This article was downloaded by: [York University Libraries]On: 14 November 2014, At: 08:49Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

Environmental TechnologyPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/tent20

Optical monitoring of activated sludge flocs in bulkingand non-bulking conditionsE. Koivuranta a , J. Keskitalo b , A. Haapala a , T. Stoor a , M. Sarén c & J. Niinimäki aa Fibre and Particle Engineering Laboratory , University of Oulu , Oulu , Finlandb Control Engineering Laboratory , University of Oulu , Oulu , Finlandc Metso Automation Inc. , Kajaani , FinlandAccepted author version posted online: 11 Jul 2012.Published online: 07 Aug 2012.

To cite this article: E. Koivuranta , J. Keskitalo , A. Haapala , T. Stoor , M. Sarén & J. Niinimäki (2013) Optical monitoringof activated sludge flocs in bulking and non-bulking conditions, Environmental Technology, 34:5, 679-686, DOI:10.1080/09593330.2012.710410

To link to this article: http://dx.doi.org/10.1080/09593330.2012.710410

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Environmental Technology, 2013Vol. 34, No. 5, 679–686, http://dx.doi.org/10.1080/09593330.2012.710410

Optical monitoring of activated sludge flocs in bulking and non-bulking conditions

E. Koivurantaa∗, J. Keskitalob, A. Haapalaa, T. Stoora, M. Sarénc and J. Niinimäkia

aFibre and Particle Engineering Laboratory, University of Oulu, Oulu, Finland; bControl Engineering Laboratory, University of Oulu,Oulu, Finland; cMetso Automation Inc., Kajaani, Finland

(Received 9 May 2012; final version received 28 June 2012 )

This study introduces a novel optical monitoring method to image and characterize activated sludge flocs and to study thedependency of sludge settling properties on the floc structure. The novel method can easily analyse thousands of particles ina short timeframe using the developed image analysis program. The main advantage of this method is its applicability forin situ use because the only required pre-treatment is sample dilution. This study tested real process samples from activatedsludge plants treating wastewater from a pulp mill. The sludge samples were collected in bulking and non-bulking situations,and the image analysis results were compared to the settling speed of the samples. The structure of the activated sludge flocswas clearly different in bulking sludge situations as characterized by more fragile and elongated flocs. Additionally, excessiveamounts of filamentous bacteria hold the flocs apart, hindering sludge settling. These results show that this method is suitablefor studying and optimizing activated sludge processes.

Keywords: floc structure; image analysis; bulking sludge; filamentous bacteria; settling properties.

IntroductionThe activated sludge process (ASP) is one of the most typ-ical processes in both municipal and industrial wastewatertreatment plants (WWTPs) [1]. Similar to most biologicalprocesses, the ASP is highly sensitive to external and inter-nal variations that may lead to large environmental conse-quences because the processed water is usually dischargedto waterways [2]. Furthermore, wastewater treatment iscontinuously facing more strict environmental demands onwastewater discharge [3,4]. New innovations are needed tomeasure wastewater treatment operations performance tomanage these demands and to improve the control of ASPsystems.

The ASP consists of two stages, a biological stage (aer-ation basin) and a physical stage (secondary clarifier). Themain objective of the process is to promote the growth ofmicroorganisms that degrade organic matter into biomass,carbon dioxide and water. Microorganisms flocculate spon-taneously in the aeration tank and are removed by sedimen-tation in the secondary clarifier [5]. The high-performanceof floc formation is critical because the effluent fromsecondary clarifiers has to achieve certain environmentalstandards [6–8].

Unsatisfactory settling is a common problem in acti-vated sludge plants worldwide and is caused by bulkingsludge or poor flocculation properties [9,10]. The term‘bulking sludge’ is typically used to describe the exces-sive growth of filamentous bacteria [5]. Industrial WWTPs,

∗Corresponding author. Email: elisa.koivuranta@oulu.fi

which process waste such as pulp and paper effluents, areprone to filamentous bacteria growth [11–13]. Furthermore,the formation of pinpoint flocs (PPs) and viscous or zooglealbulking may also cause problems in sludge settling. PPs aresmall, mechanically fragile flocs that are formed when thereis a lack of filamentous bacteria [14]. Thus, filamentousbacteria are a critical population within the ASP. Poor set-tling that is not associated with filamentous bacteria is mostprobably related to viscous or zoogleal bulking caused by anexcessive amount of extracellular polysaccharides (EPSs),which are not measurable by optical imaging methods. TheEPS composition typically retains water, making the floc’sdensity close to that of water, hindering floc settling [14–17]. Although the reasons that cause poor settling propertiesare well-known, controlling this behaviour is still difficultbecause there is lack of a proper in situ imaging system forflocs.

Controlling the floc structure can solve the observedsettling problems because the flocculation of activatedsludge is critical for the whole process and the sludgeprocess have a time-varying characteristic [18]. The cor-relation between ASP performance and floc characteristicscan be determined from floc morphology; the floc sizedistribution, density and filament length are the mainfactors associated with activated sludge settling proper-ties [14–16,19,20]. Typically the studies have relied uponmicroscopy analyses to characterize the activated sludgefloc morphology and size, although it is a time-consuming,

© 2013 Taylor & Francis

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 0

8:49

14

Nov

embe

r 20

14

680 E. Koivuranta et al.

and the published results remain inconclusive in theirattempts to explain the phenomena of sludge flocculationand settling [1–3,15,16,19,21]. This variation is likely to berelated to some shortcomings in the image analysis.

The standard recommendation for microscopy imagingis to analyse at least 625 individual particles per sample[22], but most studies did not measure that number of par-ticles, therefore, the obtained results cannot be consideredstatistically reliable. Another problem is the sample prepa-ration and the transfer of flocs into a microscope [22]. Manyresearchers have developed automated image analysis sys-tems to speed up the analysis process [3,10,19,23], but even,the imaging methods proposed so far remain difficult touse in situ, which also questions how representative thesamples are [22]. Additionally, it is challenging to take arepresentative sample during the rapidly changing dynamicprocess of the ASP. Automated sludge sampling and imag-ing are therefore necessary to make in situ imaging of flocsfeasible.

Thus, the present problem in monitoring and control-ling the ASP is the lack of proper in situ measurementmethods for monitoring floc morphology. Automated opti-cal monitoring of activated sludge flocs would allow fast,reliable control of floc formation and would prevent set-tling problems caused by bulking sludge flocs or PPs. Theaim of this paper is to present a new way to character-ize activated sludge flocs with minimal pre-treatment toenable statistically accurate in situ analysis of the ASP flocproperties.

Materials and methodsActivated sludge samplingThe activated sludge samples were taken from a WWTP thattreats effluents of a pulp mill producing oxygen bleachedchemical kraft pulp. The WWTP at the mill is a fully aerobicactivated sludge plant consisting of a primary clarifier, aer-ation basins and a secondary clarifier. The average amountof treated wastewater is 32,000 m3 per day.

Collection samples were taken from the aeration tanksand were analysed on the same day. Samples were takenwhen the process had problems in sludge settling and therewere poor purification results (described as bulking sludge)and when the process performed normally and purificationresults were normal (described as non-bulking sludge). Intotal, four samples were taken in each of the bulking andnon-bulking situations.

Table 1 presents the average values of the chemical oxy-gen demand (COD) removal percentage in the ASP, the totalsolids content of effluent from the secondary clarifier and thediluted sludge volume index (DSVI) for both the bulkingsludge and non-bulking sludge situations. These analyseswere made in the laboratory of the WWTP and clearly showthe differences in the purification results between the normaland bulking sludge situations.

Table 1. Average purification values in bulking andnon-bulking situations.

Bulking Non-bulkingsludge sludge

Total solids content of effluent[mg/l]

360 20

COD removal in the activatedsludge process [%]

35 67

DSVI [mg/l] 442 160

Floc measurement environmentThe floc measurement environment, MOFI, is a small-scale research system that includes tube flow imagingwith a high-resolution charge-coupled device (CCD) cam-era. The CCD cameras image sensor is 5.0 mm × 3.7 mm(1392 × 1040 pixels) with a pixel size of approximately3.6 μ m × 3.6 μ m. It is possible to use MOFI to mea-sure specific particle features, e.g. the total particle area,the number of particles and different shape factors. Par-ticle imaging takes place in an imaging cuvette fromwhich the diluted sample is drawn by a pump withoutaffecting the sample’s properties. In this study, the sam-ples are disposed of after imaging although they can alsobe circulated through the imaging unit. Furthermore, dif-ferent samples can be filtered to obtain on-line drainageand dewatering data, e.g. the specific filtration rate. Fil-tration analysis not only provides the filtration speed butalso the intrinsic filtrate and high consistency filter cakeproperties. However, in this study, filtration analysis wasnot used. The structure of the MOFI is presented inFigure 1.

The only necessary pre-treatment to image activatedsludge flocs with MOFI is dilution. In this study, all sam-ples were diluted with deionized water at a 1 : 200 ratio fora total volume 2 L. Dilution affects the floc sizes, so it iscritical to use the same dilution every time. After the dilu-tion, samples were gently stirred to avoid floc breakage andthen processed through the imaging unit. Approximately350–400 images from every sample were taken. In a sin-gle image there were typically approximately 80 individualflocs: thus, on average over 28,000 individual flocs wereanalysed from every sample.

Image analysisAn image processing and analysis program was developedfor the automated analysis of images obtained with theMOFI. The program was developed using MATLAB 7.8.0(The MathWorks Inc., Natick, MA). Although the devel-opment was performed in MATLAB, the program can becompiled as a standalone executable. All image processingand analysis operations can be performed from a graphicaluser interface. The image analysis results are stored in atable with comma-separated values for further analysis.

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 0

8:49

14

Nov

embe

r 20

14

Environmental Technology 681

Figure 1. The structure of the floc morphology and sludge dewatering measurement environment (MOFI).

The MOFI stores images in the JPEG format withdimensions of 1392 × 1040 pixels and a colour depth of8 bits per pixel. The images are stored in a container formatthat includes the background image. The first step in pre-processing is unpacking the images obtained from a singlesample. The background is removed by dividing the sam-ple images by the background image, taken prior to eachtest run. After background removal, image processing con-tinues by separating the floc and filament objects into twobinary images.

Separation of flocs and filamentsTwo main classes of objects can be distinguished in acti-vated sludge: flocs and filaments. The abundance and shapeof objects in these two classes is expected to yield valuableinformation on sludge quality. Therefore, the classificationof objects into flocs and filaments is a critical step in theimage analysis of activated sludge. However, in activatedsludge, flocs and filaments often overlap or intersect, andas a result, a single object defined by its contiguous regionmay have both floc and filament parts. Figure 2 shows twotypical activated sludge flocs. Additional processing stepsare required to separate the intersecting flocs and filamentsbecause the classification of contiguous objects into the twoclasses is not sufficient.

The separation of flocs and filaments was achieved witha sequence consisting of filtering, morphological opera-tions and subtractions. This image processing sequence isillustrated in Figure 2 for typical activated sludge flocs

with intersecting filaments. After the background has beenremoved in step 2, the image intensity values are adjustedin step 3 to increase the contrast. The negative of the imageis created in step 4 and median filtering is then applied toremove filaments and small debris from the image in step 5.A threshold operation with a predefined threshold value isperformed in step 6 to create a binary image. A filling oper-ation is performed in step 7 to fill inner holes in flocs in thebinary image. Note that, in the example shown in Figure 2,there are no holes to fill, and therefore, subfigures 6 and7 are identical. In step 8, the image is dilated, resulting inan increased floc size, which is necessary for the successof step 9, where the dilated image of step 8 is subtractedfrom the threshold binary image of step 3 to create the fila-ment image. When the subtraction is performed, the dilatedflocs remove debris that would otherwise be left around thefloc boundaries. The floc image in step 10 is obtained byremoving the filament image of step 9 from the thresholdbinary image of step 3. The final floc and filament imagesare obtained after further processing, which is explainedfollowing two sections. The processed floc and filamentimages are shown in subfigures 11 and 12, respectively.

Processing of floc objectsFurther processing is performed after the separation of flocsand filaments into two binary images to remove objectsthat cannot be considered proper flocs or filaments, and tocalculate the final results. The processing of floc objects isdescribed first.

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 0

8:49

14

Nov

embe

r 20

14

682 E. Koivuranta et al.

Figure 2. Illustration of the steps for separating of flocs and filaments: (1) the original image; (2) background removal; (3) adjustmentof image intensity values to increase contrast; (4) creating negative image; (5) median filtering; (6) thresholds to create binary images; (7)filling holes; (8) dilation; (9) filament image; (10) floc image; (11) processed floc image; and (12) processed filament image.

Objects touching the image border are very unlikely tobe completely within the image. Including these objects inthe analysis would give unrealistic morphological parame-ters. Therefore, the first step in processing floc objects is toremove those that touch the image border.

Air leaks in the measuring equipment cause occasionalair bubbles to appear in the images. Air bubbles can be sep-arated from flocs because micro scale bubbles are ellipticaland almost completely round while flocs have a highly irreg-ular shape. The next step in the processing of floc objects isto remove air bubbles based on their roundness value (seeEquation (2)). Objects with roundness above a predefinedvalue are removed as air bubbles. However, this procedurefor air bubble removal is not perfect because bubbles thatoverlap with other objects will not be detected.

Some of the flocs in each image appear to be out-of-focus while others are in focus with well-defined bound-aries, which is a feature of the measurement system. Someof the morphological parameters are calculated from themeasurement of the floc perimeter. Including out-of-focusflocs in further analysis would skew the results for mor-phological parameters because activated sludge flocs havea highly irregular perimeter instead of the smooth perimeterthat occurs when the image is not focused properly. Out-of-focus flocs are detected based on their mean colour intensity.Out-of-focus flocs have a lighter colour than well-focusedflocs on a white background. Flocs with grey scale colourvalues above a predefined threshold are categorized as beingout of focus and are subsequently removed from furtherprocessing.

The final step before calculating the morphologicalparameters is to remove the small debris that is presentin both flocs and filaments, and it consists of dispersed

bacteria, inorganic particles and fragments of flocs and fil-aments. Objects with less than a predefined value of pixelsare removed as small debris. The number of small objectsis counted because it may be indicative of sludge quality.

Although a number of studies [1,2,7,10,19] have eval-uated the use of information obtained with image analysisfor estimating sludge quality, the choice of appropriate mor-phological parameters is not obvious. In practice, a largenumber of morphological parameters are calculated. Themost significant parameters for predicting sludge qualityare chosen from the complete set of calculated parameters.However, the appropriate set of parameters depends on theimaging system and properties of the WWTP, such as thetype of the plant, wastewater and operation of the plant.In this study, the following shape parameters were calcu-lated for activated sludge flocs: form factor (FF), roundness(RO), aspect ratio (AR), fractal dimension (FD) and con-vexity (CO). The shape parameters were calculated as anaverage of the values for individual flocs over a singleimage. Additionally, the following size parameters werecalculated: mean, median and standard deviation of the flocarea in a single image, equivalent diameter (Deq) and flocsize distribution.

The form factor (FF) is affected by the irregularity orroughness of the object’s boundary. It is 1.0 for a perfectcircle and below 1.0 for any other shape. Objects with moreirregular boundaries have a longer perimeter per surfacearea and therefore have smaller form factors. The formfactor is calculated from Equation (1) as [24]:

FF = 4π × areaperimeter2 (1)

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 0

8:49

14

Nov

embe

r 20

14

Environmental Technology 683

Roundness (RO) is also 1.0 for a perfect circle. However,its definition differs from that of the form factor. Roundness(Equation (2)) is defined as the ratio between the area of anobject and the area of a circle with a diameter equal to theobject’s length [24]:

RO = 4 × areaπ × length2 (2)

The aspect ratio (AR) describes how elongated an objectis [24]. It is calculated from Equation (3) by:

AR = lengthwidth

(3)

Convexity (CO) is defined as the ratio between theobject’s convex perimeter and its all over perimeter. Itis 1.0 for objects with no concavities and decreases forobjects that that are concave. The equivalent diameter (Deq)

(Equation (4)) is the diameter of a circle with an area equalto the object’s area:

Deq =√

4π

× area (4)

The floc size distribution was calculated by assigningeach floc to a size category based on its equivalent diam-eter. Eleven size categories were used, starting from zeroin increments of 25 μm. The floc size distribution was cal-culated for each sample as the sum of the distributions ofindividual images.

The fractal dimension can also be used to describe thefloc structure. The box counting concept can be used tocalculate the fractal dimensions from the floc images. In thebox counting concept, the image is covered with squaresof a certain size using the minimum number of squaresto completely cover the object in the image. The squaresize is then reduced and the process is repeated. The fractaldimension is obtained from the slope of the curve from thelogarithmic plot of the size of squares against the logarithmof number of the squares [25].

Processing of filament objectsThe processing of filament objects also begins by remov-ing the small debris and objects touching the image border.Some non-filament debris is the size of smaller fila-ments and therefore remains in the filament image afterthe removal of the small debris. This larger debris can-not be removed based on size, because doing so wouldalso remove smaller filament objects. A reduced radius ofgyration (RRG) has been found to be the most suitableshape parameter in distinguishing between filamentous andnon-filamentous objects [3,9]. Therefore, an RRG-basedcriterion was applied to objects in the filament image.Objects with smaller-than-predefined values of RRG wereremoved as non-filamentous objects.

Morphological parameters were not calculated for fila-ment objects because clusters of intersecting filaments areusually detected instead of individual filaments. A globalmeasure of filament abundance, the total filament length inthe image, was calculated instead. The total filament lengthin the image was calculated by summing the skeletonizedfilament image. A measure of the filament abundance com-pared to the number of floc-forming bacteria, the totalfilament length per total floc area ratio, was also calculated.

Determining the settling velocity of sludge by analyticalcentrifugationThe multi-sample analytical centrifuge (LUMiFuge) wasused to analyse the settling velocity of samples. Thecentrifuge allows the settling process of samples to be deter-mined under the influence of various centrifugal forces. Theintensity of the transmitted near-infrared light as a functionof time and position over the entire sample length is mea-sured, which makes it possible to determine the settlingvelocity of the samples. The samples were centrifuged at79 g (800 rpm) for 8 min and the settling speed after the first10 s was calculated. At least four parallels were measuredand the standard deviation of the measurement was 5%.

Results and discussionRepeatability of image analysisThe repeatability of the floc imaging was tested. One of thesamples was filmed 30 times with MOFI so that one sessionincluded at least 20,000 individual flocs. Images were anal-ysed with an automated image analysis program. Thus, over600,000 individual flocs were characterized and the stan-dard deviations of the most critical shape factors, filaments,particle areas and number of particles were determined. Theresults are presented in Table 2.

The standard deviations of the floc morphology areminor. Thus, the measurements can be regarded as sta-tistically reliable for measuring the floc morphologicalcharacteristics. Based on these results, 28,000 individualflocs should give statistically significant results comparedto others methods where 200 objects per sample [19], 2000particles per sample [9], 50 images per sample [4] and150−200 images per sample [2,8,14] were analysed. Thus,the amount of analysed flocs per sample in this study canprovide substantially higher statistical relevance than inprevious investigations.

Comparing bulking sludge to non-bulking sludgeThe morphology of the activated sludge floc was studiedwith the floc measurement environment in both bulkingand non-bulking situations to look for any differences inthe floc structure between the situations. Real process sam-ples were taken on four different days in both situations.At least 28,000 individual flocs were analysed from each

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 0

8:49

14

Nov

embe

r 20

14

684 E. Koivuranta et al.

Table 2. Mean values and standard deviations of the image analysis.

Filament length/floc Form Aspect Mean area of all Number of all particlesarea [1/m] factor Roundness ratio particles [μ m2] per image

Mean value 0.0073 0.508 0.499 1.879 3027 312Standard deviation [%] 2.74 2.17 1.40 1.06 5.19 6.41

Figure 3. Examples of images from the bulking situation (left) and the non-bulking situation (right).

sample with the automated image analysis. Figure 3 showsexamples of images in both bulking and non-bulking sludgesamples.

The amount of filaments was clearly higher in the bulk-ing sludge situations compared to the non-bulking situationsby visual inspection of the images. Additionally, the flocswere more fragile and smaller in the bulking sludge sit-uations. Based on the images obtained from MOFI, theexcessive growth of filamentous bacteria caused the bulkingevents and poor purification results in the WWTP.

The mean values of the flocs morphological characteris-tics were calculated and compared to the settling velocity ofsamples measured by LUMiFuge. The critical results of theimage analysis and sludge settling velocity are presented inFigures 4 and 5.

Figures 4 and 5 show that the settling velocity was lowerin the bulking sludge situations than in the non-bulking sit-uations, which correlates well with the poor purificationresults of the ASP (Table 1). The shape of the flocs wasclearly different in the bulking sludge situations comparedto the non-bulking sludge situations. The aspect ratio washigher, which means that the flocs were more elongated,and the form factor and roundness were lower meaning thatflocs had more irregular boundaries and were more fragilein the bulking situations. Additionally, the filament lengthwas higher in the bulking-sludge situations, most likely dueto the extensive growth of filamentous bacteria, which is themajor reason for the bulking sludge events.

Figure 5 shows that the mean area of the particles wassmaller and that the number of particles was higher in thebulking sludge situations, which may indicate the formationof PPs. However, PPs are formed when there is a lack of fila-mentous bacteria. In this case, based on the filaments results,there were excessive amounts of filamentous bacteria, so

the bulking events were not caused by formation of the PPseven though the mean area of particles was smaller.

Based on the results in Figures 4 and 5 the structure ofthe flocs in the bulking sludge situations is clearly differentcompared to those in the non-bulking situations. The dif-ference is also significant based on the repeatability testsand standard deviations (Table 2). Thus, the bulking eventswere caused by the growth of filamentous bacteria, whichholds flocs apart and impedes the settling speed, and can beobserved as the difference in the mean area of particles, fila-mentous properties and shape factors in the bulking sludgesituation.

Although the method was mainly tested on the sam-ples from one WWTP, a preliminary test with samples fromthree other WWTPs shows that flocs have structural differ-ences between plants treating different wastewaters. Thus,the results presented in this paper cannot be directly extrapo-lated to other activated sludge plants and more comparativedata are required for comprehensive analysis. However, thismethod is suitable for many ASP plants because filamen-tous bulking or poor flocculation mainly causes the bulkingevents.

Based on this study, the novel optical method to char-acterize flocs is capable of recognizing differences in thefloc structure, but not to analyse species of bacteria. Thismethod could detect changes in floc morphology beforebulking events occur because the method is designed tobe used in situ. In the upcoming bulking events, operatingconditions at WWTPs could be examined based on the theo-ries to explain bulking sludge [5] and to look for reasons forthe growth of filamentous bacteria. Thus, upcoming bulkingevents could be counteracted earlier by making adjustmentsto the operating conditions, or adding nutrients or chemicalsin the process. Additionally, the in situ measurements could

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 0

8:49

14

Nov

embe

r 20

14

Environmental Technology 685

Figure 4. Filaments and shape factors of flocs in bulking sludge and non-bulking sludge situation.

Figure 5. The mean area and number of particles per image in the bulking sludge and non-bulking sludge situation.

provide new information on the causes of bulking sludgebecause it can characterize floc faster. However, the methodmay not be suitable in situations where viscous or zooglealbulking causes problems.

ConclusionsThe novel optical method to characterize flocs was testedusing both bulking and non-bulking samples from an acti-vated sludge plant that treats wastewater from pulp mill.The large number of analysed particles allows for statisti-cally reliable information and the method can also be usedin situ. Based on these results, the structure of the flocs inthe bulking situations indicates that the bulking events in

the studied cases were caused by the growth of filamentousbacteria. Furthermore, the flocs in the bulking sludge situ-ations were more fragile and elongated. This new methodcreates a novel way to control and operate activated sludgetreatment plants and to obtain new information about thewhole ASP.

AcknowledgementsThis research has been carried out in the Measurement, Monitor-ing and Environmental Efficiency Assessment (MMEA) researchprogram of CLEEN Ltd. - Cluster for Energy and Environment.The authors would also like to acknowledge Tekes (the FinnishFunding Agency for Technology) for the financial support andMetso Inc. for providing the floc imaging setup.

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 0

8:49

14

Nov

embe

r 20

14

686 E. Koivuranta et al.

References[1] E.N. Banadda, I.Y. Smets, R. Jenné, and J.F. Van Impe,

Predicting the onset of filamentous bulking in biolog-ical wastewater treatment systems by exploiting imageanalysis information, Bioprocess Biosyst. Eng. 27 (2005),pp. 223–348.

[2] D.P. Mesquita, O. Dias, A.L. Amaral, and E.C. Ferreira,Monitoring of activated sludge settling ability through imageanalysis: Validation on full-scale wastewater treatmentplants, Bioprocess Biosyst. Eng. 32 (2009), pp. 361–367.

[3] R. Jenné, C. Cenens, A. Geeraerd, and J. Impe, Towards on-line quantification of flocs and filaments by image analysis,Biotechnol. Lett. 24 (2001), pp. 931–935.

[4] I.Y. Smets, E.N. Banadda, J. Deurinck, N. Renders, R. Jenné,and J.F. Van Impe, Dynamic modeling of filamentous bulkingin lab-scale activated sludge processes, J. Process Contr. 16(2006), pp. 313–319.

[5] A.M.P. Martins, K. Pagilla, J.J. Heijnen, and M.C.M. vanLoosdreht, Filamentous bulking sludge – a critical review,Water Res. 38 (2004), pp. 793–817.

[6] A.E. Steiner, D.A. McLaren, and C.F. Forster, The nature ofactivated sludge flocs, Water Res. 10 (1976), pp. 25–30.

[7] D.-H. Li and J.J. Ganczarczyk, Structure of activated sludgeflocs, Biotechnol. Bioeng. 35 (1990), pp. 57–66.

[8] D. Mesquita, O. Dias, A. Dias, A. Amaral, and E. Ferreira,Correlation between sludge settling ability and image analy-sis information using partial least squares, Anal. Chm. Acta642 (2009), pp. 94–101.

[9] E.M. Contreras, L. Giannuzzi, and N.E. Zaritzky, Use ofimage analysis in the study of competition between filamen-tous and non-filamentous bacteria, Water Res. 38 (2004),pp. 2621–2630.

[10] J. Jenné, E. Banadda, I. Smets, J. Deurinck, and J. Impe,Detection of filamentous bulking problems: Developing animage analysis, Microsc. Microanal. 13 (2007), pp. 36–41.

[11] G. Thompson, J. Swain, M. Kay, and C.F. Forster, 2001.The treatment of pulp and paper mill effluent: A review,Bioresource Technol. 77 (2001), pp. 275–286.

[12] G. Thompson, and C. Forster, Bulking in activated sludgeplants treating paper mill wastes, Water Res. 37 (2003),pp. 2363–2644.

[13] J.-M. Brault, P. Whalen, and P. Stuart, Early warning signs ofbulking in an activated sludge system through interpretationof ATP data in a systems analysis context, Environ. Technol.32 (2011), pp. 1649–1660.

[14] D.P. Mesquita, A.L. Amaral, and E.C. Ferreira, Identify-ing different types of bulking in an activated sludge sys-tem through quantitative image analysis, Chemosphere 85(2011), pp. 643–652.

[15] B. Jin, B.-M. Wilén, and P. Lant, A comprehensive insightinto floc characteristics and their impact on compressibilityand settleability of activated sludge, Chem. Eng. J. 95 (2003),pp. 221–234.

[16] M. Schmid, A. Thill, U. Purkhold, M. Walcher, J.Y. Bot-tero, P. Ginest, P.H. Nielsen, S. Wuertz, and M. Wagner,Characterisation of activated sludge flocs by confocal laserscanning microscopy and image analysis, Water Res. 37(2003), pp. 2043–2052.

[17] G. Peng, F. Ye, and Y. Li, Investigation of extracellularpolymer substances (EPS) and physicochemical propertiesof activated sludge from different municipal and industrialwastewater treatment plants, Environ. Technol. 33 (2012),pp. 857–863.

[18] C.P. Chu and D.J. Lee, Comparison of dewaterability and flocstructure of synthetic sludge with activated sludge, Environ.Technol. 26 (2005), pp. 1–10.

[19] K. Grijspeerdt and W. Verstraete, Image analysis to estimatethe settleability and concentration of activated sludge, WaterRes. 31 (1997), pp. 1126–1134.

[20] A.L. Amaral and E.C. Ferreira, Activated sludge monitor-ing of a wastewater treatment plant using image analysisand partial least squares regression, Anal. Chim. Acta 544(2005), pp. 246–253.

[21] B.-M. Wilén, D. Lumley, A. Mattsson, and T. Mino,Relationship between floc composition and flocculationand settling properties studied at a full-scale acti-vated sludge plant, Water Res. 42 (2008), pp. 4404–4418.

[22] P. Jarvis, B. Jefferson, and S.A. Parsons, Measuring flocstructural characteristics, Rev. Environ. Sci. Biotechnol. 4(2005), pp. 1–18.

[23] M. Motta, M.-N. Pons, N. Roche, and H. Vivier, Charac-terisation of activated sludge by automated image analysis,Biochem. Eng. J. 9 (2001), pp. 165–173.

[24] J. Russ, Computer-assisted Microscopy, Plenum Press, NewYork, 1990.

[25] G. Bushell, Y. Yan, D. Woodfield, J. Raper, and R. Amal, Ontechniques for the measurement of the mass fractal dimen-sion of aggregates, Adv. Colloid Interface Sci. 95 (2002),pp. 1–50.

Dow

nloa

ded

by [

Yor

k U

nive

rsity

Lib

rari

es]

at 0

8:49

14

Nov

embe

r 20

14