similar to the original samples provides the most logical results. However, a reduced number of experiments were performed therefore the concluions concerning the dilution media are not totally conclusive. Effect of the hydraulic regime

The absolute values of filterability obtained in this research are dependent on the hydraulic regime defined in the tubular membrane used in the DFCi. In the DFCm, activated sludge samples are measured in laminar regime, if the MLSS concentration in the sample exceeds 5 g/L (Geilvoet 2010). Our results show that the absolute values of filterability, represented by ∆R20, measured in laminar regime are higher than the values obtained in turbulent regime. These results were expected, since a turbulent regime is understood to be preferable, for process operation in cross-flow filtration membranes, than a laminar regime (Rosenberger et al. 2002c). Nevertheless, even when filterability is measured under turbulent conditions, the extra shear provided does not eliminate completely the added resistance of “High MLSS” dilutions.

6.10 Conclusions

Our initial hypothesis was that, due to structural differences, activated sludge with moderate to good filtration quality and high MLSS concentration, i.e. exceeding a critical MLSS concentration, would retain fouling particles in its matrix, as opposed to activated sludge with low MLSS concentration. In this research, the critical MLSS concentration was found to be 10.5 g/L, so two groups were defined: the “Low MLSS” group for sets with MLSS concentration in the original sample equal to or lower than 10 g/L; the “High MLSS” group for sets with MLSS concentration in the original sample exceeding 10.5 g/L. Consequently, references to low and high MLSS concentration always consider 10.5 g/L as the critical MLSS concentration. The conclusions obtained in this research can be summarized as follows:

- The “High MLSS" group provides dilutions with decreased filterability as opposed to the “Low MLSS” group. The “Low MLSS” group provides dilutions with increased filterability.

- Filterability results from the “Low MLSS” group are independent of the filtration quality of the activated sludge. In the “High MLSS” group, dilutions with decreased filterability are obtained only when the sludge has moderate to good filtration quality.

110

- When filtered according to the DFCm, activated sludge with good to moderate quality and high MLSS concentration originates cake layers with lower specific cake resistance by the concentration of fouling particles and lower compressibility than activated sludge with low MLSS concentration.

- In this research, a direct relationship between filterability and SMP was not obtained. Furthermore, the applied SMP methods are quantifying SMP relevant and irrelevant for membrane filtration.

- SMP results of sludge samples with high MLSS concentration are slightly higher than those with low MLSS concentration.

- There are differences in the viscosity of MBR activated sludge with low and high MLSS concentration. Viscosity of sludge samples with low MLSS concentration is basically determined by temperature, shear rate and MLSS concentration. These factors do not accuratly predict the viscosity of samples with high MLSS concentration. The viscosity of samples with high MLSS concentration is likely to be affected by SMP concentration or sludge structure.

- A dilution effect was observed, in the range of 0.5-1 µm, in all diluted samples, as expected. A deflocculation effect, while diluting high MLSS concentration sludge that could confirm our hypothesis, was not obtained in the range of 0.5-1 µm. However, the variability of β values does indicate that deflocculation is likely to occur in particles smaller than 0.5 µm.

- Dilution and deflocculation are visible in the particle size range of 2-100 µm in both “High MLSS” and “Low MLSS” groups. No differences were visible, in the particle size range of 2-100 µm, between the two groups.

- The decrease in filterability, obtained when diluting high MLSS concentration sludge, is independent of membrane configuration.

- The absolute numbers of filterability presented in this research are dependent on the hydraulic regime. However, the extra shear provided by a turbulent regime does not eliminate completely the added resistance of “High MLSS” dilutions.

Apparently, our filterability results show that the effect of MLSS concentration on membrane filtration depends on the MLSS concentration range and, when the MLSS concentration exceeds a certain critical value, on the activated sludge filtration quality. The specific cake resistance by the concentration of fouling particles and compressibility of the cake layer, SMP and viscosity results showed structural differences between sludge with MLSS concentration below the critical value and above the critical value. The retention of particles in membrane tank activated sludge, with high MLSS concentration and moderate to good filtration quality, could not be measured, most probably because the retained particles are smaller than 0.5 µm.

111

7. Concentrating Membrane Bioreactor Activated Sludge

7.1 Introduction

The main goal of this thesis is to analyze the role of MLSS concentration in MBR filtration. This Chapter describes the experiments intended to manipulate MBR activated sludge, in particular the MLSS concentration, through a concentrating operation. The contradictory results provided by research concerning the impact of MLSS concentration on membrane filtration were described in Chapter 2. In particular, a decrease in fouling, for MLSS concentration between 10 and 20 g/L, coinciding with a decrease in EPS and SMP concentrations was reported (Judd 2005). In addition, Moreau (2010) described experiments in an MBR pilot where significant improvements in filterability occurred which could be clearly linked to an increase in MLSS concentration in the activated sludge. Therefore, there are indications that, in specific membrane configurations and membrane operation circumstances, an increase in MLSS concentration can lead to filterability improvement. Our own results, described in Chapter 6, indicated that the effect of MLSS concentrations on membrane filtration depended on the applied MLSS concentration range and, for activated sludge with MLSS concentration above a certain critical value, on the activated sludge quality. Due to a particular activated sludge structure, membrane tank activated sludge with high MLSS concentration is likely to entrap particles in its matrix when the filtration quality is moderate to good, as opposed to membrane tank activated sludge with low MLSS concentration. In this research, low and high MLSS concentrations are defined as below and above 10.5 g/L, respectively. In the research described in Chapter 6, the aforementioned hypothesis was evaluated by testing activated sludge with decreasing MLSS concentrations produced by dilutions. In this Chapter the same hypothesis is evaluated by testing activated sludge with increasing MLSS concentrations, produced by a concentrating operation. MBR activated sludge samples with increasing MLSS concentrations were obtained by increasing the “permeate flow/return flow” ratio of a membrane tank in the full-scale MBR of Heenvliet. During these experiments, further referred to as concentrating experiments, the aforementioned MBR was operated in series and in parallel with a CAS system. Further explanations about this configuration operation are provided in section 7.3. In the concentrating experiments performed when the MBR was operated in parallel, the activated sludge samples were collected at the restart of permeate extraction and aeration in the respective membrane tank. Collecting sludge samples upon the restart of the membrane tank allowed us to evaluate the quality of activated sludge with increasing MLSS concentrations, without disrupting the operation of the full-scale MBR. However, at the restart of the membrane tank, the activated sludge quality may also change due to the restart of aeration. The restart of aeration will improve the mixing of the tank contents, change the activated sludge particle size and provide DO for the conversion of carbonaceous material.

112

The effect of DO in MBR activated sludge filterability was studied by Geilvoet (2010). The author showed that MBR sludge filterability would decrease without aeration. An increase in the ∆R20 value from 0.05 to 3.3 x1012m-1 was measured, when the activated sludge was kept without DO for 4 days. However, the activated sludge showed a recovery rate 12 times faster than the decay rate. After a period of 6 hours of aeration, the activated sludge presented a ∆R20 of 0.7 x1012m-1. Abbassi et al. (1999) evaluated the effect of DO and mixing intensity to minimize sludge production in CAS systems. According to the latter authors, DO concentrations from 2 to 6 mg/L and mixing reduce the floc size, through floc breakage, by 10% and 12%, respectively. In an MBR system, Ji et al. (2006) provided similar conclusions: the increase in aeration lead to floc breakage. Therefore, it can be predicted that the restart of aeration in a membrane tank will provide a fast increase in filterability and reduce the activated sludge particle size by floc breakage. Moreover, our results described in Chapter 5 showed that, smaller activated sludge flocs do not necessarly lead to worse filterability. In this research, MBR activated sludge samples with increasing MLSS concentrations were collected from a full-scale membrane tank. When the MBR was operated in parallel with a CAS system, the sludge samples were collected at the restart of permeate extraction and aeration, after a maximum period of 3 days without operation. The collected samples were analyzed for filterability, MLSS concentration, SMP concentrations, viscosity, particle size in the range 2-100 µm and particle size in the range 0.4-5 µm. In addition, the last MBR sludge sample, collected on each day of the concentrating experiments, was also submitted to diluting experiments.

7.2 Methodology

The activated sludge was collected from the membrane tanks of the Heenvliet MBR (Waterboard Hollandse Delta). The characteristics of the WWTP are described in Table V-I. In particular, the WWTP is comprised of a CAS and a MBR system. The concentrating experiments took place during 2009 and the WWTP switched from series to parallel operation in March 2009. To avoid filterability variation due to lack of biomass acclimation, the concentrating experiences were performed before the switch, January 2009, and repeated 5 months after the switch, July 2009. In both series and parallel operation of the WWTP, activated sludge samples with increasing MLSS concentrations were obtained by increasing the “permeate flow/return flow” ratio of the membrane tank. In the concentrating experiments performed during series operation, the sludge samples were collected from a continuously operated membrane tank. In the concentrating experiments performed during parallel operation, the sludge samples were collected at the restart of the membrane tank. Background information on the operation and specific activated sludge characteristics of the Heenvliet MBR is presented in section 7.3. Each concentrating experiment consisted of one morning of measurements on site, followed by one afternoon of analyses in the TU Delft water lab. The Heenvliet MBR is

113

located at a driving distance of about 45 min from the TU Delft water lab. In total, 10 concentrating experiments were performed on the following days: 20-1-2009; 22-1-09; 1-7-2009; 3-7-2009; 20-7-2009; 23-7-2009; 27-10-2009; 29-10-2009; 19-11-2009 and 1-12-2009. A total of 79 sludge samples were analyzed. During one concentrating experiment, several sludge samples were collected. Each sludge sample was divided into several smaller samples to determine the following parameters:

- Filterability: performed on site according to the DFCm (Evenblij et al. 2005); - MLSS concentration: performed on site; - Particle size distribution in the range 2-100 µm: performed on site; - Viscosity: samples were transported in a cooling box and measured at the TU

Delft water lab on the same day they were collected; - Particle size distribution in the range 0.4-5 µm: samples were filtered on site,

transported in a cooling box, and measured in the TU Delft water lab on the same day they were collected;

- SMP concentrations, namely, protein and polysaccharide concentrations: samples were filtered on site, transported in a cooling box, kept in the TU Delft water lab at below 4ºC, and measured within two days after collection.

Starting from the concentrating experiments performed on 1-7-2009 and following, the last sludge sample, collected each day, was also submitted to diluting experiments, following the methodology described in Chapter 6. The dilution samples were then submitted to the same analyses as the samples collected during the concentrating experiments. Details of the materials and methods used in this research are provided in Chapter 4. Furthermore, viscosity and particle counting in the range of 0.4-5 µm, in the concentrating experiments performed during January, namely on 20-1-2009 and 22-1-09, will not be discussed because the sample preparation methodology was not optimized.

7.3 Background Information about the Heenvliet Membrane Bioreactor

The concentrating effect at the Heenvliet MBR

The Heenvliet WWTP originally consisted of a carrousel type CAS system with a population equivalents capacity of 9,000. The Heenvliet MBR, which started operation in 2006, increased the plant’s population equivalents capacity by 4,000. The MBR was designed to treat 100 m3/h, representing 25 % of the maximum hydraulic capacity. The Heenvliet MBR is also referred to as a hybrid MBR because it can be operated in series or in parallel with the CAS system, according to the amount of influent to be treated (Mulder et al. 2005). The complete layout of the Heenvliet WWTP and, in particular, of the Heenvliet MBR, is presented in Figure 7.1.

114

Figure 7.1- The schematics of Heenvliet WWTP.

Figure 7.1 shows that the Heenvliet MBR has two separate membrane tanks plus several other tanks intended for carbonaceous and nutrient removal. The function of the equalization tank, designated to as VT in Figure 7.1, is to provide extra aeration to the activated sludge before reaching the membrane tanks, so it will be also referred to as aeration tank. At the Heenvliet MBR, due to the extraction of permeate and a low return ratio (Table V-I), the MLSS concentration in the operating membrane tanks is higher than in the other tanks of the MBR. A comparison of the activated sludge quality in the aeration tank and membrane tanks, in terms of MLSS concentration and filterability, is presented in Figure 7.2. The sludge samples were collected simultaneously, in both the aeration tank and the membranes tanks, to avoid possible daily and seasonal variations.

0,0

0,4

0,8

1,2

1,6

2,0

AT

MT

AT

MT

AT

MT

∆R

20[x

10

12m

-1]

0

5

10

15

20

MLS

S (g/L

)

10-8-08 21-1-09 23-1-09

Figure 7.2- Filterability (columns) and MLSS concentration (dots) in MBR activated

sludge collected from the aeration tank (AT) and membrane tank (MT) of the Heenvliet MBR.

DT

115

The results presented in Figure 7.2 clearly show that the filterability is always better in the membrane tanks, coupled with higher MLSS concentrations, than in the aeration tank. Figure 7.1 shows that the aeration tank, referred to as VT, is immediately before the membrane tanks, following the circulation of the activated sludge flow. The results presented in Figure 7.2 show that a concentrating effect, resulting in activated sludge with higher MLSS concentration and improved filterability, does occur in the membrane tanks of Heenvliet MBR. Membrane tank operation mode

The Heenvliet MBR has two membrane tanks and, in April 2009, the membrane tanks started being operated one at a time in order to reduce energy consumption. The switch from one operating membrane tank to the other occurred no longer than every three days. The layout represented in Figure 7.3 illustrates this operational mode.

Figure 7.3- Layout of the membrane tank operation at Heenvliet MBR.

The upper part of Figure 7.3 illustrates the simultaneous operation of the two membrane tanks. Activated sludge is circulated from the activated sludge tanks to membrane tank 1 and to membrane tank 2, in parallel, returning to the activated sludge tanks. In this operation mode, aeration and permeate extraction occurs in both membrane tanks. The lower part of Figure 7.3 illustrates the operation of one membrane tank: membrane tank 2. In this operation mode, activated sludge is still circulated from the activated sludge tanks to membrane tank 1, returning to the activated sludge tanks, but aeration and permeate extraction only occurs in one membrane tank: membrane tank 2.

116

Series and parallel configuration operation

During the first three years of operation the Heenvliet MBR was operated in series configuration. In March 2009, the Heenvliet MBR, switched from series to parallel operation. Figure 7.4 illustrates the series and parallel configuration operation at the Heenvliet MBR.

At dry weather flow (dwf) At storm weather flow (swf)

CAS

MBR MBR

MBR

Para

llel

co

nfi

gu

rati

on

Seri

es c

on

fig

ura

tio

n

CAS

CAS CAS

MBR

dwf

(1-y) dwf

y dwf

swf

~1/3 swf

~2/3 swf

swf

~1/3 swf

~2/3 swf

dwf

dwf

Figure 7.4- The schematics of the Heenvliet MBR configuration operation (Witteveen

&Boss) Details of the operational characteristics, of the abovementioned configuration modes, and of effects on the quality of the MBR activated sludge are discussed below. The switch from series to parallel configuration has consequences in terms of MLSS concentrations ranges obtained in the MBR tanks. During series operation, a maximum MLSS concentration of 18 g/L was measured in the membrane tanks, while during parallel operation, the maximum MLSS concentration obtained in the membrane tanks was 12.5 g/L. During series operation, the MBR influent from the CAS system has an average MLSS concentration of 4 g/L. During parallel operation, the MBR influent is raw wastewater thus with much lower MLSS concentrations. In practice, the capacity of the installed MBR pumps defines the lower MLSS concentration range obtained during parallel configuration mode. Figure 7.5 shows the permeability and flux in membrane tank 1 during series and parallel configuration mode. Similar results (not shown) were found at membrane tank 2.

117

100

200

300

400

500

600

700

800

900

1000

1-1

-20

08

10

-4-2

00

8

19

-7-2

00

8

27

-10

-20

08

4-2

-20

09

15

-5-2

00

9

23

-8-2

00

9

1-1

2-2

00

9

Pe

rme

ab

ility

[L

/m2.h

.ba

r]

0

5

10

15

20

25

30

35

40

Flu

x [L

/m2.h

]

Permeability MT1

Flux MT1

Series Parallel

Figure 7.5- Permeability and flux variations in membrane tank 1 (MT1) of the Heenvliet

MBR. The permeability results, shown in Figure 7.5, reflect operational choices, explained as follows. During series operation, the aim was to explore the limits of MBR technology. The flux was mainly set at 20 L/m2.h or higher. After a chemical cleaning, permeability values would, initially, increase sharply and then decrease until reaching values below 200 L/m2.h.bar. When the parallel configuration was applied, the main aim was to reduce energy consumption. The production of permeate was reduced by more than half (not shown) and the remaining load of the WWTP was sent to the CAS system. The flux was set around 10 L/m2.h. Consequently, permeability increased up to 400 L/m2.h.bar. A few days after the switch, permeability started a decreasing trend, more pronounced whenever the flux was increased. Permeability results are mainly dependent on the applied flux and membrane cleaning frequency and do not provide direct information about the MBR activated sludge quality. The filterability and temperature of the membrane tank activated sludge in the Heenvliet MBR, during series and parallel configuration, are presented in Figure 7.6.

0,0

0,2

0,4

0,6

0,8

1,0

1,2

1-1

-2008

30-3

-2008

22-6

-2008

20-9

-2008

19-1

2-2

008

15-3

-2009

11-6

-2009

7-9

-2009

2-1

2-2

009

∆R

20[*

10

12m

-1]

0

5

10

15

20

25

Tem

pera

ture

[oC

]

Filterability

Temperature

Series Parallel

Figure 7.6 - Filterability and temperature variations in the activated sludge from the

Heenvliet MBR.

118

Through Figure 7.6 it is possible to observe that the switch from series to parallel configuration coincides with a deterioration of filterability. In January 2009, the activated sludge quality was moderate, while in March 2009, the activated sludge quality was poor. In March 2009, the temperature was increasing, but filterability was decreasing instead of improving. The biomass characteristics can explain these results as follows. In series configuration, the raw wastewater is first submitted to the biological treatment provided by the carrousel. In the parallel configuration, the biological treatment occurs only in the MBR. Therefore, it is likely that, after switching from series to parallel operation, MBR biomass needed to readapt. Activated sludge needs to be acclimated before some sort of steady state is reached (Cicek et al. 1998; Le-Clech et al. 2003). After one month of parallel operation, the activated sludge quality had improved and presented a moderate quality. Figure 7.7 shows the filterability vs. temperature of the membrane tank activated sludge during series and parallel configuration.

0,0

0,4

0,8

1,2

5 10 15 20 25Temperature [

oC]

∆R

20 [*1

012m

-1]

Series

Parallel

Figure 7.7- Filterability vs. temperature of the membrane tank activated sludge during

series and parallel operation of the Heenvliet MBR. Figure 7.7 clearly shows that the filterability results obtained during parallel configuration are worse that during series operation. The worse filterability results, obtained during parallel operation, provide from samples collected during the two first months of parallel configuration. During this period is likely that the activated sludge was still adapting to the new conditions. Nevertheless, the filterability results obtained at temperatures between 12 and 15ºC are clearly worse during parallel configuration. These results indicate that the biological conditions, in particular a lower MLSS concentration range coupled with a biological treatment restricted to the MBR system, imposed during parallel operation, might be detrimental to activated sludge filterability. Further discussion on this subject will be provided in section 7.4.

119

7.4 Results and Discussion

Filterability results As mentioned earlier, the activated sludge samples collected while the MBR was operated in series with the CAS system came from membrane tanks in continuous operation. However, the sludge samples, when the MBR was operated in parallel, were collected from a restarting membrane tank. Therefore, the activated sludge quality will be altered not only by the increasing extraction of permeate, but also by the restart of aeration. According to our hypothesis, the extraction of permeate, which in the Heenvliet MBR membrane tanks brings about an increase in MLSS concentration due to the low return rate, may have a positive or negative effect on filterability, depending on the MLSS concentration range. In contrast, the impact of aeration on membrane tank filterability is always positive, as explained below. The restart of aeration provides mixing of the membrane tank contents and DO to DO-depleted activated sludge. As mentioned earlier, Geilvoet (2010) showed that the filterability recovery rate, after a period of 4 days without aeration, is 12 times higher than the decay rate. After a period of 6 hours of aeration, the ∆R20 had decreased from 3.3 to 0.7 x1012m-1. In the reported experiments, mixing was provided when aeration was absent. It can be concluded that DO improves filterability and that the improvement is significantly faster than the decay, caused by the absence of aeration. At the Heenvliet MBR, the membrane tanks, when aeration is on, are designed as completely mixed tanks. Figure 7.8 represents the tracer concentration curves, in an ideal completely mixed reactor subjected to a tracer slug input.

0

0,2

0,4

0,6

0,8

1

1,2

0 0,5 1 1,5 2 2,5 3

HRT [#]

Norm

. concentratio

n [C

/Ci]

Figure 7.8- Effluent tracer concentration curve in an ideal completely mixed reactor.

Figure 7.8 shows that, after 1 HRT, the tracer concentration will be 0.37 from the initial tracer concentration injected into the reactor, meaning that 63% of the reactor’s contents will have been replaced by fresh sludge.

120

In practice, completely mixed tanks are non-ideal reactors. However, the sludge of a non-operating membrane tank, at the Heenvliet MBR, is not completely stagnant or depleted of oxygen. The MBR has an aeration tank just before the membrane tanks and the flow, containing aerobic activated sludge, circulated through the non operating membrane tank. Therefore, mixing is provided through flow circulation and DO is transported to the membrane tanks even during the non operating phase. Therefore, in this research, we assumed that after 1 HRT, the sludge in the restarting membrane tank would have been completely replaced by aerobic sludge. Consequently, it was assumed that during the first HRT of the restarting membrane tank, improvements in filterability would be mainly caused by aeration. After one HRT has passed, improvements in filterability due to aeration are still possible but less significant. The effective hydraulic retention time in the Heenvliet membrane tanks, on the days the concentration experiments, is presented in Table VII-I. Table VII-I also shows the return ratio, from the membrane tanks to the nitrification tanks, applied on those same days.

Table VII-I Hydraulic Retention Time and Return Ratio, in the Heenvliet MBR membrane tanks, on the days of the concentrating experiments

Return Ratio Concentrating Experiment

[day-month-year]

Configuration operation mode

Hydraulic Retention Time

[min]

Qr/Q [-]

X/Xr-X [-]

20-1-09 Series 20 1.7 1.8 22-1-09 Series 22 1.9 2.1 1-7-09 Parallel 39 5.8 3.7* 3-7-09 Parallel 38 5.6 3.1*

20-7-09 Parallel 36 4.9 2.3* 23-7-09 Parallel 33 4.3 1.7*

27-10-09 Parallel 71 2.1 2.5 29-10-09 Parallel 72 2.4 2.4 19-11-09 Parallel 71 2.2 2.8 1-12-09 Parallel 38 3.9 4.2

Key: Qr- Return flow [m3/d]; Q- Inflow MBR [m3/d]; X- MLSS concentration in the carbonaceous and nutrient removal tanks [g/L]; Xr- MLSS concentration in the return flow [g/L]; * X values obtained through an online sensor, which did not provide reliable results during this particular period. In this research, the daily variation of MLSS concentration is directly affected by the return ratio. A higher return ratio reduces the increase in MLSS concentration in the membrane tanks and vice versa (Equation 1.2). Therefore, the return ratio has a direct impact on the concentrating of the membrane tank sludge. Figures 7.9 and 7.10 present the membrane tank filterability and MLSS concentration in the membrane tank and aeration tank of Heenvliet MBR, on the days of the concentrating experiments. Results are identified by date and time of measurement. The filterability results not linked to the others were obtained during the first HRT of the restarting membrane tank.

121

0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,49:3

6

10:1

5

10:4

7

11:2

4

11:5

5

9:5

4

10:2

1

10:4

8

11:1

5

11:4

6

12:1

8

12:5

0

14:5

4

8:1

2

9:5

3

10:5

3

11:3

8

13:0

6

7:5

4

8:2

6

9:0

1

9:3

3

10:3

2

11:1

0

8:1

9

9:1

1

9:4

1

10:1

1

10:3

9

11:0

8

Time [h:min]

∆R

20 [x

10

12m

-1]

0

5

10

15

20

MLS

S [g/L

]

Filterability

MLSS MT

MLSS AT

20-1-09 22-1-09 1-7-09 3-7-09 20-7-09

series series parallel parallel parallel

Figure 7.9- Membrane tank filterability and MLSS concentration in the membrane tank

(MT) and aeration tank (AT), on the days of the concentrating experiments.

0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

8:2

9

9:0

6

9:4

5

10:1

0

10:3

9

11:1

4

9:2

5

10:0

6

10:4

1

11:3

3

12:1

1

9:2

9

10:0

0

10:2

2

11:3

2

12:0

9

9:5

0

10:1

2

10:3

9

11:2

9

11:5

4

9:4

0

10:0

9

10:4

0

11:0

9

12:2

0

Time [h:min]

∆R

20 [x

10

12m

-1]

0

5

10

15

20

MLS

S [g/L

]

Filterability

MLSS MT

MLSS AT

23-7-09 27-10-09 29-10-09 19-11-09 1-12-09

Figure 7.10- Membrane tank filterability and MLSS concentration in the membrane tank

(MT) and aeration tank (AT), on the days of the concentrating experiments. Figure 7.9 and Figure 7.10 show that improvements in filterability coincide with increasing MLSS concentrations in the membrane tanks. The initial activated sludge quality, represented by the first filterability result on each day, is between moderate and poor. In the concentrating experiments performed during series configuration, i.e. on 20-1-09 and 22-1-09, the improvement of filterability with increasing MLSS concentration in the membrane tank is clear. In the experiments performed during parallel configuration, during the first HRT, represented by the non-linked filterability results, filterability always improves, most probably due to the restart of aeration. In the absence of DO, worse filterability was measured (Geilvoet 2010). Nevertheless, after one HRT is completed, represented by the linked filterability results, filterability continues to improve. Simultaneously, MLSS concentrations in the membrane tanks are increasing.

122

The magnitude of the filterability improvement varies according to the concentrating experiment, but there is a clear relationship with the measured MLSS concentration in the membrane tank. In the results obtained during series operation, filterability improved on average 68% and the MLSS concentration in the membrane tanks was clearly above 10.5 g/L. In the concentrating results obtained during parallel operation, filterability improved on average 21% and the MLSS concentration was lower than during series configuration. Furthermore, Table VII-I shows that the return ratio was higher than in the experiments performed during series operation, indicating that the concentration effect was not so evident. Figure 7.11 and 7.12 present the measured ∆R20 values according to the MLSS concentration in the membrane tank sludge. Figure 7.11 presents the results of the concentrating experiments performed while the WWTP was operating in series configuration, i.e. when the MLSS concentration range in the MBR was higher. Figure 7.12 presents the results obtained during parallel configuration, i.e. when the MLSS concentration range was lower. In Figure 7.12, the results obtained during the first HRT of the restarting membrane tank are represented by non-linked white dots. Each concentrating experiment is represented by its own symbol. The results presented in Figure 7.11 clearly indicate a trend towards better filterability in samples with higher MLSS concentrations.

0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

14 15 16 17 18 19

MLSS[g/L]

∆R

20 [*

10

12m

-1]

20-Jan-09

22-Jan-09

Figure 7.11- Filterability and MLSS concentration of the membrane tank activated sludge

during series configuration.

123

0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

6 8 10 12 14

MLSS [g/L]

∆R

20 [*

10

12m

-1]

1-Jul-09 HRT<1

1-Jul-09 HRT>1

3-Jul-09 HRT<1

3-Jul-09 HRT>1

20-Jul-09 HRT<1

20-Jul-09 HRT>1

23-Jul-09 HRT<1

23-Jul-09 HRT>1

27-Oct-09 HRT<1

27-Oct-09 HRT>1

29-Oct-09 HRT<1

29-Oct-09 HRT>1

19-Nov-09 HRT<1

19-Nov-09 HRT>1

1-Dec-09 HRT<1

1-Dec-09 HRT>1

Figure 7.12- Filterability and MLSS concentration of the membrane tank activated sludge

during parallel configuration.

Figure 7.12 shows that the concentrating experiments performed during parallel configuration do not provide results as clear as those obtained during series configuration. There is a trend towards better filterability with increasing MLSS concentration in samples with MLSS concentration above 10.5 g/L. The opposite trend, i.e. better filterability with decreasing MLSS concentration, is observed in part of the samples with MLSS concentration equal to or lower than 10 g/L, but the results are not completely clear. It is noteworthy to point out that in each concentration experiment performed during parallel operation, the tested range of MLSS concentration, after the first HRT is completed, is small. The return ratio during the parallel configuration period is higher than during series configuration, thus reducing decreasing the concentrating of the sludge. Furthermore, extra stress caused by the on/off operation of the membrane tanks, applied during parallel configuration, may also contribute to deteriorate the membrane tank activated sludge. The ∆R20 results according to the corresponding αR*ci and s coefficients, obtained in all the concentrating experiments, are represented in Figure 7.13.

124

Figure 7.13- ∆R20 vs. αR*ci and s coefficient in the concentrating experiments.

Figure 7.13 shows that the added resistance, represented by the ∆R20 value, is linearly related to the αR*ci product, i.e. the product of the specific cake resistance by the concentration of fouling particles. A linear regression between the two aforementioned parameters provides a R2 fitting of 0.95. In contrast, the values of the s coefficient, i.e. the compressibility of the cake layer, are very variable. Generally, the s values vary from 0 to 0.2, indicating a low compressibility of the cake layer. It can be concluded that the αR*ci product coefficient has the major influence on the obtained value of added resistance.

Soluble Microbial products results

In this research, the SMP results in the bulk sludge for samples with MLSS concentration above 10.5 g/L are higher than for samples with MLSS concentration equal or below 10 g/L. The average proteins and polysaccharides concentration for samples with MLSS concentration above 10.5 g/L are 19 and 10.6 mg/L, respectively. For samples with MLSS concentration equal to or lower than 10 g/L, the average proteins and polysaccharides concentration is 14.5 and 3.6 mg/L, respectively. Figure 7.14 represents the filterability vs. concentration of proteins (a) and polysaccharides (b) in the bulk activated sludge and retained in the DFCi membrane.

0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

0 20 40 60 80

αR*ci [*10-3

m2]

∆R

20 [*1

012m

-1]

0,0 0,2 0,4 0,6 0,8 1,0S [-]

aR*Cf

S

αR*cii

s αR

*cii

s

125

0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

0 20 40 60

Proteins [mg/L]

∆R

20 [

*10

12m

-1]

Bulk

Retained

0,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

0 20 40 60

Polysaccharides [mg/L]

∆R

20 [

*10

12m

-1]

Bulk

Retained

a) b) Figure 7.14- Filterability vs. protein (a) and polysaccharide (b) concentrations in the

concentrating experiments. Figure 7.14 shows that, in this research, filterability and SMP concentrations are not directly related. Pearson coefficient results between filterability and protein concentrations, in the bulk activated sludge and retained in the DFCi membrane, are 0.24 and 0.3, respectively. The Pearson coefficient results between filterability and polysaccharide concentrations, in the bulk activated sludge and retained in the DFCi membrane, are 0.14 and 0.11, respectively. The lack of consistency between filterability and SMP results can be caused by the applied SMP methods, which quantify all SMP in the samples regardless of their influence on membrane filtration, as explained in Chapter 6. As mentioned in Chapter 6, several authors propose that the effect of SMP concentrations on fouling is only verified under low SRT conditions (Drews et al. 2008; Kimura et al. 2009). Given that the Heenvliet MBR applies long SRT, i.e. about 20 days (Table V-I), no relationship between filterability and SMP concentrations would be present. The lack of relationship between filterability and SMP concentrations was also verified by other authors, as follows. Lyko et al. (2008), in a long term monitoring study at a full scale MBR, concluded that SMP concentrations are inappropriate indicators of sludge filterability. Geilvoet (2010) reached similar conclusion. As obtained by the abovementioned authors, in this research there is no direct relationship between filterability and SMP concentrations.

Viscosity results

The average apparent viscosity according to the applied shear, calculated from the results of 43 samples with MLSS concentrations between 6.2 and 12.5 g/L, is presented in Figure 7.15. The vertical bars represent the standard deviation for each average result.

126

0

20

40

60

80

0 20 40 60 80 100

Shear rate [s-1

]

Appare

nt vis

cosity

[m

Pa.s

]

Figure 7.15- Average apparent viscosity vs. shear rate in membrane tank activated sludge

samples. As Figure 7.15 shows, the activated sludge always presented a non-Newtonian behavior, with decreasing apparent viscosity along with an increasing applied shear rate, as reported by other authors (Rosenberger et al. 2002c; Hasar et al. 2004; Meng et al. 2007). These authors also report a strong relationship between the apparent viscosity results and MLSS concentration. In this research, between the apparent viscosity and MLSS concentrations, a minimum value of the Pearson correlation coefficient of 0.82 was obtained for the shear rate 5 s-1 and a maximum of 0.86 for the shear rates 10 and 100 s-1, which confirms a strong positive correlation between MLSS and apparent viscosity. Figure 7.16 represents the apparent viscosity, at shear rate 100 s-1, MLSS concentration and filterability results of the activated sludge samples.

Figure 7.16- Apparent viscosity vs. MLSS concentration and filterability of membrane

tank activated sludge samples. With regard to the apparent viscosity and filterability results represented in Figure 7.16, there are differences between samples with MLSS concentrations equal to or lower than 10 g/L and samples with MLSS concentrations above 10.5 g/L, described as follows.

0,01

0,1

1

10

0 5 10 15Apparent viscosity [mPa*s]

∆R

20 [

*10

12m

-1]

0

5

10

15

ML

SS

[g

/L]

DR20

MLSS

∆R20

127

Samples with MLSS concentration equal to or lower than 10 g/L provide viscosity results between 3 and 10 mPa.s, with moderate to poor filterability. The higher values of ∆R20, thus samples with worse filterability, are obtained when the MLSS concentration is equal to or lower than 10 g/L. In contrast, samples with MLSS concentrations above 10.5 g/L provide viscosity results between 8 and 12 mPa.s, with moderate to good filterability. Samples with higher viscosity require higher cross-flow velocities to create turbulent regimes (Rosenberger et al. 2002c). However, in a laminar regime, such as that defined in the DFCm when activated sludge with MLSS concentrations above 5 g/L is filtered (Geilvoet 2010), samples with higher viscosity also provide higher shear stress on the membrane, improving sludge filterability. Figure 7.16 shows differences in the apparent viscosity results between samples with MLSS concentrations equal to or lower than 10 g/L and samples with MLSS concentrations above 10.5 g/L, as follows. Samples with MLSS concentration equal to or lower than 10 g/L show an almost linear increase in viscosity with increasing MLSS concentrations. A linear regression between the two aforementioned parameters, for samples within the designated MLSS concentration range, provides an R2 of 0.94. In contrast, samples with MLSS concentrations above 10.5 g/L, do not present a linear increase in viscosity with increasing MLSS concentrations. A linear regression between the two aforementioned parameters, for samples within the designated MLSS concentration range, provides an R2 of 0.57. As discussed in Chapter 6, the viscosity of samples with MLSS concentration above 10.5 g/L is more dependent on other factors, besides temperature, MLSS concentration and shear rate, than samples with MLSS concentration lower than or equal to 10 g/L.

Particle counting in the range of 0.4-5 µm

The average particle counting results, in the range 0.5-1 µm, of all measured samples, are shown in Figure 7.17 a). Figure 7.17 b) shows the average of the normalized particle counting data. The vertical bars represent the standard deviation in each particle size.

0

1

2

3

4

5

0,5 0,6 0,7 0,8 0,9 1,0

Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[*1

06/m

L]

0

2

4

6

8

10

-0,25 -0,2 -0,15 -0,1 -0,05 0

Log (Particle size) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

0,56 0,66 0,76 0,86 0,96

Particle size [µm]

a) b) Figure 7.17- Average number of particles/mL per particle size (a) and normalized data (b)

in the range 0.5-1 µm.

128

The number of particles in each particle size is extremely variable, as shown by the standard deviation bars in Figure 7.17 a). Nevertheless, Figure 7.17 b) shows that the particle counting results were very consistent. In Figure 7.17 b), the standard deviation, in all measured sizes, is very similar and the represented normalized data have a clear linear behavior, demonstrated by an R2 value of 0.99. We initially postulated that activated sludge with MLSS concentration above 10.5 g/L was retaining fouling particles in its matrix, as opposed to sludge with MLSS concentrations equal to or lower than 10.5 g/L. To evaluate the hypothesis, the total number of particles in the range 0.5-1 µm and the MLSS concentration, in each sample measured in the concentrating experiments, are presented in Figures 7.18 and 7.19. Samples are identified by day and time of measurement. Particle counting results obtained during parallel configuration before the first HRT was completed are not linked to the other particle counting results.

0

5

10

15

8:1

2

9:5

3

10

:53

11

:38

13

:06

7:5

4

8:2

6

9:0

1

9:3

3

10

:32

11

:10

8:1

9

9:1

1

9:4

1

10

:11

10

:39

11

:08

8:2

9

9:0

6

9:4

5

10

:10

10

:39

11

:14

Time [h:min]

To

tal n

um

be

r o

f p

art

icle

s [*1

06/m

L]

0

5

10

15

MLS

S [g/L

]

Total particles

MLSS

1-7-09 3-7-09 20-7-09 23-7-09

Figure 7.18- Total number of particles/mL, in the range 0.5-1 µm, and MLSS

concentration in activated sludge samples. Figure 7.18 represents concentrating experiments where the MLSS concentration in the sludge samples, collected after the first HRT was completed, was higher than 10.5 g/L. After the first HRT is completed, a reduction in the total number of particles, coupled with an increase in MLSS concentrations, is visible in all the concentrating experiments, represented in Figure 7.18. In the concentrating experiments performed on the 3-7-09 and 23-7-09, two measurements were taken before the first HRT was completed. In the latter measurements, the number of submicron particles is increasing. These results can be explained as follows. As mentioned earlier, increased DO concentrations and mixing conditions will break the sludge flocs (Abbassi et al. 1999; Ji et al. 2006). Thus, at the restart of aeration in a membrane tank, it is likely that the number of smaller particles will increase due to floc breakage. This speculation is confirmed by the increasing number of particles, in the

129

range 0.5-1 µm, in the samples measured on 3-7-09 and 23-7-09, before the first HRT is completed. A decrease in the number of particles, along with an increase in MLSS concentration, is obtained, in all concentrating experiments represented in Figure 7.18, after the first HRT is completed. In a membrane tank, after the first HRT, a decrease in the number of particles, in the size range 0.5-1 µm, along with an increase in MLSS concentration and an increase in filterability (Figure 7.9), can be explained by the retention of particles in the sludge matrix. Figure 7.19 shows the total number of particles, in the range 0.5-1 µm, and the MLSS concentration in the samples collected during concentrating experiments, where MLSS concentration was equal to or lower than 10 g/L. Particle counting results obtained during parallel configuration before the first HRT was completed are not linked to the other particle counting results.

0

5

10

15

9:2

5

10

:06

10

:41

11

:33

12

:11

9:2

9

10

:00

10

:22

11

:32

12

:09

9:5

0

10

:12

10

:39

11

:29

11

:54

9:4

0

10

:09

10

:40

11

:09

12

:20

Time [h:min]

To

tal n

um

be

r o

f p

art

icle

s

[*1

06/m

L]

0

5

10

15

ML

SS

[g

/L]

Total particles

MLSS

27-10-09 29-10-09 19-11-09 1-12-09

Figure 7.19- Total number of particles/mL, in the range 0.5-1 µm, and MLSS

concentration in activated sludge samples. In the results represented in Figure 7.19, after the first HRT is completed, the total number of particles in the range 0.5-1 µm is either stable or increases in all the concentrating experiments, with the exception of the experiment performed on 27-10-09. Furthermore, the improvement of filterability, after the first HRT is completed, in the concentration experiments represented in Figure 7.19, is on average 11%, while in the concentrating experiments represented in Figure 7.18, is on average 30%. The results presented in Figure 7.19, with the exception of the experiment performed on 27-10-09, are consistent with our initial hypothesis, i.e. activated sludge with MLSS concentrations equal to or lower than 10 g/L would not retain particles in its matrix. The exception to the above conclusion is the concentrating experiment performed on 27-10-09, where the total number of particles is decreasing, even when the MLSS concentration remains stable. Furthermore, the total number of particles, obtained before

130

the first HRT is completed, is decreasing over time, contrary to the expected breaking effect of aeration. It is possible that circumstances such as decreasing temperatures (Figure 7.6), continuous membrane tank operation in on/off mode (Figure 7.5) and inflow characteristics were stressing the activated sludge and accumulating smaller particles in the membrane tank. In this case, whenever the membrane tank is restarted, the improved flow circulation can carry away the smaller particles. In this particular case, aeration does not cause breakage because the activated sludge is assumed to be already fragmented. Geilvoet (2010) concluded that filterability is closely related to the amount of fine particles in the free water. Furthermore, that the submicron particle volume was a good indicator of the activated sludge filterability. The submicron volume was defined as the total volume of particles in the range 0.4 to 1 µm. The filterability and volume of submicron results obtained by Geilvoet (2010) were reproduced in Figure 2.6. In this research, the submicron volume of the results obtained in the concentrating experiments was calculated as defined by Geilvoet (2010). The filterability and volume of submicron particles, obtained in the concentrating experiments, are presented in Figure 7.20.

0,01

0,1

1

10

0,0 0,5 1,0 1,5 2,0 2,5

Volume submicron [*106µm

3/mL]

∆R

20[*

10

12m

-1]

1-Jul-09 HRT<1

1-Jul-09 HRT>1

3-Jul-09 HRT<1

3-Jul-09 HRT>1

20-Jul-09 HRT<1

20-Jul-09 HRT>1

23-Jul-09 HRT<1

23-Jul-09 HRT>1

27-Oct-09 HRT<1

29-Oct-09 HRT<1

29-Oct-09 HRT>1

19-Nov-09 HRT<1

19-Nov-09 HRT>1

1-Dec-09 HRT<1

1-Dec-09 HRT>1

27-Oct-09 HRT>1

Figure 7.20- Filterability vs. submicron particle volume.

Figure 7.20 shows that the results obtained in this research are within the range of values obtained by Geilvoet (2010). However, in this research, no extreme filterability, either very good or very poor, was measured. The majority of the samples had moderate filterability. Nevertheless, the results of submicron particle volume are within the range of results obtained by Geilvoet (2010) for samples with moderate filterability. Particle counting in the range of 2-100 µm

Figure 7.21 presents the average number of particles/mL and normalized particle counting data per particle size in the range 10-100 µm. The vertical bars represent the standard deviation in each particle size measurement.

131

0

4

8

12

16

20

0 20 40 60 80 100Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[*1

03/m

L]

0

2

4

6

1 1,2 1,4 1,6 1,8 2

Log (Particle size) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

10 100Particle size [µm]

Lower end

[----------------------]Upper end

[-----------] a) b)

Figure 7.21- Number of particles/mL per particle size (a) and normalized particle counting data (b), in the range 10-100 µm, in membrane tank activated sludge samples.

Figure 7.21 shows that the particle counting results in the range 10-100 µm were very homogenous. The normalized data, represented in Figure 7.21 b), show that all samples present a variable β, i.e. a variable slope in the normalized particle counting data. Table VII-II presents the average values of β, Log A and R2, for each concentrating experiment, obtained when a linear regression is fitted to the normalized particle counting data, at the lower and upper ends of the size range, presented in Figure 7.21 b).

Table VII-II Average values of β, Log A and R2, at the lower and upper ends of the 10-100 µm particle size range, in each concentrating experiment.

Average β Average Log A Average R2 Concentration experiment

[day-month-year]

Lower

end Upper

end

Lower

end Upper

end

Lower

end Upper

end 20-1-09 -0.30 5.50 -4.00 13.68 -0.41 0.93 22-1-09 -0.28 5.52 -4.03 13.70 -0.39 0.94 1-7-09 -0.70 5.67 -3.58 13.76 -0.76 0.92 3-7-09 -0.40 5.66 -3.94 13.77 -0.55 0.92

20-7-09 -0.67 5.74 -3.54 13.94 -0.57 0.92 23-7-09 -0.37 5.87 -3.99 14.16 -0.51 0.92 27-10-09 -0.32 6.11 -4.14 14.41 -0.44 0.92 29-10-09 -0.36 6.34 -4.08 14.77 -0.50 0.92 19-11-09 -0.34 6.46 -4.06 15.08 -0.47 0.93 1-12-09 -0.28 5.91 -4.12 14.18 -0.37 0.93

Lawler (1997) theoretically defined particle size distributions and stated that the variable β, if apparent, is less than 1 at the lower end of the size range and higher than 4 at the upper end of the size range. The values of β, at the lower and upper ends of the size range, presented in Table VII-II, comply with the limits described by Lawler (1997). It can be concluded that the obtained particle counting distributions, in the range 10-100, are reliable.

132

When the particle counting results in the range 10-100 µm are analyzed per concentrating experiment, the only noticeable difference is between samples collected before and after the first HRT is completed. An example, from the activated sludge in the concentrating experiment of 23-7-09, is presented in Figure 7.22. Similar results (not shown) were obtained in all concentrating experiments. Samples are identified by time of collection and HRT, since the restart of the membrane tank.

0

4

8

12

16

20

0 20 40 60 80 100

Particle size [µm]

Nu

mb

er

of P

art

icle

s [x1

03/m

L]

8:29 HRT<1

9:06 HRT<1

9:45 HRT>1

10:10 HRT>1

10:39 HRT>1

11:14 HRT>1

Figure 7.22- Number of particles/mL per particle size, in the range 10-100 µm, for

activated sludge samples collected in concentrating experiment of 23-7-09. Figure 7.22 shows that the only particle size distribution slightly detached from the others is the sample collected at 8:29, before the first HRT was completed. This sample has a higher number of particles, between 10 and 25 µm, coupled with lower counts, in the range 25 to 100 µm, than all the others samples. These results are consistent with the effects of deflocculation, described in Chapter 6. In this case, whenever the membrane tank is restarted, the opposite effect occurs, i.e. flocculation of the activated sludge takes place, reducing the number of particles in the range of 10 to 25 µm and increasing the number of particles between 25 and 100 µm, as obtained in the samples collected before the first HRT was completed. The abovementioned flocculation, in the suspended fraction of the sludge, occurring upon the restart of the membrane tank, is fast and does not progress after the first HRT is completed, as Figure 7.22 shows. With the exception of the sample collected at 8:29, all the other samples present similar distributions. The homogeneity of the particle size distributions in the range 10-100 µm, after the first HRT is completed, is common to all concentrating experiments (results not shown). It can be concluded, after the first HRT, aeration or concentrating effects are not visible through particle counting in the range of 10-100 µm. Figure 7.23 represents the sludge floc size, defined as the size where the maximum volume of particles in the range 10-100 µm was obtained vs. the filterability and MLSS concentration obtained in the concentrating experiments.

133

Figure 7.23- Floc size vs. filterability and MLSS concentration in activated sludge

samples. As mentioned earlier, the sludge filterability had mainly moderate quality in the concentrating experiments. There is no significant variation in the activated sludge filtration quality enabling us to provide definite conclusions, regarding the relationship between filterability and floc size, in one MBR plant. Nevertheless, Figure 7.23 shows that the floc size mainly varies between 40 and 50 µm, which is not very significant. However, the small variation in floc size does agree with the conclusions provided in Chapter 5, stating the relationship between filterability and floc size may not be visible in one MBR due to aeration restrictions.

7.5 Link between Concentrating and Diluting Results

The results of the diluting experiments, performed with the last sample collected in each concentrating experiment, are presented in Figure 7.24. Figure 7.24 shows the average filterability and average MLSS concentration obtained in all the diluting experiments. To allow direct comparison with filterability results provided in Chapter 6, the sets were divided into two groups: “High MLSS” for sets obtained from original sludge with MLSS concentration above 10.5 g/L; “Low MLSS” for sets obtained from original sludge with MLSS concentration equal to or lower than 10 g/L. The number of sets used to calculate the average results is indicated in brackets in the key. Original samples, i.e. not diluted with permeate, are represented by larger points.

0

0,2

0,4

0,6

0,8

1

0 20 40 60 80 100

Floc size [µm]

∆R

20 [x10

12m

-1]

0

5

10

15

20

MLS

S [g/L

]

DR20

MLSS

∆R20

134

0,0

0,1

0,2

0,3

0,4

0,5

0 5 10 15MLSS [g/L]

∆R

20 [*1

01

2m

-1]

High MLSS (4 sets)

Low MLSS (4 sets)

Figure 7.24- Average filterability and MLSS concentration in diluting experiments.

Figure 7.24 shows similar results to those described in Chapter 6. Dilution samples providing from sludge with high MLSS concentration, i.e. above 10.5 g/L, present lower filterability than the original samples, as opposed to dilution samples providing from sludge with low MLSS concentration, i.e. equal to or less than 10 g/L. The SMP concentrations, viscosity and particle counting results in the range of 0.4 to 5 µm and 2 to 100 µm provided similar results (not shown) to those described in Chapter 6. The results obtained in the diluting experiments, performed with the last sludge collected on each day of the concentrating experiments, show that the sludge, obtained after one morning of operation, had similar characteristics to sludge obtained after several days of membrane tank operation.

7.6 Conclusions

We initially postulated that, due to structural differences, MBR activated sludge with high MLSS concentration, i.e. above a certain critical MLSS concentration, would retain fouling particles in its matrix, as opposed to sludge with low MLSS concentration. In this research, the critical MLSS concentration was found to be 10.5 g/L. As follows, terms of high and low MLSS concentrations refer to the abovementioned critical MLSS concentration. The main conclusions obtained in this research can be summarized as follows:

- The concentrating experiments performed during series configuration, where the activated sludge has high MLSS concentration, show increasing filterability, a 68% improvement, with increasing MLSS concentration. In the concentrating experiments, performed during parallel configuration, where the activated sludge has high MLSS concentration, filterability improves 36% with increasing MLSS concentration. Particle counting results in the range 0.5-1 µm showed that the total number of particles during the aforementioned concentrating experiments, is decreasing.

135

- The concentrating experiments performed during parallel configuration, where the activated sludge had low MLSS concentration, show a much lower increase of filterability, an 11% improvement, with increasing MLSS concentration. During these experiments, with one exception, the total number of particles in the range of 0.5-1 µm does not decrease. However, probably due to extra stress in the activated sludge, resulting from decreasing temperatures and continuous membrane tank operation in on/off mode, the obtained results are not completely conclusive.

- Structural differences were identified between activated sludge with high MLSS concentration and activated sludge with low MLSS concentration. SMP concentrations in activated sludge with high MLSS concentration are higher than SMP concentrations in activated sludge with low MLSS concentration. The viscosity of sludge with high MLSS concentration is more dependent on other factors besides temperature, MLSS concentration, shear rate and shear stress.

- In this research, the submicron volume results, calculated as described by Geilvoet (2010) are within the range of values obtained by this author.

- The results of the diluting experiments, performed with the last sludge sample of each concentrating experiment, provided similar results to those described in Chapter 6. The activated sludge obtained after one morning of membrane tank operation has similar characteristics to the sludge obtained after several days of operation.

Overall, the results obtained in the concentrating experiments, confirm the initial hypothesis: activated sludge with high MLSS concentration is retaining particles in its matrix, as opposed to activated sludge with low MLSS concentration, i.e. between 6.2 and 10.5 g/L. However, the concentrating experiments of activated sludge with low MLSS concentration were not completely conclusive, probably due to extra stress in the activated sludge. Nevertheless, a concentrating effect, improving filterability of membrane tank activated sludge, was identified.

136

137

8. Assessing Membrane Integrity through Particle Counting

8.1 Introduction

Membrane Bioreactor (MBR) technology removes, next to particles, protozoan cysts and oocysts, helminthes ova, bacteria and viruses (Metcalf & Eddy 2003). However, the removal of microorganisms is only effective if the membranes are intact. Particle counting, as an integrity monitoring technique, has widespread application in drinking water treatment plants (Guo et al. 2010). In November 2006, filamentous bacteria were identified in the permeate of the Heenvliet MBR, in the Netherlands. An on site monitoring program was carried out through particle counting in the range of 2-100 µm. In January 2007, it was found that for several membranes, the membrane material and frame had become detached and consequently affected the permeate quality. The problem was solved by replacing the membranes, which occurred in October 2007. Curiosity and opportunity motivated the continuation of the on site monitoring of MBR permeate, through particle counting in the range of 2-100 µm, in another 3 full-scale and 4 pilot-scale MBRs, scattered across Europe. The methodology for counting particles is defined as a standard method to characterize physical and aggregate properties in water and wastewater (APHA 2000). For particle counters with light-blockage methods, such as the particle counter used in this research, interferences in the measurements result from contamination with particles, presence of gas bubbles and electronic noise (Hargesheimer et al. 1998; APHA 2000). Moreover, some authors claim that particle counters are not able to detect differences in water quality (Guo et al. 2010). The pilot- and full-scale MBR installations analyzed in this research mainly use UF membranes with nominal pore sizes between 0.04 and 0.08 µm. Therefore, permeate particles, with diameters between 2 and 100 µm, are 100 times larger than the nominal pore size of the membranes. Two hypotheses can explain the presence of these particles in the permeate. Firstly, the membranes are leaking, secondly a physical-chemical and/or biochemical process is occurring in the membrane pores and/or on the permeate side of the membrane, resulting in particles larger than the nominal pore size of the membrane. A membrane leakage can be identified by the presence, in the MBR permeate, of either microorganisms significantly larger than the pore size of the membrane, or activated sludge. Both possibilities were evaluated in this research. The existence of particles larger than the nominal pore size of the membrane in the MBR permeate may not directly indicate a disruption of membrane integrity. It is known that the back flush of chemicals through the permeate side of the membrane is responsible for an increase in particles with sizes between 2 and 5 µm (Poyatos et al. 2010). Furthermore, the existence of particles with sizes between 1 and 32 µm and their aggregation along the water distribution system was acknowledged (Verberk et al. 2007). The phenomenon was explained through a slow deposition of particles, followed by growth and re-suspension

138

(Vreeburg et al. 2008). According to the latter author, the process is highly dependent on the design and maintenance of the water distribution system. The permeate collection systems, analyzed in this research, have different design and maintenance than water distribution systems. Therefore, the theories formulated to explain the presence of particles in water distribution systems should not be directly applied to MBR permeate. However, a similar process where growth/aggregation of particles occurs, either in the membrane pores or in the permeate collection system, is also considered as a possibility to explain the presence of particles, with sizes between 2 and 100 µm, in the MBR permeate. In this research, 255 permeate samples collected from 4 full-scale and 4 pilot-scale MBRs scattered across Europe were measured through particle counting in the range of 2-100 µm. The methodology, results and discussion are presented below. To evaluate the reliability of the particle counter used in this research, 43 demineralized water samples, 192 process water samples and 4 CAS samples were also measured using the same methodology. The sources of the process water were drinking water, an on-site well and permeate. The results and discussion concerning this data are presented in the following methodology section.

8.2 Methodology

Samples

The permeate samples were measured on site, directly after collection. Permeate was collected in the first outlet available, i.e. preferably in the permeate outlets of each membrane module. In this research, 43 samples of demineralized water were measured. The demineralized water samples were either measured in the TU Delft water lab or transported to the WWTP and measured on site. The 4 CAS effluent samples were collected at the Heenvliet WWTP and measured on site.

The location, date of collection and number of permeate and process water samples measured in this research are presented in Table VIII-I. Table VIII-II presents a summary of the main characteristics of the MBR installations under analysis.

139

Table VIII-I Location, date of collection and number of permeate and process water

samples. Number of samples Date

(month-year)

Permeate Process water *

Schilde (Belgium) Full-scale 04-08 2 21 08-08 19 17 04-09 7 9 Monheim (Germany) Full-scale 09-08 17 3 Nordkanal Full-scale 07-07 6 27 (Germany) 11-08 33 10 Heenvliet Full-scale 03-07 63 27 (The Netherlands) 06-07 9 12 09-07 18 21 01-08 18 14 02-08 17 5 06-08 26 3 TUB/KWB (Germany) Pilot 06-07 4 7 NTNU (Norway) Pilot 09-07 8 14 UNITN (Italy) Pilot 11-07 5 1 EAWAG (Switzerland) Pilot 11-07 3 1

TOTAL 255 192

Key: * The sources of the process water were drinking water, an on-site well and permeate.

Table VIII-II Characteristics of MBR installations.

Pop. Equi.

WWTP Membrane Supplier

Membrane pore size

[µm]

Cleaning

Schilde 28,000 CAS* +

MBR

Zenon (HF)

0.04 Mechanical and chemical (once a week)

Monheim 9,700 MBR Zenon (HF)

0.04 Mechanical and chemical (twice a year)

Nordkanal

80,000 MBR Zenon (HF)

0.04 Mechanical and chemical (once a week)

Heenvliet

13,000 CAS + MBR

Toray (FS)

0.08 Mechanical and chemical (twice a year)

TUB/ KWB

250 MBR A3 (FS)

0.2 Mechanical and chemical (frequency according to need)

NTNU

8 MBBR **

Zenon (HF)

0.04 Mechanical and chemical (frequency according to need)

UNITN 200 MBR Zenon (HF) Eidos

0.04; 0.1

Mechanical and chemical (frequency according to need)

EAWAG 100 MBR Zenon, Puron (HF),

Kubota (FS)

0.04; 0.04; 0.4

Mechanical and chemical (frequency according to need)

Key: * CAS - Conventional Activated Sludge system;** MBBR- Moving Bed Biofilm Reactor.

140

Processing of the results

To compare particle counting results of different samples, statistical parameters were calculated and the data reduced to power-law coefficients, calculated from data in the range of 3.5-35 µm. The R2 fitting to a power law equation, calculated from data in the range of 3.5-35 µm, is also used to compare different samples. Ceronio et al. (2002) showed that power-law coefficients can successfully represent particle counting results when applied to particles with sizes between 2 and 50 µm. Furthermore, the published particle counting data, concerning wastewater effluents, are presented through the use of power-law model coefficients (Kaminski et al. 1997; Neis et

al. 1997; van der Graaf et al. 2001; Miska-Markush 2009). In this research, the size range selected to calculate the power-law coefficients was 3.5-35 µm, in order to increase the reliability of the obtained coefficients and allow comparison with particle data obtained from drinking water distribution systems (Verberk et al. 2007; Vreeburg et al. 2008). As mentioned earlier, a membrane leakage can be identified by the presence of either microorganisms or activated sludge in the MBR permeate. To test the first possibility, the particle size distributions were checked for increasing numbers of particles with increasing sizes, matching the sizes of possible permeate microorganisms, i.e. microorganisms larger than the membrane pore size, known to be a part of the microbial communities in the activated sludge and bacteria predators, or belonging to the following levels of the food chain. To discuss the second possibility, particle counting data of MBR activated sludge presented in Chapter 5 is used. The data is used to estimate activated sludge intrusions of 0.1% and 0.01% into the permeate side of the membrane. The estimated activated sludge distributions were then compared with the permeate data obtained in this research. Further details of the materials and methods used in this research are provided in Chapter 4. Particle counting results of demineralized water, process water and CAS effluent

The average total number of particles in the demineralized water samples, measured in a lab environment, was 37 counts/mL. The sample with less particles had a total of 6 particles/mL. Figure 8.1 shows the particle size distribution and normalized particle counting data, from the demineralized water sample, measured in a lab environment, which presented the highest number of particles, namely 78 particles/mL.

141

0

5

10

15

20

0 20 40 60 80 100

Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[#

/mL

]

-2

-1

0

1

2

3

4

0 0,5 1 1,5 2

Log (Particle size) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

1 10 100

Particle size[µm]

a) b) Figure 8.1 - Number of particles/mL per particle size (a) and normalized particle counting

data (b) from a demineralized water sample. According to APHA(2000), particle-free water is practically impossible to obtain. Therefore, a reduced number of particles in the demineralized water samples was expected. The particle counting results of the process water samples, grouped according to the origin of the process water, are characterized by statistical parameters in Table VIII-III.

Table VIII-III Statistical parameters of particle counting results from process water samples grouped according to their origin.

Process water Drinking water On-site well Permeate Number of samples # 136 39 3 Average total number of counts #/mL 285 511 253 Standard Deviation in total number of counts

% 133 78 23

Maximum value in total number of counts

#/mL 2934 2012 317

Minimum value in total number of counts

#/mL 26 166 208

Samples with R2 power-law fitting >0.95

% 24 79 33

Table VIII-III shows that all the process water samples had particles in the range of 2-100 µm, regardless of their origin. The presence of particles in the range of 1-31 µm in drinking water distribution systems was acknowledged by drinking water experts (Verberk et al. 2007; Vreeburg et al. 2008). This phenomenon was explained by a combination of post-treatment flocculation, corrosion, re-suspension and biological growth. Furthermore, it was acknowledged that during the transport of water, from the treatment plant to the distribution reservoirs, more large particles were formed (Verberk

et al. 2007). Therefore, the presence of particles in the process water collected from drinking water distribution systems was expected. On site process water requires smaller

142

distribution systems, but factors such as age and maintenance of the distribution system can affect the deposition and re-suspension of particles (Vreeburg et al. 2008). In the specific case of the process water collected from an on-site well, the gravel filtration treatment may be responsible for a higher average number of particles. In this research, 4 samples of CAS effluent were also measured. The maximum number of total counts in the CAS effluent samples was 5564 particles/mL and the minimum number was 3754 particles/mL. The average value of total counts was 4468 particles/mL, with a standard deviation of 17%. The β values of CAS effluent samples, according to published results and those obtained in this research are presented in Table VIII-IV. Table VIII-IV β values of CAS effluent samples, according to published results and those

obtained in this research. Measured range

[µm] β Literature Source

CAS 2-300 3-5 (Kaminski et al. 1997) CAS 0.1-100 1.0 (Neis et al. 1997) CAS+ AT* 0.1-100 1.0 and 0.8 (Neis et al. 1997) CAS 2-750 2.6-2.83 (van der Graaf et al. 2001) CAS 2-100 3.37-3.87 (Miska-Markush 2009) CAS 2-100 2.8-3.7 This research

Key: * AT- Advanced treatment

As Table VIII-IV shows, the values of β obtained in this research are within the range of values obtained by other researchers. Figure 8.2 shows the average particle size distribution and average normalized particle counting data of the process water, according to its origin, and of CAS effluent. To allow comparison with demineralized water, the sample results shown in Figure 8.1 are also presented in Figure 8.2.

0

100

200

300

400

500

600

0 20 40 60 80 100

Particle size [µm]

Nu

mb

er

of

pa

rtic

les

[#

/mL

]

Process (Drinking water)

Process (Permeate)

Process (On site well)

Demineralized water

CAS effluent

-2

-1

0

1

2

3

4

0 0,5 1 1,5 2

Log (Particle size) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

1 10 100

Particle size[µm]

Process (Drinking water)

Process (Permeate)

Process (On site well)

Demineralized water

CAS Effluent

a) b)

Figure 8.2 Average number of particles/mL per particle size (a) and average normalized particle counting data (b) from process water samples, according to their origin, and CAS

effluent.

143

Overall, Figure 8.2 demonstrates that, in the context of a WWTP, the particle counter used in this research is capable of distinguishing between water samples of different sources, contrary to what Guo et al. (2010) postulated. Figure 8.2 clearly shows that the CAS effluent samples are the ones with more particles and that demineralized water is the sample with smaller counts of particles, as was logically expected. Figure 8.2 also shows several differences in the process water according to its source. The process water from the on-site well has more particles than the process water from permeate and drinking water. However, as Figure 8.2 b) shows, the results of the process water from the on-site well reveal a constant decrease in the number of particles with increasing particle size, contrary to the process water from permeate and drinking water. The process water from permeate and drinking water shows an increase in the number of particles in consecutive particle sizes from 40 µm and 60 µm to 100 µm, respectively. With regard to the process water collected from the drinking water distribution systems, the abovementioned consecutive increase can be explained by flocculation, corrosion, biological growth and consequent re-suspension. A local process water distribution system is smaller in length than the drinking water distribution system, but it is possible that similar phenomena occur in the permeate distribution system of the WWTP. The results shown in Figure 8.2 were used to calculate the volume of particles, assuming that particles have a spherical shape. The number and volume distribution were then used to calculate the integral, as the area below the curves of the number and volume distribution. The obtained results are presented in Table VIII-V and will be used for further discussion. Table VIII-V Total number and volume (integral) for average process water (origin stated

in brackets), average CAS effluent and demineralized water. Total Number

[#/mL] Volume

[ppb]* Process (Drinking water) 79 768 Process (Permeate) 92 2361 Process (On-site well) 164 273 CAS effluent 1442 1398 Demineralized water 21 20

Key: * 1 ppb=1000 µm3/mL

8.3 Results

Permeate particle size distributions

Figure 8.3 presents the total number of particles/mL obtained in all permeate samples. The total count of particles/mL obtained in the process water samples and the average demineralized water results are also provided to allow a rough comparison. Figure 8.3 shows that there are particles with sizes between 2 and 100 µm in all permeate samples.

144

Demineralized

water

1

10

100

1000

10000

0 50 100 150 200 250

Number of samples [#]

Tota

l co

un

t o

f p

art

icle

s [#/m

L]

Process water

Permeate

Figure 8.3 - Total number of particles/mL in permeate and process water samples.

Figure 8.3 shows that, overall, permeate and process water samples present similar values, but maximum results are obtained in the permeate samples. The total number of particles in the permeate samples varies between 20 and 6052 particles/mL, while in the process water, it varies between 33 and 2934 particles/mL. Table VIII-VI presents the statistical parameters applied to the permeate results, grouped according to the on site campaign. Table VIII-VI Statistical parameters of particle counting results from permeate samples,

in on site campaigns, identified by location, month and year. Average

[#/mL] (1) Standard

Deviation [%](2)

Maximum [#/mL] (3)

Minimum [#/mL] (4)

R2>0.95 [%](5)

Schilde 04-08 77 25 91 64 0 08-08 158 57 392 47 63 04-09 126 66 294 51 29 Monheim 09-08 496 176 3859 123 65 Nordkanal 07-07 128 56 217 36 17 11-08 507 136 3244 93 100 Heenvliet 03-07 613 142 4364 29 40 06-07 720 70 1300 114 89 09-07 1447 100 5639 170 50 01-08 501 73 1424 99 67 02-08 501 80 1486 79 65 06-08 672 172 6056 102 81 TUB/KWB 06-07 198 45 320 120 0 NTNU 09-07 53 137 233 20 12.5 UNITN 11-07 68 50 117 28 0 EAWAG 11-07 377 116 877 66 67

Key: (1) Average total number of counts/mL; (2) Standard Deviation in the total number of counts; (3)

Maximum value of the total number of counts;(4) Minimum value of the total number of counts;(5) Samples percentage with R2 power-law fitting >0.95 in the range of 3.5-35 µm.

Table VIII-VI demonstrates the variability of the total count results in the permeate samples. The standard deviation varies between 25% and 176%. In the same location, the results also vary according to the date of measurement. In the Schilde campaign performed on 04-2008, there was no sample with a power-law fitting (R2) exceeding 0.95,

145

while in the 08-2008 campaign, 63% of the samples comply with it. Therefore, it can be concluded that the total number of counts/mL in the range of 2-100 µm varies between locations and within one location. The permeate samples with R2 power-law fitting exceeding 0.95 and below 0.95 were used to calculate average distributions, presented in Figure 8.4. Figure 8.4 shows the average number of particles/mL per particle size (a) and average normalized particle counting data (b).

0

20

40

60

80

100

0 20 40 60 80 100

Particle size [µm]

Nu

mb

er

of

pa

rtic

les

[#

/mL

]

Average R2>0,95

Average R2<0.95

-2

-1

0

1

2

3

4

0 0,5 1 1,5 2

Log (Particle size) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

1 10 100

Particle size[µm]

Average R2>0,95

Average R2<0,95

a) b) Figure 8.4 - Average number of particles/mL per particle size (a) and average normalized

particle counting data (b) for samples with R2 power-law fitting >0.95 and <0.95. In the particle number distribution presented in Figure 8.4 a), the distinction between the two distributions is not clear. In contrast, Figure 8.4 b) shows a clear distinction between the two averages. “Average R2>0.95” has mainly one value of β, i.e. shows a continuous decrease in the number of particles along with increasing particle size. In contrast, “Average R2<0.95” shows several values of β and even an increase in the number of particles with increasing particle size, in the range of 20-40 µm. Permeate with microorganisms

In 15 samples collected from the Heenvliet MBR, in the 3-2007 campaign, increased counts in size increments, matching with sizes of specific microorganisms, were obtained. As specific microorganisms, we defined microorganisms larger than the membrane pore size, known to be a part of the microbial communities in the activated sludge and bacteria predators, or belonging to the following levels of the food chain. All of the samples have an R2 power-law fitting <0.95 in the range of 3.5-35 µm. Figure 8.5 presents the average number of particles/mL per particle size and average normalized particle counting data for the abovementioned 15 samples.

146

0

5

10

15

20

0 20 40 60 80 100

Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[#

/mL

]

-4

-2

0

2

4

0 0,5 1 1,5 2

Log (Particle size) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

1 10 100

Particle size[µm]

a) b) Figure 8.5 - Average number of particles/mL per particle size (a) and average normalized

particle counting data (b) for permeate with microorganisms. Figure 8.5 a) shows that between 30 and 50 µm, a small number of particles was measured in the permeate samples. These few particles are responsible for the peaks in the normalized data presented in Figure 8.5 b). The size of 30 to 50 µm corresponds to the size of several protozoa, such as: Aspidisca sp., Chilodonella cuculatus, Lionotus sp., Hexamitus sp., amoebae and shelled amoebae (STORA 1979; Bitton 1999). It is therefore likely that the measured particles in the size range of 30 to 50 µm are protozoa. Permeate with MBR activated sludge

In 16 samples collected from the Heenvliet MBR, in the 3-2007 campaign, permeate particle counting distributions, matching with estimated activated sludge intrusions into the permeate side of membrane, were detected. The abovementioned 16 samples have an R2 power-law fitting <0.95 in the range of 3.5-35 µm. These results were obtained 3 days after the permeate results with increased counts in size increments matching with sizes of specific microorganisms. Figure 8.6 a) shows the average number of particles/mL per particle size for the aforementioned permeate samples and estimated activated sludge intrusions. Figure 8.6 b) presents the normalized particle counting data. It is noteworthy to recall that particle counting in the range of 2-100 µm of MBR activated sludge, performed with the particle counter used in this research, provided reliable results in the range of 10-100 µm. Therefore, Figure 8.6 presents the particle size range between 10 and 100 µm. The estimated sludge intrusions shown in Figure 8.6 correspond to 0.1% and 0.01% of activated sludge particles reaching the permeate side of the membrane. The normalized particle counting data presented in Figure 8.6 b) shows that the permeate particle size distribution is between the estimated particle size distributions of 0.1% and 0.01% of MBR activated sludge particles.

147

0

5

10

15

20

25

10 40 70 100

Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[#

/mL

]

99,9%Permeate+0,1% Sludge

99,99%Permeate+0,01%Sludge

Permeate

-4

-2

0

2

4

1 1,2 1,4 1,6 1,8 2

Log (Particle size) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

10 100Particle size[µm]

Permeate

99,9%Permeate+0,1% Sludge

99,99%Permeate+0,01%Sludge

a) b) Figure 8.6 - Number of particles/mL per particle size (a) and normalized particle counting

data for average permeate and estimated activated sludge intrusions. The results, presented in Figure 8.6, indicate that the measured permeate particle size distributions are probably a combination of permeate and MBR activated sludge. Figure 8.4 a) shows that the number of particles, in permeate samples with R2 power-law fitting exceeding 0.95, is considerable until particles of 20 µm. The decrease in the number of particles between 10 and 20 µm in the permeate average, shown in Figure 8.6 a), is likely to be caused by permeate particles. From 20 µm onwards, the effect of the MBR activated sludge particles becomes clearer, resulting in a maximum number of particles at 30 µm.

8.4 Discussion

Table VIII-VII presents a summary of the results obtained in this research. The percentage of permeate samples where microorganisms and sludge intrusions were found is indicated, as well as the percentage of samples with R2 power-law fitting exceeding and below 0.95, in the data range of 3.5-35 µm.

148

Table VIII-VII Percentage of permeate samples with microorganisms, activated sludge intrusions, R2 power-law fitting exceeding 0.95 and below 0.95, by on site campaign,

identified by location, month and year. Permeate particle size

distributions[%] Permeate particle size

distributions (*) [%]

With microorganisms

With sludge

R2>0.95 R2<0.95

Schilde 04-08 - - 0 100 08-08 - - 63 37 04-09 - - 29 71 Monheim 09-08 - - 65 35 Nordkanal 07-07 - - 17 83 11-08 - - 100 0 Heenvliet 03-07 24 25 40 60 06-07 - - 89 11 09-07 - - 50 50 01-08 - - 67 33 02-08 - - 65 35 06-08 - - 81 19 TUB/KWB 06-07 - - 0 100 NTNU 09-07 - - 12.5 87.5 UNITN 11-07 - - 0 100 EAWAG 11-07 - - 67 33

Key: (*) Samples percentage with R2 power-law fitting >0.95 or <0.95 in the range of 3.5-35 µm.

Table VIII-VII shows that permeate samples with particle counting results matching the sizes of specific microorganisms and activated sludge intrusions were only found at the Heenvliet MBR, in the on site campaign of 03-2007. In January 2007, a construction failure in the membrane modules, particularly unstuck membrane frames, was detected at the Heenvliet MBR. The membranes modules were replaced in October 2007. The permeate samples with microorganisms were measured in March 2007 and the previous chemical cleaning was carried out in January 2007. Protozoa feed mainly on unattached bacteria. Between January and March, abundant food was available on the permeate side of the membrane, where no competition or predation is likely to exist. Furthermore, unstuck membrane frames can let larger microorganisms enter the permeate side of the membrane through the open gaps between the membrane frames. A contamination followed by an infection is, therefore, the most likely reason for the obtained results. The permeate samples with activated sludge intrusions were also measured in March 2007 at the Heenvliet MBR and are, therefore, affected by the aforementioned membrane construction failure. These results were obtained three days after the permeate results with microorganisms. Under normal operating conditions, an MBR achieves an average of a 4 log reduction of total coliforms. A 0.1% activated sludge intrusion into the permeate side of the membrane represents the loss of 1 log, meaning that the MBR will achieve a 3 log reduction. In the permeate samples with sludge intrusions measured in this research, the permeate particle

149

size distributions were between 0.1% and 0.01% of activated sludge intrusion (Figure 8.6). In these cases, the integrity of the membranes was compromised and the membranes were leaking, confirming our first hypothesis. In all the permeate samples analyzed in this research, particles were counted in the range of 2-100 µm (Figure 8.3). The pilot- and full-scale installations where the permeate samples were collected use mainly UF membranes with nominal pore sizes in the submicron range (Table VIII-II). Therefore, the measured particles are larger than the nominal pore size of the membrane by a factor of 100. At this point, two remarks should be made. The first one is that the nominal pore size of a membrane should not be considered as a totally effective cut-off. During the manufacture of membranes, macrovoids are formed and avoiding them totally is extremely difficult (Chung et al. 2008). Furthermore, after manufacturing, probably only one membrane, in a group of thousands, is checked to confirm the nominal pore size. Secondly, particles have multiple shapes. Depending on the angle of observation, the diameter of a particle can vary by a factor of at least 10 (Boller et al. 2011). Both remarks add uncertainty to the definition of the nominal pore size and particle size, but do not seem enough to explain the difference of a factor of 100 between pore and permeate particle sizes. There is a graphic similarity in the normalized values of “Average R2<0.95” (Figure 8.4 b)) and process water with drinking water quality (Figure 8.2 b)), because a variable β is obtained in both particles size distributions. In process water with drinking water quality, the obtained results can be explained by flocculation, corrosion, biological growth and consequent re-suspension. According to Vreeburg (2008), the process occurs in drinking water distribution systems and is time-dependent and site-specific. In our research, permeate samples with R2 power-law fitting <0.95 in the range of 3.5-35 µm occur in all locations, so our results are not site-specific. However, a similar process, even if not exactly the same as the theory reported for drinking water distribution systems, may be occurring, enabling the aggregation of particles, either in the membrane pores or in the permeate collection system. Back-flushing of chemical products through the permeate side of the membrane is performed once a week, as a cleaning strategy, in the MBR installations with hollow fiber membranes. Poyatos et al. (2010) reported an increase in the number of particles, with sizes between 2 and 5 µm, in the permeate of a UF membrane after chemical cleaning. Moreover, Fenu et al. (2009) reported scaling in the permeate collection system due to precipitation of soluble salts, namely CaCO3, Fe(OH)2 and Ca3(PO4)2, during chemical cleaning at a pH of 11. It is therefore likely that the weekly back flush of chemicals in hollow fiber membranes is introducing particles into the permeate. However, flat sheet membranes are chemically cleaned with much lower frequency. During 2008, the flat sheet membranes at the Heenvliet MBR were chemically cleaned in February and August. However, in the campaign performed in June, the number of particles/mL is higher than in the campaign performed in February (Table VIII-VI). Chemical cleaning may be responsible for part of the particles, with sizes between 2 and 100 µm, present in the MBR permeate, but other phenomena seem to also play a role.

150

When a particle goes through a membrane pore, it is likely that it undergoes chemical and physical changes. Particles that were previously retained in the MBR activated sludge flocs may be released in the vicinity of the membranes, due to increased shear, and penetrate the membrane pores. In the membrane pores and in the permeate collection systems, these particles will be available again for aggregation. Our data do not provide information about the precise nature of the process, but nonetheless indicate that an aggregation of particles occurs either in the membrane pores or in the permeate collection system. Table VIII-VII shows that, with two exceptions, the majority of the visited locations present permeate samples with a constant β in the range of 3.5 to 35 µm, i.e. an R2 power-law fitting exceeding 0.95. These results have no graphic similarity with process water of drinking water quality, where flocculation, corrosion, biological growth and consequent re-suspension occur (Figure 8.2 b) and Figure 8.4 b)). Table VIII-VIII presents the power-law coefficients, averaged by location and campaign, obtained for the sample results with an R2 power-law fitting exceeding 0.95, in the range of 3.5 to 35 µm. Table VIII-VIII Average power-law coefficients for permeate samples with R2 power-law fitting >0.95 in the range of 3.5 to 35 µm, in each campaign, identified by location, month

and year. Number of

samples β Log A

Schilde 04-08 0 - - 08-08 12 2.91 3.42 04-09 2 2.82 3.41 Monheim 09-08 11 2.95 3.59 Nordkanal 07-07 1 2.63 3.11 11-08 33 2.50 3.18 Heenvliet 03-07 25 3.16 3.9 06-07 8 3.37 4.28 09-07 9 3.38 4.11 01-08 12 3.39 4.18 02-08 11 3.16 4.01 06-08 21 2.77 3.54 TUB/KWB 06-07 0 - - NTNU 09-07 1 2.76 3.37 UNITN 11-07 0 - - EAWAG 11-07 2 2.54 3.18

The results shown in Table VIII-VIII allow us to define an average distribution for permeate samples and its maximum and minimum limits. According to the results obtained in this research, if a permeate sample presents particle size distribution results within the limits defined in Figure 8.7 a), and normalized particle counting data as represented in Figure 8.7 b), there are no indications of membrane leakage or growth and aggregation in the permeate collection system.

151

0

100

200

300

400

500

0 20 40 60 80 100

Particles size [µm]

Nu

mb

er

of

pa

rtic

les

[#

/mL

]

Average

Maximum

Minimum

-4

-2

0

2

4

0 0,5 1 1,5 2

Log (Particle size) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

1 10 100

Particle size[µm]

Average

Maximum

Minimum

a) b)

Figure 8.7 - Estimated number of particles/mL per particle size (a) and normalized particle counting data for permeate samples with no apparent membrane leakage or

growth and aggregation. The results shown in Figure 8.7 were used to calculate the volume distribution and, in particular, the integral as the area below the number and volume distribution curves. The obtained results are presented in Table VIII-IX.

Table VIII-IX Total number and volume integral for estimated permeate particle size distributions with no visible membrane leakage or growth and aggregation.

Total Number

[#/mL] Volume

[ppb]* Average permeate 118 45 Maximum permeate 1099 741 Minimum permeate 13 3

Key: * 1 ppb=1000 µm3/mL

The average results presented in Table VIII-IX can be compared with the results presented in Table VIII-V. The average permeate has a similar total number to process water of drinking water quality and of permeate quality, but the total volume is much lower for the estimated average permeate. The results are justified because it was demonstrated that process water of drinking water quality and of permeate quality has higher amounts of larger particles, probably due to accumulation and growth in the water distribution system. In contrast, our average permeate aims to characterize permeate distributions without significant aggregation or growth. The research described in this chapter allows us to define relevant steps of a methodology to check MBR membrane integrity, as follows:

- Permeate grab samples should be measured on site on a regular basis to obtain particle number distributions, i.e. number of particles/mL per particle size, in the range of 2-100 µm.

- Particle number distributions should be used to calculate the integral, i.e. the area below the curve of the number of particles per mL.

152

- Particle number distributions should be normalized to check if possible increased counts in consecutive size ranges are relevant. The normalized data should be fitted to power-law functions in the range of 3.5-35 µm. If an R2 exceeding 0.95 is obtained, no membrane leakage, growth or aggregation should be observed. If an R2 below 0.95 is obtained, membrane leakage, growth or aggregation may be observed.

- If R2 is below 0.95, the normalized distribution will show significant increased counts in consecutive sizes ranges. If the increased counts match the sizes of microorganisms larger than the membrane pore size, known to be a part of the microbial communities in the activated sludge and bacteria predators, or belonging to the following levels of the food chain, there is an indirect indication of integrity loss. If the increased counts occur in extended fractions of the size range, as shown in this research for permeate with activated sludge intrusions, membrane integrity is compromised.

- The obtained integral values should be compared with the results presented in Table VII-IX. Moreover, the particle volume distribution and corresponding integral can eventually be calculated and compared with the results presented in Table VII-IX. If the obtained integral values exceed the maximum values, defined in Table VII-IX, and if the possibilities of microorganisms or sludge intrusions were already excluded, the permeate presents significant growth and/or aggregation. In this case, membrane integrity is not compromised.

8.5 Conclusions

In the beginning of this research, we proposed to assess membrane integrity through particle counting in the range of 2-100 µm in MBR permeate. The application of this methodology was motivated by a specific construction failure in a membrane module of a full-scale MBR. Curiosity and opportunity lead us to repeat the monitoring campaign at another 3 full-scale and 4 pilot MBR installations scattered across Europe. Two hypotheses were formulated to explain the presence of particles with sizes between 2 and 100 µm in the permeate of MBR installations. Firstly, the membranes are leaking and, secondly, a physical-chemical and/or biochemical process is occurring in the membrane pores and/or in the permeate, enabling submicron particles to aggregate. Our hypotheses were tested by collecting and measuring 500 samples of MBR permeate, demineralized water, CAS effluent and process water. The latter was obtained from 3 sources: drinking water, an on-site well and MBR permeate. The obtained conclusions can be summarized as follow:

- In all the permeate samples analyzed in this research, particles were counted in the range of 2-100 µm.

- Permeate with microorganisms and activated sludge intrusions were only found at one full-scale MBR, when a construction failure in the membrane modules was detected.

153

- The permeate samples with microorganisms and activated sludge intrusions indicate, indirectly and directly, respectively, the existence of a membrane leakage.

- Particle size distributions of permeate samples, with R2 power-law fitting below 0.95, in the range of 3.5 to 35 µm, and process water with drinking water quality both present normalized particle counting data with variable β. A similar process, even if not exactly the one described for drinking water distribution systems, may be occurring either in the membrane pores or in the permeate collection system, enabling the growth and/or aggregation of particles.

- The back flush of chemicals, through the permeate side of the membrane, is likely to be responsible for a fraction of the permeate particles, with sizes between 2 and 100 µm, but other phenomena seem to also play a role.

- Permeate samples with R2 power law exceeding 0.95, in the range of 3.5 to 35 µm, presented no signs of leakage, significant growth and/or aggregation. These results were used to establish an average permeate number distribution with maximum and minimum limits.

- Both hypotheses proved true. The membrane leakage was associated with a specific membrane construction failure. A physical-chemical or biochemical process occurs constantly, either in the membrane pores or in the permeate collection system, enabling particles to aggregate. However, it only partially manifests itself by significant aggregation that can be identified by an R2 power-law fitting below 0.95, in the range of 3.5 to 35 µm.

- Relevant steps of a methodology to assess MBR membrane integrity, through particle counting in the range of 2-100 µm, were defined.

154

155

9. Technological and Scientific Relevance

9.1 Introduction

The first part of this chapter aims to clarify the effect of activated sludge parameters on filterability, namely: MLSS concentration, floc size, SMP concentrations and submicron particles. In addition, the effect of operational parameters, such as return ratio, on activated sludge filterability is also discussed. A theory on the effect of MLSS concentration on MBR filtration is developed. Our results indicate that there are structural differences between activated sludge with high and low MLSS concentrations. These structural differences were discussed in terms of viscosity and SMP concentration results and excepted structures, for MBR activated sludge with high and low MLSS concentration, as described earlier. The results presented in this thesis indicate that a critical MLSS concentration for MBR filtration may exist. The concept of critical MLSS concentration and its operational consequences are presented and discussed. Our results also indicate that a specific MLSS concentration range should be applied for MBR design and operation. The design, particularly the possible need for a separate membrane tank, and operational consequences of applying each MLSS concentration range are presented and discussed. Issues of permeate quality and membrane integrity in MBRs are presented as extra information about MBR technology. The chapter ends with recommendations for further research, aiming to improve knowledge of MBR activated sludge and contribute to optimize MBR design and operation. The main conclusions obtained in this thesis are summarized in the last section of this chapter.

9.2 Effect of Mixed Liquor Suspended Solids Concentration on Filterability

The filterability and corresponding MLSS concentration in all MBR activated sludge samples measured in this research are represented in Figure 9.1. The samples are identified according to the type of experiment in which they were measured, namely monitoring, diluting and concentrating.

156

0,001

0,01

0,1

1

10

0 5 10 15 20

MLSS [g/L]

∆R

20[*

10

12m

-1]

monitoring

diluting

concentrating

Figure 9.1 - Filterability vs. MLSS concentration in MBR activated sludge samples.

Figure 9.1 shows that, in this research, three ranges of MLSS concentrations, in terms of their effect on filterability, can be distinguished as follows. In activated sludge with MLSS concentrations between 2 and 8 g/L, better filterability is obtained in activated sludge with lower MLSS concentrations; in activated sludge with MLSS concentrations between 8 and 10.5 g/L, filterability remains more or less constant, regardless of the MLSS concentration; in activated sludge with MLSS concentrations between 10.5 and 18 g/L, filterability improves with higher MLSS concentrations. In this research, designated high and low MLSS concentrations are defined as being above or below a certain critical MLSS concentration, respectively. The concept and absolute value of critical MLSS concentration, found in this research, are discussed in section 9.8. From the results represented in Figure 9.1, it can be concluded that the effect of MLSS concentration on membrane filtration depends on the applied MLSS concentration range. Our explanation for the results presented in Figure 9.1 is related to the MBR activated sludge structure. MBR activated sludge is usually understood as equal to CAS activated sludge. However, MBR activated sludge is two to three times more concentrated than CAS activated sludge and, in an MBR system, more particles are retained in the bioreactor, as opposed to CAS systems, where small particles are washed out through the clarifier. The aforementioned conditions of the MBR systems lead to a specific MBR activated sludge structure, described as follows. In MBR activated sludge with low MLSS concentration, i.e. in this research, equal to or less than 10 g/L, fouling particles, directly affecting filterability, are freely dispersed in the bulk activated sludge. In MBR activated sludge with high MLSS concentration, i.e. in this research, exceeding 10.5 g/L, and with moderate to good quality, fouling particles are more or less retained in the activated sludge matrix. In this research, the structural differences between activated sludge with low and high MLSS concentration were confirmed through filterability, viscosity and SMP concentrations. The filterability of MBR activated sludge was quantified through the ∆R20 parameter and the corresponding cake layer through the αR*ci product, i.e. the specific cake resistance by the concentration of fouling particles, and the s compressibility coefficient. Figure 9.2

157

shows the αR*ci product and s coefficient results, according to the MLSS concentration of the activated sludge samples obtained in the diluting and concentrating experiments.

0

20

40

60

80

100

0 5 10 15 20

MLSS [g/L]

αR

*ci [*

10

-3m

2]

diluting

concentrating

0

0,2

0,4

0,6

0,8

1

0 5 10 15 20

MLSS [g/L]s

[-]

diluting

concentrating

a) b)

Figure 9.2 - αR*ci product (a) and s coefficient (b) vs. MLSS concentration. In Figure 9.2, the areas below the semicircles contain more than 95% of the αR*ci product and s coefficient results obtained in this research. The semicircles represented in Figure 9.2 show that both the αR*ci product and the s coefficient, in activated sludge with low MLSS concentration, present higher results in samples with higher MLSS concentration, as opposed to the results obtained in activated sludge with high MLSS concentration, i.e. lower αR*ci product and s coefficient results in samples with higher MLSS concentration. Figure 9.2 shows that activated sludge samples with high MLSS concentration tend to have a lower αR*ci product and lower compressibility. On average, the samples with high MLSS concentration have a αR*ci product and s coefficient of 17.3 [*10-3m2] and 0.05, respectively, while samples with low MLSS concentration have a αR*ci product and s coefficient of 24.7 [*10-3m2] and 0.09, respectively. Furthermore, our results show that the ∆R20 parameter has a linear relationship with the αR*ci product, providing a R2 fitting of 0.98. Therefore, it can be concluded that cake layer resistance is mainly determined by the specific cake resistance by the concentration of fouling particles. Geilvoet (2010) concluded that the cake layer resistance can be mainly attributed to the concentration of fouling particles in the cake layer. Therefore, it can be concluded that for activated sludge with high MLSS concentration, the cake layer has a lower concentration of fouling particles and is less compressible than activated sludge with low MLSS concentration. The viscosity results showed that activated sludge with high MLSS concentration is more affected by other factors, besides temperature, shear rate and MLSS concentration, than the viscosity of activated sludge with low MLSS concentrations. In addition, activated sludge with high MLSS concentration also has higher amounts of proteins and polysaccharides than activated sludge with low MLSS concentration.

158

The ability to retain fouling particles in the activated sludge matrix is associated with the MLSS concentration in the sludge, but also with its quality. When sludge filterability is poor, fouling particles are likely to be dispersed in the bulk activated sludge. The abovementioned description of the effect of MLSS concentration on activated sludge filterability explains all the results obtained by previous researchers as follows. In this research, increased fouling with increasing MLSS concentration is obtained for low MLSS concentrations. Whenever the MLSS concentration exceeds a given critical value, two possibilities exist. When the sludge quality is moderate to good, a particular structure of the activated sludge matrix enables the retention of fouling particles, resulting in an improvement of sludge filterability. If the sludge quality is poor, fouling particles remain dispersed in the bulk activated sludge, resulting in decreased filterability. The concepts related to this hypothesis are illustrated in Figure 9.3.

Filterability

Poor

MLSS [g/L]

Low MLSS High MLSS

Moderate

Good

Critical MLSS concentration

Figure 9.3 - Effect of MLSS concentration on filterability (reproduction of Figure 6.1). According to the theory illustrated in Figure 9.3, an increasing fouling propensity with increasing MLSS concentration (Cicek et al. 1998; Han et al. 2005; Meng et al. 2007; Trussell et al. 2007) is obtained, above a possible MLSS critical value, because the filtration quality of the sludge is poor due to a fragmented structure unable to retain fouling particles in its matrix. The term “possible” is used to define the critical MLSS concentration because, in the aforementioned case, its existence will not be identified. On the other hand, a decreasing fouling propensity above a possible critical MLSS is obtained in this research and in other reported studies (Harada et al. 1994; Bouhabila et al. 1998; Defrance et al. 1999; Le-Clech et al. 2003) because, due to a particular structure, the activated sludge is able to retain fouling particles in its matrix. The theory presented in Figure 9.3 does not intend to demonstrate that filterability and MLSS concentration have a first order relationship. Our explanation for the obtained results is based on the bulk activated sludge structure. In a submerged MBR, operating in cross-filtration mode, most of the suspended particles are likely to be dragged away from the membrane. During this process, the dragged suspended particles exert a certain shear stress on the cake layer and, consequently, also contribute to the resulting cake. Smaller

159

flocs and colloids are considered as the main contributors to the cake layer (Le-Clech et

al. 2006; Meng et al. 2009). Our results show that the activated sludge structure is probably determining which particles reach the membrane. It can be concluded that MLSS concentration has a direct effect on membrane filtration, attributed to the smaller flocs constituting the cake layer, but also an indirect effect, due to the shear stress exerted by the dragged suspended particles and the bulk activated sludge structure, which, under certain circumstances, may be able to retain fouling particles. It is noteworthy to point out that, in research studies performed with artificial wastewater, improvement of filterability may not occur when MLSS concentration exceeds a possible critical value. Artificial wastewater is usually fed with easily biodegradable substances, therefore, likely to produce less colloidal and dissolved materials. An increase in MLSS concentrations in MBR installations fed with artificial wastewater is likely to be mainly due to increasing amounts of inert material, less likely to be retained by a particular activated sludge structure, which may lead to filterability improvements.

9.3 Effect of Floc size on Filterability

Our results, described in Chapter 5, showed that filterability improves with increasing floc size. These results are represented in Figure 9.4. The MBR activated sludge samples are identified by the type of experiment in which they were collected.

0,001

0,01

0,1

1

10

0 20 40 60 80 100

Floc size [µm]

∆R

20[*

10

12m

-1]

monitoring

Figure 9.4 - Filterability vs. floc size of MBR activated sludge.

The results presented in Figure 9.4 show that there is an improvement of filterability with increasing floc size. The results obtained at a full-scale MBR, when an unusual amount of filamentous bacteria was reported in the MBR activated sludge (Moreau 2010), are circled in Figure 9.4. These results show that whenever the MBR activated sludge is not stabilized, increasing floc sizes will not lead to filterability improvements. An improvement of filterability with increasing floc size is consistent with the results obtained by other researchers (Lee et al. 2003; Cho et al. 2005c). These authors showed that larger flocs, which tend to have a higher porosity and aggregate more loosely, produce cake layers of lower specific resistance, therefore resulting in improved

160

filterability. In addition, the limitation of fouling is being studied by applying coagulants/flocculants and absorbent agents (Koseoglu et al. 2008; Iversen et al. 2009b; Remy et al. 2009), which complies with the conclusion obtained in this research. The results presented in Figure 9.4 were obtained at 4 full-scale and 4 pilot-scale MBRs. Figure 9.5 shows the filterability and floc size obtained at one full-scale MBR. The MBR activated sludge samples are identified by the type of experiment in which they were collected.

0,001

0,01

0,1

1

10

0 20 40 60 80 100

Floc size [µm]

∆R

20[*

10

12m

-1]

concentrating

Figure 9.5 Filterability vs. floc size of MBR activated sludge in one full-scale MBR.

The results presented in Figure 9.5 show that the floc size, in one full-scale MBR installation, does not vary significantly. This is probably due to the aeration limitations. According to Germain et al. (2005), aeration is one of the main factors affecting floc size. In MBR installations, membrane tank aeration is used to prevent clogging, besides removing fouling and providing DO to the biomass, therefore a minimum amount of air has to be continuously supplied. Considering that aeration causes floc breakage (Abbassi et al. 1999; Ji et al. 2006), lower aeration values would be preferable for sludge filterability. However, lower aeration values may cause clogging. The small variability of floc size, in one MBR installation, may contribute to explain why lab-scale results of flux enhancers are hardly applicable to pilot-scale installations (Iversen et al. 2009a). It can be concluded that floc size, in one single full-scale MBR installation, does not vary enough to demonstrate a clear relationship between filterability and floc size.

161

9.4 Effect of Soluble Microbial Products Concentration on Filterability

Figures 9.6 and 9.7 present the filterability vs. protein and polysaccharide concentrations, respectively, of all samples measured in this research. Figures 9.6 a) and 9.7 a) present the SMP concentration in bulk activated sludge, while Figures 9.6 b) and 9.7 b) present the SMP concentration retained in the DFCi membrane. The activated sludge samples are identified by the type of experiment in which they were collected.

0,001

0,01

0,1

1

10

0 10 20 30 40 50 60

Proteins [mg/L]

∆R

20[*

10

12m

-1]

diluting

concentrating

0,001

0,01

0,1

1

10

0 10 20 30 40 50 60

Proteins [mg/L]

∆R

20[*

10

12m

-1]

diluting

concentrating

a) b)

Figure 9.6 - Filterability vs. protein concentration in bulk activated sludge (a) and retained in the DFCi membrane (b), for MBR activated sludge samples.

0,001

0,01

0,1

1

10

0 10 20 30 40 50 60

Polysaccharides [mg/L]

∆R

20[*

10

12m

-1]

diluting

concentrating

0,001

0,01

0,1

1

10

0 10 20 30 40 50 60

Polysaccharides [mg/L]

∆R

20[*

10

12m

-1]

diluting

concentrating

a) b)

Figure 9.7 - Filterability vs. polysaccharide concentration in bulk activated sludge (a) and retained in the DFCi membrane (b), for MBR activated sludge samples.

Figure 9.6 and Figure 9.7 show that no consistent relationship was found in this research between SMP concentration and filterability. The results, presented in these two figures do not define an obvious trend.

162

Figure 9.8 shows the relationship between the SMP concentration in the bulk activated sludge and retained in the DFCi membrane. The results are identified by the type of experiment in which they were obtained.

0

10

20

30

40

50

60

0 10 20 30 40 50 60

Proteins bulk [mg/L]

Pro

tein

s r

eta

ine

d [

mg

/L]

diluting

concentrating

0

10

20

30

40

50

60

0 10 20 30 40 50 60

Polysaccharides bulk [mg/L]

Po

ly.

reta

ine

d [

mg

/L]

diluting

concentrating

a) b)

Figure 9.8 – Protein (a) and polysaccharide (b) concentrations in activated sludge vs. retained in the DFCi membrane.

Figure 9.8 shows that the bulk and retained SMP results in the DFCi membrane are related. On average, 52% of the proteins contained in the bulk activated sludge are retained by the DFCi membrane, while an average of 70% of polysaccharides are retained. In our diluting experiments, the dilution samples were produced according to a defined methodology, therefore, a prediction of the expected SMP results could be made. The SMP results obtained from the dilution samples were compared against the expected SMP results. With regard to SMP concentrations retained in the DFCi membrane, a consistent relationship between filterability and SMP results was found in 24% and 21% of the protein and polysaccharide results, respectively. By “consistent relationship”, we mean a better filterability with lower SMP concentration retained in the DFCi membrane and vice versa. The abovementioned results indicate that the applied SMP methods are not sensitive enough because they quantify all the SMP contained in the samples without differentiating between relevant and irrelevant SMP for membrane filtration. The inadequacy of the applied SMP methods was already reported (Lyko et al. 2008; Geilvoet 2010). Several authors also suggest that SMP concentrations influence fouling only under low SRT conditions (Drews et al. 2008; Kimura et al. 2009). However, in this research, part of our filterability results do show a consistent relationship with SMP concentration, and samples were collected in an MBR applying high SRT conditions, according to the definition proposed by the above authors. Therefore, in this research, we consider it more likely that a consistent relationship between SMP concentrations and filterability is not obtained due to the inadequacy of the applied methods to quantify the relevant SMP for membrane filtration.

163

9.5 Effect of Submicron Particles on Filterability

Submicron particles were identified as important fouling components (Leiknes et al. 2006; Geilvoet et al. 2007a). A worse filterability was related to a higher amount of submicron particles. Geilvoet (2010) proposed the volume of submicron particles, i.e. the accumulated volume of particles in the range 0.4-1 µm (assuming that all counted particles have a spherical shape), as a good indicator of activated sludge filterability. The results of filterability and volume of submicron particles obtained by Geilvoet (2010) were presented in Figure 2.6. The volume of submicron particles was calculated, as defined by Geilvoet (2010), for all the samples measured in this research. The filterability according to the volume of submicron particles, obtained in this research, is presented in Figure 9.9. The activated sludge samples are identified by the type of experiment in which they were collected.

0,001

0,01

0,1

1

10

0,0 0,5 1,0 1,5 2,0 2,5

Volume submicron [*106µm

3/mL]

∆R

20[*

10

12m

-1]

diluting

concentrating

Figure 9.9 - Filterability vs. volume of submicron particles for MBR activated sludge

samples. The results presented in Figure 9.9 show that, in particular, the concentrating sample results are within the range of values obtained by Geilvoet (2010). Figure 9.9 shows that part of the results obtained in the diluting experiments deviate from the range of values obtained by Geilvoet (2010). In the diluting experiments, activated sludge was diluted with permeate, causing dilution and deflocculation, as our particle counting results in the range of 2-100 µm demonstrated. Our particle counting results in the range of 0.4-5 µm, from the diluting experiments, showed that the activated sludge was diluted but, in the abovementioned range, deflocculation was not directly measured. However, the particle counting results in the range of 0.4-5 µm, namely the variability of the measured β values, i.e. the slope of the normalized particle counting distribution, did indicate that deflocculation could have occurred for particles smaller than 0.5 µm, in original samples with high MLSS concentration and moderate to good filtration quality. It is therefore likely that the methodology applied in the diluting experiments caused deflocculation in a lower size range than that quantified by the volume of submicron particles, as defined by Geilvoet (2010).

164

The methodology applied in our diluting experiments does not reflect the normal operation of an MBR installation. In contrast, our concentrating results were obtained during the operation of a full-scale MBR installation. Therefore, the concentrating results are likely to be more representative of MBR activated sludge. Furthermore, our particle counting results in the range of 2-100 µm, in the concentrating experiments, showed that, after the first HRT was completed, flocculation was not visible. Nevertheless, the total number of particles in the range of 0.4-5 µm was decreasing simultaneously with an improvement of filterability. The results presented in Figure 9.9 show that, in the diluting and concentrating experiments, extreme values of filterability, i.e. activated sludge with good and poor filtration quality, were not measured in considerable numbers. Due to this fact, we are not able to provide a definitive assessment on the reliability of the submicron volume of particles as an indicator of filterability. Nevertheless, in view of the above paragraphs, we consider it very likely that the volume of submicron particles, as defined by Geilvoet (2010), is a good indicator of activated sludge filterability.

9.6 Effect of Return Ratio on Membrane Tank Filterability

Our filterability results, presented in Chapter 5, showed that a heterogeneous filterability, between all the tanks of one MBR, can exist, as opposed to a homogenous filterability. These two situations were found and represented in Figure 5.3 and 5.4. Furthermore, when heterogeneous filterability in one MBR installation was obtained, the best filterability was measured in the membrane tanks. Heterogeneous and homogenous filterability, between the various tanks of one MBR, can be explained through the return ratio, from the membrane tanks to the carbonaceous and nutrient removal tanks. A high return ratio would obviously lead to homogenous filterability in all the MBR tanks, while a low return ratio would lead to heterogeneous filterability between the various MBR tanks. Wisniewski et al. (1998) measured a decrease in permeate flux in an MBR operated at high return ratios Therefore, according to the these authors and our own results, with respect to filterability low return ratios are preferable for MBR operation. The return ratio affects the membrane tank activated sludge characteristics in at least two ways, namely the MLSS concentration and floc size of the activated sludge. The MLSS concentration in a membrane tank is dependent on, namely: the MLSS concentration in the carbonaceous and nutrient removal tanks and the ratio permeate flow/ return flow (Equation 1.2). On the other hand, Wisniewski et al. (1998) showed that high return ratios would lead to floc breakage. Our own results, discussed in section 9.3, showed that, in one full-scale MBR, the membrane tank floc size does not vary significantly, probably due to aeration limitations imposed by the need to control clogging. This conclusion is drawn from the data obtained in the concentrating experiments performed at a full-scale MBR with heterogeneous

165

filterability. In the abovementioned experiments, the return ratio varied between 1.7 and 5.6 (Table VII-I). Nevertheless, the floc size in the membrane tanks varied between 40 to 50 µm (Figure 9.4). Our own results, discussed in section 9.3, showed that a relationship between filterability and floc size is visible, in MBRs operating under stable conditions, when the floc size varies between 20 to 50 µm. Therefore, our results lead us to conclude that the return ratio, in one full-scale MBR, does not significantly affect the floc size of the activated sludge in the membrane tanks. The return ratio in an MBR with a separate membrane tank contributes to the MLSS concentration obtained in the membrane tank. Our results, in the concentrating experiments performed at a full-scale MBR with heterogeneous filterability, showed that the lowest return ratios, i.e. below 2 (Table VII-I), coincide with high filterability improvements, namely 71% and 65% in the concentrating experiments performed on 20-1-09 and 22-1-09 (Figure 7.9), i.e. when the MLSS concentration in the membrane tank exceeded the critical MLSS concentration. According to our own results, improvements of filterability with increasing MLSS concentrations only occur when the MLSS concentration in the activated sludge exceeds a critical value. Therefore, it can be concluded that low return ratios are preferable for MBR operation, in MBR installations with a separate membrane tank, applying membrane tank MLSS concentrations exceeding the critical MLSS concentration value.

9.7 Structure of Activated Sludge with High and Low Mixed Liquor Suspended Solids Concentration

In this research, the terms of high and low MLSS concentration are defined as exceeding the critical MLSS concentration and below the critical MLSS concentration, respectively. Furthermore, in this research, the critical MLSS concentration was found to be 10.5 g/L. According to the described in section 9.2, the effect of MLSS concentration on membrane filtration depends on the applied MLSS concentration range, due to structural differences between activated sludge with high and low MLSS concentration. Activated sludge with high MLSS concentration and good to moderate filtration quality may be able to retain fouling particles, contrary to activated sludge with low MLSS concentrations. According to our results, structural differences between activated sludge with high and low MLSS concentrations were obtained, such as SMP concentration and viscosity. These results are discussed as follows.

166

Soluble Microbial Products

Table IX-I shows a summary of the average SMP concentration in bulk activated sludge, obtained in this research, according to the type of experiment, for activated sludge with high and low MLSS concentration.

Table IX-I Average SMP concentration, as proteins and polysaccharides, in MBR bulk activated sludge.

Proteins [mg/L] Polysaccharides [mg/L] Low MLSS High MLSS Low MLSS High MLSS Diluting* 12 15.9 8.6 10.8 Concentrating 14.5 19 3.6 10.6

Key:* Average results from set 1 to 22 intended to characterize sludge with high and low MLSS concentration.

Table IX-I shows that activated sludge with high MLSS concentration and good to moderate filtration quality has a higher SMP concentration than activated sludge with low MLSS concentration. This result was obtained in both the diluting and concentrating experiments. Proteins and polysaccharides are macromolecules that link all particles together (Le-Clech et al. 2006). Higher concentrations of proteins and polysaccharides, in bulk activated sludge with moderate to good filtration quality and high MLSS concentration, indicate that the constituents themselves, to establish a network of activated sludge particles, are available in the bulk activated sludge. In contrast, the concentration of proteins and polysaccharides is lower in bulk activated sludge with low MLSS concentration, so the ability to establish the abovementioned network is reduced. According to our own conclusions, the applied SMP methods are not fully appropriate to characterize the effect of SMP concentrations on filterability because they quantify all SMP, regardless of their impact on MBR filtration. However, because the applied SMP methods are quantifying all the SMP present in the bulk MBR activated sludge, they are likely to be the most suitable to identify structural differences in bulk MBR activated sludge.

Viscosity

The viscosity of MBR activated sludge is usually described as a function of MLSS concentration, shear rate and temperature by several authors (Rosenberger et al. 2002c; Laera et al. 2007; Yang et al. 2009). The models proposed by these authors, represented in Table II-II, were used to estimate the viscosity of all activated sludge samples measured in this research, for a temperature of 20ºC and a shear rate of 100 s-1. The estimated and measured viscosity results, according to MLSS concentration, in all the MBR activated sludge samples measured in this research, are represented in Figure 9.10. The activated sludge samples measured in this research are identified according to the type of experiment in which they were obtained.

167

0

10

20

30

40

0 5 10 15 20

MLSS [g/L]

Ap

p. vis

co

sity

[m

Pa

*s]

diluting

concentrating

Yang et al.(2009)

Rosenberger etal. (2002)

Laera et al.(2007)

Figure 9.10 - Estimated and measured apparent viscosity vs. MLSS concentration in

MBR activated sludge samples. Figure 9.10 shows that our results are generally in between Laera et al. (2007) and Rosenberger et al. (2002c) estimated results. In Figure 9.10, it can be observed that the dispersion of the estimated viscosity, obtained by applying the various models proposed by the abovementioned authors, increases with increasing MLSS concentration. This result shows that, as the MLSS concentration in the sludge increases, the reliability of estimated viscosity results based only on MLSS concentration, shear rate and temperature decreases. Therefore, the impact of other factors, such as SMP concentration or the sludge structure, for instance, on the obtained viscosity is increasing with increasing MLSS concentration in the activated sludge sample. The estimated and measured viscosity results presented in Figure 9.10, obtained from activated sludge with high and low MLSS concentration, as defined in this research, were compared. It was observed that the estimated viscosity of MBR activated sludge with high MLSS concentration was, on average, 2.2 times more inaccurate than the estimated viscosity from MBR activated sludge with low MLSS concentration. For MBR activated sludge samples with high MLSS concentration, the most reliable estimated viscosity results are provided by the model proposed by Rosenberger et al. (2002c). These results show us that structural differences between activated sludge with high and low MLSS concentration, as defined in this research, do exist and are measurable through viscosity. Structure of activated sludge with high and low MLSS concentration

The structure of activated sludge with high and low MLSS concentration, as defined in this research, is conceptually represented in Figure 9.11.

168

Figure 9.11 – MBR activated sludge structure with low MLSS concentration (left) and high MLSS concentration (right) in one MBR installation (reproduction of Figure 6.2).

The filterability, SMP concentration and viscosity results obtained in this research indicate that the bulk MBR activated sludge structure with high MLSS concentration and moderate to good filtration quality is probably composed of a network of linked activated sludge particles behaving as a fine size mesh capable of retaining fouling components, as schematically illustrated in the right part of Figure 9.11. This network of activated sludge particles is probably linked through SMP and other forms of EPS, such as bound EPS. The aforementioned network of particles provides a more viscous activated sludge, with increased viscosity, that in order to be correctly predicted, should probably include the contribution of SMP, EPS or/and structural parameters. An MBR activated sludge structure with low MLSS concentration and an MBR activated sludge structure with high MLSS concentration but poor filtration quality are likely to be composed of fragmented particles, with free and unlinked SMP, thus being unable to retain fouling particles in the activated sludge matrix. The structure of activated sludge with low MLSS concentration is schematically illustrated in the left part of Figure 9.11.

169

9.8 Critical Mixed Liquor Suspended Solids Concentration

The filterability results obtained in this research allowed us to define a critical MLSS concentration. In the context of this research, a critical MLSS concentration is understood as a turning point in the filterability of activated sludge, i.e. the MLSS concentration above which sludge filterability improves due to the retention of fouling particles in the bulk activated sludge. According to our theory, described in section 9.2, bulk activated sludge with MLSS concentration below and exceeding the critical MLSS concentration may have different structures. In this research, continuous measurements in one MBR installation were not performed, therefore we cannot draw conclusions about the operational consequences of continuous operation with critical MLSS concentration for MBR activated sludge filterability. However, we can logically assume that operating an MBR installation that has activated sludge with unstable structure is not beneficial. A stable MBR activated sludge structure is preferable, so critical MLSS concentrations should be avoided. In this research, a critical MLSS concentration of 10.5 g/L was obtained. This critical value was obtained by the results of both the diluting and concentrating experiments. In the diluting and concentrating experiments the activated sludge samples were mainly collected at one full-scale MBR. However, samples collected at 3 other full-scale MBR did comply with the absolute number, defined in this research, as critical MLSS concentration. In addition, the results obtained by Moreau (2010), presented in Figure 3.4, in field campaigns at 5 pilot-scale and 3 full-scale MBRs, and our own results shown in Figure 9.2, showed that a maximum specific cake resistance x solid concentration and compressibility coefficient were obtained for MBR activated sludge with MLSS concentration around 10 g/L. These results indicate that the critical MLSS concentration of 10.5 g/L is not site-specific. The volume of submicron particles, as defined by Geilvoet (2010), vs. the MLSS concentration in all the samples measured in this research, are presented in Figure 9.12. The activated sludge samples are identified by the type of experiment in which they were collected.

170

0,0

0,5

1,0

1,5

2,0

2,5

0 5 10 15 20

MLSS [g/L]

Volu

me s

ubm

icro

n [*1

06 µ

m3/m

L] diluting

concentrating

Figure 9.12 - MLSS concentration vs. volume of submicron particles in MBR activated

sludge samples. The area below the semi-circle, represented in Figure 9.12, contains 94% of the volume of submicron particles results obtained this research. The remaining 6% are obtained from samples of the diluting experiments, but this may be a consequence of the applied diluting methodology, as described in section 9.5. The semi-circle represented in Figure 9.12, as well as the semi-circles represented in Figure 9.2, indicate that maximum values are obtained in activated sludge samples with MLSS concentration around 10 g/L. Geilvoet (2010) concluded that the volume of submicron particles is a good indicator of sludge filterability. If this statement is indeed correct, which our own conclusions also indicate, activated sludge samples with MLSS concentrations around 10 g/L have worse filterability, as Figure 9.1 shows, due to an increased number of submicron particles in the bulk activated sludge. Although our comments in the above paragraphs support the absolute number of 10.5 g/L for critical MLSS concentration as defined in this research, we have reservations about defining an absolute value that can be broadly applied to municipal MBR activated sludge. Our reservations derive mainly from three reasons, explained below. Our own results of the monitoring experiments, shown in Figure 9.1, indicate that improvements in filterability with increasing MLSS concentrations may be obtained for MLSS concentrations above 9 g/L. The conclusions and results obtained in this research, as well as Geilvoet (2010) and Moreau (2010), always rely on the DFCm. If filterability is measured through other methods, with different membranes and membrane operation, absolute results will also differ. Our SMP concentration and viscosity results show structural differences between activated sludge with high and low MLSS concentration, as defined in this research, but do not contribute to define the absolute value for critical MLSS concentration. In this research, the definition of the absolute value for critical MLSS concentration, only derived from the filterability results, measured through the DFCm. Therefore, it is likely that filterability, measured through other methods, indicates different absolute numbers as critical MLSS concentration.

171

Our concept of critical MLSS concentration derives from the activated sludge structure that is directly related to the feed water characteristics. Variations in feed water will mostly likely lead to different MBR activated sludge structures. Consequently, the absolute number for critical MLSS concentration will also vary. It can be concluded that, in this research, a critical MLSS concentration of 10.5 g/L was found through the filterability results of diluting and concentrating experiments, obtained by applying the DFCm. The absolute number of 10.5 g/L can be used as reference but not extrapolated, without further research, to all MBR installations handling municipal wastewater.

9.9 Consequences of the Applied Mixed Liquor Suspended Solids Concentration Range

Our hypothesis to explain the effect of MLSS concentration on MBR filtration, described in section 9.2, allows us to define a low and high MLSS concentration range for MBR activated sludge. The “Low MLSS concentration” range is composed of MLSS concentrations values below the critical MLSS concentration; while the “High MLSS concentration” range is constituted by MLSS concentration values exceeding the critical MLSS concentration. As mentioned earlier, our research does not allow us to draw conclusions about the consequences of operating continuously with critical MLSS concentrations. However, to increase operational control when the MBR is stable, it is always preferable to avoid structural changes in the bulk activated sludge. Therefore, for the design and operation of one MBR installation, a clear choice between “Low MLSS concentration” and “High MLSS concentration” ranges should be made. Table IX-II summarizes the main design and operational consequences of applying the abovementioned low and high MLSS concentration ranges, according to the results obtained in this research. The design consequences mainly refer to the question whether to apply a separate membrane tank or not.

Table IX-II Design and operational consequences of applying low and high MLSS concentration ranges in MBR activated sludge.

Low MLSS concentration High MLSS concentration Design Separate membrane tank not required Separate membrane tank preferable Operation Less clogging More clogging Less air required to scour the membrane and provide DO to the biomass

More air required to scour the membrane and provide DO to the biomass

Less air preferable to promote floc growth Less air preferable to promote floc growth=> Air flow requires optimization

Applied return ratio is irrelevant Low return ratio, i.e. lower than 2, preferable to achieve improved filterability

172

Designing MBR installations to operate activated sludge with low MLSS concentration means, according to our research, that a separate membrane tank is not required. These MBR installations will have tanks with larger volumes, but if the membranes are placed in an aerobic tank, the number of tanks in one MBR will also be reduced. A separate membrane tank, in an MBR installation, is usually applied to assure that carbonaceous and nutrient removal occurs before the activated sludge reaches the membrane tank. However, given the high return ratios applied in full-scale MBRs (Table V-I), unprocessed activated sludge will always reach the membranes. Therefore, a more pragmatic approach should be adopted and membranes should be placed in aerobic tanks of the MBR installation, similar to the design of the Nordkanal MBR. In the aforementioned design, it is likely that cleaning procedures have to be optimized. Furthermore, in MBR installations designed without a separate membrane tank MLSS concentrations as low as possible should be applied. Applying high MLSS concentrations can be an advantage in terms of filterability, when the MBR installation has a separate membrane tank. In this case, filterability can be improved by applying a low return ratio and a higher MLSS concentration, thus stimulating the retention of fouling particles in the bulk activated sludge matrix. However, operating with activated sludge of high MLSS concentration also means more clogging and increased air rates to scour the membranes. Moreover, air flow should be kept at minimum levels to reduce costs, but also to promote larger flocs. Therefore, air flow, in MBR installations treating activated sludge with high MLSS concentration, should be carefully optimized. Overall, MBR installations treating activated sludge with low MLSS concentrations provide more operational control and are likely to optimize operating costs by reducing the amount of air, needed for oxygen supply.

173

9.10 Suspended Particles in the Membrane Bioreactor Permeate

Our research, described in Chapter 8, showed that MBR permeate always has particles in the size range of 2 to 100 µm. The total number of counts/mL in MBR permeate, in the size range of 2-100 µm, obtained at 4 full-scale and 4 pilot-scale MBRs, is presented in Figure 9.13. Each box illustrated in Figure 9.13 shows, through the top and bottom lines, the average and minimum number of total counts/mL, respectively, obtained in each MBR installation; the top of the line, attached to each box, shows the maximum number of total counts/mL.

1

10

100

1000

10000

Sch

ilde

Mo

nh

eim

EA

WA

G

TU

B/K

WB

No

rdka

na

l

NT

NU

UN

ITN

He

en

vlie

t

To

tal c

ou

nt o

f p

art

icle

s [#

/mL

]

Figure 9.13 – Total number of particles/mL in permeate samples from MBR installations. The results shown in Figure 9.13 lead us to conclude that the cut-off of UF membranes should not be defined according to the membrane nominal pore size. The presence of particles with sizes between 1 and 32 µm and their aggregation in drinking water distribution systems has been already acknowledged (Verberk et al. 2007). The phenomenon was explained by a slow deposition of particles, followed by growth and re-suspension, occurring in water distribution systems according to their design and maintenance (Vreeburg et al. 2008). Drinking water samples, used in the visited WWTP as process water, were also measured in this research by particle counting in the range of 2-100 µm. Similarities in particle counting parameters, such as the β value, i.e. the slope of the normalized distribution, were found between permeate and drinking water samples. It can be concluded that a similar phenomenon to that identified for drinking water distribution systems, even if not exactly the same but where growth of particles occurs, may contribute to explain the existence of particles in MBR permeate. Poyatos et al. (2010) reported an increase of particles with sizes between 2 and 5 µm in the permeate of an MBR pilot, after chemical cleaning. Hollow fiber membranes, such as those used in the aforementioned MBR installation, require a weekly back flush of chemicals through the permeate side of the membrane. Moreover, Fenu et al. (2009) reported scaling in the permeate collection system due to precipitation of soluble salts, namely CaCO3, Fe(OH)2 and Ca3(PO4)2, during chemical cleaning at a pH of 11. Changes in pH during chemical cleaning can therefore justify the existence of particles in the

174

permeate. However, in this research, MBR installations with flat sheet membranes, where chemical cleaning is carried out twice a year (Table VIII-II), were also analyzed and increasing numbers of permeate particles were measured without being preceded by a chemical cleaning operation. Therefore, we can conclude that chemical cleaning may be responsible for the existence of permeate particles in MBR installations, but other phenomena are also playing a role. Our research, described in Chapter 8, showed that under normal MBR operation, i.e. when membrane integrity is not compromised, suspended particles in MBR permeate are always present. The existence of these particles can be explained by a physical-chemical and/or biochemical process, occurring either in the membrane pores or in the permeate collection system.

9.11 Methodology for Verifying Membrane Integrity

Our research, described in Chapter 8, allowed us to define a methodology for verifying membrane integrity and simultaneously evaluate the amount of growth or accumulation of particles in permeate collection systems. The proposed methodology is based on the analysis of counts and shapes of particle counting distributions. It consists of 8 steps, described as follows: 1) The reliability of the particle counter should be checked through a zero count test. If

the zero count tests provide results above the allowable limits, flushing of demineralized water or suitable chemicals should be performed. Allowable zero count limits and suitable chemicals are defined in the specifications of the particle counting equipment.

2) A sample of demineralized water should be measured before and after the measurements of permeate samples, in order to control the drift of the particle counter.

3) Permeate grab samples should be measured on site on a regular basis to obtain particle number distributions, i.e. number of particles/mL per particle size, in the range of 2-100 µm. The maximum and minimum allowable limits of particle number distribution, where no membrane leakage, excessive growth or aggregation is visible, are presented in Figure 9.14.

175

0

100

200

300

400

500

0 20 40 60 80 100

Particles size [µm]

Num

ber of partic

les [#/m

L]

Average

Maximum

Minimum

Figure 9.14 - Particle number distributions of MBR permeate with no membrane leakage,

excessive growth or aggregation (reproduction of Figure 8.7 a)).

4) Particle number distributions should be used to calculate the integral, i.e. the area below the curve of the number of particles per mL.

5) Particle number distributions should be normalized to check if possible increased counts in consecutive size ranges are relevant. The normalized data should be fitted to power-law functions in the range of 3.5-35 µm. If the obtained R2 exceeds 0.95, no membrane leakage, growth or aggregation is visible. If R2 is below 0.95, membrane leakage, growth or aggregation may be visible.

6) If R2 is below 0.95, the normalized distribution will show significant increased counts in consecutive size ranges. If the increased counts match the sizes of specific microorganisms, i.e. microorganisms larger than the membrane pore size, known to be a part of the microbial communities in the activated sludge and bacteria predators, or belonging to the following levels in the food chain, there is an indirect indication that membrane integrity is compromised. If the increased counts occur in extended fractions of the size range, there is a direct indication that membrane integrity is compromised, due to sludge intrusion. Examples of permeate with microorganisms and activated sludge are presented in Figure 9.15.

176

-4

-2

0

2

4

0 0,5 1 1,5 2

Log (Particle size) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

1 10 100

Particle size[µm]

-4

-2

0

2

4

1 1,2 1,4 1,6 1,8 2

Log (Particle size) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/mL

/ µm

]

10 100

Particle size[µm]

Permeate

99,9%Permeate+0,1% Sludge

99,99%Permeate+0,01%Sludge

a) b)

Figure 9.15 - Normalized particle counting data of MBR permeate with microorganisms (a) and sludge intrusions (b) (reproduction of Figures 8.5 b) and 8.6b), respectively).

7) The obtained integral values, i.e. the total number, should be compared with the

results presented in Table IX-III. Moreover, the volume distribution and the corresponding integral can eventually be calculated and compared with the results presented in Table IX-III. If the obtained integral values exceed the maximum values and if the possibilities of microorganisms or sludge intrusions were already excluded, the permeate presents significant growth and/or aggregation. In this case, membrane integrity is not compromised, but the permeate collection system should be cleaned.

Table IX-III Total number and volume for MBR permeate

(reproduction of Table VIII-IX). Total Number

[#/mL] Volume [ppb]*

Average permeate 118 45 Maximum permeate 1099 741 Minimum permeate 13 3

Key: * 1 ppb= 1000 µm3/mL

8) The particle counter should be cleaned with a continuous flow of demineralized water,

until the total number of particles measured in the demineralized water is similar to that obtained before measuring the permeate samples. If this step is not achievable with demineralized water, suitable chemicals should be used.

177

9.12 Recommendations for Further Research

The research described in this thesis allows us to identify several subjects that would be greatly improved by further research. A description of these subjects is provided as follows. Filterability measurements within one MBR installation

Our research showed that filterability can vary between MBR tanks of one MBR installation. Therefore, it is very important that published research concerning MBR activated sludge filterability indicates exactly where the measured sludge samples were collected. In addition, the applied return ratio, between membrane tanks and carbonaceous and nutrient removal tanks, in the case of an MBR installation with a separate membrane tank, should also be mentioned. Analyzing the variation of filterability between the various tanks of one full-scale MBR installation, in case the MBR installation has heterogeneous filterability, in a systematic way could be very valuable. If filterability measurements were coupled with suitable parameters to study flocculation conditions of the activated sludge, the development of the flocs, within one MBR installation, would be known. This knowledge could contribute to the description of optimal flocs for MBR filtration, adapted to the operating conditions of full-scale MBR installations. Effect of submicron particles on filterability

The effect of submicron particles on activated sludge filterability should be further investigated. Fundamental research on the effect of submicron particles on MBR filtration would, at this point, provide the most reliable results to identify which submicron particle size is relevant for MBR filtration. Our diluting experiments indicated that deflocculation may occur for particles smaller than 0.5 µm. Therefore, the applied measuring methods should also provide information concerning submicron particles smaller than 0.5 µm. The effect of submicron particles on filterability should be analyzed, after a fundamental approach, through an applied approach based on real MBR activated sludge. In this situation, the exact definition of the sample preparation methodology and the effect of waiting time between sample preparation and measurement should be thoroughly analyzed. SMP measuring methods

The effect of SMP on filterability should be preceded by an evaluation of reliable measuring methods capable of distinguishing between relevant and irrelevant SMP for MBR filtration. Such evaluation should be performed with MBR activated sludge collected from the membrane tanks of MBR installations.

178

Critical MLSS concentration

In this research, a critical MLSS concentration, in terms of activated sludge structure, was defined. This definition relied on filterability measurements of activated sludge collected mainly from one full-scale MBR installation, with heterogeneous filterability and operating with high MLSS concentrations. Filterability measurements in other MBR installations, with a separate membrane tank and applying high MLSS concentrations, should be repeated to evaluate the absolute value of critical MLSS concentration found in this research. Filterability measurements, resulting from other methods besides the DFCm, would also provide valuable information. The aforementioned research could contribute to define a standard methodology to identify the critical MLSS concentration. In this research, the critical MLSS concentration was identified through diluting and concentrating experiments, but other possibilities also exist, such as evaluating the filterability in all tanks of one MBR installation with heterogeneous filterability between the various MBR tanks. MBR activated sludge structure

In this research, a different activated sludge structure for activated sludge with high and low MLSS concentrations was identified through filterability, SMP concentrations and viscosity measurements. The different activated sludge structures, in full-scale MBR activated sludge, should be further investigated through morphological parameters. Considering that our hypothesis on the effect of the MLSS concentration on MBR filtration relies on the structure of the bulk activated sludge matrix, it is relevant that other research investigations on the subject select analytical methods where the matrix of the analyzed activated sludge is preserved. MBR activated sludge quality with high MLSS concentration

At present, MBRs with separate membrane tanks are common, so an optimization of the operating conditions at these installations can be applied. In this research, an improvement of filterability in MBR activated sludge with high MLSS concentrations was measured. However, activated sludge with high MLSS concentrations is not advisable, mainly due to clogging of the aeration ports and increased aeration. Therefore, aeration ports, specifically designed for MBR technology applying high MLSS concentrations, should be further investigated. Aeration in an MBR installation with a separate membrane tank is used for multiple purposes, such as scouring the membranes and providing DO to the activated sludge. However, if the functions of aeration could be separated, similar to what occurs in side-stream MBR systems, aeration for membrane scouring could be optimized. These optimizations would also have to consider that the removal of carbonaceous material and nutrients should be successful in the MBR tanks specifically designed for those purposes. Furthermore, the return ratio, between membrane tanks and carbonaceous and nutrient removal tanks, should be low to promote floc growth in the membrane tanks and avoid homogenization of the activated sludge characteristics between all tanks of the MBR. The design and operation of MBR installations, with a separate membrane tank intended to

179

operate at high MLSS concentrations, require further optimization, particularly in terms of aeration and return ratio. MBR permeate quality

In this research, MBR permeate quality was investigated through particle counting in the range of 2-100 µm. The applied method does not provide information about the nature of the measured permeate particles. However, the exact nature of the permeate particles is relevant, especially to optimize cleaning procedures in MBR installations and when the MBR permeate is reused. MBR technology is particularly suitable for reuse purposes, so the exact nature of the permeate particles should be investigated.

9.13 Conclusions

The main conclusions drawn from the research described in this thesis can be summarized as follows:

- The effect of MLSS concentration on MBR filterability depends on the applied MLSS concentration range, which can be explained through the MBR activated sludge structure. MBR activated sludge with good to moderate filtration quality and high MLSS concentration, i.e. exceeding a critical value, is capable of retaining fouling particles in its matrix. In the aforementioned case increasing the MLSS concentration contributes to improve filterability. In contrast, activated sludge with low MLSS concentration, i.e. below a critical MLSS concentration, is unable to retain fouling particles in its matrix. In the abovementioned case, filterability improves by applying MLSS concentrations as low as possible.

- A critical MLSS concentration of 10.5 g/L was obtained in this research. - The structure of bulk activated sludge with high MLSS concentration and good to

moderate filtration quality is probably composed of a network of particles linked by SMP and/or bound EPS, behaving as a fine size mesh. The structure of activated sludge with low MLSS concentration is likely to be composed of fragmented particles with unlinked SMP and/or bound EPS.

- Filterability improves with increasing floc size. However, in the membrane tank of one MBR installation, variations in floc size are limited, probably due to aeration restrictions.

- Improved MBR design and operation control would be achieved by placing the membranes in aeration tanks, intended for carbonaceous and nutrient removal, and operating at low MLSS concentration.

- Low return ratios are preferable for MBR operation, in MBR installations with a separate membrane tank operating at high MLSS concentration.

- A methodology was defined to verify membrane integrity and simultaneously evaluate growth and accumulation of particles in MBR permeate.

- There are always particles, in the size range of 2-100 µm, in MBR permeate. - SMP methods are probably quantifying all SMP present in the samples, without

distinguishing between relevant and irrelevant SMP for MBR filtration. - The volume of submicron particles is very likely to be a good indicator of

activated sludge filterability.

180

XVII

Bibliography

Abbassi, B., S. Dullstein and N. Rabinger (1999). "Minimization of excess sludge

production by increase of oxygen concentration in activated sludge flocs; experimental and theoretical approach." Water Reseach 34: 139-146.

APHA (2000). Standard methods for the examination of water and wastewater, American Public Health Association, USA.

Atasoy, E. (2008). Filtration by a flat sheet submerged filtration unit, Technical University of Delft, The Netherlands,Tubitak, Turkey.

Bae, T. H. and T. M. Tak (2005). "Interpretation of fouling characteristics of ultrafiltration membranes during the filtration of membrane bioreactor mixed liquor." Journal of Membrane Science 264(1-2): 151-160.

Belfort, G., R. H. Davisb and A. L. Zydney (1994). "The behavior of suspensions and macromolecular solutions in crossflow microfiltration." Journal of Membrane Science (96): 1-58.

Bitton, G. (1999). Wastewater microbiology, Wiley-Liss, Inc, New York. Boller, M. and R. Kaegi (2011). Characterization of aquatic particles. Handbook on

particle separation processes. A. van Nieuwenhuijzen and J. van der Graaf, IWA:

288. Bouhabila, E. H., R. Ben Aim and H. Buisson (1998). "Microfiltration of activated sludge

using submerged membrane with air bubbling (application to wastewater treatment)." Desalination 118(1-3): 315-322.

Brepols, C., H. Schafer and N. Engelhardt (2009). Economic aspects of large scale membrane bioreactors. Final MBR-network workshop, Berlin.

Ceronio, A. D. and J. Haarhoff (2002). "Dealing with large particle counting data sets." Water Supply 2(5-6): 6.

Ceronio, A. D. and J. Haarhoff (2005). "An improvement on the power law for the description of particle size distributions in potable water treatment." Water Research (39): 305–313.

Chang, I. S. and S. N. Kim (2005). "Wastewater treatment using membrane filtration - effect of biosolids concentration on cake resistance." Process Biochemistry 40(3-4): 1307-1314.

Chang, I. S., C. H. Lee and K. H. Ahn (1999). "Membrane Filtration Characteristics in Membrane-Coupled Activated Sludge System: The Effect of Floc Structure on Membrane Fouling." Separation Science and Technology 34(9): 1743-1758.

Cho, B. D. and A. G. Fane (2002). "Fouling transients in nominally sub-critical flux operation of a membrane bioreactor." Journal of Membrane Science 209(2): 391-403.

Cho, J., K. G. Song, H. Yun, K. H. Ahn, J. Y. Kim and T. H. Chung (2005a). "Quantitative analysis of biological effect on membrane fouling in submerged membrane bioreactor." Water Science and Technology 51(6-7): 9-18.

Cho, J. W., K. G. Song and K. H. Ahn (2005b). "The activated sludge and microbial substances influences on membrane fouling in submerged membrane bioreactor: unstirred batch cell test." Desalination 183(1-3): 425-429.

XVIII

Cho, M. H., C. H. Lee and S. Lee (2005c). "Influence of floc structure on membrane permeability in the coagulation-MF process." Water Science and Technology 51(6-7): 143-150.

Chung, T. S., N. Widjojo, N. Peng and S. Bonyadi (2008). The evolutional understanding of membrane formation: from the origins of macrovoids and irregular membrane contour to the formation of macrovoid-free membrane. Engineering with Membranes 2008, EMS, Vale de Lobo, Portugal.

Cicek, N., M. T. Suidan, P. Ginestet and J.-M. Audic (2003). "Impact of soluble organic compounds on permeate flux in an aerobic membrane bioreactor." Environmental Technology 24(2): 249-256.

Cicek, N., H. Winnen, M. T. Suidan, B. E. Wrenn, V. Urbain and J. Manem (1998). "Effectiveness of the Membrane Bioreactor in the biodegradation of high molecular weight compounds." Water Reseach 32(5): 1553-1563.

De la Torre, T., V. Inversen, A. A. Moreau and J. Stuber (2009). "Filtration characterization methods in MBRs systems: a pratical comparison." Desalination and water treatment 9: 15-21.

Defrance, L. and M. Y. Jaffrin (1999). "Reversibility of fouling formed in activated sludge filtration." Journal of Membrane Science 157(1): 73-84.

Drews, A., J. Mante, V. Iversen, M. Vocks, B. Lesjean and M. Kraume (2007). "Impact of ambient conditions on SMP elimination and rejection in MBRs." Water Research 41(17): 3850-3858.

Drews, A., M. Vocks, U. Bracklow, V. Iversen and M. Kraume (2008). "Does Fouling in MBR Depend on SMP?" Desalination 231: 141-149.

Drews, A., M. Vocks, V. Iversen, B. Lesjean and M. Kraume (2006). "Influence of unsteady membrane bioreactor operation on EPS formation and filtration resistance." Desalination 192(1-3): 1-9.

Dubois, M., K. A. Gilles, J. K. Hamilton, P. A. Rebers and P. Smith (1956). "Colorimetric method for determination of sugars and related substances." Analytical Chemistry 28: 350-356.

EC (2010a). 2009 Environment Policy Review, European Commission Environment, Luxemburg.

EC (2010b). Environment statistics and accounts in Europe, European Commission Eurostat, Luxemburg.

Evenblij, H. (2006). Filtration Characteristics in Membrane Bioreactors. Water Management Department Delft University of Technology. PhD: 209.

Evenblij, H., S. Geilvoet, J. H. J. M. van der Graaf and H. F. van der Roest (2005). "Filtration characterisation for assessing MBR performance: three cases compared." Desalination 178: 115-124.

Fang, H. H. P. and X. L. Shi (2005). "Pore fouling of microfiltration membranes by activated." Journal of Membrane Science 264(1-2): 161-166.

Fenu, A., O. Braimah, J. W. Mulder, M. Kennedy and G. Amy (2009). Organic matter in MBR sludge waters: foulant identification and cleaning considerations. Final MBR-network workshop, Berlin, Germany.

Geilvoet, S. (2010). The Delft Filtration characterisation method, Assessing membrane bioreactor activated sludge filterability. Water Management Department. Delft, Delft University of Technology. PhD: 243.

XIX

Geilvoet, S., A. A. Moreau, M. Lousada-Ferreira, A. van Nieuwenhuijzen and J. H. J. M. van der Graaf (2007a). Filtration characterization, SMP analyses and particle size distribution in the submicron range of MBR activated sludge. IWA Particle separation 2007, Toulouse, France.

Geilvoet, S. P., A. A. Moreau, M. Lousada-Ferreira, A. F. van Nieuwenhuijzen and J. H. J. M. van der Graaf (2007b). Filtration Characterization and sludge quality monitoring at a full-scale MBR during its first year of operation. IWA International conference on membranes for water and wastewater treatment, Harrogate, United Kingdom.

Germain, E. and T. Stephenson (2005). "Biomass characteristics, aeration and oxygen transfer in membrane bioreactors: their interrelations explained by a review of aerobic biological processes." Reviews in Environmental Sience and Bio/Technology 4: 223-233.

Gregory, J. (1998). "The Role of Floc Density in Solid-liquid Separation." Filtration & Separation(May): 367-366.

Guo, H., Y. Wyart, J. Perot, F. Nauleau and P. Moulin (2010). "Low-pressure membrane integrity tests for drinking water treatment: A review." Water Reseach 44(210): 41-57.

Gur-Reznik, S., I. Katz and C. G. Dosoretz (2008). "Removal of dissolved organic matter by granular-activated carbon adsorption as a pretreatment to reverse osmosis of membrane bioreactor effluents." Water Research 42(6-7): 1595-1605.

Han, S. S., T. H. Bae, G. G. Jang and T. M. Tak (2005). "Influence of sludge retention time on membrane fouling and bioactivities in membrane bioreactor system." Process Biochemistry 40(7): 2393-2400.

Harada, H., K. Momonoi, S. Yamazaki and S. Takizawa (1994). "Application of anaerobic membrane reactor for treatment of a wastewater containing high strenght particulate organics." Water Science and Technology 30(12): 307-319.

Hargesheimer, E. E. and C. M. Lewis (1998). Particle Counting: How, Why, Where & What Equipment, Alberta Water & Wastewater Operators Association, Calgary, Canada.

Hasar, H., C. Kinaci, A. Ünlü, H. Togrul and U. Ipek (2004). "Rheological properties of activated sludge in a sMBR." Biochemical Engineering Journal (20): 1-6.

Huang, X., P. Gui and Y. Qian (2001). "Effect of sludge retention time on microbial behaviour in a submerged membrane bioreactor." Process Biochemistry 36(10): 1001-1006.

Huisjes, E. H., K. Colombel and B. Lesjean (2009). The European Market: specificities and future trends. Final MBR-Network workshop, Berlin, Germany.

Huyskens, C., E. Brauns, E. Van Hoof and H. De Wever (2008). "A new method for the evaluation of the reversible and irreversible fouling propensity of MBR mixed liquor." Journal of Membrane Science 323(1): 185-192.

Itonaga, T., K. Kimura and Y. Watanabe (2004). "Influence of suspension viscosity and colloidal particles on permeability of membrane used in membrane bioreactor (MBR)." Water Science and Technology 50(12): 301-309.

Ivanovic, I. and T. Leiknes (2008). "Impact of aeration rates on particle colloidal fraction in the biofilm membrane bioreactor (BF-MBR)." Desalination 231(1-3): 182-190.

XX

Iversen, V., S. Hermann, A. Drews, R. E. Mehrez, M., M. Jekel, B. Lesjean and M. Kraume (2009a). Effects of various flux enhancers in MBR: lab scale vs. pilot scale operation. Final MBR-network workshop, Berlin, Germany.

Iversen, V., H. Koseoglu, N. O. Yigit, A. Drews, M. Kitis, B. Lesjean and M. Kraume (2009b). "Impacts of membrane flux enhancers on activated sludge respiration and nutrient removal in MBRs." Water Reseach 43(3): 822-830.

Jefferson, B., A. Brookes, P. Le Clech and S. J. Judd (2004). "Methods for understanding organic fouling in MBRs." Water Science and Technology 49(2): 237-244.

Ji, L. and J. Zhou (2006). "Influence of aeration on microbial polymers and membrane fouling in submerged membrane bioreactors." Journal of Membrane Science 276: 168-177.

Judd, S. (2005). "Fouling control in submerged membrane bioreactors." Water Science and Technology 51(6-7): 27-34.

Judd, S. (2006). The MBR book- Principles and Applications of Membrane Bioreactors in Water and Wastewater Treatment, Elsevier, Great Britain.

Judd, S. (2008). "The status of membrane bioreactor technology." Trends in Biotechnology 26(2): 109-116.

Kaminski, I., N. Vescan and A. Adin (1997). "Particle size distribution and watewater filter performance." Water Science and Technology 36(4): 217-224.

Kimura, K., T. Naruse and Y. Watanabe (2009). "Changes in characteristics of soluble microbial products in membrane bioreactors associated with different solid retention times: Relation to membrane fouling." Water Research 43(4): 1033-1039.

Koseoglu, H., N. O. Yigit, V. Iversen, A. Drews, M. Kitis, B. Lesjean and M. Kraume (2008). "Effects of several different flux enhancing chemicals on filterability and fouling reduction of membrane bioreactor (MBR) mixed liquors." Journal of Membrane Science 320(1-2): 57-64.

Kraume, M., D. Wedi, J. Schaller, V. Iversen and A. Drews (2009). "Fouling in MBR: What use are lab investigations for full scale operation?" Desalination 236(1-3): 94-103.

Laera, G., C. Giordano, A. Pollice, D. Saturno and G. Mininni (2007). "Membrane bioreactor sludge rheology at different solid retention times." Water Research 41(18): 4197-4203.

Lawler, D. F. (1997). "Particle size distributions in treatment processes:theory and practice." Water Science and Technology(36): 15-23.

Le-Clech, P., V. Chen and T. A. G. Fane (2006). "Fouling in membrane bioreactors used in wastewater treatment." Journal of Membrane Science (284): 17-53.

Le-Clech, P., B. Jefferson and S. Judd (2003). "Impact of aeration, solids concentration and membrane characteristics on the hydraulic performance of a membrane bioreactor." Journal of Membrane Science (218): 117–129.

Lebleu, N., C. Roques, P. Aimar and C. Causserand (2009). "Potable water production by membrane processes: membrane characterization using a series of bacterial strains." Water Science and Technology: Water Supply 9(4): 405-412.

Lee, J., W. Y. Ahn and C. H. Lee (2001). "Comparison of the filtration characteristics between attached and suspended growth microorganisms in submerged membrane bioreactor." Water Research 35(10): 2435-2445.

XXI

Lee, S. A., A. G. Fane, R. Amal and T. D. Waite (2003). "The effect of floc size and structure on specific cake resistance and compressibility in dead-end microfiltration." Separation science and technology 38(4): 869-887.

Lee, Y., J. Cho, Y. Seo, J. W. Lee and K. H. Ahn (2002). "Modeling of submerged membrane bioreactor process for wastewater treatment." Desalination 146(1-3): 451-457.

Leiknes, T., I. Ivanovic and H. Ødegaard (2006). Investigating the effect of colloids on the performance of a biofilm membrane reactor (BF-MBR) for treatment of municipal wastewater. Water Institute of South Africa (WISA) Biennial Conference. Durban, South Africa, Water SA (on-line): 708-714.

Leiknes, T. and H. Odegaard (2007). "The development of a biofilm membrane bioreactor." Desalination 202(1-3): 135-143.

Lesjean, B. and E. H. Huisjes (2008). "Survey of the European MBR market: trends and perspectives." Desalination 231: 71-81.

Lesjean, B., S. Rosenberger, C. Laabs, M. Jekel, R. Gnirss and G. Amy (2005). "Correlation between membrane fouling and soluble/colloidal organic substances in membrane bioreactors for municipal wastewater treatment." Water Science and Technology 51(6-7): 1-8.

Liu, R., X. Huang, Y. F. Sun and Y. Qian (2003). "Hydrodynamic effect on sludge accumulation over membrane surfaces in a submerged membrane bioreactor." Process Biochemistry(39): 157-163.

Lowry, O. H., N. J. Rosebourgh, A. R. Farr and R. J. Randall (1951). "Protein measurement with the folin phenol reagent." Journal of Biological Chemistry 193: 265-275.

Lyko, S., T. Wintgens, D. Al-Halbouni, S. Baumgarten, D. Tacke, K. Drensla, A. Janot, W. Dott, J. Pinnekamp and T. Melin (2008). "Long-term monitoring of a full-scale municipal membrane bioreactor - Characterisation of foulants and operational performance." Journal of Membrane Science 317(1-2): 78-87.

Madaeni, S. S., A. G. Fane and D. E. Wiley (1999). "Factors influencing critical flux in membrane filtration of activated sludge." Journal of Chemical Technology and Biotechnology(74): 539-543.

Masse, A., M. Sperandio and C. Cabassud (2006). "Comparison of sludge characteristics and performance of a submerged membrane bioreactor and an activated sludge process at high solids retention time." Water Research 40(12): 2405-2415.

Meng, F., B. Shi, F. Yang and H. Zhang (2007). "New insights into membrane fouling in submerged membrane bioreactor based on rheology and hydrodynamics concepts." Journal of Membrane Science(302): 8.

Meng, F. G., S. R. Chae, A. Drews, M. Kraume, H. S. Shin and F. L. Yang (2009). "Recent advances in membrane bioreactors (MBRs): Membrane fouling and membrane material." Water Research 43(6): 1489-1512.

Meng, F. G., H. M. Zhang, F. L. Yang, S. T. Zhang, Y. S. Li and X. W. Zhang (2006). "Identification of activated sludge properties affecting membrane fouling in submerged membrane bioreactors." Separation and Purification Technology 51(1): 95-103.

Metcalf & Eddy, I. (2003). Wastewater engineering, Treatment and Reuse, Mc Graw-Hill, New York.

XXII

Min, K. N., S. J. Ergas and A. Mermelstein (2008). "Impact of dissolved oxygen concentration on membrane filtering resistance and soluble organic matter characteristics in membrane bioreactors." Water Science and Technology 57(2): 161-165.

Miska-Markush, V. (2009). Effluent filtration for more than particle removal. Water Management Department. Delft, Delft University of Technology. PhD: 256.

Miyoshi, T., T. Tsuyuhara, R. Ogyu, K. Kimura and Y. Watanabe (2009). "Seasonal variation in membrane fouling in membrane bioreactors (MBRs) treating municipal wastewater." Water Research 43(20): 5109-5118.

Moreau, A. A. (2010). Filterability Assessment of Membrane Bioreactors at European Scale. Water Management Department Delft, Delft University of Technology. PhD: 211.

Mulder, J. W., F. Kramer and E. van Sonsbeek (2005). Hydrid MBR-the perfect upgrade for Heenvliet. H2O, Rinus Visser. April: 66-68.

Neis, U. and A. Tiehm (1997). "Particle size analysis in primary and secondary waste water effluents." Water Science and Technology 36(4): 151-158.

Ng, H. Y., T. W. Tan and S. L. Ong (2006). "Membrane fouling of submerged membrane bioreactors: Impact of mean cell residence time and the contributing factors." Environmental Science & Technology 40(8): 2706-2713.

Oxtoby, S. (2003). "Experiences in integrity testing ultrafiltration membranes at a large potable water treatment works." Water Science and Technology: Water Supply 3(5-6): 1-7.

PBL. (2003). "Nature 2000 and the National Ecological Network." Retrieved 29-3-2011, 2011.

Phattaranawik, J., A. G. Fane and F. S. Wong (2008). "Novel membrane-based sensor for online membrane integrity monitoring." Journal of Membrane Science 323(1): 113-124.

Poyatos, J. M., M. Molina-Munoz, J. Gonzalez-Lopez and E. Hontoria (2010). "Particle Size Distribution in Permeate from a Pilot-Scale Membrane Bioreactor System under Actual Operating Conditions." Particulate Science and Technology 28(4): 309-318.

Psoch, C. and S. Schiewer (2006). "Anti-fouling application of air sparging and backflushing for MBR." Journal of Membrane Science 283(1-2): 273-280.

Qin, H. J., K. A. Kekre, G. H. Tao, M. H. Oo, M. N. Wai, T. C. Lee, B. Viswanath and H. Seah (2006). "New option of MBR-RO process for production of NEWater from domestic sewage." Journal of Membrane Science 272(1-2): 70-77.

Reid, E., X. R. Liu and S. J. Judd (2008). "Sludge characteristics and membrane fouling in full-scale submerged membrane bioreactors." Desalination 219(1-3): 240-249.

Remy, M., P. van der Marel, A. Zwijnenburg, W. Rulkens and H. Temmink (2009). "Low dose powdered activated carbon addition at high sludge retention times to reduce fouling in membrane bioreactors." Water Research 43(2): 345-350.

Rosenberger, S., H. Evenblij, S. T. Poele, T. Wintgens and C. Laabs (2005). "The importance of liquid phase analyses to understand fouling in membrane assisted activated sludge processes - six case studies of different European research groups." Journal of Membrane Science 263(1-2): 113-126.

XXIII

Rosenberger, S. and M. Kraume (2002a). "Filterability of activated sludge in membrane bioreactors." Desalination(146): 373-379.

Rosenberger, S., U. Kruger, R. Witzig, W. Manz, U. Szewzyk and M. Kraume (2002b). "Performance of a bioreactor with submerged membranes for aerobic treatment of municipal waste water." Water Research(36): 413-420.

Rosenberger, S., K. Kubin and M. Kraume (2002c). "Rheology of Activated Sludge in Membrane Bioreactors." Eng. Life Sci. 2(9): 269-275.

Sombatsompop, K., C. Visvanathan and R. Ben Aim (2006). "Evaluation of biofouling phenomenon in suspended and attached growth membrane bioreactor systems." Desalination (201): 12.

STORA (1979). Handleiding voor microscopisch slibonderzoek (in Dutch), Stichting Toegepast Onderzoek Reiniging Afvalwater, Rijswijk: 87.

Stricot, M., A. Filali, N. Lesage, M. Sperandio and C. Cabassud (2010). "Side-stream membrane bioreactors: Influence of stress generated by hydrodynamics on floc structure, supernatant quality and fouling propensity." Water Research 44(7): 2113-2124.

te Poele, S. (2005). Foulants in ultrafiltration of wwtp effluent. Water Management Department Delft University of Technology. PhD.

Trussell, R. S., R. P. Merlo, S. W. Hermanowicz and D. Jenkins (2007). "Influence of mixed liquor properties and aeration intensity on membrane fouling in a submerged membrane bioreactor at high mixed liquor suspended solids concentrations." Water Research 41(5): 947-958.

UNITN (2006). State of the art and definition on a common basis: parameter definition for sludge components, Universita degli Studi di Trento, Trento, Italy.

van der Graaf, J. H. J. M., J. de Koning and J. Tang (2001). "Particle size analysis as a tool for performance measurements in high rate effluent filtration." Water Science and Technology 43(10): 303-310.

van Nieuwenhuijzen, A. F., H. Evenblij, C. A. Uijterlinde and F. L. Schulting (2008). "Review on the state of science on membrane bioreactors for municipal wastewater treatment." Water Science and Technology 57(7): 979-986.

Verberk, J. Q. J. C., K. J. O’Halloran, L. A. Hamilton, J. H. G. Vreeburg and J. C. J. van Dijk (2007). "Measuring particles in drinking water transportation systems with particle counters." Journal of Water Supply: Research and Technology-AQUA 56(5): 345-355.

Vreeburg, J. H. G., D. Schippers, J. Q. J. C. Verberk and J. C. van Dijk (2008). "Impact of particles on sediment accumulation in a drinking water distribution system." Water Research 42(16): 4233-4242.

Wedi, D. (2006). The effect of cold temperatures on MBR performances. EUROMBRA workshop: Bio-Fouling in Membrane Systems, NTNU, Trondheim, Norway.

WEF (2006). Membrane systems for wastewater treatment, Water Enviroment Federation Press, McGraw-Hill, USA.

WHO. (2009). "Water scarcity- Fact 2." 10 Facts about water scarcity Retrieved 29-3-11, 2011.

Wisniewski, C. and A. Grasmick (1998). "Floc size distribution in a membrane bioreactor and consequences for membrane fouling." Colloids and Surfaces A: Physicochemical and Engineering Aspects(138): 403-411.

XXIV

Wu, J. L. and X. Huang (2009). "Effect of mixed liquor properties on fouling propensity in membrane bioreactors." Journal of Membrane Science 342(1-2): 88-96.

Yamamoto, K., M. Hiasa, T. Mahmood and T. Matsuo (1989). "Direct Solid-Liquid Separation Using Hollow Fiber Membrane in an Activated-Sludge Aeration Tank." Water Science and Technology 21(4-5): 43-54.

Yang, F., A. Bick, S. Shandalov, A. Brenner and G. Oron (2009). "Yield stress and rheological characteristics of activated sludge in an airlift membrane bioreactor." Journal of Membrane Science 334(1-2): 83-90.

Yang, Z., X. F. Peng, M. Y. Chen, D. J. Lee and J. Y. Lai (2007). "Intra-layer flow in fouling layer on membranes." Journal of Membrane Science 287(2): 280-286.

XXV

Curriculum vitae

Surname(s): Lousada Ferreira First name(s): Maria da Conceição Date of birth: 11/11/72 Nationality: Portuguese

Maria (da Conceição) Lousada Ferreira completed high school at Liceu de Oeiras (High School of Oeiras), Portugal, in 1990. She graduated as an Environmental Engineer- Branch Environment in 1995 at Faculty of Science and Technology of Universidade

Nova de Lisboa (New University of Lisbon). After obtaining her licentiate degree she worked as an Environmental Engineer, at Associação National de Criadores de Raça

Churra (National Association for Breeders of Churra Strain) and Parque Natural do

Douro International (Natural Park of International -river- Douro). In 2000, Maria Lousada Ferreira started as teaching assistant/researcher at the Civil Engineering Department of the University of Beira Interior, Portugal, where she investigated biodegradation kinetics of wastewater treatment effluents through respirometry. In 2006, Maria Lousada Ferreira started her PhD studies at the Water Management Department of the Delft University of Technology under the supervision of Prof. Ir. J. H. J. M. van der Graaf. Publications (Peer reviewed journals)

Lousada-Ferreira M., Moreau A., van Lier J.B., van der Graaf J.H.J.M. (2011) Particle counting as a tool to predict filterability in membrane bioreactors activated sludge?, Water Science and Technology 64-1, 139-146 Lousada-Ferreira M., Geilvoet S.P, Moreau A.A., Atasoy A.E., Krzeminski P.W., van Nieuwenhuijzen A., van der Graaf J.H.J.M. (2010) MLSS concentration: still a poorly understood parameter in MBR filterability, Desalination 250, 618-622 Van den Broeck R., Krzeminski P., Van Dierdonck J., Gins G., Lousada-Ferreira M., Van Impe J.F.M., van der Graaf J.H.J.M f, Smets I.Y.,van Lier, J.B. (-) Activated sludge characteristics affecting sludge filterability in municipal and industrial MBRs: unraveling correlations using multi-component regression analysis, Accepted for publication in May 2011, Journal of Membrane Science, doi:10.1016/j.memsci.2011.05.010 Silva A.F., Carvalho G., Oehmen A., Lousada-Ferreira M., van Nieuwenhuijzen A., Reis M.A.M., Crespo T.B. (-) Microbial population analysis of nutrient removal-related organisms in membrane bioreactors, Accepted for publication in July 2011, Applied Microbiology and Biotechnology, doi: 10.1007/s00253-011-3499-5

XXVI

Publications (Conference Proceedings)

Lousada-Ferreira M., Langhorst W., de Kreuk M., van der Graaf J. (2010) The Heenvliet hybrid membrane bioreactor: comparison between series and parallel operation, Singapore water week, SWW-IWA, Singapore, June/July. Lousada-Ferreira M., Moreau A., van Nieuwenhuijzen A., van der Graaf J.H.J.M. (2009) Characterizing municipal MBR activated sludge quality, Membrane technology conference, IWA, Beijing, China, September. Moreau A., Lousada-Ferreira M., van Nieuwenhuijzen A., van der Graaf J.H.J.M. (2009) Overview of MBR activated sludge filterability at European Scale, Membrane technology conference, IWA, Beijing, China, September. Lousada-Ferreira M., van Nieuwenhuijzen A., van der Graaf J.H.J.M. (2009) Characterizing full scale municipal MBR activated sludge quality through filterability and particle counting, Particle separation and nanoparticles conference, IWA, Durham, USA, June. van der Graaf J.H.J.M., van Nieuwenhuijzen A., Geilvoet S., Moreau A., Lousada-Ferreira M., Krzeminski P. (2009) The European MBR tour of the Delft Filtration Characterization method, Final MBR network workshop, Berlin, Germany, March/April. Moreau A., Lousada-Ferreira M., van Nieuwenhuijzen A., van der Graaf J.H.J.M (2009) Seasonal effect: impact on MBR sludge filterability depending on membrane configurations, Final MBR network workshop, Berlin, Germany, March/April. Silva A.F., Carvalho G., Lousada-Ferreira M., van Nieuwenhuijzen A., Guglielmi, Crespo J.G., Reis M.A.M., Crespo M.T.B. (2009) Microbial population structure of pilot- and full-scale membrane bioreactors, Final MBR network workshop, Berlin, Germany, March/April. Lousada-Ferreira M., Moreau A., Geilvoet S.P., Krzeminski P., van Nieuwenhuijzen A., van der Graaf J.H.J.M. (2008) In-situ monitoring of sludge filterability, MBR- network workshop at Engineering with Membranes conference, EMS, Vale do Lobo, Portugal, May. Geilvoet S., Moreau A.A., Lousada-Ferreira M., van Nieuwenhuijzen A., van der Graaf J.H.J.M. (2007) MBR activated sludge filterability and quality alteration under stress conditions, IMSTEC 07, 6th International membrane science and technology conference, IWA, Sydney, Australia, November. Geilvoet S., Moreau A.A., Lousada-Ferreira M., van Nieuwenhuijzen A., van der Graaf J.H.J.M. (2007) Filtration characterization, SMP analyses and particle size distribution in the submicron range of MBR activated sludge, Particle separation conference, IWA, Toulouse, France, July.

XXVII

Moreau A.A., Geilvoet S., Lousada-Ferreira M., van Nieuwenhuijzen A., van der Graaf J.H.J.M. (2007) Effect of various parameters on the filtration characteristics of MBR sludge, 4th International conference on membranes for water and wastewater treatment, IWA, Harrogate, UK, May. Geilvoet S., Moreau A.A., Lousada-Ferreira M., van Nieuwenhuijzen A., van der Graaf J.H.J.M. (2007) Filtration characterization and sludge quality monitoring at a full-scale MBR during its first year of operation, 4th International conference on membranes for water and wastewater treatment, IWA, Harrogate, UK, May.

XXVIII

XXIX

Annex A

Relevant Operational Parameters in Submerged Membrane Bioreactors4

- Trans-membrane pressure (TMP): The driving force for permeation for most

water treatment membrane filtration applications

feed permeateTMP P P P= ∆ = −

- Flux (J): Quantity of material passing through a unit area of membrane per unit time. Flux as described by Darcy’s law:

tP R

PJ

η

∆=

Where: ∆P - Trans-membrane pressure [Pa or bar];

ηp - permeate dynamic viscosity [Pa.s], usually considered equal to water;

Rt - total filtration resistance [m-1]; J - flux[L/m2.h or m/s].

- Total Resistance (Rt): Includes the following components: membrane resistance and resistance of the fouling layer, adsorbed by the membrane surface.

fmt RRR +=

Where: Rm - clean membrane resistance [m-1]; Rf - fouling resistance [m-1].

- Critical flux (Jc): There are different definitions for critical flux: the flux obtained during sub-critical flux is equal to the clean water flux measured under the same conditions; the sub-critical flux is the flux rapidly established and maintained during start-up of filtration, but is not necessarily equal to clean water flux; sub-critical operation is when a constant permeability is maintained for an extended period of time, even when preceded by an initial decline in flux.

- Selectivity: Defined by the pore size of the membrane; sometimes referred to as the Molecular Weight Cut-Off (MWCO) i.e. the molecular weight of a solute corresponding to a 90% rejection factor for a given membrane, where the rejection factor (R) is defined as follows:

feedi

permeatei

c

cR

,

,1−=

Where: R - rejection factor [-]

ci,permeate – concentration of component i in permeate [mg/L] ci,feed – concentration of component i in feed [mg/L]

MWCO is mainly used for membranes where the determining factor is not particle size, but difference in diffusivity.

4 Mainly referring to definitions proposed by Evenblij, H. (2006). Filtration Characteristics in Membrane Bioreactors. Water Management Department Delft University of Technology. PhD: 209. ; and Judd, S. (2006). The MBR book- Principles and Applications of Membrane Bioreactors in Water and Wastewater Treatment, Elsevier, Great Britain.

XXX

- Permeability (P): The ratio between the flux and TMP [L/m2.h.bar], as follows:

TMP

JP =

Permeability can be corrected for temperature by incorporating viscosity, as follows:

ref

act

cTMP

JP

η

η⋅=

Where: Pc = permeability corrected for reference temperature [L/m2.h.bar];

ηact = actual viscosity [Pa.s]; ηref = viscosity at reference temperature [Pa.s].

- Compressibility and compressibility coefficient (s): compressibility can be understood as the compress potential of a certain cake layer expressed by the compressibility coefficient, s, varying between 0 and 1. A compressibility coefficient of 0 is obtained when no compression occurs, i.e. when the resistance is independent from compression. In contrast, a compressibility coefficient of 1 is obtained when the resistance is dependent from compression, therefore when the cake layer is highly compressible.

XXXI

Annex B

Flat Sheet Membrane Unit 5

The flat sheet membrane installation (FSu) layout is shown in Figure B.1.

Figure B.1 - Flat sheet unit layout (Atasoy 2008).

The flat sheet unit (FSu) consists of a single submerged flat sheet Toray membrane with a pore size of 0.08 µm.

FSu measuring protocol

The sludge filtration tests in the FSu are performed using a constant flux of 40 L/m2.h and an aeration rate of 0.6 Nm3/h.

5 The FSu was developed by Elif Atasoy, researcher at the Marmara Research Center, Turkey, during her one-year stay at the Department of Sanitary Engineering of TU Delft. The FSu is described herein only briefly to protect the author’s rights.

XXXII

Annex C

Results of Diluting Experiments

Comparison between Membrane and Aeration Tank Activated Sludge

Figure C.1 - ∆R20 vs. αR*ci product and s coefficient of “High MLSS” and “Low MLSS”

sets.

0

20

40

60

80

100

0 5 10 15 20

MLSS [g/L]

αR

*ci [

*10

-3m

2]

High MLSS

Low MLSS

0,0

0,1

0,2

0,3

0,4

0 5 10 15 20

MLSS [g/L]

s [

-]

High MLSS

Low MLSS

a) b)

Figure C.2 - αR*ci product (a) and s coefficient (b) vs. MLSS concentration of “High MLSS” and “Low MLSS” sets.

0,0

0,4

0,8

1,2

1,6

2,0

0 20 40 60 80 100

αR*ci [*10-3

m2]

∆R

20 [

*10

12m

-1]

0,0 0,2 0,4 0,6 0,8 1,0s [-]

aR*ci (High MLSS)

aR*ci (Low MLSS)

s (High MLSS)

s (Low MLSS)

αR*ci

αR*ci

XXXIII

0,0

0,4

0,8

1,2

1,6

2,0

0 10 20 30 40

Proteins [mg/L]

∆R

20 [

*10

12m

-1]

Bulk

Retained

0,0

0,4

0,8

1,2

1,6

2,0

0 10 20 30 40 50

Polysaccharides [mg/L]

∆R

20 [

*10

12m

-1]

Bulk

Retained

a) b)

Figure C.3 – Filterability vs. proteins a) and polysaccharides concentrations b) in bulk activated sludge and retained on DFCi membrane.

0

10

20

30

40

0 10 20 30 40

Proteins bulk [mg/L]

Pro

tein

s r

eta

ine

d [

mg

/L]

0

10

20

30

40

50

0 10 20 30 40 50

Polysaccharides bulk [mg/L]

Po

ly.

reta

ine

d [

mg

/L]

a) b)

Figure C.4 – Proteins a) and polysaccharides concentrations b) retained on the DFCi membrane vs. protein and polysaccharide concentrations in bulk activated sludge.

XXXIV

Figure C.5 - Apparent viscosity results vs. MLSS concentration and filterability.

0

10

20

30

40

50

60

70

0,5 0,6 0,7 0,8 0,9 1,0

Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[#

*103

/mL

]

Original (MT) High MLSS (2 samples)

Dilution1 (MT) High MLSS (2 samples)

Dilution 2 (MT) High MLSS (2 samples)

0

10

20

30

40

50

60

70

0,5 0,6 0,7 0,8 0,9 1,0

Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[#

*103

/mL

]

Original (AT) Low MLSS (1 sample)

Dilution1 (AT) Low MLSS (1 sample)

Dilution 2 (AT) Low MLSS (1 sample)

a) b)

Figure C.6 - Average particles/mL per particle size, for “High MLSS” sets (a) and for “Low MLSS” sets (b), by type of sample, in membrane tank (MT) and aeration tank (AT).

3

4

5

6

-0,25 -0,2 -0,15 -0,1 -0,05 0

Log (Particle size) [µm]

No

rma

lize

d p

art

icle

co

un

t[#

/(m

L* µ

m)]

0,56 0,66 0,76 0,86 0,96

Particle size [µm]

Original (MT) High MLSS(2 samples)

Dilution1 (MT) High MLSS(2 samples)

Dilution 2 (MT) High MLSS(2 samples)

Original (AT) Low MLSS (1sample)

Dilution1 (AT) Low MLSS(1 sample)

Dilution 2 (AT) Low MLSS(1 sample)

Figure C.7 - Normalized particle counting data for “Low MLSS” sets and for “High

MLSS” sets by type of sample in membrane tank (MT) and aeration tank (AT).

0

5

10

15

20

0 5 10 15 20 25 30Apparent viscosity [mPa*s]

ML

SS

[g/L

]

0,1

1

10

∆R

20 [

*10

12m

-1]

MLSS

DR20∆R20

XXXV

0,001

0,01

0,1

1

10

0 5 10 15 20

MLSS [g/L]

∆R

20[*

10

12m

-1]

0

0,02

0,04

0,06

0,08

0,1

Vo

lum

e s

ub

mic

ron

[*1

06 µ

m3/m

L] DR20 (MT) High MLSS

(2 sets)

DR20 (AT) Low MLSS(1 set)

V.sub micron (AT)LowMLSS (1 set)

V. sub micron (MT) HighMLSS (2 sets)

Figure C.8 - Average filterability and average volume of submicron particles vs. average MLSS concentration of MBR activated sludge, in membrane tank (MT) and aeration tank

(AT).

0

5

10

15

20

25

0 20 40 60 80 100

Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[#

*103

/mL

]

Original MT High MLSS (3 samples)

Dilution 1 MT High MLSS (3 samples)

Dilution 2 MT High MLSS (3 samples)

0

5

10

15

20

25

0 20 40 60 80 100

Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[#

*103

/mL

]

Original AT Low MLSS (2 samples)

Dilution 1 AT Low MLSS (2 samples)

Dilution 2 AT Low MLSS (2 samples)

a) b)

Figure C.9 - Average particles/mL per particle size, for “High MLSS” sets (a) and for “Low MLSS” sets (b) by type of sample in membrane tank (MT) and aeration tank (AT).

XXXVI

1

2

3

4

5

1 1,2 1,4 1,6 1,8 2

Log (dp) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

10 100

Particle size [µm]

Original MT High MLSS (3 samples)

Dilution 1 MT High MLSS (3 samples)

Dilution 2 MT High MLSS (3 samples)

Lower end

[-----------------------------]

Upper end

[------------]

1

2

3

4

5

1 1,2 1,4 1,6 1,8 2

Log (dp) [µm]

No

rma

lize

d p

art

icle

co

un

t [#

/(m

L* µ

m)]

10 100

Particle size [µm]

Original AT Low MLSS (2 samples)

Dilution 1 AT Low MLSS (2 samples)

Dilution 2 AT Low MLSS (2 samples)

Lower end

[-----------------------------]Upper end

[------------]

a) b)

Figure C.10 - Normalized particle counting data for “High MLSS” sets (a) and for “Low MLSS” sets (b) by type of sample in membrane tank (MT) and aeration tank (AT).

XXXVII

Comparison between Membrane Tank Activated Sludge from different Membrane

Bioreactors

Figure C.11 - ∆R20 vs. αR*ci product and s coefficient of “High MLSS” and “Low

MLSS” sets.

0

40

80

120

160

200

0 5 10 15 20

MLSS [g/L]

αR

*ci [

*10

-3m

2]

High MLSS

Low MLSS

0,0

0,1

0,2

0,3

0,4

0 5 10 15 20

MLSS [g/L]

s [

-]

High MLSS

Low MLSS

a) b)

Figure C.12 - αR*ci product (a) and s coefficient (b) vs. MLSS concentration of “High MLSS” and “Low MLSS” sets.

0

1

2

3

4

5

0 40 80 120 160 200

αR*ci [*10-3

m2]

∆R

20 [

*10

12m

-1]

0,0 0,2 0,4 0,6 0,8 1,0s [-]

aR*ci (High MLSS)

aR*ci (Low MLSS)

s (High MLSS)

s (Low MLSS)

αR*ci

αR*ci

XXXVIII

0,0

0,4

0,8

1,2

1,6

2,0

0 10 20 30

Proteins [mg/L]

∆R

20 [

*10

12m

-1]

Bulk

Retained

0,0

0,4

0,8

1,2

1,6

2,0

0 10 20Polysaccharides [mg/L]

∆R

20 [

*10

12m

-1]

Bulk

Retained

a) b)

Figure C.13 – Filterability vs. proteins a) and polysaccharides concentrations b) in bulk activated sludge and retained on DFCi membrane.

0

10

20

0 5 10 15 20 25

Proteins bulk [mg/L]

Pro

tein

s r

eta

ine

d [

mg

/L]

0

10

20

0 5 10 15 20

Polysaccharides bulk [mg/L]

Po

ly.

reta

ine

d [

mg

/L]

a) b)

Figure C.14 – Proteins a) and polysaccharides concentrations b) retained on the DFCi membrane vs. protein and polysaccharide concentrations in bulk activated sludge.

XXXIX

0

10

20

30

40

50

0,5 0,6 0,7 0,8 0,9 1,0

Particle size [µm]

Nu

mb

er

of

Pa

rtic

les

[#

*103

/mL

]

Original (Ootmarsum) (1 sample)

Dilution 1 (Ootmarsum) (1 sample)

Dilution 2 (Ootmarsum) (1 sample)

Original (Vaarseveld) (2 samples)

Dilution 1 (Vaarseveld) (2 samples)

Dilution 2 (Vaarseveld) (2 samples)

2

3

4

5

6

-0,25 -0,2 -0,15 -0,1 -0,05 0

Log (Particle size) [µm]

No

rma

lize

d p

art

icle

co

un

t[#

/(m

L* µ

m)]

0,56 0,66 0,76 0,86 0,96

Particle size [µm]

Original (Ootmarsum)(1 sample)

Dilution 1(Ootmarsum) (1sample)Dilution 2(Ootmarsum) (1sample)Original (Vaarseveld)(2 samples)

Dilution 1(Vaarseveld) (2samples)Dilution 2(Vaarseveld) (2samples)

a) b)

Figure C.15 - Average particles/mL per particle size (a) and normalized particle counting data (b) in sets obtained from the Ootmarsum and Varsseveld MBRs.

0,00

0,01

0,10

1,00

10,00

0 5 10 15 20

MLSS [g/L]

∆R

20[*

10

12m

-1]

0

0,02

0,04

0,06

0,08

0,1

Vo

lum

e s

ub

mic

ron

[*1

06 µ

m3/m

L] DR20 Ootmarsum (1

set)

DR20 Vaarseveld (2sets)

V.sub micronVaarseveld (2 sets)

V. sub micronOotmarsum (1 set)

Figure C.16 - Average filterability and average volume of submicron particles vs. average MLSS concentration of MBR activated sludge in sets obtained from the Ootmarsum and

Varsseveld MBRs.

XL

Comparison between Permeate, Tap Water and Conventional Activated Sludge

Effluent as Dilution Media

Figure C.17 - ∆R20 vs. αR*ci product and s coefficient of “High MLSS” and “Low

MLSS” sets.

0

20

40

60

80

100

0 5 10 15 20

MLSS [g/L]

αR

*ci [

*10

-3m

2]

High MLSS

Low MLSS

0,0

0,1

0,2

0,3

0,4

0 5 10 15 20

MLSS [g/L]

s [

-]

High MLSS

Low MLSS

a) b)

Figure C.18 - αR*ci product (a) and s coefficient (b) vs. MLSS concentration of “High MLSS” and “Low MLSS” sets.

0

0,2

0,4

0,6

0,8

1

0 20 40 60 80 100

αR*ci [*10-3

m2]

∆R

20 [

*10

12m

-1]

0,0 0,2 0,4 0,6 0,8 1,0s [-]

aR*ci (High MLSS)

aR*ci (Low MLSS)

s (High MLSS)

s (Low MLSS)

αR*ci

αR*ci

XLI

0

10

20

0 5 10 15 20 25 30

Proteins bulk [mg/L]

Pro

tein

s r

eta

ine

d [

mg

/L]

0

10

20

30

0 10 20 30 40

Polysaccharides bulk [mg/L]

Po

ly.

reta

ine

d [

mg

/L]

a) b)

Figure C.19 – Filterability vs. proteins a) and polysaccharides concentrations b) in bulk activated sludge and retained on DFCi membrane.

XLII

Comparison between Constant and Variable Cross-Flow Velocity

Figure C.20 - ∆R20 vs. αR*ci product and s coefficient.

0

20

40

60

80

100

0 5 10 15 20

MLSS [g/L]

αR

*ci [

*10

-3m

2]

High MLSS

Low MLSS

0,0

0,1

0,2

0,3

0,4

0 5 10 15 20

MLSS [g/L]

s [

-]

High MLSS

Low MLSS

a) b)

Figure C.21 - αR*ci product (a) and s coefficient (b) vs. MLSS concentration of “High MLSS” and “Low MLSS” sets.

0,0

0,5

1,0

1,5

2,0

0 20 40 60 80 100

αR*ci [*10-3

m2]

∆R

20 [

*10

12m

-1]

0,0 0,2 0,4 0,6 0,8 1,0s [-]

aR*ci (Low MLSS)

aR*ci (High MLSS)

s (Low MLSS)

s (High MLSS)

αR*ci

αR*ci