Combining Group-Contribution Concept and Graph Neural Networks Toward Interpretable Molecular Property Models.

Quantitative structure-property relationships (QSPRs) are important tools to facilitate and accelerate the discovery of compounds with desired properties. While many QSPRs have been developed, they are associated with various shortcomings such as a lack of generalizability and modest accuracy. Albeit various machine-learning and deep-learning techniques have been integrated into such models, another shortcoming has emerged in the form of a lack of transparency and interpretability of such models. In this work, two interpretable graph neural network (GNN) models (attentive group-contribution (AGC) and group-contribution-based graph attention (GroupGAT)) are developed by integrating fundamentals using the concept of group contributions (GC). The interpretability consists of highlighting the substructure with the highest attention weights in the latent representation of the molecules using the attention mechanism. The proposed models showcased better performance compared to classical group-contribution models, as well as against various other GNN models describing the aqueous solubility, melting point, and enthalpies of formation, combustion, and fusion of organic compounds. The insights provided are consistent with insights obtained from the semiempirical GC models confirming that the proposed framework allows highlighting the important substructures of the molecules for a specific property.

Determining the linear correlation between dielectric spectroscopy and viable biomass concentration in filamentous fungal fermentations.

OBJECTIVES: Dielectric spectroscopy is commonly used for online monitoring of biomass growth. It is however not utilized for biomass concentration measurements due to poor correlation with Cell Dry Weight (CDW). A calibration methodology is developed that can directly measure viable biomass concentration in a commercial filamentous process using dielectric values, without recourse to independent and challenging viability determinations. RESULTS: The methodology is applied to samples from the industrial scale fermentation of a filamentous fungus, Acremonium fusidioides. By mixing fresh and heat-killed samples, linear responses were verified and sample viability could be fitted with the dielectric [Formula: see text] values and total solids concentration. The study included a total of 26 samples across 21 different cultivations, with a legacy at-line viable cell analyzer requiring 2 ml samples, and a modern on-line probe operated at-line with 2 different sample presentation volumes, one compatible with the legacy analyzer, a larger sample volume of 100 ml being compatible with calibration for on-line operation. The linear model provided an [Formula: see text] value of 0.99 between [Formula: see text] and viable biomass across the sample set using either instrument. The difference in ∆C when analyzing 100 mL and 2 mL samples with an in-line probe can be adjusted by a scalar factor of 1.33 within the microbial system used in this study, preserving the linear relation with [Formula: see text] of 0.97. CONCLUSIONS: It is possible to directly estimate viable biomass concentrations utilizing dielectric spectroscopy without recourse to extensive and difficult to execute independent viability studies. The same method can be applied to calibrate different instruments to measure viable biomass concentration. Small sample volumes are appropriate as long as the sample volumes are kept consistent.

Integrated Model for Understanding N2O Emissions from Wastewater Treatment Plants: A Deep Learning Approach.

This study aims to demonstrate the application of deep learning to quantitatively describe long-term full-scale data observed from wastewater treatment plants (WWTPs) from the perspectives of process modeling, process analysis, and forecasting modeling. Approximately, 750,000 measurements including the influent flow rate, air flow rate, temperature, ammonium, nitrate, dissolved oxygen, and nitrous oxide (N2O) collected for more than a year from the Avedøre WWTP located in Denmark are utilized to develop a deep neural network (DNN) through supervised learning for process modeling, and the optimal DNN (R2 > 0.90) is selected for further evaluation. For process analysis, global sensitivity analysis based on variance decomposition is considered to identify the key parameters contributing to high N2O emission characteristics. For N2O forecasting, the proposed DNN-based model is compared with long short-term memory (LSTM), showing that the LSTM-based forecasting model performs significantly better than the DNN-based model (R2 > 0.94 and the root-mean-squared error is reduced by 64%). The results account for the feasibility of data-driven methods based on deep learning for quantitatively describing and understanding the rather complex N2O dynamics in WWTPs. Research into hybrid modeling concepts integrating mechanistic models of WWTPs (e.g., ASMs) with deep learning would be suggested as a future direction for monitoring N2O emissions from WWTPs.

Unravelling the environmental and economic impacts of innovative technologies for the enhancement of biogas production and sludge management in wastewater systems.

The retrofitting of wastewater treatment plants (WWTPs) should be addressed under sustainability criteria. It is well known that there are two elements that most penalize wastewater treatment: (i) energy requirements and (ii) sludge management. New technologies should reduce both of these drawbacks to address technical efficiency, carbon neutrality and reduced economic costs. In this context, the main objective of this work was to evaluate two real plants of different size in which major modifications were considered: enhanced recovery of organic matter (OM) in the primary treatment and partial-anammox nitrification process in the secondary treatment. Plant-wide modelling provided an estimate of the input and output flows of each process unit as well as the diagnosis of the main performance indicators, which served as a basis for the calculation of environmental and economic indicators using the LCA methodology. The combination of high-rate activated sludge (HRAS) + partial nitrification Anammox can decrease the environmental impacts by about 70% in the climate change (CC) category and 50% in the eutrophication potential (EP) category. Moreover, costs can be reduced by 35-45% depending on the size of the plant. In addition, the enhanced rotating belt filter (ERBF) can also improve the environmental profile, but to a lesser extent than the previous scenario, only up to 10% for CC and 15% for EP. These positive results are only possible considering the production of energy through biogas valorization according to the waste-to-energy scheme.

Nitrous oxide production in autotrophic nitrogen removal granular sludge: A modeling study.

The sustainability of autotrophic granular system performing partial nitritation and anaerobic ammonium oxidation (anammox) for complete nitrogen removal is impaired by the production of nitrous oxide (N2 O). A systematic analysis of the pathways and affecting parameters is, therefore, required for developing N 2 O mitigation strategies. To this end, a mathematical model capable of describing different N 2 O production pathways was defined in this study by synthesizing relevant mechanisms of ammonium-oxidizing bacteria (AOB), nitrite-oxidizing bacteria, heterotrophic bacteria (HB), and anammox bacteria. With the model validity reliably tested and verified using two independent sets of experimental data from two different autotrophic nitrogen removal biofilm/granular systems, the defined model was applied to reveal the underlying mechanisms of N 2 O production in the granular structure as well as the impacts of operating conditions on N 2 O production. The results show that: (a) in the aerobic zone close to the granule surface where AOB contribute to N 2 O production through both the AOB denitrification pathway and the NH 2 OH pathway, the co-occurring HB consume N 2 O produced by AOB but indirectly enhance the N 2 O production by providing NO from NO 2- reduction for the NH 2 OH pathway, (b) the inner anoxic zone of granules with the dominance of anammox bacteria acts as a sink for NO 2- diffusing from the outer aerobic zone and, therefore, reduces N 2 O production from the AOB denitrification pathway, (c) operating parameters including bulk DO, influent NH 4+ , and granule size affect the N 2 O production in the granules mainly by regulating the NH 2 OH pathway of AOB, accounting for 34-58% of N 2 O turnover, and (d) the competition between the NH 2 OH pathway and heterotrophic denitrification for nitric oxide leads to the positive role of HB in reducing N 2 O production in the autotrophic nitrogen removal granules, which could be further enhanced in the presence of a proper level of influent organics.

CFD predicted pH gradients in lactic acid bacteria cultivations.

The formation of pH gradients in a 700 L batch fermentation of Streptococcus thermophilus was studied using multi-position pH measurements and computational fluid dynamics (CFD) modeling. To this end, a dynamic, kinetic model of S. thermophilus and a pH correlation were integrated into a validated one-phase CFD model, and a dynamic CFD simulation was performed. First, the fluid dynamics of the CFD model were validated with NaOH tracer pulse mixing experiments. Mixing experiments and simulations were performed whereas multiple pH sensors, which were placed vertically at different locations in the bioreactor, captured the response. A mixing time of about 46 s to reach 95% homogeneity was measured and predicted at an impeller speed of 242 rpm. The CFD simulation of the S. thermophilus fermentation captured the experimentally observed pH gradients between a pH of 5.9 and 6.3, which occurred during the exponential growth phase. A pH higher than 7 was predicted in the vicinity of the base solution inlet. Biomass growth, lactic acid production, and substrate consumption matched the experimental observations. Moreover, the biokinetic results obtained from the CFD simulation were similar to a single-compartment simulation, for which a homogeneous distribution of the pH was assumed. This indicates no influence of pH gradients on growth in the studied bioreactor. This study verified that the pH gradients during a fermentation in the pilot-scale bioreactor could be accurately predicted using a coupled simulation of a biokinetic and a CFD model. To support the understanding and optimization of industrial-scale processes, future biokinetic CFD studies need to assess multiple types of environmental gradients, like pH, substrate, and dissolved oxygen, especially at industrial scale.

Assessment of Full-Scale N2O Emission Characteristics and Testing of Control Concepts in an Activated Sludge Wastewater Treatment Plant with Alternating Aerobic and Anoxic Phases.

This work aims to obtain full-scale N2O emission characteristics translatable into viable N2O control strategies and conduct full-scale testing of the proposed N2O control concepts. Data of a long-term monitoring campaign was first used to quantify full-scale N2O emission and probe into the seasonal pattern. Then trends between N2O production/emission and process variables/conditions during typical operating cycles were revealed to explore the dynamic N2O emission behavior. A multivariate statistical analysis was performed to find the dependency of N2O emission on relevant process variables. The results show for the first time that relatively low/high N2O emission took place in seasons with a decreasing/increasing trend of water temperature, respectively. Aerobic phase contributed to N2O production/emission probably mainly through the hydroxylamine pathway. Comparatively, heterotrophic bacteria had a dual role in the anoxic phase and could be responsible for both net N2O production and consumption. Incomplete denitrification might contribute mainly to the N2O production/emission in the anoxic phase and the accumulation of N2O to be significantly emitted in the following cycle due to the competition between different denitrification steps for electron donors. Therefore, properly extending the length of anoxic phase could serve as a potential control means to regulate N2O accumulation in the anoxic phase. The full-scale testing not only verified the efficacy of reduced dissolved oxygen set-point in reducing N2O emission by 60%, but also confirmed the proposed concepts of control over the aerobic and anoxic phases collectively.

Application of a mechanistic model as a tool for on-line monitoring of pilot scale filamentous fungal fermentation processes-The importance of evaporation effects.

A mechanistic model-based soft sensor is developed and validated for 550L filamentous fungus fermentations operated at Novozymes A/S. The soft sensor is comprised of a parameter estimation block based on a stoichiometric balance, coupled to a dynamic process model. The on-line parameter estimation block models the changing rates of formation of product, biomass, and water, and the rate of consumption of feed using standard, available on-line measurements. This parameter estimation block, is coupled to a mechanistic process model, which solves the current states of biomass, product, substrate, dissolved oxygen and mass, as well as other process parameters including kL a, viscosity and partial pressure of CO2 . State estimation at this scale requires a robust mass model including evaporation, which is a factor not often considered at smaller scales of operation. The model is developed using a historical data set of 11 batches from the fermentation pilot plant (550L) at Novozymes A/S. The model is then implemented on-line in 550L fermentation processes operated at Novozymes A/S in order to validate the state estimator model on 14 new batches utilizing a new strain. The product concentration in the validation batches was predicted with an average root mean sum of squared error (RMSSE) of 16.6%. In addition, calculation of the Janus coefficient for the validation batches shows a suitably calibrated model. The robustness of the model prediction is assessed with respect to the accuracy of the input data. Parameter estimation uncertainty is also carried out. The application of this on-line state estimator allows for on-line monitoring of pilot scale batches, including real-time estimates of multiple parameters which are not able to be monitored on-line. With successful application of a soft sensor at this scale, this allows for improved process monitoring, as well as opening up further possibilities for on-line control algorithms, utilizing these on-line model outputs. Biotechnol. Bioeng. 2017;114: 589-599. © 2016 Wiley Periodicals, Inc.

A novel model-based control strategy for aerobic filamentous fungal fed-batch fermentation processes.

A novel model-based control strategy has been developed for filamentous fungal fed-batch fermentation processes. The system of interest is a pilot scale (550 L) filamentous fungus process operating at Novozymes A/S. In such processes, it is desirable to maximize the total product achieved in a batch in a defined process time. In order to achieve this goal, it is important to maximize both the product concentration, and also the total final mass in the fed-batch system. To this end, we describe the development of a control strategy which aims to achieve maximum tank fill, while avoiding oxygen limited conditions. This requires a two stage approach: (i) calculation of the tank start fill; and (ii) on-line control in order to maximize fill subject to oxygen transfer limitations. First, a mechanistic model was applied off-line in order to determine the appropriate start fill for processes with four different sets of process operating conditions for the stirrer speed, headspace pressure, and aeration rate. The start fills were tested with eight pilot scale experiments using a reference process operation. An on-line control strategy was then developed, utilizing the mechanistic model which is recursively updated using on-line measurements. The model was applied in order to predict the current system states, including the biomass concentration, and to simulate the expected future trajectory of the system until a specified end time. In this way, the desired feed rate is updated along the progress of the batch taking into account the oxygen mass transfer conditions and the expected future trajectory of the mass. The final results show that the target fill was achieved to within 5% under the maximum fill when tested using eight pilot scale batches, and over filling was avoided. The results were reproducible, unlike the reference experiments which show over 10% variation in the final tank fill, and this also includes over filling. The variance of the final tank fill is reduced by over 74%, meaning that it is possible to target the final maximum fill reproducibly. The product concentration achieved at a given set of process conditions was unaffected by the control strategy. Biotechnol. Bioeng. 2017;114: 1459-1468. © 2017 Wiley Periodicals, Inc.

A generic methodology for the optimisation of sewer systems using stochastic programming and self-optimizing control.

The design of sewer system control is a complex task given the large size of the sewer networks, the transient dynamics of the water flow and the stochastic nature of rainfall. This contribution presents a generic methodology for the design of a self-optimising controller in sewer systems. Such controller is aimed at keeping the system close to the optimal performance, thanks to an optimal selection of controlled variables. The definition of an optimal performance was carried out by a two-stage optimisation (stochastic and deterministic) to take into account both the overflow during the current rain event as well as the expected overflow given the probability of a future rain event. The methodology is successfully applied to design an optimising control strategy for a subcatchment area in Copenhagen. The results are promising and expected to contribute to the advance of the operation and control problem of sewer systems.

pH variation and influence in an autotrophic nitrogen removing biofilm system using an efficient numerical solution strategy.

A pH simulator consisting of an efficient numerical solver of a system of nine nonlinear equations was constructed and implemented in the modeling software MATLAB. The pH simulator was integrated in a granular biofilm model and used to simulate the pH profiles within granules performing the nitritation-anammox process for a range of operating points. The simulation results showed that pH profiles were consistently increasing with increasing depth into the granule, since the proton-producing aerobic ammonium-oxidizing bacteria (AOB) were located close to the granule surface. Despite this pH profile, more NH3 was available for AOB than for anaerobic ammonium oxidizers, located in the center of the granules. However, operating at a higher oxygen loading resulted in steeper changes in pH over the depth of the granule and caused the NH3 concentration profile to increase from the granule surface towards the center. The initial value of the background charge and influent bicarbonate concentration were found to greatly influence the simulation result and should be accurately measured. Since the change in pH over the depth of the biofilm was relatively small, the activity potential of the microbial groups affected by the pH did not change more than 5% over the depth of the granules.

Estimation of environment-related properties of chemicals for design of sustainable processes: development of group-contribution+ (GC+) property models and uncertainty analysis.

The aim of this work is to develop group-contribution(+) (GC(+)) method (combined group-contribution (GC) method and atom connectivity index (CI) method) based property models to provide reliable estimations of environment-related properties of organic chemicals together with uncertainties of estimated property values. For this purpose, a systematic methodology for property modeling and uncertainty analysis is used. The methodology includes a parameter estimation step to determine parameters of property models and an uncertainty analysis step to establish statistical information about the quality of parameter estimation, such as the parameter covariance, the standard errors in predicted properties, and the confidence intervals. For parameter estimation, large data sets of experimentally measured property values of a wide range of chemicals (hydrocarbons, oxygenated chemicals, nitrogenated chemicals, poly functional chemicals, etc.) taken from the database of the US Environmental Protection Agency (EPA) and from the database of USEtox is used. For property modeling and uncertainty analysis, the Marrero and Gani GC method and atom connectivity index method have been considered. In total, 22 environment-related properties, which include the fathead minnow 96-h LC(50), Daphnia magna 48-h LC(50), oral rat LD(50), aqueous solubility, bioconcentration factor, permissible exposure limit (OSHA-TWA), photochemical oxidation potential, global warming potential, ozone depletion potential, acidification potential, emission to urban air (carcinogenic and noncarcinogenic), emission to continental rural air (carcinogenic and noncarcinogenic), emission to continental fresh water (carcinogenic and noncarcinogenic), emission to continental seawater (carcinogenic and noncarcinogenic), emission to continental natural soil (carcinogenic and noncarcinogenic), and emission to continental agricultural soil (carcinogenic and noncarcinogenic) have been modeled and analyzed. The application of the developed property models for the estimation of environment-related properties and uncertainties of the estimated property values is highlighted through an illustrative example. The developed property models provide reliable estimates of environment-related properties needed to perform process synthesis, design, and analysis of sustainable chemical processes and allow one to evaluate the effect of uncertainties of estimated property values on the calculated performance of processes giving useful insights into quality and reliability of the design of sustainable processes.

Modelling of Harvesting Techniques for the Evaluation of the Density of Microalgae.

A better estimation of the density of cells has great relevance in the design of harvesting units. In the case of microalgae, the density is a function of the internal composition, which in turn is affected by external environmental conditions. The density of microalgae is often regarded as a constant or a generic value is retrieved from literature. This study proposes a procedure to evaluate the density of Chlorococcum sp. with simple sedimentation and centrifugation experiments coupled with the population balance equation (PBE), which is solved numerically. The density of cells is not constant; instead, it is a function of the size of particles, which in turn changes with the cells' phase of their life cycle. The calculated cellular density ranged between 1000 and 1100 kg m-3 in function of the cell size in both the sedimentation and centrifugation tests. The method can be extended to other microalgae species as well as to other types of cells.

Plant-wide assessment of alternative activated sludge configurations for biological nutrient removal under uncertain influent characteristics.

This study presents an extensive plant-wide model-based assessment of four alternative activated sludge (AS) configurations for biological nitrogen (N) and phosphorus (P) removal under uncertain influent loads and characteristics. Zeekoegat wastewater treatment plant (WWTP) in South Africa was chosen as case study due to its flexible design that enables operation in four different AS configurations: 3-stage Bardenpho (A2O), University of Cape Town (UCT), UCT modified (UCTM), and Johannesburg (JHB). A metamodeling based global sensitivity analysis was performed on a steady-state plant-wide simulation model using Activated Sludge Model No. 2d with the latest extension of physico-chemical processes describing the plant-wide P transformations. The simulation results showed that the predictions of effluent chemical oxygen demand (COD), N and P using the proposed approach fall within the interquartile range of measured data. The study also revealed that process configuration can affect: 1) how influent uncertainty is reflected in model predictions for effluent quality and cost related performances, and 2) the parameter rankings based on variance decomposition, particularly for effluent phosphate, sludge disposal and methane production. The results identified UCT and UCTM as more robust configurations for P removal (less propagated uncertainty and less sensitivity to N load) in the expense of incomplete denitrification. Moreover, based on the results of Monte-Carlo based scenario analysis, the balanced SRT for N and P removal is more sensitive to influent load variation/uncertainty for the A2O and JHB configurations. This gives a more operational flexibility to UCT and UCTM, where a narrow SRT range can ensure both N and P removal.

Modified calibration protocol evaluated in a model-based testing of SBR flexibility.

The purpose of this paper is to refine the BIOMATH calibration protocol for SBR systems, in particular to develop a pragmatic calibration protocol that takes advantage of SBR information-rich data, defines a simulation strategy to obtain proper initial conditions for model calibration and provides statistical evaluation of the calibration outcome. The updated calibration protocol is then evaluated on a case study to obtain a thoroughly validated model for testing the flexibility of an N-removing SBR to adapt the operating conditions to the changing influent wastewater load. The performance of reference operation using fixed phase length and dissolved oxygen set points and two real-time control strategies is compared to find optimal operation under dynamic conditions. The results show that a validated model of high quality is obtained using the updated protocol and that the optimization of the system's performance can be achieved in different manners by implementing the proposed control strategies.

Multivariate Principal Component Analysis and Case-Based Reasoning for monitoring, fault detection and diagnosis in a WWTP.

The main idea of this paper is to develop a methodology for process monitoring, fault detection and predictive diagnosis of a WasteWater Treatment Plant (WWTP). To achieve this goal, a combination of Multiway Principal Component Analysis (MPCA) and Case-Based Reasoning (CBR) is proposed. First, MPCA is used to reduce the multi-dimensional nature of online process data, which summarises most of the variance of the process data in a few (new) variables. Next, the outputs of MPCA (t-scores, Q-statistic) are provided as inputs (descriptors) to the CBR method, which is employed to identify problems and propose appropriate solutions (hence diagnosis) based on previously stored cases. The methodology is evaluated on a pilot-scale SBR performing nitrogen, phosphorus and COD removal and to help to diagnose abnormal situations in the process operation. Finally, it is believed that the methodology is a promising tool for automatic diagnosis and real-time warning, which can be used for daily management of plant operation.

A new extant respirometric assay to estimate intrinsic growth parameters applied to study plasmid metabolic burden.

Start-up phenomena in microbial biokinetic assays are not captured by the most commonly used growth-related equations. In this study we propose a new respirometric experimental design to estimate intrinsic growth parameters that allow us to avoid these limitations without data omission, separate mathematical treatment, or wake-up pulses prior to the analysis. Identifiability and sensitivity analysis were performed to confirm the robustness of the new approach for obtaining unique and accurate estimates of growth kinetic parameters. The new experimental design was applied to establish the metabolic burden caused by the carriage of a pWW0 TOL plasmid in the model organism Pseudomonas putida KT2440. The metabolic burden associated was manifested as a reduction in the yield and the specific growth rate of the host, with both plasmid maintenance and the over-expression of recombinant proteins from the plasmid contributing equally to the overall effect.

Assessment of the fate of organic micropollutants in novel wastewater treatment plant configurations through an empirical mechanistic model.

Novel wastewater treatment plants (WWTPs) are expected to be less energetically demanding than conventional ones. However, scarce information is available about the fate of organic micropollutants (OMPs) in these novel configurations. Therefore, the objective of this work is to assess the fate of OMPs in three novel WWTP configurations by using a plant-wide simulation that integrates multiple units. The difference among the three configurations is the organic carbon preconcentration technology: chemically enhanced primary treatment (CEPT), high-rate activated sludge (HRAS) combined or not with a rotating belt filter (RBF); followed by a partial-nitritation (PN-AMX) unit. The simulation results show that the three selected novel configurations lead mainly to comparable OMPs removal efficiencies from wastewater, which were similar or lower, depending on the OMP, than those obtained in conventional WWTPs. However, the presence of hydrophobic OMPs in the digested sludge noticeably differs among the three configurations. Whereas the configuration based on sole HRAS to recover organic carbon leads to a lower presence of OMPs in digested sludge than the conventional WWTP, in the other two novel configurations this presence is noticeable higher. In conclusion, novel WWTP configurations do not improve the OMPs elimination from wastewater achieved in conventional ones, but the HRAS-based WWTP configuration leads to the lowest presence in digested sludge so it becomes the most efficient alternative.

Uncertainty analysis in WWTP model applications: a critical discussion using an example from design.

This study focuses on uncertainty analysis of WWTP models and analyzes the issue of framing and how it affects the interpretation of uncertainty analysis results. As a case study, the prediction of uncertainty involved in model-based design of a wastewater treatment plant is studied. The Monte Carlo procedure is used for uncertainty estimation, for which the input uncertainty is quantified through expert elicitation and the sampling is performed using the Latin hypercube method. Three scenarios from engineering practice are selected to examine the issue of framing: (1) uncertainty due to stoichiometric, biokinetic and influent parameters; (2) uncertainty due to hydraulic behaviour of the plant and mass transfer parameters; (3) uncertainty due to the combination of (1) and (2). The results demonstrate that depending on the way the uncertainty analysis is framed, the estimated uncertainty of design performance criteria differs significantly. The implication for the practical applications of uncertainty analysis in the wastewater industry is profound: (i) as the uncertainty analysis results are specific to the framing used, the results must be interpreted within the context of that framing; and (ii) the framing must be crafted according to the particular purpose of uncertainty analysis/model application. Finally, it needs to be emphasised that uncertainty analysis is no doubt a powerful tool for model-based design among others, however clear guidelines for good uncertainty analysis in wastewater engineering practice are needed.

Comparison of the modeling approach between membrane bioreactor and conventional activated sludge processes.

Activated sludge models (ASM) have been developed and largely applied in conventional activated sludge (CAS) systems. The applicability of ASM to model membrane bioreactors (MBR) and the differences in modeling approaches have not been studied in detail. A laboratory-scale MBR was modeled using ASM2d. It was found that the ASM2d model structure can still be used for MBR modeling. There are significant differences related to ASM modeling. First, a lower maximum specific growth rate for MBR nitrifiers was estimated. Independent experiments demonstrated that this might be attributed to the inhibition effect of soluble microbial products (SMP) at elevated concentration. Second, a greater biomass affinity to oxygen and ammonium was found, which was probably related to smaller MBR sludge flocs. Finally, the membrane throughput during membrane backwashing/relaxation can be normalized and the membrane can be modeled as a continuous flow-through point separator. This simplicity has only a minor effect on ASM simulation results; however, it significantly improved simulation speed.