Mustafa Evren Ersahin1

Guclu Insel1

Recep Kaan Dereli1

Izzet Ozturk1

Cumali Kinaci1

1Istanbul Technical University, CivilEngineering Faculty, EnvironmentalEngineering Department, AyazagaCampus, Maslak, Istanbul, Turkey.

Research Article

Model Based Evaluation for the Anaerobic Treatmentof Corn Processing Wastewaters

The objective of this study was to implement a process model to simulate thedynamic behavior of a full-scale anaerobic reactor treating corn processing waste-waters under mesophilic conditions. For this purpose, the IWA Anaerobic DigestionModel No.1 (ADM1) was applied to an expanded granular sludge bed reactor of anindustrial wastewater treatment plant. The plant has a wastewater treatment capacityof 1594 m3/day. Sensitivity analysis showed that the influent characterization, basedon disintegration into carbohydrates, proteins, lipids and inert components, had alarge impact on the model outputs. An efficient correlation obtained between themeasured and the simulated data, including effluent COD, pH and methane produc-tion, highlights good parameter optimization and wastewater characterization. Thispaper demonstrates a practical example for the application of the ADM1 to an agro-industrial full-scale anaerobic reactor.

Keywords: Anaerobic Digestion Model No.1 (ADM1); Anaerobic treatment; Corn processing waste-waters; Expanded granular sludge bed reactor (EGSBR); Modeling;

Received: August 1, 2007; revised: October 19, 2007; accepted: October 24, 2007

DOI: 10.1002/clen.200700105

1 Introduction

Effluent from the corn milling industry is known as a high strengthwastewater due to its high protein and starch content. Wastewaterfrom corn wet mill industries has a high COD (Chemical OxygenDemand), mainly of soluble and biodegradable character, with aninitial inert COD content of less than 15% [1]. This character has pro-moted the application of biological processes as appropriate treat-ment technology. Anaerobic and/or aerobic biological treatmentsystems have been used to treat these types of effluents [2, 3].

Corn based glucose products are key ingredients in the growinginternational markets (food, biochemical, pharmaceutical, etc.).Intermediate products, such as vegetable oil, protein or/and whole-wheat and fructose obtained from starch have been utilized as rawmaterial by catering factories, breeding of animals, and processingindustries for sweeteners and beverages, respectively. The investi-gated industry is located in Bursa province, Turkey. Around 264 000tons of corn is processed annually by the industry. The mean spe-cific wastewater production was calculated as about 2 m3/t-corn.The industry has a three-stage advanced wastewater treatmentplant (WWTP) including an anaerobic expanded granular sludgebed reactor (EGSBR), intermittently aerated single sludge activatedsludge system for biological nitrogen (N) removal and chemical posttreatment unit for phosphorus (P) removal (see Fig. 1). Simultaneous

C and N removal can be achieved by the aerobic treatment stage.Chemical P removal has been performed by using FeCl3 as the coagu-lant in the third stage. The quality of the effluent meets the dis-charge limits of the European Union (EU) Urban Wastewater Direc-tive for Sensitive Regions. More information about the investigatedindustrial WWTP can be found elsewhere [3].

Considering the high costs and time consumed in experimentalstudies, mathematical models have been developed for the betterunderstanding of anaerobic treatment processes to overcome theobstacles, such as operational problems for the investigation of thedynamic behavior and more efficient design of anaerobic digesters.Modeling of anaerobic digestion processes have been investigatedand developed during the last 40 years [4]. One of the most sophisti-cated and complex mathematical model, the IWA Anaerobic Diges-tion Model No.1 (ADM1) was developed for the simulation and anal-ysis of anaerobic processes by the IWA Task Group [5]. ADM1 is ageneric model with consistent nomenclature, units and equationstructure. However, practical industrial use of the model has beenrelatively low. Some of the reasons for this may be the lack of imple-mentation in widely used software packages, lack of clearly detailedmethods for implementation and application of the models and rel-atively low numbers of case studies, and lack of clearly known bene-fits available from modeling [6].

Only a few studies have been reported for the application ofADM1 on industrial wastewater treatment. Batstone and Keller(2003) applied the ADM1 to an upflow anaerobic sludge bed reactortreating recycling paper mill wastewater [6]. They investigatedwhether acid (HCl) dosing was economically viable for the preven-tion of excessive precipitation of calcium carbonate (CaCO3) in thereactor. They concluded that pH reduction by HCl addition wasprobably not economical for the industry. Furthermore, in this

Correspondence: Mustafa Evren Ersahin, Istanbul Technical University,Civil Engineering Faculty, Environmental Engineering Department, Aya-zaga Campus, 34469, Maslak, Istanbul, Turkey.E-mail: ersahin@itu.edu.tr

Abbreviations: ADM1, Anaerobic Digestion Model No.1; COD, ChemicalOxygen Demand; EGSBR, Expanded granular sludge bed reactor; WWTP,Wastewater treatment plant

i 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.clean-journal.com

576 Clean 2007, 35 (6), 576 –581

Clean 2007, 35 (6), 576 – 581 Anaerobic Treatment of Wastewater 577

study they also investigated whether free ammonia inhibition wasalleviated by changing the operation conditions from mesophilic tothermophilic in an anaerobic solid digester fed with solids and con-centrated liquid streams from gelatin processing. According to sim-ulation results, they predicted that changing to thermophilic condi-tions would have no significant impact on ammonia inhibition forthe reactor. Kalfas et al. (2006) implemented the ADM1 for the simu-lation of the anaerobic digestion of olive pulp in continuous stirredtank reactors (CSTR) operated at mesophilic and thermophilic con-ditions [7]. They reported that ADM1 was able to simulate the anae-robic digestion of olive pulp in CSTR type digesters and also theyestimated key parameters such as specific maximum uptake rateand saturation constants for the volatile fatty acids degradation.

Treatment of industrial wastewaters such as corn processingeffluents will benefit from direct implementation of models for con-trol, operation and optimization of full-scale treatment plants andassisting modelling studies transfer to industry. In the literature,there is no available study for the application of a general anaerobicmodel, such as ADM1, for corn processing wastewaters as an indus-trial application [8]. Thus, the results obtained from this studypresent an example for application of a structured model for indus-trial wastewater treatment.

The main purpose of this study is to compare the ADM1 resultswith long-term data from a full-scale anaerobic reactor treatingwastewater originated from a corn processing plant. For this pur-pose, model based wastewater characterization was performed,then calibration and verification studies were carried out using aone year continuous set of data of the anaerobic treatment system.The main operating parameters, including effluent COD, pH andmethane production rate, were predicted by the ADM1 consideringdaily data from the wastewater plant.

2 Materials and Methods

2.1 Corn Wet Milling Industry and Anaerobic Reactor

The investigated corn processing plant has two main productionsteps including wet mill and starch slurry derivatives production.The starch slurry is further processed to produce glucose, fructoseand dextrin in the starch slurry derivatives units. The simplified

product flow diagram of the plant is shown in Fig. 2. Average pollu-tion profile data is presented in Tab. 1.

The strong effluents from the wet mill processing have beentreated by the investigated anaerobic reactor. The operating dataobtained from the EGSBR, with an effective volume of 1226 m3,served as the main material for the modelling study (see Tab. 2). Theorganic loading rate (OLR) and hydraulic retention time (HRT) val-ues for the EGSBR were 3.57 kg COD/m3 N day and 0.77 day, respec-tively. The solids retention times in the anaerobic reactor are veryhigh (A100 days, in general) and the VSS/TSS ratios of the granularbiomass averaged 80%. The subsequent aerobic unit has an effective

i 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.clean-journal.com

Figure 1. Process scheme of WWTP. 1 Equalization tanks; 2 Conditioning tank; 3 EGSB reactor; 4 Separator; 5 Selector; 6 Aeration tanks; Settlingtank; 8 Chemical treatment; 9 Settling tank; 10 Belt filter; 11 Flare.

Table 1. Average pollution profile of the industry*.

Source Effluent formation COD

m3/day m3/t-corn mg/L kg/day kg/t-corn

Wet mill 640 0.64 4140 2650 2.65Refenery 804 0.80 1755 1411 1.41Total 1444 1.44 2812 4061 4.06

* The wastewater coming from auxiliary units were not included inthe pollution profile.

Table 2. Monthly average operating data of the EGSBR in year 2004.

Month Flow rate(m3/day)

Influent Effluent Methaneproduc-

COD(kg/day)

TKN(kg/day)

COD(mg/L)

pH tion*(m3/day)

1 1313 2.76 0.063 0.43 6.97 10462 1187 2.64 0.040 0.41 6.95 9913 1719 2.63 0.037 0.41 6.97 13204 1576 2.84 0.042 0.41 6.97 13395 1573 3.31 0.065 0.47 6.96 15436 1649 3.13 0.046 0.48 6.95 15197 1354 2.05 0.025 0.37 6.92 8508 1570 3.08 0.039 0.43 6.86 14239 1814 2.59 0.024 0.38 6.85 1384

10 1864 2.74 0.040 0.44 6.90 147811 1936 2.61 0.040 0.44 6.88 145312 1569 2.59 0.035 0.46 6.93 1168Average 1594 2.75 0.041 0.43 6.93 1293

* Estimated values considering COD removals in the reactor.

578 M. E. Ersahin et al. Clean 2007, 35 (6), 576 –581

volume of 4500 m3. COD removal rates of the anaerobic and aerobicunits were same at 85%.

The operational parameters monitored during the period of thestudy were carried out according to AWWA standard methods [9] atthe wastewater laboratory of the company. Control analyses werealso performed at the Environmental Engineering Department ofIstanbul Technical University on monthly basis.

2.2 Model Implementation

The COD, pH and methane production in the reactor were used todetermine the consistency of ADM1 simulation results. Initialparameter values were derived from the literature and the ADM1Technical Scientific Report [5] in order to initiate model simula-tions. The Aquasim 2.1 [10] simulation program was used to per-form the simulations.

COD fractionation of the raw wastewater fed to the anaerobicreactor into protein, starch, lipids and inerts was the most criticalstep for the implementation of the ADM1 to any type of organicwaste. The composition of the corn processed in the plant (see Tab.

3) was used for the determination of COD components of the waste-water. The main characteristics of the raw wastewater wereassumed with the fractions given in Tab. 4, considering the valuesin earlier studies on composition of the corn and process informa-tion of the industry [11, 1].

Due to the lack of reliable biogas measurements in the plant,methane production data was calculated considering expectedmethane yields at 358C and about 95% conversion of removed CODto CH4.

2.3 Model Matrix Modification

The ADM1 has state variables including inorganic carbon (SIC) andinorganic nitrogen (SIN). These may act as source or sink terms toclose mass balances. An addition was made to the ADM1 in order toincorporate stoichiometric coefficients for the biomass decay proc-ess expressing N and C release due to biomass decay. This was causedby a differentiation between the carbon and nitrogen content in thebiomass and the composite material. To resolve the carbon andnitrogen discrepancy, the following terms were introduced to themodel matrix [12].

Cbac – Cxc (i = 10, j = 12–19 in the model matrix) (1)

Nbac – Nxc (i = 11, j = 12 –19 in the model matrix) (2)

Another addition was the inclusion of the stoichiometric coeffi-cient for the disintegration process expressing N and C release [7].During simulation studies, it was shown that, without this addi-tion, the carbon and nitrogen balances calculated using the originalparameters given in the ADM1 report can not be closed (m L –0.0029 kmole C/kg COD, n L 0.00128 kmole N/kg COD). The addi-tions into the original matrix [5] are summarized in Tab. 5. In thistable:

m = Cxc – fxI_xcCinert – fSI_xc Cinert – fpr_xc Caa – fch_xc Csu – fli_xc Cli (3)

n = Nxc – fxI_xc Ninert – fSI_xc Ninert – fpr_xc Naa (4)

In order to guarantee the carbon and nitrogen balances in case ofapplication of the model for anaerobic treatment of different wastetypes, balance terms should be taken into account.

i 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.clean-journal.com

Figure 2. Product flow diagram of the corn processing plant.

Table 3. Composition of corn processed in the plant.

Component Value (%)

MoistureProteinStarchOil

15.48.6

70.94.9

Foreign material 0.2

Table 4. Accepted wastewater characterization fractions for corn pro-cessing wastewater.

Parameter Name UsedValue

Range

fSI_xc

fXI_xc

Soluble inerts from compositesParticulate inerts from compo-sites

0.10.05

f 0.1f 0.1

fch_xc

fpr_xc

fli_xc

Carbohydrates from compositesProteins from compositesLipids from composites

0.770.060.02

0.7 – 0.80.05 – 0.1f 0.02

Clean 2007, 35 (6), 576 – 581 Anaerobic Treatment of Wastewater 579

3 Results and Discussion

3.1 Model Calibration and Sensitivity Analysis

The model calibration was performed by the average effluent COD,pH and methane gas flow values for a six month period of the inves-tigated anaerobic reactor. Some of the model parameters were cali-brated in order to improve the consistency of simulation resultswith the real plant data (see Tab. 6). In this study, the recursivemodel calibration method was applied, considering the most sensi-tive parameters to all model simulation outputs (COD, pH, etc.) [13,14].

COD fractions that have major impacts on the model outputswere identified based on the sensitivity analysis. Effects of changesin these parameters on the COD, pH and CH4 gas flow outputs areshown in Tab. 7.

The values in Tab. 7 indicate the relative changes pertaining tothe model parameters. For example, a 20% change in the value offSI_xc results in an 11.8% increase in the COD value. Effluent COD ismost sensitive to fSI_xc among these three parameters. fch_xc alsoincreases the pH value and biogas flow significantly. Biogas flow isdecreased by the increases in the other two parameters.

3.2 Model Validation

A validation study was performed to assess the quality and applic-ability of the optimized parameters and influent characterizationvalues. The simulation outputs were compared with a completelydifferent data obtained for another six months period from theinvestigated full-scale anaerobic reactor.

Steady state conditions were reached after a start-up phase (A200days). After consistency had been obtained between the simulationresults and real data, varying input was applied as a feed regime toobserve the dynamic process behavior.

The comparison of model outputs and reactor data for the COD,methane production and pH is shown in Fig. 3 (a, b, c). The verifica-tion results performed for dynamic loadings have shown that cali-brated kinetic parameters are consistent and the model predicts thebehavior of the anaerobic process with high accuracy, except formethane production.

i 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.clean-journal.com

Table 5. Modified matrix format.

Component fi i 10 11 Ratej Process z SIC SIN

1 Disintegration. ………………

m n kdis N Xc

13 Decay of Xsu

14 Decay of Xaa

15 Decay of Xfa

16 Decay of Xc4

17 Decay of Xpro

18 Decay of Xac

19 Decay of Xh2

Cbac – Cxc

Cbac – Cxc

Cbac – Cxc

Cbac – Cxc

Cbac – Cxc

Cbac – Cxc

Cbac – Cxc

Nbac – Nxc

Nbac – Nxc

Nbac – Nxc

Nbac – Nxc

Nbac – Nxc

Nbac – Nxc

Nbac – Nxc

kdec,xsu N Xsu

kdec,xaa N Xaa

kdec,xfa N Xfa

kdec,xc4 N Xc4

kdec,xpro N Xpro

kdec,xac N Xac

kdec,xh2 N Xh2

Table 6. Calibrated parameter values.

Parameter Suggestedvalue (*)

Used value

kdis (day – 1)khyd,CH (day – 1)

0.40.25

10.45

KS-fa (kg COD/m3)KS-ac (kg COD/m3)KS-pro (kg COD/m3)km-pro (COD/COD N day)km-fa (COD/COD N day)km-ac (COD/COD N day)km-c4 (COD/COD N day)km-su (COD/COD N day)

0.40.150.3

1368

2030

0.30.10.2

169

121535

* The values were taken from Tab. 6.2 in ref. [5].

Figure 3. Measurement and simulation results for (a) effluent COD, (b)predicted methane gas production, and (c) pH.

580 M. E. Ersahin et al. Clean 2007, 35 (6), 576 –581

The response of the model to the dynamic variation of measuredvalues of the variables to be estimated were consistent with changesin increasing and decreasing trends seen in Fig. 3 (a, b, c). Accordingto the comparison of the model predictions with actual digesterparameter values, the mean absolute relative error values are 18,10, and 1%, respectively for methane production, COD and pH. Highrelative error for methane production may be explained by practi-cal difficulties in the measurement and/or estimation of methaneproduction from the full-scale anaerobic reactors. Another possiblereason for such deviations can also be attributed to non-optimiza-tion of several model parameters. Similar findings have been experi-enced by other studies related to ADM1 applications [12].

During the simulation studies, it was found that influent waste-water characterization parameters (fSI_xc, fXI_xc, fch_xc, fpr_xc, fli_xc) hadgreat impact on the model outputs. In order to achieve accuratemodel predictions it is important to define the properties of thewaste stream entering the anaerobic reactor appropriately.

4 Conclusions

In this study, a model based influent characterization was per-formed and ADM1 was implemented to the anaerobic treatment ofwastewater from a corn processing company. According to simula-tion results, the ADM1 was able to simulate the anaerobic treatmentperformance of a full-scale EGSBR under dynamic conditions. Theresults showed that the estimated wastewater characterization wasconsistent. One of the most common problems experienced inapplying ADM1 to full-scale anaerobic reactors is unreliable biogasproduction data due to the lack of gas meters or inaccurate gasmeasurement. More precise parameter optimization study isrequired to increase the consistency of model predictions for meth-ane production. The verified model effectively predicted pH andCOD removal efficiency and can be used for process control pur-poses. A further optimization study is a valuable tool to determinethe effects of different operation alternatives or maximization ofmethane gas produced in the system. The effect of different processalternatives and shock loadings on the system may be investigatedby using the verified model. The results of this study have clearlyshowed that ADM1 can be successfully applied to modeling of full-scale anaerobic reactors treating agro-industry effluents.

Acknowledgements

The authors are thankful to the Cargill Orhangazi Plant and CemBilge for providing operational data used in the study.

i 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.clean-journal.com

Table 7. Sensitivity of parameters to model outputs.

Relative change in parameterfi

Changes in the model outputs (relative changes %)

20% 50% 100% 20% 50% 100% 20% 50% 100%

Model parameters zfSI_xc CODfch_xc

fXI_xc

11.83.50.2

33.510.40.4

40.224.61.0

pH 0.42.90.2

1.016.70.6

2.341.51.3

CH4 flow – 2.64.1

– 1.6

– 6.211.1– 4.1

– 13.832.8– 8.6

Symbols used

Caa

Cbac

Cinert

Cli

Csu

Cxc

(kmole C/kg COD)(kmole C/kg COD)(kmole C/kg COD)(kmole C/kg COD)(kmole C/kg COD)(kmole C/kg COD)

Carbon content of amino acidsCarbon content of biomassCarbon content of inertsCarbon content of lipidsCarbon content of sugarsCarbon content of composite material

fch_xc

fli_xc

fpr_xc

fSI_xc

fXI_xc

HRTkdec,xaa

kdec,xac

kdec,xc4

kdec,xfa

kdec,xh2

kdec,xsu

kdec,xpro

kdis

km-c4

km-su

KS-fa

Naa

Nbac

Ninert

Nxc

OLRSIC

SIN

Xaa

Xac

Xc

Xc4

Xfa

Xh2

Xpro

Xsu

Yac

Ysu

(kg COD/kg COD)(kg COD/kg COD)(kg COD/kg COD)(kg COD/kg COD)(kg COD/kg COD)(day)(day – 1)

(day – 1)

(day – 1)

(day – 1)

(day – 1)

(day – 1)

(day – 1)

(day – 1)(COD/COD N day)

(COD/COD N day(kg COD/m3)(kmole N/kg COD)

(kmole N/kg COD)(kmole N/kg COD)(kmole N/kg COD)(kg COD/m3 N day)(kmole C/kg COD)(kmole N/kg COD)(kg COD/m3)(kg COD/m3)(kg COD/m3)(kg COD/m3)

(kg COD/m3)

(kg COD/m3)(kg COD/m3)(kg COD/m3)(COD/COD)(COD/COD)

Carbohydrates from compositesLipids from compositesProteins from compositesSoluble inerts from compositesParticulate inerts from compositesHydraulic retention timeFirst order decay rate for amino acid de-grading bacteriaFirst order decay rate for acetate degrad-ing bacteriaFirst order decay rate for valerate-buty-rate degrading bacteriaFirst order decay rate for fatty acid de-grading bacteriaFirst order decay rate for hydrogen de-grading bacteriaFirst order decay rate for sugars degrad-ing bacteriaFirst order decay rate for propionate de-grading bacteriaDisintegration rate coefficientMaximum specific uptake rate for valer-ate-butyrateMaximum specific uptake rate for sugarsHalf saturation value for fatty acidsNitrogen content of amino acids and pro-teinsNitrogen content of biomassNitrogen content of inertsNitrogen content of compositesOrganic loading rateInorganic carbonInorganic nitrogenConcentration of amino acid degradersConcentration of acetate degradersConcentration of composite materialsConcentration of valerate and butyratedegradersConcentration of long chain fatty acid de-gradersConcentration of hydrogen degradersConcentration of propionate degradersConcentration of sugar degradersYield of biomass on acetateYield of biomass on sugars

Clean 2007, 35 (6), 576 – 581 Anaerobic Treatment of Wastewater 581

References

[1] G. Eremektar, O. Karahan-Gul, F. Germirli-Babuna, S. Ovez, H. Uner,D. Orhon, Biological Treatability of a Corn Wet Mill, Wat. Sci. Tech.2002, 45 (12), 339 – 346.

[2] A. R. Howgrave-Graham, H. I. Isherwood, F. M. Wallis, Evaluation ofTwo Upflow Anaerobic Digesters Purifying Industrial WastewatersHigh in Organic Matter, Wat. Sci. Tech. 1994, 29 (9), 225 – 229.

[3] C. Y. Gomec, E. Erdim, I. Ozturk, C. Bilge, Advanced Full-Scale PostTreatment Experiences of Anaerobically Pretreated Effluent from aCorn Milling Industry, in Proc. of the 10th World Congress on AnaerobicDigestion (AD10), Montreal, Canada, 29 August – 3 September 2004, 4,2207 – 2210.

[4] H. N. Gavala, I. Angelidaki, B. K. Ahring, Kinetics and Modeling ofAnaerobic Digestion Process, Adv. Biochem. Eng. Biotechnol. 2003, 81,57 – 93.

[5] D. J. Batstone, J. Keller, I. Angelidaki, S. V. Kalyuzhnyi, S. G. Pavlosta-this, A. Rozzi, W. T. M. Sanders, H. Siegrist, V. A. Vavilin, AnaerobicDigestion Model No.1, Scientific and Technical Report No.13, IWA Publish-ing, London 2002.

[6] D. J. Batstone, J. Keller, Industrial Application of the IWA AnaerobicDigestion Model No.1 (ADM1), Wat. Sci. Tech. 2003, 47 (12), 199 – 206.

[7] H. Kalfas, I. V. Skiadas, H. N. Gavala, K. Stamatelatou, G. Lyberatos,Application of ADM1 for the Simulation of Anaerobic Digestion ofOlive Pulp Under Mesophilic and Thermophilic Conditions, Wat. Sci.Tech. 2006, 54 (4), 149 – 156.

[8] M. E. Ersahin, Implementation of ADM1 to Anaerobic Treatment ofCorn Processing Wastewaters, M.Sc. Thesis, Institute of Science andTechnology, Istanbul Technical University 2005.

[9] American Public Health Association, Standard Methods for the Examina-tion of Water and Wastewater, American Water Work Association,Washington 2000.

[10] P. Reichert, J. Ruchti, W. Simon, Aquasim 2.0, Swiss Federal Institutefor Environmental Science and Technology (EAWAG), CH-8600, Due-bendorf, Switzerland 1998.

[11] P. H. Blanchard, Technology of Corn Wet Milling, Elsevier, Amsterdam1992.

[12] F. Blumensaat, J. Keller, Modeling of a Two-Stage Anaerobic Diges-tion Using the IWA Anaerobic Digestion Model No.1 (ADM1), Wat.Res. 2005, 39 (1), 171 – 183.

[13] G. Insel, G. Sin, D. S. Lee, I. Nopens, P. A. Vanrolleghem, A CalibrationMethodology and Model-based System Analysis for Nutrient Remov-ing SBRs Under Reduced Aeration, J. Chem. Tech. Biotechnol. 2006, 81 (4),679 – 687.

[14] P. A. Vanrolleghem, G. Insel, B. Petersen, G. Sin, D. De Pauw, I.Nopens, H. Dovermann, S. Weijers, K. Gernaey, A comprehensivemodel calibration procedure for activated sludge models, in Proc. ofWEFTEC 2003, 76th Annual Technical Exhibition and Conference, October11 – 15, Los Angeles, CA 2003.

i 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.clean-journal.com